Imitation learning of robots by integrating Microsoft kinect and PID controller with a sensor for angular displacement in a robot joint

The process of gesturing in Human Robot Interaction has a complex character and can't be pre-programmed explicitly. Nowadays, teaching robots is a well established concept ranging from demonstration by variants of teleoperation to imitation by external observations. We propose an innovative approach for imitation learning of robots by integrating Microsoft Kinect motion-sensing device and PID Controller with direct feedback from an originally developed sensor for angular displacement mounted in the robot joint. The robot learns by external observations of the teacher poses by Kinect sensor. By analytic approach for inverse kinematics, joint angles that satisfy the desired pose over time are calculated and transmitted to PID controller. The angular displacement measured in joint space is used to observe the performance of the system as an approximation for the deviation of end-effectors from the desired trajectory.