The Research Of Belief Change In Intelligent Agent Action Reasoning

The ability of gathering information of the world and revising its original beliefs based on the new information is important for intelligent agent. Belief revision is the process that changing original beliefs of an agent to accept new, more precise and more reliable information which may inconsistent with existing beliefs. The main idea of action reasoning is that each element caused the change of the system states can be regarded as action. Belief revision and action reasoning have been two important branches of Artificial Intelligence.This paper focuses on belief revision and action reasoning. For belief revision, a dependent belief revision algorithm is put forward, which satisfies Ind assumption and covers the shortage of reinforcement revision methods. For action reasoning, an improved ARBC model is proposed, which introduces the reinforcement revision algorithm in the process of model construction. Several instances are employed to validate the feasibility of the raised method. The main work of this paper is stated as follows:(1) Fully investigate and analyze the existing belief revision assumption, and then select the proper hypothesis combination suited to guide belief revision algorithm as the research foundation of belief revision. In addition, main existing belief revision algorithms are studied while their shortages are pointed out.(2) Basic assumption and framework are researched in details. On this basis, dependent belief change algorithm is proposed, which revises beliefs based on the dependent belief sets in order to iteratively revises beliefs according to recover the discarded beliefs. Belief update and beliefs revision are talked about from action and acknowledge of agent. Finally, the program code is provided while some instances are used to verify the presented algorithms and the improve of reinforcement revision.(3) ARBC model is presented according to inheriting and improving the axiomatics, such as fluent calculus and state calculus. In this model, formulas denote beliefs and formula sets denote beliefs sets. Dependent belief revision algorithm is adopted in belief revision and action reasoning. In architecture, the ARBC model is divided into decision level,axiom level and reasoning level that denote strategy of planning and designing, working axiom and reasoning inference engine, respectively. Dependent belief revision algorithm is injected into ARBC model which is estimated by some instances.finally the related works are compared.