基于神经网络的机器人逆运动学算法

Arithmetic of inverse kinematics of robots based on neural networks

文章提出了一种基于神经网络改进算法的机器人逆运动学的求解方法。在讨论机器人逆运动学求解问题以及基于BP算法的神经网络基本原理的基础上,分析了几种BP算法改进方法,出了基于神经网络的机器给人逆运动学求解的具体步骤和设计神经网络的相关注意事项。与传统算法进行比较,通过对PR实验室机器人仿真研究表明,该算法具有简单、学习收敛速度快等特点,且避免了传统算法的许多棘手问题。

A new method of solving inverse kinematics of robots based on improvement of neural networks is presented in this paper. We discuss the problem about solving inverse kinematics of robot and the principle of ANN based on BP arithmetic, moreover, we discuss the principle and method of some kinds of improvement of BP arithmetic. Detailed step of this arithmetic and the proceeding that should be noticed are also discussed here. Compared with traditional methods, using simulation of PR robot (4-joint robot in lab), we can draw a conclusion that this arithmetic is a simple and fast-leaning method and many intractable problems in traditional way can be avoided.