Abstract
This paper shows a ball-pass algorithm for a multi- agent system problem, RoboCup Middle Size League, where many autonomous mobile move simultaneously. In the environment, all robots have to estimate the ball position and opponent robots, and make decision to score goals. However, the estimated ball and robot's positions include errors and need some methods to revise their positions. In order to make the decision from the estimated position data with uncertainty, the positions of both side robots, strategy, environmental data are expressed using probability distributions, which are combined to the ball-pass probability distribution. Based on the distribution, a mobile robot decides where to kick the ball. The proposed algorithm is evaluated through simulations.
