Improved Situation Interpretation Metrics in Context-Based Reasoning Simulations

The current implementation of situation interpretation metrics (SIMs) in context-based reasoning (CxBR) simulations makes it difficult to effectively and efficiently characterize situations in the simulation. While previous work in the field has dealt to a certain extent with SIMs, the primary focus of such work has been in other areas. Thus, while the current implementation of SIMs is functional, there is much room for improvement. This paper describes a series of changes and improvements to the current implementation of SIMs that allow a more structured and modular approach to characterizing situations. By creating a more clearly defined structure for the SIMs, it is possible to separate the SIMs that are pertinent to the simulation at large and all the agents from the SIMs that are pertinent only to individual agents. This clear separation allows agents to set up a dependency tree between their own local SIMs and the global SIMs. By being aware of the dependencies between the SIMs, the agents can make more efficient decisions. Because the situation is only reevaluated when necessitated by a change in a particular SIM, only the affected bits of knowledge represented by the changed SIMs are reevaluated. The implementation of this structure is in progress, thus experimental results are not yet available. But, it is believed that this approach will provide a more efficient means of situation evaluation and decision making in CxBR simulations.