University of British Columbia
Understanding behaviour of vulnerable road users in mixed and less organized traffic environments
Abstract
dc:descriptionVulnerable road users (VRUs), including pedestrians, motorized two-wheelers (2Ws), and three-wheelers (3Ws), account for a disproportionate share of global traffic fatalities, particularly in mixed and less-organized traffic environments where road users interact without strict lane discipline or formal control. Traditional crash-based safety assessments are reactive and insufficient for capturing the behavioural mechanisms that govern safety-critical interactions. This thesis uses behaviour-aware proactive safety frameworks that both quantify crash risk and interpret decision-making mechanisms for VRU-vehicle interactions. First, a probabilistic Extreme Value Theory (EVT) framework is developed to estimate crash risk from traffic conflicts extracted using trajectory data. A Full Bayesian non-stationary Peak Over Threshold (POT) approach is applied to pedestrian-vehicle, 2W-vehicle, and 3W-vehicle interactions. Microscopic behavioural variables, such as speeds and minimum distances, are incorporated as covariates to model risk. The framework enables mode-specific crash risk estimation and severity comparisons across different conflict types, identifying distinct behavioural tendencies in high-risk interactions. Second, a Multi-Agent Adversarial Inverse Reinforcement Learning (MAAIRL) framework is implemented to infer latent reward structures and the conflict-negotiation strategies adopted by road users, based on conflict trajectories. Using Markov Games (MGs), the models capture interaction-level behavioural preferences without predefined rules, enabling the realistic simulation of conflict-resolution strategies. The learned policies are evaluated through accuracy metrics and used for a simulation-based crossing conflict analysis for road users. The findings demonstrate that crash risk in mixed and less-organized traffic is strongly behaviour-driven, with distinct vulnerability patterns across VRU categories. By linking microscopic behavioural dynamics to probabilistic crash risk and learned decision-making policies, the proposed frameworks enable targeted, mode-specific safety interventions. These insights support behaviour-driven safety analysis and offer practical implications for safety systems, infrastructure design, driver/rider assistance systems, and intelligent transportation systems in mixed and less-organized traffic environments.
Degree
thesis:*- Name thesis:degree_name
- Master of Applied Science - MASc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Civil Engineering
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Goyal, Sambhav
- Contributors dc:contributor
-
- Sayed, Tarek, 1965-
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/94939