A behavior-based implicit planning method in competitive environment

The environment of a middle-size autonomous robot soccer system (MARSS) is highly dynamic, competitive and partially observed. To decide quickly, behave smoothly and speedily, the proposed MARSS used a behavior-based implicit planning, and select suitable task according to the hidden state. To overcome imprecise perceptions and actions, several subtle but simple goal-driven behaviors are designed. Tightly integrated with the simple perceptions, these behaviors switch flexibly and robustly by continuous feedback. In the competitive environments, spontaneous interleaving of these behaviors implicitly plans effective behavior sequences to fulfill the game´s tasks, and exhibits some important emergent behaviors making the system design more simple, robust and competitive. The method´s effectiveness is verified by experiments and games.