International Construction Specialty Conference of the Canadian Society for Civil Engineering (ICSC) (5th : 2015)

Multi-agent system for improved safety and productivity of earthwork equipment using real-timelocation… Vahdatikhaki, Faridaddin; Hammad, Amin; Langari, Seied Mohammad Jun 30, 2015

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

Item Metadata

Download

Media
52660-Vahdatikhaki_F_et_al_ICSC15_315_Multi_Agent_System.pdf [ 1.98MB ]
52660-Vahdatikhaki_F_et_al_ICSC15_315_Multi_Agent_System_slides.pdf [ 3.33MB ]
Metadata
JSON: 52660-1.0076351.json
JSON-LD: 52660-1.0076351-ld.json
RDF/XML (Pretty): 52660-1.0076351-rdf.xml
RDF/JSON: 52660-1.0076351-rdf.json
Turtle: 52660-1.0076351-turtle.txt
N-Triples: 52660-1.0076351-rdf-ntriples.txt
Original Record: 52660-1.0076351-source.json
Full Text
52660-1.0076351-fulltext.txt
Citation
52660-1.0076351.ris

Full Text

5th International/11th Construction Specialty Conference 5e International/11e Conférence spécialisée sur la construction    Vancouver, British Columbia June 8 to June 10, 2015 / 8 juin au 10 juin 2015   MULTI-AGENT SYSTEM FOR IMPROVED SAFETY AND PRODUCTIVITY OF EARTHWORK EQUIPMENT USING REAL-TIME LOCATION SYSTEMS  Faridaddin Vahdatikhaki1, Amin Hammad2, 3 and Seied Mohammad Langari1 1 Department of Building, Civil and Environmental Engineering, Concordia University, Canada 2 Concordia Institute for Information Systems Engineering, Concordia University, Canada 3 hammad@ciise.concordia.ca Abstract: The growing complexity and scope of construction projects is making productivity and safety of earthwork of a great concern for project and site managers. In earthwork operations, where heavy machines are being used, various safety and risk issues put the timely completion of a project at stake. Additionally, the construction working environment is heavily susceptible to unforeseen changes and circumstances that could impact the project, both cost and schedule wise. As a response to the looming safety threats or unforeseen changes of working conditions, re-planning is almost always required. In order for re-planning to yield the optimum results, real-time information gathering and processing is a must. GPS and other Real-time Location Systems (RTLSs) have been used for the purpose of real-time data gathering and decision-making in recent years. Similarly, Location-based Guidance Systems (LGSs), e.g., Automated Machine Control/Guidance (AMC/G), are introduced and have been employed mainly for the purpose of high-precision earthwork operations. However, the current application of LGS is limited to the machine-level productivity optimization, which is not sufficient to address the project-level monitoring and decision-making needs. In the context of complex earthwork operations where several teams are concurrently working towards different ends, the globally optimized operations should coordinate the actions of multiple teams of equipment to eliminate the productivity lost by organizational, logistics and operational management. Therefore, the objective of this paper is to develop a Multi-agent System (MAS) structure to orchestrate the machine-level information (i.e. states and poses) induced based on RTLSs to a coherent project-level system committed to support operations towards the enhanced productivity and safety of the overall project. In the proposed MAS, several layers of agents are processing and managing the huge amount of collected sensory data into useful information that can be used in decision making at different operational levels. The proposed MAS has a semi-distributed structure to strike a balance between the optimality of the outputs and the required computational efforts. A case study is developed to demonstrate the applicability of the proposed MAS. Also, a two-layer safety mechanism is proposed based on which near real-time collision-free path planning and real-time collision avoidance can be performed. In the light of the results of the case study, it is found that the the proposed MAS structure is able to effectively address the team-level coordination of different pieces of equipment and improve the safety of construction site using the proposed two-layer safety mechanism.  1 INTRODUCTION The construction industry is concerned with Improving the productivity and safety of construction projects (Beavers et al. 2006). In earthwork operations, where heavy machines are being used, various safety and 315-1 risk issues put the timely completion of a project at stake. Additionally, the construction working environment is heavily susceptible to unforeseen changes and circumstances that could impact the project, both cost and schedule wise. As a response to the looming safety threats or unforeseen changes of working conditions, re-planning is almost always required. In order for re-planning to yield the optimum results, real-time information gathering and processing is a must. The Global Positioning System (GPS) and other Real-time Location Systems (RTLSs) have been used for the purpose of real-time data gathering and decision-making in recent years (Perkinson et al. 2010). Similarly, Location-based Guidance Systems (LGSs), e.g., Automated Machine Control/Guidance (AMC/G), are introduced and have been employed mainly for the purpose of high-precision earthwork operations. LGS integrates geo-positioning technologies with 3D design models and Digital Terrain Models (DTMs) to either (1) support the machine operator through the provision of continuous guidance on a digital screen mounted in the cabin of the machine, or (2) control the position and movements of the equipment (or part of it). While GPS and total stations are the main tracking technologies used in AMC/G, other types of emerging Real-time Location Systems (RTLS), e.g., Ultra-Wideband (UWB), can be integrated with similar monitoring mechanisms to provide monitoring and guidance capabilities for earthwork equipment.   The current application of LGS is limited to the machine-level productivity optimization in large projects, which is not sufficient to address the project-level monitoring and decision-making needs.There are several challenges that have to be overcome in order to maximize the benefits of using this technology in the 3D surveying-design-contract-construction-inspection workflow (Dunston and Monty 2009, Torres and Ruiz 2011, Vonderohe 2009). The problem of providing near real-time guidance or control support for the operators of earthwork equipment based on the consideration of the entire fleet can become complex, in line with the fleet size and equipment interactions. For such complex problems, the conventional approach of central problem solving becomes far-fetched, attributable to the fact it is difficult or impractical to globally grasp and analyze the multi-dimensionality and dynamisms of such problems. Distributed intelligent systems are designed to address such complex problems in terms of several collaborating intelligent agents, who try to solve the overall problem by synthesizing limited views of individual agents (Ferber 1999). Such systems are referred to as Multi-Agent Systems (MASs), which consist of several intelligent agents capable of interaction.  Furthermore, despite the growing availability of LGS, its application for safety is limited to real-time proximity-based object detection and warnings. In the existing systems, the increasingly affordable advanced sensing and location systems are used to mitigate the collision risks by warning the operators against the potential dangerous proximities in real time (Burns 2002, Carbonari et al. 2011, Zhang and Hammad 2012, Guenther and Salow 2012, Wu et al. 2013, Zolynski et al. 2014, Vahdatikhaki and Hammad 2015a). Cheng (2013) proposed to use the pose and speed data for the generation of the workspaces. This method does not consider the equipment state as a means to economize the use of space around the equipment and does not cover the equipment with rotary movements, e.g., excavators. Therefore, there is a need for a solution that is able to reliably predict the operation of the equipment for a long-enough time window to enable different pieces of equipment to adjust their planned paths to avoid collisions in near-real time. Therefore, the objective of this paper is to develop a MAS structure to orchestrate the machine-level information (i.e. states and poses) induced based on RTLSs to a coherent project-level system committed to support operations towards the enhanced productivity and safety of the overall project. The paper also aims to develop a two-layer safety mechanism: the first layer of which enables the equipment to plan a collision-free path considering the predicted movement of all other equipment, and the second layer is acting as a last line of defense in view of possible discrepancies between the predicted paths and actual paths. The structure of the paper is as follows. First, the proposed method is introduced, followed by the explanation of the implementation and a case study. Finally, the conclusions and future work are presented.       