A high-precision joint prediction method for robotic machining error decomposition embedded with error decomposition

Résumé fourni par la source

Industrial robots have become another important manufacturing equipment, following machine tools due to their large working space and other advantages. Affected by geometric and compliance errors, robotic machining errors are very large, significantly impacting the application of robotic machining in scenarios that rely on precision. This paper focuses on the systematic components and condition-related components of machining errors, proposing a high-precision joint prediction method that embeds the decomposition of machining errors. The original machining errors are decomposed into low-frequency systematic components and higher-frequency condition-related components based on their contribution. For the former, a state-based Gaussian process regression model is constructed, mapping joint states and eigenvalues of generalized forces to the machining error space; for the latter, a regression model based on time convolutional neural networks is constructed, taking joint angles and sequences of generalized forces as the raw input, and predicting machining errors related to working conditions through high-dimensional feature extraction and mapping. The proposed method was tested in the task of planar groove milling. The advantages of the proposed method over current methods were verified through the construction of interpolation and extrapolation tasks. In the two constructed test tasks, the maximum RMSE was only 2.241[Formula: see text][Formula: see text]m, which further confirms the effectiveness of the proposed method.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.