Applying a Structured Dispersion Parameter to Multiclass Stochastic User Equilibrium Assignment Model

In recent practical transportation planning in Japan, the adoption of stochastic user equilibrium assignment models for traffic assessment has been discussed. Such models include a dispersion parameter that accounts for drivers' errors in perception of travel time at the time of route choice decision. In practice, setting a rational value for this parameter is an issue. Generally, the dispersion parameter is exogenously set to the same value for all origin destination (O-D) pairs. However, it is not guaranteed that errors in drivers' perception of travel costs are equal for all O-D pairs. This study examines how the parameter is set and applies a multiclass stochastic user equilibrium assignment model to consider differences in drivers' errors in perception of travel costs in using an existing general-purpose road network. Results show that changes in the dispersion parameter affect the user class flows that make up the link flows rather than the link flows themselves. Setting a dispersion parameter according to the travel cost of an O-D pair rather than setting a constant parameter improves the reproducibility of the user class flows on each link. These findings encourage practical transportation planners to evaluate qualitative measures of traffic flows in the road network, such as the utilization of road links.