2 PROPOSED METHOD Figure 1 shows an overview of the scope for the proposed MAS framework. The main assumptions are that every piece of equipment on the construction site has a sufficient number of RTLS Data Collectors 315-2 (DCs) attached at specific locations to track its movement, and that every equipment operator is supported by an agent that can communicate with other agents in a MAS framework. The proposed MAS supports the project at three different levels: (1) Planning, (2) execution and monitoring, and (3) re-planning. At the planning level, the MAS is able to streamline the operation and task assignments to different equipment as well as to perform equipment path planning (Zhang and Hammad 2012), which is operationalized in terms of strategic and tactical planning. At the execution and monitoring level, MAS is committed to (i) provide visual guidance to equipment operators, (ii) collect and process RTLS data, (iii) apply appropriate  error correction techniques to identify the pose of the equipment (Vahdatikhaki et al. 2015), (iv) identify the state of the equipment (Vahdatikhaki and Hammad 2014), (v) apply the Near Real-time Simulation (NRTS) (Vahdatikhaki and Hammad 2014), (vi) generate equipment workspaces, i.e., Dynamic Equipment Workspaces (DEWs) (Vahdatikhaki and Hammad 2015a) and Look-Ahead Equipment Workspaces (LAEWs) (Vahdatikhaki and Hammad 2015b), and (vii) report the necessary information to pertinent agents. The aforementioned two types of workspaces differ in that while DEWs are generated based on the equipment pose and speed in real time to form a safety buffer around the equipment that can help prevent collisions, LAEWs are built based on the predicted future motion of equipment and operator visibility in near-real time to help find a collision-free path for the equipment, as explained in Section 2.2. Finally, at the re-planning level, the proposed MAS framework addresses the need for task-reassignment, path re-planning, and design change requests, which may become necessary in view of the potential unforeseen safety risks identified at the monitoring level. As can be seen in Figure 1, while the proposed MAS framework offers advantages at both the operational and managerial levels, only the operational aspects of framework are addressed in this paper.   Equipment State and Pose IdentificationWorkspace Generation Strategic and Tactical Path Planning Collecting and Processing RTLS DataPlanning Operation and Task AssignmentExecution and Monitoring Near-real-time SimulationReporting (Progress Tracking, Safety Warnings, and Delay Notice)Re-PlanningTask Re-assignmentPath Re-planningDesign Change RequestsVisual Guidance to Equipment Operators Figure 1: Overview of the Scope for the Proposed MAS Framework  The authors have previously presented the overview of the proposed MAS (Hammad et al., 2013). This paper extends this research by providing a more in-depth discussion of the agents’ functionality in the MAS and how the LAEWs are being used by different agents to avoid collisions.  A multi-layer agent architecture is proposed in which agents supporting the operators of single machines constitute the lowermost layer of the agent hierarchy. These agents process and manage the huge amount of sensory data, provided by an UWB system or other types of location systems, into useful information that can be used in decision making at different operational and managerial levels. Figure 2 315-3 architecture is based on breaking the activities of an agent in vertical modules where every module has limited responsibilities and the results of the higher modules always supersede those of the lower modules, if there is a conflict between various modules (Ferber 1999). In a nutshell, OAs constantly monitor the operations and perform the routine calculations for the equipment condition monitoring, pose and state-identifications, cycle time, generation of tactical plans, generation of risk maps, detecting underground utilities, and generating DEWs.    Avoid Safety RisksExecute TasksStartAny Safety Risks ?YesNew or Incomplete Task ?YesInput InformationNoMonitor the SiteUpdate DTMMonitor the EquipmentPose and State-IdentificationReport to TCA and other OAsGenerate DEWGenerate the Risk mapCalculate the Cycle TimeDetect UtilitiesGenerate Tactical PlanEndNo  Inform all Affected OAs, TCAs and GCA Assign TasksStartAny Safety Risks ?YesOngoing Operations?YesInput InformationNoMonitor and Update the OAsResolvable  Locally?YesReschedule TaskNoDelayed ?YesNew Operation?YesNoEndPerform NRTSReport to GCA or Higher TCAMeasure the Progress Generate Strategic PlanGenerate LAEWsNoNo (a) (b) Figure 3: High-Level Flowchart of (a) the OA Functionalities, and (b) the TCA Functionalities 2.1.2 Coordination Agents Coordination encompasses agents representing team coordinators who are responsible for making critical decisions, e.g., new work schedules or command for the suspension of the operation, using data from all other agents, and further communicating their decisions with the appropriate OAs for the execution. Essentially, this component consists of one GCA and several TCAs. However, depending on the characteristics of the project, the phase of the project and simultaneous operations, several layers of teams and sub-teams can be formed. Each team is coordinated and supported by a TCA. The role of a TCA is to assign tasks to the subordinate OAs or sub-TCAs and to collect information from them. Figure 3(b) shows the high-level flowchart of the TCA functionalities. The main functionality of a TCA is to assign and monitor the tasks of the OAs. At the top of the flowchart, the TCA determines whether a new operation is assigned or an operation is ongoing. In the first case, the operation is broken into OA-executable tasks and assigned to the relevant available OAs. Next, in view of the reports from subordinate OAs, the progress monitoring, NRTS, and LAEWs (if any risk is identified), either the tasks are rescheduled if the problem can be resolved locally, or the GCA (or higher level TCA) is informed for directions. Local resolvability means that the problem can be solved by the information present to a single TCA, without the need to engage into negotiations with other TCAs. The negotiation between agents in a decentralized MAS structure is outside the scope of the present paper. The GCA is responsible for monitoring and controlling the operations to ensure the smooth execution of the project. The GCA also generates the operations’ schedule and the resource distribution based on the available resources, project schedule, the chosen construction methods and available sub-contractors. The functionalities of the GCA are realized through the accumulation of information about the project and the progress of different operations. The project information is the combination of all essential documents/information based on which an earthwork project is executed. At a high-level abstraction, safety regulations, available resources, project schedule, construction methods, and available sub-315-5 contractors, all of which are coming from the PDA, are the main ingredients of the project information. Safety regulations are used to derive basic safety rules that need to be observed throughout the project. Available resources and available sub-contractors are used for the resource configurations and distribution. The project schedule is used for the generation of operation schedules that can be assigned to different TCAs. The Construction methods provide the GCA with the initial information needed to retrieve the right operation procedures.  2.2 Safety Management in MAS As stated in Section 2, the safety of earthwork operation in the proposed MAS structure is supported through a two-layer mechanism which includes near real time collision-free path (re-) planning using LAEWs and real-time collision avoidance using DEWs. These two layers are running independently in parallel with different update rates. Given the nature and functionality of DEWs, they are updated in real time with the same rate offered by the tracking technology (dt). LAEWs, on the other hand, require intensive computations and communications between various agents, and thus they are updated with a rate less than DEWs. The LAEWs are generated over every Δt and whenever a deviation from the predicted path of various equipment is observed. While the details of the two types of workspace are presented in the previous work of the author (Vahdatikhaki and Hammad 2015a, Vahdatikhaki and Hammad 2015b), a brief explanation of each workspace is presented in the following sections.  