Professors of Nara Institute of Science and Technology developed a robot table tennis player who develops the skill by learning the idiosyncrasies of the opponent. A man who carries the sensor is the opponent, and the robot estimates the position of the returned table tennis ball by analyzing how he swings the racket to increase the probability of hitting back the ball. The system is made up of a robot arm, a PC for information processing, two cameras, and a sensor to be worn by the opponent. The camera instantly knows the orbit of the ball hit by the opponent, and the PC analyzes the orbit together with the data of the opponent’s moves, and estimates the orbit and let the robot swing the racket. Although the area to which the robot can hit the ball is limited because he cannot move as fast as a man, he successfully hit back about 70% of balls that were returned to the area about 20 cm around the racket. Experienced table tennis players have their own idiosyncrasies in their moves of the neck, shoulders, and wrists because they developed their own shapes of the swing. It is 20% easier to estimate the orbit of the ball hit by an experience player than by an inexperienced payer. The developers plan to apply this technology to an industrial robot.
Related web page: http://www.naist.jp/en/