Motion Planning to Catch a Moving Object with Human Arm Characteristics.

Abstract

This paper proposes a method to generate human like arm movement to catching a moving object. Firstly, human arm movement is analyzed to catch a constant velocity's object and discuss several typical characteristics. Then a time-varying LQG controller is designed to take account human characteristics into robot motion. This controller holds the values of non-linear terms in the dynamic model of robot arm during one sampling interval, which are updated at each sampling step. Adopting this criteria, we can get an optimal control input easily in real-time, and modify the controller with arranging the weight matrices in LQG criterion. Using these time-varying weight matrices, arm movement is generated considering with the characteristics of human behavior. Finally, the valodity of this method is confirmed with some simulation results.