Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evaluation of transit signal priority options for rapid transit and light rail transit in the city of… Barton, Michael 2003

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2003-0480.pdf [ 7.73MB ]
Metadata
JSON: 831-1.0063904.json
JSON-LD: 831-1.0063904-ld.json
RDF/XML (Pretty): 831-1.0063904-rdf.xml
RDF/JSON: 831-1.0063904-rdf.json
Turtle: 831-1.0063904-turtle.txt
N-Triples: 831-1.0063904-rdf-ntriples.txt
Original Record: 831-1.0063904-source.json
Full Text
831-1.0063904-fulltext.txt
Citation
831-1.0063904.ris

Full Text

EVALUATION OF TRANSIT SIGNAL PRIORITY OPTIONS FOR RAPID TRANSIT AND LIGHT RAIL TRANSIT IN THE CITY OF RICHMOND by: M I C H A E L B A R T O N B.E.S., University of Waterloo, 2001 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF M A S T E R O F A P P L I E D S C I E N C E in THE F A C U L T Y OF G R A D U A T E STUDIES Department of Civil Engineering We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA June 2003 © Michael Barton, 2003 Authorization: In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Civil Engineering The University of British Columbia Vancouver, Canada Date Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page ii Transit and Light Rail Transit in the City of Richmond Abstract Transit signal priority (TSP) involves modifications to traffic signal cycle plans in order to give preferential treatment to transit vehicles to reduce travel delay and, consequently, average travel times and variability in travel times. This can make transit a more attractive mode of transportation, which can result in travelers choosing public transit over low occupancy vehicles. One of the main goals of this thesis is to develop a strategy that maximizes the benefits of TSP by finding the best values for certain parameters, including TSP parameters such as length of the green extension provided to transit vehicles, implementation of phase skipping and unconditional versus conditional priority. There are also corridor properties, such as locations of transit stops, locations of check-in detectors and whether transit has a shared or exclusive travel corridor. The literature reviewed in this paper indicates that TSP can provide significant travel time improvements compared to a network of fixed or actuated traffic signals. However, there can be considerable impacts on cross street traffic operations since signal phases are disrupted with each action taken to provide TSP. The other main goal of this thesis is to compare the influence of TSP implementation on transit operations with an express bus system and a light rail transit system. In order to study the influence of different parameters TSP performance, a model has been developed to simulate traffic and transit operations in the No. 3 Road corridor in the City of Richmond. The results indicate that transit vehicle delay and travel time are inextricably linked, meaning that reductions in travel time are largely the result of reductions in delay. The results also indicate that use of phase skipping and the choice between conditional and unconditional TSP are among the most important considerations in TSP implementation. Phase skipping results in less transit delay and lower travel times than the same scenarios without skipping. However, phase skipping increases adverse impacts as more cross streets experience delay increases and the scale of the increases is greater than without skipping. Unconditional TSP provides lower transit travel times since every vehicle is given priority at every intersection regardless of the operational situation existing at the time. However, conditional TSP is more sensitive to cross street traffic operations and other operational parameters, such as signal coordination. The light rail transit system provides lower transit travel times than express bus since light rail vehicles (LRVs) experience less delay in traveling through the corridor than buses. This is mainly the result of shorter average dwell times for LRVs at transit stations and the fact that an LRT would operate in an exclusive right-of-way for the entire corridor, free of the influence of automobile traffic experienced by buses in shared travel segments of the corridor. Based on the study findings, the following recommendations are made. Far side transit stops should be used at all locations to limit transit delay and travel times and allow for efficient use of green extensions by transit vehicles. A green extension should be selected based on detector-to-intersection distance and travel speed of a transit vehicle. Phase skipping should be employed when the objective is to minimize delay experienced by transit vehicles. An LRT line will offer significant transit operational benefits if implemented. A cost-benefit analysis should be undertaken to determine if the considerable delay and travel time benefits of LRT relative to express bus justify the capital expenditures required for implementation. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page iii Transit and Light Rail Transit in the City of Richmond Table of Contents Section Section Name Page # ABSTRACT ii TABLE OF CONTENTS iii LIST OF TABLES vii LIST OF FIGURES x ACKNOWLEDGEMENTS xi 1. INTRODUCTION 1 1.1 Objectives of Research 2 1.2 Problem Definition 2 1.3 Scope of Research 3 1.4 Thesis Structure 3 2. TRANSIT SIGNAL PRIORITY LITERATURE 4 REVIEW 2.1 Transit Signal Priority 5 2.1.1 Types of Transit Signal Priority 5. 2.1.1.1 Passive Priority 6 2.1.1.2 Active Priority 7 2.1.1.2.1 Conditional Priority 9 2.1.1.2.2 Unconditional Priority \ \ 2.1.1.2.3 Pre-Signals 12 2.1.2 Transit Vehicle Detection 13 2.1.3 Bus Schedule Reliability 16 2.2 Important Considerations in TSP Implementation 19 2.2.1 Signal Coordination 19 2.2.2 Compensation 19 2.2.3 Location of Transit Stops 19 2.3 Results from Implementation of Transit Signal Priority 20 2.3.1 Impacts of Transit Signal Priority on Travel Time 20 2.3.2 Impacts of Transit Signal Priority on Delay 21 2.4 Summary of Transit Signal Priority Literature Results 25 University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page iv Transit and Light Rail Transit in the City of Richmond 2.5 Light Rail Transit Signal Priority Considerations 27 2.5.1 LRT Operating Properties 29 2.5.1.1 Rates of Acceleration and Deceleration 29 2.5.1.2 Travel Speed 30 2,5.1,3 Headways 30 2.5.2 LRT Design Parameters 32 2,5,2.1 LRV Specifications 32 2.5.2.2 Station Design and Spacing 34 2.5.2.3 Station Placement 36 2.5.2.4 Integration with Other Transportation Modes 37 2.5.3 Light Rail Transit Signal Priority 38 2.5.4 Bus Rapid Transit—An Alternative to LRT 40 2.6 Literature Review Chapter Summary 41 STUDY DESIGN 42 3.1 TSP Parameters 42 3.1.1 Green Extension 42 3.1.2 Phase Skipping 42 3.1.3 Check-In Detector Location 42 3.1.4 Transit Stop Location 43 3.1.5 Type of Transit Signal Priority 43 3.2 Measures of TSP Effectiveness 43 3.2.1 Travel Time 43 3.2.2 Stop Delay 43 3.2.3 Total Delay 43 3.2.4 Cross Street Delay 44 3.2.5 Green Extension Effectiveness 44 3.3 VISSIM Model 44 3.3.1 Introduction 44 3.3.2 Operating Parameters 45 3.3.3 Network Geometry 46 3.3.4 Existing Upstream Check-In Detector Locations 47 3.3.5 Transit Stop Locations 48 3.3.6 Existing Signal Control Logic 48 3.3.7 Model Calibration 50 3.3.8 Simulation Scenarios 51 3.3.9 Challenges and Assumptions 52 University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page v Transit and Light Rail Transit in the City of Richmond 4. RESULTS AND ANALYSIS 54 4.1 Unconditional Signal Priority, Express Bus Service 54 4.1.1 Transit Vehicle Travel Time and Delay 55 4.1.1.1 Transit Stop and Check-In Detector Locations 55 4.1.1.2 Length of Green Extension 56 4.1.1.3 Phase Skipping 59 4.1.1.4 Summary 59 4.1.2 Transit Vehicle Travel Time Variability 60 4.1.2.1 Transit Stop and Check-In Detector Locations 60 4.1.2.2 Length of Green Extension 61 4.1.2.3 Phase Skipping 61 4.1.2.4 . Summary 62 4.1.3 Cross Street Traffic Operations 63 4.1.3.1 Total Delay 63 4.1.3.2 Level of Service 66 4.1.4 Green Extension Effectiveness 68 4.1.4.1 Detector and Transit Stop Location. 68 4.1.4.2 Green Extension Length 71 4.1.4.3 Phase Skipping 72 4.1.4.4 Summary 73 4.2 Express Bus Service with Conditional Transit Signal 74 . Priority 4.2.1 Transit Vehicle Travel Time and Delay 75 4.2.1.1 Transit Stop and Detector Location 75 .. 4.2.1.2 Phase Skipping 76 4.2.1.3 Unconditional versus Conditional Priority 77 4.2.1.4 Summary 78 4.2.2 Travel Time Variability 78 4.2.2.1 Unconditional Priority v. Conditional Priority 78 4.2.3 Cross Street Traffic Operations 81 4.2.3.1 Total Delay 81 4.2.3.3. Level of Service 82 4.3 Summary and Discussion of Results for Express Bus 84 Signal Priority 4.3.1 Transit Vehicle Travel Time and Delay 84 4.3.2 Travel Time Variability 86 4.3.3 Cross Street Traffic Operations 87 4.3.4 Green Extension Effectiveness 88 University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page vi 4.4 Light Rail Transit with Unconditional Transit Signal 89 Priority 4.4.1 Transit Vehicle Travel Time and Delay 89 4.4.1.1 Detector Locations 89 4.4.1.2 Green Extension Length 90 4.4.1.3 Phase Skipping 92 4.4.1.4 LRT versus Unconditional and Conditional Express Bus 92 Priority 4.4.1.5 Summary 95 4.4.2 Travel Time Variability 96 4.4.2.1 Detector Locations 96 4.4.2.2 Green Extension Length 96 4.4.2.3 Phase Skipping 97 4.4.2.4 LRT Priority v. Express Bus Priority 97 4.4.2.5 Summary 99 4.4.3 Cross Street Delay Analysis 100 4.4.3.1 Total Delay 100 4.4.3.2 Unconditional TSP v. Conditional TSP 103 4.4.3.3 Level of Service 104 4.4.4 Green Extension Effectiveness 105 4.4.4.1 Transit Station and Detector Location 105 4.4.4.2 Green Extension Length and Phase Skipping 106 4.4.4.3 Summary 107 4.5 Summary and Discussion of Results for Unconditional 108 LRT Priority 4.5.1 Transit Vehicle Travel Time and Delay 108 4.5.2 Travel Time Variability 109 4.5.3 Cross Street Traffic Operations 110 4.5.4 Green Extension Effectiveness 110 5. CONCLUSIONS AND RECOMMENDATIONS 112 5.1 Conclusions 112 5.1.1 Transit Vehicle Travel Time and Delay 112 5.1.2 Travel Time Variability 113 5.1.3 Cross Street Traffic Operations 113 5.1.4 Green Extension Effectiveness 114 5.2 Recommendations 115 References 116 University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page vii Transit and Light Rail Transit in the City of Richmond List of Tables Table Table Title Page# 2-1 Average Delay (sec/veh) 9 2-2 Travel Time and Delay Results—Entire Network- 9 2-3 Guidelines for Use of Unconditional Priority During Off- 12 Peak Hours 2-4 Total Delay Results for TSP Simulation 21 2-5 Average Travel Time Comparison with Different TSP 22 Strategies 2-6 Standard Deviation of Travel Time for Different TSP 22 Strategies 2-7 Average Delay Comparison Between Field and Model 23 Observations 2-8 Total Intersection Delay Comparison 24 2-9 Summary of Literature Review Study Results 25 2-10 Comparison of LRT Acceleration and Deceleration Rates 29 2-11 Comparison of Average LRT Travel Speed in Different 30 Cities 2-12 Comparison of LRT Headways in Different Cities 31 2-13 Comparison of Peak Headways in Different Cities 32 2-14 Comparison of Light Rail Vehicle Dimensions 33 2-15 Average Station Spacing 35 2-16 Typical LRT Station Spacing 35 2-17 LRT Line Comparison for Different Systems 35 2-18 Average Stopped Delay 39 2-19 Maximum Queue Lengths 39 3-1 Approximate Check-In Detector Locations 47 3-2 Transit Stop Locations 48 3-3 Existing Signal Timing Parameters in No. 3 Road Corridor 49 4-1 Total Transit Vehicle Delay Comparison—Detector Location 56 Analysis 4-2 Total Transit Vehicle Delay Comparison—Transit Stop 56 Location Analysis 4-3 Transit Vehicle Travel Time Comparison—Detector 56 Location Analysis 4-4 Transit Vehicle Travel Time Comparison—Transit Stop 56 Location Analysis 4-5 Total Transit Vehicle Delay Comparison—Phase Skipping 59 Analysis 4-6 Transit Vehicle Travel Time Comparison—Phase Skipping 59 Analysis 4-7 Transit Vehicle Travel Time Variability Comparison— 60 Detector Location Analysis University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page viii Transit and Light Rail Transit in the City of Richmond 4-8 Transit Vehicle Travel Time Variability—Transit Stop 61 Location Analysis 4-9 Transit Vehicle Travel Time Variability Comparison 61 4-10 Transit Vehicle Travel Time Variability Comparison—Phase 62 Skipping Analysis 4-11 Cross Street Delay Comparison 63 4-12 Cross Street Delay Comparison—Phase Skipping Analysis 64 4-13 Cross Street Delay Comparison 64 4-14 Cross Street Delay Comparison—Phase Skipping Analysis 64 4-15 Change in Cross Street Delay 65 4-16 Cross Street Delay Comparison—Green Extension Length 65 4-17 Cross Street Delay Comparison—Green Extension Length 65 4-18 Cross Street Stop Delay and Level of Service 67 4-19 Green Extension Efficiency Comparison—Near v. Far Side 70 Stop 4-20 Green Extension Efficiency Comparison—Near v. Far Side 70 Stop 4-21 Transit Vehicle Travel Time Comparison—Detector 75 Location Analysis 4-22 Transit Vehicle Travel Time Comparison—Detector 75 Location Analysis 4-23 Total Transit Vehicle Delay Comparison—Detector Location 75 Analysis 4-24 Total Transit Vehicle Delay Comparison—Transit Stop 76 Location Analysis 4-25 Total Transit Vehicle Delay Comparison—Phase Skipping 76 Analysis 4-26 Transit Vehicle Travel Time Comparison—Phase Skipping 76 Analysis 4-27 Southbound Travel Time Comparison—Unconditional v. 77 Conditional Bus Priority 4-28 Northbound Travel Time Comparison—Unconditional v. 77 Conditional Bus Priority 4-29 Travel Time Variability Comparison—Unconditional v. 79 Conditional Bus Priority 4-30 Number of Cross Street Approaches Experiencing Increases 82 in Delay with TSP 4-31 Cross Street Stop Delay and Level of Service 83 4-32 Transit Vehicle Travel Time Comparison—Detector 90 Location Analysis 4-33 LRT Total Delay Comparison—Phase Skipping Analysis 92 4-34 LRT Travel Time Comparison—Phase Skipping Analysis 92 4-35 Southbound Total Delay Comparison—Unconditional Bus v. 93 Unconditional LRT Priority 4-36 Northbound Total Delay Comparison—Unconditional Bus v. 93 Unconditional LRT Priority University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page ix Transit and Light Rail Transit inthe City of Richmond 4-37 Southbound Travel Time Comparison—Unconditional Bus 93 v. Unconditional LRT Priority 4-38 Northbound Travel Time Comparison—Unconditional Bus 93 v. Unconditional LRT Priority 4-39 Travel Time Variability Comparison—Detector Location 96 Analysis 4-40 Transit Vehicle Travel Time Variability Comparison 97 4-41 Travel Time Variability Comparison—Phase Skipping 97 Analysis 4-42 Travel Time Variability Comparison 99 4-43 Range of Cross Street Delay Increase Values 100 4-44 Cross Street Delay Comparison 101 4-45 Cross Street Delay Comparison—Phase Skipping 101 4-46 Cross Street Delay Comparison 102 4-47 Cross Street Delay Comparison—Phase Skipping 102 4-48 Cross Street Delay Comparison—Green Extension Length 102 4-49 Cross Street Delay Comparison—Green Extension Length 102 4-50 Number of Cross Street Approaches Experiencing Increases 103 in Delay with TSP 4-51 Cross Street Stop Delay and Level of Service 104 4-52 Overall Green Extension Efficiency Comparison—Phase 107 Skipping Analysis University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Pagex List of Figures Figure Figure Title Page# 2A . Normal Phasing v. Split Phasing 7 2B Signal Timing Comparison 13 4A Average Travel Time—Green Extension Length Analysis 58 4B Comparison of Green Extension Effectiveness Between 69 Detector Locations 4C Comparison of Green Extension Effectiveness Between 70 Transit Stop Locations 4D Green Extension Effectiveness—Green Extension Length 71 Comparison 4E Green Extension Effectiveness—Green Extension Length 72 Comparison 4F Green Extension Effectiveness: Phase Skipping Comparison 73 4G Travel Time Variability Comparison—50-Metre Detectors, 79 Far Side Stops 4H Travel Time Variability Comparison—50-Metre Detectors, 80 Existing Stops 41 Light Rail Transit Travel Time Comparison—Southbound 91 4J Light Rail Transit Travel Time Comparison—Northbound 91 4K Travel Time Comparison—50-Metre Detectors, Existing 94 Transit Stops 4L Travel Time Comparison—50-Metre Detectors, Far Side 94 Transit Stops 4M Green Extension Effectiveness Comparison—Detector 106 Location University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page xi Transit and Light Rail Transit in the City of Richmond Acknowledgements Well, here it is. The time I thought would never come. As I sit here writing what will be the final segment of my Master's thesis, I cannot help but think back to the days when I was sitting in front of my computer running simulation after simulation, wondering how I was going to transform all of the data that I was generating into a comprehensible thesis paper. Not only was there no light at the end of the tunnel, but I was waiting for the walls to come crashing down around me. But here I am—I hope a better person for having gone through the last 13 months. This period brought about many changes in my personal life and relationships, as well as in myself as a person. As in my life, my thesis was an up-and-down battle that went from excitement at figuring out a piece of code that had stopped me in my tracks for days to the depression of realizing the hundreds of simulations that would have to be completed and the subsequent analysis of that data that would be necessary. But I can truly say that I wouldn't have made it without the support of some notable people. First of all, my advisor, Dr. Tarek Sayed, kept me on track to complete my research some time close to the pre-planned completion data. He helped me to focus my efforts when the amount of material and the research opportunities seemed almost overwhelming. If not for his help, I would be attaching this section to a 500-page encyclopedia of signal priority results and analysis. Fortunately, we were able to decide what was important and what I stubbornly held onto without any real purpose. My family and friends were pivotal in my not breaking down over the past year. Events that those concerned know of made the road to the present even tougher to walk. Without the support I received, it would be been easy to give up or not give my best. While I was basically on my own when it came to developing my model, running the simulations and analyzing the results, these people helped me to release stress and to realize that there was more to life than transit data. Along the way, there are things I wish I had said and done, as well as many things I wish I didn't say or do. Some relationships have blossomed while others, sadly, have seen their demise. But life is about learning by doing, and my experiences over this year will stay with me forever and undoubtedly affect the remaining course of my life. TS, MB, JB, MB, AW, AB, PB University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 1 1. INTRODUCTION Transit signal priority involves modifications to traffic signal cycle plans in order to give preferential treatment to transit vehicles to reduce travel delay and, consequently, average travel times and variability in travel times. By improving transit travel times and reliability, the overall quality of transit service is also improved. This can make transit a more convenient and attractive mode of transportation, which can result in travelers choosing public transit over low occupancy vehicles. Reducing dependence on low occupancy vehicles can help to reduce congestion, thus improving travel times and level of service in the transportation network. This can also decrease the need to build new road infrastructure or expand existing infrastructure to deal with increased travel demand. Traffic congestion occurs when the number of vehicles traveling in a transportation network approaches or exceeds the capacity of the network. The traditional approach to dealing with congestion has been to build new roadway facilities rather than managing the traffic demand. According to the United States Department of Transportation, it is difficult for construction of new highway capacity to keep pace with growth in demand. For example, between 1980 and 1999, the U.S. highway network increased by 1.5% while vehicle kilometers of travel increased by 76% (www.fhwa.dot.gov'). Increased traffic congestion leads to economic losses due to wasted fuel and lost productivity. Vehicles burn fuel unnecessarily when stopped or traveling at low speeds in congestion conditions. Lost productivity occurs when people are sitting in their vehicles rather than working or participating in other activities. Moreover, congestion results in unreliable travel times, which requires travelers to build more time into their travel schedules. These conditions also lead to increases in stress and lost leisure time. There are also environmental costs associated with congestion. Vehicles tend to burn more fuel when traveling at low speeds, which reduces air quality. Furthermore, land and environmental resources are sacrificed to build new highway facilities in order to increase traffic capacity. Shifting a portion of the automobile traffic demand to public transit will help to reduce the pressure on roadway facilities and the associated costs and problems. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 2 1.1 Objectives of Research The main objectives of this research project are the following: • To study transit signal priority strategies to determine which combinations of geometrical design, physical configuration, vehicle operating and TSP characteristics result in the most efficient and effective transit operations; • To compare the effects of unconditional and conditional transit signal priority on transit operations; and • To compare the results of transit signal priority implementation with an express bus system and a light rail transit system. 1.2 Problem Definition Although the majority of the literature reviewed regarding the implementation of TSP indicates that signal priority provides significant benefits for transit service quality, there are many variables that can affect the effectiveness and efficiency of TSP provision. The goal is to implement a strategy that maximizes the benefits of TSP. This requires finding the best values for certain parameters and the optimal combination of these parameters. These parameters include variables in the TSP logic itself, as well as physical and operational properties of the travel corridor. TSP logic parameters that are considered in this research are the length of the green extension provided to transit vehicles, the implementation of phase skipping and unconditional versus conditional priority. The corridor properties that are considered are the locations of transit stops, the locations of vehicle check-in detectors and whether transit has a shared or exclusive travel corridor. While the major objective is to improve transit operations, it is also important to consider how TSP affects traffic operations on streets crossing the main corridor. Thus, the challenge is to find a balance between improving transit service while maintaining an acceptable level of service for cross street traffic operations. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 3 1.3 Scope of Research The case study considered in this research project focuses on transit operations in the No. 3 Road corridor in the City of Richmond. The particular study segment stretches approximately 3 km from Bridgeport Road at the northern limit to Granville Avenue at the southern limit. There are sixteen signalized intersections in this corridor, which are listed below: TSP is applied to the intersections from Sea Island Way to Park Road. The Bridgeport Road signal timings are not modified because transit vehicles enter the corridor on Sea Island Way. Similarly, transit vehicles exit the corridor at Anderson Road before reaching Granville Avenue. This research project examines the effectiveness of a transit signal priority strategy with different physical and operational characteristics. The most significant characteristics affecting TSP are whether green extension is employed, the length of the green extension, the use of phase skipping, the location of check-in detectors, the location of transit stops and the use of hard signal priority versus conditional priority. 1.4 Thesis Structure This thesis is composed of five major chapters, including this introductory section. Following the introduction is the literature review component of this research project. Chapter 3 discusses the case study used to examine the different components of transit signal priority and the model employed to simulate the proposed transit signal priority strategies. Chapter 4 contains the simulation results and analysis and discussion of these results. Finally, chapter 5 outlines the conclusions drawn from the results and analysis component and provides recommendations regarding TSP implementation. Yohan Mall Access Leslie Street Alderbridge Way Bridgeport Road Sea Island Way Capstan Road Cambie Road Browngate Road Lansdowne Mall Access Lansdowne Road Ackroyd Road Westminster Highway Saba Road Cook Road Park Road Granville Avenue University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 4 Transit and Light Rail Transit in the City of Richmond 2. TRANSIT SIGNAL PRIORITY LITERATURE REVIEW Transit signal priority (TSP) involves the implementation of certain adjustments to a transportation network to improve transit efficiency, productivity and, ultimately, attractiveness. By reducing transit travel times and delays, transit operators aim to increase the share of total trips made by public transit relative to low occupancy vehicles. The objective of TSP control strategies is to operate traffic signals so as to minimize total person delay. Total person delay is typically the focus rather than vehicle delay since transit vehicles carry significantly greater numbers of people per vehicle than private automobiles. Moreover, the success of TSP usually requires that any increase in delay on cross streets be more than offset by the reduction in delay on the prioritized roadway. TSP can be implemented actively, which involves making adjustments to signal timings based on the detection of transit vehicles. TSP can also be implemented passively by setting fixed signal timings to favour transit vehicles, rather than making adjustments based on the presence of a transit vehicle. The effectiveness of a TSP scheme can be gauged with the following performance measures: total intersection delay; delay to minor movements; minor movement cycle failures (when vehicles arrive on the red and are unable to clear the intersection during the following green signal); bus travel time; bus schedule reliability (even i f no significant decrease in average travel times is detected after TSP implementation, schedule reliability may improve if the variation in travel times decreases); intersection bus delay (the time difference between when the bus stops at the end of a queue at a signalized intersection and the time it passes the stop line); and intersection person delay. By reducing the time that transit vehicles spend delayed at intersections, TSP initiatives are intended to reduce transit delay and travel time and improve service reliability while enhancing service quality. These initiatives are implemented with the accompanying objective of providing benefits with a minimum impact on other facility users, including cross street traffic and pedestrians. According to the Intelligent Transportation Society of America (2002), transit vehicles spend an average of 15% of trip time waiting at traffic signals. Reducing this delay by an average of 40% would reduce a 60-minute round trip to 55 minutes. Moreover, if 5-minute headways were required, only 11 buses would be necessary, compared to 12 under the 60-minute trip length. Sunkari et al. (1995) suggest that delay at traffic signals comprises between 10 and 20% of overall bus trip times and nearly 50% of the delay experienced by a bus. The concepts of signal priority and signal pre-emption are often used interchangeably. The Intelligent Transportation Society of America (2002) makes the following distinction University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 5 Transit and Light Rail Transit in the City of Richmond between these two concepts: signal priority modifies normal signal operation to better accommodate transit vehicles while signal pre-emption interrupts the normal process for special events—in this case, a bus arrival—and uses a special timing plan that requires the traffic signal controller to transition out of and back into the coordinated operation of the normal signal timing plan. According to Nelson and Bullock (1999), pre-emption temporarily disrupts coordinated operation by removing coordination constraints to service the pre-emption call. Once the call is dropped, the signal then returns back to normal operation after a period of offset seeking. However, this process may take anywhere from 0 to 5 cycles depending on the selected transition algorithm and the length of the cycle. Pre-emption reallocates some of the green time to the priority movement, effectively stealing green time from other movements. If one of the movements that loses green time is already saturated, then significant delay can result on that approach. 2.1 Transit Signal Priority Although passenger vehicles greatly outnumber transit vehicles on road networks, the number of persons transported per vehicle is much greater in the case of transit vehicles. As Lin et al. (1995) explain, due to the large number of passengers in transit vehicles compared to passenger vehicles, signal-related delay to transit vehicles can greatly increase the total traffic operating costs at an intersection. Therefore, optimizing travel efficiency of transit vehicles is beneficial for the overall effectiveness of a transportation system in terms of persons transported. Furthermore, Chang et al. (1996) suggest that preferential treatment of bus users, such as providing signal priority or pre-emption, encourages the use of public transit systems and is one of the most promising strategies for relieving urban congestion. Route travel times for transit vehicles are increased when transit vehicles are regularly required to stop at traffic signals. This also reduces the travel efficiency and person-carrying capacity of the transit system. Once a transit vehicle falls behind schedule, it will typically have more than the average number of passengers to pick up at subsequent stations, which causes further delays. Another drawback with fixed traffic signals with respect to transit efficiency is that some transit vehicles will be delayed more than others, due to random variations in the number of signals at which individual transit vehicles are required to stop. This causes increases and decreases in the service headways between transit vehicles, which reduce the efficiency of the transit service. 2.1.1 Types of Transit Signal Priority Transit signal priority can be implemented through either passive or active strategies. A l -Sahili and Taylor (1996) describe passive priority as determining and making appropriate adjustments based on the intensity of bus movements from background traffic data while University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 6 active priority involves detection of individual buses and making priority decisions based on the information provided by detectors. Chada and Newland (2001) indicated the following characteristics of an effective transit signal priority strategy: • The incorporation of a wide variety of priority techniques for different situations; • The ability to minimize delay to non-transit traffic and offer compensation; and • The means to estimate cost to both passengers and the transit agency associated with enacting any given TSP method. 2.1.1.1 Passive Priority Passive priority strategies implement fixed signal settings that favour transit vehicles in order to reduce delay experienced by these vehicles. According to Skabardonis (2000), fixed time plans that favour buses are effective for high bus volumes and fairly predictable arrival times at the intersection. But uncertainty in bus arrival times may substantially reduce the benefits of transit-weighted fixed-time plans. For example, priority could be provided by adjusting the offset between two coordinated signals to account for the slower speed of the bus and the mid-block dwell time so that the bus could travel through both signals unimpeded. However, this strategy can be extremely difficult to implement since transit stops prevent vehicles from moving at a constant speed and dwell times at stops vary. Signalization plans are not directly affected by the presence of transit vehicles. Many of these strategies involve adjustments to cycle lengths, offsets and splits. One approach is to allocate more green time to the street with the prioritized transit movement by increasing the split size for the prioritized phase. Another option is to shorten the cycle length to reduce delay caused by waiting for the next green phase. However, this comes at the expense of reduced intersection capacity due to the increased percentage of lost time in a particular time period. Shortening cycle length along an arterial reduces stopped time delay to transit vehicles and private vehicles, but the merits must be weighed against the capacity reduction on the arterial. As Garrow (1998) explains, using short cycle lengths is appealing because benefits to transit are realized with little monetary cost and the strategy does not penalize vehicles along the cross streets by using a portion of their green time to favour transit vehicles. Split phasing is a passive priority technique in which the green phase for the transit corridor occurs twice in the same cycle. This reduces wait time between green phases. Figure 2A compares the signal timing plan under normal phasing with split phasing. According to Bauer et al. (1995), the degree of cross street saturation and the amount of green time taken from the cross street are important considerations in deciding whether to use TSP. For example, taking 5 to 10 seconds of green time from cross streets with saturation over 0.9 and 0.8 respectively can cause signal plan failure. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 7 A B C Norma l Phas ing A B A C Spl i t Phas ing Figure 2A: Normal Phasing v. Split Phasing 2.1.1.2 Active Priority Active priority strategies modify signal settings dynamically and in real-time in order to reduce delay to an oncoming transit vehicle. Under full priority schemes, the objective is to provide zero delay to transit vehicles. Partial priority employs only the least disruptive TSP techniques, including green extension and red truncation, and strict limits are often placed on extension lengths. Relative priority forces transit vehicles to compete with other traffic for green time and priority. The well being of traffic operations on cross streets is considered before granting priority to the transit vehicle's approach. Active strategies include green extension, red truncation (early start), phase insertion and phase skipping. Green extension involves extending the green phase for an approaching transit vehicle beyond its normal setting to allow a bus to pass through an intersection. The length of the green extension is normally subject to a maximum duration, in order to reduce delay to non-priority traffic. Chada and Newland (2001) suggest a maximum green extension of 10 seconds. These authors also suggest that, under highly congested conditions, green extension should only be implemented in every other cycle. Research by Garrow (1998) indicates that a 10-second green extension offers superior performance over 20-second extension across all bus saturation levels because larger green extensions place the signal timing of an intersection farther away from the original timing intended for automobiles. Red truncation (or early start) advances the bus street green phase by prematurely terminating all other non-bus phases. This occurs when a prioritized transit vehicle is approaching an intersection that is showing a red interval to its approach. Another option is to insert a short phase into the phase sequence to serve the priority transit movement when other phases need to be served before the normal return to green on the priority approach. That is, an extra phase is inserted if a bus is detected during the red phase and it is not due to receive green in the next scheduled phase. The signal controller then returns to the normal sequence once the bus has cleared the intersection. The special green phase is injected into the normal phase sequence while all other phases are stopped. As Davol (2001) explains, there is a time window in which the extra phase can be inserted. The earliest time the inserted phase can start is determined by the University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 8 Transit and Light Rail Transit in the City of Richmond minimum guaranteed time for the preceding phase. The latest time that the inserted phase can end is determined by the latest time the following signal phase can start and still receive its minimum required time. Phase skipping is an active TSP strategy that involves skipping one or more non-priority phases to provide priority to a bus phase. This strategy is usually employed when the bus movement is not due to receive a green signal in the next phase. In such a situation, a red truncation of the current phase would not provide priority to the bus. Consequently, the phase scheduled to follow the current phase is skipped so that the signal controller can move to the phase serving the transit vehicle. Phase rotation involves rotating the order of signal phases to provide TSP. For example, a lagging left turn phase may be changed to a leading phase when a bus requests priority, and then revert back to a lagging phase. According to Garrow (1998), the use of green extensions alone, without red truncations, tends to have the best overall impact on traffic. While the addition of red truncations to the use of green extensions reduces delays to transit, there is a high cost to general traffic. Garrow (1998) suggests that the greater the ratio of arterial traffic volumes to cross street traffic volumes, the less significant the introduction of signal priority appears to be in terms of increases in delay per vehicle. When traffic volume is light, active priority may result in little or no impact on other traffic due to excess available capacity. However, the effects can be significant when the intersection is near or at capacity. The system-wide value of active priority may only be worthwhile i f transit has a high ridership in the corridor, causing benefits per person to outweigh the costs. Al-Sahili and Taylor (1996) used simulation to evaluate the effects of different combinations of active priority strategies. These plans included the following: 1. green extension and red truncation without compensation; 2. green extension and red truncation with compensation; 3. phase skipping without providing compensation: when extension or truncation was not sufficient to let the bus pass through a green signal, the phase was completely skipped for one cycle; and 4. skipping a phase while providing subsequent compensation. These researchers considered the average delay to vehicles travelling along the test link under pre-existing conditions (base case) and each of the four plans described above. Table 2-1 outlines the average delay per vehicle over the 45-minute simulation period. