Algorithm based on neural network for inverse kinematics
of redundant manipulator

OU Qunwen1, YUN Chao1, YANG Xuebing2, HU Zhangwen3

(1. School of Mechanical Engineering and Automation, Beihang University, Beijing 100191, China;

2. Beijing Wise Welding Technology & Engineering Co. Ltd., Beijing 100076, China;

3. Beijing SANY Heavy Industry Co. Ltd., Beijing 100220, China)

Abstract: Aiming at the difficulties of gaining inverse kinematics solution of redundant manipulator, a new method based on RBF neural network was proposed. The forward kinematics equation was established for a new 7DOF manipulator. A rule of “best compliance” based on weighted least square method was supposed. Based on this rule, genetic algorithm was used to search for the global optimal solution, which was used as the training data of artificial neural network. Then the neural network was trained to achieve a stable state. The simulation and experiment were designed to test the RBF neural network, and the performance of the neural network was analyzed. The results indicate that the RBF neural network has high precision and fast convergence speed, and it can provide a new method for solving the inverse kinematics solution of any redundant manipulator.

Key words: genetic algorithm; RBF neural network; redundant degree of freedom; inverse kinematics