Network-Level Safety Prediction Models for Long-range Transportation Planning
Date
2019-09-25
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This research develops a holistic, proactive approach for integrating traffic safety into transportation planning to effectively evaluate the network-level safety impacts of transportation plans and policies. The research enhances the regional transportation model (RTM) framework by incorporating a network-based collision prediction model (NCPM) as a fifth step in the traditional four-step RTM modelling structure, allowing the model to predict the number of collisions on major and local roads at the planning stage. In addition to traditional estimates of traffic demand, the developed integrated RTM-NCPM framework also predicts the number of collisions for base and future planning horizons. The obtained results show promise in their ability to replicate observed collision frequencies and types in the base year, pointing to the robustness of the proposed framework. The approach developed in this thesis was rigorously evaluated for various future scenarios using the City of Calgary’s RTM as a case study. A sensitivity analysis was conducted to test model performance under several congestion pricing and transit fare policies. The RTM-NCPM framework showed a decrease in Property Damage only (PDO) and fatal and injury (FI) collisions of 13% and 6%, respectively, on local roads and of 8.5% and 8.6%, respectively, on major roads when fuel price was doubled. PDO and FI collisions decreased by 8% and 5%, respectively, on local roads and by 4.1% and 3.6%, respectively, on major roads when parking costs were doubled. When transit fares were reduced by half, PDO and FI collisions decreased by 5% and 2%, respectively, on local roads and by 3.8% and 3.4%, respectively, on major roads. The research develops a system dynamics (SD) modelling approach that models future impacts of autonomous vehicles (AVs) on the number of collisions. This approach captures the complex interactions resulting from the introduction of AVs while taking inputs from the RTM model. The developed model was used to examine the effectiveness of potential future AV-related policies and scenarios in reducing collisions. These scenarios and policy changes include: higher AV penetration rates, shared AVs with higher passenger occupancy, and improvements to sensing and communication technologies. The SD model for a scenario with shared autonomous vehicles (SAVs) with an average occupancy rate of 1.4 showed an increase in collisions through the year 2060, followed by a decrease in collisions. This scenario’s results suggested that the extreme assumption regarding the highest level of SAV mode share, with an average occupancy rate of 12 and a total shift of AVs to SAVs, would result in the lowest number of collisions compared to other scenarios.
Description
Keywords
Collision Prediction Models, Autonomous Vehicle, System Dynamics, Transportation Planning
Citation
Farhan, A. (2019). Network-Level Safety Prediction Models for Long-range Transportation Planning (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.