Playing billard with an anthropomorphic robot arm

Abstract

The intention of this work is to investigate the possibility of playing billiard with the DLR Light Weight Robot III (DLR-LWR III). Playing billiard with such a robot is due to the necessity to combine strategic planning capabilities, precise vision, and accurate robot control a challenging task to be performed. In contrast to existing approaches, where an active cue acceleration device is used to achieve the velocity for striking a ball, we directly generate this with the robot. For this purpose an experimental billiard table was set up. Furthermore, a set of randomly placed billiard balls are localized and identified with a computer vision system. In order to properly plan shots according to the particular situation, an artificial intelligence was developed, which is capable of analyzing a given table state and choose an appropriate shot. In order to accomplish this task, a custom made cue end effector was constructed. To show the systems abilities, experiments are performed, verifying the accuracy and success rate of the vision system and simulation results for performing the entire loop are presented.