Adaptive Artificial Intelligence for Next-Generation Conflict Simulation

Period of Performance: 11/04/2003 - 11/04/2004

$100K

Phase 1 STTR

Recipient Firm

Information Extraction & Transport, Inc.
1911 N. Ft. Myer Drive, Suite 600
Arlington, VA 22209
Principal Investigator
Firm POC

Research Institution

University of Massachusetts Lowell
1 University Ave.
Lowell, MA 01854
Institution POC

Abstract

Existing turn-based strategy war games have the potential to act as useful military training tools. However, their utility is greatly limited by the shortfalls of the artificial intelligence (AI) routines used to provide computer opponents. The key limitation in state-of the-art AI used in commercial wargames is the inability of the computer to perform situation assessment - to recognize the human player's tactics and strategy. Information Extraction and Transport, Inc. (IET) will develop a sophisticated architecture for automated situation assessment as part of a powerful and adaptive AI engine for commercial wargames. This capability will be implemented as a system of intelligent agents that collaborate using a Bayesian Blackboard. A number of decision theoretic models will be enabled that provide computer opponents the capability to effectively simulate a wide range of behaviors that translate into greatly varying tactics, actions, and strategies. Genetic algorithms and Bayesian learning methods will be explored for purposes of creating a strategist that is capable of adapting to changes in opponent strategy and responding agilely when the human locates weaknesses in the computer-generated strategies. This research will result in a capability to provide an adaptive AI engine for commercial wargames. There is an immediate application to commercial wargame designers and consumers of commercial wargames. This includes military training institutions that use commercial wargames to supplement formal strategic and operational planning. There is additional potential for sales to DoD contractors developing military simulations for US or allied military training. Finally, the AI engine developed by this research will provide a testbed for advanced tactical fusion research