基于学习型滑模预测控制的无人驾驶车辆非结构化环境轨迹跟踪及稳定性控制

Learning-based Sliding Mode Predictive Trajectory Tracking and Stability Control for Autonomous Vehicle in Unstructured Environments

针对高速无人驾驶车辆在复杂非结构化环境下轨迹跟踪精度和横摆稳定性之间难以协调平衡的难题,提出一种基于学习型滑模预测控制的轨迹跟踪及稳定性协调控制方法。首先,基于高斯过程回归方法构建数据学习型预测模型,以解决复杂非结构化环境下无人驾驶车辆的剩余模型不确定性和环境噪声干扰问题。其次,提出一种基于高斯-滑模预测控制的轨迹跟踪及稳定性协调控制方法,将基础模型与高斯不确定性预测相结合作为控制模型,设计基于滚动预测优化的滑模控制方法,满足多约束下控制器的实时性和鲁棒性要求。此外,构建车辆未来时刻行驶风险预测模型,利用基于预测时域内相对残差的递归贝叶斯定理提前决策多目标融合型函数的权重系数,满足全局性能最优。仿真结果表明,所提方法有效提高了存在路面噪声干扰的非结构道路下高速无人驾驶车辆轨迹跟踪精度和动力学稳定性。

Aiming at the difficulty in balancing trajectory tracking accuracy and yaw stability of high-speed autonomous vehicles in complex unstructured environments,a trajectory tracking and stability coordination control method based on learning sliding mode predictive control(L-SMPC)was proposed.Firstly,a data learning prediction model is constructed based on Gaussian process regression(GPR)to solve the problem of residual model uncertainty and noise interference in the complex unstructured environment.Then,a trajectory tracking and stability coordination control method based on Gaussian-sliding mode predictive control(GP-SMPC)is proposed,the basic model and Gaussian uncertainty prediction are combined as the control model,and the sliding mode control method based on rolling prediction optimization is designed,thus satisfying the real-time performance and robustness of the controller under multiple constraints.In addition,the future vehicle driving risk prediction model is constructed,and the weight coefficients of the multi-objective fusion function are decided in advance by recursive Bayes theorem based on the relative residuals in the prediction horizon,which satisfies the global optimal performance.The simulation results show that,the proposed method effectively improves the trajectory tracking accuracy and dynamics stability of high-speed autonomous vehicles on unstructured roads with road noise interference.