Vehicle Active Stabilizer Bar Composite Control and Optimization Based on Reinforcement Learning

During actual vehicle operation, working conditions are highly complex, involving both body roll induced by steering centrifugal force and attitude fluctuations caused by random road irregularities or sudden lateral wind disturbances. By optimizing the control of the active stabilizer bar (ASB), its torque compensation capability can be more effectively utilized, thereby improving vehicle ride quality and handling stability under extreme conditions. This paper first establishes a vehicle roll model with a passive stabilizer bar. Then, an active disturbance rejection control (ADRC) controller, a linear active disturbance rejection control (LADRC) controller, and a fuzzy proportional–integral and proportional–derivative (PI-PD) controller are designed and verified through simulation. The results show that all three active control methods improve roll stability compared with the passive system, and the ADRC controller achieves better control performance than the fuzzy PI-PD and LADRC controllers. Furthermore, a control strategy for the active stabilizer bar model is developed based on the deep deterministic policy gradient (DDPG) algorithm. The simulation results show that, using deep reinforcement learning for feedforward optimization, the fuzzy PI-PD, LADRC, and ADRC control methods reduce the body roll angle by 3.8%, 27.1%, and 25.0%, respectively. The front-axle anti-roll moments are reduced by 13.4%, 14.0%, and 16.5%, respectively, while the rear-axle anti-roll moments are reduced by 14.8%, 13.4%, and 14.5%, respectively.