倒立摆的神经网络自适应滑模控制

Neural Network Adaptive Sliding Mode Control for the Cart-Inverted Pendulum System

针对倒立摆系统的欠驱动、非线性和存在外界干扰的特点,提出一种基于RBF神经网络自适应滑模控制方法。首先将倒立摆系统的数学模型转换成符合欠驱动系统特征的标准形式,然后采用分层滑模控制方法设计控制律。RBF神经网络自适应方法对系统的不确定参数以及小车与导轨之间摩擦力等外界非线性干扰引起的不确定上界进行补偿,减小控制器的控制输出量,提高了控制精度。仿真结果表明,该方法具有很好的鲁棒性和抗干扰能力,能满足系统要求。

A novel Adaptive sliding mode control method based on RBF neural network is proposed for the cart-inverted pendulum（CIP） system against its characteristics of under actuation, nonlinearity and external disturbances. The mathematical model of CIP is converted into the norm expression of a class of underactuated systems, and then a stable hierarchical sliding-mode method can be adopted to design the CIP control law. RBF network is utilized to adaptively learn and compensate the upper bound of uncertainties which include the parameter variations and external disturbances such as the friction between the cart and rail. The controller output is reduced and the control accuracy is improved. The simulation results show that the controller can guarantee good robustness and anti-interference capability and meet the system requirement.