基于主成分分析法和核主成分分析法的机器人全域性能综合评价

Comprehensive Evaluation of Robotic Global Performance Based on Principal Component Analysis and Kernel Principal Component Analysis

在机器人运动学和动力学性能评价中,表示机器人运动学和动力学性能的指标众多,全域性能指标是其中一项重要的评价指标,而全域性能指标又包括:线速度全域性能指标、角速度全域性能指标等指标.不同指标间往往存在不同程度的相关性,其中有些相关性非常显著,这使它们提供的信息有可能发生重叠.引入统计学原理,依据线性降维与非线性降维原则,应用主成分分析法(principal component analysis,PCA)和核主成分分析法(kernel principal component analysis,KPCA)对不同尺度的PUMA560机器人的全域性能进行综合评价,从而选择综合全域性能最优的机器人.计算结果表明:KPCA方法较PCA方法有更好的降维效果,能够更有效地处理多个单一性指标间的非线性关系,提供更多的综合全域性能评价信息,可为建立机器人综合全域性能与其尺度之间的数值计算关系,为基于综合全域性能指标最佳尺度选取的研究提供科学的参考依据.

In evaluation of robotic kinematic and dynamic dexterity performance, there are many different indexes to express robotic kinematic and dynamic dexterity, the global performance index is one of the important evaluation index, including acceleration, angular acceleration and linear acceleration performance index. Those different indexes tend to have different degrees of correlations among them. Some correlations are very remarkable, therefore, the provided information may be overlapped. This paper describes comprehensive evaluation of global performance of the PUMA560 robot with different scales by using principal component analysis (PCA) and kernel principal component analysis (KPCA), which is characterized by linear dimension reduction and nonlinear dimension reduction principle, then the best robotic scales with comprehensive global performance can be selected. Results show that KPCA method has more effective reduction effects, and can reveal the nonlinear relationship among different single performance indexes to provide more comprehensive global performance evaluation information, which can reveal the numerical calculation retationships among comprehensive global performance and scales, and provide scientific reference for selection of the best robotic scales based on comprehensive global performance indexes.