An Equilibrium Model for Schedule-Based Transit Networks with Hard Vehicle Capacities

TL;DR

This paper develops a model for schedule-based transit networks incorporating hard vehicle capacities, characterizes equilibria, and proposes algorithms for computing them, with practical testing on real-world data.

Contribution

It introduces a quasi-variational inequality framework for modeling user equilibria with vehicle capacities and provides algorithms for single and multi-commodity cases.

Findings

Existence of equilibria proven for fixed departure times.

Polynomial time algorithm for single-commodity fixed departure times.

NP-hardness of equilibrium existence in multi-commodity with departure time choice.

Abstract

Modelling passenger assignments in public transport networks is a fundamental task for city planners, especially when deliberating network infrastructure decisions. A key aspect of a realistic model is to integrate passengers' selfish routing behaviour under limited vehicle capacities. We formulate a side-constrained user equilibrium model in a schedule-based transit network, where passengers are modelled via a continuum of non-atomic agents that travel from their origin to their destination. An agent's route may comprise several rides along given lines, each using vehicles with hard loading capacities. We give a characterization of (side-constrained) user equilibria via a quasi-variational inequality and prove their existence for fixed departure times by generalizing a well-known result of Bernstein and Smith (Transp. Sci., 1994). We further derive a polynomial time algorithm for single-commodity instances with fixed departure times. For the multi-commodity case with departure time choice, we show that deciding whether an equilibrium exists is NP-hard, and we devise an exponential-time algorithm that computes an equilibrium if it exists, and signals non-existence otherwise. Using our quasi-variational characterization, we formulate a heuristic for computing multi-commodity user equilibria in practice, which is tested on multiple real-world instances. In terms of social cost, the computed user-equilibria are quite efficient compared to a system optimum.