Development of a Dynamic Transport Simulator for Policy Evaluation: Applications to Ride-Sharing and Low Emission Zone in Paris

Abstract

This thesis investigates the use of dynamic transport simulations to evaluate urban mobility policies, focusing on ride-sharing and Low Emission Zones in the Paris metropolitan area. It relies on two transport simulators: the existing METROPOLIS1 and METROPOLIS2, a novel simulator developed as part of this PhD. Both simulators employ discrete-choice theory to estimate responses of commuters to policy changes and assess the impact on utility. Using a mesoscopic framework, the simulators efficiently model road congestion at the city and regional scales. METROPOLIS2 builds upon its predecessor, enabling the evaluation of a broader range of policies, such as vehicle-based restrictions, with enhanced efficiency and accuracy. The first chapter explores a ride-sharing scheme with inflexible drivers who maintain fixed departure times and routes regardless of whether they carry passengers. METROPOLIS1 is used to model the departure time and route chosen by each commuter, as well as the resulting congestion levels. The optimal matching of drivers with passengers, as well as the pick-up and drop-off points, are obtained by solving an integer linear programming problem. Simulations of Paris’s morning commute demonstrate that, even with low participation, ride-sharing can reduce congestion, fuel consumption, and CO2 emissions. Additional benefits can be achieved by increasing vehicle capacity or offering monetary incentives, without compromising driver inflexibility. The second chapter introduces METROPOLIS2, a mesoscopic agent-based transport simulator capable of modeling travel decisions (mode, departure time, and route) based on discrete-choice theory within a dynamic, continuous-time framework. The simulator improves upon its predecessor METROPOLIS1 by incorporating trip chaining, multiple vehicle types, greater flexibility in utility specification, etc. Its efficiency and accuracy are validated through two case studies: replicating analytical results from the standard single-road bottleneck model and demonstrating superior speed and convergence compared to its predecessor on a large-scale scenario of the Paris region. The third chapter applies METROPOLIS2 to evaluate the Low Emission Zone (LEZ) in Paris. Open data are used to generate a synthetic population and road network, while machine-learning techniques (such as Lasso regression and Bayesian optimization) calibrated the simulation to replicate observed travel times and behaviors. The analysis assesses the LEZ’s effects on air quality, congestion and inequalities, highlighting benefits for city-center residents but revealing potential disadvantages for suburban populations dependent on older vehicles.

Keywords

@phdthesis{Javaudin2024,
  author  = "Lucas Javaudin",
  title   = "Development of a Dynamic Transport Simulator for Policy Evaluation: Applications to Ride-Sharing and Low Emission Zone in Paris",
  school  = "CY Cergy Paris Université",
  year    = "2024",
  type    = "Phd Thesis",
  address = "Cergy, France",
}