Resource-efficient direct yaw moment control for four-wheel independent drive electric vehicles based on dual event-triggered adaptive dynamic programming with off-policy mechanism

Abstract

This paper aims to design a resource-efficient direct yaw moment control (DYC) scheme for four-wheel independent drive electric vehicles using reinforcement learning. Firstly, the DYC problem is constructed based on the nonlinear vehicle model and described as an optimization problem by constructing performance indicators and Hamiltonian functions. Then, a dual event-triggered mechanism is proposed to save communication and computational costs. Next, the Hamilton–Jacobi–Bellman equation of the optimization problem is constructed using the adaptive dynamic programming method. Furthermore, an experience replay buffer incorporating dual historical data is designed for weight iteration to relax the persistence of excitation condition and eliminate the need for importance sampling. Finally, the approximate optimal control law is obtained through online tuning with an off-policy mechanism. Simulation and multiple hardware-in-the-loop tests demonstrate that the proposed scheme is robust and effectively enhances vehicle lateral stability while reducing communication and computational costs by approximately 75% and 25%, respectively, compared to traditional time-triggered methods.