不确定机器人的优化神经网络滑模控制策略

Optimal Neural Network-based Sliding Mode Control for Robot Manipulators with Uncertainties

滑模控制响应快,对系统参数和外部扰动呈不变性,可保证系统的渐进稳定性,但它要求控制系统的不确定性的上界值必须已知。在实际中,不确定性的上界值是无法测量的,针对这个问题,采用RBF神经网络来对干扰的上界进行自适应学习,同时也降低了滑模产生的抖振现象。通过对单关节机器人的仿真研究表明:在存在模型误差和外部扰动的情况下,该方案既能达到高精度快速跟踪的目的,又能削弱滑模控制的抖动问题。

The sliding mode control has quick response and takes on invariability to system parameters and external disturbances, which can assure the asymptotic stability of system, but it demands that the industry value of the uncertainty must be known. However, it is impossible in the practice system. The RBF neural network was introduced to the common sliding mode control. The control law can guarantee fast convergence of trajectory tracking error as well as robustness for external disturbances and parameter uncertainties. The results of simulation about a single-link robotic manipulator show that the scheme can achieve tracking effect with high precision and speediness, diminish chattering of control under the condition of existing model error and external disturbance.