Pedestrian velocity and direction choice problem based on probabilistic activity domain

In this study, we propose a sequential model of pedestrian time-space use in urban space, describing choice behavior based on rough spatial cognition. We describe the behavior using an individual-OD-based coordinates, and assume that each individual make decision based on the utility of dynamic activity domain under time constraints. The model using the generalized coordinates makes if possible to analyze spatial characteristics without predefined links or zones. The model is estimated with GPS-based probe person data obtained within the 1km-area around Shibuya station and reasonable results that show the tendency of characteristic behaviors are obtained.