AN HEURISTIC BASED APPROACH TO TEAM BASED BEHAVIOURS IN REAL TIME STRATEGY GAMES

This paper explores aspects of high level strategic cooperation and coordination between AI players. A cooperative team-based AI player for the commercial RTS game StarCraft is created using team-based heuristics, state machines and influence maps. Effective teamwork inherently improves the performance of teams in team-based games; as a result, the effectiveness of the AI implementation is determined by measuring the improvement in the AIs overall performance when playing team-based games. Various experiments are carried out to test the effectiveness of each team-based behaviour. Results demonstrate that the coordinated and cooperative team of AI players gain a decisive advantage over non-cooperative teams, and that, as expected, teamwork is vital to the success of a team in RTS games. It is concluded from these results that an AI which is capable of team-based cooperation can be successfully created for an RTS using a heuristic based approach.