基于神经网络补偿的机器人稳定自适应控制

STABLE ADAPTIVE CONTROL FOR ROBOT MANIPULATORS BASED ON NEURAL NETWORK COMPENSATION CONTROLLER

针对建模不精确的机器人,提出了一种基于神经网络补偿的机器人轨迹跟踪稳定自适应控制方法,文中通过设计神经网络补偿器和自适应鲁棒控制项,有效地补偿了模型的不确定性部分和网络逼近误差。由于算法包含有补偿神经网络逼近误差的鲁棒控制项,实际应用中对神经网络规模的要求可以降低;而且神经网络连接权是在线调整的,不需要离线学习过程。理论表明算法能够保证跟踪误差及神经网络连接权估计最终一致有界,仿真结果也验证了算法的有效性。

A stable adaptive control approach based on neural network (NN) compensation controller is developed for tracking the trajectory of robot manipulator whose dynamic model is not exactly known. A NN compensation controller and an additional adaptive robust control input item are incorporated with the computed torque control scheme to effectively compensate NN approximation error and dynamic model uncertainties. By applying our control algorithm, the number of hidden-layer neurons can be reduced since the additional robust control input item has the ability to compensate NN approximation error. The weights of NN are tuned on-line, and no offline learning phase is required. The uniformly ultimately bounded (UUB) stability of tracking errors and NN weights is guaranteed. Simulation results show the effectiveness of the proposed controller.