Managing Multi-modal Transportation System Dynamics with Adaptive Pricing

This study models and manages a multi-region and multi-modal transportation system, given that travelers can adjust their mode choices over (calendar) time, and traffic dynamics in the network evolve from day to day. In particular, it considers that the city network can be partitioned into two regions: the city center and the periphery. There are park-and-ride facilities located at the boundary of the city center. For travelers living in the periphery, they can either drive to the city center, or take public transit, or drive to the park-and-ride facilities and then transfer to public transit. For travelers living in the city center, they can either drive or take public transit. Travelers can “learn” from their experience, as well as information about traffic condition, thus will adjust their choices. It follows that the dynamic traffic pattern (within day) in the city network will evolve over (calendar) time. To improve traffic efficiency in the network, we propose to update parking pricing (or congestion pricing) from period-to-period (e.g., one period can be one month), which is adaptive to system state. More specifically, in the end of a period, after “learning” from past experience during the period, travelers’ choices will evolve to an equilibrium state, as well as the traffic dynamics (within day) in the network. These equilibrium dynamic traffic conditions at the end of the period can be measured or observed. We then can update the pricing strategies for next period based on these observed or measured conditions. The proposed approach is practical and suitable for implementation in large-scale networks, and can help to reduce total social cost effectively, as shown in our numerical experiments.