作者：乔贵方1,2, 田荣佳1, 张颖1, 王保升3, 宋光明2, 宋爱国2
作者单位：1. 南京工程学院自动化学院，江苏 南京 211167;
2. 东南大学仪器科学与工程学院，江苏 南京 210096;
3. 南京工程学院 智能制造装备研究院，江苏 南京 211167
关键词：工业机器人;运动学误差;差分进化算法;机器人标定;径向基神经网络
摘要：
针对串联型工业机器人的绝对定位误差较大的问题，该文提出一种两级误差标定方法，该方法融合了误差模型法和基于径向基(radial basis function, RBF)神经网络的非模型标定方法。首先，基于M-DH（modified DH）运动学模型建立串联型工业机器人的位姿误差模型，并基于差分进化 (differential evolution, DE)优化算法实现M-DH运动学参数误差的辨识，将TX60机器人的平均综合位置/姿态误差从(0.5368 mm，0.1745°)降低为(0.1772 mm，0.0875°)。其次，为了进一步提升机器人的精度性能，利用RBF神经网络拟合预测TX60机器人的剩余误差，该方法将机器人的平均综合位置/姿态误差从(0.2178 mm，0.0863°)降低为(0.1044 mm，0.0411°)。最后，通过实验验证基于两级误差标定方法的精度提升效果要优于单一的基于RBF神经网络的误差标定方法，平均综合位置/姿态误差降低比例分别是4.9%和14.9%。因此，该文提出的两级误差标定方法能够有效地提升机器人的精度性能。
Research on double-stage positional error calibration method of serial industrial robot
QIAO Guifang1,2, TIAN Rongjia1, ZHANG Ying1, WANG Baosheng3, SONG Guangming2, SONG Aiguo2
1. Automation Department, Nanjing Institute of Technology, Nanjing 211167, China;
2. School of Instrument Science and Engineering, Southeast University, Nanjing 210096, China;
3. Research Department of Intelligent Manufacturing Equipment, Nanjing Institute of Technology, Nanjing 211167, China
Abstract: Aiming at the problem of large absolute positioning error of serial industrial robots, this paper presents a double-stage error calibration method which combines the error model based method and the non-model calibration method based on RBF neural network. Firstly, the pose error model of the serial industrial robot is established based on the M-DH kinematics model. The errors of M-DH kinematic parameter are identified based on the DE optimization algorithm. The average comprehensive position/attitude errors of TX60 robot are reduced from (0.5368 mm, 0.1745°) to (0.1772 mm, 0.0875°). Secondly, the RBF neural network is used to fit and predict the residual error which can further improve the accuracy performance of TX60 robot. This method reduces the average comprehensive position/attitude errors of TX60 robot from (0.2178 mm, 0.0863°) to (0.1044 mm, 0.0411°). Finally, the experiments have been done to verify the accuracy improvement effect based on the double-stage error calibration method is better than the single-stage error calibration method based on the RBF neural network. The average comprehensive position/attitude errors respectively decrease by 4.9% and 14.9%. The experimental results prove that the double-stage error calibration method proposed in this paper can effectively improve the accuracy performance of robot.
