Neural learning and Kalman filtering enhanced teaching by demonstration for a Baxter robot

Abstract

In this paper, Kalman filter has been successfully carried out to fuse the data obtained from a Kinect sensor and a pair of MYO armbands. To do this, the Kinect sensor is used to capture movements of operators which is programmed by Microsoft Visual Studio. Operator wears two MYO armbands with the inertial measurement unit (IMU) embedded to measure the angular velocity of upper arm motion for the human operator. Additionally a neural networks (NN) control upgraded Teaching by Demonstration (TbD) technology has been designed and it also has been actualized on the Baxter robot. A series of experiments have been completed to test the performance of the proposed technique, which has been proved to be an executed approach for the Baxter robot's TbD has been designed.