基于神经网络逆系统的机器人分数阶滑模控制∗

摘要

To improve the control performance of the industrial robot, an ANN-inversion based fractional-or-der sliding mode control( FOSMC) scheme is proposed. Firstly, the BP neural network is used for the inver-sion of the industrial robot, and approximate decoupling and linearization of the industrial robot is got. Sec-ondly, the composite pseudo linear system, which is composed of the ANN-Inversion system and the con-trolled industrial robot, is equivalent to a linear system with disturbance in view of the industrial robot uncer-tainties and the BP neural network’ s approximation error. Then, the fractional-order sliding control( FOS-MC) scheme is proposed based on the SMC theory and fractional calculus for the linear system with disturb-ance, and the stability analysis is given. Finally, case study is fulfilled for a two-DOF robot under different conditions, and results show the effectiveness of the proposed control scheme.%为提高工业机器人的控制性能，提出了基于神经网络逆系统的分数阶滑模控制方法。首先，使用BP神经网络逼近机器人的逆系统，实现了工业机器人系统的近似解耦线性化；考虑到工业机器人存在的不确定性和BP神经网络的逼近误差，将神经网络逆与工业机器人组成的复合伪线性系统等效为含有扰动的线性系统；在此基础上，基于滑模控制和分数阶微积分理论设计了分数阶滑模控制器，证明了闭环系统的稳定性。针对二自由度机器人多种不同工况的仿真研究表明了所提方法的有效性。