Abstract: In the trajectory tracking task of autonomous mining trucks, the terrain information in open-pit mines is intricate and the road surface types are diverse, resulting in a pronounced conflict between tracking accuracy and lateral stability. Furthermore, the dynamic characteristics of the tire-ground interaction in autonomous trucks are complex, presenting significant challenges in modeling for trajectory tracking. To address the above issues, a Stability-based Model Free Adaptive Control (SMFAC) method is proposed, which achieves coordinated control of driving stability and tracking accuracy in a data-driven manner. Based on Pacejka tire model, the influence of road friction coefficient on the system is introduced. The centroid side slip angle versus centroid side slip angle velocity phase plane of mining trucks is constructed by the simplified dynamic equation. The boundary of stability region under different environmental conditions is identified dynamically to solve the stability coefficient. In addition, an Actor-Critic framework is constructed using radial basis function networks for the model-free trajectory tracking controller. The Actor network calculates the control quantity based on the system state, while the Critic network evaluates the value of the real-time control quantity and fits the value function. The stability coefficient is employed to design the network error during the learning process, aiming to minimize the error function to update weights of the hidden layer. By iteratively interacting with the environment through the mining truck, an optimal strategy for trajectory tracking is developed. A trajectory tracking simulation system is built on the CarSim and MATLAB/Simulink integrated platform to validate the effectiveness of proposed method. The results indicate that, compared with the model free PID algorithm and the model-based MPC and LQR algorithms, the proposed method can take into account both the angle tracking accuracy and the lateral tracking accuracy at both low and high speeds. And proposed method achieves the optimal trajectory tracking performance in double lane shift, single lane shift, and curve conditions. In addition, the SMFAC method can suppress the fluctuation of the output action, generate a smoother driving trajectory, and ensure the control stability of the unmanned mining truck.