The Prediction of Positioning shift for a Robot Arm Using Machine Learning Techniques

Abstract

This study presents an Artificial Intelligence (AI) approach for estimating the Cartesian positioning shift of a wafer handling robot arm to prevent the occurrence of unexpected event, drop of wafers. First, a Charge-coupled Device (CCD) based robot arm fault diagnostic system was built to measure the target positions of the robot arm when handling wafers. An ensemble-based machine learning model with time series cross validation technique from a commercial software called Decanter AI (Mobagel Inc.) was applied to predict the quantity of the maximum position shift with respect to X and Y axis for next one minute. The prediction results by the test datasets through 38,417 minutes show that the Root Mean Square Error (RMSE) is 4.351 μm to validate the trained model is appropriate for predicting the positioning shift of the handling robot arm.