Abstract: Aiming at the problems of unstable attitude of mountain unmanned operation chassis under complex roads and poor adaptability and robustness of traditional control methods,a chassis attitude control strategy based on Newton-Raphson-Based Optimizer (NRBO) algorithm,eXtreme Gradient Boosting (XGBoost) algorithm and Twin Delayed Deep Deterministic policy gradient (TD3) algorithm was proposed. Firstly,the seven-degree-of-freedom active suspension vibration model environment was built;then the state prediction model of NRBO-XGBoost was trained,the state prediction model was added to the TD3 algorithm and the attention mechanism was added to the network to enhance the decision-making ability and adaptive ability of the TD3 intelligences in complex environments,and at the same time,the reward function was designed and the TD3 intelligences were trained to realize the chassis attitude control in complex road environments; finally,simulations were carried out based on Matlab 2023a/Simulink software. The simulation results show that the chassis attitude strategy method based on the improved TD3 can effectively suppress the attitude change of unmanned operation chassis under complex roads,and the pitch angle,lateral inclination angle,and vertical displacement are suppressed by 61.4 %,84.9 %,and 84.9 %,respectively,which significantly improves the smoothness;compared with the traditional DDPG,PPO,and TD3 reinforcement learning control strategies,with the improved TD3 algorithm the pitch angle is improved by 49.1 %,7.4 % and 37.2 %,respectively,the lateral inclination angle is improved by 83.3 %,36.5 % and 34.7 %,respectively,the vertical displacement is improved by 70.7 %,77.5 % and 64.0 %,respectively,and the vertical displacement acceleration is improved by 67.7 %,42.1 % and 49.7 %,respectively,which provides a better control effect with better adaptability and robustness.

Key words: mountain unmanned operation chassis, active suspension control, improved TD3 algorithm, self-attention mechanism