2.2.1   Look-Ahead Equipment Workspace (LAEW) The flowchart of the proposed method for the generation of the LAEW of one piece of equipment (equipment q) is shown in Figure 4(a). As shown in this figure, the input of this method comprises the sensory data, the equipment specifications and its accurate 3D model, the current pose and state data generated by the OA of the equipment q (OAq), and future state data coming from the NRTS that is performed by the TCA. A rule-based system is used to identify the states of different equipment with a high accuracy by leveraging a set of equipment proximity and motion rules that determine the states of the equipment (Vahdatikhaki and Hammad 2014). Also, a robust optimization-based method that uses geometric and operational characteristics of the equipment is used to improve the quality of the pose estimation (Vahdatikhaki et al. 2015). Additionally, the updated 3D model of the site, and the project’s detailed plan (including the location of different scheduled tasks, their time frame, and the site layout) are available through the Information Agent. Finally, a set of heuristic rules that define the operation of a skilled operator is also required to be available to each OA. The generation of LAEW is based on the discretization of the entire site space into cells, and then calculating the risk associated with each cell given the future expected states of different pieces of equipment, which is performed by each OA. As shown in Figure 4(b), the pose data are used to identify the current state, which is then passed on to the TCA to perform the NRTS in order to generate the operational pattern of each OA. These data are then communicated with the OAq who will first integrate the equipment pose with its 3D model and the updated 3D model of site to situate the equipment in the virtual environment. Then, the OA will use the project plan, and the rules that govern the operation of the machine by a skilled operator to generate the risk map of the equipment. Finally, the OAs transfer their individual risk maps to the TCAs who will first combine these risk maps and then use the tolerable risk level of each OA to generate the LAEW. It should be highlighted that LAEWp for equipment p is generated based on the combination of the risk maps from all pieces of equipment surrounding equipment p, excluding equipment p itself. LAEWp can be used by the OAp to perform path re-planning, if required. Similarly, the path-replanning performed by the OAq at the end of the flowchart shown in Figure 4(a) is realized using LAEWq. 2.2.2 Dynamic Equipment Workspace (DEW) DEWs aim to use the pose, state, and speed characteristics of the equipment to generate a space around the equipment that would allow the prevention of immediate collisions with other pieces of equipment or obstacles on site, considering the equipment stoppage time (ts). ts  can be used to determine how much of the space in the moving direction of equipment is unsafe after the operator becomes aware of a potential collision considering the operator reaction time and braking time. In addition to the DEWs of the equipment, semi-dynamic obstacles (such as trenches, temporary or permanent structures, etc.), also 315-6 need to be represented by their own corresponding safety zones to enable effective collision avoidance at the global level.   Apply Pose and State IdentificationApply NRTSGenerate the Risk MapGenerate LAEWp based on Specific Risk Threshold Apply Path Re-planningStartEndSensory DataUpdated 3D Model of SiteEquipment Specs and 3D ModelSituate the Equipment in Virtual Environment Skilled Operator Rules for Equipment Path PlanningProject PlanTCA OA of Equipment q Information AgentCombine risk maps of different OAs  (a) (b) Figure 4: (a) Flowchart for the Generation of LAEW, and (b) Schematic Representation of LAEW Generation Process  For the DEWs to be effectively used for the purpose of collision detection and avoidance, every OA needs to be able to generate its own DEW and have near-real-time information about the DEWs of other OAs. Figure 5(a) shows the flowchart for the generation of the proposed DEWs. With the 3D model of the equipment and its pose and state information available, the method proceeds to determine the linear and angular speeds of the equipment. For instance, an excavator can travel on its tracks with the linear speed of , move its bucket with the linear speed of , or swing with the angular speed of . Upon the determination of the speed vectors, the DEW can be generated based on the type of the equipment and the equipment state. For example, two distinct types of states can be identified for an excavator, namely stationary states (swinging, loading, dumping, and waiting) and traversal states (relocating, maneuvering). Figure 5(b) shows different DEWs of an excavator for different states. Next, to avoid redundant computation, an OA can perform pairwise comparisons of DEWs only with the OAs that are in its vicinity. To determine the equipment in vicinity, the multi-layer workspace concept (Chae 2008, Wang and Razavi 2015) can be applied. In this method, the pairwise distances between every two pieces of equipment are calculated and if the two pieces of equipment have a distance less than a threshold, then the collision detection between their DEWs is performed. In order to further reduce the computation efforts and avoid redundant calculations, the priorities of the different equipment can be used to delegate the calculation to the OA of the equipment with the lower priority. If a collision is detected between the two, the equipment with the lower priority will stop and send a warning to the OA of the other equipment. If both pieces of equipment have the same priority, then the OAs of both should perform the collision detection and if a collision is detected they should both stop.  3 IMPLEMENTATION AND CASE STUDY In order to demonstrate the feasibility of the proposed MAS approach in improving safety using LAEWs and DEWs, a prototype system is implemented using Unity3D game engine (2015) and two simulated scenarios are examined. The scenarios used for the case study consider an excavation operation for a 315-7  StartDigging done at DS?YesGenerate path to digging Generate path to digging using RRTGenerate risk mapReceive LAEWs from TCASafe?Higher priority?NoNoMove to diggingYesYesNoDumpGenerate path to truck Generate path to truck using RRTGenerate risk mapReceive LAEWs from TCASafe?Higher priority?NoNoMove to truckYesYesEnd  Excavator A Truck A Excavator B Truck B  (b)  (a) (c)  Figure 6: (a) Algorithm Representing the Operation Logic of Excavator, (b) Current Poses and Initial Paths of Excavator, (c) LAEW of Excavator B and Final Path of Equipment B    Truck B Truck A   Truck B Truck A  Truck B Truck A  (a) (b) (c) Figure 7: (a) The Layout of the Second Scenario, (b) Collision Detection between DEWs, and (c) Collision Avoidance Decision made by OAs  4 CONCLUSIONS AND FUTURE WORK In this paper, a MAS structure is introduced for improving the safety and productivity of automated guidance and control of earthwork equipment. In the proposed MAS structure, every piece of equipment is supported by an operator agent to oversee the task and provide guidance whenever needed. A multi-layer agent hierarchy assigns monitors and coordinates the task executions, and a set of three types of agents feed the system with the relevant information. The functionalities, jurisdictions and the input-output scheme of every type of agents were discussed. A two-layer safety mechanism was introduced, where the first layer enables the equipment to plan a collision-free path considering the predicted movement of all other equipment and the second layer acts as a last-line-of-defense in view of possible discrepancies between the predicted paths and actual paths.  In view of the results of the case study, it is shown that the MAS is capable of effectively handling the harmonization of various pieces of equipment on the site beyond what is available by the conventional 315-9 LGSs. The combination of LAEWs and DEWs are found to be an efficient approach to deal with collision-free path planning and real-time collision avoidance. The authors are planning to investigate the negotiation between different levels of agents as part of their future work.   References  Beavers, J. E., Moore, J. R., Rinehart, R. and Schriver, W. R. 2006. Crane-related fatalities in the construction industry. Journal of Construction Engineering and Management, 132(9):901-910. Burns, R. L. 2002. Dynamic Safety Envelope For Autonomous-Vehicle Collision Avoidance System. U.S., Patent No. US 6393362. Carbonari, A., Giretti, A. and Naticchia, B. 2011. A proactive system for real-time safety management in construction sites. Automation in Construction, 20(6), 686-698. Chae, S. 2009. Development of warning system for preventing collision accident on construction site. In Proceedings of the 26th Int. Symposium on Automation and Robotics in Construction. Cheng, T., 2013. Automated Safety Analysis of Construction Site Activities Using Spatio-temporal Data, Doctoral dissertation, Georgia Institute of Technology. Dunston, P. S. and Monty, J. 2009. Practices for seamless transmission of design data from design phase to construction equipment operation - a synthesis study, Purdue University, West Lafayette, Indiana. Ferber, J. 1999. Multi-agent systems: an introduction to distributed artificial intelligence. Addison-Wesley, Harlow, UK. Guenther , N. and Salow, H. 2012. Collision avoidance and operator guidance innovating mine vehicle safety. Queensland Mining Industry Health & Safety Conference, Townsville. Hammad, A., El Ammari, K., Langari, S. M., Vahdatikhaki, F., Soltani, M., AlBahnassi, H., Paes, B. 2014. Simulating Macro and Micro Path Planning of Excavation Operations Using Game Engine. Proceedings of the 2014 Winter Simulation Conference, Tolk, S. Y. Diallo, I. O. Ryzhov, L. Yilmaz, S. Buckley, and J. A. Miller (eds.), Savannah, GA. Hammad, A., Vahdatikhaki, F. and Zhang, C. 2013. A novel integrated approach to project-level automated machine control/guidance systems in construction projects. ITcon, 162-181. La Valle, S. M. 2006. Planning Algorithms. Cambridge University Press, New York. Perkinson, C. L., Bayraktar, M. E. and Ahmad, I. 2010. The use of computing technology in highway construction as a total jobsite management