In order to solve the problem of parameter tuning in the design of sliding mode variable structure control law, a parameter optimization method based on reinforcement learning is proposed. Firstly, according to the system, the corresponding sliding mode control law is designed, and the range of parameter selection is given. Based on the Actor-Critic structure, the parameter online adjuster is designed. TD-error method is used to solve the problem and gradient descent method is used to calculate the updated value of neural network weight. Finally, the longitudinal channel system of fixed wing vehicle is taken as an example to simulate and Verify through experiments, which shows that the proposed control method reduces the blindness of control parameter tuning and improves the dynamic performance of the system significantly.