Trajectory solutions for a game-playing robot using nonprehensile manipulation methods and machine vision

The need for autonomous systems designed to play games, both strategy-based and physical, comes from the quest to model human behaviour under tough and competitive environments that require human skill at its best. In the last two decades, and especially after the 1996 defeat of the world chess champion by a chess-playing computer, physical games have been receiving greater attention. Robocup TM, i.e. robotic football, is a well-known example, with the participation of thousands of researchers all over the world. The robots created to play snooker/pool/billiards are placed in this context. Snooker, as well as being a game of strategy, also requires accurate physical manipulation skills from the player, and these two aspects qualify snooker as a potential game for autonomous system development research. Although research into playing strategy in snooker has made considerable progress using various artificial intelligence methods, the physical manipulation part of the game is not fully addressed by the robots created so far. This thesis looks at the different ball manipulation options snooker players use, like the shots that impart spin to the ball in order to accurately position the balls on the table, by trying to predict the ball trajectories under the action of various dynamic phenomena, such as impacts. A 3-degree of freedom robot, which can manipulate the snooker cue on a par with humans, at high velocities, using a servomotor, and position the snooker cue on the ball accurately with the help of a stepper drive, is designed and fabricated. [Continues.]