No abstract

No abstract

to finish a task agent needs to do two things: planning and acting. Actions included in the classical planning are usually primitive actions. For complex tasks the planning must contain sensing actions and knowledge of the agent. High-level intelligent agents need to have the ability to sense the external environment independent of the designer and to apply their knowledge in planning. Action theories in situation calculus is a typical approach for describing the planning with sensing action and knowledge, where the change of knowledge is regarded as an epistemic fluent described by situation accessibility relation ( , ) K s s ′ . But since agent can not determine which situation is accessible from a given situation, this approach depends on the designer who writes the axioms for agent. Zhou proposed an extended situation calculus language SCS L for describing action reasoning independent of the designer. In this paper, we give the concept about individual knowledge of agent, and research on the properties of agent's individual knowledge based on SCS L . We also give an example to illustrate our approach can solve more issues than the existing approach.

No abstract

Qualitative reasoning is a very efficient method that people often use to solve problems. Recently, the literature about qualitative reasoning as a causal analysis and decision-making tool has been emerging. However, the existing qualitative reasoning methods are mainly used for the modeling of imprecise problems. There are not many studies on formalization tools of qualitative “inference”. This paper proposes a logical system for qualitative reasoning (QRL). The main research contents include: the grammar and semantic structure of QRL are given; the meaning of formula logic truth value in QRL and its assignment rules are discussed; and the reliability and completeness of QRL are proved. Compared with fuzzy logic, probability logic and other uncertain reasoning methods, the advantage of QRL method is that it can use axiomatic reasoning method, and does not need to construct membership functions and collect a large number of samples.