ATTUnited-2001: Using Heterogeneous Players

ATTUnited-2001 is a successor of the CMUnited teams: CMUnited-97, CMUnited98, CMUnited-99, and ATT-CMUnited-2000. It is built mainly upon CMUnited99 [4]. It also incorporates the team action architecture from ATT-CMUnited2000 [3], but not the dynamic planning of set-plays capability [2]. The main research focus of ATTUnited-2001 is the a ect of heterogeneous players in the RoboCup simulator [1]. Heterogeneous players were introduced to the RoboCup simulator for the rst time this year, in version 7.0 of the simulator [1]. In particular, in any given game, each team is able to select from identical pools of players, including the default player type from years past and six randomly generated players. At startup, teams are con gured with all default players. However, the autonomous on-line coach may substitute in the randomly generated players for any player other than the goalie. The only restriction is that each random player type can be assigned to at most 3 teammates. The random players are generated by the simulator at startup time by adjusting ve parameters, each representing a trade-o in player abilities: 1. Maximum speed vs. stamina recovery 2. Speed vs. turning ability 3. Acceleration vs. size 4. Leg length vs. kick accuracy 5. Stamina vs. maximum acceleration For the remainder of this paper, the parameters will be referred to by their associated numbers above. At the outset, it was not known whether using heterogeneous players would be advantageous. The research reported here was designed to generate a prediction of the advantage of including a player with a speci c player type on a team, and as a secondary goal, characterizing the utility of playing a given player in a given position (location on the eld).