This paper presents a dynamic mixed discrete-continuous choice approach to modeling pedestrian travel and activity choice behavior in public facilities. The approach views revealed behavior as a manifestation of pedestrians' preferences by assuming that pedestrians choose the alternative that maximizes expected (subjective) utility, while taking into account the uncertainty in expected traffic conditions. The choice dimensions are trajectories between origin and subsequent destinations, areas where activities are performed (multiple vs. fixed destination), execution of discretionary activities, and finally activities completion times and order. The disutility of a trajectory determines the trajectory choice of the traveler. Destination area choice is included in the modeling by determining time-dependent and destination-specific arrival costs. Furthermore, penalties for not executing a planned activity are introduced into the modeling framework. The resulting modeling approach has a clear analogy with stochastic control theory and dynamic programming in continuous time and space. The main innovations presented here is the relaxation of the assumption that routes are discrete sets of travel links. The approach relaxes the need to build a discrete network, while routes (trajectories) are continuous functions in time and space. At the same time, destination choice is included in the modeling framework.
