The Research And Application Of Plan Recognition Based On Dynamic Probabilistic Relational Models

Plan recognition has been developed over30years. It can be applied in the fields ofnatural language understanding, military coordination, intrusion detection, anti-terrorist,and so on. Plan recognition based on Dynamic Probabilistic Relational Model(DPRMs) wasproposed by Sumit S in2003. The dynamic probabilistic relational model theory is based onprobabilistic relational model theory. The main goal of probabilistic relational model is todeal with the uncertainty in modeling domain. When a frame structure is given,probabilistic relation model attempts to define a complete probabilistic distribution. Datalearning method is difficult to express the data, and it may lose much relation structureinformation. Probabilistic relational model is proposed to solve this problem.Agents work in an uncertain environment usually. There are many time-dependentobjects and relations. The model need a rich expression ability, it can carry out probabilisticreasoning and change from time to time. Probabilistic relational mode is difficult to solvethese changing situations. Dynamic probabilistic relational models overcomes theshortcomings of the probabilistic relational models. It can identify the target of machineoperation, according to the comparision of the target and the initial target. In this paper, theDynamic probabilistic relational model and its application in plan recognition is studied.Research topics in this thesis:(1) Discussion of the fundamental theory in the temporal constraints of Kautz Planrecognition methods. a practical application is presented: Intelligent flower cultivationsystem.(2) Discussion of the integration of the Bayesian network and object relation.(3) Discussion of abstraction tree and planning of Dynamic probabilistic relationalmodel.(4) The general practical filter and RB partical filter algorithm is compared in matlab,the validity of Dynamic probabilistic relational models is provided through visual graphicsand data.(5) A Top-down plan recognition algorithm flow char is proposed. The. Algorithmdevelops the Kautz’s down-top algorithm.