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 9 Transit and Light Rail Transit in the City of Richmond Table 2-1: Averag e Delay (sec/veh) Direction Base Case Plan 1 Plan 2 Plan 3 Plan 4 W B 418.5 409.0 416.6 380.2 421.1 EB 305.0 306.3 310.8 330.5 343.3 It is important to notice that Plan 4 results in increased delay in both directions. This is due to the fact that skipping a signal phase causes the greatest disturbance to the signal-timing plan. As a result, a significant amount of time is required to recover from the phase sequence interruption. In this situation, no pre-emption is the best option in terms of overall delay when the signal network is optimized. T a b l e 2-2 illustrates the total travel time, delay and average delay for all road users in the network from the simulation. Table 2-2: Travel Time and Delay Results—Entire Network Case Travel Time (person-minutes) Delay (person-minutes) Average Delay (sec/trip) No pre-emption 20,264 12,042 48.4 1 20,350 12,174 49.0 2 20,409 12,221 49.2 3 22,758 13,769 49.6 4 23,230 14,398 52.3 2.1.1.2.1 Conditional Priority Under conditional priority schemes, a priority request is evaluated based on pre-determined criteria and priority is awarded based on how adequately the criteria are satisfied. Control logic that automatically and immediately favours transit vehicles can cause delays to other vehicular movements at the same intersection, disrupt signal coordination and even cause problems in transit operations by causing transit vehicles to run ahead of schedule. These and other problems associated with unrestricted transit signal priority can propagate each time priority is given at an intersection. To address these deficiencies, the signal control system should employ real-time information about transit operations and general traffic conditions and adapt to changing conditions while minimizing disruptions to a road network. The signal priority algorithm should decide whether to give transit priority based on the following considerations: • Traffic volumes at all approaches to an intersection; • Queue lengths and potential spillbacks that might block lanes or intersections; • Expected arrival times of transit vehicles at an intersection; • Deviations from schedule; • Deviations from proper service headways between transit vehicles; and • Expected arrival times of connecting transit vehicles at downstream stations. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 10 Transit and Light Rail Transit in the City of Richmond According to Zhang (2001), appropriate conditions for consideration in a priority decision include: the degree of saturation on approaches that will disbenefit from TSP; schedule adherence; transit vehicle ridership; time since priority was last given; and the number of queued cross street vehicles. Skabardonis (2000) suggests a number of criteria for granting conditional transit signal priority. The first is that there should be spare green time in the cycle so that signal pre-emption or priority will not result in significant impacts to saturated movements at a signalized intersection or loss of coordination. Spare green time is calculated as G e = X Gj (1 - Xj), where G e = spare green time in the cycle; N = number of phases; Gj = green time for phase i ; and Xj = degree of saturation for the critical link moving in phase i . Another criterion is schedule adherence. That is, transit priority should not be provided i f it may result in buses being ahead of schedule, and some proposed strategies provide priority only to those buses that are behind schedule. Tobin (2002) suggests using a bus lateness threshold of 3 minutes while Balke (1999) suggests a threshold of 5 minutes. Kelman (2002) argues that transit signal priority should not be implemented at intersections with a volume-to-capacity ratio of greater than 0.90. This position is supported by Balke (1999), who argues that the increase in non-priority delay cannot be offset by reductions in bus delays when the v/c ratio exceeds 0.90. Furth and Muller (2001) point out that the frequency of priority interruptions is an important consideration since an intersection may need time to recover after granting priority; that is, some time may be necessary to clear the queues built up during the priority interruption or to return to the background cycle that allows for signal progression. If priority interruptions occur too frequently, an intersection may be unable to recover, resulting in cycle failures when queues cannot be cleared. Consequently, some conditional priority schemes inhibit priority requests for a fixed period of time after an interruption, or when long queues exist. Chang et al. (1996) applied a performance index approach where, upon detection of a bus, a performance index is calculated to compute the net benefit of providing a green extension to the bus movement. If the performance index (PI) is negative, then giving an extension is not favourable and TSP is not granted. The PI is a sum of the trade-offs that result due to the signal control decision, including: passenger delay (Cpd ); vehicle delay (Coc'); and schedule delay (Csd ). The formula applied by these researchers was PI = Z PI1', where PI1' = C p / + Coc'' + Csd1'. Conditional priority schemes make use of real-time traffic signal control systems. In general, the information provided by vehicle detectors is used by the signal controllers to decide whether to give an individual vehicle signal priority so as to maintain regular headways. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 11 Transit and Light Rail Transit in the City of Richmond 2.1.1.2.2 Unconditional Priority Unconditional priority involves the provision of transit signal priority each and every time it is requested, regardless of the resulting impacts on overall traffic operations or the actual need for priority. According to the U . S. Federal Transit Administration (2000), such basic control logic can cause serious problems to other traffic, including increased delays and travel times. According to Garrow (1998), unconditional priority is usually necessary only at high volume intersections. Where the cross street volume is light, unconditional priority will rarely be triggered by the bus since the bus approach already receives a large fraction of the intersection green time. When the cross street intersection saturation levels drop below 0.25, one might consider using unrestricted signal priority because signal timing should already heavily favour the bus approach. Chada and Newland (2001) recommend that unconditional signal priority be reserved for express bus service during off-peak hours and that it should be regulated by placing limits on the duration of green extension and red truncation periods, especially at intersections with busy cross streets. Garrow (1998) concurs with Chada and Newland (2001) in suggesting that unconditional priority should be reserved for express bus service during off-peak hours since this service uses longer headways than local bus service, which results in fewer priority calls. Off-peak traffic volumes enable cross streets to recover from each priority call more quickly than during the peak period. Implementation of transit signal priority during peak time periods is more difficult than during off-peak periods since both cross streets and arterials are likely to be operating at higher degrees of saturation than during off-peak times. As a result, less excess network capacity is available. Findings reported by Garrow (1998) indicate that when 10-second green extensions are used in conjunction with a cross street saturation of 0.8, signal priority does not result in substantial increases in delay per vehicle along the cross street approaches. When cross street saturation is raised to 0.9 while maintaining the 10-second extensions, the cross streets begin to feel more substantial delay increases. In addition, the increases in delay felt by cross streets do not readily dissipate with time. Garrow (1998) suggested that green time should not be taken from cross streets operating at saturation levels of 1.0 to award priority to transit vehicles along the arterial. Garrow (1998) also found that when green extensions were increased to 20 seconds, larger delay increases were encountered along the cross streets. However, when cross streets were operating at saturation levels of 0.8, enough excess capacity was available to allow them to recover from the impacts of the priority signal timing within 2 to 3 signal cycles following each priority call. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 12 Table 2-3 displays the guidelines provided by Garrow (1998) regarding the appropriate green extension and red truncation lengths to use depending on cross-street saturation. Table 2-3: Guidelines for Use of Unconditional Priority During Off-Peak Hours Cross Street Saturat ion Level Recommended Green Extension or Red T runca t i on Length Below 0.25 Unbounded 0.25-0.35 20 seconds 0.35-0.70 10 seconds 2.1.1.2.3 Pre-Signals A pre-signal gives buses priority access to a bus advance area at an intersection stop line to avoid the traffic queue and avoid bus delay. In most cases, a reserved bus lane is provided on the intersection approach. While pre-signals are prominent in the United Kingdom, they are not widely used in North America. Wu and Hounsell (1998) identified the following categories of pre-signal design: • Category A: the pre-signal controls only the non-priority traffic while buses are uncontrolled; • Category B: buses are also controlled by a pre-signal at the end of the bus lane. Once the transit vehicle has cleared the intersection, the pre-signal in the bus lane turns red and non-priority vehicles receive right of way; and • Category C: similar layout to B, except that detectors are installed in the bus lane. When a bus approaches the pre-signal, a red signal is shown to the non-priority traffic stream to enable the bus to access to the main signal without impedance. The bus lane pre-signal then turns red, allowing non-priority traffic to fill the bus advance area behind the bus after it has selected the appropriate lane at the main signal. Full priority at a pre-signal occurs when buses can call the green in both the pre-signal and the main signal and, thus, may experience no delay at the intersection. If buses can only call the green at the pre-signal, they may be delayed by a red light at the main signal, depending on the timing of the main signal. Wu and Hounsell (1998) identify two potential problems with pre-signals. The first is that there is potential for waste of intersection capacity by not releasing traffic from the pre-signal sufficiently early to make use of the green time at the main intersection. The second is that the relocated traffic queue may block back to the upstream intersection. Figure 2B illustrates the signal timings for parallel main and pre-signals. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 13 .^4 P-^ 9 P • r n < rm c 9 m • rp: red time at the pre-signal; gm: green time at the main signal; rm: red time at the main signal; C: cycle length. gp: green time at the pre-signal; Figure 2B—Signal Timing Comparison Oakes and Metzger (1995) studied the use of pre-signals in Ealing, a suburb of London, England. The objective was to introduce pre-signals to allow westbound buses to bypass the queue on Uxbridge Road at Park View Road and rejoin the regular traffic stream at Lady Margaret Road. The researchers noted that an important issue to consider is that the bus lane must be long enough so that a bus will not be blocked from accessing the lane by a spillback of vehicles from the pre-signal. At the highest level of priority, buses would be recognized on their approach and the control system would make adjustments to ensure that the bus receives green before the driver needs to slow down to stop at the pre-signal. A bus gate detects approaching buses in the bus lane and affords maximum priority to buses by providing a green signal on arrival at the stop line. 2.1.2 Transit Vehicle Detection An important attribute of transit signal priority systems is the accuracy of vehicle detection, determination of vehicle arrival and departure times and current and future queue length calculation. The accuracy of the information available to transit agencies can greatly impact the efficiency and effectiveness of transit service and operations. Transit vehicles must be detected before any priority decision process can be initiated. Arrival and departure times indicate whether a particular bus is running on schedule and, if not, how far it has deviated from the schedule. Schedule adherence and vehicle queuing are two important considerations in making a TSP decision under a conditional priority scheme. Consequently, the reliability of the system that provides schedule, travel and other traffic data to a transit agency is critical to the effectiveness of a TSP system. Vehicle detection and traffic monitoring have traditionally been accomplished using vehicle presence detectors. According to May (1990), presence detectors detect the presence and passage of vehicles over a short segment of roadway. "When a vehicle enters the detection zone, the sensor is activated and remains so until the vehicle leaves the detection zone. The 'on' time referred to as the vehicle occupancy time requires the University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 14 Transit and Light Rail Transit in the City of Richmond vehicle to travel a distance equivalent to its length plus the length of the detection zone" (May, 1990, 181). The length of time the detection zone is occupied by a vehicle is used to determine the vehicle speed and length. The headway between vehicles can be determined simply by looking at the time between successive periods of occupancy. Presence detectors are used for many purposes, such as detection loops at traffic-actuated signals and in real-time adaptive signal control systems such as SCOOT. However, the information from presence detectors is representative only of the situation that prevailed at the time the measurements were taken. Presence detectors cannot provide constant, real-time information. As a result, a series of detectors must be used to predict vehicle arrivals. Vehicle detectors are placed at the stop line to estimate queue lengths. Detectors are also placed upstream of an intersection to estimate arrivals at the next downstream intersection. However, there is uncertainty in such estimates since vehicles may be delayed between the detector and the intersection. Automatic vehicle location (AVL) systems offer an alternative to presence detectors. They have been developed more recently in conjunction with advancements in transportation technology. A V L involves constant monitoring of the location and travel time characteristics of a transit vehicle. As such, the uncertainty of a bus arrival and its location on a link is reduced. According to Chang et al. (1996), A V L can provide information on the following: the number of buses on a link; the location of a bus on a link; the schedule delay of a bus; and the speed of a bus. Gillen et al. (2001) identify the following benefits of A V L : increased dispatching and operating efficiency; more reliable service; quicker response to service disruptions; reliable information for signal priority decisions; and improved data collection. There are a number of vehicle location technologies that can be used separately, or in combination, to provide information to the A V L system. The most common technologies are dead reckoning, ground-based radio, signpost systems and global positioning systems (GPS). In dead reckoning, a transit vehicle is informed of its starting location and the vehicle then calculates the distance traveled from the starting location using its odometer and measures direction of travel with compass headings. But as Casey et al. (1998) explain, due to the need to frequently reset the equipment at known locations, dead reckoning is usually used in conjunction with other technologies. According to the Transportation Research Board (1997), a dead reckoning system is relatively inexpensive since all of the technology is contained on the transit vehicle and no external infrastructure is required. However, the accuracy of the location information degrades with distance traveled and is adversely affected by uneven road surfaces, steep hills and magnetic interference. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 15 Ground-based radio systems use a system of transmitters and receivers to triangulate the location of a transit vehicle. According to Casey et al. (1998), this type of system can locate a vehicle with a margin of error of approximately 150 feet. The Transportation Research Board (1997) identified a number of advantages and disadvantages with these systems. On the plus side, ground-based radio technology involves low in-vehicle costs. Furthermore, there are no blind spots or interference with transmissions. However, these systems require a well-equipped infrastructure of transmitters and receivers, and no data is available for areas outside of this infrastructure. In fact, the frequency of location updates depends on the density of signposts along transit routes. In a signpost system, radio beacons are placed along the transit route and are detected by receivers on passing transit vehicles. The bus then reports its position to the dispatch centre according to distance traveled beyond the previous beacon, which is taken from the odometer. Casey et al. (1998) explain that GPS uses signals transmitted by a network of 24 orbiting satellites to triangulate the location of a transit vehicle. A problem with such a system is that the satellite signals can be interrupted by tall buildings and foliage on the ground. Differential GPS (DGPS) is used to improve the position location accuracy of GPS systems. According to the Transportation Research Board (2000), a GPS receiver is placed at a stationary site where the precise location is known. The difference between the known location and its GPS-measured location is used to correct the GPS-determined position. Fox et al. (1998) describe the use of DGPS in the TSP system in Aalborg, Denmark. It is claimed that using a tachometer with DGPS reduces the error in estimating bus position to less than 10 centimetres in the vicinity of known "entry line" points. Data collected through A V L systems are transmitted to the dispatch centre through either polling or exception reporting. Casey (2000) explains that, under polling, the dispatch computer asks each transit vehicle in turn for its location. Once all the vehicles being monitored have been contacted, the computer starts again with the first vehicle and repeats the cycle. Thus, the frequency of location updates depends on the number of vehicles being monitored and the number of available communication frequencies. On the other hand, with exception reporting, each bus reports its location to dispatch only at specified points or when the vehicle is running off schedule beyond a specified tolerance level. In Milwaukee, Wisconsin, the schedule adherence criterion is 2 minutes ahead or 3 minutes behind schedule. In Portland, Oregon, the Tri-Met Transit District uses a variable on-time window that can be changed for individual buses, several buses, all buses on a route or all buses in the fleet (Casey et al., 1998). Casey (2000) explains that exception reporting makes more efficient use of available radio channels, which are often limited in urban areas. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 16 Transit and Light Rail Transit in the City of Richmond Gillen et al. (2001) studied over twenty A V L systems that have been implemented in different U.S. municipalities. In general, A V L was found to provide significant benefits to both transit agencies and consumers. Productivity gains came from better use of capital, the need for fewer buses, more efficient use of fuel and energy and reduction in the amount of labour needed for maintenance and operations. For consumers, benefits came in the form of more reliable service, shorter trip times and less waiting time at transit stops. A study by Casey (2000) found that A V L systems tend to improve transit schedule adherence. The Milwaukee County Transit System reported a 4.4% increase in schedule adherence, from 90 to 94%. Kansas City, Missouri reported a 12.5% increase in adherence, from 80 to 90%. Denver, Colorado reported a reliability increase between 12 and 21% on various routes. As "intelligent" transportation systems become more widely adopted in North America, A V L systems will become more commonplace. As a result, the data available to transit agencies will improve upon the information provided by use of traditional presence detectors. 2.1.3 Bus Schedule Reliability Levinson (1991) defines reliable bus service as service where buses run on time along a route, where the spacing between successive buses is uniform and where variations in schedule adherence are kept to a minimum. Improving the adherence of transit vehicles to pre-determined arrival and departure times is one of the most important objectives of transit signal priority because reliability is an important element in the attractiveness of public transit. Poor schedule adherence increases waiting times, makes transferring more difficult, causes uncertainty in arrival times and can erode ridership over time. Levinson (1991) explains that on bus routes with long headways, unreliable service makes transfers undependable and requires riders to take earlier trips to assure a connection. On routes with short headways, late buses are likely to be overcrowded and delays can propagate along the line so that bunching may occur. Rietveld et al. (2001) identify variability in bus travel time as a cause for concern since many travelers travel via a chain of elements, and the problem becomes particularly acute if one of the connections is missed. The problem is amplified in situations where the frequency of transit service is low and increases in waiting time are substantial. Unreliability not only leads to an increase in total travel time, but also to an increase in the percentage of time that travelers spend in the most comfortable situation—waiting at the transit stop. According to Ding and Chien (2001), unreliability is a major deterrent to transit use as riders who have long or unpredictable wait times will be discouraged from University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 17 using transit. This is important since, according to Khasnabis et al. (1993), while transit demand is elastic in terms of travel cost and travel time, transit demand tends to be more sensitive to travel time than travel cost. According to Adamski and Turnau (1998), poor transit service reliability is attributable to complexity in transit scheduling and planning due to the following: • The dynamics of traffic during the operation of transit vehicles along the routes; • High level of uncertainty in demand and human behaviour; • Unpredictable operational events and random disturbances; • Unreliable operation of the driver-vehicle entity; and • Complex interactions with other vehicles and the environment. Levinson (1991) attributes poor bus reliability mainly to traffic factors and transit factors. Traffic factors include bus delay at traffic signals, curb parking impeding bus flow, variable traffic conditions, unexpected occurrences and weather conditions. Transit factors include high complexity and long routes, stops spaced too closely or too far apart, variations in passenger arrival rates and variable ridership. Ceder (2001) explains that bus frequency is commonly set to ensure that adequate space is available to accommodate the maximum number of passengers along the entire route during a given time period. The assumption that passengers will adjust to set timetables instead of transit operators adjusting timetables to passenger demand is a main cause of unreliable service. This is critical since headway regularity is critical to prevent overcrowding and long wait times. Maintaining punctuality of both arrivals and departures can be difficult due to natural variability in the time to travel each link, congestion, differences in passenger boarding times and many other factors. Carey (1998) identifies one strategy to improve transit reliability that involves putting "slack" or "recovery time" into the schedule to allow activities to get back on schedule if they happen to be running late. Running ahead of schedule is usually much easier to remedy than running behind schedule since operators can drive more slowly or hold at a stop so as to arrive at the next stop on time. However, Carey (1998) identifies what is referred to as "behavioural response", which describes how operators and other components of the transit system respond to having more time available. In effect, adding time to complete an activity tends to result in an increase in the time required to complete the activity. Consequently, injecting "slack" into the schedule is not a very effective means of improving the efficiency and reliability of transit service. Furth and Muller (2001) explain that improving schedule adherence and headway regularity through holding vehicles at particular locations along a route is a one-sided University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 18 strategy as it can only be applied to vehicles running ahead of schedule. However, placing a lateness condition on the granting of conditional priority directly corrects both early and late schedule deviations. Vehicles running ahead of schedule are denied priority and will tend to be delayed at signals. Vehicles running late, on the other hand, are given priority and tend to move ahead. The criterion of schedule adherence with respect to transit signal priority is basically that those transit vehicles that are running behind schedule receive priority while vehicles that are on or ahead of schedule do not. The objective is to maintain uniform headways to provide stable bus frequency and occupancy. According to the U.S. Federal Transit Administration (2000), headways are naturally unstable due to probabilistic variations in dwell times at transit stations and travel times and speeds along routes. This results in bunching of buses. When a bus falls behind schedule, it tends to have more than the average number of passengers to pick up at the next downstream station. This results in a longer dwell time, which causes further delay. This continues along the route and the bus keeps falling further behind schedule. The following bus then encounters fewer than the average number of passengers and shorter dwell times, allowing it to catch up to the preceding bus. Remedying this problem is challenging because it is difficult to speed up a bus that is running behind schedule. Transit signal priority offers a solution. Headways can be regulated by giving priority to vehicles that are running behind schedule and not to those on or ahead of schedule. By regulating headways, transit operators can reduce the standard deviations in bus route travel times and arrival and departure times at transit stops. While there are other methods of regulating headways other than TSP, they are not as effective. Deliberately delaying an oncoming vehicle that is running ahead of schedule increases delays to onboard passengers and vehicle travel times. Eberlein et al. (1998) discuss "deadheading", where a transit vehicle skips a number of stations in order to save time and reduce the headway between it and the preceding vehicle. This technique usually begins at the terminal once all of the passengers have disembarked and the operator is beginning the return route. While waiting passengers beyond those stops being skipped benefit, the same cannot be said for passengers at the skipped stops. Ding and Chien (2001) explain that deadheading sacrifices passenger wait times at the skipped stops and frustrates onboard passengers whose destinations are skipped. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 19 Transit and Light Rail Transit in the City of Richmond 2.2 IMPORTANT CONSIDERATIONS IN TSP IMPLEMENTATION 2.2.1 Signal Coordination In order to coordinate traffic signals to allow for efficient progression, a fixed common cycle length and consistent green periods are used at each intersection in the corridor in order to create a through band. However, providing TSP under these conditions can result in breaking some of these constraints, which disturbs the underlying control strategy. This can adversely affect vehicle progression in a corridor, leading to increased delays and travel times. 2.2.2 Compensation Some transit signal priority strategies employ methods to compensate for time taken from other non-priority, non-bus phases in the next cycle, in order to limit the adverse effects that priority causes to non-priority traffic. Compensation is the re-allocation of green time to a non-priority phase that was truncated or skipped to give transit priority. Chada and Newland (2001) note that when pre-emption takes place at highly congested and saturated intersections, pre-emption effects often continue for several cycles. If priority interruptions are frequent, the signal controller may never be able to fully recover. The Intelligent Transportation Society of America (2002) recommends the use of "signal recovery" so signals can transition back to normal signal operation or compensate signal phases that were cut short or skipped during the priority event. This transition can take several cycles and is one reason why pre-emption is not a popular strategy. 2.2.3 Location of Transit Stops Several researchers recommend using far side bus stops in conjunction with active signal priority to ensure that signal priority calls are not wasted as transit vehicles dwell at bus stops. Garrow (1998) suggests that the presence of a near side bus stop greatly hinders the effectiveness of green extensions since a significant portion, i f not all, of the green extension is wasted while passengers board and deboard at the near-side bus stop. Transit signal priority is much more successful when used with far side bus stops since priority is no longer a function of bus dwell time. According to Zhang (2001), the system-wide disbenefits of TSP increase as dwell times at near side bus stops increase. If the bus stops within the detection zone, the signal timings will be adjusted to allow the vehicle to pass. However, the loading and unloading of passengers at the near side stop means that some or all of the additional green time is not utilized. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 20 Transit and Light Rail Transit in the City of Richmond 2.3 RESULTS FROM IMPLEMENTATION OF TRANSIT SIGNAL PRIORITY The main objectives of transit signal priority initiatives are to reduce travel times, standard deviations of travel times and delays to vehicles and persons. The following cited studies deal with these objectives. 2.3.1 Impacts of Transit Signal Priority on Travel Time A study by Chada and Newland (2001) in Seattle, Washington concluded that TSP produced an average reduction in stops by buses of 24%, total time savings of 8% and average reduction in bus delay of 34%. These researchers concluded that cross street traffic was not significantly impacted as the reduction in cross street green time did not cause any cross street vehicles to wait for more than one green signal. Chada and Newland (2001) employed a strategy with the following rules: traffic signals shall not be shortened beyond minimum or clearance intervals; traffic signals shall not skip any phases; and traffic signals shall not break coordination to provide TSP. However, they suggested that any effective TSP must be designed for each specific region and no one plan can be effectively implemented in all situations. The Metropolitan Atlanta Rapid Transit Authority (MARTA) (1999) Signal Priority Control System (SPCS) uses red truncation to reduce bus travel times. In a three-week test of the system, there were reductions in both inbound and outbound trips to the region. After implementation of the SPCS, inbound trips were reduced to an average of 28 minutes from 41.8 minutes, or a 33% reduction. Outbound trips were reduced by 17% from 33.1 minutes to 27.5 minutes. Kloos (2002) describes a pilot project undertaken in Portland, Oregon. The implementation of TSP reduced bus travel time in the peak direction during the peak periods by 10% and improved on-time transit arrival performance by 8 - 10%-According to Fox et al. (1998), a conditional TSP system in Lyon and Toulouse, France uses various criteria, including delay to transit vehicles while minimizing the disruption to the background signal timing plan. Upon implementation of this system, statistically significant reductions in transit travel time of 11 to 14% were recorded with very little change in general traffic travel times. Moreover, there was a 19 to 29% reduction in the standard deviations of bus journey times. Hu et al. (2000) discuss the TPS project on the Metro Rapid Bus System in Los Angeles, California. The project covers two of the most heavily traveled transit corridors in the city. The system uses early green/red truncation, green extension and phase call/insertion strategies. This project recorded bus travel time savings ranging from 8 to 10% in these corridors as a result of TSP implementation. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 21 2.3.2 Impacts of Transit Signal Priority on Delay Chang et al. (1995) used TRAE-NETSIM to simulate a test intersection. The researchers employed a conditional TSP strategy with the use of a performance index method. They compared the total delay at the intersection for low, medium and high volume scenarios while toggling between a bus frequency of 1 every 3 minutes and 1 every 2 minutes. For the low volume scenarios, the algorithm without pre-emption produced 80 - 90% more delay than with pre-emption. Under high volumes, the benefits of TSP were less noticeable. This is due to the fact that, in high congestion, bus passengers must compete with the passenger car queues for priority and, therefore, absolute priority is not given to a bus. Table 2-4 illustrates the total delay experienced by a bus at the simulated intersection. Table 2-4 Total Delay Results for TSP Simulat ion Traffic Volume Delay (vphpl) Mean Bus Discharge Headway (sec) Total Delay (sec) w/o pre-emption With pre-emption 300 180 24,342 3,483 300 120 25,470 4,833 500 180 34,044 18,609 500 120 33,303 22,020 1000 180 57,990 57,195 1000 120 60,372 55,869 Davol (2001) studied the effects of implementing TSP in a bus rapid transit network in Stockholm, Sweden. The buses may use green extension and non-priority phase truncation at all intersections. However, phase insertion is available only at a limited number of intersections. Coordination was designed based on automobile progression, but the low volumes on the side streets create a wide green band that transit vehicles can potentially use. Davol (2001) suggested that i f impacts on non-priority signal phases are too great, increasing minimum times instead of prohibiting TSP will lessen the impact while maintaining travel time savings for buses. Table 2-5 illustrates the average travel time results of the case study. Table 2-6 provides details of the standard deviation of travel time under each of the TSP strategies. The standard deviation of travel time is a direct indicator of the reliability of transit service. The results of the study indicated that the use of non-priority phase shortening with insertion of an extra priority phase provides the lowest standard deviation of travel time and, therefore, the most reliable transit service. