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.

翻译：现代商用地面车辆日益配备多种运行模式（如人工驾驶、高级驾驶辅助、远程遥控操作、完全自动驾驶）。这些模式通常依赖于异构的感知基础设施和不同的路径规划算法，可能导致对交通环境的认知偏差和路径偏好错位。尽管此类技术加速了向智能化交通网络的转型，但当前部署方案仍无法规避由自私路由行为引发的挑战——驾驶员或自动化智能体倾向于选择个体最优路径，而非考虑全网拥堵缓解。现有交通流管理策略虽能处理路径规划中的领导者-跟随者动态博弈，但未考虑车辆在多运行模式间动态切换的能力。本文将车辆控制仲裁系统与多模态车辆的交互建模为具有信息不对称特性的重复单领导者-多跟随者斯塔克尔伯格博弈。针对该场景下精确解计算的难处理性，我们提出一种基于遗憾匹配算法的信任感知控制交易策略，通过自适应更新仲裁系统在连续动态路径决策中的混合策略。理论分析给出了TACTS算法实现的总网络行程时间相对于系统最优总行程时间的边界。在多种真实交通网络及不同拥堵水平下的仿真实验表明，TACTS能持续降低全网拥堵水平，其性能普遍优于其他路径规划与控制分配策略，在高拥堵和强诱导车流场景下表现尤为突出。