基于神经网络的机器人自学习控制器

SELF LEARNING CONTROLLER BASED ON NEURAL NETWORKS FOR ROBOTIC MANIPULATOR

提出一种神经网络与ＰＩＤ控制相结合的机器人自学习控制器．为加快神经网络的学习收敛性，研究了有效的优化学习算法．以两关节机器人为对象的仿真表明，该控制器使机器人跟踪希望轨迹，其系统响应、跟踪精度和鲁棒性优于常规的控制策略．

This paper presents a new self learning controller based on neural networks for robotic manipulator.A fast learning algorithm of neural networks is proposed to improve both speed and convergence of the learning process.Simulation results of a two link robot show that the proposed method can give more significant performance and robustness than conventional approaches.