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 22 Transit and Light Rail Transit in the City of Richmond Table 2-5 Average Travel Time Comparison wi th Different TSP Strategies TSP Implemented Average Travel Time (seconds) Buses Other Vehicles None 321 118 Green extension 320 117 Shortening of Phase 319 118 Insertion of Phase 312 118 Extension + shortening 313 118 Extension + insertion 309 117 Shortening + insertion 288 118 Table 2-6 Standard Deviation of Travel Time for Different TSP Strategies TSP Implemented Standard Deviation of Travel Time (seconds) Buses Other Vehicles None 45 46 Green extension 40 46 Shortening of Phase 42 46 Insertion of Phase 35 46 Extension + shortening 38 46 Extension + insertion 34 45 Shortening + insertion 27 46 Hunter-Zaworski et al. (1995) tested the effectiveness of green extension and queue jump as priority techniques. A queue jump refers to situation similar to a pre-signal where a bus at a red light at the stop bar receives an advanced green so it can pull in front of the parallel stopped queue. When these techniques were used concurrently, bus travel times decreased in the peak period in the peak direction as follows: 5% in A M (inbound); and 7.8% in P M (outbound). There was also a 12.3% decrease in person delay for bus passengers. Moreover, there was no significant change in overall total intersection delay. Khasnabis et al. (1997) studied the impact of signal pre-emption on bus travel times. They applied the criterion that an approaching bus must simultaneously need and qualify for preferential treatment before pre-emption can be granted. The system logic recognizes that not all buses that need pre-emption will qualify for it because of maximum specified limits on pre-emption. The compensation strategies used were the following: • No compensation; • One cycle compensation: decrease in green time on cross street is compensated in the next cycle; the ratio of green times on the main and cross streets remains unchanged over a period of 2 cycles; and • Two-cycle compensation. In 10 test cases, the bus route travel times all decreased, with a range from 0.3 to 13.5%. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 23 The King County Department of Transportation (2002), in its Route 358-Aurora Avenue North Transit Signal Priority Study, reported the following results in the Rainer Avenue South Corridor upon implementation of TSP: bus delay reduced by 34%; traffic signal-related stops reduced by 24%; travel time savings of 8%; no side street cycle failures; and less than 4 seconds per vehicle increase in side street delay. Skabardonis (2000) used a 6.7 km test segment that included 21 signalized intersections. Optimal timing plans to favour buses along the arterial reduced the delay to buses by 14%, reduced the number of stops by 1% and improved the average bus speed by about 4%. This translates into delay savings of about 2 seconds per bus per intersection. Minimal impacts were observed on the rest of the traffic stream (1% increase in total delay). Sensitivity analyses performed showed that estimated transit improvements are insensitive to a range of bus volumes up to 30 buses per hour. Sunkari et al. (1995) studied the following TSP techniques: 1. No priority given; 2. Pre-defined maximum extension to the priority phase (10 seconds in this model); 3. Maximum early start for the priority phase. The model was developed using the delay equation in the 1985 H C M . The magnitude of delay savings to the bus depends on the time at which it arrives at the intersection or is detected. If it arrives during the green portion and can pass safely through the intersection without an extension, then there are no delay savings. The researchers found that the model gives results that are reasonably close to those observed in the field. In the test, the southbound direction on the link was given bus priority. Table 2-7 illustrates compares the average delay as measured in the field and simulated in the model for each of the three TSP techniques. Table 2-8 illustrates the total intersection delay for each technique. Table 2-7 Average Delay Comparison Between Field and Model Observat ions Case Approach V/C Ratio Field Delay (sec/veh) Model Delay (sec/veh) 1 NB 0.39 11.7 12.0 SB 0.43 6.8 12.0 EB 0.73 21.7 41.8 WB 0.71 35.5 46.9 2 NB 0.37 13.1 11.0 SB 0.49 4.4 11.0 EB 0.97 39.9 77.7 WB 0.79 39.1 55.4 3 NB 0.39 9.9 10.4 SB 0.42 4.9 10.9 EB 0.86 55.2 57.8 WB 0.85 40.6 69.7 University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 24 Transit and Light Rail Transit in the City of Richmond Table 2-8 Total Intersection Delay Comparison Case Field Model 1 19.1 21.1 2 18.1 26.0 3 19.2 20.2 In a study cited by Zhang (2001) and undertaken by Westwood Professional Services (1995), low and medium priority schemes that use special priority phases and extensions produced no reductions in travel times. A high priority scheme giving signal pre-emption resulted in a 38% reduction in bus travel times, but also a 23% increase in automobile delay. A study of a conditional priority strategy employed in Lyon and Toulouse, France by Forges and Henry (1994) attempted to minimize the delay to transit vehicles while minimizing the disruption to the background phase sequence. The study produced reductions in transit travel times of 11 to 14%. Furth and Muller (2001) studied the bus priority system in Eindhoven, Netherlands. A l l the buses in the system have on-board computers that track the vehicle location and trip time details. Since traffic is mixed, granting TSP means giving priority to an entire regular traffic stream. Constraints ensure that minimum 6-second green times and minimum red and yellow clearance times are enforced. The TSP scheme uses green extension, red truncation and phase skipping. Priority interruptions are compensated for by giving streams the minimum green times until the signal timing program recovers. The researchers compared absolute priority with conditional priority. While absolute priority increased average vehicle delay by 40 seconds compared to no priority, conditional priority did not significantly change the average delay. In addition, absolute priority increases the percentage of lost time in each cycle by 34%, from 29% to 39%. On the other hand, conditional priority only increases the lost time by 4% in the a.m. peak period and 2% during the p.m. peak period. Finally, under absolute priority, non-priority approaches lost more than 20% capacity while only 5% of capacity was lost under conditional priority. Head (2002) evaluated the effectiveness of TSP using simulation. With the baseline network and coordinated signals, the average bus delay at signals was 81.62 seconds per route. The use of green extension for transit reduced the delay to 77.35 seconds per route. Bus travel time was reduced from 417 seconds per route with baseline conditions to 401.12 seconds with green extension implementation. Green extension also reduced average car delay from 33.48 seconds to 32.76 seconds. Although full signal pre-emption reduced bus travel time to 362.6 seconds, the average car delay was increased from 33.48 seconds to 34.94 seconds. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 25 The Selective Priority Network Technique (SPRINT) has been implemented in London, England. This technique attempts to reduce bus delay at an intersection, subject to the following constraints: a maximum number of cycles than can be run during a time interval with signal timings different than the base plan; a maximum time difference of a stage from the base plan; and maximum levels of saturation. In addition, TSP is not implemented in consecutive cycles. According to Fox et al. (1998), on a 3 kilometre segment of Uxbridge Road containing eight intersections, the technique produced a 2-second reduction in bus delay per intersection on the main road and a 6.4-second reduction in bus delay per intersection on the side road links. 2.4 Summary of Transit Signal Priority Literature Results Table 2-9 summarizes the results of some of the studies performed in the literature reviewed above. Table 2-9 Summary of Literature Review Study Results Reference Strategy Results Al-Sahili and Taylor (1996) • Green extension/red truncation with no compensation • Green extension/red truncation with compensation • Phase skipping without compensation • Phase skipping with compensation • 2.3% reduction in bus travel time; 0.4% increase in total travel time; 1.1% increase in total delay • 0.5% reduction in bus travel time; 0.7% increase in total travel time; 1.5% increase in total delay • 9.2% reduction in bus travel time; 12.3% increase in total travel time; 14.3% increase in total delay • 0.6% increase in bus travel time; 14.6% increase in total travel time; 19.6% increase in total delay Boie and Nookala (1996) • Green extension and red truncation. • No noticeable change in bus delay; 23% increase in traffic delay. Chada and Newland (2001) • Green extension and red truncation with maximum durations • average reduction in stops for buses of 24%; total bus travel time savings of 8%; average reduction in intersection bus delay of 34%. Davol (2001) • Green extension (1) • Red truncation (2) • Phase insertion (3) • Green extension with red truncation (4) • Green extension with phase insertion (5) • Red truncation with phase insertion (6) • (1) 0.3% reduction in bus travel time; 11.1% reduction in standard deviation of travel time; • (2) 0.6% reduction in bus travel time; 6.7% reduction in standard deviation of travel time; • (3) 2.8% reduction in bus travel time; 22.2% reduction in standard deviation of travel time; • (4) 2.5% reduction in bus travel time; 15.6% reduction in standard deviation of travel time; • (5) 3.8% reduction in bus travel time; 24.4% reduction in standard deviation of travel time; University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 26 • (6) 10.3% reduction in bus travel time; 40% reduction in standard deviation of travel time; • none of the strategies had a noticeable impact on non-priority traffic. Duncan and Mirabdal (1996) • Green extension and red truncation. • 6 - 25% reduction in transit signal delay. Forges and Henry (1994) • reductions in transit travel times of 11 to 14%. Fox et al. (1998) • Green extension, red truncation. • 11 - 14%o reduction in transit travel time; • 19 - 29% reduction in variability of bus travel times. Furth and Muller (2001) • Green extension, red truncation and phase skipping; • Absolute priority v. conditional priority. • absolute priority increased average vehicle delay by 40 seconds compared to no priority; conditional priority did not significantly change the average delay; absolute priority increases the percentage of lost time in each cycle by 34%; conditional priority only increases the lost time by 4% in the a.m. peak period and 2% during the p.m. peak period; under absolute priority, non-priority approaches lost more than 20% capacity while only 5% of capacity was lost under conditional priority. Head (2002) • Green extension, red truncation. • 5.2% reduction in bus delay; • 3.8% reduction in bus travel time. Hu et al. (2000) • Green extension, red truncation, phase insertion. • 8 - 1 0 % reduction in bus travel time. Hunter-Zaworski (1995) • Green extension + queue jump • 5% decrease in inbound bus travel time in A M ; and 7.8% decrease in outbound bus travel time in P M ; 12.3% decrease in person delay for bus passengers; no significant change in overall total intersection delay. Khasnabis et al. (1997) • Absolute priority through pre-emption • Inconsistent results: range from 0.3% to 13.5% reductions in bus travel times. Khasnabis et al. (1996) • Green extension, red truncation. • 14% reduction in bus delay; 1% reduction in number of stops; average bus speed increased by 4%; minimal impacts on rest of the traffic stream (1% increase in total delay). King County Department of Transportation (2002) • Bus delay reduced by 34%; traffic signal related stops reduced by 24%; travel time savings of 8%; no side street cycle failures; and less than 4 seconds per vehicle increase in side street delay. King County DOT and City of Seattle (Transit Signal Priority . . .) (2000) • Green extension, red truncation. • 24% reduction in stops for TSP eligible buses; 8% reduction in travel times; 34% reduction in average intersection bus delay for TSP eligible buses. King County DOT and City of Seattle • Green extension, red truncation. • 50%) reduction in signal-related stops; 51% reduction in average traffic signal delay for prioritized buses; 13.5% University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 2 7 Transit and Light Rail Transit in the City of Richmond (Preliminary Transit. . .) (1999) decrease in intersection average person delay; insignificant side street effects. Kloos (2002) • Green extension, red truncation. • 10% reduction in bus travel time in peak direction during peak periods; • 8 - 1 0 % improved schedule adherence. Lewis, V . (1996) • Green extension, red truncation. • bus travel time savings of 1.4 - 6.4%; average bus signal delay reduction of 20%. Metropolitan Atlanta Rapid Transit Authority (1999) • Red truncation • inbound trips reduced by an average of 33%; outbound trips were reduced by 17%. Vahidi, H . (2000) • Green extension, red truncation. • 15 - 49% reduction in transit signal delay; schedule reliability improved. Westwood Professional (1995) • Insertion of special priority phases; green extension. • Signal pre-emption • medium priority scheme using special priority phases and extensions produced no reductions in travel times; high priority scheme giving signal pre-emption resulted in a 38% reduction in bus travel times, but also a 23% increase in automobile delay. Wu and Hounsell (1998) • Pre-signals: category A • Pre-signals: category B • 4.2 - 9.6% reduction in bus delay; 0.3 -1.14% increase in non-priority vehicle delay. • 89 - 154% reduction in bus delay; 3.08 -4.16% increase in non-priority vehicle delay. 2.5 Light Rail Transit Signal Priority Considerations The Transportation Research Board (2000) defines light rail transit as: " A metropolitan electric railway system characterized by its ability to operate single cars or short trains along exclusive rights-of-way at ground level, on aerial structures, in subways, or occasionally, in streets and to board and discharge passengers at track or car floor level" (p. 3). Light rail has a smaller passenger capacity than traditional heavy rail. Heavy rail can haul 6 to 10 cars, providing capacity for up to 50,000 people per hour. LRT, on the other hand, has a maximum capacity of 10,000 to 15,000 people per hour per direction. According to the U.S. Department of Transportation (2000), heavy rail can be cost-effective when employment and residential densities are high enough to promote very high ridership. However, LRT is often more practical in cities with lower densities. LRT fell into relative obscurity in the second half of the 20 t h century but it is beginning to exude new life. Light rail transit (LRT) has seen resurgence in many cities around the world. New LRT lines are being built and existing lines are being upgraded, reconstructed or extended. The Transportation Research Board (2000) differentiates between "First Generation" and "Second Generation" systems. First Generation systems are those that evolve from earlier trolley and tramway lines that remained in constant University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 28 Transit and Light Rail Transit in the City of Richmond operation over time. Second Generation systems are newly designed, sometimes using portions of abandoned trolley or railway lines. The Transportation Research Board (2000) outlines the following basic elements of all LRT systems: • Infrastructure: tracks, stations, storage yards, tunnels and bridges; • Rolling Stock: the rail car fleet; • Fixed Equipment: an operations and maintenance center, the electric power supply, signal and communications facilities (p. 10). A surface LRT line may be completely physically separated from other transportation modes and their users with the use of bridges and overpasses. On the other hand, rights-of-way of other transportation modes may cross the LRT line at grade, which can lead to serious conflicts if proper traffic control measures are not in place. One of the advantages of LRT is the flexibility of design, including track geometry, station location, station design and other characteristics. For example, an on-street LRT line is affected by the presence of buildings and other structures, which may require the line to include sharp turns and steep grades. According to the TRB (2000), it is possible to design LRT cars that can negotiate curves with radii as short as 11 metres. The cars themselves vary in size depending on the necessary passenger capacity of the system and the geometry of the line itself. According to Rymer et al. (1989), the attractiveness of LRT lies in its potentially lower implementation costs compared to heavy rail. One of the key elements of this lower cost is the fact that light rail transit lines do not need to be completely grade separated from automobile traffic. Parkinson (1989) outlines some of the other advantages of light rail. Because light rail vehicles are smaller than heavy rail vehicles, tunnels and underpasses can be smaller, which reduces capital construction costs. LRT can usually be planned and built in less time than heavy rail, with fewer environmental impacts than other types of rapid transit. Furthermore, wide, multiple doors allow for faster boarding and alighting than buses, which decreases dwell and travel times and makes transit service more reliable. Finally, the rapid acceleration and deceleration capabilities of LRVs allow an LRT line to share a corridor with other transportation modes. The economy of LRT comes largely from this ability to use existing rights-of-way, including railroads, power lines and medians of streets and freeways. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 29 Transit and Light Rail Transit in the City of Richmond 2.5.1 LRT Operating Properties 2.5.1.1 Rates of Acceleration and Deceleration The rate at which transit vehicles accelerate and decelerate depends heavily on the operator. The operator may increase or decrease the rate of acceleration or deceleration continuously throughout the course of a trip. While it is possible to determine the maximum acceleration and deceleration capabilities of a transit vehicle, and even the average rates over a number of trips, it is difficult to analyze the acceleration or deceleration of an individual transit vehicle at a certain period in time. This is significant since variation in the rates of acceleration and deceleration affects travel times and arrival times at transit stations, which can adversely affect the reliability of the transit service. Bauer et al. (1995) suggest that the service acceleration rate for a typical L R V is 2 2 approximately 1.1 m/s with a deceleration rate of approximately 1.4 m/s . These values are comparable to those of the Transportation Research Board (1995), which suggested a 2 2 service acceleration rate of 1.3 m/s and a service brake rate of 1.5 m/s . Table 2-10 outlines the service acceleration and deceleration rates of a number of LRT systems currently in operation. It is important to note that the service acceleration and deceleration rates are significantly lower than the L R V is capable of applying. Table 2-10 Comparison of LRT Accelerat ion and Deceleration Rates City Service Acceleration Rate (m/s2) Service Deceleration Rate (m/s2) Calgary 1.1 1.3 Grenoble 0.92 1.2 Nantes 1.1 1.2 Kuah and Allen indicated that LRT cruise speeds in the range of 40 to 50 km/h are generally high while low cruise speeds range from 25 to 32 km/h. The typical rate of acceleration is approximately 0.84 m/s and the typical rate of deceleration is approximately 0.76 m/s . In order to effectively implement transit signal priority, It is necessary to have some method of regulating acceleration and deceleration rates. A relatively accurate estimate of these rates is necessary for calculation of bus arrival and travel times, as well as the appropriate location of transit vehicle detectors in the roadway. The accuracy of travel time estimates is also important in absolute priority strategies that modify signal timings so that the traffic controller cycles through signal phases to provide a green phase to the transit vehicle before it begins decelerating. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 30 Transit and Light Rail Transit in the City of Richmond 2.5.1.2 Travel Speed Travel speed, like acceleration and deceleration rates, determine bus travel times and transit vehicle arrival times. However, speed is not a constant. Rather, it varies between drivers and transit vehicle types. The reliability of transit service can be negatively affected i f actual travel speeds are significantly different from those used for travel time and schedule calculations. Average travel speed is often used in calculating travel times. Thus, it is important that operators adhere as closely as possible to the modeled travel speed, as is the case with acceleration and deceleration rates. Table 2-11 summarizes average speeds observed on various LRT systems. Table 2-11 Compar ison of Average LRT Travel Speed in Different Cities City Average Speed of System (km/h) Calgary 29 Cleveland 30 Edmonton 30 Newark 34 Philadelphia 26 Portland 30 Sacramento 34 San Diego 29 San Jose 32 Grenoble 17-20 Nantes 21 -23 Nieuwigen 29 Hannover 24 Stuttgart 25 Stockholm 26 The LRT systems in Calgary, Ottawa, Nieuwigen, Hannover, Cologne, London and Stockholm operate with a maximum speed of 80 km/h. According to Imhoff (1989), the Denver LRT vehicles are capable of reaching speeds of 88.5 km/h. The systems in the French cities of Grenoble and Nantes operate at maximum speeds of 70 km/h. However, in all of these systems, the regular cruising speed is lower than the maximum, in the range of 50 to 60 km/h. 2.5.1.3 Headways Operating headways between transit vehicles vary with the time of day, depending on passenger demand. The headways are intended to provide a certain level of service to passengers during a particular time period. When these headways are disturbed, the level of transit service declines. When buses begin to bunch up, the lead vehicle is forced to University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 31 Transit and Light Rail Transit in the City of Richmond carry heavier-than-average passenger loads while following vehicles carry lighter-than-normal loads. This results in a continual decrease in the headways between transit vehicles, which causes buses to be late and increase passenger wait times. Headways are important because they determine bus occupancies, passenger wait times and the number of transit vehicles required to provide a particular level of transit service. Consequently, the data that is used to calculate headways is critical and must be as true to life as possible. Table 2-12 compares the headways used in a number of LRT systems during different time periods. Table 2-13 compares peak period headways of different LRT systems. Table 2-12 Comparison of LRT Headways in Different Cities City Headways Calgary Peak = 5 min Off-peak = 15 min Min. practical = 2 min Ottawa Peak = 20 min Off-peak / weekend = 40 min Edmonton Peak = 5 min Off-peak = 10 min Evening and weekend = 15 min Denver Peak = 5 min Off-peak = 10 min Morning = 15 min Evening = 30 min Sacramento Peak = 15 min Off-peak = 15 min Weekend and Holiday = 30 min Salt Lake City Peak = 10 min Off-peak = 15 min TRAX (Utah) Peak =10 min Off-peak = 15 min Evening = 20 min University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 32 Transit and Light Rail Transit in the City of Richmond Table 2-13 Comparison of Peak Headways in Different Cities Ci ty Peak Headway (minutes) L A - Blue Line 6 L A - Green Line . 5 Portland 3 Baltimore 15 Buffalo 5 Denver 5 Sacramento 15 San Diego 4.25 St. Louis 7.5 Boston 7.5 Philadelphia 3 San Francisco 10 Cleveland 6 Pittsburgh 3 San Jose 10 Calgary 5 Ottawa 20 Edmonton 5 Salt Lake City 10 2.5.2 LRT Design Parameters 2.5.2.1 L R V Specifications The physical design of a transit vehicle is important in determining transit station size and configuration, passenger capacity and necessary headways to provide a desired level of service. Boarding and alighting times for passengers are also based on floor height, door width and number of doors. The LRVs used on different LRT systems vary slightly in their design parameters. According to the Transportation Research Board (2000), body widths vary from 2.6 to 2.9 metres and lengths of one-piece cars range from 15 to 20.4 meters. Table 2-14 compares the size of different LRT system vehicles. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 33 Transit and Light Rail Transit in the City of Richmond Table 2-14 Comparison of Light Rail Vehicle Dimensions City Vehicle Type Dimensions Calgary U2 24.4 m * 2.65 m Calgary SD160 24.82 m * 2.65 m Ottawa Bombardier BR643 48 m * 2.9 m Britain - 15 m * 7.5 m Stockholm - 30 m * 2.65 m Porto - 34.75 m * 2.65 m Grenoble - 29.4 m * 2.3 m Nantes - 28.5 m * 2.3 m Nieuwigen - 28.5 m * 2.3 m Hannover - 28 m * 2.4 m Cologne High floor (K5000) 29.4 m * 2.65 m Cologne Low floor (K4000) 28.4 m * 2.65 m London High floor 38.8 m * 2.65 m London Low floor 30.1 m * 2.65 m Rotterdam - 30.5 m * 2.664 m Minneapolis - 28.65 m * 2.668 m Saarbrucken - 37.1m* 2.65 m Stockholm - 29.7 m * 2.65 m An important consideration in the design of an L R V is whether a low floor or high floor design is implemented. Conventional LRT vehicles have a high floor, which requires steps down to low platforms or on-street stations. Conventional LRVs have a disadvantage compared to low floor vehicles when passengers are boarding from street level or low platforms, especially elderly or disabled passengers. Conventional LRVs require longer dwell times than low floor LRVs at stops with low-level platforms, which increases the overall travel time. The Transportation Research Board (1995) provides a general dwell time estimate of 8 to 20 seconds, depending on the load situation. However, a single wheelchair boarding or alighting to a high-floor L R V from low-level stations can take 2 to 4 minutes. This can be problematic where train headways are very small, causing trains to run behind schedule and with inadequate headways. According to the Transportation Research Board (1995), when implementing a new LRT system, using low floor LRVs in conjunction with low-level platforms, such as on-street and curb stations, can provide more reliable service and shorter trip times. This is due to shorter dwell times at stations. In addition, i f the reduction in trip time is equal to the operating headway, an L R V can be eliminated without affecting the system level of service, which can reduce capital costs. Operating costs would also be reduced from savings in labour no longer required to operate and maintain the eliminated train. Low floor LRVs can be built to almost any size, depending on the required passenger capacity and any restrictions imposed by the infrastructure. Existing stations and University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 34 platforms may have to be modified to increase platform lengths and to prevent blocking of intersections and access points. 2.5.2.2 Station Design and Spacing The Transportation Research Board (1996) indicates that, as the first point of contact between passenger and transit service, the transit station is critical to the system's overall goal of providing timely, safe and convenient transportation. Parkinson (1995) discusses flexibility in LRT station design and location, which can be seen in the fact that a station can be as short as one L R V car, it can be located on a curve in the track and the station can be as minimal as a transit stop sign. An important consideration in designing an LRT line is the number of stations and/or stops that need to be provided. This will be a function of the passenger demand along the line and the locations of major residential and economic districts. For example, Claflin (1995) explains that the Denver LRT line is 8.5 km (5.3 mi) in length, with 14 passenger stations and 24 platforms. The Edmonton LRT line is 12.3 km in length and consists of 6 underground stations and 4 surface stations, in conjunction with several park-and-ride lots (McLachlan, 1995). According to Bertini, Botha and Nielsen (1995), the Silicon Valley LRT in San Jose had an initial line of 33.8 km in length and 33 stations and 11 park-and-ride lots. The Tasman Corridor extension is 20 km in length and has 18 new stations. The distance between stations and stops on a transit line is important because it affects the level of transit service, passenger walk times, travel time and capital and operating costs. The spacing must be based on the objective of the transit service provider. Determining the appropriate spacing between transit stops involves a trade-off between the following situations: 1. Close stops, short walk distances, but more frequent stops and, thus, longer trips; 2. Stops farther apart, longer walk distances, but more infrequent stops, higher speeds and shorter travel times. Schumann (1988) studied the average spacing between LRT stations on different lines, and his results are summarized in Table 2-15. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 35 Transit and Light Rail Transit in the City of Richmond Table 2-15 Average Station Spacing City Avg. Stn. Spacing (km) Calgary 0.9 Cleveland 0.8 Edmonton 1.3 Newark 0.6 Philadelphia 0.4 Portland 1.0 Sacramento 1.0 San Diego 1.5 San Jose 1.0 Average 0.944 Station spacing will vary depending on the operating environment. The Transportation Research Board (1996) suggests the spacing parameters outlined in Table 2-16 for different operating environments: Table 2-16 Environment Spacing Range Typical Spacing C B D 92 - 305 m 183 m Urban 152-366 m 229 m Suburban 183-762 m 305 m Rural 198-805 m 381 m Table 2-17 City Track Length (km) # Stations Average Spacing (km) Baltimore 81.9 32 2.56 Boston 124.7 95 1.31 Buffalo 22.7 14 1.62 Cleveland 53.1 34 1.56 Dallas 75.2 20 3.76 Denver 45.9 20 2.295 Galveston 7.9 3 2.63 Kenosha 3.1 1 3.1 Los Angeles 137.9 36 3.83 Memphis 9.8 28 0.35 New Orleans 22.0 9 2.44 Newark 40.1 23 1.74 Philadelphia 275.2 64 4.3 Pittsburgh 72.1 13 5.55 Portland 115.7 47 2.46 Sacramento 63.4 29 2.19 St. Louis 58.3 18 3.24 Salt Lake City 39.6 16 2.48 San Diego 155.5 49 3.17 San Francisco 112.7 11 10.25 San Jose 90.6 47 1.93 Seattle 3.4 9 0.38 University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 36 Transit and Light Rail Transit in the City of Richmond 2.5.2.3 Station Placement According to the TRB (1996), the critical factors in transit stop placement, other than ridership potential, are safety and avoidance of conflicts that would impede bus, car or pedestrian flows. From a safety perspective, transit stops should be located so as to protect passengers from passing road traffic, to provide access for people with disabilities and to be in close proximity to passenger crosswalks and curb ramps. The stops should also be located close to major trip generators. In addition, the transit stop location should allow for convenient passenger transfers to other routes and modes with nearby stops. This would include the stop for the same transit route in the opposite direction. Transit stops should also be located with consideration for transit and automobile traffic operations. There should be adequate curb space at the stop for the number of transit vehicles expected to be at the stop at the same time. The proximity of on-street parking and truck delivery zones, type of traffic control at nearby intersection, traffic volumes and the proximity and traffic volumes of nearby driveways and access points must also be taken into consideration in determining the transit stop location. A transit stop can be located on the near or far side of an intersection or at a mid-block location. Each of these locations has advantages and disadvantages. The use of far side transit stops minimizes conflicts between right-turning vehicles on the same approach and transit vehicles moving through the intersection and provides additional right turn capacity. According to the TRB (1996), far side bus stops reduce the deceleration distances of transit vehicles since they can use the intersection as part of the necessary deceleration distance. Far side bus stop placement also creates gaps in traffic for buses to rejoin the traffic flow. However, far side bus stops can cause stopping buses to back up and block the intersection during peak periods. Also, stopped buses may obscure sight distances for crossing vehicles and pedestrians. Far side stops can also cause transit vehicles to stop twice: first at a red light and then at the far side stop. Finally, far side stops can increase rear-end motor vehicle accidents if drivers do not expect buses in front of them to stop. Near side transit stops minimize interference when automobile traffic is heavy on the far side of an intersection. Furthermore, near side stop location allows for boarding and alighting while the transit vehicle is stopped at a red light. However, near side stops increase conflicts with right-turning vehicles on the same approach and reduce the capacity of the right-turn lane. In addition, near side stops can render transit signal priority ineffective i f the stop dwell time causes green extensions to expire before transit vehicles can clear the intersection. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 3 7 Transit and Light Rail Transit in the City of Richmond The final option is to place transit stops at mid-block locations. While this location minimizes sight distance problems for vehicles and pedestrians, it encourages mid-block passenger crossings (jaywalking) and increases the walking distance for passengers that want to cross at the nearest intersection. 2.5.2.4 Integration with Other Transportation Modes Unless a transit vehicle operates in an exclusive, segregated right-of-way, its operation will have some impact on other modes of transportation. Many transit systems operate in shared or semi-exclusive rights-of-way where an L R V must interact with automobiles, pedestrians, cyclists, buses and other transportation modes. The TRB (1996) outlines the following planning principles and guidelines for LRT systems: 1. Respect the existing urban environment; 2. Comply with motorist, pedestrian and L R V operator expectancy; 3. Strive to simplify decisions and minimize road-user confusion; 4. Clearly transmit the level of risk associated with the surrounding environment; and 5. Provide recovery opportunities for errant pedestrians and motorists. This interaction between modes can lead to conflicts that adversely affect traffic safety and operations. Thus, it is important that measures are taken to provide transit service with the least disruption to other traffic operations. Korve, Farran and Mansel (1995) explain that the design and control of an LRT system should comply with motorists and pedestrians' expectations. Violation of these expectations can result in conflicts between the LRT and other modes. The decisions that users of other modes have to make in interacting with the L R T should be simplified as much as possible to prevent confusion. Korve et al. (1995) suggest physical separation of automobile operations and LRT operation when these modes share a street right-of-way. As Wang et al. (1998) explain, when an L R V shares an on-street right-of-way, and the L R V does not have an exclusive travel lane, disruption to the traffic flow is initiated when the L R V stops at a station. The scale of the disruption increases with dwell time, bus frequency and traffic flow. The disruption can increase travel time and delay for users of transportation modes other than LRT. Bates and Lee (1989) identify the following potential impacts of an at-grade LRT system: reduced roadway space; reduced on-street parking space; reduced accessibility to adjacent land uses; left turn parking restrictions; diversion of automobile traffic to parallel arterials; and increased automobile travel time and delays. