摘要
利用径向基函数(RBF)神经网络逼近能力较优和收敛速度较快的特点将运动学逆解过程转换为神经网络权值的训练过程。在训练RBF网络时,采用正交最小二乘(OLS)算法来确定网络中心,并将正解结果作为训练样本,实现了冗余扫查机器人运动学逆解的计算,避免了传统方法的繁琐公式推导及数值病态问题。
The RBF (radial basis function) network adopted transformed the solution of inverse kinematics into the training process of network weights. For selecting the center of RBF network,an orthogonal least square (OLS) algorithm, which can avoid oversize dimensions of RBF network and value ill conditioning, was utilized. Meanwhile, a lot of forward kinematial results were applied as the training data set of RBF network, with which the inverse kinematics results of joint variables were obtained.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2006年第17期1765-1768,共4页
China Mechanical Engineering
基金
国家863高技术发展研究计划资助项目(2002AA442110)
上海电机学院青年教师科研基金资助项目