Utilising the See-and-Follow Method for Enhancing Robot Learning Ability

Abstract

Abstract The rapid growth and development of robots today can be reflected from variations of research in the robotic field, and one of them that has been flourishing in the area is robot learning. Learning is one of the most important aspect needs to be implemented in a robot that allows the robot to improve its performance through experience-driven knowledge. There are several forms of learning methods and one of the most favourable approach is the see-and-follow method. This paper aims to develop robots that can learn to do the same movements with human movements through the see-and-follow method. Kinect sensor is used as a technology to support the process of recognizing the human body and its movements; which later to be transformed into the robot dimension. The results show that Kinect sensor can be used to measure the distance and recognize the joints of the human body along with the movements produced within a distance of 1 meter to 3.75 meters with an average reading error of 0.04 meter. The average delay time for the whole movement is 2.6 seconds. The average percentage of similarity between robot and human movements for the whole movement scenarios is 93.45%. Hence, high level of similarity in movement concludes that the method of see-and-follow successfully increases the learning skills of the robot.

Access to Document

Other files and links

• Link to publication in Scopus
• Open Access Version Available

Fingerprint