Comprehensive Evaluation Of Robotic Performance Based On Multivariate Statistical Analysis

The key to improve the robotic properties and intelligent controlling level is tooptimize the robotic performance constantly, while the core problems are how toevaluate the robotic performance reasonably and how to control the robot fine.Therefore, it’s important to research the comprehensive evaluation method for roboticperformance based on single performance indexes. The principal research contents areas follows:(1) Applicability of Principal Component Analysis (PCA). The commonly usedstatistical evaluation methods are analyzed. Due to the correlation and diversity of therobotic single performance indexes, PCA is much more applicable for comprehensiveevaluation for robotic performance. Further, for the shortcomings of traditional PCA,the Kernel Principal Component Analysis (KPCA) is introduced in comprehensiveevaluation for robotic performance. Gray relative analysis (GRA) is a comprehensiveselection problem, so GRA is introduced to prove PCA is feasible and practicable.(2) Comprehensive evaluation of the robotic performance. Kinematic dexterityand global performance indexes of robot are summarized, while their calculationmethod and physical meaning are explained. Comprehensively evaluates the roboticperformance as the follow two kinds of situations: first, known the dimension andconfiguration of robot, comprehensively evaluates the robotic kinematic dexterityperformance when the robot completes task at the different positions in the workspace,second, known the operation task, comprehensively evaluates the performance ofrobot with the different configurations or the same configuration with differentdimensions.(3) Comprehensive performance evaluation index. Through the comprehensiveevaluation of the above two kinds of situations of the robotic performance, thecorrelation among single robotic indexes are found, and the evaluation method forrobotic comprehensive performance are built, which initially provide scientificresearch basis for the optimum order of configurations and tasks positions based oncomprehensive performance evaluation index.The study above has developed a set of theory and method for roboticperformance’s comprehensive evaluation based on PCA or its extension methods,analyzing the correlations among each single performance indexes, and providingscientific research basis for the optimum order of task positions and the robotic dimensional.