Abstract: With the development of intelligent driving technology, lane keeping has become an important function for active safety in commercial vehicles. However, under the influence of load variations, delayed steering response, and complex road disturbances, traditional control methods struggle to balance dynamic tracking performance and robustness. Active Disturbance Rejection Control (ADRC) exhibits good system adaptability, however, its parameter tuning relies heavily on empirical experience and lacks a dynamic optimization mechanism. To address this issue, a lane-keeping strategy combining parameter tuning with adaptive ADRC is proposed. This strategy optimizes controller parameters using the Adaptive Particle Swarm Optimization (APSO) algorithm and introduces a self-learning coefficient module to enhance system adaptability and stability in correcting lane deviations under different loads and road conditions. Simulation results show that, compared with traditional ADRC and Model Predictive Control (MPC) methods, the proposed method reduces vehicle lateral deviation more quickly while ensuring steering safety. Additionally, the proposed method exhibits lower computational complexity and smaller convergence values.