基于分层贝叶斯网络的计划识别方法

分析了态势估计中计划识别问题。采用分层贝叶斯网络对计划进行分级表示,依据发生的战场事件和目标行为,动态构建分层贝叶斯网络结构,并将其作为证据进行推理。为了解决动态构建贝叶斯网络的推理问题,提出构建虚拟节点的方法,将底层贝叶斯网络的推理结果作为不确定证据,输入到对应的上层网络节点中,以代替该底层贝叶斯网络。最后,文中通过仿真实验,对基于分层贝叶斯网络的计划识别方法进行了实验验证,结果表明该方法是有效可行的。

Method for plan recognition based on hierarchical Bayesian networks

The plan recognition in situation assessment is analysed.The hierarchical Bayesian networks is utilized to represent the plan.According to the event and target action,the hierarchical Bayesian networks for plan recognition is dynamically constructed.In order to solve the inference for dynamic Bayesian networks,the approach of adding virtue node is proposed.In the approach,the inference result of replacement Bayesian networks is considered as the uncertainty evidence of corresponding node in the higher networks.At last,a simple example of plan recognition is given.The results show that the method is available and efficient.