Task learning utilizing curve fitting method for kinect based humanoid robot

Tele-operation system had been proven useful in various fields, either in researches, robotics, medical or even military. This system is capable of aiding human in carrying out tasks, by minimizing human limitation caused by environment or capability factor. However, it is a big challenge to obtain a robust system for controlling a robot especially when the robot has a high degree of freedom. In this paper, human mimicry as a teleoperation method is presented using a semi-humanoid robot, with Microsoft Kinect sensor as its interface. Vector calculations are used to obtain joint angles of human body to be imitated by the humanoid robot, focusing on upper body imitation. Then, a task learning methodology is proposed for the humanoid robot to carry out a task independently by using curve-fitting method. The experimental result shows that the robot system is able to perform human movement imitation as well as task learning.