tool. Automation in Construction, 19(7), 884-897. Torres, H. N. and Ruiz, M. 2011. Improving Highway Project Delivery. TRB Annual Meeting, Washington, D.C.,  Unity3D, 2015. Unity - Game Engine. Retrieved on 03/26, 2014, from http://unity3d.com/ Vahdatikhaki, F. and Hammad, A. 2014. Framework for near real-time simulation of earthmoving projects using location tracking technologies. Automation in Construction, 42:50–67. Vahdatikhaki, F. and Hammad, A. 2015a. Dynamic Equipment Workspace generation for improving earthwork safety using real-time location system. Advanced Engineering Informatics, (Accepted with modifications). Vahdatikhaki, F. and Hammad, A. 2015b. Risk-based Look-ahead Workspace Generation for Earthwork Equipment Using Near Real-time Simulation. Automation in Construction (Accepted with modifications). Vahdatikhaki, F., Hammad, A. and Siddiqui, H. 2015. Optimization-based Excavator Pose Estimation Using Real-time Location Systems. Automation in Construction (accepted with modifications). Vonderohe, A. P. 2009. Report of status and plans for implementing technologies for design and con-struction in WisDOT, Construction Materials and Support Center. Wang, J. and Razavi, S. N. 2015. Low False Alarm Rate Model for Unsafe-Proximity Detection in Construction. Journal of Computing in Civil Engineering. Wu, H., Tao, J., Li, X., Chi, X., Hua, X., Yang, R., Wang, S., Chen, N. 2013. A location based service approach for collision warning systems in concrete dam construction. Safety Science , 51(1):338-346. Zhang, C. and Hammad, A. 2012. Improving lifting motion planning and re-planning of cranes with consideration for safety and efficiency. Journal of Advance Engineering Informatics, 26(2):396-410. Zolynski, G., Schmidt, D. and Berns, K. 2014. Safety for an Autonomous Bucket Excavator During Typical Landscaping Tasks. New Trends in Medical and Service Robots, 20: 357-368.  315-10  5th International/11th Construction Specialty Conference 5e International/11e Conférence spécialisée sur la construction    Vancouver, British Columbia June 8 to June 10, 2015 / 8 juin au 10 juin 2015   MULTI-AGENT SYSTEM FOR IMPROVED SAFETY AND PRODUCTIVITY OF EARTHWORK EQUIPMENT USING REAL-TIME LOCATION SYSTEMS  Faridaddin Vahdatikhaki1, Amin Hammad2, 3 and Seied Mohammad Langari1 1 Department of Building, Civil and Environmental Engineering, Concordia University, Canada 2 Concordia Institute for Information Systems Engineering, Concordia University, Canada 3 hammad@ciise.concordia.ca Abstract: The growing complexity and scope of construction projects is making productivity and safety of earthwork of a great concern for project and site managers. In earthwork operations, where heavy machines are being used, various safety and risk issues put the timely completion of a project at stake. Additionally, the construction working environment is heavily susceptible to unforeseen changes and circumstances that could impact the project, both cost and schedule wise. As a response to the looming safety threats or unforeseen changes of working conditions, re-planning is almost always required. In order for re-planning to yield the optimum results, real-time information gathering and processing is a must. GPS and other Real-time Location Systems (RTLSs) have been used for the purpose of real-time data gathering and decision-making in recent years. Similarly, Location-based Guidance Systems (LGSs), e.g., Automated Machine Control/Guidance (AMC/G), are introduced and have been employed mainly for the purpose of high-precision earthwork operations. However, the current application of LGS is limited to the machine-level productivity optimization, which is not sufficient to address the project-level monitoring and decision-making needs. In the context of complex earthwork operations where several teams are concurrently working towards different ends, the globally optimized operations should coordinate the actions of multiple teams of equipment to eliminate the productivity lost by organizational, logistics and operational management. Therefore, the objective of this paper is to develop a Multi-agent System (MAS) structure to orchestrate the machine-level information (i.e. states and poses) induced based on RTLSs to a coherent project-level system committed to support operations towards the enhanced productivity and safety of the overall project. In the proposed MAS, several layers of agents are processing and managing the huge amount of collected sensory data into useful information that can be used in decision making at different operational levels. The proposed MAS has a semi-distributed structure to strike a balance between the optimality of the outputs and the required computational efforts. A case study is developed to demonstrate the applicability of the proposed MAS. Also, a two-layer safety mechanism is proposed based on which near real-time collision-free path planning and real-time collision avoidance can be performed. In the light of the results of the case study, it is found that the the proposed MAS structure is able to effectively address the team-level coordination of different pieces of equipment and improve the safety of construction site using the proposed two-layer safety mechanism.  1 INTRODUCTION The construction industry is concerned with Improving the productivity and safety of construction projects (Beavers et al. 2006). In earthwork operations, where heavy machines are being used, various safety and 315-1 risk issues put the timely completion of a project at stake. Additionally, the construction working environment is heavily susceptible to unforeseen changes and circumstances that could impact the project, both cost and schedule wise. As a response to the looming safety threats or unforeseen changes of working conditions, re-planning is almost always required. In order for re-planning to yield the optimum results, real-time information gathering and processing is a must. The Global Positioning System (GPS) and other Real-time Location Systems (RTLSs) have been used for the purpose of real-time data gathering and decision-making in recent years (Perkinson et al. 2010). Similarly, Location-based Guidance Systems (LGSs), e.g., Automated Machine Control/Guidance (AMC/G), are introduced and have been employed mainly for the purpose of high-precision earthwork operations. LGS integrates geo-positioning technologies with 3D design models and Digital Terrain Models (DTMs) to either (1) support the machine operator through the provision of continuous guidance on a digital screen mounted in the cabin of the machine, or (2) control the position and movements of the equipment (or part of it). While GPS and total stations are the main tracking technologies used in AMC/G, other types of emerging Real-time Location Systems (RTLS), e.g., Ultra-Wideband (UWB), can be integrated with similar monitoring mechanisms to provide monitoring and guidance capabilities for earthwork equipment.   The current application of LGS is limited to the machine-level productivity optimization in large projects, which is not sufficient to address the project-level monitoring and decision-making needs.There are several challenges that have to be overcome in order to maximize the benefits of using this technology in the 3D surveying-design-contract-construction-inspection workflow (Dunston and Monty 2009, Torres and Ruiz 2011, Vonderohe 2009). The problem of providing near real-time guidance or control support for the operators of earthwork equipment based on the consideration of the entire fleet can become complex, in line with the fleet size and equipment interactions. For such complex problems, the conventional approach of central problem solving becomes far-fetched, attributable to the fact it is difficult or impractical to globally grasp and analyze the multi-dimensionality and dynamisms of such problems. Distributed intelligent systems are designed to address such complex problems in terms of several collaborating intelligent agents, who try to solve the overall problem by synthesizing limited views of individual agents (Ferber 1999). Such systems are referred to as Multi-Agent Systems (MASs), which consist of several intelligent agents capable of interaction.  Furthermore, despite the growing availability of LGS, its application for safety is limited to real-time proximity-based object detection and warnings. In the existing systems, the increasingly affordable advanced sensing and location systems are used to mitigate the collision risks by warning the operators against the potential dangerous proximities in real time (Burns 2002, Carbonari et al. 2011, Zhang and Hammad 2012, Guenther and Salow 2012, Wu et al. 2013, Zolynski et al. 2014, Vahdatikhaki and Hammad 2015a). Cheng (2013) proposed to use the pose and speed data for the generation of the workspaces. This method does not consider the equipment state as a means to economize the use of space around the equipment and does not cover the equipment with rotary movements, e.g., excavators. Therefore, there is a need for a solution that is able to reliably predict the operation of the equipment for a long-enough time window to enable different pieces of equipment to adjust their planned paths to avoid collisions in near-real time. Therefore, the objective of this paper is to develop a MAS structure to orchestrate the machine-level information (i.e. states and poses) induced based on RTLSs to a coherent project-level system committed to support operations towards the enhanced productivity and safety of the overall project. The paper also aims to develop a two-layer safety mechanism: the first layer of which enables the equipment to plan a collision-free path considering the predicted movement of all other equipment, and the second layer is acting as a last line of defense in view of possible discrepancies between the predicted paths and actual paths. The structure of the paper is as follows. First, the proposed method is introduced, followed by the explanation of the implementation and a case study. Finally, the conclusions and future work are presented.       