4 présentations

• 
  14h45 - 15h07

  Modeling and Optimization of Charging Infrastructure in Bike-Sharing Networks

  • Ghazaleh Mohseni, prés.,
  • Mehdi Nourinejad, York University
  • Peter Park, York University

  The increasing popularity of e-bike sharing systems has introduced new challenges in managing bike availability and optimizing charging infrastructure. Efficient placement of chargers improves service consistency, minimizes downtime, and maximizes ridership. This paper proposes the problem of placing P chargers in a micromobility network with electric bikes. The proposed model introduces a Markovian system to track system state changes in terms of the number of bikes and their state-of-charge at each station. The state transitions consider bike returns, rentals, and charging events, introducing a complex transition mechanism. We identify an equilibrium by deriving the steady-state probabilities and key performance metrics that describe the system’s behavior. Subsequently, we develop a heuristic optimization model to determine the optimal locations for chargers across the network. The optimization aims to maximize ridership using the outcomes of the steady-state model. The proposed framework enhances system reliability, supports sustainable urban transportation planning, and provides guidance for planners and operators in optimizing e-bike infrastructure to meet the growing demands.

• 
  15h07 - 15h29

  Evaluating Drone-Delivery Efficiency in Different Urban Settings Using Graph Neural Networks (GNNs).

  • Bahar D Viniche, prés., York University
  • Opher Baron, Rotman School of Management, University of Toronto
  • Mehdi Nourinejad, York University

  Last-mile delivery is the time-sensitive and costly leg of the logistic process, necessitating innovative solutions to optimize deliveries in urban and suburban areas. Drone delivery improves last-mile logistics by reducing delivery time, decreasing traffic congestion in dense urban regions, and ensuring timely deliveries to remote locations. It is, therefore, essential to evaluate the effectiveness of drone delivery across diverse urban settings and scenarios. We employ Graph Neural Networks (GNNs) to evaluate the effectiveness of drone delivery in different urban settings. GNNs are powerful framework for applying deep learning to graph-structured data, such as city networks. We developed and leveraged the "Drone Sidekick Tool," an interactive platform designed to analyze and collect data on travel distance, time, and environmental factors associated with drone implementation. This tool, coupled with our algorithmic approach, provides comprehensive insights into the potential of drone-assisted deliveries and supports the development of more efficient and adaptable solutions.

• 
  15h29 - 15h51

  Charging Station Location and Sizing for Electric Vehicles Under Congestion

  • Kinay Omer Burak, Amazon
  • Fatma Gzara, University of Waterloo
  • Sibel Alumur, prés., University of Waterloo

  This paper studies the problem of determining the strategic location of charging stations and their capacity levels under stochastic electric vehicle flows and charging times taking into account the route choice response of users. A bilevel optimization model is developed where the network planner (leader) minimizes infrastructure costs while meeting probabilistic service requirements for waiting times to charge. Electric vehicle users (followers) minimize their route lengths, affecting charging demand and wait times. The bilevel problem is reduced to a single-level mixed-integer model when stations operate as M/M/c queues and users cooperate. A decomposition-based solution method is developed using a logic-based Benders algorithm. Computational experiments on benchmark and real-world highway networks analyze the impact of route choice, service requirements, and deviation tolerance on decisions.

• 
  15h51 - 16h13

  Solving Extreme-Scale EV Charger Location-Allocation Problems using High-Performance Parallel Computing

  • Waquar Kaleem, Pennsylvania State University
  • Taner Cokyasar, Argonne National Laboratory
  • Jeffrey Larson, Argonne National Laboratory
  • Omer Verbas, Argonne National Laboratory
  • Tanveer Hossain Bhuiyan, University of Texas at San Antonio
  • Anirudh Subramanyam, prés., Pennsylvania State University

  This talk addresses electric vehicle (EV) charger location and allocation problems for last-mile delivery operations. The goal is to decide where to open charging stations and how many chargers of each type to install in each station, subject to budgetary and waiting-time constraints. We formulate the problem as a mixed-integer quadratic program, where each station–charger pair is modeled as a multi-server queue with stochastic Poisson arrivals. The model is extremely large with millions of variables and constraints for a typical metropolitan area and even the relaxed model cannot be loaded in solver memory, let alone solved to optimality. To address this challenge, we develop a Lagrangian dual decomposition framework that decomposes the problem by station and leverages parallelization on high-performance computing systems where each subproblem is solved using a cutting plane method. We also develop a three-step rounding heuristic to transform the fractional subproblem solutions into feasible integral solutions. Computational experiments on data from the Chicago metropolitan area with hundreds of thousands of households and thousands of candidate stations show that our approach produces high-quality solutions in cases where existing methods cannot even load the model in memory. We also analyze various policy scenarios, demonstrating that combining existing depots with newly built stations under multi-agency collaboration substantially reduces costs and congestion. These findings offer a scalable and efficient framework for developing sustainable large-scale EV charging networks.