基于扰动观测的客车ESC自适应神经网络滑模控制

为了提高客车电子稳定性控制系统（ESC）的控制精度，针对实际车辆系统建模中存在各种非线性扰动项以及传统滑模控制（Sliding Mode Control，SMC）中抖振较大的问题，提出一种自适应神经网络滑模控制算法。基于2自由度车辆模型，首先设计一个二阶滑模（Second-order Sliding Mode，SOSM）估计器对车辆的质心侧偏角进行估计，然后利用径向基（Radial Basis Function，RBF）神经网络对车辆系统建模中的各种非线性扰动项进行实时估计，并进行Lyapunov稳定性证明，RBF神经网络估计车辆系统建模的各种非线性扰动项可以有效减小滑模控制符号项的系数，从而减小滑模抖振水平。为了更进一步优化传统滑模控制的参数调节过程，减小滑模抖振并提高系统控制精度，再次利用RBF神经网络对传统滑模控制中的关键参数进行自适应调节。最后为了验证算法的有效性，搭建客车电控气压制动系统硬件在环试验台，在硬件在环试验台上对算法的有效性和精度进行试验验证。研究结果表明：客车ESC在自适应神经网络滑模算法的控制下，横摆角速度和质心侧偏角能够较好地跟随上理想的横摆角速度和理想质心侧偏角，横摆角速度和质心侧偏角的跟随误差降低；利用RBF神经网络估计客车建模中的各种非线性扰动项和利用RBF神经网络自适应调节传统滑模控制的关键参数，可以有效提高客车ESC的控制精度。

Bus ESC Adaptive Neural Network Sliding Mode Control Based on Disturbance Observation

In order to improve the ESC control accuracy of bus,aiming at the problems of various nonlinear disturbances in the modeling of actual vehicle systems and the large chattering in traditional sliding mode control(SMC),an adaptive neural network sliding mode control algorithm was proposed.Based on the two-degree-of-freedom vehicle model,a second-order sliding mode(SOSM)estimator was first designed to estimate the vehicle's sideslip angle,and then radial basis function(RBF)neural network was used to estimate various nonlinear disturbance terms in vehicle system modeling in real time.Lyapunov stability was proved.The RBF neural network estimates various nonlinear disturbance terms for vehicle system modeling,which can effectively reduce the coefficients of the sliding mode control symbol term,thereby reducing the sliding mode chattering level.In order to further optimize the parameter adjustment process of traditional sliding mode control,reduce sliding mode chattering,and improve system control accuracy,the RBF neural network was used to adaptively adjust key parameters in traditional sliding mode control.Finally,in order to verify the effectiveness of the algorithm,a hardware-in-the-loop test bench for the bus's electronically controlled pneumatic brake system was set up.The effectiveness and accuracy of the algorithm were experimentally verified on the hardware-in-the-loop test bench.The test results of hardware-in-the-loop show that under the control of the adaptive neural network sliding mode algorithm,the ESC of the bus can better follow the ideal yaw rate and ideal sideslip angle,the following errors of yaw rate and sideslip angle are reduced.Therefore,this paper uses RBF neural network to estimate various nonlinear disturbance terms in bus modeling,and uses RBF neural network to adaptively adjust key parameters of traditional sliding mode control,which can effectively improve the ESC control accuracy of bus.