2 PROPOSED METHOD Figure 1 shows an overview of the scope for the proposed MAS framework. The main assumptions are that every piece of equipment on the construction site has a sufficient number of RTLS Data Collectors 315-2 (DCs) attached at specific locations to track its movement, and that every equipment operator is supported by an agent that can communicate with other agents in a MAS framework. The proposed MAS supports the project at three different levels: (1) Planning, (2) execution and monitoring, and (3) re-planning. At the planning level, the MAS is able to streamline the operation and task assignments to different equipment as well as to perform equipment path planning (Zhang and Hammad 2012), which is operationalized in terms of strategic and tactical planning. At the execution and monitoring level, MAS is committed to (i) provide visual guidance to equipment operators, (ii) collect and process RTLS data, (iii) apply appropriate  error correction techniques to identify the pose of the equipment (Vahdatikhaki et al. 2015), (iv) identify the state of the equipment (Vahdatikhaki and Hammad 2014), (v) apply the Near Real-time Simulation (NRTS) (Vahdatikhaki and Hammad 2014), (vi) generate equipment workspaces, i.e., Dynamic Equipment Workspaces (DEWs) (Vahdatikhaki and Hammad 2015a) and Look-Ahead Equipment Workspaces (LAEWs) (Vahdatikhaki and Hammad 2015b), and (vii) report the necessary information to pertinent agents. The aforementioned two types of workspaces differ in that while DEWs are generated based on the equipment pose and speed in real time to form a safety buffer around the equipment that can help prevent collisions, LAEWs are built based on the predicted future motion of equipment and operator visibility in near-real time to help find a collision-free path for the equipment, as explained in Section 2.2. Finally, at the re-planning level, the proposed MAS framework addresses the need for task-reassignment, path re-planning, and design change requests, which may become necessary in view of the potential unforeseen safety risks identified at the monitoring level. As can be seen in Figure 1, while the proposed MAS framework offers advantages at both the operational and managerial levels, only the operational aspects of framework are addressed in this paper.   Equipment State and Pose IdentificationWorkspace Generation Strategic and Tactical Path Planning Collecting and Processing RTLS DataPlanning Operation and Task AssignmentExecution and Monitoring Near-real-time SimulationReporting (Progress Tracking, Safety Warnings, and Delay Notice)Re-PlanningTask Re-assignmentPath Re-planningDesign Change RequestsVisual Guidance to Equipment Operators Figure 1: Overview of the Scope for the Proposed MAS Framework  The authors have previously presented the overview of the proposed MAS (Hammad et al., 2013). This paper extends this research by providing a more in-depth discussion of the agents’ functionality in the MAS and how the LAEWs are being used by different agents to avoid collisions.  A multi-layer agent architecture is proposed in which agents supporting the operators of single machines constitute the lowermost layer of the agent hierarchy. These agents process and manage the huge amount of sensory data, provided by an UWB system or other types of location systems, into useful information that can be used in decision making at different operational and managerial levels. Figure 2 315-3 architecture is based on breaking the activities of an agent in vertical modules where every module has limited responsibilities and the results of the higher modules always supersede those of the lower modules, if there is a conflict between various modules (Ferber 1999). In a nutshell, OAs constantly monitor the operations and perform the routine calculations for the equipment condition monitoring, pose and state-identifications, cycle time, generation of tactical plans, generation of risk maps, detecting underground utilities, and generating DEWs.    Avoid Safety RisksExecute TasksStartAny Safety Risks ?YesNew or Incomplete Task ?YesInput InformationNoMonitor the SiteUpdate DTMMonitor the EquipmentPose and State-IdentificationReport to TCA and other OAsGenerate DEWGenerate the Risk mapCalculate the Cycle TimeDetect UtilitiesGenerate Tactical PlanEndNo  Inform all Affected OAs, TCAs and GCA Assign TasksStartAny Safety Risks ?YesOngoing Operations?YesInput InformationNoMonitor and Update the OAsResolvable  Locally?YesReschedule TaskNoDelayed ?YesNew Operation?YesNoEndPerform NRTSReport to GCA or Higher TCAMeasure the Progress Generate Strategic PlanGenerate LAEWsNoNo (a) (b) Figure 3: High-Level Flowchart of (a) the OA Functionalities, and (b) the TCA Functionalities 2.1.2 Coordination Agents Coordination encompasses agents representing team coordinators who are responsible for making critical decisions, e.g., new work schedules or command for the suspension of the operation, using data from all other agents, and further communicating their decisions with the appropriate OAs for the execution. Essentially, this component consists of one GCA and several TCAs. However, depending on the characteristics of the project, the phase of the project and simultaneous operations, several layers of teams and sub-teams can be formed. Each team is coordinated and supported by a TCA. The role of a TCA is to assign tasks to the subordinate OAs or sub-TCAs and to collect information from them. Figure 3(b) shows the high-level flowchart of the TCA functionalities. The main functionality of a TCA is to assign and monitor the tasks of the OAs. At the top of the flowchart, the TCA determines whether a new operation is assigned or an operation is ongoing. In the first case, the operation is broken into OA-executable tasks and assigned to the relevant available OAs. Next, in view of the reports from subordinate OAs, the progress monitoring, NRTS, and LAEWs (if any risk is identified), either the tasks are rescheduled if the problem can be resolved locally, or the GCA (or higher level TCA) is informed for directions. Local resolvability means that the problem can be solved by the information present to a single TCA, without the need to engage into negotiations with other TCAs. The negotiation between agents in a decentralized MAS structure is outside the scope of the present paper. The GCA is responsible for monitoring and controlling the operations to ensure the smooth execution of the project. The GCA also generates the operations’ schedule and the resource distribution based on the available resources, project schedule, the chosen construction methods and available sub-contractors. The functionalities of the GCA are realized through the accumulation of information about the project and the progress of different operations. The project information is the combination of all essential documents/information based on which an earthwork project is executed. At a high-level abstraction, safety regulations, available resources, project schedule, construction methods, and available sub-315-5 contractors, all of which are coming from the PDA, are the main ingredients of the project information. Safety regulations are used to derive basic safety rules that need to be observed throughout the project. Available resources and available sub-contractors are used for the resource configurations and distribution. The project schedule is used for the generation of operation schedules that can be assigned to different TCAs. The Construction methods provide the GCA with the initial information needed to retrieve the right operation procedures.  2.2 Safety Management in MAS As stated in Section 2, the safety of earthwork operation in the proposed MAS structure is supported through a two-layer mechanism which includes near real time collision-free path (re-) planning using LAEWs and real-time collision avoidance using DEWs. These two layers are running independently in parallel with different update rates. Given the nature and functionality of DEWs, they are updated in real time with the same rate offered by the tracking technology (dt). LAEWs, on the other hand, require intensive computations and communications between various agents, and thus they are updated with a rate less than DEWs. The LAEWs are generated over every Δt and whenever a deviation from the predicted path of various equipment is observed. While the details of the two types of workspace are presented in the previous work of the author (Vahdatikhaki and Hammad 2015a, Vahdatikhaki and Hammad 2015b), a brief explanation of each workspace is presented in the following sections.  