Congestion Mitigation in Vehicular Traffic Networks with Multiple Operational Modalities
Modern commercial ground vehicles are increasingly equipped with multiple operational modalities (e.g., human driving, advanced driver assistance, remote tele-operation, full autonomy). These often rely on heterogeneous sensing infrastructures and distinct routing algorithms, which can yield misaligned perceptions of the traffic environment and route preferences. While such technologies accelerate the transition toward increasingly intelligent transportation networks, their current deployment fails to avoid challenges associated with selfish routing behavior, in which drivers or automated agents prioritize individually optimal routes instead of network-wide congestion mitigation. Existing traffic flow management strategies can address leader-follower dynamics in traffic routing problems but are not designed to account for vehicles capable of dynamically switching between multiple operational modes. This paper models the interaction between a vehicle control arbitration system and a multi-modal vehicle as a repeated single-leader, multiple follower Stackelberg game with asymmetric information. To address the intractability of computing an exact solution in this setting, we propose a Trust-Aware Control Trading Strategy (TACTS) utilizing a regret matching-based algorithm to adaptively update the arbitration system's mixed strategy over sequential, dynamic routing decisions. Theoretical results provide bounds on the realized total network travel time under TACTS algorithm relative to the system-optimal total network travel time. Experimental results of simulations between the system and a vehicle in several real-world traffic networks under various different congestion levels demonstrate that TACTS consistently reduces network-wide congestion and generally outperforms alternative routing and control-allocation strategies, particularly under high congestion and heavy induced vehicle flows.