Research And Implementation Of Plan Recognition Algorithm Based On Plan Library

The question of plan recognition is the issues which cross the field between psychology and artificial intelligence, It includes work on story understanding, psychological modelling, natural language pragmatics, and intelligent computer system interfaces. After more than 20 years of development, the application of plan recognitio is increasing constantly, there have already presented a lot of models and method of plan recognitio at present, plan recognition has good development prospects. The method in traditional plan recognition is to set up a plan library according to some observed actions, according to certain search and match mechanism to obtain the right plan.Among them method of Kautz regards as representatives mainly.But the plan library is static, work out by hand , need updating and safeguarding, and the search and match mechanism of plan library does not consider the fuzzy treatment of knowledge.And Jun Hong finds inspiration in graphplan, utilizing the goal graph to recognise plans. Because the goal graph is produced dynamically, the goal graph is equivalent to an implicit plann library.But it can only explain that has happened actions, and actual meaning of plan recognition is it can predict future actions to lie in. Traditional planning describes that has not considered the influence degree of each atcions that making up the plan to plan appearing when those happens,so it often makes the result have very great deviations . We gave every action of plan a a right value according to their influence degree the plan appearing, and this make plan recognizer can make recognising fast and accurate. We use fuzzy method in search and match mechanism,which making the result received according with the objective fact further. This text has not adopted such artificial intelligence language programming as LISP ,etc. , but make use of OOP method to realize the algorithm of the plan recognition based on plan libary, we apply the method of Class and Object to planning recognizer . The experiment has shown that our method is efficient and robust.