Decoupling Passenger Flows for Improved Load Prediction - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

Decoupling Passenger Flows for Improved Load Prediction

Résumé

This paper continues our work on perturbation analysis of multimodal transportation networks (TNs) by means of a stochastic hybrid automaton (SHA) model. We focus here on the approximate computation , in particular on the major bottleneck consisting in the high dimensionality of systems of stochastic differential balance equations (SDEs) that define the continuous passenger-flow dynamics in the different modes of the SHA model. In fact, for every pair of a mode and a station, one system of coupled SDEs relates the passenger loads of all discrete points such as platforms considered in this station, and all vehicles docked to it, to the passenger flows in between. In general, such an SDE system has many dimensions, which makes its numerical computation and thus the approximate computation of the SHA model intractable. We show how these systems can be canonically replaced by lower-dimensional ones, by decoupling the passenger flows inside every mode from one another. We prove that the resulting approximating passenger-flow dynamics converges to the original one, if the replacing set of balance equations set up for all decoupled passenger flows communicate their results among each other in vanishing time intervals.
Fichier principal
Vignette du fichier
Paper3.pdf (315.26 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01330136 , version 1 (10-06-2016)

Identifiants

  • HAL Id : hal-01330136 , version 1

Citer

Stefan Haar, Simon Theissing. Decoupling Passenger Flows for Improved Load Prediction. 13th International Conference on Quantitative Evaluation of SysTems (QEST 2016) , Aug 2016, Québec City, Canada. ⟨hal-01330136⟩
251 Consultations
116 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More