A novel agent-based modelling framework for travel time reliability analysis

Abstract

ABSTRACTUnder transportation network supply and demand uncertainty, travel time reliability emerges from non-linear interactions among numerous travellers who are driven by self-interest, learn and adapt in changing situations, and therefore can be modelled as an emergent network property. This paper proposes a novel theoretical framework for the study of complexity regarding travel time reliability on the basis of empirically derived individual decision rules, and develops an agent-based evolutionary model for travel time reliability analysis. Findings show that actual route choice behaviours related to network reliability are often non-optimal, and that these behaviours themselves are important determinants of travel time reliability under network supply and demand uncertainty. While many travellers search for alternative routes under uncertainty, few travellers actually change routes. Route choice rules employed by travellers can successfully improve travel time reliability in most tested uncertainty scenarios, but they do not necessarily reduce absolute travel time. While reducing demand uncertainty improves travel time and planning under uncertainty, the model provides a theoretically sound tool for analysing the impact of traveller information systems on individual travel behaviour and network performance. It can also serve as an empirically estimated router for microscopic traffic simulators or act as a behavioural assignment algorithm for activity-based travel demand models.