Choice set robustness and internal consistency in correlation-based logit stochastic user equilibrium models

Abstract

This is a repository copy of Choice set robustness and internal consistency in correlationbased logit stochastic user equilibrium models.

Formulation and solution method of bounded path size stochastic user equilibrium models – consistently addressing route overlap and unrealistic routes

“…It remains to identify a less strict proof that can be applied. Nevertheless, after searching for cases with multiple solutions in numerous experiments in the current study as well as in Duncan et al (2022) for other path size SUE models, we could not find any cases where multiple solutions existed. Again, this is an entirely different prospect to e.g.…”

“…In this section, some numerical experiments are conducted to compare the computational performance and flow results of bounded SUE models (namely BSUE, BBPS, BAPS, & BAPS′ SUE) as well as with other SUE models (namely MNL, PSL, GPSL, GPSL′, APSL, & APSL′ SUE, as defined in Duncan et al (2022) with flow-dependent path size terms). We examine results in the case where there are pre-generated approximated universal choice sets.…”

“…We examine results in the case where there are pre-generated approximated universal choice sets. Bounded SUE models are solved with the BPS SUE algorithm in Algorithm 1, and non-bounded SUE models are solved with a Flow-Averaging Algorithm (FAA) (see Duncan et al, 2022). The BPS SUE algorithm can be seen as an extension of the FAA, where column generation, flow allocation for new routes, and bound condition phases are added for the equilibration of choice sets.…”

“…For the small example network, the pre-generated choice set is the actual universal set of all 4 routes, where the routes are Route 1: 1 → 3, Route 2: 1 → 4, Route 3: 2 → 3, Route 4: 2 → 4. For Sioux Falls and Winnipeg, we utilise choice sets generated in Duncan et al (2022). For the Sioux Falls network, the choice sets were obtained by generating all routes with a free-flow travel time less than 2.5 times as much as the free-flow travel time on the quickest route for each OD movement.…”

“…The solution techniques developed in Duncan et al (2022) for solving APSL SUEwhich is a similar bi-level fixedpoint problemprovide promise for efficiently solving BAPS SUE, but it is uncertain how well these techniques will transfer given the used route choice sets also need to be equilibrated. For example, at each iteration of the solution algorithm there may be a different selection of routes, which compromises the information gained from the previous iteration.…”