Abstract: In response to the difficulty in selecting the weight matrices Q and R for the Model Predictive Control (MPC) algorithm in existing lateral vehicle control, which results in poor vehicle trajectory tracking performance, a vehicle trajectory tracking control method based on the multi-strategy Pelican Algorithm (POA) for optimizing the MPC weight matrices Q and R is proposed. Firstly, a simplified vehicle dynamics model is established, and a lateral MPC controller is designed. Secondly, to address the limitations of the Pelican Algorithm, an Improved Pelican Algorithm (IPOA) with multiple strategies is proposed. It introduces sociological learning strategies and a forgetting mechanism to establish interactive channels among Pelican individuals, enhancing the algorithm's convergence speed. Furthermore, to address the issues of insufficient population diversity and the tendency to fall into local optima, a dynamic adaptive t-distribution mutation operator with a feedback mechanism is designed to expand the search space and enhance the local optimization capability in later stages. Finally, the IPOA is used to optimize the MPC weight matrices, and joint simulation is conducted on the Simulink and Carsim platforms. Results show that the proposed IPOA-MPC controller effectively improves the accuracy and stability of vehicle tracking along reference trajectories, reducing the maximum lateral error by 60.74% and demonstrating strong generalization capabilities in the optimized controller.

Key words: Pelican Algorithm, Trajectory tracking, Model predictive control, Weight matrix amnesic mechanism