Adaptive event-triggered soft actor-critic control for MR suspension subject to actuator smoothness constraints

Magnetorheological (MR) suspension control faces high computational and communication burdens, large actuator force fluctuations, and reduced damper service life. This study proposes an adaptive event-triggered soft actor-critic (AET-SAC) control strategy subject to actuator smoothness constraints. First, a quarter-vehicle MR suspension dynamic model incorporating first-order actuator dynamics and hard constraints on the control-force change rate is established. Next, a dual smoothness constraint mechanism is incorporated into the SAC framework. A multi-objective reward function is designed to balance vibration suppression performance, actuator energy consumption, and damper service life. Then, an adaptive event-triggered mechanism based on the relative error of body acceleration is incorporated into the policy inference process, with the triggering threshold adjusted according to the vibration intensity to reduce unnecessary control updates. Lyapunov theory is used to establish uniform ultimate boundedness of the closed-loop system and the absence of the Zeno phenomenon. Finally, comprehensive simulations are conducted under bump and random road conditions, together with hardware-in-the-loop experiments on an STM32H753 embedded platform. The results show that, compared with the passive suspension, the proposed AET-SAC strategy reduces the RMS value of body acceleration by more than 60% and unnecessary control updates by approximately 58%. In addition, the proposed strategy provides smoother control-force output and lower actuator energy consumption than the AET-TD3, AET-DDPG, and conventional LQR controllers.