Incremental Learning Method for Robot Error Based on GWO-XGBoost Algorithm

In this paper, a method based on the GWO-XGBoost incremental learning model is proposed to improve the robot’s positioning accuracy. It is crucial to reduce the positioning error of the robot. In addition, the trajectory and position of the robot are always changing during the working process. Therefore, to address this problem, inspired by the idea of online learning, the XGBoost incremental learning model optimized with improved grey wolf algorithm is used to predict the positioning error of industrial robots. Firstly, the Grey Wolf algorithm is used to determine the optimal hyper-parameters of XGBoost model, then XGBoost is used for modelling prediction, and finally the accuracy and generalization ability is improved by online incremental learning. The GWO-XGBoost model has good fitting and prediction ability, the model is validated on Universal robots by online incremental training with good results.