2.2.1   Look-Ahead Equipment Workspace (LAEW) The flowchart of the proposed method for the generation of the LAEW of one piece of equipment (equipment q) is shown in Figure 4(a). As shown in this figure, the input of this method comprises the sensory data, the equipment specifications and its accurate 3D model, the current pose and state data generated by the OA of the equipment q (OAq), and future state data coming from the NRTS that is performed by the TCA. A rule-based system is used to identify the states of different equipment with a high accuracy by leveraging a set of equipment proximity and motion rules that determine the states of the equipment (Vahdatikhaki and Hammad 2014). Also, a robust optimization-based method that uses geometric and operational characteristics of the equipment is used to improve the quality of the pose estimation (Vahdatikhaki et al. 2015). Additionally, the updated 3D model of the site, and the project’s detailed plan (including the location of different scheduled tasks, their time frame, and the site layout) are available through the Information Agent. Finally, a set of heuristic rules that define the operation of a skilled operator is also required to be available to each OA. The generation of LAEW is based on the discretization of the entire site space into cells, and then calculating the risk associated with each cell given the future expected states of different pieces of equipment, which is performed by each OA. As shown in Figure 4(b), the pose data are used to identify the current state, which is then passed on to the TCA to perform the NRTS in order to generate the operational pattern of each OA. These data are then communicated with the OAq who will first integrate the equipment pose with its 3D model and the updated 3D model of site to situate the equipment in the virtual environment. Then, the OA will use the project plan, and the rules that govern the operation of the machine by a skilled operator to generate the risk map of the equipment. Finally, the OAs transfer their individual risk maps to the TCAs who will first combine these risk maps and then use the tolerable risk level of each OA to generate the LAEW. It should be highlighted that LAEWp for equipment p is generated based on the combination of the risk maps from all pieces of equipment surrounding equipment p, excluding equipment p itself. LAEWp can be used by the OAp to perform path re-planning, if required. Similarly, the path-replanning performed by the OAq at the end of the flowchart shown in Figure 4(a) is realized using LAEWq. 2.2.2 Dynamic Equipment Workspace (DEW) DEWs aim to use the pose, state, and speed characteristics of the equipment to generate a space around the equipment that would allow the prevention of immediate collisions with other pieces of equipment or obstacles on site, considering the equipment stoppage time (ts). ts  can be used to determine how much of the space in the moving direction of equipment is unsafe after the operator becomes aware of a potential collision considering the operator reaction time and braking time. In addition to the DEWs of the equipment, semi-dynamic obstacles (such as trenches, temporary or permanent structures, etc.), also 315-6 need to be represented by their own corresponding safety zones to enable effective collision avoidance at the global level.   Apply Pose and State IdentificationApply NRTSGenerate the Risk MapGenerate LAEWp based on Specific Risk Threshold Apply Path Re-planningStartEndSensory DataUpdated 3D Model of SiteEquipment Specs and 3D ModelSituate the Equipment in Virtual Environment Skilled Operator Rules for Equipment Path PlanningProject PlanTCA OA of Equipment q Information AgentCombine risk maps of different OAs  (a) (b) Figure 4: (a) Flowchart for the Generation of LAEW, and (b) Schematic Representation of LAEW Generation Process  For the DEWs to be effectively used for the purpose of collision detection and avoidance, every OA needs to be able to generate its own DEW and have near-real-time information about the DEWs of other OAs. Figure 5(a) shows the flowchart for the generation of the proposed DEWs. With the 3D model of the equipment and its pose and state information available, the method proceeds to determine the linear and angular speeds of the equipment. For instance, an excavator can travel on its tracks with the linear speed of , move its bucket with the linear speed of , or swing with the angular speed of . Upon the determination of the speed vectors, the DEW can be generated based on the type of the equipment and the equipment state. For example, two distinct types of states can be identified for an excavator, namely stationary states (swinging, loading, dumping, and waiting) and traversal states (relocating, maneuvering). Figure 5(b) shows different DEWs of an excavator for different states. Next, to avoid redundant computation, an OA can perform pairwise comparisons of DEWs only with the OAs that are in its vicinity. To determine the equipment in vicinity, the multi-layer workspace concept (Chae 2008, Wang and Razavi 2015) can be applied. In this method, the pairwise distances between every two pieces of equipment are calculated and if the two pieces of equipment have a distance less than a threshold, then the collision detection between their DEWs is performed. In order to further reduce the computation efforts and avoid redundant calculations, the priorities of the different equipment can be used to delegate the calculation to the OA of the equipment with the lower priority. If a collision is detected between the two, the equipment with the lower priority will stop and send a warning to the OA of the other equipment. If both pieces of equipment have the same priority, then the OAs of both should perform the collision detection and if a collision is detected they should both stop.  3 IMPLEMENTATION AND CASE STUDY In order to demonstrate the feasibility of the proposed MAS approach in improving safety using LAEWs and DEWs, a prototype system is implemented using Unity3D game engine (2015) and two simulated scenarios are examined. The scenarios used for the case study consider an excavation operation for a 315-7  StartDigging done at DS?YesGenerate path to digging Generate path to digging using RRTGenerate risk mapReceive LAEWs from TCASafe?Higher priority?NoNoMove to diggingYesYesNoDumpGenerate path to truck Generate path to truck using RRTGenerate risk mapReceive LAEWs from TCASafe?Higher priority?NoNoMove to truckYesYesEnd  Excavator A Truck A Excavator B Truck B  (b)  (a) (c)  Figure 6: (a) Algorithm Representing the Operation Logic of Excavator, (b) Current Poses and Initial Paths of Excavator, (c) LAEW of Excavator B and Final Path of Equipment B    Truck B Truck A   Truck B Truck A  Truck B Truck A  (a) (b) (c) Figure 7: (a) The Layout of the Second Scenario, (b) Collision Detection between DEWs, and (c) Collision Avoidance Decision made by OAs  4 CONCLUSIONS AND FUTURE WORK In this paper, a MAS structure is introduced for improving the safety and productivity of automated guidance and control of earthwork equipment. In the proposed MAS structure, every piece of equipment is supported by an operator agent to oversee the task and provide guidance whenever needed. A multi-layer agent hierarchy assigns monitors and coordinates the task executions, and a set of three types of agents feed the system with the relevant information. The functionalities, jurisdictions and the input-output scheme of every type of agents were discussed. A two-layer safety mechanism was introduced, where the first layer enables the equipment to plan a collision-free path considering the predicted movement of all other equipment and the second layer acts as a last-line-of-defense in view of possible discrepancies between the predicted paths and actual paths.  In view of the results of the case study, it is shown that the MAS is capable of effectively handling the harmonization of various pieces of equipment on the site beyond what is available by the conventional 315-9 LGSs. The combination of LAEWs and DEWs are found to be an efficient approach to deal with collision-free path planning and real-time collision avoidance. The authors are planning to investigate the negotiation between different levels of agents as part of their future work.   