摘要
传统方法求解6DOF斜交非球型手腕喷漆机器人的逆解运算量大、精度低,不利于实现实时、精确控制;基于三层标准BP神经网络,针对其收敛速度慢、易陷入极小值等问题,采用增加动量项的方法进行改进,建立斜交非球型手腕喷漆机器人末端执行器位姿与各关节角之间的映射关系;利用Matlab对神经网络模型进行训练和验证,结果表明,BP神经网络学习精度高、收敛速度快,可满足6DOF喷漆机器人逆运动学的求解。
A three layers Back Propagation （BP）Articial Neural Network （ANN）has been proposed here to overcome the traditionally inverse kinematics problem,which are mainly heavy computation,low precision and being can＇t realize real-time ,accurate control in 6R serial spraying manipulator with non-spherical wrist.Considering the defect of slow convergent speed and easy convergence to a local minimun point of error function in the typical feed-forwad ANN,an improved algorithm by adding items of the momentum was established to map the relation between the 6R serial spraying manipulator end-effectar＇s pose and joint angles.Then the improved BP ANN was trained and verified through Matlab,and the result shows that the P ANN has a higher accuracy,slow convergent speed and the capacity of the network could satisfy the solution to the inverse kinematics problem of 6R werial wpraying manipulator.
出处
《机械设计与制造》
北大核心
2012年第9期170-172,共3页
Machinery Design & Manufacture
基金
河北省教育厅项目(2009335)
