用于战术意图识别的动态序列贝叶斯网络

针对意图规划过程的层次性和序列性，提出了序列贝叶斯网络(series Bayesian network, SBN)模型，并将动态贝叶斯网络(dynamic Bayesian network, DBN)模型和SBN模型结合起来，构建了动态SBN(dynamic SBN, DSBN)模型进行规划识别，模型的DBN部分用于由特征序列推理元意图，SBN部分用于由序列意图逐层推理父意图。推导了模型的算法，分析了模型在规划识别问题中的表达和推理能力。实验表明，DSBN模型能够有效根据特征序列识别战术意图。

DSBN used for recognition of tactical intention

The series Bayesian network (SBN) model aiming at the characteristics of hierarchy and sequentiality of the planning process is proposed and the dynamic series Bayesian network (DSBN) made up of dynamic Bayesian network (DBN) and SBN for plan recognition is constructed. DBN part of the DSBN model is used for reasoning meta intentions by a series of characters, and SBN part of DSBN model is for reasoning the upper intention by links of children intentions. The paper infers the deduction algorithm of the DSBN model, and analyzes the ability of depict and reference of the DSBN model in the domain of plan recognition. Experimental results show that the DSBN model can recognize the tactical intention by a series of characters efficiently.