摘要
针对传统BP神经网络算法应用于机器人逆运动学求解时存在的因易陷入局部极值导致输出误差偏大的问题,该文提出了一种基于PSO优化的BP神经网络在求解机器人逆运动学中的应用。首先通过PSO算法迭代计算粒子适应度;其次,依据个体极值和群体极值的不断更新得到最优的BP网络初始权值阈值。该方法避免了局部极值问题并且加快了BP网络训练过程的收敛速度。实验结果表明,采用论文提出的方法对机器人逆运动学的求解得到误差小于0.1°的关节角输出。
Aiming at the traditional BP neural network algorithm is applied to the robot inverse kinematics solution is due to fall into the local extremum of the output error caused,this paper proposed a BP neural network based on PSO optimizing in solving robot inverse kinematics should be used. Firstly through the PSO algorithm iteration calcu- lation of particle fitness. Secondly,according to the individual and group extreme constantly updated get the optimal initial weights of BP neural network threshold. The method avoid local extremum problem and accelerate the conver- gence speed of the BP network training process. The experimental results show that the proposed method is used to solve the inverse kinematics of the robot,and the output error is less than 0.1 degrees.
作者
赵建强
刘满禄
王姮
ZHAO Jian-qiangl LIU Man-lu WANG Heng(Special Environment Robot Technology Key Laboratory of Sichuan Province,Southwest University of Science and Technology,Mianyang 621000,China School of Information Science and Technology,University of Science and Tech- nology of China,Hefei 230026,China)
出处
《自动化与仪表》
2016年第11期1-6,共6页
Automation & Instrumentation
基金
四川省科技支撑计划项目(2015GZ0027)
