Robot teaching by teleoperation based on visual interaction and neural network learning

Traditional methods of Robot teaching require human demonstrators to program with a teaching pendant, which is a complex and time-consuming exercise. In this paper, we propose a novel method based on teleoperation which allows a demonstrator to train robot in an intuitive way. More specifically, at the beginning the demonstrator controls a robot by visual interaction. And then a learning algorithm based on radial basis function (RBF) network is used to transfer the demonstrator's motions to the robot. To verify the effectiveness of this developed methods, several simulation experiments have been carried out which based on Microsoft Kinect Sensor and the Virtual Robot Experimentation Platform (V-REP). The experimental results show that this method has achieved satisfactory performance. With the help of this method, the robot can not only complete the task autonomously after teaching, but also can learn the details of demonstrator's behavior.