A case-based reasoning (CBR) system that continuously interacts with an environment must be able to autonomously create new situation cases based on its perception of the local environment in order to select the appropriate steps to achieve the current mission goal. Although many continuous problem domains seem appropriate for case-based reasoning, a general formal framework is still missing. This paper presents a step in the direction of developing such a formal model of autonomous creation of new situation cases. The model is based on the notion of the step for attentional shift. This notion allows us to define the representation scheme for situation cases. We have introduced two types of situation cases: contextual cases and action cases. The solution component of contextual cases, also called a contextual behavior routine, is used as a resource to direct the attention of the CBR system to the relevant aspects of the local environment. The solution component of action cases, also called an action behavior routine, is used to guide selection of manipulative steps. There are two key roles of steps for attentional shift in our model. The first one is that steps for attentional shift represent a description structure of situation cases. The second role is that steps for attentional shift represent an abstract representation of actions by which the CBR system moves the attention to the relevant aspects of a local environment.
展开▼
