Human motion based training & execution robotic system with a prototype industrial application

Abstract

Frequent robot re-training is required if the robot has to perform variety of tasks. Human Motion is the most natural and fluent form of input to train a robot, and robot motion has always been programmed in an un-natural way. This paper presents a method of training of robots by real time emulation of human motion of different limbs. The Human Motion data is also recorded and the motion can be regenerated at a later time. The purpose is to bridge the gap between human motion and robotic motion and make it more natural and apply it on a prototype of an Image Processing based Industrial Sorting/Inspection System. Lego Mindstorms NXT Robotic kit is used as the robot.