摘要
采用分类神经网络形式,利用运动学逆解,通过遗传算法结合Levenberg-Marquardt训练方法,可实现机器人位置从关节变量空间到工作变量空间的非线性映射,从而求得并联机器人运动学正解估计值,然后通过拟牛顿迭代计算可求得精确解,将此方法应用于6-PRRS并联机器人,结果表明:该方法计算精度高,耗时少,可应用于并联机器人的任务空间实时控制或求解并联机器人的工作空间。
The solution of FK involves the solving of a series of simultaneous non-linear equations and, usually, non-unique, multiple sets of solutions are obtained from one set of data. In this paper, a classed neural network was trained by genetic arithmetic combined with Levenberg - Marquardt method to recognize the relationship between the joints space and the Cartesian space of parallel robot, and provided the evaluation of Fk in a range with a little error. By performing a few iterations using the Quasi-Newton method, the solution can be obtained fast and reliably. This method is applied in 6- PRRS parallel robot, the result shows this method can get a high precision value and spends a little time, so it can by applied in the real time control on task space or in getting the work space of parallel robot.
出处
《吉林大学学报（工学版）》
EI
CAS
CSCD
北大核心
2008年第3期731-734,共4页
Journal of Jilin University:Engineering and Technology Edition
基金
长江学者和创新团队发展计划项目(IRT0423)
关键词
自动控制技术
并联机器人
正运动学
神经网络
拟牛顿法
任务空间
automatic control technology
parallel robot
forward kinematics
neural network
quasi-Newton method
task space
