To improve the adaptive ability of intelligent driving vehicles to cope with different working conditions when using MPC for path tracking, a predictive control method based on RBF neural network and SSA optimization algorithm for variable time domain model is proposed. Build MPC controller based on vehicle three degree of freedom dynamic model. Based on this, RBF neural network is used to estimate the lateral error and heading angle error values under different vehicle speeds, adhesion coefficients, and time-domain parameters. Transfer the estimated results to SSA optimization to achieve real-time optimization of controller prediction and control time domains. A Simulink/Carsim simulation model was established to verify the proposed adaptive MPC controller. The experimental results showed that compared with fixed time-domain MPC and vehicle speed based adaptive time-domain Speed-MPC controller, the control strategy significantly reduced the average and maximum lateral and heading angle errors under three operating conditions: low-speed low adhesion, medium speed medium adhesion, and high-speed high adhesion. Therefore, the improved AMPC controller has good control performance, which can ensure the accuracy and stability of vehicle path tracking.