基于改进霜冰算法的并联机器人误差建模与参数辨识

Error Modeling and Parameter Identification of Parallel Robots Based on Modified Rime Optimization Algorithm

摘要:

针对当前并联机器人运动学标定存在误差模型复杂和参数辨识效率不高等问题，本文提出了一种并联机器人误差建模方法和高效稳定的参数辨识算法。基于闭环矢量法完成了并联机器人运动学分析；在此基础上提出等效误差思想并建立相应误差模型；根据参数辨识算法高辨识精度要求对霜冰算法（Rime optimization algorithm，RIME）收敛精度低的缺陷进行改进，提出了均分法、莱维选择算子以及交替正余切策略来改进其初始化性能、全局优化能力以及局部优化能力，并以改进霜冰算法（Modified rime optimization algorithm，MRIME）进行误差参数辨识；根据辨识结果对机器人驱动输入进行补偿。以Delta机器人为研究对象进行标定实验，实验结果表明，改进霜冰算法提升了寻优效率、精度、稳定性，参数辨识平均耗时0.126s，标定后机器人平均位置精度提升41.96%，验证了所提误差模型和参数辨识算法的有效性。

Abstract:

Aiming at the problems of complex error model and low efficiency of parameter identification in the kinematic calibration of parallel robots, an error modeling method and an efficient, stable parameter identification algorithm were proposed for parallel robots. Kinematic analysis of the parallel robot was completed based on the closed-loop vector method. Building upon this foundation, the concept of equivalent errors was introduced, and a corresponding error model was established. To meet the high-precision requirements of parameter identification algorithms, improvements were made to overcome the low convergence accuracy deficiency of the rime optimization algorithm (RIME). Specifically, the bisection method, Lévy selection operator, and alternating sine-cosine strategy were proposed to enhance its initialization performance, global optimization capability, and local optimization capability. The improved algorithm, termed modified RIME (MRIME), was then employed for error parameter identification. Based on the identification results, compensation was applied to the robot’s drive inputs. Calibration experiments were conducted by using a Delta robot as the study object. Experimental results demonstrated that the modified RIME algorithm significantly improved optimization efficiency, accuracy, and stability. The average parameter identification time was 0.126s. After calibration, the robot’s average positional accuracy was improved by 41.96%. These results validated the effectiveness of the proposed error model and parameter identification algorithm.