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 38 Transit and Light Rail Transit in the City of Richmond Another important element of integrating LRT with other transportation modes involves access mode planning to allow for transfers between modes. Hubbell et al. (1989) stress the importance of planning at LRT stations for effective coordination between feeder bus lines and LRT to optimize transfers. This planning strategy also includes park-and-ride lots that allow travelers to drive to LRT stations and then use LRT and other public transit modes for the remainder of the journey. According to the Transportation Research Board (2000): "Adding LRT trunk lines and coordinating them with a region's buses to create a multimodal, multidestination transit system results in growth for both modes— even in the low-density, auto-oriented cities of the American West" (p. 20). While completely grade separating an LRT line would eliminate conflicts with other transportation modes, the economic and environmental cost would be much greater than an at-grade LRT system. When an LRT shares a right-of-way with automobile traffic, the LRVs can run either in the median of a roadway or along the side. Bates and Lee (1989) recommend using median-running on a two-way street and side-running on a one-way street. Side-running on two-way streets can create conflicts between the LRVs and automobiles entering the street from minor streets or driveways. Moreover, on-street parking would have to be eliminated. According to Taylor, Lee and Tighe (1989), conflicts between LRVs and automobiles are avoided by either prohibiting automobile movements across the tracks or by using separate traffic signals to control LRVs. This would involve exclusive left turn phases for vehicles moving on the same street as the L R V at an intersection, and these vehicles would not be able to turn during the green phase for the L R V . This prevents automobiles from turning in front of an L R V . 2.5.3 Light Rail Transit Signal Priority Dale et al. (1995) discussed the implementation of TSP in the Westside Light Rail Transit extension running from downtown Portland to downtown Hillsboro in Oregon. A VISSIM model was used to analyze the impacts of granting full signal priority to LRT vehicles in Hillsboro. The study found that full signal priority could be implemented with the existing LRT headways and traffic volumes while producing only slight increases in average intersection delay. However, back-to-back train arrivals compounded queue spillbacks downstream and caused increased delay to successive LRVs. Furthermore, the maximum queues at some intersections in the study area began to exceed storage capacity with implementation of full TSP. As a result, the researchers recommended using a conservation TSP strategy that prevented back-to-back train scenarios. Tables 2-18 and 2-19 outline average intersection stopped delay and maximum queue lengths for various locations in the study area. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 39 Transit and Light Rail Transit in the City of Richmond Table 2-18 Average Stopped Delay Intersection A M Peak Period PM Peak Period Without LRT With LRT Without LRT With LRT l s , /Main 10.4 11.2 18.3 20.8 1st /Washington 8.6 8.6 13.4 18.4 1 s t/Baseline 8.8 9.1 10.8 10.8 First/Oak 7.8 7.8 8.3 8.8 10"7Main 40.0 43.9 53.2 46.2 10"VWashington 12.6 16.8 22.9 22.0 10"7Baseline 12.5 12.6 16.6 17.7 10"7Oak 5.8 5.8 9.4 9.9 Table 2-19 Maximum Queue Lengths (feet) Intersection Direction A M Peak period PM Peak Period Without LRT With LRT Without L R T With LRT l s ' /Main N B 195 225 220 238 rv Washington N B 109 185 107 184 I s '/ Washington SB 214 298 237 400 l s t/Baseline SB 232 258 227 275 6 t h/Main N B 173 194 231 307 6'V Washington N B 112 113 193 265 6"7 Washington SB 169 219 314 376 10"7Main N B 317 365 282 422 10 th/ Washington N B 144 329 178 363 10 th/ Washington SB 236 262 211 357 10,h/Baseline SB 252 304 573 632 Korve et al. (1996) suggest that the goal of transit signal priority should be to minimize delays for transit vehicles while: maintaining essential arterial and cross street progression; providing safe clearances for vehicles and pedestrians; and minimizing delay of pre-empted motor vehicles or pedestrian movements. In order to avoid conflicts between transit vehicles and automobiles that intend to turn left in front of the transit vehicle, Korve et al. (1996) suggest separating these phases into separate signal phases operating at different times. According to the Federal Transit Administration (1998), transit signal priority will be ineffective if not implemented in conjunction with exclusive bus lanes or queue bypass lanes that allow buses to circumvent automobile traffic on an intersection approach. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 40 Transit and Light Rail Transit in the City of Richmond 2.5.4 Bus Rapid Transit—An Alternative to LRT According to Sislak (2000), BRT is a step towards developing improved public transportation systems that have the service delivery characteristics and appeal of a rail transit line at the fraction of the capital cost. The underlying concept is to duplicate the reliability, level of service, comfort and appeal of a modern LRT line while achieving the flexibility and cost-effectiveness of bus systems. Sislak (2000) suggests that BRT services should include the following: exclusive bus lanes or busways; bus signal priority or pre-emption; reliability; improved fare collection and boarding/alighting patterns; appealing bus designs and seating arrangements; and improved facilities and passenger amenities such as stations and stops. Sislak (2000) found improving bus service to be more cost-effective than building a new subway or LRT line along Euclid Avenue in Cleveland, Ohio. The improved bus service would be a 6.76 km exclusive transitway constructed in the median of Euclid Avenue, featuring automatic vehicle location, transit signal priority and passenger information systems. According to the Federal Transit Adminstration (1998), BRT would provide significantly faster operating speeds, greater service reliability and increased convenience, often matching the quality of rail transit when implemented in appropriate settings. BRT would be expected to include the following features: bus lanes or busways; bus signal priority or pre-emption; improved fare payment system to reduce boarding times; and improved boarding to make entering and exiting buses easier and faster. The FTA insists that transit signal priority can only be implemented effectively in conjunction with dedicated bus lanes or queue bypass lanes that allow buses to circumvent traffic on an intersection approach. As the FTA (1998) explains, BRT involves a trade-off between improvement in travel times by reducing the number of stops and convenient access made possible by frequent stops. However, success BRT system can be expected to produce improvements in bus service, operations and ridership. Travel times will be reduced by removing impediments from bus travel path by operating in exclusive lanes. Ridership would be expected to increase due to improved bus speeds and schedule adherence. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 41 2.6 Literature Review Chapter Summary The purpose of this chapter is to introduce the concept of transit signal priority and to discuss all of the important considerations necessary in developing and implementing a TSP strategy. This is done by reviewing an extensive number of studies undertaken with respect to TSP and providing an overview of the general findings. Transit signal priority involves modifications to background signal timing plans in order to provide priority to transit vehicles. Passive TSP involves implementing fixed signal setting that favour transit vehicles in order to reduce transit delay. On the other hand, active TSP involves dynamic signal timing modifications in real-time to reduce delay to oncoming transit vehicles. Active priority can be either conditional or unconditional. With conditional priority, TSP requests are evaluated based on pre-determined criteria and only those transit vehicles that satisfy the criteria are granted priority. With unconditional priority, all transit vehicles are granted priority, regardless of impacts on traffic operations. A major objective of transit signal priority is to improve the reliability of transit vehicles travel times, as well as arrival and departure times. Unreliable travel time make transit operations inefficient and unattractive relative to low occupancy modes. Important considerations in TSP implementation include impacts on traffic signal coordination and options for recovery, compensation for signal phases affected by TSP provision, appropriate transit stop/station locations and check-in detector locations. The literature reviewed in this paper indicates that, i f implemented properly, TSP can provide significant transit delay and travel time improvements compared to a network of fixed or actuated traffic signals. However, there can be considerable impacts on cross street traffic operations since these signal phases are disrupted with each action taken to provide signal priority. There are different considerations depending on whether a transit system uses buses or light rail vehicles. These vehicles have different rates of acceleration, speed properties, passenger capacities, station requirements and physical dimensions. A l l of these characteristics affect how TSP must be implemented. How effectively a TSP strategy operates will depend on how well the strategy accounts for the characteristics of the different transit vehicles. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 42 3. STUDY DESIGN In order to study the influence of different physical and operational parameters on the performance of transit signal priority, a model has been developed to simulate traffic and transit operations in the No. 3 Road corridor from Bridgeport Road to Granville Avenue. The following sections describe the properties of the corridor as modeled. This model is used to compare the relative effectiveness of different proposed TSP strategies in improving transit operations. This is done by manipulating the important TSP parameters and observing the resulting effects on the measures of effectiveness. 3.1 TSP Parameters 3.1.1 Green Extension Green extension is a TSP tactic intended to allow oncoming transit vehicles to clear an intersection before the end of a green phase. When a bus is detected during a green phase, an extension is applied to the length of the green phase to allow the bus to clear the intersection. The effectiveness of the green extension in clearing buses depends on the length of the green extension, the distance of the detector from the intersection and the speed of the bus. The existing TSP strategy does not provide green extensions. 3.1.2 Phase Skipping Phase skipping is a TSP tactic aimed at reducing the delay to transit vehicles approaching an intersection. This strategy involves immediately cycling to green for the bus phase once the minimum green of the current phase has expired. The phases in the cycle that precede the bus phase are skipped and not served until the next cycle. While this technique reduces the delay that buses experience in waiting for a green phase, there can be significant adverse impacts on cross street traffic operations. The current TSP strategy employs phase skipping with red truncation. 3.1.3 Check-In Detector Location The location of the upstream check-in detectors significantly influences the efficiency of TSP provision. It is the detection of oncoming transit vehicles that initiates the TSP process. If detectors are located too far from an intersection, then a green phase may expire before a bus can clear the intersection, even if the green phase has been extended. If the detectors are located too close to an intersection, transit vehicles may experience significant delay as the signal controller cycles to green for the bus phase. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 43 Transit and Light Rail Transit in the City of Richmond 3.1.4 Transit Stop Location In general, transit stops can be located on the near or far side of intersections or at mid-block locations. This research considers how different transit stop locations affect the effectiveness of TSP provision. 3.1.5 Type of Transit Signal Priority Unconditional and conditional priority will be compared in this study. Unconditional TSP gives priority to every transit vehicle at every intersection. Conditional TSP, on the other hand, only gives priority to those transit vehicles that satisfy certain pre-defined criteria. These two types of TSP are evaluated for their impacts on the measures of effectiveness. 3.2 Measures of TSP Effectiveness This section describes the methods used to evaluate and compare the different TSP scenarios simulated in this study. 3.2.1 Travel Time One measure of the relative effectiveness of different TSP strategies is the change in northbound and southbound travel times resulting from each strategy. The most effective TSP strategy will minimize delays to transit at signalized intersections and, consequently, minimize the time required to travel the length of the line. Differences in the characteristics of the northbound and southbound corridors may cause one TSP strategy to be the most effective in one direction but not in the other. 3.2.2 Stop Delay Stop delay measures the time that a transit vehicle is not moving, whether it is at an intersection or a transit stop. By reducing the amount of time that a transit vehicle is stopped, the travel time and regularity of travel time and arrivals at stops can be improved. The most effective TSP strategies will function so that the bus receives a green phase before it begins to decelerate when approaching an intersection. Thus, the majority of time that the bus spends not moving at the desired operating speed will be at transit stop locations. 3.2.3 Total Delay Total delay measures the total time a bus is not moving at the desired operating speed. This includes stop delay and also time when a bus is accelerating and decelerating. An inefficient TSP strategy will require a bus to decelerate significantly or stop at an intersection before the signal controller provides the green phase. Therefore, the objective University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 44 Transit and Light Rail Transit in the City of Richmond is to provide TSP that minimizes or eliminates the delay to transit vehicles as they move through intersections. 3.2.4 Cross Street Delay When evaluating a TSP strategy, it is important to consider the effect on automobile traffic operations as well as the effect on transit operations. A strategy that provides efficient transit service may cause extreme adverse impacts on the level of service of traffic operations on streets crossing the main corridor. It is important to consider whether the benefits of the TSP strategy provided to transit service justify the impact on cross street operations. Ideally, a TSP strategy should provide significant transit operational improvements with minimal impacts on cross street traffic operations. This research examines both total delay and stop delay to cross street traffic. Stop delay per vehicle is typically used as the criteria for determining the level of service on an intersection approach. 3.2.5 Green Extension Effectiveness The success with which green extensions are provided is also an important indication of how effectively a strategy provides TSP. Green extension effectiveness is measured as the ratio of transit vehicles that clear an intersection during a green extension to the total number of green extensions requested. A low ratio indicates that the parameters of the TSP scenario do not allow transit vehicles to efficiently use green extensions. This measure of effectiveness is affected mainly by detector location, transit stop location and the length of the green extension. These factors are also interdependent. 3.3 VISSIM Model 3.3.1 Introduction The simulation model used in this study was designed using the VISSIM simulation software package. The manufacturer of this software describes VISSIM as a microscopic, behaviour-based simulation model of urban traffic and public transit operations. The advantage of this simulation software is the control it allows the user in developing models that accurately represent a real situation. This allows for coding of a variety of different traffic operational configurations, including transit signal priority schemes. The software also provides an impressive visual animation of the modeled scenario. There are a number of transit operating parameters that must be assigned values in order for the VISSIM simulation to accurately model transit movements. These include operating headways, speed and acceleration properties, dwell times at transit stops and University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 45 Transit and Light Rail Transit in the City of Richmond automobile traffic composition. Other parameters that must be modeled are the locations of check-in detectors, transit stop locations and the existing signal control logic. The accuracy with which the model represents the study corridor affects how applicable the simulation results are to the corridor. If the corridor is not accurately modeled, the conclusions and recommendations made based on the simulation results cannot be expected to be appropriate for the corridor upon implementation of TSP. Thus, it is important to calibrate the model to ensure that it accurately reflects the situation being modeled. 3.3.2 Operating Parameters The corridor was modeled with the existing bus schedule used from 3 p.m. to 4 p.m. on weekdays. Thus, 98 B-Line buses operate on 5 or 6-minute headways in the northbound direction and 6 or 7-minute headways in the southbound direction. The 98 B-Line buses operate at a desired speed of 50 km/h (13.9 m/s). The desired acceleration is 2 2 approximately 1.0 m/s and he desired deceleration rate is 0.9 m/s . Because dwell times vary over time and from stop to stop, a normal distribution with a mean of 20 seconds per stop and a standard deviation of 3 seconds per stop has been employed. Finally, No. 3 Road automobile traffic has been modeled with a traffic composition of 5% heavy vehicles and a desired speed range of 45 to 60 km/h. This corridor has also been modeled with a hypothetical at-grade light rail transit (LRT) system that operates in the middle of the No. 3 Road corridor in the right-of-way currently used by buses. However, the LRT will operate in an exclusive right-of-way for the entire corridor, unlike the current bus system, which only has exclusive lanes north of Westminster Highway. Thus, reconstruction of the road geometry will be necessary. The physical specifications of the LRT line simulated in this research are based on designs produced by private consulting firms in Vancouver to provide options for the corridor. The hypothetical LRT line has five stations in each direction, as outlined below: 1. NORTHBOUND Richmond Centre Station: immediately downstream of Park Road; Westminster Station: immediately downstream of Westminster Highway; Alderbridge Station: immediately upstream of Alderbridge Way; Cambie Station: midway between Browngate and Cambie Road; and Capstan Station: midway between Capstan Way and Sea Island Way. 2. SOUTHBOUND Capstan Station: immediately downstream of Capstan Way; University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 46 Transit and Light Rail Transit in the City of Richmond • Cambie Station: immediately downstream of Cambie Road; • Alderbridge Station: immediately downstream of Alderbridge Way; • Westminster Station: immediately downstream of Westminster Highway; and • Richmond Centre Station: immediately upstream of Park Road. This physical design and operating parameters of the LRT system are based on information provided by TransLink and private consulting firms currently undertaking work in the No. 3 Road corridor. Each light rail vehicle (LRV) is composed of three cars, each 2.65 metres by 27 metres. Therefore, each train will be 81 metres in length. This is accommodated by station platforms that are 90 metres in length. The desired operating speed is 50 km/h with a desired acceleration rate of 1.2 m/s2 and a desired deceleration rate of 0.9 m/s2. The LRVs operate with peak headways of 6 minutes, an average dwell time of 14 seconds and a standard deviation of dwell time of 2.1 seconds. The dwell time values are 70% of the values used for express bus service. LRT dwell time has been assumed to be some proportion less than express bus for a few reasons. Firstly, LRT passengers purchase their tickets prior to boarding and passengers board the train without having to feed the ticket into a fare collection box. This reduces per passenger boarding times. Secondly, there are two or more doors on each car, which reduces the time required for a group of passengers to enter the vehicle. Finally, the platforms are designed to be level with the L R V floor, which allows for faster and easier boarding, especially for the elderly and handicapped. These advantages should reduce average travel times and travel time variability of LRT compared to express bus service. In addition, the provision of an exclusive right-of-way for the entire length of the corridor should improve travel times by allowing LRVs to travel at the desired operating speed, rather than traveling at the speed dictated by automobile traffic flow. 3.3.3 Network Geometry The express bus TSP scenarios maintain the existing network geometry with only the transit stops and check-in detector locations being manipulated. The existing geometry requires buses to share travel lanes with automobile traffic between Granville Avenue and Westminster Highway. From Ackroyd Road to Sea Island Way, the buses travel in an exclusive right-of-way in the middle of the corridor. Transit stops are currently located at various near side, mid-block and far side locations. They are moved to far side locations in some scenarios to study the influence of transit stop location on TSP effectiveness. The LRT system involves modifications to the network geometry. These modifications are based on design plans completed by a private consulting firm in developing options for the corridor. The LRT line will have an exclusive right-of-way for the entire corridor. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 47 This requires re-alignment of the roadway south of Westminster Highway. The transit stations will be located at locations where the 90-metre platforms can be accommodated. Thus, transit station location is not manipulated in the LRT scenarios. 3.3.4 Existing Upstream Check-In Detector Locations Bus check-in detectors are located upstream of all intersections in the corridor in both the northbound and southbound directions. These detectors identify transit vehicles as they move past particular points in the corridor. These detectors can be used for a number of purposes, but they are particularly useful in transit signal priority as the detection of a transit vehicle initiates the TSP strategy. Table 3-1 outlines the approximate current location of the check-in detectors upstream of the intersections, based on information provided by TransLink, the transit authority for the Greater Vancouver Regional District. Table 3-1 Intersection Northbound Southbound Park 85 54 Cook 73 151 Saba 118 65 Westminster 63 112 Ackroyd 137 176 Lansdowne 176 15 Lansdowne Mall 108 146 Alderbridge 11 190 Leslie 167 167 Browngate 167 109 Cambie 122 190 Yohan Mall 167 179 Capstan 200 170 Sea Island 116 220 of intersection) 3.3.5 Transit Stop Locations There are currently eight 98 B-Line stations in the northbound corridor and six stations in the southbound corridor. There is a mix of near side, mid-block and far side stations. In the northbound corridor, the Anderson, Cook and Westminster stations are located in a right-of-way that is shared with automobile traffic. From the Lansdowne Road to Sea Island Way, the bus line operates in an exclusive corridor. In the southbound direction, the Westminster, Cook and Anderson stations share the travel right-of-way with automobile traffic. On the other hand, the Capstan, Cambie and Alderbridge stations are situated in the exclusive transit corridor. The relative location of these transit stops at their respective intersections is summarized in Table 3-2. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 48 Transit and Light Rail Transit in the City of Richmond Table 3-2 Transit Sto p Locations Northbound Southbound Intersection Location Intersection Location Anderson Far Side Capstan Far Side Cook Far Side Cambie Mid-Block Westminster Near Side Alderbridge Far Side Lansdowne Far Side Westminster Mid-Block Alderbridge Near Side Cook Far Side Cambie Near Side Anderson EB on Anderson Capstan Far Side Sea Island Far Side 3.3.6 Existing Signal Control Logic Based on information provided by the City of Richmond, it was determined that the signalized intersections in the study corridor are semi-actuated with minimum and maximum green times. The north-south and east-west phases are guaranteed the minimum green time in every cycle, but the length of the phase is extended up to the maximum if traffic volumes are high and the gap between vehicles is less than the maximum allowable gap. Thus, the intersections tend to operate as pre-timed signals with the maximum green times during peak periods. The left turn phases are also actuated so that they may be skipped if there are no vehicles waiting for the phase. Table 3-3 summarizes the existing signal timing parameters for the intersections in the study corridor. These are the parameters that are manipulated by the TSP strategy when a TSP call is received. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 49 Transit and Light Rail Transit in the City of Richmond Table 3-3 Existing Signal Timing Parameters in No. 3 Road Corridor Intersection Phase Min. Green FDW Max. Green M a x . G a p NSLT 5 - 12 2.5 Granville NS 5 1 35 4 EWLT 7 - 15 2.5 EW 5 1 30 4 Park NS 15 11 55 5 EW 5 1 35 4 NSLT 5 - 15 3 Cook NS 15 11 55 5 EW 5 1 40 4 NSLT 5 - 15 5 Saba NS 15 11 55 4 EW 5 1 35 4 NSLT 5 1 20 3 Westminster NS 8 4 45 4 EWLT 5 1 20 3 EW 8 4 45 4 NSLT 3 - 16 3 Ackroyd NS 15 11 45 4 EW 8 4 25 4 NSLT 3 - 15 3 Lansdowne NS 15 11 45 4 EWLT 5 - 15 3 EW 8 3.5 30 4 NSLT 3 - 15 3 Lansdowne Mall NS 15 11 50 4 ' EW 6 2 25 4 NSLT 3 - 15 3 Alderbridge NS 15 11 45 4 EWLT 5 - 18 3 EW 8 3 35 4 NSLT 3 - 20 3 Leslie NS 15 11 50 4 EWLT 5 - 15 3 EW 8 3 30 4 NSLT 3 - 15 3 Browngate NS 15 11 50 4 EW 6 2 35 4 NSLT 3 - 20 3 Cambie NS 15 11 45 4 EWLT 5 - 15 3 EW 8 3.5 30 4 NSLT 3 - 17 3 Yohan NS 15 11 50 4 EW 6 2 25 4 NSLT 3 - 15 3 Capstan NS 15 11 50 4 EW 8 3.5 35 4 Sea Island NS 7 2 32 3 EW 10 5. 35 3 Bridgeport NS 7 2 17 3 EW 10 5 25 3 There is currently a signal priority strategy in place. The TSP logic operates as signal pre-emption, meaning that if a transit vehicle is detected during a red phase for the main corridor, the signal controller will cycle to green for the main corridor immediately after University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 50 the minimum green of the current phase has expired. The existing strategy uses phase skipping, meaning that the signal controller skips over phases that precede the transit phase in the scheduled cycle in order to provide green to the oncoming transit vehicle. However, green extension is not used to allow transit vehicles arriving during the green phase to clear before the green time expires. 3.3.7 Model Calibration While the corridor segment simulated in this research extends along No. 3 Road from Bridgeport Road to Granville Avenue, the existing 98 B-Line exits and enters the southern boundary of the corridor just north of Granville Avenue at Anderson Road. The study segment from Sea Island Way to Granville Avenue is approximately 3,000 metres in length. The six-lane cross-section is composed of two northbound shared travel lanes and two southbound shared travel lanes separated by an exclusive two-lane bus right-of-way in the middle. The exclusive bus corridor extends between Ackroyd Road and Sea Island Way in both directions. It is important to determine how accurately the simulation model represents the real situation existing in the corridor. This is known as model calibration and it involves comparing real travel times to those resulting from the simulation model. Travel time data with the existing corridor characteristics and transit signal priority strategy were obtained from TransLink. The simulations were run under the traffic conditions during the heaviest weekday traffic volumes, which occur between 3 p.m. and 4 p.m. According to travel time studies undertaken by TransLink on November 4, 28 and 29, 2002, the northbound bus travel time from Anderson Road to the Airport Station is approximately 18 minutes during the 3 p.m. to 4 p.m. traffic volume peak. Based on discussions with TransLink staff, it was determined that the approximate northbound travel time between Sea Island Way and Airport Station is 7 minutes. Consequently, the travel time between Anderson Road and Sea Island Way is around 11 minutes. In the southbound direction, the TransLink travel time studies indicated that the travel time between Airport Station and Anderson Road is approximately 11 minutes. The travel time between Airport Station and Sea Island Way was determined to be approximately 3 minutes during the 3 p.m. to 4 p.m. traffic volume peak. Therefore, the existing Sea Island Way to Anderson Road southbound travel time is roughly 8 minutes. With the existing transit stops, detector locations, traffic volumes and transit signal priority strategy, the VISSIM model produced average travel times of 662 seconds (11.03 minutes) in the northbound corridor and 511 seconds (8.52 minutes) in the southbound corridor. Consequently, the travel times simulated with the VISSIM model correspond quite closely to the existing real travel times observed in the corridor. It is important to University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 51 Transit and Light Rail Transit in the City of Richmond note that these are average travel times. Each travel time is different and the difference between real and simulated travel times varies with every transit vehicle. Trave l Di rect ion Real Trave l T ime Simulated Trave l T ime Northbound Southbound 11 minutes (660 seconds) 8 minutes (480 seconds) 11.03 minutes (662 seconds) 8.52 minutes (511 seconds) 3.3.8 Simulation Scenarios A number of scenarios employing different combinations of the study parameters were simulated in order to study the influence of each parameter on the pre-determined measures of effectiveness. The parameters manipulated in each scenario are transit stop location, detector location, green extension length and phase skipping. Transit stops are either located at their existing locations (near side, mid-block or far side) or all at far side locations. For the LRT scenarios, transit station location is held constant since the size of the platforms limits possible locations. The check-in detectors are situated at their existing locations, or 50 metres, 100 metres or 150 metres upstream of each intersection. The green extension options applied by the TSP strategy are 10, 15, 20 and 30 seconds. Each TSP scenario is simulated with and without phase skipping. Each of the express bus scenarios is also simulated with hard (unconditional) and soft (conditional) priority to allow for comparison between the results of these two types of TSP. In the express bus scenarios, each of the possible detector configurations (existing, 50 metres, 100 metres and 150 metres) were simulated with each of the possible green extensions, with existing transit stops and far side transit stops. This amounts to 32 scenarios. The same scenarios were simulated with and without phase skipping, which increases the number of simulation scenarios to 64. Furthermore, each of these scenarios was simulated with hard and conditional TSP, which increases express bus scenarios to 128. Finally, each scenario was simulated five times in order to determine average values for the measures of effectiveness. Thus, 640 simulation runs were completed to obtain the data for the express bus TSP analysis. Fewer simulations were necessary for the LRT scenarios. Since transit station location is held constant, each scenario did not need to be simulated twice to determine the effect of transit station location. In addition, the existing detector locations are not appropriate for the segment of the corridor that is re-aligned for the LRT line since the existing detectors will be moved. Therefore, 12 scenarios (50, 100 and 150-metre detectors, each with four green extensions) were simulated with the planned station locations. These scenarios were simulated with and without phase skipping, which increases the number of scenarios to 24. With five simulation runs for each scenario, 120 simulations were University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 52 Transit and Light Rail Transit in the City of Richmond completed to obtain data for the LRT analysis. Consequently, a total of 760 simulation runs completed for this research project. 3.3.9 Challenges and Assumptions 1. The dwell time at a transit stop is variable since it depends on the number of boarding and alighting passengers and the time needed for each passenger to enter and exit the transit vehicle. This makes it impossible to accurately model the dwell time of each transit vehicle at every stop. Dwell times have been modeled using a normal distribution with mean and standard deviation values based on discussions with TransLink staff and information from the literature reviewed. LRT dwell times have been assumed to be less than bus dwell times, with less variability, for the following reasons: • Passengers board without having to feed tickets into a fare collection box; • Multiple doors allow for faster boarding; and • Platforms are level with L R V floors, which allows for faster boarding, especially for the elderly and handicapped. 2. Speed and acceleration of transit vehicles varies depending on the operator and the environment. The results of this research will have to be based on certain acceleration/deceleration and speed assumptions. It has been assumed that the modeled values will be close to average speed and acceleration values and operators will be encouraged to adhere as closely as possible to the parameters used in developing the TSP system. 3. There is variability in the simulation results from run to run depending on the random seed used to initiate the probability distributions used in modeling certain variables. This makes it difficult to determine the most appropriate results to use for the measures of effectiveness. To control variability in the results, a trimmed average of travel time and delay values was calculated. Five simulation runs were completed for each scenario. The high and low values were eliminated and an average was taken of the remaining three values. 4. For the first few minutes of a simulation run, the network is not fully flooded with vehicle traffic. This can produce travel time and delay results that are lower than they actually should be. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 53 Transit and Light Rail Transit in the City of Richmond To counter this, delay and travel time results are not collected for the first 15 minutes of the 1.5-hour simulation period. This allows the network to regulate itself before data collection begins. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 54 Transit and Light Rail Transit in the City of Richmond 4. RESULTS AND ANALYSIS 4.1 Unconditional Signal Priority, Express Bus Service In this module, express bus service in the No. 3 Road corridor is studied with unconditional transit signal priority. The existing TSP strategy can basically be described as signal pre-emption. When a bus is detected during a red phase, the signal controller cycles to green for the bus after the expiration of the minimum green time of the current phase. Therefore, it is a phase skipping strategy that employs red truncation. However, green extensions are not provided in the existing TSP strategy to allow buses arriving during green phases to clear intersections. The unconditional TSP strategies proposed in this research grant unconditional priority to all buses traveling in the corridor with the use of red truncation and green extension. Green extension is used in the proposed strategies to determine the effects of green extension provision and the length of the green extension provided. Other TSP parameters are also modified in order to study the influence of these parameters on bus operations. These parameters include transit stop location, check-in detector location, length of green extension and the use of phase skipping. The TSP strategies simulated for this research employ different combinations and values of these parameters. The effectiveness of the different TSP strategies is evaluated using the following measures of effectiveness: • Average transit vehicle travel times in the northbound and southbound corridors; • Total delay to transit vehicles as they travel through the corridors; • Variability in travel times from one transit vehicle to another; • Impact on cross street automobile traffic operations, measured in terms of delay per vehicle and level of service; and • Green extension effectiveness as measured by green extension efficiency ratios. