在目标图算法的基础上引入推演扩展和推演检测过程,并进一步扩展了目标图的构造和分析过程,使得目标图可以处理带有推理规则的规划识别问题,从而在初始世界状态空间未完全已知的情况下依然可以识别目标.该算法不需要规划库,从而避免了利用规划库所带来的诸如手工编码和搜索时具有指数代价的问题.
A novel algorithm for inference-rule-introduced plan recognition was introduced. Using a method of inference expansion, goal graph was reconstructed to extend and improve original algorithm to allow plan recognition with inference rule. Thus the inference can be assure d to be true. The algorithm based on goal graph needs no plan library, so it can' t be suffered from the problems in the acquisition and hand coding of large plan libraries. It also has no problem in searching the plan space of exponential size.