Abstract: In order to improve the accuracy and anti-interference ability of intelligent trolley of intelligent vehicle trajectory tracking, a fuzzy control method based on feed-forward type variable theory domain was proposed. Taking the automated guided vehicle (AGV) as the research object, the kinematic model of the vehicle was established to analyze the dynamics of the vehicle. Under the condition of keeping the main control parameters unchanged, the feedforward control was added into the variable-theoretic domain fuzzy control to build the feedforward variable-theoretic domain fuzzy PID control system, and then the parameters were adjusted through the variable-theoretic domain fuzzy PID control to obtain a better stability, and then the target tracking performance of the system was optimized through the adjustment of the feedforward control parameters. The system was adjusted by the variable domain fuzzy PID control parameters to obtain better stability, and then the feedforward control parameters were adjusted to make the system's target tracking performance reach the best. An S-shaped curve tracking route was set up and the simulation experiments on four control models, including traditional PID, fuzzy PID, variable universe fuzzy PID, and feedforward variable universe fuzzy PID were conducted on the Matlab/Simulink platform to compare and verify the control performance of the proposed method.The results show that under the condition of no interference, the maximum lateral deviation of vehicle turning of the four control methods are 0.258 0, 0.198 8, 0.179 2 and 0.112 5 m, respectively, and the method of this paper has the smallest deviation and the highest trajectory tracking accuracy. Under the simulation of the real road conditions, the maximum lateral deviation of vehicle turning of the four control methods are 0.891 7, 0.820 9, 0.791 7, 0.683 3 m, and the deviation of this paper’s method is still the smallest, and the anti-interference ability is the strongest, which verifies that the composite control method proposed in the paper can take tracking and stability into account.