Babak Mehran

Demand Responsive Transit (DRT) systems rely on effective spatial zoning to balance service quality, operational efficiency, and cost control. This is particularly relevant when serving user groups with heightened sensitivity to service reliability such as older adults. Designing optimal DRT zones is a highly complex combinatorial problem as the number of possible zoning configurations grows exponentially with the number of spatial units, and must satisfy operational and spatial feasibility constraints. Traditional evolutionary algorithms, while widely used for DRT zoning, rely on static search operators that do not learn from past search outcomes or prior infeasible actions, limiting both the efficiency and constraint awareness of the search process. This study proposes a novel Deep Reinforcement Learning–Genetic Algorithm (DRL–GA) framework for optimizing elderly-oriented DRT zoning. The framework embeds a Double Deep Q-Network (DDQN) within a genetic algorithm to guide zoning modifications through action selection based on reinforcement learning. The agent observes current zoning configurations, proposes boundary reassignment actions, and receives rewards based on reductions in a multi-objective cost function capturing passenger waiting time, vehicle-kilometers traveled, and rejected demand. Connectivity constraints are explicitly enforced to ensure spatially contiguous service zones. The proposed approach is evaluated using a large-scale real-world case study in Winnipeg, Canada, based on elderly travel demand estimated from real life transit boarding data captured over a period of one year. Results demonstrate that the DRL–GA framework consistently achieves comparable or better zoning solutions than a GA-inspired baseline, while reducing computational time by approximately 15%, and increasing the proportion of cost-improving actions. Seasonal experiments further show that the framework adapts zoning structures to winter and summer demand patterns, achieving over 70% reductions in waiting time and travel time relative to a single-zone scenario.

Understanding how demographic attributes influence risky driving and support for traffic safety laws is essential for developing targeted regulations. This study analyzes self-reported data from Canadian drivers to identify high-risk groups using accident history and acceptance of unsafe behaviors. Two cases are examined: Case 1 focuses on accidents and demerit points; Case 2 focuses on acceptance of risky behaviors. The analysis involves k-means clustering to classify risk groups, factor analysis to group safety regulations into three categories—speeding, distracted/intoxicated driving, and red-light violations—and logistic regression to explore demographic associations. Key findings show that driving experience, income, and region influence risk in Case 1, while vehicle size, driving frequency, gender, age, and income are significant in Case 2. Senior drivers tend to support stricter safety laws opposing distracted and intoxicated driving, over-speeding, and red-light violation. The study’s results can inform targeted driver education, address high-risk groups, and enhance traffic regulation policies.

This study attempts to develop (i) truck safety performance functions (SPFs), and (ii) hazard-specific crash modification factors (CMFs), for cold-region rural highways. Police-reported truck-involved crashes on rural highway segments of Alberta, Canada, were used to develop truck SPFs for four crash severity levels: total, fatal, personal injury (PI), and property damage only. Three settings of the Poisson–Tweedie regression modeling approach representing Poisson, geometric Poisson, negative binomial distributions were used to develop truck SPFs; the negative binomial distribution was deemed as the most appropriate distribution to model truck-involved crashes for all crash severity levels. The CMF for poor visibility (CMF = 1.5) suggests that poor visibility increases PI-type truck-involved crashes on rural two-lane two-way highway segments by 50% as compared to the number of such crashes attributed to crash causes other than transportation hazards. Road safety researchers may adopt the methodology to effectively rank hazard risks to highway freight transportation systems.

The availability of both boarding and alighting counts is crucial for transit planning, as it allows for the estimation of Origin-Destination patterns that reflect demand. However, automatic data collection methods often provide only boarding counts. This paper investigates the use of transit symmetry to address the lack of alighting counts by analyzing cluster-level symmetry, which captures the dynamics of trip exchanges between nearby stops to achieve balanced ingoing and outgoing flows. An optimization problem is formulated to maximize symmetry across clusters, evaluating the effectiveness of clustering based on stop location using k-means. A novel deep reinforcement learning algorithm utilizing the Double Deep Q-Network framework is proposed to identify optimal clusters, demonstrating an 8% improvement in symmetry optimization over k-means. Notably, k-means clusters still achieve an average symmetry level of 62% when accounting for geographical proximity and walking distance limits. Through the proposed clusters, transit planners can better understand passenger flow dynamics and predict alighting counts with incomplete data.

Passenger boarding and alighting data are critical for understanding travel patterns in transit systems, yet many systems only record boarding, making it difficult to track alighting. This study introduces a method to estimate alighting counts from boarding data using transit symmetry at a cluster level. Bus stops are grouped into clusters, assuming equal inflow and outflow, and a gravity model is calibrated to estimate an origin–destination (OD) matrix. Using data from three bus routes in Regina, Canada (2017), the method achieved a symmetry ratio of 60% across clusters and a mean structural similarity measure score of 0.84, validating the accuracy of the OD estimates. This approach provides a practical solution to the challenge of missing alighting data. While effective for single-mode transit systems, future research will expand its application to multi-modal systems and address factors like seasonal and land-use variations, offering a cost-effective alternative for OD estimation.

The present study demonstrated the potential applications of reliability models for crash analysis of a large highway network. Specifically, three major outcomes of reliability models were investigated: temporal distributions of crashes, reliability score, and expected number of crashes, using 20-year data (2001–2020) of crashes recorded on the Saskatchewan highway network. A series of reliability models were developed for crashes by crash severity, vehicle involvement, and highway type. First, the temporal distributions of crashes on each segment were fit to a statistical distribution. Second, the reliability scores were used to rank the high crash risk segments. Third, the mean expected crash frequency was used to develop network-wide safety performance functions for total and fatal crashes in urban and rural highway segments using Poisson–Tweedie (PTw) regression models. The developed PTw models showed that the presence of trucks in the traffic composition has a significant effect on crash frequency, especially for urban highway segments.