This research presents an evolvable cooperation strategy based on the fuzzy reinforcement learning in the interactive robot soccer game. The interactive robot soccer game has been developed to let human join into the game dynamically and reinforce entertainment characteristics. Accordingly, a cooperation strategy between humans and autonomous robots is very important to make game more realistic.
Until now, many researchers in MIROSOT or RoboCup have proposed various strategies. Because the robot soccer system is hard to model and its environment changes dynamically, it is very difficult to pre-program cooperation strategies between robot agents. So, the programmer must have special experience and expert knowledge about robot soccer system in order to make a strategy work well. Therefore, many learning strategy methods using reinforcement learning, genetic algorithm, or genetic programming have also been proposed by many researchers. Between them, reinforcement learning is highly suited for solving problems dynamically without explicit knowledge of the system. So, that method has been frequently utilized in robot soccer strategy development. Accordingly this paper proposes an evolving strategy with simple rules in the developed robot soccer game with fuzzy reinforcement learning method. Additionally, to evaluate the usefulness of the proposed strategy for the cooperation between humans and robots, some experiments have been carried out.
