基于仿生优化神经网络的6DOF工业机械手标定  

Calibration of 6DOF Industrial Manipulator Based on Bionic Optimized Neural Network

作　　者：曹俊 任吉慧 邓绯[3] CAO Jun;REN Jihui;DENG Fei(Education and Teaching Supervision Office,Sichuan TOP IT Vocational Institute,Chengdu Sichuan 611743,China;School of Mechanical Engineering,Sichuan University Jinjiang College,Meishan Sichuan 620860,China;School of Computer Engineering,Sichuan Vocational and Technical College,Suining Sichuan 629000,China)

机构地区：[1]四川托普信息技术职业学院教育教学督导办公室,四川成都611743 [2]四川大学锦江学院机械工程学院,四川眉山620860 [3]四川职业技术学院计算机工程学院,四川遂宁629000

出　　处：《机床与液压》2024年第15期69-74,共6页Machine Tool & Hydraulics

基　　金：四川省教育厅2022年度科研计划项目(22ZB0407)。

摘　　要：针对六自由度工业机械手标定中的定位精度问题,提出一种基于仿生优化神经网络的机械手标定方法。研究六自由度工业机械手的运动学模型,并给出其D-H参数。通过将关节挠度模型和传统的运动学模型标定技术相结合,来同步识别机械手的运动学参数和柔度参数,以提高定位精度。然后,构造人工神经网络对未建模误差进行进一步补偿,如摩擦、机械传动误差和热膨胀。此外,采用入侵杂草优化算法对神经网络的权值和偏置进行优化。最后,采用六自由度机械手HX300对所提方法进行了实际测试,验证其可行性。研究结果表明:标定后机械手的定位精度得到了显著提高,平均误差、最大误差和标准差分别为0.345、0.6374、0.1624 mm,均小于其他标定方法;与GA-BP神经网络标定方法相比,所提方法具有更好的收敛能力,平均误差降低了15.92%,适用于高精度的各种工业应用。Aiming at the positioning accuracy problem of 6DOF industrial manipulator calibration,a manipulator calibration method based on bionic optimization neural network was proposed.The kinematics model of 6DOF industrial manipulator was studied,and its D-H parameters were given.By combining the joint deflection model with the traditional kinematics model calibration technology,the kinematics parameters and flexibility parameters of the manipulator could be identified synchronously to improve the positioning accuracy.Then,an artificial neural network was constructed to further compensate the unmodeled errors,such as friction,mechanical transmission error and thermal expansion.In addition,the invasive weed optimization algorithm was used to optimize the weight and bias of neural network.Finally,a 6DOF manipulator HX300 was used to test the proposed method,and its feasibility was verified.The research results show that the positioning accuracy of the manipulator is significantly improved after calibration.The average error,maximum error and standard deviation are 0.345 mm,0.6374 mm,0.1624 mm,respectively,and they are all smaller than other calibration methods.Compared with GA-BP neural network calibration method,the proposed method has better convergence ability,and the average error is reduced by 15.92%,which is suitable for various high-precision industrial applications.

关 键 词：工业机械手 参数标定 关节柔度 误差补偿 仿生优化算法 

分 类 号：TH133.33[机械工程—机械制造及自动化] TP242.2[自动化与计算机技术—检测技术与自动化装置] V231.92[自动化与计算机技术—控制科学与工程]