Eye-In-Hand Visual Servoing for Accurate Shooting in Pool Robotics

Abstract

Deep Green is a robotic pool system whose objective is to play the game of pool competitively against skillful human opponents. To enhance the playing accuracy of the system, a visual servoing algorithm using a wrist-mounted camera was developed to correct the absolute positioning error of the robot. The novel technique considers an ideal line defined by the intersection of the object and cue ball centers at their ideal locations from the vantage of this camera. The ideal line represents an imaginary straight shot trajectory projected onto the image plane. The transformation of 2-D image points into 3-D robotic motions is done through estimation (in the image-based approach) or by a 2-D to 3-D spatial mapping (in the position-based approach). While the first option is simpler, the later is more effective, requiring fewer iterations and therefore less time to converge. Experiments were designed to measure the accuracy of the system. Using the wrist-mounted camera, the system increased its shooting accuracy by a factor of three, with high consistency.