Abstract
In recent digital games, the control of characters has become complicated, and behavior trees tend to be adopted as a behavior decision method in character AI. However, it is not easy to create many character AI using behavior tree and adjust behavior well. In the present character AI adjustment, it is a work requiring a lot of labor to adjust based on the result of actually moving the character by AI, and if the adjustment is insufficient, to readjust it. And, it is difficult to find the part which needs the adjustment, especially in behavior tree AI. In this paper, we proposed a method to probabilistically find out a part required for adjustment at carrying out adjustment of behavior tree AI by using Bayesian estimation. In this method, nodes in the middle of behavior tree used for action decision are evaluated using Bayesian estimation.