Keywords: industrial robot;kinematic errors;DE algorithm;robot calibration;RBF neural network
2022, 48(7):134-139,146 收稿日期: 2021-11-09;收到修改稿日期: 2022-01-17
基金项目: 国家自然科学基金项目（5190525）；中国博士后科学基金（2019M650095）；江苏省高等学校自然科学研究面上项目（17KJD460006）
作者简介: 乔贵方（1987-），男，江苏新沂市人，副教授，博士，主要从事机器人标定等研究
参考文献
[1] 中华人民共和国国务院. 国务院关于印发《中国制造2025》的通知[EB/OL]. (2015-5-19). http://www.gov.cn/zhengce/content/2015-05/19/content_9784.html
[2] LIN Y, ZHAO H, DING H. Posture optimization methodology of 6R industrial robots for machining using performance evaluation indexes[J]. Robotics and Computer-Integrated Manufacturing, 2017, 48: 59-72
[3] JIAO J, TIAN W, LIAO W, et al. Processing configuration off-line optimization for functionally redundant robotic drilling tasks[J]. Robotics and Autonomous Systems, 2018, 110: 112-123
[4] LEHMANN C, PELLICCIARI M, DRUST M, et al. Machining with industrial robots: the COMET project approach[J]. Robotics in Smart Manufacturing. Springer, 2013, 371(4): 27-36
[5] SUN T, LIAN B, YANG S, et al. Kinematic calibration of serial and parallel robots based on finite and instantaneous screw theory[J]. IEEE Transactions on Robotics, 2020, 36(3): 816-834
[6] YANG P, GUO Z, KONG Y. Plane kinematic calibration method for industrial robot based on dynamic measurement of double ball bar[J]. Precision Engineering, 2020, 62: 265-272
[7] LI Z, LI S, LUO X, et al. An overview of calibration technology of industrial robots[J]. IEEE/CAA Journal of Automatica Sinica, 2021, 8(01): 27-40
[8] SANTOLARIA J, CONTE J, PUEO M, et al. Rotation error modeling and identification for robot kinematic calibration by circle point method[J]. Metrology and Measurement Systems, 2014, 21(1): 85-98
[9] ZHANG T, SONG Y, WU H, et al. A novel method to identify DH parameters of the rigid serial-link robot based on a geometry model[J]. Industrial Robot, 2020, 48(1): 157-167
[10] HAYATI S, MIRMIRANI M. Improving the absolute positioning accuracy of robot manipulators[J]. Journal of Robotic Systems, 1985, 2(4): 397-413
[11] JIANG Z, GAO W, YU X. An improved robot calibration method using the modified adjoint error model based on POE[J]. Advanced Robotics, 2020, 34(19): 1229-1238
[12] 乔贵方, 万其, 吕仲艳, 等. 基于ZRM-MDH模型转换的串联机器人运动学参数标定[J]. 农业机械学报, 2021, 52(3): 382-389
[13] 陆艺, 沈添秀, 郭小娟, 等. 线结构光视觉传感器机器人手眼关系标定[J]. 中国测试, 2019, 255(10): 6-9
[14] 刘湛基, 王晗, 陈桪, 等. 机器人与激光跟踪仪的坐标系转换方法研究[J]. 中国测试, 2017, 232(11): 102-107
[15] 乔贵方, 吕仲艳, 张颖, 等. 基于BAS-PSO算法的机器人定位精度提升[J]. 光学精密工程, 2021, 29(4): 763-771
[16] 温秀兰, 崔俊宇, 芮平, 等. 工业机器人几何参数混合优化辨识与标定[J]. 组合机床与自动化加工技术, 2018, 533(7): 23-26
[17] 李宇飞, 田威, 李波, 等. 机器人铣削系统精度控制方法及试验[J]. 航空学报, 2022, 43(5): 625815-625815
[18] BAI Y. On the comparison of model-based and modeless robotic calibration based on a fuzzy interpolation method[C]//IEEE Conference on Robotics, Automation and Mechatronics. IEEE, 2004.
[19] CAI Y, YUAN P, SHI Z, et al. Application of universal kriging for calibrating offline programming industrial robots[J]. Journal of Intelligent & Robotic Systems, 2019, 94: 339-348
[20] LI B, TIAN W, ZHANG C, et al. Positioning error compensation of an industrial robot using neural networks and experimental study[J]. Chinese Journal of Aeronautics, 2022, 35(2): 346-360
[21] 董成举, 刘文威, 李小兵, 等. 六轴工业机器人工作空间分析及区域性能研究[J]. 中国测试, 2020, 262(5): 154-160
[22] 田威, 廖文和, 工业机器人精度补偿技术与应用[M]. 北京: 科学出版社, 2021.