References  Beavers, J. E., Moore, J. R., Rinehart, R. and Schriver, W. R. 2006. Crane-related fatalities in the construction industry. Journal of Construction Engineering and Management, 132(9):901-910. Burns, R. L. 2002. Dynamic Safety Envelope For Autonomous-Vehicle Collision Avoidance System. U.S., Patent No. US 6393362. Carbonari, A., Giretti, A. and Naticchia, B. 2011. A proactive system for real-time safety management in construction sites. Automation in Construction, 20(6), 686-698. Chae, S. 2009. Development of warning system for preventing collision accident on construction site. In Proceedings of the 26th Int. Symposium on Automation and Robotics in Construction. Cheng, T., 2013. Automated Safety Analysis of Construction Site Activities Using Spatio-temporal Data, Doctoral dissertation, Georgia Institute of Technology. Dunston, P. S. and Monty, J. 2009. Practices for seamless transmission of design data from design phase to construction equipment operation - a synthesis study, Purdue University, West Lafayette, Indiana. Ferber, J. 1999. Multi-agent systems: an introduction to distributed artificial intelligence. Addison-Wesley, Harlow, UK. Guenther , N. and Salow, H. 2012. Collision avoidance and operator guidance innovating mine vehicle safety. Queensland Mining Industry Health & Safety Conference, Townsville. Hammad, A., El Ammari, K., Langari, S. M., Vahdatikhaki, F., Soltani, M., AlBahnassi, H., Paes, B. 2014. Simulating Macro and Micro Path Planning of Excavation Operations Using Game Engine. Proceedings of the 2014 Winter Simulation Conference, Tolk, S. Y. Diallo, I. O. Ryzhov, L. Yilmaz, S. Buckley, and J. A. Miller (eds.), Savannah, GA. Hammad, A., Vahdatikhaki, F. and Zhang, C. 2013. A novel integrated approach to project-level automated machine control/guidance systems in construction projects. ITcon, 162-181. La Valle, S. M. 2006. Planning Algorithms. Cambridge University Press, New York. Perkinson, C. L., Bayraktar, M. E. and Ahmad, I. 2010. The use of computing technology in highway construction as a total jobsite management tool. Automation in Construction, 19(7), 884-897. Torres, H. N. and Ruiz, M. 2011. Improving Highway Project Delivery. TRB Annual Meeting, Washington, D.C.,  Unity3D, 2015. Unity - Game Engine. Retrieved on 03/26, 2014, from http://unity3d.com/ Vahdatikhaki, F. and Hammad, A. 2014. Framework for near real-time simulation of earthmoving projects using location tracking technologies. Automation in Construction, 42:50–67. Vahdatikhaki, F. and Hammad, A. 2015a. Dynamic Equipment Workspace generation for improving earthwork safety using real-time location system. Advanced Engineering Informatics, (Accepted with modifications). Vahdatikhaki, F. and Hammad, A. 2015b. Risk-based Look-ahead Workspace Generation for Earthwork Equipment Using Near Real-time Simulation. Automation in Construction (Accepted with modifications). Vahdatikhaki, F., Hammad, A. and Siddiqui, H. 2015. Optimization-based Excavator Pose Estimation Using Real-time Location Systems. Automation in Construction (accepted with modifications). Vonderohe, A. P. 2009. Report of status and plans for implementing technologies for design and con-struction in WisDOT, Construction Materials and Support Center. Wang, J. and Razavi, S. N. 2015. Low False Alarm Rate Model for Unsafe-Proximity Detection in Construction. Journal of Computing in Civil Engineering. Wu, H., Tao, J., Li, X., Chi, X., Hua, X., Yang, R., Wang, S., Chen, N. 2013. A location based service approach for collision warning systems in concrete dam construction. Safety Science , 51(1):338-346. Zhang, C. and Hammad, A. 2012. Improving lifting motion planning and re-planning of cranes with consideration for safety and efficiency. Journal of Advance Engineering Informatics, 26(2):396-410. Zolynski, G., Schmidt, D. and Berns, K. 2014. Safety for an Autonomous Bucket Excavator During Typical Landscaping Tasks. New Trends in Medical and Service Robots, 20: 357-368.  315-10  Multi-agent System for Improved Safety and Productivity of Earthwork Equipment Using Real-time Location Systems Faridaddin VahdatikhakiProf. Amin HammadSeied Mohammad Langari2 of 29Introduction• Earthwork refers to a set of operations leading to reshaping ofthe natural surface of earth (de Athayde Prata et al. 2008) andaccounts for:• More than 20% of the total cost of the road building projects(Smith et al. 1996).• A considerable amount of fatalities, i.e. 74 out of 775 only in2012 (BLS 2012).Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions3 of 29Introduction (cont.)• A Location-based Guidance System (LGS) is a systemthat combines a location tracking system and other sensorydata with On-Board Instrumentation (OBI) to performcomplex real-time monitoring of the equipment and providenecessary guidance.GPS Antenna Cabin DisplayControl Device SensorsAMC/G Hardware (Adapted from Leica Geosystems 2013)Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions4 of 29• Current application of LGSs is limited to the machine-levelproductivity improvement.• The high cost of procuring available LGSs limits theiravailability for small and medium size contractors.• Real-time data coming from LGS is not efficiently use toupdate the cycle time of operation.• LGS application for safety is limited to real-time proximity-based object detection and warnings.Problem StatementIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions5 of 29• To enable the fleet-level application of LGS.• To provide a method for improving the pose estimationperformance of low-cost RTLSs for use in LGSs.• To devise a generic framework for Near Real-timeSimulation (NRTS) based on data from LGS.• To develop a mechanism for improving the safety ofearthwork operations using LGS.ObjectivesIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions6 of 29• Operation Management System and Autonomous Excavation(Connected Site, Singh and Cannon 1998, Bradley and Seward 1998, Stentz et al. 1999)• Near-real Time Simulation (NRTS)(Lu et al. 2007, Akhavian and Behzadan 2012, Song and Eldin 2012, Pradhananga and Teizer 2013, Akhavianand Behzadan 2013)• Collision Avoidance using Workspaces(Cheng 2013, Marks and Teizer 2013, Luo et al. 2014, Pradhananga 2014, Wang and Razavi 2015)• MAS Application in Construction Industry(Ren and Anumba 2002, Kim and Russel 2003a and 2003b, Lee and Bernold 2008, Zhang and Hammad2012a).Literature ReviewIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions7 of 29Operator AgentPoseStateConditionSchedule…Operator AgentTeam Coordinator AgentOperator AgentOperator AgentTeam Coordinator AgentGeneral Coordinator AgentTeam Safety ManagementTask AssignmentProgress Monitoring…Project Safety ManagementOperation AssignmentProgress Monitoring…Site State AgentProject Document AgentDesign Document AgentImages are extracted from multiple sources (Stanford Business Mapping, Wikipedia, Google 3D Warehouse, Autodesk)Multi-agent ArchitectureIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions8 of 29Pose/State IdentificationTactical Planning Generation of DEWSafety CheckTask ExecutionLook-ahead Risk CalculationCycle-time CalculationEquipment MonitoringUtility DetectionUpdating DTMTask AssignmentProgress MonitoringNRTSSafety CheckLAEW GenerationTask Rescheduling OA TCAMAS Functions Covered in this Research Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions9 of 29𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 = 𝑀𝑀𝐹𝐹𝐹𝐹 [𝑃𝑃 × �𝑗𝑗=14𝐶𝐶𝑗𝑗𝑟𝑟+ �𝑖𝑖=16𝐷𝐷𝑖𝑖 − 𝑑𝑑𝑖𝑖𝐷𝐷𝑖𝑖+ 𝛼𝛼 − 𝛽𝛽𝛼𝛼+ 𝑅𝑅𝑟𝑟]rCorrected DCCaptured DCrCnSolution spaceAmount of CorrectionCRcal Calculated CRMeasured CRCRmActual DCd1d2d3d5 d6βCRcalCRm   D1 D2 D3 D5 D6 α CRm DC1 DC2 DC3 DC4 Pose EstimationIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions10 of 29   D1 D2 D3 D5 D6 α CRm DC1 DC2 DC3 DC4 D1= 12.5 cmD2= 5.5 cmD3= 18 cmD4= 16 cmD5= 1.5 cmD6= 14 cmα= 14.77˚Pose Estimation (cont.)Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions11 of 29Return zoneHauling zoneLoading zoneDumping ZoneIf the truck is in the dumping zoneit is the closest to the conveyor beltits velocity is zeroFixed Zones for TruckThen the truck is dumpingbucket is relatively stationary its locati n intersects with the truck’s bedExcavatorThen the excavator is dumpingState IdentificationIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions12 of 29Agent-based Safety Management SystemDynamic Equipment Workspace Look-Ahead Equipment WorkspaceIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions13 of 29Swinging StateTraversal State(moving on the tracks)Loading/Dumping StateR3R4R2R5𝐹𝐹𝐹𝐹bLeb Web?⃗?𝑣bDynamic Equipment Workspace (cont.)Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions14 of 29Look-ahead Equipment WorkspaceIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions15 of 29Look-ahead Equipment Workspace (cont.)Risk Map Look-Ahead Equipment WorkspaceIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions16 of 29    Bucket Left Bucket Right Track front Cab Body Track back Crane front Crane back Truck Left Truck Right Ubisense UWBDome CameraEquipment Used in Case StudiesIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions17 of 290204060801000 10 20 30 40 50 60 70Percentage of Data in the Error Range (%)Location Error (cm)Before Correction Correction with All ConstraintsCase Study 1: Error Analysis for Pose0204060801000 20 40 60 80 100 120 140 160 180Orientation Error (degree)81.398.553.528.6Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions18 of 29Case Study 2: Lab TestIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions19 of 29Comparison of captured and corrected dataVideoCase Study 2: Excavator Pose EstimationIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions20 of 29Case Study 2: State IdentificationIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions21 of 29Case Study 2: Dynamic Equipment WorkspaceIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions22 of 29Case Study 2: Dynamic Equipment Workspace (Cont.)0102030405060708090100False Positive Space Savings Compared toCylindrical DEWDetection Clearance Comparedto Cylindrical DEWProposed DEWSymmetric DEWBuffer DEWCylindrical DEWIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions23 of 29Case Study 2: Look-ahead Equipment WorkspaceRisk Map Look-ahead Equipment WorkspaceIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions24 of 29• The feasibility of DEW-based safety monitoring is studied in this case study.