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 55 Transit and Light Rail Transit in the City of Richmond 4.1.1 Transit Vehicle Travel Time and Delay 4.1.1.1 Transit Stop and Check-In Detector Locations In the No. 3 Road corridor, there is currently a mix of near side, mid-block and far side transit stop locations. In order to test the influence of transit stop location on travel time and delay, each TSP scenario was simulated with the existing stop configuration, as well as with a configuration employing far side stops at all locations. Upstream vehicle check-in detectors are currently situated at varying distances from their respective intersections. The influence of check-in detector location has been tested by simulating each TSP scenario with the existing detector locations, as well as with detectors located 50, 100 and 150 metres upstream of the corridor intersections. The delay experienced by transit vehicles is one of the most important factors affecting corridor travel time. The simulation results indicate a strong positive correlation between travel time and total delay. That is, as total delay increases or decreases, travel time tends to do the same. For this reason, travel time and total delay will be discussed concurrently and it will be illustrated how total delay and travel time reductions correspond with one another. In the southbound direction, the 100-metre detectors with the far side transit stop configuration tend to provide the lowest travel times of the tested detector configurations, with and without phase skipping. Although with the existing transit stops the 50-metre detectors provide the lowest travel times, the best travel time performance in the southbound travel corridor is achieved with check-in detectors located 100 metres upstream of intersections and with transit stops located at far side locations. The use of far side transit stops helps to improve travel time by reducing bus delay. At a near side stop location, buses can be forced to double stop: first at the near side stop and then at the red signal. The near side stop can cause a green extension to expire while a bus dwells at the stop, thus wasting the green time provided to the bus by TSP. In the northbound direction, the lowest travel times are achieved under the far side stop configuration with the 50- and 100-metre detectors. The use of far side transit stops results in noticeable travel time reductions compared to the same scenarios with the existing stops under all green extensions. The 100-metre detectors appear to be the most effective in terms of travel time. The 50-metre detectors may be too close to the intersections to provide effective TSP without causing significant delay to oncoming transit vehicles. The 150-metre detectors may be located too far upstream to allow buses to reliably clear intersections during green extensions. The results in Tables 4-1 through 4-4 illustrate that total transit delay and travel time are typically lowest with the 50 and 100-metre detectors and that the use of far side transit University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 56 stops results in significantly lower total delay and travel time values than the existing transit stop configuration. There is the obvious trend that the scenario with the least bus delay also has the lowest travel times. It should be noted that the northbound delay and travel time values are consistently higher than the southbound direction. There are three main reasons for this result. Firstly, there are currently eight transit stops in the northbound direction and only six transit stops in the southbound direction. Consequently, a transit vehicle will experience more delay traveling northbound than southbound in allowing passengers to board and alight. Secondly, all of the transit stops in the southbound direction are either mid-block or far side stops. However, in the northbound direction, three of the eight transit stops are near side stops, which can result in significant delay due to double stopping and, consequently, greater travel time than in the southbound direction. Finally, traffic volume is higher in the northbound corridor than the southbound. This can result in transit vehicle travel speeds in the shared travel segments of the corridor being lower in the northbound direction than the southbound. Table 4-1 Total Transit Vehicle Delay Compar ison—Detec to r Location Analysis Detector Location Southbound Northbound 50 Metres 313.3 sec 435.8 sec 100 Metres 296.4 sec 445.7 sec 150 Metres 303.2 sec 485.5 sec Table 4-2 Total Transit Vehicle Delay Compar ison—Trans i t Stop Location Analysis Detector Location Southbound Northbound Existing Stops Far Side Stops Change in Delay Due to Far Side Stops Existing Stops Far Side Stops Change in Delay Due to Far Side Stops 50 Metres 313.3 306.6 -2.1% 435.8 443.8 + 1.8% 100 Metres 296.4 272.4 -8.1% 445.7 446.2 + 0.1% 150 Metres 303.2 300.5 - 0.9% 485.5 460.8 -5.1% Table 4-3 Transit Vehicle Travel Time Compar ison—Detec to r Location Analysis Detector Location Southbound Northbound 50 Metres 531.3 sec 670.3 sec 100 Metres 515.7 sec 680.7 sec 150 Metres 528.6 sec 723.2 sec University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 57 Table 4-4 Transit Vehicle Travel Time Compar ison—Trans i t Stop Location Analysis 15-Second Green Extension, No Phase Skipping Detector Location Southbound Northbound Existing Stops Far Side Stops Change in Travel Time Due to Far Side Stops Existing Stops Far Side Stops Change in Travel Time Due to Far Side Stops 50 Metres 531.3 521.1 - 1.9% 670.3 679.3 + 1.3% 100 Metres 515.7 480.9 - 6.7% 680.7 672.7 - 1.8% 150 Metres 528.6 519.6 - 1.7% 723.2 694.2 - 4.0% 4.1.1.2 Length of Green Extension Prior to running the simulations, travel time and delay values were expected to be lower with longer green extensions since they should increase the probability of clearing an intersection during an extension. However, under all of the studied TSP scenarios, both with and without phase skipping and in both travel directions, there is no definite correlation between the length of the green extension and the corridor travel times. For the most part, there is minor up-and-down fluctuation in travel times as the green ( extension increases from 10 to 30 seconds. Figure 4A shows the travel time trend lines for the different scenarios as the green extension increases in length. The average travel times under different green extensions do not appear to be significantly different. There are a couple of possible explanations for this unexpected result. As will be seen later, it is obvious that a longer green extension improves the efficiency of green extension provision. Thus, it is odd that a longer green extension does not translate into lower travel times since the higher green extension efficiency should reduce delay to transit vehicles arriving during green phases. The lack of correlation between green extension length and travel time may be due variability in the simulation model. Green extension is only given to transit vehicles that arrive during a green phase. The phase in which a transit vehicle arrives depends on the random probability distributions that control different factors in the simulation. Hence, it is quite possible that, during the 1.5-hour simulation period, a significant number of TSP calls may be made during red phases, thus not requiring a green extension. Therefore, transit travel time and delay would depend on other parameters, such as phase skipping, and green extension length would be inconsequential. As a result, the unexpected lack of correlation between transit travel time/delay and the length of green extension may be due to only a small number of TSP calls that required green extension being included in the sample ' used to calculate average travel times. This can dilute the relationship between travel time and green extension length. It is possible that a strong relationship may exist between these two parameters, but it is simply not visible in the simulation results. This is the most reasonable explanation. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 58 Transit and Light Rail Transit in the City of Richmond Another explanation is that once green extensions reach a certain length, the majority of transit vehicles are able to clear intersections and increasing the green extension length does little to improve travel time or reduce delay. Consequently, there would be no apparent relationship between green extension length and transit travel time. Ideally, there is a maximum necessary green extension length in which all buses would be able to clear an intersection. This green extension length is based on the detector-to-intersection distance and the travel speed of the bus. In this research, buses have been assumed to have a desired travel speed of 50 km/h (13.9 m/s). Thus, in a protected right-of-way with buses maintaining the desired travel speed, the maximum necessary green extension would be approximately 4 seconds, 8 seconds, 12 seconds and 16 seconds for detector-to-intersection distances of 50, 100, 150 and 200 metres respectively. Consequently, the 20 and 30-second green extensions studied in this thesis would theoretically be unnecessary. However, longer green extensions will be required when traffic congestion in the shared travel segments of the corridor causes buses to travel at speeds lower than desired. A longer extension will also be necessary where there is a near side transit stop requiring a bus to stop for boarding and alighting passengers before clearing the intersection. In general, any situation where a bus is not traveling at the desired travel speed for whatever reason will necessitate a longer green extension for intersection clearance. This, is the case at most of the intersections in the study corridor. There is a definite trend that increasing the length of the green extension increases the ratio of buses clearing an intersection during a green extension, even where 50-metre detectors are employed. This explanation applies to all the study scenarios with both express bus and light rail transit service. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 59 Travel Time Comparison Effect of Green Extension Length 700 | 450 I 10 15 20 30 Green Extension (seconds) «^»100-Metre Detectors, Southbound. No Skipping — • — 100-Metre Detectors. Southbound, Skipping " * " 5 0 - M e t r e Detectors, Northbound, No Skipping —A—50-Metre Detectors, Northbound, Skipping Figure 4A: Average Travel Time—Green Extension Length Analysis 4.1.1.3 Phase Skipping Under typical unconditional priority scenarios, once a TSP request is initiated, the signal controller will cycle through the minimum green times of the signal phases preceding the bus phase in the cycle. However, in scenarios that implement phase skipping, when a TSP call is received, the signal controller immediately cycles to the bus phase once the minimum green time of the current phase has expired. The use of phase skipping is expected to improve travel time performance by reducing delay in transitioning from red to green in the bus phase. This will reduce the total delay experienced by transit vehicles while traveling in the corridor. Each of the studied scenarios has been simulated with and without phase skipping in order to determine the effect of phase skipping on bus delay and travel times. From review of the simulation results, it is clear that the use of phase skipping provides significant travel time benefits relative to the same TSP strategies without phase skipping. However, it should be noted that when a TSP call is made in the phase immediately preceding the bus phase, phase skipping provides no benefit over no phase skipping since the bus will be served at the same time in both situations. The advantage of phase skipping comes when there are many different signal phases and a TSP call is received a significant period of time before the bus phase is scheduled to be served. It University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 60 should be noted, and will be discussed in more depth later, that phase skipping should result in more consistent travel times since the variability in the time to cycle to green for the bus phase after a TSP request will likely be reduced. Tables 4-5 and 4-6 outline the differences in total delay and travel times between TSP scenarios with and without phase skipping. The travel time benefit of phase skipping relative to no phase skipping is clear under all of the studied detector location configurations. It is also clear that the travel time improvement is largely due to reductions in total transit delay. Table 4-5 Total Transit Vehicle Delay Compar ison—Phase Skipping Analysis Detector Location Southbound Northbound No Skipping Skipping Change in Delay Due to Phase Skipping No Skipping Skipping Change in Delay Due to Phase Skipping 50 Metres 306.6 258.2 - 15.8% 443.8 392.2 - 13.0% 100 Metres 272.4 237.7 - 12.7% 446.2 401.4 - 10.0% 150 Metres 300.5 285.5 - 5.0% 460.8 447.6 - 2.9% Table 4-6 Transit Vehicle Travel Time Compar ison—Phase Skipping Analysis Detector Location Southbound Northbound No Skipping Skipping Change in Travel Time Due to Phase Skipping No Skipping Skipping Change in Travel Time Due to Phase Skipping 50 Metres 521.1 475.7 - 8.7% 679.3 625.9 - 7.9% 100 Metres 480.9 455.8 - 5.2% 672.7 635.5 - 5.5% 150 Metres 519.6 506.3 - 2.6% 694.2 673.5 - 3.0% 4.1.1.4 Summary The use of far side transit stops provides significant travel time reductions relative to a mix of near side, mid-block and far side locations. This is largely due to reduction in delay that can be caused by double-stopping at near side transit stop locations. The 100-metre detectors provide the best overall travel time performance in the corridor with 50-metre detectors providing comparable results. The length of the green extension applied in the TSP strategy was not found to be correlated to bus travel times as no discernible positive or negative relationship is noticeable in the simulation results. Finally, phase skipping provides significant travel time benefits relative to the same TSP scenarios without skipping by reducing delay to transit vehicles that occurs while the signal controller cycles from red to green for the bus phase. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 61 Transit and Light Rail Transit in the City of Richmond 4.1.2 Transit Vehicle Travel Time Variability One of the aims of TSP is to improve transit reliability by reducing the variability in average travel time. This would result in better schedule adherence and, thus, more reliable and attractive transit service. 4.1.2.1 Transit Stop and Check-In Detector Locations In the no phase skipping TSP scenarios, none of the studied detector configurations definitively provides more reliable travel times than the others. The detector location with the lowest travel time variability varies with different green extensions. However, it is possible to look specifically at a green extension of 15 seconds, since this is close to the value used in the TSP strategies discussed in the literature review. In the southbound corridor, the 100-metre detectors provide the lowest travel time variability both with and without phase skipping. In the northbound corridor, the 50-metre detectors provide the lowest travel time variability with and without phase skipping. It is clear that the 150-metre detectors result in the most variable travel times, especially when used with the existing transit stop configuration. These results are summarized in Table 4-7. Table 4-7 Transit Vehicle Travel Time Variabil i ty Compar ison—Detec to r Locat ion Analysis 15-Second Green Extension, No Phase Skipping Detector Location Southbound Northbound Standard Std. Dev. As Standard Std. Dev. As Deviation % of Mean Deviation % of Mean 50 Metres 11.38 2.14% 8.11 1.21% 100 Metres 5.27 1.02% 10.81 1.59% 150 Metres 15.40 2.95% 15.53 2.15% While one particular check-in detector location does not consistently result in the lowest travel time variability, it is clear that transit stop location is a critical factor. As illustrated in Table 4-8, in all but one scenario, the use of far side stops results in lower travel time variability than the same scenario with the existing transit stop configuration. This can be explained by reduction in the probability of a bus experiencing double stopping due to near side transit stops, which improves the consistency of time required to clear an intersection. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 62 Transit and Light Rail Transit in the City of Richmond Table 4-8 Transit Vehicle Travel Time Variabil i ty Compar ison—Trans i t Stop Locat ion Analysis 15-Second Green Extension, No Phase Skipping Detector Location Southbound Northbound Existing Stops Far Side Stops Existing Stops Far Side Stops Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M 50 Metres 11.38 2.14% 11.00 2.11% 8.11 1.21% 7.22 1.07% 100 Metres 5.27 1.02% 6.14 1.28% 10.81 1.59% 5.95 0.88% 150 Metres 15.40 2.95% 8.18 1.56% 15.53 2.15% 8.99 1.29% 4.1.2.2 Length of Green Extension In neither the southbound nor the northbound corridor does length of green extension appear to have a significant impact on travel time variability. The standard deviation as percentage of mean (SD%M) values fluctuate up and down with increasing green extension duration without any discernible pattern. The erratic pattern is most likely due to variability in the simulation results rather than the influence of green extension length. Table 4-9 compares the S D % M values for different detector locations across the range of green extensions. Table 4-9 Transit Vehicle Travel Time Variabil i ty Compar ison (Standard Deviation as % of Mean) Far Side Transit Stops Detector Location Southbound Northbound GE = 10 GE = 15 GE = 20 GE = 30 GE = 10 GE = 15 GE = 20 GE = 30 No Phase Skipping 50 Metres 2.71% 2.11% 1.32% 2.96% 3.00% 1.07% 1.18% 0.72% 100 Metres 1.89% 1.28% 1.47% 2.06% 2.75% 0.88% 1.19% 1.85% 150 Metres 1.73% 1.56% 2.93% 1.34% 1.21% 1.29% 0.57% 1.55% Phase Skipping 50 Metres 2.14% 0.97% 1.70% 2.27% 0.77% 0.35% 1.07% 1.07% 100 Metres 1.34% 0.45% 2.15% 1.87% 1.30% 0.90% 0.51% 0.29% 150 Metres 1.61% 1.75% • 1.60% 0.64% 0.30% 1.73% 1.77% 0.89% 4.1.2.3 Phase Skipping As was expected prior to running the simulations, phase skipping appears to have a positive influence on travel time variability relative to no phase skipping. The use of phase skipping appears to narrow the range of S D % M values among the different TSP scenarios. In other words, the difference between the different scenarios in terms of travel time variability is lessened with the implementation of phase skipping. Phase skipping reduces variability in travel time relative to the same scenario without phase skipping in many of the studied scenarios. These results are most likely due to phase skipping University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 63 Transit and Light Rail Transit in the City of Richmond reducing variability in the time for a signal controller to cycle to green for the bus phase since the bus phase is served immediately after the minimum green of the current phase. Table 4-10 illustrates the effect of phase skipping with a 15-second green extension. Table 4-10 Transit Vehicle Travel Time Variability Comparison—Phase Skipping Analysis 15-Second Green Extension, Far Side Transit Stops Detector Location Southbound Northbound No Skipping Phase Skipping No Skip ping Phase Skipping Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M 50 Metres 11.00 2.11% 4.61 0.97% 7.22 1.07% 2.19 0.35% 100 Metres 6.14 1.28% 2.06 0.45% 5.95 0.88% 5.75 0.90% 150 Metres 8.18 1.56% 8.85 1.75% 8.99 1.29% 11.63 1.73% 4.1.2.4 Summary None of the studied detector configurations provides the most reliable travel times in all TSP scenarios. In the southbound corridor, the 50-metre detectors provide the lowest S D % M values of the studied detector locations when there is no phase skipping. When phase skipping is implemented, the 100-metre detectors provide the lowest travel time variability. In the northbound corridor, the existing detector locations provide the most consistent travel times without phase skipping while the 50-metre detectors are more effective when phase skipping is in place. However, it is clear that the 150-metre detectors produce the most variable travel times. The location of transit stops significantly affects travel time variability. The use of far side transit stops consistently results in less travel time variability than the same scenarios with the existing stop configuration. Similarly, the use of phase skipping is beneficial in terms of reducing travel time variability relative to the same scenarios without phase skipping. However, the length of the green extension is not correlated in the simulated scenarios to travel time variability. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 64 Transit and Light Rail Transit in the City of Richmond 4.1.3 Cross Street Traffic Operations 4.1.3.1 Total Delay An important consideration in analyzing different TSP scenarios is how cross street automobile traffic operations are affected by the implementation of TSP in the main corridor. Prior to running the simulations, it was expected that phase skipping and green extension length would have considerable effects on delay experienced by traffic on streets crossing the main corridor. The implementation of phase skipping was expected to be critical since, when skipping is in place, a TSP request results in all signal phases preceding the bus phase being skipped in order to serve the bus phase. The length of the green extension was expected to be critical because the longer the extension, the longer cross street traffic has to wait for the termination of an extended bus phase. The simulation results indicate that cross street delay would not be significantly increased relative to the existing situation with the implementation of the proposed TSP scenarios. For the most part, the differences between existing vehicle delay and delay under each of the proposed TSP scenarios are less than 10 seconds, which would result in only minor changes in operational efficiency. The location of transit stops and check-in detectors does not appear to significantly affect the delay experienced by automobile traffic on cross street approaches. Table 4-11 demonstrates that the range in delay values with different detector and transit stop configurations is small. Table 4-11 Cross Street Delay Comparison 15-second Green Extension, No Phase Skipping Detector Location Existing Transit Stops Far Side Transit Stops 50 Metres 45.0 sec 47.3 sec 100 Metres 50.4 sec 54.2 sec 150 Metres 58.9 sec 46.2 sec However, as expected, the implementation of phase skipping results in noticeably greater impacts on cross street traffic operations than no phase skipping. The simulation results indicate that phase skipping causes more cross street approaches to experience increases in delay relative to the existing situation than without phase skipping and the scale of increases is greater than without phase skipping. The following are some of the most significant reductions in cross street delay, as measured at the westbound approach to Westminster Highway in the no phase skipping scenarios: 30.4% (existing detectors, existing transit stops); 30.4% (50-metre detectors, far side transit stops); 30.7% (50-metre detectors, existing transit stops); and 31.2% (50-metre detectors, existing transit stops). The greatest reductions in cross street delay are noticeably smaller with phase skipping: University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 65 20.6% (50-metre detectors, far side transit stops); 21.8% (100-metre detectors, far side transit stops); and 27.7% (100-metre detectors, far side transit stops). Table 4-12 illustrates how cross street delay values change with implementation of phase skipping. Table 4-12 Cross Street Delay Compar ison—Phase Skipping 15-Second Green Extension, Far Side Transit Stops Detector Location No Phase Skipping Phase Skipping Change in Delay 50 Metres 47.3 sec 61.9 sec + 30.9% 100 Metres 54.2 sec 61.2 sec + 12.9% 150 Metres 46.2 sec 54.5 sec + 18.0% The most significant increases in cross street delay were recorded on the westbound approach to the intersection with Leslie Street. The simulation results for this location are summarized in Tables 4-13 and 4-14. It is clear that the location of the transit stops is not a critical factor in cross street delay while the use of phase skipping is critical. Without phase skipping, there were the following notable increases: 10.1% (100-metre detectors, existing transit stops); 11.3% (100-metre detectors, far side stops); 15.5% (50-metre detectors, existing stops); and 17.4% (50-metre detectors, far side stops). With the implementation of phase skipping, the greatest delay increases become significantly larger than without phase skipping: 28.3% (150-metre detectors, far side stops); 28.4% (existing detectors, existing transit stops); 32.3% (150-metre detectors, far side transit stops); 33.8% (50-metre detectors, far side transit stops); 38.4% (existing detectors, existing transit stops); and 41.1% (50-metre detectors, far side transit stops). Table 4-13 Cross Street Delay Compar ison 15-second Green Extension, No Phase Skipping Detector Location Existing Transit Stops Far Side Transit Stops 50 Metres 33.9 sec 36.6 sec 100 Metres 35.1 sec 35.1 sec 150 Metres 34.7 sec 34.0 sec Table 4-14 Cross Street Delay Compar ison—Phase Skipping 15-Second Green Extension, Far Side Stops Detector Location No Phase Skipping Phase Skipping Change in Delay 50 Metres 36.6 sec 48.2 sec + 31.7% 100 Metres 35.1 sec 37.1 sec + 5.7% 150 Metres 34.0 sec 42.0 sec + 23.5% Table 4-15 compares changes in delay at three sample cross street approaches with different detector locations. The results summarized in this table indicate that the changes in cross street delay relative to existing are small and the implementation of phase University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 66 Transit and Light Rail Transit in the City of Richmond skipping increases cross street delay relative to the same scenarios without phase skipping. Table 4-15 Change in Cross Street Delay Location No Phase Skipping Phase Skipping Lansdowne Road Westbound Approach (50-Metre Detectors) - 4.3 sec - 2.7 sec Lansdowne Road Eastbound Approach (100-Metre Detectors) - 4.9 sec + 2.2 sec Park Road Westbound Approach (150-Metre Detectors) + 0.4 sec + 1.5 sec Contrary to preliminary expectations, the length of the green extension applied in the TSP strategy does not appear to significantly affect cross street traffic operations. While there are some intersections that show a general pattern of increasing cross street delay with increasing green extension, for the most part, delay fluctuates up and down as the green extension increases from 10 to 30 seconds. This is shown for the Westminster Highway and Leslie Street westbound approaches in Tables 4-16 and 4-17. Table 4-16 Cross Street Delay Compar ison—Green Extension Length (Existing Transit Stops) Detector Location G E = 10 G E = 15 GE = 20 G E = 30 Delay Change from Existing Delay Change from Existing Delay Change from Existing Delay Change from Existing 50 m 62.9 sec -3.8% 65.4 sec 0% 68.6 sec + 4.8% 57.9 sec - 11.5% 100 m 59.7 sec - 8.8% 68.9 sec - 5.3% 75.6 sec + 15.9% 65.2 sec - 0.4% 150 m 57.5 sec - 12.0% 56.5 sec - 13.6% 60.8 sec -7.1% 52.4 sec - 19.9% Table 4-17 Cross Street Delay Compar ison—Green Extension Length (Existing Transit Stops) Detector Location GE = 10 GE = 15 GE = 20 GE = 30 Delay Change from Existing Delay Change from Existing Delay Change from Existing Delay Change from Existing 50 m 38.4 sec + 12.5% 42.8 sec + 25.3% 39.7 sec + 16.4% 35.4 sec + 3.7% 100 m 41.2 sec + 20.6% 35.7 sec + 4.6% 37.9 sec + 11.1% 39.5 sec + 15.7% 150 m 37.5 sec + 10.0% 42.7 sec + 25.0% 38.3 sec + 12.2% 39.5 sec + 15.7% 4.1.3.2 Level of Service The above analysis of cross street traffic operations considers the total delay experienced by automobiles. However, the stop delay experienced by an average vehicle is also an important measure of transit operational effectiveness. The average stop delay per vehicle University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 6 7 Transit and Light Rail Transit in the City of Richmond is used to determine the level of service (LOS) on a cross street approach. LOS evaluates the efficiency of traffic operations on an approach. Each of the potential TSP strategies is evaluated for its impact on level of service of cross street approaches in the corridor. Below are the stop delay values that correspond to each level of service designation: Level o f Service Stop Delay per Vehicle A <= 5 seconds B 5.1 - 15.0 seconds C 15.1-25.0 seconds D 25.1 -40.0 seconds E 40.1-60.0 seconds F > 60.0 seconds Under the studied TSP scenarios, the change in stop delay relative to the existing situation at the majority of cross street approaches is insufficient to change the level of service. In the majority of cases where level of service does change, the LOS improves. Where there is reduction in LOS, it is usually the result of existing stop delay being close to the upper limit of the current LOS category. Consequently, even a small increase in stop delay worsens the LOS. However, this does not mean that automobile traffic will experience a noticeable change in traffic operations. Table 4-18 shows examples of the following: • Small increases in stop delay can result in level of service increases where the existing stop delay is near the upper threshold of the current LOS category; • Some cross streets have existing stop delay levels so high that even decreases of over 20 seconds do not improve the level of service from LOS F; and • Implementing certain TSP scenarios can produce significant reductions in stop delay relative to the existing situation. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 68 Transit and Light Rail Transit in the City of Richmond Table 4-18 Cross Street Stop Delay and Level of Service Approach Existing Stop Delay Existing Level of Service GE = 10 GE = 15 GE = 20 GE = 30 Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS Change in Stop Delay (sec) • LOS SO-m Detectors, Far Side Stops, Phase Skipping Westminster EB 96.87 F - 10.63 F -21.40 F - 15.33 F -20.87 F Lansdowne WB 27.57 D -3.77 C -2.73 C -2.93 C -2.70 C Lansdowne EB 44.20 E -4.40 D - 4.50 D -2.13 E -5.30 D Leslie WB 27.97 D + 8.43 D + 13.27 E + 14.23 E + 8.43 D 100-m Detectors, Far Side Stops, Phase Skipping Westminster WB 51.90 E + 0.07 E - 12.10 D - 15.07 D -4.53 D Lansdowne EB 44.20 E + 2.20 E -4.30 D + 3.30 E -2.03 E Alderbridge EB 58.00 E + 3.47 F + 1.37 F + 2.50 F + 2.60 F I50-m Detectors, Far Side Stops, Phase Skipping Park WB 14.57 B + 1.00 C + 1.53 C -0.40 B + 1.47 C Saba WB 22.77 C + 2.77 D + 2.47 D + 0.10 C -0.13 C Westminster EB 96.87 F - 10.90 F - 10.47 F - 17.40 F -7.37 F University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 69 Transit and Light Rail Transit in the City of Richmond 4.1.4 Green Extension Effectiveness Green extension effectiveness is measured as the ratio of transit vehicles that clear an intersection during a green extension provided by the signal controller after receiving a TSP call to the total number of transit vehicles that request a green extension. This is a critical measure of how effectively green extension is provided by the TSP strategy and is affected most significantly by detector and transit stop location and the length of the green extension. 4.1.4.1 Detector and Transit Stop Location It is clear that the 50-metre detector locations provide the greatest green extension effectiveness in the corridor as a whole, both with and without phase skipping. The effectiveness decreases as the detector locations move further upstream of the intersections. This is illustrated in Figure 4B. This is logical and was expected prior to running the simulations since the closer a transit vehicle is to an intersection when a green extension is requested, the greater should be the probability that the transit vehicle will clear the intersection during the green extension. However, as was noted earlier, closer detector locations can result in increased delay to transit when the TSP call is received during the red phase and the signal controller must cycle to green. Therefore, the configuration that provides the most efficient green extensions may not be the same as the one that provides the lowest average travel times. When far side transit stop locations are used, the green extension effectiveness is higher than the same TSP scenarios with the existing transit stop configuration. However, the difference between the existing and far side transit stop scenarios is relatively minor. This is illustrated in Figure 4C. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 70 Green Extension Effectiveness Detector Location Comparison—No Phase Skipping 75% | 10 15 20 30 Green Extension (seconds) | ^ ^ M 5 0 m Detectors, Far Side Stops ^ ^ ^ 1 0 0 m Detectors, Far Side Stops — ^ ^ 1 5 0 m Detectors, Far Side Stops | Figure 4B: Compar ison of Green Extension Effectiveness Between Detector Locat ions The overall green extension effectiveness in the corridor is low, ranging from 1.87% to 69.92% without phase skipping and from 0% to 65.76% with phase skipping. The low average green extension efficiency ratios are the result of variation in effectiveness among individual intersections. Some intersections experience efficiency ratios of 100% while others have ratios of 0%, depending on the TSP scenario. It is important to consider why some intersections are able to provide more effective green extensions than others. Detector and transit stop locations are the major variables affecting green extension effectiveness. If a check-in detector is located too far upstream of an intersection, then the oncoming transit vehicle will not be able to travel to and through the intersection before expiration of the extension. This is illustrated by the fact that the closer detector locations result in higher green extension efficiency ratios. Similarly, i f there is a near side transit stop, a bus may dwell at the stop for the duration of the green extension and then be forced to stop at the red light. This is evident from the fact that the far side stop scenarios provide better green extension effectiveness than the existing stop configuration. There are intersections with near side transit stops that do not improve in green extension effectiveness even with an increased green extension length. However, when the transit stop is moved to the far side, a dramatic improvement in effectiveness is realized. In other words, while the green extension effectiveness benefit of far side transit stops is relatively small when the travel corridor as a whole is considered, at individual intersections, the benefit can be significant. This is illustrated in Tables 4-19 and 4-20 University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 71 Transit and Light Rail Transit in the City of Richmond and F i g u r e 4C. These figures show two locations that have near side transit stops. When this stop is moved to the far side of the intersection, there is noticeable improvement in green extension effectiveness at this location. Table 4-19 Green Extension Efficiency Compar ison—Near v. Far Side Stop No. 3 Road Northbound at Westminster Highway—100-Metre Detector Conf igurat ion Green Extension Length Near Side Stop Far Side Stop 10 seconds 0% 16.7% 15 seconds 10.0% 14.3% 20 seconds 11.1% 33.3% 30 seconds 10.0% 100% Green Extensior No. 3 Road Northbound Table 4-20 l Eff iciency Compar ison—Near v. Far Side Stop at Alderbr idge Way—50-Metre Detector Conf igurat ion Green Extension Length Near Side Stop Far Side Stop 10 seconds 0% 50.0% 15 seconds 0% 66.7% 20 seconds 9.1% 75.0% 30 seconds 9.1% 100% Green Extension Effectiveness Transit Stop Location Comparison-No Phase Skipping 75% Figure 4C: Compar ison of Green Extension Effectiveness Between Transit Stop Locat ions University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 72 4.1.4.2 Green Extension Length In all the studied TSP scenarios, with and without phase skipping, there is a positive relationship between green extension length and green extension effectiveness. In other words, the greater the length of the green extension, the higher the green extension efficiency ratio tends to be. This was expected since a longer green extension should increase the probability that a bus will be able to clear an intersection during the extension. Figures 4D and 4E indicate that increasing the length of the green extension increases the efficiency of green extension provision in the travel corridor under all TSP scenarios. Green Extension Effectiveness Compar ison-No Phase Skipping 80% 70% 60% 50% 20% 10% — • — Existing Detectors, Existing Stops ^ ^ ^ 5 0 m Detectors. Existing Stops ^ ^ " 5 0 m Detectors, Far Side Stops - ^ " 1 0 0 m Detectors, Existing Stops ^ ^ • " 1 0 0 m Detectors, Far Side Stops — * — 1 5 0 m Detectors, Existing Stops • 150 m Detectors, Far Side Stops ^ ^ ^ ^ 10 15 20 30 Green Extension (seconds) Figure 4D: Green Extension Effectiveness—Green Extension Length Comparison University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 73 Transit and Light Rail Transit in the City of Richmond Green Extension Effectiveness Comparison-Phase Skipping 10 15 20 30 Green Extension (seconds) Figure 4E: Green Extension Ef fect iveness—Green Extension Length Compar ison 4.1.4.3 Phase Skipping The implementation of phase skipping does not appear to provide noticeable change in green extension effectiveness relative to the same scenarios without skipping. However, phase skipping was not expected to be a major influence on this measure of effectiveness. Figure 4F shows that green efficiency ratios are similar under different detector configurations with and without phase skipping. This is because phase skipping affects the time to green for the bus phase rather than the ability of a bus to use a green extension. Phase skipping is important when a bus is detected during a red phase while green extension is utilized when the bus phase is green upon receipt of a TSP request. