PERSONALIZED TOLLING POLICIES
Team: Yifei Xie, Ravi Seshadri, Moshe Ben-Akiva
Model heterogeneous managed lanes travel behavior
Design and evaluate a bi-level optimization framework tailored for real-time personalized pricing
Managed lanes (ML) is a common road pricing scheme that offers separate tolled lanes adjacent to free general-purpose lanes. Toll charges are expected to reduce road congestion, provide more reliable travel, and generate revenue for transportation network improvement.
To better achieve revenue and traffic management objectives, we propose a real-time prediction-based personalized tolling system. In addition to system-level displayed toll rates, it offers personalized discounts tailored to individual-specific preferences. A traveler who wishes to receive discounts is required to download a specialized mobile app, through which a discount is offered upon arrival.
For effective real-time solutions of personalized discounts and displayed toll rates, two connected optimization subproblems are formulated with consistent objectives: user optimization and system optimization. The user optimization is triggered upon detection of each subscriber (a traveler with the discount app). It offers a discount based on individual-level choice prediction, the displayed toll rates, and a control parameter. The control parameter affects how much discount to be offered considering network congestion. The system optimization works on a rolling horizon and is triggered periodically (e.g. every five minutes). It performs simulation-based traffic prediction, and optimizes the displayed toll rates and the control parameter accordingly. The pricing system is backed with a choice model fully capturing preference heterogeneity and a dynamic traffic assignment (DTA) system, both of which are updated with real-time data from the mobile app and network surveillance in an online setting.
The proposed framework has promising potentials to be applied to other transportation (e.g. ride-hailing services) or non-transportation applications (e.g. e-commerce) where personalized pricing is applicable.
Zhang, Y., 2019. Real-time personalized toll optimization based on traffic predictions (Doctoral dissertation, Massachusetts Institute of Technology).
Zhang, Y., Atasoy, B., Akkinepally, A. and Ben-Akiva, M., 2019. Dynamic toll pricing using dynamic traffic assignment system with online calibration. Transportation Research Record, 2673(10), pp.532-546.
Xie, Y., Seshadri R., Zhang, Y. and Ben-Akiva, M. Real-time personalized tolling with long-term objectives, working paper.