Improvements to Autonomous Forces Through the Use of Genetic Algorithms and Rule Base Enhancement
Abstract
This thesis discusses two approaches to enhancing the performance of intelligent autonomous agents in a computer combat simulation environment so that their performances more closely model the tactical decisions made by human players. The first approach addresses incorporating a genetic algorithm (GA) into the NPSNET Autonomous Force Expert System (NPSNET AF), while the second approach focuses on enriching the existing rule base and decision strategies. First, we develop a functional genetic algorithm with the intent of providing dynamic, real-time learning within the NPSNET AF. However, we conclude that the GA is better suited for a static problem, such as artillery battery registering of fires, rather than for the dynamic battlefield of the NPSNET. Second, we enrich the NPSNET AF expert system by enabling it to choose from among four formations and by providing a mechanism for transitioning between them. We enable the expert system to make formation decisions based upon general terrain characteristics and target location.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1993
- Accession Number
- ADA275033
Entities
People
- John P. Steiner
- Robert A. Jacobs
Organizations
- Naval Postgraduate School