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 74 Transit and Light Rail Transit in the City of Richmond Green Extension Effectiveness Phase Skipping Comparison 10 15 20 30 Green Extension (seconds) ^ ^ " 5 0 m Detectors, Far Side Stops--No Skipping • ^ • ' " 5 0 m Detectors, Far Side Stops-Phase Skipping • 100 m Detectors, Far Side Stops-No Skipping —4—100 m Detectors, Far Side Stops-Phase Skipping Figure 4F: Green Extension Effectiveness: Phase Skipping Compar ison 4.1.4.4 Summary The closer the check-in detectors are to an intersection, the greater the green extension effectiveness appears to be. Moreover, the far side transit stop configuration improves green extension effectiveness relative to the existing transit stop configuration under all of the studied TSP scenarios. Green extension length and effectiveness appear to be positively related. That is, as the length of the green extension increases, the efficiency of green extension provision improves. However, the implementation of phase skipping does not significantly improve green extension efficiency in any of the studied TSP scenarios. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 75 4.2 Express Bus Service With Conditional Transit Signal Priority The difference between unconditional and conditional TSP for express bus service is frequency of priority provision. This is discussed extensively in the literature review component of this paper. Under unconditional TSP, every bus is given priority at every intersection after it passes over the check-in detector. However, under conditional TSP, a bus receives priority only i f certain conditions are satisfied. These conditions are set by the transit operator and may include bus occupancy levels, traffic volumes, schedule adherence or other criteria. Schedule adherence has been selected as the TSP criteria for this research. A bus receives priority only if it is 30 seconds or more behind the scheduled arrival time at a transit stop. This lateness threshold is arbitrary but has been set based on input from TransLink and private consultants. One of the aims of conditional TSP is to provide priority only to those buses that actually need it. Signal priority can help put buses that are running late back on schedule by limiting delay at traffic signals. By only providing TSP to a select group of transit vehicles, it is anticipated that there will be less disruption to the overall background signal cycle plan and background automobile traffic operations than unconditional priority. This is due to the fact that cross street signal phases will not be affected by every bus traveling in the corridor, but only those that satisfy pre-defined criteria. The more stringent the criteria, the smaller the number of buses that receive TSP. Consequently, disruption to the signal cycle will be lessened. Conditional TSP is also expected to result in higher average travel times and levels of delay than unconditional TSP since some buses will receive priority and others do not. This scenario is evaluated based on average corridor travel times, travel time variability, delay to transit vehicles and impacts on cross street traffic operations, as in the unconditional bus priority section. However, green extension effectiveness is not evaluated for one main reason. The only difference between the conditional strategy and the unconditional strategy is the frequency of TSP provision. The operating parameters of the buses, the detector and transit stop locations and the physical parameters of the travel corridor remain the same as in the unconditional TSP scenario. Therefore, the green extension efficiency will not be significantly affected by the change from unconditional to conditional priority. Similarly, the detector and transit stop location and the length of the green extension will not affect conditional TSP any differently than unconditional TSP and, thus, it is not necessary to discuss these measures of effectiveness. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 76 Transit and Light Rail Transit in the City of Richmond 4.2.1 Transit Vehicle Travel Time and Delay 4.2.1.1 Transit Stop and Detector Location The simulation results indicate that the use of the 50 and 100-metre detectors results in the lowest travel time and delay values, as was the case in the unconditional signal priority scenarios. This was expected prior to running the simulations since frequency of TSP provision is the only modification from the unconditional TSP strategy. Consequently, the influence of detector and transit stop location was expected to be relatively the same under conditional and unconditional TSP. However, the influence of transit stop location is not as noticeable as in the unconditional TSP scenarios. In some scenarios, the use of far side transit stops results in the lowest travel time and delay results while, in others, the existing stop configuration provides the lowest values. This is due to the fact that some buses do not receive priority. While far side transit stops are beneficial for buses receiving TSP in terms of travel time, this may not be the case for buses not receiving priority. Thus, average travel time is not consistently lower with far side stops than the existing configuration. Table 4-21 Transit Vehicle Travel Time Compar ison—Detec to r Location Analysis Detector Location Southbound Northbound 50 Metres 598.4 sec 700.7 sec 100 Metres 622.0 sec 721.8 sec 150 Metres 610.7 sec 722.0 sec Table 4-22 Transit Vehicle Travel Time Compar ison—Trans i t Stop Location Analysis Detector Location Southbound Northbound Existing Stops Far Side Stops Time Benefit of Far Side Stops Existing Stops Far Side Stops Time Benefit of Far Side Stops 50 Metres 598.4 593.1 + 5.3 700.7 708.8 -8.1 100 Metres 622.0 573.5 + 48.5 721.8 734.9 -13.1 150 Metres 610.7 601.3 + 9.4 722.0 759.9 -37.9 Table 4-23 Total Transit Vehicle Delay Compar ison—Detec to r Location Analysis Detector Location Southbound Northbound 50 Metres 377.8 sec 463.8 sec 100 Metres 402.6 sec 485.4 sec 150 Metres 391.4 sec 486.1 sec University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 77 Transit and Light Rail Transit in the City of Richmond Table 4-24 Total Transit Vehicle Delay Compar ison—Trans i t Stop Locat ion Analysis 15-Second Green Extension, No Phase Skipping Detector Location Southbound Northbound Existing Far Side Time Benefit of Existing Far Side Time Benefit of Stops Stops Far Side Stops Stops Stops Far Side Stops 50 Metres 377.8 378.2 -0.4 463.8 476.0 -12.2 100 Metres 402.6 353.5 + 49.1 485.4 499.2 -13.8 150 Metres 391.4 382.2 + 9.2 486.1 527.0 -40.9 4.2.1.2 Phase Skipping It is quite clear that the implementation of phase skipping results in significant transit vehicle travel time and delay improvements over the same scenarios without phase skipping. In all other scenarios, travel time reductions range from 32.2 seconds to 124.9 seconds. These reductions are largely the result of decreases in the average time to cycle to green for the bus phase upon receipt of a TSP call during a red phase. Tables 4-25 and 4-26 show the effects of phase skipping on total delay values and travel times with a 15-second green extension. It is clear that a large proportion of the reductions in travel time that result from the implementation of phase skipping are due to reductions in total transit vehicle delay. Table 4-25 Total Transit Vehicle Delay Compar ison—Phase Skipping Analysis Detector Location Southbound Northbound No Skipping Skipping Time Benefit of Skipping No Skipping Skipping Time Benefit of Skipping 50 Metres . 378.2 288.6 89.6 sec 476.0 403.6 72.4 sec 100 Metres 353.5 275.9 77.6 sec 499.2 402.4 96.8 sec 150 Metres 382.2 285.5 96.7 sec 527.0 447.6 79.4 sec Table 4-26 Transit Vehicle Travel Time Compar ison—Phase Skipping Analysis Detector Location Southbound Northbound No Skipping Time Benefit of No Skipping Time Benefit of Skipping Skipping Skipping Skipping 50 Metres 593.1 507.9 85.2 sec 708.8 636.9 71.9 sec 100 Metres 573.5 497.3 76.2 sec 734.9 636.9 98.0 sec 150 Metres 601.3 538.5 62.8 sec 759.9 686.3 73.6 sec University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 78 Transit and Light Rail Transit in the City of Richmond 4.2.1.3 U n c o n d i t i o n a l versus C o n d i t i o n a l P r i o r i t y Unconditional TSP for express bus service results in lower average travel times than conditional TSP with the 50- and 100-metre detector locations and with both the existing and far side transit stop configurations. This is the case with and without phase skipping and under all green extensions. When 150-metre detector locations are used, the travel times between unconditional and conditional TSP strategies are more comparable. However, this is more likely the result of the relative ineffectiveness of 150-metre detectors in providing reliable and efficient transit signal priority rather than the differences between unconditional and conditional TSP strategies. It should be noted, and will be further emphasized later, that unconditional priority with light rail transit results in significantly lower travel times than express bus service with both unconditional and conditional priority. Tab les 4-27 a n d 4-28 compare travel times with unconditional and conditional priority under a 15-second green extension. The majority of scenarios experience lower travel times with unconditional priority than conditional priority. Those scenarios that have lower travel times with conditional TSP only realize a few seconds of travel time benefit. It is likely that the conditional TSP scenarios with the lowest travel times also had the highest number of TSP provisions. Conditional TSP is expected to have higher travel times than unconditional TSP because the transit vehicles that do not receive priority will have greater delay at intersections than those vehicles that do receive priority. Table 4-27 Southbound Travel Time Compar ison—Uncond i t iona l v. Condit ional Bus Priority Detector Location Unconditional TSP Conditional TSP Difference from Unconditional to Conditional 50 Metres 521.1 sec 593.1 sec + 13.8 % 100 Metres 480.9 sec 573.5 sec + 19.3% 150 Metres 519.6 sec 601.3 sec + 15.7% 50 Metres 475.7 sec 507.9 sec + 6.8% 100 Metres 455.8 sec 497.3 sec + 9.1% 150 Metres 506.3 sec 538.5 sec + 6.4% Table 4-28 Northbound Travel Time Compar ison—Uncond i t iona l v. Condit ional Bus Priority Detector Location Unconditional TSP Conditional TSP Difference from Unconditional to Conditional 50 Metres 679.3 sec 708.8 sec + 4.3% 100 Metres 672.7 sec 734.9 sec + 9.2% 150 Metres 694.2 sec 759.9 sec + 9.5% 50 Metres 625.9 sec 636.9 sec + 1.7% 100 Metres 635.5 sec 636.9 sec + 0.2% 150 Metres 673.5 sec 686.3 sec + 1.9% University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 79 Transit and Light Rail Transit in the City of Richmond 4.2.1.4 Summary The most effective detector locations in terms of travel time are 50 and 100 metres upstream, depending on the TSP scenario considered. However, it is important to note that the majority of scenarios result in greater travel times than the existing TSP strategy for buses. The use of far side stops results in lower travel times than the existing stops in the southbound corridor without phase skipping and in the northbound corridor with phase skipping. However, southbound with phase skipping and northbound without skipping experience lower travel times with the existing stop configuration. As was noted in the unconditional bus TSP section, green extension appears to have little or no correlation with travel time. Finally, phase skipping results in significant travel time benefits relative to no phase skipping in the majority of scenarios. 4.2.2 Travel Time Variability 4.2.2.1 Unconditional Priority v. Conditional Priority There is the general trend that conditional priority results in higher travel time variability than unconditional priority. This is illustrated in Table 4-29 and Figures 4G and 4H. This is logical since unconditional TSP always provides priority to a bus at every intersection while conditional TSP only gives priority when certain conditions are satisfied. Therefore, at some locations a bus will receive priority but not at others. This will differ between buses and result in high travel time variability. Unconditional priority does not provide lower S D % M values than conditional in all scenarios, but the difference between these values tends to be smaller when conditional priority results in lower S D % M values than when unconditional priority results in lower values. It is important to note that in many of the scenarios where conditional travel time variability is lower than unconditional TSP, the difference is relatively minor. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 80 Travel Time Variability Comparison-Unconditional v. Conditional Bus Priority 50-Metre Detectors, Far Side Transit Stops-Southbound 5.5% Figure 4G: Travel Time Variabil i ty Compar ison—50-Metre Detectors, Far Side Stops Table 4-29 Travel Time Variabil i ty Compar ison—Uncond i t iona l Priority v. Condit ional Priority Southbound Corridor, 15-Second Green Extension, Phase Skipping Detector Location Existing Transit Stops Far Side Transit Stops Unconditional TSP Conditional TSP Unconditional TSP Conditional TSP Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M 50 Metres 10.05 2.10% 21.67 4.07% 4.61 0.97% 26.00 5.12% 100 Metres 6.84 1.40% 14.40 2.90% 2.06 0.45% 9.63 1.94% 150 Metres 10.16 1.95% 5.51 1.07% 8.85 1.75% 8.37 1.55% University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 81 T r a v e l T i m e V a r i a b i l i t y C o m p a r i s o n - U n c o n d i t i o n a l v . C o n d i t i o n a l B u s P r i o r i t y 50-Metre Detectors, Existing Transit Stops-Southbound 6.0% • 0.0% I GE = 10 GE = 15 GE = 20 GE = 30 Green Extension (seconds) |^^ MUncondit ional, No Skipping —•—Conditional, No Skipping ^^^Uncondit ional, Skipping —•—Conditional. Skipping"] Figure 4H: Travel Time Variability Comparison—50-Metre Detectors, Existing Stops University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 82 Transit and Light Rail Transit in the City of Richmond 4.2.3 Cross Street Traffic Operations 4.2.3.1 Total Delay The cross street traffic operational results for conditional TSP are quite similar to the unconditional priority scenarios, and will be only lightly touched upon here. For the most part, there are no significant changes in delay to cross street traffic associated with implementing the proposed conditional TSP strategies. Where delay increases, it usually amounts to only a few seconds. As in unconditional TSP, the location of transit stops and check-in detectors has little or no influence on the delay experienced by vehicles on cross street approaches. There appears to be no noticeable relationship between the length of the green extension applied by a TSP strategy and the amount of delay experienced by cross street traffic. That is, whether there is an increase or decrease in delay, and the scale of the increase or decrease, is not influenced by the length of the green extension. Phase skipping causes more approaches to experience increases in delay relative to existing than no phase skipping, and the increases are greater with phase skipping. The results indicate that more approaches experience delay increases under unconditional TSP than conditional TSP. This is logical since all of the approaches are having their signal cycles modified by the arrival of every transit vehicle under unconditional TSP. However, with conditional TSP, only some intersections experience TSP modifications with the arrival of a transit vehicle. Therefore, the amount of disruption experienced by cross street traffic is less under conditional than unconditional TSP. The simulation results do not indicate that either unconditional or conditional TSP consistently provides smaller increases and larger decreases in cross street delay relative to the existing situation than the other. While it is clear that unconditional TSP causes more cross street approaches to experience increases in delay than conditional priority, the scale of these increases cannot be definitively said to be greater. However, the use of phase skipping increases the number of cross street approaches experiencing increases in delay, reduces delay reductions and increases delay increases compared to the existing situation under unconditional and conditional TSP strategies. Table 4-30 provides a comparison of the number of cross street approaches that experience delay increases relative to existing under the different TSP scenarios. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 83 Table 4-30 Number of Cross Street Approaches Experiencing Increases in Delay with TSP Far Side Transit Stops Scenario Conditional TSP Unconditional TSP No Skipping Phase Skipping No Skipping Phase Skipping 50-Metre Detectors, Far Side Transit Stops, G E = 10 4 9 1 1 50-Metre Detectors, Far Side Transit Stops, G E = 15 8 8 9 8 50-Metre Detectors, Far Side Transit Stops, G E = 20 5 11 3 7 50-Metre Detectors, Far Side Transit Stops, G E = 30 5 8 7 5 100-Metre Detectors, Far Side Transit Stops, G E = 10 4 9 6 13 100-Metre Detectors, Far Side Transit Stops, G E = 15 4 7 7 8 100-Metre Detectors, Far Side Transit Stops, G E = 20 6 9 7 12 100-Metre Detectors, Far Side Transit Stops, G E = 30 5 7 5 6 150-Metre Detectors, Far Side Transit Stops, G E = 10 4 9 5 8 150-Metre Detectors, Far Side Transit Stops, G E = 15 5 11 5 10 150-Metre Detectors, Far Side Transit Stops, G E = 20 5 12 8 12 150-Metre Detectors, Far Side Transit Stops, G E = 30 3 8 7 9 4.2.3.3 Level of Service The level of service of traffic operations on the cross street approaches is based on the average stop delay per vehicle. The simulation results for conditional bus TSP are similar to unconditional bus TSP. In general, the implementation of the proposed TSP strategies does not greatly affect cross street traffic operations relative to the existing situation. Transit stop and check-in detector location do not greatly affect cross street traffic delay and there is no trend indicating that any transit stop/detector configuration results in lower cross street traffic delay than the others. Furthermore, the length of the green extension applied in the TSP strategy is not significantly related to the level of delay experienced by cross street traffic. As in the unconditional bus TSP scenarios, the use of phase skipping causes greater increases and smaller decreases in cross street delay than the same scenarios without skipping. It is clear that conditional bus TSP causes fewer cross street approaches to experience reductions in level of service than unconditional bus TSP or unconditional LRT TSP. This is due to the fact that only transit vehicles that qualify for priority will have impacts on traffic signal cycles rather than all of the transit vehicles traveling in the corridor. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 84 Transit and Light Rail Transit in the City of Richmond Table 4-31 outlines examples of how the different TSP strategies affect cross street stop delay. At some locations, the implementation of TSP causes significant reductions in stop delay, resulting in level of service improvement. For example, this is seen at the westbound approach to the No. 3 Road corridor on Westminster Highway with 50-metre detectors. Some locations have existing stop delay levels that are so high that even significant reductions in stop delay do not improve level of service from LOS F. This is illustrated by the eastbound Westminster Highway approach. Finally, some locations experience reductions in level of service with only small increases in stop delay. This is the result of the existing stop delay being at the upper limit of the current LOS designation. This can be seen on the westbound Park Road approach. Table 4-31 Cross Street Stop Delay and Level of Service Far Side Transit Stops, Phase Skipping Approach Existing Stop Delay Existing Level of Service G E = 10 G E = 15 GE = 20 GE = 30 Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS 50-m Detectors, Far Side Stops, Phase Skipping Park WB 14.57 B + 2.43 C -0.03 B -0.60 B -0.50 B Westminster EB 96.87 F -9.63 F -21.13 F - 13.60 F - 11.90 F Alderbridge WB 50.8 E + 24.7 F + 30.47 F + 31.80 F + 38.97 F 100-m Detectors, Far Side Stops, Phase Skipping Park WB 14.57 . B + 2.10 C + 1.47 C + 0.53 C -0.53 B Westminster EB 96.87 F - 13.20 F -8.53 F - 14.77 F - 15.80 F Alderbridge WB 50.80 E + 18.03 F + 14.10 F + 22.53 F + 18.77 F ISO-m Detectors, Far Side Stops, Phase Skipping Park WB 14.57 B + 1.70 C -0.37 B + 1.83 C + 0.20 B Westminster EB 96.87 F - 12.03 F -5.70 F - 10.27 F -3.00 F Alderbridge WB 50.80 C + 13.77 F + 12.57 F + 5.70 E + 2.77 E University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 85 Transit and Light Rail Transit in the City of Richmond 4.3. SUMMARY AND DISCUSSION OF RESULTS FOR EXPRESS BUS SIGNAL PRIORITY 4.3.1 Transit Vehicle Travel Time and Delay Transit travel time and delay are inextricably linked. The majority of travel time reductions are the result of reductions in the delay that transit vehicles experience while traveling through the corridor. Thus, the trends noticed in travel time with respect to detector location, transit stop location, phase skipping and green extension length are basically the same as in the total delay results. Bus travel time and delay under transit signal priority is greatly affected by the locations of transit stops and check-in detectors. There is currently a mix of near side, mid-block and far side transit stops in the No. 3 Road corridor. The simulation results indicate that the use of far side stops at all intersections would result in lower travel times than the same scenarios with the existing transit stop configuration. This is most likely the consequence of a reduction in double stopping, which occurs when a bus must stop at a near side transit stop and then again at the intersection for a red phase. The use of near side transit stops becomes a significant problem when a bus is detected during a green phase, particularly if the check-in detector is located upstream of the transit stop, which is usually the case. When the bus crosses the detectors, a green extension is applied. However, the bus must first stop at the near side transit stop to allow passengers to board and alight. In many cases, except where there are very long green extensions, the bus will dwell at the stop for the duration of the extension, resulting in double stopping. There is no one check-in detector configuration that provides the lowest travel times in all the studied TSP scenarios. This is because the effectiveness of detector location is influenced by many factors. Bus travel speed is critical as it determines the time necessary for a bus to travel from the detector to the intersection. This is an important consideration in TSP because the aim is to minimize the time required for a bus to clear an intersection. If a detector is located too close to an intersection, the bus will be required to decelerate or stop while the signal controller works to provide green for the bus phase. In general, the slower the travel speed, the closer the detectors can be to the intersection without causing increased delay to transit and, thus, increased travel time. This is seen in the simulation results. The southbound corridor experienced the lowest travel times with detectors located 100 metres upstream while the northbound corridor experienced the best travel time performance with 50-metre detectors. This difference is most likely accounted for by the fact that the northbound traffic flow is heavier in the shared travel segments of the No. 3 Road corridor. The use of phase skipping is also an important factor when determining appropriate detector locations. In general, when phase skipping is used, the detector locations can be closer to the intersections than without University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 86 Transit and Light Rail Transit in the City of Richmond phase skipping without causing greater delay to transit vehicles. This is because the use of phase skipping reduces the average time to cycle to green after receipt of the TSP call. When detectors are located significantly far from an intersection, the green phase may be provided and then expire before the bus can clear the intersection, especially i f there is a near side transit stop. When there is no phase skipping, the average time to cycle to green is longer than with phase skipping. Overall, detectors located relatively close to the intersections provide less average travel time with phase skipping than detectors located further upstream. The length of a green extension provided by the TSP strategy is not as influential on travel times as was expected prior to running the simulations. In all the studied scenarios in both travel directions, there is no discernible relationship between the green extension length and travel time. Travel times tend to fluctuate up and down as the green extension increases from 10 to 30 seconds. As explained in the analysis, the lack of correlation between green extension length and travel time may be due to variability in the simulation model. Due to the effects of random probability distributional properties of the model, the majority of TSP calls that take place during a simulation run may occur during a red phase so that green extension is not necessary. The travel time would then depend on the delay experienced in cycling from red to green for the bus phase. Any relationship that exists between green extension length and travel time would be diluted by travel time results that did not incorporate green extension. If many of the travel times used to calculate the average travel time do not depend on length of green extension, then any relationship between these variables is not evident. Phase skipping is a critical element of all of the studied TSP strategies. It is obvious that the use of phase skipping reduces average travel times as this was observed under all the studied TSP strategies with all green extensions. This is the result of the reduction in average time to cycle to green after receipt of a TSP call. Without phase skipping, the signal controller must cycle through the minimum green times of all phases preceding the bus phase. This allows for high variability in the time to green because it depends on where in the cycle the TSP call is received. However, with phase skipping, the green bus phase is given immediately after the end of the minimum green of the current phase. This reduces the average time to green and variability in this time. This allows buses to clear with less average delay and, consequently, provides lower travel times. These trends prevail in both unconditional and conditional bus signal priority scenarios. The difference is that travel times are generally higher with conditional TSP than unconditional TSP. The northbound delay and travel time values are consistently higher than the southbound direction. There are three main reasons for this result. Firstly, there are currently eight transit stops in the northbound direction and only six transit stops in the southbound direction. Consequently, a transit vehicle will experience more delay University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 87 Transit and Light Rail Transit in the City of Richmond traveling northbound than southbound in allowing passengers to board and alight. Secondly, all of the transit stops in the southbound direction are either mid-block or far side stops. However, in the northbound direction, three of the eight transit stops are near side stops, which can result in significant delay due to double stopping and, consequently, greater travel time than in the southbound direction. Finally, traffic volume is higher in the northbound corridor than the southbound. This can result in transit vehicle travel speeds in the shared travel segments of the corridor being lower in the northbound direction than the southbound. 4.3.2 Travel Time Variability Variability in corridor travel time from one trip to another adversely affects the consistency and reliability of arrival times at transit stops and the regularity of headways between transit vehicles. In general, travel time variability affects the overall attractiveness and convenience of transit as a mode of transportation. A major aim of transit signal priority is to reduce variability in travel time and thus improve the reliability of transit arrivals and trip times. This allows for better trip planning and transfers to other routes and transportation modes. There were certain expectations prior to running the simulations: • Phase skipping should provide less variation in travel time since there should be less variability in the time to cycle to green for the bus phase after a TSP call is received than in no phase skipping scenarios. • Longer green extensions should provide more consistent travel times because longer extensions should increase the probability of a transit vehicle clearing an intersection before the extension expires. • Some detector locations should provide more efficient TSP than others; however, it was not known which location would be best and how this would affect travel time variability. None of the studied detector locations consistently provides the lowest travel time variability under all TSP scenarios and green extensions. In general, the 50-metre detectors provide the lowest standard deviation as percentage of mean (SD%M) values in the southbound corridor while the existing detectors provide the lowest S D % M values in the northbound. The location of the transit stops is not a major factor in travel time variability as the existing and far side configurations result in comparable travel time variability values. Similarly, the length of the green extension does not appear to have any considerable influence on travel time variability. However, this may be due to the simulation model variability discussed earlier. The implementation of phase skipping University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 88 Transit and Light Rail Transit in the City of Richmond appears to reduce the variability in travel times for unconditional bus TSP, which can most likely be explained by a reduction in the delay in cycling to green upon receipt of the TSP call and reduction in variability in this time. However, this is not as evident in conditional bus TSP, which is likely the result of a smaller percentage of transit vehicles being granted priority than in unconditional TSP. This is likely due to the fact that a smaller percentage of transit vehicles is granted priority under conditional priority compared to unconditional priority. 4.3.3 Cross Street Traffic Operations The impact of each of the TSP strategies on cross street traffic operations was gauged in terms of average total delay per vehicle and level of service as measured by stop delay. The simulation results indicate that the majority of cross street approaches would not experience significant changes in operational efficiency relative to the existing situation with the implementation of any of the proposed TSP strategies. The transit stop and check-in detector locations are not the critical parameters affecting cross street delay. There is no consistency in which transit stop or detector configurations result in the greatest delay to cross street traffic. The most important parameter is the use of phase skipping, as was expected prior to running the simulations. When phase skipping is implemented, the number of cross street locations that experience increases in delay relative to existing increases. At locations where there are delay reductions with phase skipping, the reductions are smaller than at the same locations without phase skipping. The scale of increases in delay are heightened by phase skipping and some locations that experienced less delay than the existing situation without phase skipping experience increases with phase skipping. The implementation of phase skipping results in greater cross street delay because cross street phases are skipped to serve the bus phase when a TSP call is received. Traffic on the approaches that have their phase skipped must then wait for the next scheduled phase. Contrary to preliminary expectations, the length of the green extension applied in the TSP strategy does not significantly affect cross street traffic operations. While there are some intersections that show a general pattern of increasing cross street delay with increasing green extension, there is generally up and down fluctuation across the range of green extensions. As discussed in previous sections, this may be due to variability in the simulation model. With respect to level of service on the cross streets, the increases and decreases in stop delay are, for the most part, not great enough to change the level of service from existing, except in a few cases. The only large changes in delay are instances where stop delay is reduced from existing levels. Reductions in level of service occur at locations where the University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 89 Transit and Light Rail Transit in the City of Richmond existing stop delay is close to the upper limit of the current LOS designation so that even a small increase in stop delay results in a change in LOS. These trends are applicable to both unconditional and conditional TSP. It is clear that unconditional TSP causes more cross street approaches to experience delay increases than conditional TSP. However, it is not clear whether unconditional or conditional TSP results in greater scale of increases in delay. This cannot be inferred from the simulation results. 4.3.4 Green Extension Effectiveness The simulation results indicate that the closer the check-in detectors are located to intersections, the higher will be the green extension effectiveness. This was expected prior to running the simulations since the closer the detectors are to the intersection, the greater should be the probability that a bus will be able to clear an intersection during a green extension. However, this parameter alone does not indicate which TSP strategy is the best. Strategies that provide the highest green extension effectiveness could, for example, also result in the greatest adverse impacts on cross street traffic operations. Transit stop location is an important consideration at individual locations, but not in the corridor as a whole. At an individual location, a near side transit stop can make clearing an intersection during a green extension nearly impossible, unless the extension is very long. In these cases, far side stops improve green extension effectiveness. However, in this corridor, with the mix of transit stop locations, the overall green extension efficiency is not greatly improved with the use of all far side transit stop locations. There appears to be a positive relationship between the length of the green extension and the green extension effectiveness. That is, as the green extension length increases, so does the green extension efficiency ratio. This was expected since the longer green extension, the greater the probability that an oncoming bus will be able to clear the intersection during the extension. Although phase skipping is a critical element of TSP in terms of travel time and delay, it is relatively unimportant in terms of green extension effectiveness. This is most likely due to the fact that phase skipping mainly influences the time to provision of a green bus phase when a TSP call is made during a red bus phase. On the other hand, green extension effectiveness is only a factor when the TSP call is made during a green bus phase. