基于MEBN的战术意图识别

战术意图识别是不确定性战场环境中理解战场态势、预测敌军将来行动过程的关键。意图识别模型的构建是作战仿真系统中实体行为建模的难点之一，如何表示不确定的意图元素及其不确定性关系并进行有效推理是意图识别的核心问题。分析了影响敌人战术意图的因素及各因素之间不确定性表示途径，提出了采用多实体贝叶斯网（multi-entity Bayesian network, MEBN）来描述敌人的战术意图，采用一阶谓词来表示意图影响因素，大大扩展了贝叶斯网表示不确定性问题的能力。MEBN采用实体片断集描述敌人的作战知识，方便重用。基于描述的敌人意图知识库，给出了基于知识的模型构建算法来动态构建战术意图识别模型。最后，通过仿真实验，对基于MEBN的战术意图识别方法进行了实验验证，结果表明该方法是有效可行的。

Tactical intention recognition based on multi-entity Bayesian network

In uncertain battlefield situations, tactical intention recognition is a critical element for understanding the tactical situation and predicting the coming course of action of the enemies. How to integrate intention recognition into a behavior model of simulated combat entities is a challenging problem and how to represent the uncertain elements relating to enemy’s intention and their uncertain relations is essential issues in intention inference. Through analyzing the factors that impact tactical intentions of the enemy and the methods that denote uncertain relations among these factors, the multi-entity Bayesian network (MEBN) is presented to denote tactical intention of the enemy. MEBN expresses the influence factors by first order logic and enhances the ability of describing the uncertain problem. MEBN adopts entity fragments for describing the military domain knowledge and so it is easy reused. A knowledge-based constructing model algorithm is proposed for dynamic constructing a tactical intention model based on the knowledge base of enemy intention. Finally, a simple example of tactical intention recognition is given. The results show that the method is available and efficient.