Multi-Resolutional Knowledge Representation Using Prototypes and Properties

In the course of developing a distributed logistics command and control application based on the Cougaar agent architecture@ttp://www.cougaar.org/). we were faced with a Iorge knowledge representation problem. The logistics agents needed to reason about many thousands of different types of logistics assets, where each asset had hundreds of attribuies. The problem is factorable, however, in that each specialized logistics agent nee& to know only a relatively small amount of information about only a relatively small number of assets. By using techniques of prototypes and delegation, adapting the paradigm suggested by Lieberman [ I ] and others [2, 31. we developed a logical data model (the Cougaar LDM) for logistics that effectively factors this representation problem. This factoring provides a basis for multi-resolutional representations of the entities in the logistics system; the information about any entity at each logistics agent can be limited to only that subset in which the agent has interest.