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 90 4.4. Light Rail Transit with Unconditional Transit Signal Priority Introduction This scenario is based on a hypothetical at-grade light rail transit (LRT) system that operates in the middle of the No. 3 Road corridor in the right-of-way currently used by buses, as discussed earlier. The physical specifications of the LRT line simulated in this research are based on designs created by a private consulting firm in Vancouver to provide options for the corridor. Since the objective of LRT is to allow LRVs to travel unimpeded through the corridor, stopping only to allow passengers to board and alight, unconditional priority with and without phase skipping has been used in simulating LRT operations. Thus, it is possible to compare the relative efficiency of bus service and LRT under unconditional priority. Detector location, green extension length and phase skipping will be analyzed as in the bus scenarios. However, transit stop location is not as easily modified with LRT as in bus scenarios. Therefore, the planned station locations are maintained in all scenarios. 4.4.1 Transit Vehicle Travel Time And Delay 4.4.1.1 Detector Locations There does not appear to be a significant difference in delay and, consequently, travel times, among the different studied detector locations. Without phase skipping, the 100-metre detectors provide the lowest travel times in the southbound corridor, except with a green extension 15 seconds. However, the range of travel times among detector locations is only 13 seconds. Thus, it appears that detector location is not a critical factor in southbound travel time. In the northbound corridor, the 100-metre detectors without phase skipping result in the lowest travel times with extensions of 15 seconds or greater. As in the southbound corridor, the difference in travel time values among the studied detector locations is relatively small. The range between the highest and lowest travel times under any green extension is 24.5 seconds. With phase skipping, the southbound and northbound corridors experience similar travel times with the 50- and 100-metre detector configurations. The range from highest to lowest travel time under any green extension between the 50-metre and 100-metre detectors in the southbound direction is 15.5 seconds. In the northbound, the range is only 9 seconds. On the other hand, the 150-metre detectors provide noticeably higher travel times, which indicates that 150 metres may be too far upstream to provide reliable and University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 91 Transit and Light Rail Transit in the City of Richmond efficient signal priority. The total delay and travel time results with each detector location under a 15-second green extension are summarized in Tables 4-32. Table 4-32 Transit Vehicle Travel Time Compar ison—Detec to r Locat ion Analysis 15-Second Green Extension, No Phase Skipping Detector Location Southbound Northbound 50 Metres 378.8 sec 503.1 sec 100 Metres 378.8 sec 488.2 sec 150 Metres 374.9 sec 512.7 sec 4.4.1.2 Green Extension Length The simulation results do not indicate any definite correlation between the length of the green extension and the average L R V travel time. This is illustrated in Figures 41 and 4J. While there are patterns in some scenarios, there is no one pattern that consistently appears in a significant number of scenarios. It is highly unlikely that there is no relationship between transit travel time and the length of the green extension applied by a TSP strategy. As explained earlier, the lack of correlation in the simulation results between travel time and green extension length is likely due to a limited number of simulation runs that incorporated green extensions being included in the sample used to calculate average travel time. As a result, the travel times would be based on factors other than green extension length and the relationship between these variables would become diluted. In other words, inherent variability in the simulation model is likely responsible for these results. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 92 Light Rail Transit Travel Time Comparison Southbound Corridor 400 Figure 41: Light Rail Transit Travel Time Compar i son—Sou thbound Light Rail Transit Travel Time Comparison Northbound Corridor Green Extension (seconds) •50 m Detectors, No Skipping -50 m Detectors, Skipping •100 m Detectors, No Skipping -100 m Detectors, Skipping •150 m Detectors, No Skipping -150 m Detectors, Skipping Figure 4J: Light Rail Transit Travel Time Compar i son—Nor thbound University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 93 Transit and Light Rail Transit in the City of Richmond 4.4.1.3 Phase Skipping It is clear that the implementation of phase skipping provides significant reductions in transit delay and travel time under all scenarios and in both directions relative to the same scenarios without phase skipping. The greatest differences between skipping and no skipping occur with the 50-metre detector configuration. This is due to the high delay that can result without phase skipping with detectors located very close to an intersection when there are several phases that must be served before the bus phase. In the southbound corridor, the travel time benefit of phase skipping ranges from 49 to 62 seconds and in the northbound, the range is 50 to 62 seconds. Travel time reductions of approximately one minute are significant with overall corridor travel times are 7 to 9 minutes. These results are summarized in Tables 4-33 and 4-34. It is clear that the implementation of phase skipping produces significant reductions in transit delay, which are largely responsible for the corresponding reductions in travel time. T a b l e 4-33 L R T T o t a l De lay C o m p a r i s o n — P h a s e S k i p p i n g A n a l y s i s ( 1 5 - s e c o n d G r e e n E x t e n s i o n ) Detector Conf igura t ion Southbound N o r t h b o u n d No Sk ipp ing Sk ipp ing Benefi t o f Sk ipp ing No Sk ipp ing Sk ipp ing Benefi t o f Sk ipp ing 50 metres 159.1 113.0 46.1 sec 275.7 218.6 57.1 sec 100 metres 158.7 116.1 42.6 sec 261.0 226.2 34.8 sec 150 metres 155.1 127.1 28.0 sec 289.0 241.8 47.2 sec T a b l e 4-34 L R T T r a v e l T i m e C o m p a r i s o n — P h a s e S k i p p i n g A n a l y s i s ( 1 5 - s e c o n d G r e e n E x t e n s i o n ) Detector Conf igura t ion Southbound N o r t h b o u n d No Sk ipp ing Sk ipp ing Benef i t o f Sk ipp ing No Sk ipp ing Sk ipp ing Benefi t o f Sk ipp ing 50 metres 378.8 332.6 46.2 sec 503.1 445.4 57.7 sec 100 metres 378.8 335.7 43.1 sec 488.2 454.3 33.4 sec 150 metres 374.9 346.5 28.4 sec 512.7 470.0 42.7 sec 4.4.1.4 LRT versus Unconditional and Conditional Express Bus Priority Light rail transit with unconditional priority results in significantly lower transit delay and travel times than express bus service with either unconditional or conditional TSP. Tables 4-35 through 4-38 compare delay and travel time results for unconditional TSP with express bus service and LRT service with the planned station locations for a 15-second green extension. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 94 Table 4-35 Southbound Total Delay Comparison—Unconditional Bus v. Unconditional LRT Priority 15-second Green Extension, Far Side Transit Stops, Phase Skipping Unconditional Bus TSP Unconditional L R T TSP Difference 50 Metres 258.2 113.0 145.2 100 Metres 237.7 116.1 121.6 150 Metres 285.5 127.1 158.4 Table 4-36 Northbound Total Delay Comparison—Unconditional Bus v. Conditional LRT Priority 15-second Green Extension, Far Side Transit Stops, Phase Skipping Detector Location Unconditional Bus TSP Unconditional LRT TSP Difference 50 Metres 392.2 218.6 173.6 100 Metres 401.4 226.2 175.2 150 Metres 447.6 241.8 205.8 Table 4-37 Southbound Travel Time Comparison—Unconditional Bus v. Unconditional LRT Priority 15-second Green Extension, Far Side Transit Stops, Phase Skipping Detector Location Unconditional Bus TSP Unconditional LRT TSP Difference 50 Metres 475.7 sec 332.6 sec 143.1 sec 100 Metres 455.8 sec 335.4 sec 120.4 sec 150 Metres 506.3 sec 346.5 sec 159.8 sec Table 4-38 Northbound Travel Time Comparison—Unconditional Bus v. Unconditional LRT Priority 15-second Green Extension, Far Side Transit Stops, Phase Skipping Detector Location Unconditional Bus TSP Unconditional LRT TSP Difference 50 Metres 625.9 sec 445.4 sec 180.5 sec 100 Metres 635.5 sec 454.3 sec 181.2 sec 150 Metres 673.5 sec 470.0 sec 203.5 sec In the southbound corridor, without phase skipping, the LRT travel times range from 21.2% to 29.6% less than unconditional bus priority and 29.8% to 39.1% less than conditional bus priority. With phase skipping, LRT travel times range from 26.4% to 32.2% less than unconditional bus TSP and from 29.4% to 37.7% less than conditional bus TSP. In the northbound corridor, the LRT travel times without phase skipping range from 29.4% to 31.4% less than unconditional bus TSP and from 26.9% to 33.6% less than conditional bus priority. With phase skipping, LRT travel times range from 28% to 33.8% less than unconditional bus TSP and from 27.6% to 35.1% less than conditional bus TSP. Figures 4K and 4L compare travel times for LRT, unconditional bus TSP and conditional bus TSP in the southbound corridor with 50-metre detectors across the range of studied green extensions. These scenarios show the typical relationship between these TSP strategies in terms of travel time. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 95 Transit and Light Rail Transit in the City of Richmond T r a v e l T i m e C o m p a r i s o n - S o u t h b o u n d Hard and Conditional Bus TSP v. Hard LRT TSP-50-Metre Detectors, Existing Transit Stops 550 525 500 400 375 350 325 Green Extension (seconds) "Hard Bus Priority, No Skipping - Hard Bus Priority, Skipping •Conditional Bus Priority, No Skipping • - Conditional Bus Priority, Skipping •Hard LRT Priority, No Skipping -Hard LRT Priority. Skipping Figure 4K: Travel Time Comparison—50-Metre Detectors, Existing Transit Stops T r a v e l T i m e C o m p a r i s o n - S o u t h b o u n d Hard and Conditional Bus TSP v. Hard LRT TSP-SO-Metre Detectors, Far Side Transit Stops Green Extension (seconds) •Hard Bus Priority, No Skipping - Hard Bus Priority, Skipping •Conditional Bus Priority, No Skipping • -Conditional Bus Priority, Skipping -Hard LRT Priority (50 m) -Hard LRT Priority, Skipping (50m) Figure 4L: Travel Time Comparison—50-Metre Detectors, Far Side Transit Stops University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 96 Transit and Light Rail Transit in the City of Richmond 4.4.1.5 Summary The studied detector locations provide comparable LRT travel times with and without phase skipping. The length of the green extension has little or no influence on the travel time and delay values in the corridor, which is most likely due to variability in the simulation model. The most critical factor is whether phase skipping is implemented. Phase skipping provides reductions in travel time up to and exceeding one minute relative to the same scenarios without skipping due to significant reductions in total transit delay. Finally, LRT travel times are significantly lower than under unconditional or conditional bus TSP under all the studied scenarios. 4.4.2 Travel Time Variability 4.4.2.1 Detector Locations From the simulation results, it appears that none of the studied detector configurations consistently results in the lowest travel time variability. In the southbound corridor, it is clear that the 150-metre detectors result in the highest travel time variability. Without phase skipping, the 100-metre detector configuration tends to produce the least variability in travel times while the 50-metre detectors result in the least variable travel times with phase skipping. In the northbound corridor, the 150-metre detector configuration results in the lowest travel time variability when there is no phase skipping in place. With phase skipping, the lowest travel time variability results from the use of 100-metre detectors. These results are summarized in Table 4-39 for a green extension of 15 seconds. Table 4-39 Travel Time Variability Comparison—Detector Location Analysis 15-Second Green Extension Detector Locat ion Southbound N o r t h b o u n d No Phase Skipping Standard Deviat ion S D % M Standard Deviat ion S D % M 50 Metres 7.54 1.99% 7.87 1.56% 100 Metres 3.05 0.81% 9.49 1.94% 150 Metres 25.02 6.67% 7.16 1.40% Phase Skipping 50 Metres 3.32 1.00% 8.14 1.83% 100 Metres 7.03 2.09% 3.19 0.70% 150 Metres 10.61 3.06% 12.10 2.57% 4.4.2.2 Green Extension Length In general, there is no correlation evident between the length of the green extension provided in the TSP strategy and the variability of travel times based on the simulation University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 97 Transit and Light Rail Transit in the City of Richmond results, as measured by S D % M values. While there are some localized patterns between . the change in travel time variability and the length of the green extension, this is not a prominent trend. For the most part, the results indicate an up-and-down fluctuation in travel time variability with an increase in green extension length. This is illustrated in Table 4-40. Table 4-40 Transit Vehicle Travel Time Variabil i ty Compar ison (Standard Deviation as % of Mean) Far Side Transit Stops Detector Location Southbound Northbound GE = 10 GE = 15 G E = 20 G E = 30 GE = 10 GE = 15 G E = 20 G E = 30 No Phase Skipping 5 0 M e t r e s 0 . 8 6 % 1 .99% 2 . 5 0 % 2 . 2 6 % 1 . 0 6 % 1 .56% 0 . 6 6 % 1 .36% 100 M e t r e s 1 .74% 0 . 8 1 % 1 .16% 1 . 0 4 % 1 . 8 5 % 1 . 9 4 % 1 . 0 7 % 0 . 9 3 % 150 M e t r e s 1 .92% 6 . 6 7 % 0 . 3 5 % 3 . 5 3 % 1 . 3 3 % 1 .40% 0 . 8 0 % 0 . 4 3 % Phase Skipping 5 0 M e t r e s 0 . 6 1 % 1 .00% 0 . 6 6 % 1 .38% 1 .72% 1 .83% 1 .69% 1 .19% 100 M e t r e s 2 . 2 3 % 2 . 0 9 % 2 . 5 5 % 0 . 3 1 % 1.46% 0 . 7 0 % 2 . 6 1 % 0 . 1 6 % 150 M e t r e s 4 . 2 5 % 3 . 0 6 % 3 . 3 1 % 3 . 2 0 % 2 . 7 3 % 2 . 5 7 % 1 . 3 5 % 1 .22% 4.4.2.3 Phase Skipping From the simulation results, it is clear that the implementation of phase skipping does not significantly affect travel time variability. There are scenarios where phase skipping results in lower travel time variability than the same scenarios without phase skipping. For example, this is seen with 150-metre detectors in the southbound corridor. However, this trend is not consistently seen throughout the studied scenarios. This indicates that phase skipping is not a critical factor in travel time variability but rather that other factors are exerting more significant influence on the value of the standard deviation as a percentage of the mean. Table 4-41 illustrates these results for a 15-second green extension. Table 4-41 Travel Time Variabil i ty Compar ison—Phase Skipping Analys is 15-Second Green Extension Detector Location Southbound Northbound No Skipping Phase Skipping No Skip] ping Phase Skipping Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M Standard Deviation SD%M 5 0 M e t r e s 7 .54 1 .99% 3 .32 1 .00% 7.87 1 .56% 8.14 1 .83% 100 M e t r e s 3 .05 0 . 8 1 % 7.03 2 . 0 9 % 9 . 4 9 1 . 9 4 % 3 .19 0 . 7 0 % 150 M e t r e s 2 5 . 0 2 6 . 6 7 % 10 .61 3 . 0 6 % 7 . 1 6 1 .40% 12 .10 2 . 5 7 % University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 98 Transit and Light Rail Transit in the City of Richmond 4.4.2.4 L R T Priority v. Express Bus Priority Inspection of the simulation results indicates that neither the L R T system nor the express bus service consistently produces less variable travel times than the other. It has already been noted in the discussion on conditional bus TSP that unconditional TSP results in less variable travel times than conditional TSP. Therefore, unconditional LRT TSP will be compared to unconditional bus TSP. With the 50-metre detector configuration, express bus service results in less travel time variability than LRT with existing transit stops and no phase skipping in the southbound corridor. However, when phase skipping is implemented, LRT results in lower travel time variability than bus service, with either existing or far side transit stop configurations. In the northbound corridor, without phase skipping, there is no travel time variability benefit of either LRT or express bus service. The advantage of one system over the other shifts back and forth across the range of studied green extension lengths. However, with phase skipping, LRT clearly results in more variable travel times than express bus service. It is important to note that the location of transit stops in bus service does not significantly affect travel time variability. In the southbound corridor, with 100-metre detectors and no phase skipping, the LRT system results in less variable travel times than bus service with either the existing or far side transit stop configuration. Without phase skipping, express bus service with far side transit stops is the least variable in travel times, except with a green extension of 30 seconds. In the northbound corridor, with no phase skipping, there is no clear travel time variability benefit of LRT or bus service as S D % M values tend to vary with the length of the green extension. That is, with some green extensions, L R T values are lower and with other extensions, bus S D % M values are lower. When phase skipping is implemented, LRT appears to provide more reliable travel times than bus service. These results are illustrated in Table 4-42. When 150-metre detectors are used without phase skipping in the southbound corridor, the express bus service with far side transit stops provides the least variable travel times, except with a 20-second green extension. When phase skipping is implemented, express bus service provides less travel time variability than LRT, especially when far side transit stops are used. In the northbound corridor, without phase skipping, bus service with far side transit stops provides travel time variability values comparable to LRT. Thus, one transit system does not provide more reliable travel times than the others. However, when phase skipping is implemented, express bus service with far side transit stops generally provides the most consistent travel times. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 99 Transit and Light Rail Transit in the City of Richmond Table 4-42 Travel Time Variability Comparison Unconditional Express Bus TSP v. Unconditional LRT TSP Detector Location Bus, Existing Stops Bus, Far Side Stops Light Rail Transit No Skipping Phase Skipping No Skipping Phase Skipping No Skipping Phase Skipping Standard Deviation Southbound 5 0 M e t r e s 11.38 10.05 11.00 4.61 7.54 3 .32 100 M e t r e s 5 .27 6.84 6 .14 2 .06 3 .05 7.03 150 M e t r e s 15 .40 10 .16 8.18 8.85 25 .02 •10.61 Northbound 5 0 M e t r e s 8.11 10 .40 7.22 2 .19 7 .87 8.14 100 M e t r e s 10.81 17.29 5 .95 5.75 9 .49 3 .19 150 M e t r e s 15.53 18.05 8.99 11.63 7 .16 12 .10 Standard Deviation as % of Mean Southbound 5 0 M e t r e s 2 . 1 4 % 2 . 1 0 % 2 . 1 1 % 0 . 9 7 % 1.99% 1.00% 100 M e t r e s 1.02% 1.40% 1.28% 0 .45% 0 . 8 1 % 2 . 0 9 % 150 M e t r e s 2 . 9 5 % 1.95% 1.56% 1.75% 6 . 6 7 % 3 . 0 6 % Northbound 5 0 M e t r e s 1.21% 1.61% 1.07% 0 .35% 1.56% 1.83% 100 M e t r e s 1.59% 2 . 6 1 % 0 .88% 0 . 9 0 % 1.94% 0 . 7 0 % 150 M e t r e s 2 . 1 5 % 2 . 5 5 % 1.29% 1.73% 1.40% 2 . 5 7 % 4.4.2.5 Summary According to the simulation results, travel time variability is not greatly affected by detector location or the length of the green extension. While phase skipping has significant impacts on travel time, travel time variability is not noticeably different between no skipping and skipping scenarios. Finally, neither LRT nor express bus service consistently results in more reliable travel times than the other. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 100 Transit and Light Rail Transit in the City of Richmond 4.4.3 Cross Street Delay Analysis 4.4.3.1 Total Delay For the most part, the implementation of the proposed LRT TSP strategies does not greatly affect existing traffic operations on cross street approaches. Where there are increases in delay, they are, for the most part, only a matter of a few seconds. Without phase skipping, the 50-metre detector configuration results in, at most, seven cross street approaches experiencing increases in delay, with the greatest increase being only 6.7 seconds. With the 100-metre detectors, up to 8 of the 23 cross street approaches experience increases in delay. While the greatest increase is 9.6 seconds, the majority of the increases are under three seconds. With the 150-metre detectors, at most 7 cross street approaches experience an increase in delay relative to existing. The greatest delay increase is only 4.8 seconds, with the majority of increases being less than one second. Moreover, there is no apparent relationship between the length of the green extension applied in any TSP strategy and the amount of delay experienced by cross street approaches in the study corridor. When phase skipping is implemented, there is an increase in the number of cross street approaches that experience increases in delay relative to the existing situation. The scale of delay increases also tends to be greater with phase skipping. With the 50-metre detectors, there are scenarios in which up to 12 of the 23 cross street approaches experience increases in delay, with increases of up to 5.7 seconds. The 100-metre detectors also cause up to 12 approaches to experience increases in delay. Most of the increases are under 8 seconds, except for the Alderbridge Way eastbound approach with a delay increase of nearly 20 seconds. Finally, with 150-metre detectors, up to 12 approaches experience delay increases, in some cases over 20 seconds. However, the majority of approaches experience delay increases of less than three seconds. Table 4-43 outlines the range of cross street delay increases in each scenario with and without phase skipping. It is quite clear that the use of phase skipping increases the range and scale of delay increases. Table 4-43 Range of Cross Street Delay Increase Values TSP Scenario No Sk ipp ing Phase Skipping 50-Metre Detectors 0.74-19.5% 0.26-43.95% 100-Metre Detectors 0.69-28.13% 0.09-31.5% 150-Metre Detectors 0.26 - 9.57% 0.13-25.8% The impacts of TSP implementation on cross street delay are most readily examined by looking at individual intersection approaches. Tables 4-44 and 4-45 provide information University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 101 Transit and Light Rail Transit in the City of Richmond on the westbound approach to No. 3 Road on Westminster Highway. It is clear that, while the 50-metre detector configuration has the lowest total delay results, detector location is not a critical factor in cross street delay. The range of delay values among the detector configurations is only 8.4 seconds. However, it should be noted that all of the detector locations result in delay values that are significantly lower than under the existing TSP strategy. It is also apparent that the implementation of phase skipping increases delay experienced on cross street approaches relative to the same TSP scenarios without phase skipping. As mentioned earlier, this is due to cross streets having their phases skipped in order to serve oncoming transit vehicles. Table 4-44 Cross Street Delay Comparison Detector Location Delay Change from Existing 50 Metres 51.2 seconds -21.7% 100 Metres 55.2 seconds - 15.6% 150 Metres 46.8 seconds - 28.5% Table 4-45 Cross Street Delay Compar ison—Phase Skipping (15-Second Green Extension) Detector Location No Phase Skipping Phase Skipping Delay Change from Existing Delay Change from Existing 50 Metres 51.2 sec -21.7% 56.0 sec - 14.3% 100 Metres 55.2 sec - 15.6% 63.7 sec - 2.7% 150 Metres 46.8 sec - 28.5% 62.2 sec - 4.9% Tables 4-46 and 4-47 outline cross delay results for the westbound approach to No. 3 Road on Leslie Street. The 50-metre detector configuration appears to result in the least delay to cross street traffic of the studied locations. However, the range in delay values among the different configurations is only 10.2 seconds, which indicates that detector location is not a critical factor. Phase skipping increases not only the number of cross street approaches that experience delay increases, but also the scale of those increases. While this is not as apparent at Leslie Street as at Westminster Highway, there is still a noticeable trend that phase skipping increases delay to cross street traffic compared to no phase skipping TSP strategies. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 102 Table 4-46 Cross Street Delay Compar ison Leslie S t ree t—Westbound Approach (15-second Green Extension) Detector Location Delay Change from Existing 50 Metres 33.5 seconds -1.8% 100 Metres 43.7 seconds + 28.1% 150 Metres 34.7 seconds + 1.6% Table 4-47 Cross Street Delay Compar ison—Phase Skipping (15-Second Green Extension) Leslie S t ree t—Westbound Approach Detector Location No Phase Skipping Phase Skipping Delay Change from Existing Delay Change from Existing 50 Metres 33.5 sec - 1.8% 37.2 sec + 9.0% 100 Metres 43.7 sec + 28.1% 34.5 sec + 1.1% 150 Metres 34.7 sec + 1.6% 38.6 sec + 13.2% There does not appear to be any significant relationship between the length of the green extension applied in the TSP strategy and the amount of delay experienced by cross street traffic. There is no trend in cross street delay across the range of studied green extension lengths. This is illustrated in Tables 4-48 and 4-49, which outline cross street delay results for the Westminster Highway westbound approach and Leslie Street westbound approach. Table 4-48 Cross Street Delay Compar ison—Green Extension Length (Existing Transit Stops) Detector Location G E = 10 GE = 15 GE = 20 G E = 30 Delay Change from Existing Delay Change from Existing Delay Change from Existing Delay Change from Existing 50 m 54.1 sec - 17.3% 56.0 sec - 14.3% 64.4 sec - 1.6% 58.7 sec - 10.3% 100 m 59.1 sec -9.6% 63.7 sec - 2.7% 56.4 sec - 13.8% 62.7 sec - 4.2% 150 m 56.7 sec - 13.3% 62.2 sec - 4.9% 64.6 sec - 1.2% 63.7 sec - 2.6% Cross Street Delay Comparison-Table 4-49 -Green Extension Length (Existing Transit Stops) Detector Location G E = 10 GE = 15 GE = 20 G E = 30 Delay Change from Existing Delay Change from Existing Delay Change from Existing Delay Change from Existing 50 m 39.6 sec + 15.9% 37.2 sec + 9.0% 31.5 sec - 7.6% 43.2 sec + 26.5% 100 m 40.7 sec + 19.3% 34.5 sec + 1.1% 39.7 sec + 16.3% 42.1 sec + 23.2% 150m 39.8 sec + 16.6% 38.6 sec + 13.2% 39.5 sec + 15.6% 42.9 sec + 25.8% University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 103 Transit and Light Rail Transit in the City of Richmond 4.4.3.2 Unconditional TSP v. Conditional TSP It is quite clear from the simulation results summarized in Table 4-50 that unconditional transit signal priority causes more cross street approaches to experience increases in delay relative to existing than the same scenarios with conditional TSP. Unconditional LRT priority also results in more cross street approaches experiencing delay increases than conditional bus TSP, but the number of locations is similar to under unconditional bus TSP. Table 4-50 Number of Cross Street Approaches Experiencing Increases in Delay with TSP Scenario Conditional Bus TSP Unconditional Bus TSP Unconditional L R T TSP No Skipping Phase Skipping No Skipping Phase Skipping No Skipping Phase Skipping 50-Metre Detectors, Far Side Transit Stops, GE = 10 4 9 7 7 4 12 50-Metre Detectors, Far Side Transit Stops, GE= 15 8 8 9 8 4 10 50-Metre Detectors, Far Side Transit Stops, GE = 20 5 11 3 7 7 7 50-Metre Detectors, Far Side Transit Stops, GE = 30 5 8 7 5 6 8 100-Metre Detectors, Far Side Transit Stops, GE= 10 4 9 6 13 8 6 100-Metre Detectors, Far Side Transit Stops, GE = 15 4 7 7 8 5 6 100-Metre Detectors, Far Side Transit Stops, GE = 20 6 9 7 12 6 10 100-Metre Detectors, Far Side Transit Stops, GE = 30 5 7 5 6 4 -150-Metre Detectors, Far Side Transit Stops, GE= 10 4 9 5 8 5 12 150-Metre Detectors, Far Side Transit Stops, GE = 15 5 11 5 10 6 8 150-Metre Detectors, Far Side Transit Stops, GE = 20 5 12 8 12 7 11 150-Metre Detectors, Far Side Transit Stops, GE = 30 3 8 7 9 7 12 The simulation results do not indicate that either unconditional or conditional bus priority or unconditional LRT priority consistently provides smaller increases and larger decreases in cross street delay than the others relative to the existing situation. While it is clear that unconditional bus and LRT TSP cause more cross street approaches to experience increases in delay than under conditional priority, the scale of these increases cannot be definitively said to be greater. However, what is clear is that the use of phase skipping increases the number of cross street approaches experiencing an increase in delay, reduces delay reductions and increases delay increases under both unconditional and conditional TSP strategies. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 104 Transit and Light Rail Transit in the City of Richmond 4.4.3.3 Level of Service In addition to total delay experienced on cross street approaches, it is also important to analyze level of service, which is measured by average stop delay per vehicle. There are locations where level of service is reduced, but it is only because these locations have existing stop delay values that are at the upper limit of the current LOS designation and even a small delay increase reduces the level of service. Some locations have such high existing stop delay that even significant reductions in delay do not improve level of service from LOS F. Finally, the implementation of certain TSP strategies results in stop delay reductions at some cross street locations. Table 4-51 compares these observations at some of the cross street approaches in the corridor. T a b l e 4-51 C r o s s S t ree t S t o p D e l a y a n d L e v e l o f S e r v i c e P h a s e S k i p p i n g Approach Existing Stop Delay Existing Level of Service GE = 10 GE = 15 GE = 20 GE = 30 Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS Change in Stop Delay (sec) LOS SO-m Detectors, Phase Skipping Westminster WB 51.90 E -9.30 E -8.00 E - 1.00 E -5.50 E Westminster EB . 96.90 F - 12.00 F - 9.40 F - 12.60 F - 12.00 F Alderbridge WB 50.8 E + 24.50 F + 19.60 F + 10.80 F + 9.30 F Alderbridge EB 58.0 E + 6.30 F + 2.90 F + 5.40 F + 0.60 F 100-m Detectors, Phase Skipping Alderbridge WB 50.80 E + 0.80 E -4.20 E + 4.20 F + 17.30 F Bridgeport WB 79.10 F -2.00 F - 1.10 F -5.10 F - 11.50 F I SO-m Detectors, Phase Skipping Lansdowne WB 27.60 D -0.60 D - 1.00 D -2.80 C -2.70 C Alderbridge WB 50.80 E + 0.30 E + 4.90 E + 18.20 F + 1.60 E Westminster EB 96.87 F -7.60 F -7.00 F - 22.00 F -9.80 F The eastbound approach on Westminster Highway is an example of significant reductions in stop delay not improving the level of service because the existing delay is high with LOS F. The westbound approach on Alderbridge Way indicates how phase skipping can cause significant increases in delay relative to the same scenarios without skipping. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 105 Transit and Light Rail Transit in the City of Richmond 4.4.4 Green Extension Effectiveness 4.4.4.1 Transit Station and Detector Location A critical element of green extension effectiveness at individual intersections appears to be the presence of an LRT station. Many locations have green extension efficiency ratios of 0% or close to 0% under all green extensions, with and without phase skipping, when a transit station is located nearby. This is the case at Park Road and Cook Road, which are south and north of Richmond Centre Station respectively. This is also seen at Westminster Highway, Ackroyd Road, Cambie Road and Capstan Way, which are adjacent to LRT stations. This can be explained as the LRT station dwell times causing the green extension to expire before an L R V can clear the intersection. With and without phase skipping, the 50-metre check-in detectors provide the greatest overall green extension effectiveness in the corridor. This is illustrated in Figure 4M. This was expected since the closer the detector to the intersection, the greater the probability should be that an L R V will clear an intersection during a green extension. Overall corridor green extension ratios tend to be low because some intersections have very high green extension efficiency ratios while others are as low as 0%. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 106 Transit and Light Rail Transit in the City of Richmond Green Extension Effectiveness Average Green Extension Efficiency Ratio in the Travel Corridor Figure 4M: Green Extension Effectiveness Comparison—Detector Location 4.4.4.2 Green Extension Length and Phase Skipping As was noted in the express bus scenarios, there is a strong correlation between the length of the green extension and the green extension efficiency ratio. This relationship illustrated in Figure 4M in the previous section. As the green extension length increases, so does the green extension effectiveness. This is logical since a longer extension should provide a higher probability of an L R V being able to clear an intersection during a green extension. This is seen in all the studied TSP scenarios, with and without phase skipping. However, phase skipping is not a critical factor in green extension effectiveness. With all detector locations, the no skipping and skipping scenarios provide similar green extension efficiency ratios. This is seen in Figure 4M as the trend lines of green extension effectiveness for no skipping and skipping are relatively close together across the green extension range. This result was expected since phase skipping affects time to cycle to green for the bus phase rather than extending a green phase in which a TSP call is received. Table 4-52 compares the overall green extension effectiveness for the corridor with and without phase skipping and with each detector configuration. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 107 Transit and Light Rail Transit in the City of Richmond Table 4-52 Overall Green Extension Efficiency Comparison—Phase Skipping Analysis 15-Second Green Extension Detector Location No Phase Skipping Phase Skipping 50 metres 26.0% 20.8% 100 metres 12.6% 10.7% 150 metres 14.5% 13.3% 4.4.4.3 Summary Green extension effectiveness can be maximized at individual intersections when stations are located at far side locations, when detectors are located relatively close to intersections and long green extensions are employed. However, the use of phase skipping has little impact on green extension effectiveness. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 108 Transit and Light Rail Transit in the City of Richmond 4.5. SUMMARY AND DISCUSSION OF RESULTS FOR UNCONDITIONAL LRT PRIORITY 4.5.1 Transit Vehicle Travel Time and Delay As noted in the unconditional and conditional express bus TSP scenarios, there is a very strong link between the delay experienced by transit vehicles in traveling in the corridor and the average corridor travel time. The simulation results indicate that changes in travel time correspond quite closely to similar changes in transit delay. Unlike in the express bus TSP scenarios, transit vehicle delay and travel times under unconditional LRT TSP are not greatly affected by the location of check-in detectors. The range in delay values and average travel times among the different detector configurations is small relative to the range with express bus TSP. This could be due to the fact that light rail vehicles (LRVs) travel in an exclusive right-of-way for the entire corridor while express buses share the corridor with automobile traffic south of Ackroyd Road. The traffic in the shared segments of the corridor may cause delay to transit vehicles and make the detector location an important factor in efficiency of TSP provision and, consequently, travel time. However, LRVs are free to travel at the desired speed for the entire corridor, which may reduce the discrepancy between the different detector locations in terms of TSP provision efficiency. This may also be the result of the LRT line incorporating mainly far side and mid-block transit stop locations. This reduces problems caused when a transit vehicle must stop at the near side stop after passing over the detector, which can cause double stopping. Without near side stops, the detector locations do not greatly affect whether a vehicle will be able to clear an intersection. Transit station location was not modified in this scenario as in the bus scenarios since the physical size of the platforms limits the number of possible locations. In general, the pre-planned locations are the most logical and could not be easily relocated without moving a significant distance up or downstream. However, based on the express bus results, it is reasonable to assume that placing LRT stations immediately downstream of an intersection rather than on the near side would reduce delay and improve average travel times. As was observed in the express bus TSP scenarios, with both unconditional and conditional priority, the length of the green extension does not appear to be significantly related to average travel time. One explanation is that, since not all transit vehicles requesting TSP will arrive during a green phase, not all TSP actions will involve provision of green extension. In these cases, travel time will not be affected by the length of the green extension. Therefore, i f many of the individual travel times used to determine average travel times relied only minimally on green extensions, the University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 109 Transit and Light Rail Transit in the City of Richmond relationship between travel time and length of green extension could be lost. Another explanation is that the LRT travel properties allow the average L R V to efficiently clear the intersection during the shortest studied green extension so that increasing the length of the green extension does little to improve travel time. The patterns noticed in travel time across the green extension range would likely be consequences of the other variables influencing travel time. Phase skipping clearly results in significantly lower travel times than the same TSP scenarios without skipping. As noted in express bus TSP, this is due to a reduction in the time to cycle to green for an L R V after sending a TSP request, which reduces delay and, consequently, travel time. The travel time benefit of LRT relative to express bus is the result of the following factors. LRVs have shorter dwell times than buses since less time is spent allowing passengers to board and alight. The LRT line has an exclusive right-of-way for the entire corridor, unlike bus service. This allows LRVs to travel at the desired travel speed for the entire corridor rather than at a speed dictated by automobile traffic. Finally, there are fewer stations than on the bus line, which reduces total dwell time for the corridor. A l l of these factors result in less delay to LRVs than is experienced by express buses. 