A Machine Learning-Enhanced Digital Twin Approach for Human-Robot-Collaboration

Abstract

A key problem in human robot collaboration is a safe movement of the robot. The reason for this lies mainly in the variety of possible different events that can occur in an unstructured environment. Especially the description of a variable working space and the movements of humans are difficult to represent deterministically. In this paper, an approach to machine learning to enable industrial robots to bypass obstacles or people in the workspace is presented. First, a machine learning-enhanced robot control strategy is presented, which combines a nearest neighbor approach for path planning, clustering analysis and artificial neural networks for obstacle detection. Finally, a proof of concept is presented describing adaptive path planning for the protection of a human being.