This article describes the computational model underlying the AGILO autonomous robot soccer team, its implementation, and our experiences with it. According to our model the control system of an autonomous soccer robot consists of a probabilistic game state estimator and a situated action selection module. The game state estimator computes the robot's belief state with respect to the current game situation using a simple off-the-shelf camera system. The estimated game state comprises the positions and dynamic states of the robot itself and its teammates as well as the positions of the ball and the opponent players. Employing sophisticated probabilistic reasoning techniques and exploiting the cooperation between team mates, the robot can estimate complex game states reliably and accurately despite incomplete and inaccurate sensor information. The action selection module selects actions according to specified selection criteria as well as learned experiences. Automatic learning techniques made it possible to develop fast and skillful routines for approaching the ball, assigning roles, and performing coordinated plays. The paper discusses the computational techniques based on experimental data from the 2001 robot soccer world championship.