基于计算转矩控制结构的机械手鲁棒神经网络补偿控制

提出了一种新的不确定性机器人跟踪控制策略，文中基于计算转矩控制结构，采用了函数链网络实现一个神经网络补偿器，并叠加一个鲁棒控制项，以补偿模型的不确定性部分，另外，还考虑了神经网络逼近误差非一致有界的情形，设计了自适应的鲁棒控制项，算法可保证跟踪误差及神经网络权估计最终一致有界，与其它有关基于计算转矩控制的方法相比，该算法既不需要测量关节角加速度，也不要求惯性矩阵已知，理论和仿真均证明了算法和可靠性和有效性。

Robust Neural-Network Compensating Control for Robot Manipulator Based on Computed Torque Control

This paper proposes a new controller design approach for trajectory tracking of robot manipulator with uncertainties. The proposed controller is based on the computed torque control structure, and incorporates a compensator, which is realized by Functional Link Neural Network, and a robustifying term. In addition, when neural newtork reconstruction error is not uniformly bounded, an adaptive robustifying term is designed. It is shown that all the signals in the closed loop system are uniformly ultimately bounded. Compared with other approaches, no joint acceleration measurement and exactly known inertia matrix are required. Both theory and simulation results show the effectiveness of the proposed controller.