Modeling tactical level combat using a Multi-agent System Design Paradigm (GI Agent)

摘要

In the past 60 years the Army has undergone a major reorganization eight times at the divisional level and numerous more times at unit levels below the division. Each time the Army reorganized it's divisions a major testing program was involved. But when a change in organization is done at unit levels below division often very little attention is paid to how the change will affect the unit. When this happens, unit leaders are forced to undertake one of the most difficult jobs in today's military incorporating new equipment into a unit or reorganizing a unit without an understanding of how the changes will affect the unit. The Military modeling and simulation community has attempted to fill this need but the current set of single entity simulations are limited in their ability to replicate dynamic complex behavior. This thesis is attempting to create a Multi-Agent Simulation that will allow analysts and leaders to gain an understanding of the tactical employment affects of changing the organization of a company level infantry unit. GIAgent is a simulation tool allowing the analyst and leader to experiment with the complex relationship between maneuver and unit organization without putting the unit in the field. Combining agent based artificial intelligence techniques with artificial intelligence research from the computer gaming industry, GI Agent creates a new paradigm for combat simulation. The GIAgent software uses the RELATE architecture designed by LCDR Kim Roddy, USN and Lt Mike Dixon, USN.