Modelling public transport on‐board congestion: comparing schedule‐based and agent‐based assignment approaches and their implications

Abstract

Summary Transit systems are subject to congestion that influences system performance and level of service. The evaluation of measures to relieve congestion requires models that can capture their network effects and passengers' adaptation. In particular, on‐board congestion leads to an increase of crowding discomfort and denied boarding and a decrease in service reliability. This study performs a systematic comparison of alternative approaches to modelling on‐board congestion in transit networks. In particular, the congestion‐related functionalities of a schedule‐based model and an agent‐based transit assignment model are investigated, by comparing VISUM and BusMezzo, respectively. The theoretical background, modelling principles and implementation details of the alternative models are examined and demonstrated by testing various operational scenarios for an example network. The results suggest that differences in modelling passenger arrival process, choice‐set generation and route choice model yield systematically different passenger loads. The schedule‐based model is insensitive to a uniform increase in demand or decrease in capacity when caused by either vehicle capacity or service frequency reduction. In contrast, nominal travel times increase in the agent‐based model as demand increases or capacity decreases. The marginal increase in travel time increases as the network becomes more saturated. Whilst none of the existing models capture the full range of congestion effects and related behavioural responses, existing models can support different planning decisions. Copyright © 2016 John Wiley & Sons, Ltd.

Recovery time and propagation effects of passenger transport disruptions

“…For more details about the traffic and public transport supply representation see Toledo et al [18]. The dynamic public transport path choice model and related congestion effects is described in Cats and Hartl [17] and the modelling of unplanned service disruptions and their impacts on passenger distribution are explained in Cats and Jenelius [14]. The reader is referred to these studies for further details while the essential modelling principles are briefly provided below.…”

“…Once on-board a vehicle, a traveller makes an alighting decision which may be reconsidered in light of new information. For more information on the dynamic route choice model and the consideration of congestion effects, the reader is referred to Cats and Hartl [17].…”

“…When it equal to the upper limit of 1, some passengers may experience denied boarding and thus prolonged waiting times. Unlike static assignment models in which volumes can (unrealistically) exceed capacity (VOC >1) implying that the network cannot absorb the loaded demand, the dynamic assignment model used in this study enforces strict capacity constraints at individual vehicle level and hence ensures 0 ≤ VOC ≤ 1 [17].…”

“…We illustrate our approach by an application to the Stockholm rapid transit system (Fig. 1), made up of seven Metro lines (lines 10,11,13,14,17,18,19), four high-demand trunk bus lines (lines 1, 2, 3, 4) and one light rail train line (line 22). The Stockholm network consists of a mature and dense multi-modal core that offers route choice alternatives and the system is subject to recurrent disruptions, especially under adverse weather conditions.…”

Learning and Adaptation in Dynamic Transit Assignment Models for Congested Networks

“…These models are used for strategic planning such as evaluating network design implications. The modeling tools available for strategic transit planning purposes typically lack explicit representation of the dynamics in transit systems, which may lead to unrealistic passenger loads or travel times, as illustrated elsewhere (2)(3)(4).…”