A PA-PBT Based Behavior Modeling Framework

With its reactivity and modularity, Behavior Tree (BT) is recognized as one of the key instruments for behavior modeling. But for behavior modeling in military operations, the strong correlation between action targets and conditional judgment results in standard PABT vertical expansion does not adequately depict troop behavior model. This paper suggests a behavior modeling framework based on Planning and Action Probability Behavior Tree (PA-PBT) to address the problem. The automatic horizontal expansion operation mechanism and the probabilistic selection node are added to this framework, which applies randomization through probabilistic selection node and resolves the issue of non-strong correlation between action goals and condition judgment results under typical PA-BT vertical expansion. Accuracy of proposed framework is confirmed by experiments conducted in various settings.