线控转向系统的自适应神经网络滑模控制

Adaptive Neural Network Sliding Mode Control for Steer-by-Wire System

传统滑模控制(SMC)方法应用到线控转向系统时,需要预先获取系统干扰的上界值,且系统干扰变化会导致转角控制稳定性变差。为了提高线控转向系统的车轮转角跟踪性能,提出了一种考虑系统干扰的自适应神经网络滑模控制(RBFSMC)方法。RBFSMC先采用径向基神经网络对系统的不确定性和电机力矩扰动进行实时估计,再结合传统滑模控制设计车轮转角控制器,以提高转角控制的自适应性和稳定性。Matlab/Simulink、CarSim对SMC和RBFSMC的联合仿真结果对比表明,在车辆维持转向、连续转向和单移线/双移线工况下,RBFSMC能更好地维持0°车轮转角和实现动态车轮转角快速稳定的跟踪,较SMC具有更好的转角响应和跟踪性能,说明RBFSMC比SMC具有更好的鲁棒性和稳定性。

When the traditional sliding mode control(SMC)method is applied to steer-by-wire(SBW)system,it is necessary to obtain the upper bound value of system disturbance in advance,and the change of system distur-bance will lead to poor stability of angle control.In order to improve the wheel angle tracking performance of SBW system,an adaptive neural network sliding mode control(RBFSMC)method considering system disturbance was proposed.Firstly,the RBF neural network was used to estimate the system uncertainty and motor torque distur-bance in real time.Secondly,the wheel angle controller was designed by combining RBF with SMC in order to improve the adaptability and stability of the angle control.The joint simulation results of MATLAB/Simulink and CarSim on SMC and RBFSMC show that,RBFSMC can better maintain 0°wheel angle and realize fast and stable tracking of dynamic wheel angle under the conditions of vehicle maintaining steering,continuous steering and single/double lane changing,and it has better corner response and tracking performance than SMC.The study indicates that RBFSMC has better robustness and stability than SMC.