Neural network control of networked redundant manipulator system with weight initialization method

摘要

In this paper, a novel controller is proposed to reach tracking synchronization in task space for networked redundant manipulators in circumstance of unknown system parameters, disturbance and sub-task requirements. We notice that the initial neural weights of neural network controller in existing literatures are sloppily selected which may have influence on tracking performance. In the proposed controller, a universal method is proposed to carefully assign the initial neural weights that are commonly close to the ideal values, and consequently the tracking performance can be improved. Meanwhile, input dimension of neural network is reduced and approximability of neural network is ensured. Simulations are given to show the effectiveness of the proposed controller. (C) 2018 Elsevier B.V. All rights reserved.