Game-Theoretic Control With Learning-Aware Dynamic Event-Triggered Communication for Vehicle Active Suspension Systems

One critical technology enabling the autonomous operation of intelligent vehicles is the active suspension control, which is responsible for regulating ride comfort, road handling and so on. This paper concentrates on this problem by presenting a novel learning-aware dynamic event-triggered control scheme under limited communication resources and parameter uncertainties. First, the dynamics model of quarter-vehicle suspension is established and its active control problem is formulated as a two-player zero-sum differential game problem containing a joint cost function. Then, relying on event-triggered communication, an actor-critic adaptive learning algorithm is proposed through integral reinforcement learning and weighted experience replay technique. This algorithm can online approximate the optimal control policy against disturbance in the case of unknown internal dynamics. Furthermore, two learning-aware dynamic event-triggered mechanisms are designed for more flexible scheduling of communication resources, enabling the system to dynamically adjust the triggering process based on the learning status and state variations. A distinctive feature of these two mechanisms is their seamless integration into the learning algorithm via dynamic threshold parameters, thereby balancing multiple performance requirements. Finally, the performance of the proposed scheme is demonstrated through a nonlinear example, and further validated on an active suspension system via various simulation scenarios and comparative experiments.