Open Collections

Abstract

Cycling is promoted in cities worldwide due to its benefits over motor vehicle travel—such as reduced emissions, alleviated traffic congestion, and improved health—as part of efforts to achieve ambitious active transportation goals. To support this growth, transportation professionals need behaviourally-grounded tools to design infrastructure and policy, which require understanding cycling behaviour and preferences across the population. Cycling speed is an important and yet poorly understood aspect of travel behaviour and crucial in designing safe facilities and planning accessible networks. This dissertation aims to develop new understanding of the strength of preferences for speed and energy expenditure during utilitarian cycling based on observed behaviour. Behavioural analysis of high-resolution GPS data from cycling trips requires first developing new analytical tools to process data and reliably extract speed- and energy-related features. Hence, three early chapters evaluate methods to identify trips in active travel GPS data (Chapter 2), map-match trips to a street network (Chapter 3), and extract road grade information for cycling trips (Chapter 4). Results indicate that machine-learning algorithms outperform heuristic algorithms in trip identification. Existing map-matching algorithms can be improved for cycling trips by modifying the link cost function to determine the most likely route and allowing wrong-way travel. Additionally, road grade profiles can be extracted from a combination of LIDAR cloud point data, high-resolution elevation models, and crowdsourced GPS data. Chapter 5 proposes a unique cyclist typology to incorporate a range of self-reported preferences and behaviours into a probabilistic dichotomy of Dedicated and Casual riders, as an influencing factor for speed preferences. Chapter 6 develops a method to identify cruising events in GPS data accounting for temporal dependency between observations. Chapter 7 uses these cruising events in utility-based speed choice model to characterize the trade-offs cyclists make between energy expenditure and travel time, depending on a range of personal, trip, and contextual factors. This research creates new knowledge of cyclist preferences that is crucial for cycling speed choice modelling. The findings also enable the incorporation of speed and energy preferences into route choice models, which can improve their behavioural grounding and representation of heterogeneity in the cycling population.