• Truck A is hauling the material to the dumping spot and Truck B is returning to the excavation points. Truck BTruck ACase Study 3: Workspaces in Simulated Construction siteSite Layout ScenarioIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions25 of 29Case Study 3: Dynamic Equipment WorkspaceIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions26 of 29• Applying LAEWs for collision-free path planning.Case Study 3: Look-Ahead Equipment WorkspaceExcavator 1 Truck 1Excavator 2Truck 2PipesActual Site Current poses and Initial PathsRisk Map of Excavator  1 Final Path of Excavator 2Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions27 of 29Introduction Problem Statement Objectives Literature Review Proposed Method Case studies ConclusionsCase Study 3: Look-Ahead Equipment Workspace28 of 29• Safety issues and conflict-prone activities are addressedusing a two-layer mechanism that accounts for a widerange of human factors and uncertainties.• The proposed approach, in addition to the demonstratedoperational advantages, can offer benefits at the manageriallevel, allowing managers to make informed decisionsabout the project using real-time data and simulation data.• The MAS structure can offer faster conflict resolution,owing to the faster identification of the problem area andcommunication and negotiation between the agents.ConclusionsIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions30 of 29Task Re-assignmentPath Re-planningDesign Change RequestOperation and Task AssignmentStrategic and Tactical PlanningVisual Guidance to Equipment OperatorsCollecting and Processing RTLS DataEquipment Pose IdentificationEquipment State IdentificationNear Real-time SimulationWorkspace GenerationReporting (Progress Tracking, Safety Warnings, and Delay Notice)PlanningExecution and MonitoringRe-planningThe Scope of the Proposed MethodIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions31 of 29Alt[Equipment in Vicinity ==True] Send Priority and DEW2Send Priority and DEW1Alt[Eq1 has priority==True][Eq2 has priority==True][Else]Introduction Problem Statement Objectives Literature Review Proposed Method Case studies ConclusionsDynamic Equipment Workspace (cont.)Send LocationSend LocationSend LocationSend LocationSend LocationSend LocationAlt[Collision Detected==True]Collision DetectionCollision DetectionStopSend WarningSend WarningCollision DetectionAlt[Collision Detected==True]StopSend WarningCollision DetectionAlt[Collision Detected==True]StopSend WarningEquipment iEquipment 1 Equipment 2Send LocationSend LocationSend LocationTo help protect your privacy, PowerPoint has blocked automatic download of this picture.Send Priorities and DEWsTo help protect your privacy, PowerPoint has blocked automatic download of this picture.WarningStop32 of 29Look-ahead Equipment Workspace (cont.)Introduction Problem Statement Objectives Literature Review Proposed Method Case studies ConclusionsProximity Risks Visibility RisksBased on : (a) Shortest Distance to Equipment (Li)(b) Time to Shortest Distance (ti)based on:Equipment Blind Spot33 of 29Look-ahead Equipment Workspace (cont.)Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions(a) Shortest Distance to Equipment (Li) (c) Time in the Equipment Blind Spots(b) Time to Shortest Distance (ti)Equipment Risk Map34 of 29• Vahdatikhaki F., Hammad A., (2015) “Dynamic Equipment Workspace Generation for Improving the Earthwork Safety Using Real-time Location Systems”, Journal of Advanced Engineering Informatics (Available online).• Vahdatikhaki F., Hammad A., (2015) “Risk-based Look-ahead Workspace Generation for Earthwork Equipment Using Near Real-time Simulation”, Journal of Automation in Construction (accepted with modifications).• Vahdatikhaki F., Hammad A., Siddiqui H., (2015) “Optimization-based Excavator Pose Estimation Using Real-time Location Systems”, Journal of Automation in Construction (Available online).• Vahdatikhaki F., Hammad A., (2014) “Framework for Near Real-Time Simulation of Earthmoving Projects Using Location Tracking Technologies”, Journal of Automation in Construction, vol. 42, pp. 50-67.• Hammad A., Vahdatikhaki F., Cheng Z., (2013) “A Novel Approach to Project-Level Automated Machine Control/Guidance Systems in Construction Projects”, Journal of Information Technology in Construction, vol. 18, pp. 161-181.Publications: Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions35 of 29Introduction Problem Statement Objectives Literature Review Proposed Method Case studies ConclusionsDynamic Equipment Workspace (cont.)ExcavatorStationary States Traversal StatesTruckStationary StatesExcavator should avoid the truckExcavator should avoid the truckTraversal statesTruck should avoid the excavatorEquipment with the loweroperation costs and productivityrate (usually truck) should avoidthe other one36 of 29Look-ahead Equipment Workspace (cont.)Introduction Problem Statement Objectives Literature Review Proposed Method Case studies ConclusionsCurrent and Final Pose of Excavator Motion Path During LoadingMotion Path During Swigging Motion Path During Dumping37 of 29Look-ahead Equipment Workspace (cont.)Introduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusionsθ1θ2θ3θ4𝜃𝜃1𝜃𝜃2𝜃𝜃3𝜃𝜃4Value of DOFjTime𝐹𝐹0 𝐹𝐹1 𝐹𝐹2𝜃𝜃1,0𝜃𝜃1,1& 𝜃𝜃1,2𝜃𝜃3,2& 𝜃𝜃3,3𝜃𝜃4,0& 𝜃𝜃4,1𝜃𝜃2,0𝜃𝜃2,1𝜃𝜃2,2 & 𝜃𝜃2,3𝜃𝜃4,2& 𝜃𝜃4,3𝜃𝜃3,0& 𝜃𝜃3,1a2,0×t0𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋 𝐒𝐒𝐒𝐒𝐋𝐋𝐋𝐋𝐋𝐋 𝐭𝐭𝐋𝐋 𝐓𝐓𝐓𝐓𝐓𝐓𝐓𝐓𝐓𝐓 𝐃𝐃𝐓𝐓𝐃𝐃𝐃𝐃𝐋𝐋𝐋𝐋𝐋𝐋𝜃𝜃1,3Curl the bucketPitch the stickPitch the boomSwing the upper bodyc2,0×t0a1,0×t0 c1,2×t2c2,1×t1a2,1×t1a3,1×t1 c3,1×t1c4,1×t138 of 29𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝑇𝑇𝑇𝑇𝑇𝑇𝐹𝐹 1 = 𝑀𝑀𝐹𝐹𝐹𝐹 [𝑃𝑃 × �𝑗𝑗=14𝐶𝐶𝑗𝑗𝑟𝑟+ �𝑖𝑖=16𝐷𝐷𝑖𝑖 − 𝑑𝑑𝑖𝑖𝐷𝐷𝑖𝑖+ 𝛼𝛼 − 𝛽𝛽𝛼𝛼+ �𝑗𝑗=14𝐹𝐹𝑗𝑗𝐹𝐹𝑚𝑚𝑚𝑚𝑚𝑚+ 𝑅𝑅𝑟𝑟]   D1 D2 D3 D5 D6 α CRm DC1 DC2 DC3 DC4 Data Processor𝑆𝑆𝐹𝐹𝑆𝑆𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝐶𝐶𝑗𝑗 ≤ 𝑟𝑟d1d2 d3d5 d6rβCRcalCR*𝐷𝐷𝐶𝐶2𝑘𝑘−1′𝐷𝐷𝐶𝐶1𝑘𝑘−1′𝐷𝐷𝐶𝐶3𝑘𝑘−1′𝐷𝐷𝐶𝐶4𝑘𝑘−1′S139 of 29Case Study 2: LAEW 024681012141618200 5 10 15 20 25 30 35Dimension of the cell (cm)Computation time per simulated second (s)1.5 S2.5 S3.5 S4.5 S5.5 S40 of 29UWB Sensor 1UWB Sensor 2UWB Sensor 3UWB Sensor 4CameraCase Study 1: LayoutIntroduction Problem Statement Objectives Literature Review Proposed Method Case studies Conclusions41 of 29• The conventional simulation models are based on thestatistical data• Near-real Time Simulation (NRTS) (Lu et al. 2007, Akhavian andBehzadan 2012, Song and Eldin 2012, Pradhananga and Teizer 2013,Akhavian and Behzadan 2013)Near Real-time Simulation42 of 29• Autonomous excavation (Singh and Cannon 1998, Bradley andSeward 1998, Stentz et al. 1999)• Automated compaction (Kaufmann and Anderegg 2008)LGS in Research    Dig face Right sensor scan plane Truck Left sensor scan plane    Right scanner  Left scanner      Dig Region Range Sensors             Floor h Strategic Planning Tactical Planning Dump Planning43 of 29• Automated Machine Control/Guidance (AMC/G)– Input Data– Hardware– Tracking TechnologiesDTM (Stanford Business Mapping 2013)GPS Antenna Total Station Prism Cabin Display Control DeviceAMC/G Hardware (Leica Geosystems 2013)LGS in Industry44 of 29   (a) Excavation (b) Grading (c) Milling     (d) Paving (f) Curb installation (e) Barrier installation  • AMC/G–supported operations (Singh 2010)LGS in Industry (cont.)45 of 29Project Monitoring and Control• Monitoring and control of the project is needed to containdelays and variations in the earthwork projectsPerformance Control Cycle (Navon 2007)46 of 29• Conventional methods for the measurement of ProjectPerformance Indicators (PPI) are manual.• New methods use variety of data capturing technologies:– RFID (Motamedi and Hammad 2009, Montaser and Moselhi 2012)– GPS (Alshibani and Moselhi 2007, Perkinson et al. 2010)– UWB (Teizer et al. 2008, Zhang et al. 2011a)– Computer Vision (Rezazadeh Azar and McCabe 2011, Golparvar-Fardet al. 2013)Project Monitoring and Control (cont.)47 of 29Safety of Earthwork Operations• Safety systems using different types of technologies:– RFID (Chae and Yoshida 2010, Yang et al. 2012)– GPS (Oloufa et al. 2003, Sun et al. 2010, Wu et al. 2013)– UWB (Teizer et al. 2008, Carbonari et al. 2011, Zhang and Hammad 2012a)– Computer Vision (Talmaki et al. 2010, Chi and Caldas 2011)48 of 29Multi-agent Systems (MASs)• An agent is defined as an entity situated in an environmentwith the capability to form a perception of the environmentand act upon it (Russell and Norvig 2003) Environment Percept Actions SensorActuatorAgent ? Agents’ interaction with Environment (Russell and Norvig 2003)49 of 29MAS areas of Application• Enhanced decision making: agents are used to partiallyor completely substitute humans in decision making• Simulation of complex process: the prevalent type ofavailable information is the behavior of actors in a process50 of 29MAS in Construction• Claim negotiation (Ren and Anumba 2002)• Task Planning for earthwork (Kim and Russel 2003a and 2003b)• Data communication (Lee and Bernold 2008)• Collision avoidance and path planning (Zhang and Hammad 2012a)51 of 29Cade Study on Construction Site in Vancouver52 of 29Cade Study on Construction Site in Vancouver UWB Covered AreaSensor 2Sensor 1Sensor 320 m36.5 m53 of 29Cade Study on Construction Site in Vancouver54 of 29Cade Study on Construction Site in Vancouver

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.52660.1-0076351/manifest

Comment

Related Items