Research on Multi-joints Motion Planning Method by Online Auto-Learning Mode Based on Neural Network

Abstract

The multi-joints motion planning method by online autolearning mode based on neural network which can realize new trajectories' planning and control of multi-joints manipulator is extensively used in the field of trajectory tracking, planning and control of intelligent robot, especially for rigid manipulator to achieve the discovery of new trajectory, real-time self-learning and control itself. Such method can achieve the real-time self-learning and control of nonlinear complex trajectories, by taking advantage of the global optimal approximation performance of the neural network. The functional relationship can be established between the current trajectory information and trajectory information at the previous N moments by neural network, and because of this, information related to trajectory can be predicted, thereby realizing on-line self-learning of multi-joints. This method is used for the real-time control of intelligent mechanical arm which can reduce the difficulty of numerical solution greatly, improve the efficiency of calculation and boost the ability of real-time self-learning.