Incorporating temporal and spatial dependencies in a stochastic time-dependent multi-state supernetwork for individual activity-travel scheduling

Multi-state supernetworks have been expanded in space and time for modeling individual activity-travel scheduling (ATS) behavior. Any path in a time-expanded supernetwork represents an activity-travel pattern (ATP) at a high level of detail. To alleviate the limitation of a deterministic network representation, time uncertainty has been incorporated in multi-state supernetworks. However, the extension unrealistically assumed that all uncertain travel times are time-invariant and spatially independent. In this study, temporal and spatial dependencies among uncertain link travel times are considered in a stochastic time-dependent (STD) context using uncertain support points. ATS under uncertainty is formulated as a path finding problem in a stochastic multi-state supernetwork, given the individual’s decision rule, which for illustrative purposes is assumed to be the minimization of expected disutility in this paper. Recursive dynamic programming is applied to find the non-dominated paths, subject to time window constraints at the activity locations.