China Mechanical Engineering ›› 2023, Vol. 34 ›› Issue (20): 2456-2465.DOI: 10.3969/j.issn.1004-132X.2023.20.008

Previous Articles     Next Articles

Adaptive Kalman Filtering and PSO-GA-BP Algorithm for Robot Error Compensation

LI Guangbao1,2;GAO Dong1;LU Yong1;PING Hao2;ZHOU Yuanyuan2   

  1. 1.School of Mechanical and Electrical Engineering,Harbin Institute of Technology,Harbin,150000
    2.Shanghai Aerospace Precision Machinery Research Institute,Shanghai,201600
  • Online:2023-10-25 Published:2023-11-20

自适应卡尔曼滤波与PSO-GA-BP算法的机器人误差补偿

李光保1,2;高栋1;路勇1;平昊2;周愿愿2   

  1. 1.哈尔滨工业大学机电工程学院,哈尔滨,150000
    2.上海航天精密机械研究所,上海,201600
  • 作者简介:李光保,男,1995年生,博士研究生。研究方向为航空航天制造技术。发表论文15篇。E-mail:18363998150@163.com。
  • 基金资助:
    国家重点研发计划(2018YFB1306803)

Abstract:  The cutting holes of a certain type of launcher were machined by a seven-axis robot device clamping laser. In the machining processes due to the low trajectory accuracy and absolute positioning accuracy, it was easy to cause damage and error cutting to the launcher of the model products. The ideal model of the seven-axis robot was established using the D-H algorithm, and the ideal model was verified by the numerical algorithm of forward and inverse kinematics. Based on Sage-Husa adaptive Kalman filter, the theoretical pose parameters of the ideal model and the measured pose parameters of the laser tracker were used to solve the real pose coordinate information of the seven-axis robots, and the joint errors of the ideal pose parameters and the real pose coordinate information were obtained. Then, the error prediction model of the seven-axis robots was established by combining the PSO-GA-BP joint algorithm. The theoretical pose parameters of the seven-axis robots were used as input samples. The joint angle differences between the real pose and the theoretical pose were taken as the output sample. The joint angle values of the seven-axis robots were compensated according to the model output values by KUKA Robot Workvisual 5.0 software. Through simulation experiments and machining processes, the trajectory errors and absolute positioning errors of the seven-axis robots after the joint error compensation are decreased by 72%, meeting the production requirements. 

Key words: laser cutting, seven-axis robot, error compensation, particle swarm optimization-genettic algorithm-back propagation (PSO-GA-BP), Sage-Husa adaptive Kalman filtering

摘要: 采用七轴机器人设备夹持激光器的方式对某型号发射筒进行切割开孔加工。在加工过程中,因轨迹精度和绝对定位精度较低,容易对型号产品发射筒产生损伤和误差切割等问题,运用D-H算法建立七轴机器人理想模型,运用正逆运动学数值算法对理想模型进行验证,运用理想模型的理论位姿参数和激光跟踪仪的测量位姿参数基于Sage-Husa自适应卡尔曼滤波求解七轴机器人真实位姿坐标信息,得到理想位姿参数和真实位姿坐标信息的关节误差,然后结合粒子群优化-遗传算法-BP神经网络联合算法对七轴机器人建立误差预测模型,采用七轴机器人理论位姿参数作为输入样本,真实位姿与理论位姿的各关节角度差作为输出样本,通过库卡机器人Workvisual 5.0软件按照模型输出值对七轴机器人的各关节角度值进行补偿。经过仿真实验和加工,各关节误差补偿后的七轴机器人轨迹误差和绝对定位误差减小72%,满足工艺要求。

关键词: 激光切割, 七轴机器人, 误差补偿, 粒子群优化-遗传算法-BP, Sage-Husa自适应卡尔曼滤波

CLC Number: