Research And Implementation Of State Calculus Based On Knowledge

Reasoning about action and knowledge representation are important research areas in artificial intelligence. Reasoning about Action has been applied in many areas, such as cognitive robotics, web services, workflow and so on. The main task of reasoning about action include to predict the result of an action or an action sequence given the system's initial state and rules of changing, and to explain what action the system has experienced given the system's initial state and resultant state.Modal logic is an important tool for knowledge representation. Using modal operator to represent knowledge is well-established in theory, but it is an ideal model that requires unlimited computing resources in practice. State Calculus is a formalized system for reasoning about action that is used to describe the transformation of dynamic systems, but it can only represent the knowledge of dynamic systems to some extent.The main contents and contributions of this dissertation are as follows:(1) Based on possible world semantics in modal logic, we introduce the fluent disjunctive normal form to represent the knowledge, which enhanced State Calculus's knowledge representation ability and alleviated the difficulty in implementing knowledge representation based on modal logic.(2) Based on the State Calculus theory, we designed a new logic programming language, STEX. The reasoning in STEX by using the underlying principle of progression is more efficient than Golog, Congolog and IndiGolog by using the underlying principle of regression based on the Situation Calculus. STEX can provide a number of important knowledge representations and reasoning that cannot be achieved by FLUX based on the Fluent Calculus, and hence can be applied to a broader scope of situations.(3) We implemented the path planning example of robot taking the gold using the new logic programming language STEX. The experimental results show that the STEX enabled agent can update its own state by perceiving the dynamic environmental information and its own action, and achieve its goal through action reasoning.