基于径向基函数神经网络的并联机器人运动学正问题

以一般形式的Stewart型并联机器人为例,由机器人的位置反解问题引出机器人运动学正解问题,在分析BP网络与径向基函数网络的特点基础上,采用基于径向基函数神经网络的算法,利用最近邻聚类方法获得径向基函数中心,求解并联机器人运动学正解问题·通过对训练样本的学习,确定神经网络权系数,能够准确地求解并联机器人的位置和姿态,算法具有运算简单,求解效果好等特点·同Newton Raphson算法比较,能获得相同的效果且位置和姿态误差近似恒定,而神经网络算法避免迭代初值及额定循环次数的影响·因此该方法可作为并联机构系统运动学轨迹跟踪控制的运动学模型辨识器·

Direct Kinematic Problem Based on RBFNN of Parallel Manipulator

The kinematic model of paraller manipulators presents an inherent complexity due to their closed-loop structure and kinematic constraints and, correspondingly, the algorithm for the solution to relavant direct kinematic problems is complex. An algorithm of RBFNN for kinematic problems is thus presented in a nearest neighbor-clustering way to solve simply the positions and orientations of Stewart platform with needed precision, of which the process of solution is simple with good result provided. Comparing with the Newton-Raphson algorithm, the errors of neural network are appproximatively invariable without the effects of initial values and rated number of cycles on them. The results show that this approach can be used for the in-line control of parallel manipulators.