Search

Bounded Path Size (BPS) route choice models (Duncan et al, 2021) offer a theoretically consistent and practical approach to dealing with both route overlap and unrealistic routes. It captures correlations between overlapping routes by including correction terms within the probability relations, and has a consistent criterion for assigning zero probabilities to unrealistic routes while eliminating their path size contributions. The present paper establishes Stochastic User Equilibrium (SUE) conditions for BPS models, where the choice sets of realistic routes are equilibrated along with the route flows. Solution existence/uniqueness are addressed. A generic solution algorithm is proposed, where realistic route choice sets are equilibrated from a pre-generated approximated universal set of routes. Numerical experiments on the Sioux Falls and Winnipeg networks show that BPS SUE models can be solved in feasible computation times compared to nonbounded versions, while providing potential for significantly improved robustness to the adopted master route choice set.

The Stochastic User Equilibrium (SUE) traffic assignment model is a well-known approach for investigating the behaviours of travellers on congested road networks. SUE compensates for driver/modelling uncertainty of the route travel costs by supposing the costs include stochastic terms. Two key challenges for SUE modelling, however, are capturing route correlations and dealing with unrealistic routes. Numerous correlation-based SUE models have been proposed, but issues remain over both internal consistency and choice set robustness. This paper formulates (and proves solution existence for) new internally consistent SUE formulations of GEV structure and correction term logit route choice models, where the functional forms in the correlation components are based upon generalised, flow-dependent congested costs, rather than e.g. length / free-flow travel time as done typically. Numerical experiments are then conducted on the Sioux Falls and Winnipeg networks, where computational feasibility for obtaining internally consistent solutions, choice set robustness, and internal consistency are assessed/compared.