Projects


MVP-CA Analysis of IMMACCS (Integrated Marine Multi-Agent Command & Control System)
  • Agency: Office of Naval Research.
  • Duration: August 2000 - March 2001.


   Capabilities Being Demonstrated:
  • Exposition of the Software Architecture of IMMACCS
  • Organization of Agent Rules through Intermediate Concept Generation
  • Aid to Knowledge Entry through Template Formation
  • Exposition of common functionalities across multiple agents for reuse


IMMACCS (Integrated Marine Multi-Agent Command & Control System) was built by CDM Technologies at San Luis Obespo, CA under a contract from the Office of Naval Research. In order to render integrated systems such as IMMACCS reliable, a human should be able to analyze the knowledge contained in these systems. Since intelligent agents encapsulate intuitive concepts such as beliefs, commitments and rules, bulk of effort in building these multi-agent systems lies in specifying the underlying ontology for a given problem domain.

An ontology is a belief system for the agent that allows the problem to be formulated in concrete terms. The aim is to capture the details of the complexities of the domain in the ontology design, so as to simplify the problem solving aspects of the agents. Even though the goal of an ontological engineer is to try and formulate the ontology in a general manner, in reality the level of detail for an object definition in a given ontology, its type specification, and its placement in the ontological hierarchy to reflect its relationships to other objects in the ontology, all get influenced by the overall problem solving goals for the agent(s) that will use the ontology. Computer-assisted analysis tools, such as the MVP-CA tool, are needed desperately to perform efficient analysis for such systems, because they are often too complex for manual analysis.

During our analysis of IMMACCS, our results exposed the software architecture of one of the IMMACCS agents, the FIRES agent. In addition, it helped generate intermediate concept nodes through clustering to make the software more comprehensible, as well as helped in the detection of templatizable regions that could render the software more extensible in future. Moreover, the inter-agent clustering done by the MVP-CA tool, performed across 13 agents, showed significant portions of the software that had common functionalities across these agents. Such expositions pave the way for identifying the bottle-neck nodes in the ontology, that is, those nodes whose semantics need to be unambiguous, as well as those that may become communication bottle necks when all agents need to access the same data.

Click here to access password protected Pragati's IMMACCS presentations.


« Back | Top
 
2007 Pragati Synergetic Research, Inc. Research Focus | Products | Projects | Publications | Market Sectors | About Us