基于BP神经网络-牛顿迭代法的6-DOF并联机器人正解  

Forward Kinematics Solution of 6-DOF Parallel Robot Based on BP Neural Network and Newton Hybrid Algorithm

作　　者：段志琴 高宏力[1] 董林威 DUAN Zhiqin;GAO Hongli;DONG Linwei(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学机械工程学院,四川成都610031

出　　处：《机械制造与自动化》2023年第5期24-27,41,共5页Machine Building & Automation

摘　　要：由于6自由度并联机器人基座与末端执行器之间存在多条运动链,使其运动学正解难度较大,并且存在多解。针对并联机器人的运动学高效求解,提出一种基于神经网络和牛顿迭代法混合算法,利用神经网络模型的非线性映射能力,将输入杆长映射到上平台位姿,但映射出来的位姿精度较低,再利用迭代法求解,最后在Matlab中建立物理模型进行运算仿真。仿真结果表明:神经网络-牛顿混合求解能提高运算效率,并且有效降低误差,具有广泛的应用价值。To overcome the difficulty in forward kinematics solution of 6-DOF parallel robot due to its complicated kinematic chains between the bases of the parallel robot and the end effector,and aimed at the efficient solution of kinematics of the parallel robot against its multiple solutions,a hybrid algorithm based on neural network and Newton iteration is proposed.With the nonlinear mapping ability of the network model,the input rod length is mapped to the pose of the upper platform,and furthermore,the iterative method is used to solve the low accuracy of the mapped pose.The physical model is established in MATLAB for calculation simulation.The simulation results show that the neural network-Newton hybrid solution can improve the computational efficiency and effectively reduce errors,which has a wide range of application values.

关 键 词：6自由度并联机器人 BP神经网络 Newton-raphson迭代法 运动学正解 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]