4.5.2 Travel Time Variability Travel time variability does not appear to be significantly affected by detector location. This is illustrated by the fact that the detector configuration providing the lowest S D % M values differs from scenario to scenario. This indicates that there are other factors influencing travel time variability more greatly than detector location. Furthermore, since the length of the green extension has little influence on average travel times, it stands to reason that the length of the green extension should have no influence on travel time variability. This is observed in the simulation results as S D % M values fluctuate up and down across the green extension range. While phase skipping appears to be the most significant factor in the average L R V travel time, it is relatively inconsequential in terms of travel time variability. Phase skipping scenarios do not consistently provide higher or lower travel time variability than no phase skipping. Furthermore, travel times are not more reliable with LRT or bus service relative to each other. In some scenarios, S D % M values are lower with LRT and in other scenarios, bus S D % M values are lower. Since none of the parameters studied in this research appear to have any significant influence on travel time variability, it is reasonable to assume that this parameter is influenced by other factors. It may be that variability in travel time is largely based on the nature of traffic operations during a specific time period. For example, travel time University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 110 variability may be greatly influenced by the time in the cycle when a TSP request is received at intersections. This differs at every intersection and under each TSP scenario. Thus, travel time variability may be largely dictated by the variability in the simulation model based on the random distributions applied for certain parameters and the random seed when initiating a simulation run. 4.5.3 Cross Street Traffic Operations The delay experienced by vehicles traveling on streets crossing the No. 3 Road corridor is not greatly influenced by the location of the check-in detectors or the length of the green extension. Phase skipping, on the other hand, is a critical parameter. The implementation of phase skipping causes more cross street approaches to experience delay increases relative to the existing situation than without phase skipping. The scale of these increases is also greater. This is quite obviously the result of traffic having its phase skipped to serve transit vehicles and then having to wait for the next cycle to receive green time. However, none of the studied TSP strategies adversely impacts existing cross street traffic operations. With and without phase skipping, the majority of cross street approaches experience equal or less delay compared to the existing scheme. Those locations that experience greater delay only amount to a few seconds more delay. Thus, the implementation of the proposed TSP strategies would not adversely affect existing traffic operations on streets crossing the main corridor. Furthermore, the level of service on cross street approaches is not greatly affected by the proposed TSP strategies. In most cases, the change in delay is insufficient to change the LOS designation. Where LOS changes, the majority of cases are LOS improvements. Where LOS does decrease, it is usually the result of a small delay increase causing a location currently near the upper limit of an LOS designation to move into the next designation. The impact of LRT TSP implementation on cross street approaches is noticeably less than unconditional bus TSP, but less than conditional bus TSP. This is simply due to the lower frequency of traffic signal cycle disruption under the conditional TSP strategy. 4.5.4 Green Extension Effectiveness The simulation results for green extension effectiveness are basically as were expected prior to running the simulations. When there is a transit station in close proximity to an intersection, the green extension efficiency tends to be lower than intersections far from a station. This may be due to dwell times and deceleration when approaching a station causing green extensions to expire before an L R V can clear the intersection. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 111 The closer the check-in detector is located to an intersection, the higher the green efficiency ratio tends to be. This is due to the probability of an L R V clearing an intersection during a green extension being increased the closer an L R V is to the intersection when the TSP request is received. The longer the green extension, the greater the green extension efficiency ratio. This for the same reason as closer detector locations providing greater green extension efficiency: higher probability of clearing the intersection during the green extension. However, phase skipping does not affect green extension effectiveness. This was expected since phase skipping improves the time to cycle from red to green for the bus phase and has little to do with how efficiently LRVs use green extensions. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 112 Transit and Light Rail Transit in the City of Richmond 5. Conclusions and Recommendations 5.1 Conclusions The following conclusions apply to unconditional and conditional express bus signal priority and unconditional light rail transit signal priority. 5.1.1 Transit Vehicle Travel Time and Delay 1. Transit vehicle delay and travel time are inextricably linked, meaning that reductions in travel time are largely the result of reductions in delay experienced by transit vehicles in traveling through the corridor. In all the studied scenarios, increases and decreases in transit travel time were accompanied by corresponding increases and decreases in stop delay and total delay. 2. The 50 and 100-metre detectors provide comparably low levels of delay and, consequently, travel times in the unconditional bus TSP scenarios while the 100-metre detectors clearly provide the lowest travel times with conditional bus TSP. In unconditional LRT TSP, the differences in delay and travel times among the 50, 100 and 150-metre detector locations are small, indicating that, detector location is not as critical a parameter in travel time as in express bus service. 3. The far side transit stop configuration results in lower levels of delay and travel times than the existing stop configuration in all the studied TSP scenarios. 4. The simulation results indicate that the length of the green extension is not correlated to transit delay or average corridor travel time. However, this is most likely the result of averaging many individual travel times. Variability in the simulation model affects the phase in which transit vehicles request signal priority and only in green phases will a green extension be applied. Travel times that are based on TSP actions that mainly involved cycling from red to green to serve the bus phase will not be influenced by green extension length. When individual travel time results that did not incorporate green extensions are combined with results that did incorporate green extensions, any correlation between green extension and travel time may be diluted. 5. The implementation of phase skipping results in less transit delay and lower transit travel times than the same scenarios without phase skipping. The use of phase skipping reduces the average time required for the signal controller to cycle from red to green for the bus phase upon receipt of a TSP call. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 113 Transit and Light Rail Transit in the City of Richmond 6. LRT travel times are significantly lower than express bus service with either unconditional or conditional TSP. For express bus service, unconditional priority results in lower average travel times than conditional priority. 5.1.2 Travel Time Variability 7. The 50-metre detector configuration appears to provide the lowest travel time variability for unconditional and conditional TSP. However, the differences in variability results for LRT TSP among the studied detector locations are relatively small. 8. While transit stop location is not a consideration in an LRT system, the far side transit stop configuration results in less travel time variability than the existing stop configuration for express bus service. 9. The length of the green extension appears to have little or no correlation with travel time variability. However, this may be due to simulation model variability. Any correlation that does exist may be diluted by averaging many different transit vehicle travel time results. 10. The implementation of phase skipping does not significantly reduce variability in transit vehicle travel times. 11. While conditional bus TSP results in greater variability than unconditional bus TSP, the S D % M values are comparable between unconditional bus and LRT TSP scenarios. 5.1.3 Cross Street Traffic Operations 12. A l l of the proposed TSP strategies maintain or improve the existing traffic operations on the approaches crossing the No. 3 Road corridor. Some approaches experience less delay per vehicle with the proposed TSP scenarios than the existing strategy. Approaches that do experience delay increases relative to existing only experience increases of a few seconds. 13. Transit stop and detector locations are not critical parameters affecting the delay experienced by cross street traffic. 14. Phase skipping clearly increases adverse impacts on traffic as more cross street approaches experience delay increases and the scale of the increases is greater than without skipping. 15. The length of the green extension does not appear to be significantly related to the levels of delay experienced by cross street traffic. 16. In general, the level of service at cross street approaches does not change relative to existing with the implementation of the proposed TSP strategies. Where level of service changes, most often there is an LOS improvement due to significant stop delay reduction. Where LOS is reduced, it is usually the University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 114 Transit and Light Rail Transit in the City of Richmond result of a small stop delay increase at a location where existing stop delay is at the upper limit of the current LOS designation. 5.1.4 Green Extension Effectiveness 17. The closer detectors are located to an intersection, the greater the green extension efficiency tends to be due to increased probability of clearing an intersection during a green extension. 18. The far side transit stop configuration results in better green extension effectiveness than the existing combination of near side, mid-block and far side transit stops. 19. Green extension length and green extension effectiveness are positively related, meaning that as the length of the green extension is increased, so too is the green extension efficiency. 20. Phase skipping does not affect green extension effectiveness since it reduces average time to cycle to a green phase from red, rather than affecting the extension of a green phase. 21. There is no considerable difference between an express bus system and an LRT system in terms of green extension effectiveness. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 115 Transit and Light Rail Transit in the City of Richmond 5.2 Recommendations 1. A detector configuration with detectors located between 50 and 100 metres upstream of intersections should be employed to provide effective transit signal priority. This configuration will minimize delay to transit and, consequently, average transit vehicle travel times, reduce variability in travel times and provide efficient green extensions. 2. Far side transit stops should be used at all locations to limit delay to transit, minimize travel times and allow for efficient use of green extensions by transit vehicles. 3. A green extension should be selected based on the detector-to-intersection distance and the travel speed of a transit vehicle. A buffer can be added to ensure that transit vehicles will clear during the green extension. This requires the use of far side stops to prevent transit vehicles from having to stop before clearing an intersection. 4. Phase skipping should be employed when the objective is to minimize delay experienced by transit vehicles. Skipping will provide significant delay reductions with minimal adverse impacts on cross street traffic operations. 5. An LRT line will offer significant transit operational benefits i f implemented. A cost-benefit analysis should be undertaken to determine if the considerable delay and travel time benefits of LRT relative to express bus service justify the capital expenditures required to implement the system. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 116 References 1. Adamski, Andrzej and Andrzej Turnau. "Simulation Support for Real-Time Dispatching Control in Public Transport." Transportation Research Part A . Vol . 32, No. 2. pp. 73 - 87. 1998. Elsevier Science Ltd. 2. Al-Sahili, Khaled A. and William C. Taylor. "Evaluation of Bus Priority Signal Strategies in Ann Arbor, Michigan." Transportation Research Record 1554. pp. 74 - 80. Transportation Research Board. Washington, D . C : 1996. 3. Andrews, Christina M . , S. Manzur Elaki and James E. Clark. "Evaluation of New Jersey Route 18 OPAC/MIST Traffic Control System." Transportation Research Record 1603. Transportation Research Board. National Academy Press. Washington, D . C : 1997. pp. 150-155. 4. Balke, Kevin N . "Development and Laboratory Testing of an Intelligent Bus Priority Approach." Presented to the TRB Traffic Signal Systems Committee 1999 Summer Meeting. 5. Bates, Michael and Leo Lee. "At-Grade or Not At-Grade: The Early Traffic Question in Light Rail Transit Route Planning." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp. 351 - 367. 6. Bauer et al. "Testing of Light Rail Signal Control Strategies by Combining Transit and Traffic Simulation Models." Transportation Research Record 1494. Transportation Research Board. National Academy Press. Washington, D . C : 1995. pp. 155- 160. 7. Bertini, Robert L., Jan L. Botha and K. Odila Nielsen. "Light Rail Transit Implementation Perspectives for the Future: Lessons Learned in Silicon Valley." Conference Proceedings: Seventh National Conference on Light Rail Transit. 1995. pp. 23 - 32. 8. Boje, B.F. and M . Nookala. "Signal Priority for Buses: An Operational Test at Louisiana Avenue, Minneapolis." Compendium of Technical Papers for the 66 t h ITE Annual Meeting, Washington, D . C , 1996, pp 309-313. 9. Bottoms, Glen D. "Light Rail Transit Developments in Western Europe." Conference Proceedings: Seventh National Conference on Light Rail Transit. Volume 2. 1995. pp. 3 - 14. 10. Carey, Malachy. "Optimized Schedule Times, Allowing for Behavioural Response." Transportation Research Part B. Vol . 32, No. 5. pp. 329 - 342. 1998. Elsevier Science Ltd. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 117 11. Casey, Robert F. "What Have We Learned About Intelligent Transportation Systems." John A. Volpe National Transportation Systems Center. 1999. 12. Casey, Robert F., Lawrence N . Labell, Elisabeth J. Carpenter, Joseph A . LoVecchio, Leisa Moniz, Robert S. Ow, Jackson W. Royal and Judith C. Schwenk. "Advanced Public Transportation Systems: The State of the Art. Update '98." Federal Transit Administration. January 1998. www.fta.dot.gov/researcri/pdf/aptssoa98.pdf 13. Ceder, Avishai. "Bus Timetables with Even Passenger Loads as Opposed to Even Headways." Transportation Research Record 1760. Transportation Research Board. National Academy Press. Washington, D . C : 2001. pp. 3 - 9 . 14. Chada, Shireen and Robert Newland. "Effectiveness of Bus Signal Priority." Prepared for the 2001 ITE Annual Meeting and Exhibit, August 2001. 15. Chang, Gang-Leu, Meenakshy Vasudevan and Chih-Chang Su. "Bus Pre-emption Under Adaptive Signal Control Environments." Transportation Research Record 1494. Transportation Research Board. National Academy Press. Washington, D . C : 1995. pp. 146- 154. 16. Chang, Gang-Len, Meenakshy Vasudevan and Chih-Chiang Su. "Modelling and Evaluation of Adaptive Bus Pre-Emption Control With and Without Automatic Vehicle Location Systems." Transportation Research Part A: Policy and Practice. Volume 30, Number 4. 1996. pp. 251-268. 17. Claflin, John D. "Light Rail - A Mile High: The Denver Experience." Conference Proceedings: Seventh National Conference on Light Rail Transit. 1995. pp. 33-41. 18. Conrad, Mark, Francois Dion and Sam Yagar. "Real-Time Traffic Signal Optimization with Transit Priority: Recent Advances in the SPPORT Model." Prepared for the 77 t h Annual Meeting of the Transportation Research Board, January 1998. 19. Dale, James J., Thomas Bauer, Ronald J. Atherley and Len Madsen. " A Transit Signal Priority Impact Assessment Methodology—Greater Reliance on Simulation." Prepared for the 78 t h Annual Meeting of the Transportation Research Board, January 1999. 20. Dale, Jim, Thomas Bauer and Dan Odermort. "Light Rail Transit Signal Priority in a Downtown Environment." Conference Title: Transportation Frontiers for the Next Millennium: 69 th Annual Meeting of the ITE. Institute of Transportation Engineers. Washington: 1999. 21. Davol, Angus P. "Modeling of Traffic Signal Control and Transit Priority Strategies in a Microscopic Simulation Laboratory." Master's thesis. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 118 Transit and Light Rail Transit in the City of Richmond Massachusetts Institute of Technology. 2001. (http://web.mit.edu/its/papers/TRAFF.pdf) 22. Ding, Yuqing and Steven I. Chien. "Improving Transit Service Quality and Headway Regularily with Real-Time Control." Transportation Research Record 1760. Transportation Research Board. National Academy Press. Washington, D . C : 2001. pp. 161 - 170. 23. Dion, Francois and Bruce Hellinga. " A Rule-Based Real-Time Traffic Responsive Signal Control System with Transit Priority: Application to an Isolated Intersection." Transportation Research Part B. Vol 36. 2002. pp. 325 - 343. 24. Duncan, W. and Mirabdal, J. "Transit Preferential Streets Program in San Francisco." Compendium of Technical Papers for the 66 t h ITE Annual Meeting, Washington, D . C , 1996, pp 314-318. 25. Dunphy, Robert T. "Review of Recent American Light Rail Experience." Conference Proceedings: Seventh National Conference on Light Rail Transit. 1995. pp. 104 -113. 26. Eberlein, X u Jun, Nigel H . M . Wilson, Cynthia Barnhart and David Bernstein. "The Real-Time Deadheading Problem in Transit Operations Control." Transportation Research Part B. Vol . 32, No. 2. pp. 77 - 100. 1998. Elsevier Science Ltd. 27. Federal Transit Administration. "Automatic Vehicle Location Successful Transit Applications—A Cross-Cutting Study." August 2000. Federal Transit Administration. Washington, D.C. 28. Forges, J.L. and J.J. Henry. "CELTIC Bus Priority in Lyon." International Conference on Road Traffic Monitoring and Control." Institute of Electrical Engineers. London: 1994. 29. Fox, Gerald D. "Designing At-Grade Light Rail Transit." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp. 606 -620. 30. Fox, Ken, Haibo Chen, Frank Montgomery, Mike Smith and Simon Jones. "Selected Vehicle Priority in the U T M C Environment." 1998. Leeds: Institute for Transport Studies, University of Leeds. http://www.its.leeds.ac.uk/projects/spruce/utmclrev.html 31. Fox, Ken, Stephen Clark, Rachel Boddy, Frank Montgomery and Margaret Bell. "Some Benefits of a SCOOT UTC System." Traffic Engineering + Control. September 1998. pp. 484-489. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 119 Transit and Light Rail Transit in the City of Richmond 32. Furth, Peter G. and Theo H.J. Muller. "Conditional Bus Priority at Signalized Intersections: Better Service with Less Traffic Disruption." Transportation Research Record 1731. Transportation Research Board. National Academy Press. Washington, D . C : 2001. pp. 23-30. 33. Garrow, Michael. "Development and Evaluation of Transit Signal Priority Strategies." Center for Transportation Research, University of Texas. Prepared for the 77 t h Annual Meeting of the Transportation Research Board, January 1998. 34. Gartner, Nathan H., Philip J. Tarnoff and Christina M . Andrews. "Evaluation of Optimized Policies for Adaptive Signal Control Strategy." Transportation Research Record 1324. Transportation Research Board. National Academy Press. Washington, D . C : 1991. pp. 105-114. 35. Gerhart, Richard L. "Lessons Learned from New LRT Start-Ups." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp. 317-330. 36. Gibson, David. "A New British Tram Concept." Tramways and Urban Transit. February 2001. pp. 48 - 49. 37. Gillen, David, Elva Chang and Doug Johnson. "Productivity Benefits and Cost Efficiencies from Intelligent Transportation System Applications to Public Transit." Transportation Research Record 1747. Transportation Research Board. National Academy Press. Washington, D . C : 2001. pp. 89 - 96. 38. Greenough, John C. and W. Les Kelman. "Metro Toronto SCOOT: Traffic Adaptive Control Operation." ITE Journal. May 1998. Vol. 68, No. 5. ITE. Washington, D . C : 1998. 39. Hansen, Blake C , Peter T. Martin and H. Joseph Perrin Jr. "SCOOT Real-Time Adaptive Control in a CORSIM Simulation Environment." Transportation Research Record 1727. National Academy Press. Washington, D . C : 2000. pp. 27-31. 40. Head, Larry. "TCRP A-16A: Improved Traffic Signal Priority for Transit." . Presentation to the Transportation Research Board Transit Signal Priority Workshop, January 13, 2002. (http ://si gnal systems .tamu.edu/documents/Si gnal Control Workshop2002/TCRP_A-16A Overview Head.pdf) 41. Hu, Kang, Sean Skehan and Rex Gephart. "Implementing a Smart Transit Priority System for Metro Rapid Bus in Los Angeles." Prepared for the Transportation Research Board's 80 th Annual Meeting, November 2000. http://www.path.berkeley.edu/%7Eleap/itsdecision resources/reports/TRB00505.pdf University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 120 42. Hubbell, John, Dave Colquhoun, Dan Bolger and John Morrall. "Light Rail Transit in Calgary, 1981 - 1995: A Retrospective Review." Conference Proceedings: Seventh National Conference on Light Rail Transit. Volume 2. 1995. pp. 15 - 26. 43. Hunter-Zaworksi, Katharine M . , William C. Kloos and Alan R. Danaher. "Bus Priority at Traffic Signals in Portland: The Powell Boulevard Pilot Project." Transportation Research Record 1503. pp. 29 - 38. Transportation Research Board. Washington, D . C : 1995. 44. Imhoff, Mark. "The Denver Experience: Starting Small." Conference Proceedings: Seventh National Conference on Light Rail Transit. 1995. pp. 83 - 87. 45. Intelligent Transportation Society of America (Advanced Traffic Management Systems Committee and Advanced Public Transportation Systems Committee). "An Overview of Transit Signal Priority." April 15, 2002. rhttp://www.itsa.org/committe.nsf/4c2dbl6e002c25928525620600031772/493aa4e21 d77827585256b9d00463eec/$FILE/Signal Priority Paper Open Comments.pdf) 46. Jacobsen, Kern L . "Transit Signal Priority Treatments in the Puget Sound Region." Pacific Rim TransTech Conference Proceedings, Volume I. pp. 272 -278. July 25-28 , 1993. 47. Jagannathan, Ramesh and Ata M . Khan. "Methodology for the Assessment of Traffic Adaptive Control Systems." ITE Journal. June 2001. pp. 28-33. 48. Kelman, L. "Signal Control Priority for Transit Vehicles." Presentation to the Transportation Research Board Transit Signal Priority Workshop, January 13, 2002. (http://signalsvstems.tamu.edu/documents/SignalControlWorkshop2002/TRB_Januar v 2002 Kelman.pdf) 49. Khasnabis, Snehamay, Rama Rudraraju and Mirza F. Baig. "Economic Evaluation of Signal Pre-emption Projects." Journal of Transportation Engineering. Volume 125, Issue 2. pp. 160 - 167. 50. Khasnabis, Snehamay, Gangula V . Reddy and Syed Khurshidul Hoda. "Evaluation of the Operating Cost Consequences of Signal Pre-emption as an IVHS Strategy." Transportation Research Record 1390. National Academy Press. Washington, D . C : 1993. pp. 3 - 9 . 51. Khasnabis, Snehamay, Rajashekar R. Karnati and Rama K. Rudraraju. "NETSIM-Based Approach to Evaluation of Bus Pre-emption Strategies." Transportation Research Record 1554. pp. 80 - 89. Transportation Research Board. Washington, D . C : 1996. 52. Khasnabis, Snehamay and Rama K. Rudraraju. "Optimum Bus Headway for Pre-emption." Transportation Research Record 1603. Transportation Research Board. National Academy Press. Washington, D . C : 1997. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 121 53. King County Department of Transportation. "Route 358-Aurora Avenue N . Transit Signal Priority Study." March 2002. (www.transit.metrokc.gov). 54. King County Department of Transportation and City of Seattle Transportation. "Preliminary Transit Signal Priority Assessment of S. Genesee Street and Rainier Avenue South." Seattle. 1999. 55. King County Department of Transportation and City of Seattle Transportation. "Transit Signal Priority System Assessment Study: Rainier Avenue South Field Evaluation Draft Report." Seattle. 2000. 56. Kloos, B i l l . "Bus Priority in Portland—Lessons Learned." Presentation to the Transportation Research Board Transit Signal Priority Workshop, January 13, 2002. (http://signalsvstems.tamu.edu/documents/SignalControlWorkshop2002/TRB_TSP_S EMINAR DEC SUBMIT B 20011203 Kloos 2.pdf) 57. Korve, Hans W., Jose I. Farran and Douglas M . Mansel. "Integration of Light Rail Transit into City Streets." Conference Proceedings: Seventh National Conference on Light Rail Transit. 1995. pp. 294 - 303. 58. Korve, Hans W., Brent D. Ogden, Joaquin T. Siques, Douglas M . Mansel, Hoy A. Richards, Susan Gilbert, Ed Bani, Michelle Butchko, Jane C. Stutts and Ronald G. Hughes. "Light Rail Service: Pedestrian and Vehicular Safety." TCRP Report 69. Transportation Research Board. Washington: 2001. 59. Korve, Hans W., Jose I. Farran, Douglas M . Mansel, Herbert S. Levinson, Ted Chira-Chavala and David R. Ragland. TCRP Report 17—Integration of Light Rail Transit into City Streets. Transportation Research Board. Washington: 1995. 60. Korve, Hans W., Jose I. Farran, Douglas M . Mansel, Herbert S. Levinson, Ted Chira-Chavala and David R. Ragland. TCRP Report 17—Integration of Light Rail Transit into City Streets. Appendix F. Transportation Research Board. Washington: 1995. 61. Kuah, Geok K. and Jeffrey B. Allen. "Designing At-Grade LRT Progression: Proposed Baltimore Central Light Rail." Transportation Research Record 1361. Transportation Research Board. Washington: 1992. pp. 207 - 216. 62. Levinson, Herbert S. "Supervision Strategies for Improved Reliability of Bus Routes." National Cooperative Transit Research and Development Program Synthesis of Practice 15. Transportation Research Board. National Research Council. Washington, D . C : 1991. 63. Lewis, V . "Bus Priority Study: Tualatin Valley Highway." Tri-Met, Portland. 1996. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 122 64. Lin, Guey-Shii, Ping Liang, Paul Schoenfeld and Robert Larson. "Adaptive Control of Transit Operations." United States Federal Highway Administration. Washington, D . C : 1995. 65. Lin, G, P. Liang, P. Shonfield and R. Larson. Federal Transit Administration Office of Technical Assistance and Safety. "Adaptive Control of Transit Operations." FTA Report No. MD-26-7002. 1995. (www.fta.dot.gov/library/technology/APTS/ITS). 66. May, Adolf. "Traffic Flow Fundamentals." Prentice Hall. Upper Saddle River: 1990. 67. McLachlan, Larry. "Edmonton Light Rail Transit Experience." Conference Proceedings: Seventh National Conference on Light Rail Transit. 1995. pp. 52 - 65. 68. McNally, Michael G., Stephen P. Mattingly, James E. Moore, His-Hwa Hu, C. Arthur MacCarley and E. Jayakrishnan. "Evaluation of Anaheim Adaptive Control Field Operational Test." Transportation Research Record 1683. Transportation Research Board. National Academy Press: 1999. pp. 67-77. 69. Meadow, Linda J. and James P. Curry. "New Technologies for Improving Light Rail Grade Crossing Safety." Conference Proceedings: Seventh National Conference on Light Rail Transit. Volume 2. 1995. pp. 46 - 54. 70. Metropolitan Atlanta Rapid Transit Authority (MARTA). "Transit Signal Priority Shaves Time Off Bus Route." M A R T A Rider's Digest. Vol . 27, No. 20. September 25 - October 17, 1999. 71. Nelson, Eric J. and Darcy Bullock. "Impact Evaluation of Emergency Vehicle Preemption on Signalized Corridor Operation." Prepared for the 78 t h Annual Meeting of the Transportation Research Board, January 1999. 72. Nisar, Mohammad and Ata M . Khan. "Transitway: An Innovation in Public Transportation." ITE Journal. July 1992. pp. 35 - 39. 73. Oakes, Johnathan and Daniel Metzger. "Park View Pre-Signals in Uxbridge Road, Ealing." Traffic Engineering + Control. February 1995. Vol . 36. pp. 62 - 67. 74. Ospina, Diana, Fang Zhao and L. David Shen. "Overview of Light Rail Train Control Technologies." Conference Proceedings: Seventh National Conference on Light Rail Transit. 1995. pp. 103 - 113. 75. Owen, Larry E. and Charlie M . Stollard. "Rule-Based Approach to Real-Time Distributed Adaptive Signal Control." Transportation Research Record 1683. Transportation Research Board. National Academy Press: 1999. pp. 95-101. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 123 Transit and Light Rail Transit in the City of Richmond 76. Parkinson, Tom. "Advocacy for Conventional Light Rail." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp.66 - 78. 77. Rakka, H. and M . Van Aerde. "REAL TRAN: An Off-Line Emulator for Estimating the Effects of SCOOT." Transportation Research Record 1494. Transportation Research Board. National Academy Press. Washington, D . C : 1995. pp. 124-128. 78. Rietveld, P. F.R. Bruinsma and D.J. van Vuuren. "Coping with Unreliability in Public Transport Chains: A Case Study for Netherlands." Transportation Research Part A. Vol. 35. 2001. pp. 539-559. 79. Rymer, Bruce, Thomas Urbanik II and James C. Cline Jr. "Delay at Light Rail Transit Grade Crossings." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp. 621 - 634. 80. Schumann, John W. "RT Metro: From Sacramento's Community Dream to Operating Reality." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp. 387-407. 81. Schumann, John W. "What's New in North American Light Rail Transit Projects." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp. 8 - 42. 82. Sislak, Kenneth G. "Bus Rapid Transit as a Substitute for Light Rail Transit." Light Rail: Investment for the Future—8th Joint Conference on Light Rail Transit. Transportation Research Board. Washington: 2000. 83. Skabardonis, Alexander. "Control Strategies for Transit Priority." Institute of Transportation Studies, University of California. 1999. For the 79 t h Annual Meeting of the Transportation Research Board, January 2000. 84. Skehan, Sean. "Traffic Signal Priority for Metro Rapid Buses: The Los Angeles Experience." Prepared for the 2001 ITE Annual Meeting and Exhibit, August 2001. 85. Sunkari, Srinivasa R., Phillip S. Beasley, Thomas Urbanik II and Daniel B. Fambro. "Model to Evaluate the Impacts of Bus Priority on Signalized Intersections." Transportation Research Record 1494. Transportation Research Board. National Academy Press. Washington, D . C : 1995. pp. 117 - 123. 86. Taylor, Paul C , Leo K. Lee and Warren A. Tighe. "Operational Enhancements: Making the Most of Light Rail." Light Rail Transit: New System Successes at Affordable Prices. Papers Presented at the National Conference on Light Rail Transit. Transportation Research Board. Washington: 1989. pp. 578 - 592. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Transit and Light Rail Transit in the City of Richmond Page 124 87. Taylor, William C. and Ahmed S. Abdel-Rahim. "Final Report on Analysis of Corridor Delay Under SCATS Control." Michigan State University. April 1998. 88. Tobin, Bil l . "Transit Signal Priority Evaluation." Presentation to the Transportation Research Board Transit Signal Priority Workshop, January 13, 2002. (http://signalsvstems.tamu.edu/documents/SignalControlWorkshop2002/TRBPresent ation_Tobin.pdf) 89. Transportation Research Board. TCRP Report 2: Applicability of Low Floor Light Rail Vehicles in North America. TRB. Washington: 1995. pp. 55 - 66. 90. Transportation Research Board. TCRP Report 19: Guidelines for the Location and Design of Bus Stops. Transportation Research Board. Washington: 1996. 91. Transportation Research Board. "TCRP Synthesis 24: A V L Systems for Bus Transit." Transportation Research Board. National Research Council. 1997. www. fta. dot, gov/research/pdf/s24.pdf 92. Transportation Research Board. "This Is Light Rail Transit."8th Joint Conference on Light Rail Transit. Transportation Research Board. Washington: 2000. 93. Tyson, W.J. "Evaluating Alternative Proposals." Light Transit Systems. The Institution of Civil Engineers. Chatham: 1990. 94. U.S. Department of Transportation, Federal Transit Administration and Sacramento Regional Transit. "Downtown Sacramento-Folsom Corridor Project." USDOT. Washington: 2000. 95. Vahidi, H . "Transit Signal Priority: A Comparison of Recent and Future Implementations." Presented at 70 t h Annual ITE Meeting in Nashville, TN, 2000. 96. Vigrass, J.W. "Salt Lake City: US's Newest Light Rail Line." Tramways and Urban Transit. February 2001. pp. 57 - 58. 97. Vigrass, J. W. "TRAX US's Newest Light Rail Network." Tramways and Urban Transit. April 2001. pp. 138- 139. 98. Wang, S.C, Hai Yang, W.S. Au Yeung, S.L. Cheuk and M.K. Lo. "Delay at Signal-Controlled Intersections with Bus Stop Upstream." Journal of Transportation Engineering. May/June 1998. Volume 124, Issue 3. pp. 229 - 234. 99. Wansbeek, C J . "Porto—Europe's Largest Light Rail Project." Tramways and Urban Transit. December 2001. pp. 448 - 451. 100. Wansbeek, C J . "Stockholm: Extensions are Planned Already." Tramways and Urban Transit. March 2001. pp. 88 - 92. University of British Columbia M.A.Sc. Thesis Michael Barton Evaluation of Transit Signal Priority Options for Rapid Page 125 Transit and Light Rail Transit in the City of Richmond 101. Westwood Professional Services Inc. "Signal Priority for Buses: An Operational Test at Louisiana Avenue." December 1995. 102. Wolshon, Brian and William C. Taylor. "Impact of Adaptive Signal Control on Major and Minor Approach Delay." Journal of Transportation Engineering. January/February 1999. Vol. 125, No. 1. ASCE. Pp. 30-38. 103. Wu, Jianping and Nick Hounsell. "Bus Priority Using Pre-Signals." Transportation Research A. Vol . 32, No. 8 pp. 563 - 583. Elsevier Science Ltd. 1998. 104. Wu, Jianping, Mike McDonald and Nick Hounsell. "Characteristics of Light Rail Travel Time: Examples from France." Journal of Public Transportation. Winter 1997. Tampa: National Urban Transit Institute, pp. 73 - 92. 105. Zhang, Yihua. "An Evaluation of Transit Signal Priority and SCOOT Adaptive Signal Control." Master's Thesis. Virginia Polytechnic Institute. 2001. University of British Columbia M.A.Sc. Thesis 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0063904/manifest

Comment

Related Items