4 présentations

• 
  15h45 - 16h10

  Bayesian Calibration for Travel Demand Calibration of Urban Traffic Simulators via Traffic-Theoretic Priors

  • Seung Min Baik, prés., HEC
  • Giovanni De Nunzio, IFP Energies nouvelles
  • Azise Oumar Diallo, IFP Energies nouvelles
  • Delphine Sinoquet, IFP Energies nouvelles
  • Mohamed Reda El Amri, IFP Energies nouvelles
  • Carolina Osorio, HEC Montréal & GERAD

  Calibrating travel demand in urban traffic simulators is often an underdetermined problem. This leads to multiple origin-destination (OD) matrices producing nearly identical simulation outputs, such as traffic counts on certain links of interest. The most popular calibration methods typically aim to estimate a single OD matrix by minimizing discrepancies between simulated and observed traffic counts, but they do not account for the inherent uncertainty in the calibration process. Bayesian calibration (BC) is a general probabilistic framework designed to address this issue by inferring a distribution over plausible calibration parameters rather than identifying a single deterministic value. When applied to travel demand calibration, BC provides a principled approach for uncertainty quantification. However, it struggles to capture the complex dynamics of the transportation network if a conventional Gaussian process (GP) surrogate is used as is without domain-specific modifications. To alleviate this, we incorporate the traffic theory into the design of a more context-aware prior for the GP, improving its modeling performance even with a limited number of simulated instances. This enables a calibration process with improved accuracy and interpretability. We validate our method by estimating the demand for a case study of the metropolitan area of Lyon, France, using real-world traffic data, demonstrating its ability to quantify uncertainty. Our findings highlight the potential of integrating the BC framework into travel demand calibration in urban transportation systems.

• 
  16h10 - 16h35

  ROUTE CHOICE USING CROWD-GENERATED TRAVEL TIME INFORMATION

  • Elham Heydarigharaei, prés.,
  • Mohammadamir Ahmadian ,
  • Sina Bahrami,
  • Mehdi Nourinejad,

  Smartphones and onboard vehicle sensors are increasingly used to gather crowd-generated travel data (CTD), providing a cost-effective alternative to traditional data collection methods such as stationary sensors and cameras. However, the reliability of CTD depends on travelers’ willingness to share their data. This paper examines how CTD can be strategically utilized to optimize network-wide travel times while promoting public access to travel information and preserving user privacy in sharing travel data. Specifically, we highlight the role of two factors in decision-making and the accuracy of the travel time estimation: the contribution ratio, which represents the proportion of travelers sharing their data, and the observation window, which defines the period over which CTD is collected and analyzed. We consider a transportation network with two routes: one with fixed travel times, referred to as the safe road, and another with variable, stochastic travel times, called the random road, which is modeled as an M/M/1 queue. The expected travel time on the random road is estimated using the central limit theorem, which is applied to a sample of CTD collected within the observation window. A choice model is proposed to calculate the probability of travelers selecting each road. The study examines how variations in contribution ratios and observation windows influence route choices and overall travel times. An optimal contribution ratio is identified that minimizes overall travel time in a fixed-demand setting. Furthermore, it is demonstrated that the optimal proportion of travelers selecting the random road increases with the square root of the travel time on the safe road. Additionally, the impact of risk on route choice is explored by incorporating a risk aversion function linked to the contribution ratio. To tackle this risk, a scenario is studied where travelers can pay to access real-time travel information on the random road. We establish a threshold on the contribution ratio that determines when travelers shift from paying the access cost to using free CTD, which is a function of the access cost and the variability of travel time on the random road. Finally, it is demonstrated that introducing a reward for travelers who share their data can lead to a further reduction in overall travel time compared to the base model, where the reward is zero.

• 
  16h35 - 17h00

  A Data-Driven Platform for Visualizing and Analyzing Public Transit Schedule Deviations

  • emna boudabbous, prés.,
  • Julio Montecinos, Polytechnique Montréal

  This paper presents an automated approach for classifying and visualizing delays in public transit systems. The
  approach is applied to the Montreal public transit system, offering two key contributions: (1) the integration of spatial analysis techniques leveraging the Hexagonal Hierarchical Spatial Index (H3) system. (2) the combination of H3’s multi-resolution capabilities with KeplerGL’s visualization framework enables the identification of spatial patterns across all operational bus routes in Montreal of both systematic and stochastic deviations. Our approach reveals significant schedule deviations across Montreal’s transit network, including early and late arrivals. The approach we developed
  has the potential to provide valuable insights and guidance for transit optimization and urban planning, particularly in identifying and addressing delay-prone areas and understanding the spatial propagation of delays throughout the Montreal network.

• 
  17h00 - 17h25

  Bayesian Inference of Time-Varying Origin-Destination Matrices from Boarding and Alighting Counts for Transit Services

  • Xiaoxu Chen, prés., McGill University
  • Zhanhong Cheng, McGill University
  • Lijun Sun, McGill University

  Origin-destination (OD) demand matrices are crucial for transit agencies to design and operate transit systems. This paper presents a novel temporal Bayesian model designed to estimate transit OD matrices at the individual bus-journey level from boarding/alighting counts at bus stops. Our approach begins by modeling the number of alighting passengers at subsequent bus stops, given a boarding stop, through a multinomial distribution parameterized by alighting probabilities. Given the large scale of the problem, we generate alighting probabilities with a latent variable matrix and factorize it into a mapping matrix and a temporal matrix, thereby substantially reducing the number of parameters. To further encode a temporally-smooth structure in the parameters, we impose a Gaussian process prior on the columns of the temporal factor matrix. For model inference, we develop a two-stage algorithm with the Markov chain Monte Carlo (MCMC) method. In the first stage, latent OD matrices are sampled conditional on model parameters using a Metropolis-Hastings sampling algorithm with a Markov model-based proposal distribution. In the second stage, we sample model parameters conditional on latent OD matrices using slice and elliptical slice sampling algorithms. We assess the proposed model using real-world data collected from three bus routes with varying numbers of stops, and the results demonstrate that our model achieves accurate posterior mean estimation and outperforms the widely used iterative proportional fitting (IPF) method. Additionally, our model can provide uncertainty quantification for the OD demand matrices, thus benefiting many downstream planning/operational tasks that require robust decisions.