Selecting strategy for agent behavior based on fuzzy algorithm and Q-learning

摘要

In robot soccer simulation team, how to select a proper behavior for a player among shoot, dribbling and passing is a key issue. This paper proposes a more flexible behavior selecting strategy. In the strategy, the fuzzy-algorithm is used to deal with behavior selecting issue. Because the environment of the robot soccer is complicated, with this algorithm it's not necessary to build a precise mathematic model about the environment. Simultaneously, Q-learning is used to modify the fuzzy rules. The experimental results show that this algorithm is more efficient and robust which can improve the success rate of robot player in shoot, passing and dribbling.