Online route choice modeling for Mobility-as-a-Service networks with non-separable, congestible link capacity effects

TL;DR

This paper introduces a novel route choice model for MaaS networks that accounts for non-separable, flow-dependent link capacities, improving the understanding of congestion effects on traveler utility and system performance.

Contribution

It develops a new concept of congestible capacity with an offline-online estimation method, validated with real data, enhancing route choice modeling under capacity constraints.

Findings

Model accurately captures congestion effects on capacities.

Capacible capacities significantly influence route utility.

Model outperforms baseline approaches ignoring capacity effects.

Abstract

With the prevalence of MaaS systems, route choice models need to consider characteristics unique to them. MaaS systems tend to involve service systems with fleets of vehicles; as a result, the available service capacity depends on the choices of other travelers in different parts of the system. We model this with a new concept of "congestible capacity"; that is, link capacities are a function of flow instead of link costs. This dependency is also non-separable; the capacity in one link can depend on flows from multiple links. An offline-online estimation method is introduced to capture the structural effects that flows have on capacities and the resulting impacts on route choice utilities. The method is first applied to obtain unique congestible capacity shadow prices in a multimodal network to verify the capability to capture congestion effects on capacities. The capacities are shown to vary and impact the utility of a route. The method is validated using real system data from Citi Bike in New York City. The results show that the model can fit to the data quite well and performs better than a baseline modeling approach that ignores congestible capacity effects. By relating the route choice to congestible capacities using a random utility model, modelers can monitor and quantify the impacts to traveler consumer surplus in real time. Applications of the model and online method include monitoring capacity effects on consumer surplus, using the model to direct incentives programs for rebalancing and other revenue management strategies, and to guide resource allocation to mitigate consumer surplus impacts due to disruptions from incidents.