Lawrence Duncan: Incorporating bounds within discrete choice models: Derivation and applications to modelling route choices during public transport disruptions and optimising transport networks

Abstract.

Route choice models are important for understanding the route preferences of travellers and forecasting travel behaviour in hypothetical transport policy scenarios. Route choice modelling traditionally consists of two sequential steps: first generating a set of routes that travellers consider, and then calculating choice probabilities of these routes. This results, however, in inconsistency between route generation and route choice probability criteria, as well as discontinuity of choice probabilities as routes enter/exit the choice set. In this presentation I will mathematically derive a Conjunctive Bounded Choice Model (CBCM) that leverages upper-bounds that travellers impose on route attributes (e.g. walking/waiting/in-vehicle time) to implicitly determine the set of considered routes in the choice probability calculation stage. Crucially, it does so in a way that maintains a continuous and differentiable choice probability function. The CBCM will then be employed to model route preferences during unplanned public transport disruptions, where it is shown that preferences and bounds differ to during normal operating conditions. And, it will be employed to optimise continuous public transport network designs (e.g. line frequency, stop location, fare pricing), where the CBCM allows route choice sets to consistently adapt to the network design in a continuous way.