A HAMs-based Method for Designing Behavior
Jian-Jun Han · Jisuanji fangzhen · 2008
Hierarchical finite state machine(HFSM) is a traditional method for designing behavior and has been widely applied in game fields. However, it lacks of efficiency because it needs full executing details provided by the game designers. This paper makes a preliminary research on solving this problem: applying hierarchical reinforcement learning of HAMs method to design behavior and implementing an experiment in Quake2. The preliminary experiment shows that this method can improve the efficiency of designing behavior. In addition, this paper shows that this method has faster speed of convergence than the flat reinforcement learning based method for designing behavior.