3 Presentations

• 
  03:30 PM - 03:55 PM

  Tradeoffs between optimal and equitable bicycle infrastructure placement in Toronto

  • Madeleine Bonsma-Fisher, presenter, University of Toronto
  • Bo Lin, University of Toronto
  • Shoshanna Saxe, University of Toronto
  • Timothy C.Y. Chan, University of Toronto

  Cycling is an affordable, healthy, and sustainable method of travel, but access to important destinations on low-stress (a proxy for safe) cycling routes in cities like Toronto is both limited and unevenly distributed. When proposed infrastructure locations are optimized to provide the highest average access to opportunities, marginalized groups and locations may be further left behind since the greatest gains to network connectivity come from expansions in already-dense network areas.

  To understand how to design spatially equitable bicycle infrastructure, using Toronto as a case study, we develop an optimization model to identify road segments where new cycling infrastructure will provide the largest increase in access to opportunities. We find that optimizing accessibility to opportunities in pre-amalgamation regions of Toronto instead of city-wide leads to an infrastructure plan that is more spatially distributed with more service for equity seeking groups but lower average city-wide accessibility gains, indicating the presence of an equity-efficiency tradeoff in the location of infrastructure. We also find that infrastructure projects that maximize a region's accessibility to opportunities are often located outside that region which could challenge political perceptions of infrastructure benefits. These results inform planning and shed light on spatial equity tradeoffs in deciding where to add cycling infrastructure.
• 
  03:55 PM - 04:20 PM

  Data-driven hub network design for ridesharing

  • Gita Taherkhani, Loyola University Chicago
  • Bissan Ghaddar, Ivey Business School
  • Sibel Alumur, presenter, University of Waterloo
  • Timmy Hsu, Hong Kong University

  This paper studies the design of ridesharing hub networks to promote shared transportation. Given a set of passenger trips in an urban area, the problem is to determine the origins and destinations of a fixed number of ridesharing connections to maximize the potential users of the system. The problem is formulated as a maximal covering hub arc location model and solved to optimality using Benders decomposition. Several algorithmic enhancements, including using reduction tests to eliminate variables and adding multiple Pareto-optimal cuts, are proposed to improve the convergence of the Benders decomposition algorithm. Additionally, two data-driven clustering-based methodologies are adapted and implemented to compare with the solutions of the optimization model. All methodologies are tested using the New York City taxi trip data. Several computational experiments are conducted to compare optimization and data-driven approaches under key performance metrics that include the total number of commuters that use the ridesharing system, the percentage of satisfied trips by ridesharing, the utilization of ridesharing trips, and the driving and walking distances. The results from the optimization model yield more satisfied trips through the ridesharing system and also a more balanced utilization of the ridesharing connections compared with the results obtained from either of the clustering-based methodologies. Our results show that operating even a few ridesharing connections in the city of New York (around 20 connections) can result in reducing a significant number of individual rides and the total driving distance of the commuters without incurring a walking distance above 1 km per passenger.
• 
  04:20 PM - 04:45 PM

  How to Predict the Intention to Buy Electric Vehicles - An Interpretable Machine Learning Method with A Metaheuristic Approach

  • Hamed Naseri, Polytechnique Montréal
  • Owen Waygood, Polytechnique Montréal
  • Bobin Wang, presenter, Université Laval
  • Zachary Patterson, Concordia University

  To address the problem of climate change emissions from the transport sector, many countries are promoting electric vehicles (EVs). In order to support such efforts, it is essential to know what influences the choice of an EV over a traditional Internal Combustion Engine Vehicle (ICEV). However, two key issues were addressed with respect to its application. First, a practical question related to how to best split the training and testing data was examined. Second, machine learning techniques are hard to interpret. To address this, a discrete choice experiment was developed, and 2015 valid responses were gathered from Canadian adults with a driver’s license. A new technique based on Coyote Optimization Algorithm (COA) is developed that automatically determines the split that leads to the greatest prediction accuracy. eXtreme Gradient Boosting and Accumulated Local Effect were employed to determine the strength and direction of influence of the variable. The policy-relevant results of the analysis found that an individual’s Climate Change-Stage of Change (CC-SoC) and the price ratio of EVs to ICEVs are the most important direct influences.