论文部分内容阅读
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。
Aiming at the robot modeling with inaccurate modeling, a robot trajectory tracking adaptive control method based on neural network compensation is proposed. In this paper, by designing neural network compensator and adaptive robust control term, the uncertainty of the model is effectively compensated Part and network approximation error. Because the algorithm contains robust control items that compensate the approximation error of neural network, the requirements on the size of neural network can be reduced in practical application. And the neural network connection right is adjusted online without the need of offline learning process. The theory shows that the algorithm can guarantee the tracking error and the neural network connection weight finally consistent bounded, the simulation results also verify the effectiveness of the algorithm.