This paper proposes a goal oriented action planning method that plans the behaviors of computer generated forces and a replanning method that replans new actions when the situations are changed in war game. This new method provides good expression because it is flexible and can do more realized description unlike the conventional finite state machines. As a result, proposed method has an advantage that it can describe the behaviors of computer generated forces as those of real soldier. However, since it is not deterministic it has some difficulties in analysing the decision processing of behaviors and making the computer generated forces do some specific actions. We employed combat plane models of air forces in order to verify the feasibility of our method. Finally, we could find that our method produced very similar behaviors to those of a real soldier. This paper describes our planning method, experimental results, and future works.