Knowledge-based role allocation in robot soccer

Robot soccer teams consist of a number of robots each performing a different role within the team. The roles discussed in this paper are: goalie, defender, attacker, supporter and centre. These roles are too often statically assigned to the robots at the start of the game. Knowledge-based techniques can be used to assign these roles dynamically to allow the team to adopt the optimal behaviour for each situation. Dynamic strategy choice can also be implemented within the same knowledge-based system. A strategy (defend or attack) can be chosen based on robot and ball location which in turn determines which roles should be used in play. Once the roles are defined, they will be assigned to the best robot for each role in turn based on role importance. Testing within the Teambot simulator shows a significant in score and ball control dominance over the same team with static role allocation. This paper presents the knowledge-based role assignment approach employed and the favourable results obtained.