Research Of Plan Recognition Based On Collection Of Objects Dynamically Variable

Since Goodson， Schmidt and Sridharan proposed Plan Recognition as aresearch topic for the first time in1978， more and more scholars have engaged inresearching it，trying to solve some of the problems. The Plan Recognition is dividedto Plan Recognition with a Plan Library and Plan Recognition without a Plan Libraryaccording to whether the Plan Library is given in advance or not.Currently there existing some problems in the System of Plan Recognitionwithout a Plan Library:1. Provided that the objects collection of instantiation is static,namely, the object of the problem domain can neither be created nor be removed bythe action;2. The Planning Library by dynamically constructed can’t handle the newplan or predict the next step that the agent will take.Therefore, we constructed a new plan recognition model. The model includes:the idea of transforming object to proposition, Algorithm of Extending P-graph, NewPlan Operation Algorithm and Goal Plan Search Algorithm. The idea of transformingobject to proposition means that the object is given in the form of proposition, whichprovides a resolution to the issue that objects collection of the problem domain isdynamic and variable; Algorithm of Extending P-graph refers to dynamicallyconstructing the planning libraries based on the observed actions or facts; New PlanOperation Algorithm is that match the observed actions and its relevant knowledgewith the actions and relevant knowledge of the current Planning Library. Thematching result determines whether the P-graph gives the recognition results after theexpansion or immediately, which solves the inexistent plans in the current planninglibrary and make a prediction for the next plan; the Goal Plan Search Algorithm isused to search for the causal link to determine the goal and plan the agent trying toachieve; the FSas algorithm starts from the proposition layer and action layer and usesthe observed action effects, actions and mutually exclusive computing to reduce thespace of supported actions, which can determine the actions that support the actioneffect and further realize planning recognition.The construction of this theoretical model is of academic significance forexpanding the application of Plan Recognition. This system has a promising future inthe application of intrusion detection and network security.