一种鲁棒的机械臂动力学参数辨识方法

Identification method for dynamic parameters of manipulators based on robust

针对机械臂动力学参数在传统辨识方法下存在辨识精度不高且易受异常数据点影响的问题,提出一种鲁棒的机械臂动力学参数辨识方法。首先,采用牛顿-欧拉法建立机械臂动力学方程,推导得到机械臂动力学线性化重组模型,确定需要辨识的惯性参数最小集合;其次,设计激励轨迹,采用遗传算法优化激励轨迹参数;再次,采用Tent混沌映射对传统粒子群优化算法(PSO)的初始种群位置进行改进,自适应惯性权重和学习因子,同时设计残差权重策略来剔除辨识过程中的异常数据点;最后,采集数据进行参数辨识试验。辨识结果表明:所提方法增强了对异常数据点的鲁棒性,有效提高了辨识精度,与随机权重粒子群算法(RWPSO)相比,文中所提改进粒子群优化算法(IPSO)的残差均方根(RMS)平均减小了10.0643%,相关系数ρ平均增大了0.8273%。

Since it is not accurate to identify the dynamic parameters of manipulators and the results are susceptible to abnormal data points with the help of the traditional identification method,in this article,the identification method for dynamic parameters of manipulators based on robust is proposed.Firstly,the Newton-Euler method is used to set up the dynamic equation of the manipulator;then,the linearized recombination model is derived,and the minimum set of the inertial parameters to be identified is worked out.Secondly,the excitation trajectory is designed,and the genetic algorithm is applied to optimizedtheexcitation-trajectoryparameters.Thirdly,the initial population location of the traditional PSO is improved through the Tent chaos map,adaptive inertia weights and learning factors,and a residual weighting strategy is designed to eliminate anomalous data points in the identification process.Finally,the data is collected for the experiments of parameter identification.The results show that this method is helpful to enhance robustness of abnormal data points and ensure higher accuracy in identification.The RMS of the improved IPSO reduces by 10.0643%on average and the averageρimproves by 0.8273%,compared with those ofthe RWPSO.