A Novel Robot Teaching System Based on Mixed Reality

Abstract

In this paper, we have developed a robot teaching system where the robot learns the point-to-point motions from human demonstrations. In the demonstration phase, the operator's gestures and palm position are used to guide the robot to complete the task. At the same time, the scene of robots and its work environment, and the virtual model of the palms are integrated into Unity. Then the mixed reality scene is transferred to a virtual reality(VR) helmet, which provides operator real-time visual feedback. In the learning and reproduction phase, the Extreme Learning Machine(ELM) is used to generate a new trajectory from the training data. The experimental result shows that the robot teaching system based on mixed reality has more realistic and natural interaction. And the learning algorithm based on ELM has a good fitting ability.