基于遗传-拟牛顿算法的机器人几何参数辨识

Research on Geometric Parameter Identification of Industrial Robot Based onGenetic and Quasi Newton Algorithms

为提升工业机器人绝对定位精度,提出了一种遗传-拟牛顿算法来辨识机器人几何运动参数并进行机器人绝对定位误差补偿的联合算法。首先,利用MD-H方法建立机器人运动模型,并通过微分变化原理建立机器人各个关节误差转换到机器人末端误差模型;其次,通过激光跟踪仪测得KR16-2型机器人在指定工作空间50处采样点信息,通过传统遗传算法和遗传-拟牛顿联合算法分别对机器人几何参数进行辨识和比较,并在此算法中提出小区间生成初始种群的方式来提高迭代效率,通过算数交叉和动态变异完成新个体的生成,解决了传统遗传算法出现过早收敛问题。结果表明,此算法对机器人进行补偿后的精度分别提高71.28%和53.23%,直线制孔平均误差是0.45 mm,弧面制孔平均误差是0.42 mm,验证了联合算法的优越性和实用性。

In order to improve the absolute positioning accuracy of industrial robots,a combination algorithm of genetic and quasi-Newton algorithm was proposed to identify the geometric motion parameters of robots and compensate the absolute positioning errors.Firstly,the MD-H method is used to establish the robot motion model,and the error model of each joint is established by differential change principle.Secondly,the information of 50 sampling points in the designated work space of the KR16-2 robot was measured by the laser tracker.The geometric parameters of the robot were identified and compared respectively by the traditional genetic algorithm and the combined Gene-quasi-Newton algorithm.In this algorithm,the method of generating initial population between cells was proposed to improve the iteration efficiency.New individuals are generated by arithmetic crossing and dynamic mutation,which solves the problem of premature convergence in traditional genetic algorithm.The results show that the accuracy of the compensated robot is improved by 71.28%and 53.23%,respectively.The average error of linear hole making is 0.45 mm,and the average error of curved hole making is 0.42 mm,which verifies the superiority and practicability of the combined algorithm.