车辆紧急避让路径跟踪神经网络自抗扰控制

Neural Network Auto Disturbance Rejection Control for Vehicle Emergency Avoidance Path Tracking

高速车辆自动紧急避让技术能够提高车辆行驶的安全性。在高速车辆紧急避让过程中,由于外界干扰不确定等因素,参数固定的自抗扰控制器存在控制精度较差、效果不尽人意的问题,针对这一问题提出了一种基于神经网络的自抗扰控制方法。以车辆二自由度模型为基础,设计了二阶自抗扰控制器,利用神经网络在线整定三阶扩张状态观测器参数,并嵌入到自抗扰控制器中,同时考虑车辆避让过程中存在侧向加速度过大、曲率不连续等问题,采用Sigmoid函数进行路径再规划。Carsim/Simulink联合仿真结果表明,在不同车速下外界不同侧向风速干扰时,神经网络自抗扰控制器较常规自抗扰控制器路径跟踪精度高、鲁棒性好,且在100km/h车速下60km/h侧向风干扰时,两者最大跟踪误差分别为9.82%、58.70%。

High speed vehicle automatic emergency avoidance technology can improve the safety of vehicle running.In the process of high-speed vehicle emergency avoidance,due to external interference uncertainty and other factors,the fixed parameter active disturbance rejection controller has poor control accuracy and unsatisfactory effect.In view of this problem,a method of active disturbance rejection control based on neural network is proposed.Designing the two order active disturbance rejection controller based on two degree of freedom vehicle model,online tuning of three order extended state observer parameters by using neural network,and embedded into the active disturbance rejection controller at the same time,Sigmoid function was used to solve the excessive lateral acceleration and curvature discontinuity problems in path planning.Carsim/Simulink joint simulation results show that the neural network active disturbance rejection controller has higher tracking accuracy and better robustness than the conventional ADRC emergency avoidance path under different lateral wind speed disturbances of different vehicle speed,the maximum tracking error is 9.82%and 58.70%when the lateral wind disturbance of 60km/h of 100km/h vehicle speed.