A bounded path size route choice model excluding unrealistic routes: formulation and estimation from a large-scale GPS study

This paper develops a new route choice modelling framework that deals with both route correlations and unrealistic routes in a consistent and robust way. To do this, we explore the integration of a correlation-based Path Size Logit model with the Bounded Choice Model (BCM) (Watling et al. 2018. “Stochastic User Equilibrium with a Bounded Choice Model.” Transportation Research Part B: Methodological 114: 254–280). We find, however, that the natural integration of these models leads to behavioural inconsistencies and/or undesirable mathematical properties. Solving these challenges, we derive a mathematically well-defined Bounded Path Size (BPS) model form that utilises a consistent criterion for assigning zero choice probabilities to unrealistic routes while eliminating their path size contributions. A closed-form and fixed-point BPS model are consequently formulated.  Subsequently, we consider parameter estimation in a simulation study and on a real-life large-scale network using GPS data, where computational feasibility is demonstrated. Estimation results show the potential of the BPS models to give improved fit relative to non-bounded versions (and BCM).