Vehicle suspension system is crucial for enhancing driving comfort,maneuverability and safety. However,traditional suspension control methods have limitations in dealing with complex road conditions and variations in system parameters. To address this,an efficient adaptive controller for vehicle active suspension system is designed based on Soft Actor-Critic (SAC) algorithm. This controller utilizes maximum entropy reinforcement learning to optimize the stochasticity and gain of the strategy,demonstrating improved adaptability under different road conditions. The performance of SAC algorithm is compared and analyzed with traditional skyhook control and Deep Deterministic Policy Gradient (DDPG) algorithm by establishing a quarter-car active suspension model and incorporating a stochastic road simulation environment. The results show that the SAC-based controller significantly reduces key performance indices such as body vertical acceleration,suspension dynamic deflection and tire dynamic load. It also exhibits superior stability and control effect across various road conditions and vehicle speeds. This study verifies the strong potential and practicality of SAC algorithm in the field of active suspension control,providing a new direction for the intelligent development of vehicle suspension systems.