基于RBF神经网络的机械臂自适应控制方法

Adaptive Control Method of Manipulators Based on RBF Neural Network

摘要:

针对机械臂受内部摩擦和时变扰动等不确定性因素的影响,其轨迹跟踪控制系统的跟踪精度会下降,且影响系统的稳定性,提出一种基于径向基函数神经网络的自适应控制方法。首先,利用RBF神经网络采用离线训练和在线学习的方式对机械臂的动力学模型进行辨识；其次针对机械臂控制系统中的摩擦,设计RBF神经网络自适应控制算法对其进行逼近得到补偿控制量。针对时变扰动和神经网络逼近误差设计鲁棒项,以克服众多不确定性因素带来的影响,同时通过构造李亚普诺夫函数对所设计的控制系统进行稳定性分析；最后,仿真实验结果证明提出的控制方法具有较高的跟踪精度、抗干扰能力和较强的鲁棒性。

Abstract:

Aiming at the manipulator is affected by uncertainties such as internal friction and time-varying disturbance, the tracking accuracy of its trajectory tracking control system will decrease and affect the stability of the system, an adaptive control method based on radial basis function (RBF) neural network is proposed. Firstly, the RBF neural network is used to identify the dynamic model of the manipulator by offline training and online learning. Secondly, the RBF neural network adaptive control algorithm is designed to approach the friction in the manipulator control system to obtain the compensation control. The robust term is designed for time-varying disturbance and neural network approximation error to overcome the influence of many uncertain factors. At the same time, the Lyapunov function is constructed to analyze the stability of the designed control system. Finally, simulation results show that the proposed control method has higher tracking accuracy, anti-interference ability and stronger robustness.