A general multi-agent modelling framework for the transit assignment problem - A learning-based approach

Abstract

This paper presents the conceptual development of an innovative modelling framework for the transit assignment problem, structured in a multi-agent way and inspired by a learning-based approach. The proposed framework is based on representing passengers and both their learning and decision-making activities explicitly. The underlying hypothesis is that individual passengers are expected to adjust their behaviour (i.e. trip choices) according to their knowledge and experience with the transit system performance, and this decision-making process is based on a "mental model" of the transit network conditions. The proposed framework, with different specifications, is capable of representing current practices. The framework, once implemented, can be beneficial in many respects. When connected with urban transportation models - such as ILUTE - the effect of different land use policies, which change passenger demand, on the transit system performance can be evaluated and assessed. © Springer-Verlag Berlin Heidelberg 2006.