EINSTein Model Validation
Abstract
The use of Agent Based Models (ABM) to simulate behaviors in combat is gaining increasing recognition and interest across the Operations Research community of both the Marine Corps and now the Army. This paper presents our attempt to 'validate' EINSTein (Enhanced ISAAC Neural Simulation Tool) by comparing its outputs to those of another well-known combat simulation model, JANUS. The experiment first establishes the combat effectiveness of EINSTein entities executing a National Training Center (NTC)-type scenario. The scenario is designed to replicate one armored company of 14 'friendly' tanks versus a similar size force of 14 'enemy' main battle tanks. We allowed one set of 'friendly' entities to gain knowledge, or 'learn', by using the genetic algorithm incorporated in EINSTein. Another set of friendly entities was not allowed to 'learn'. For both cases, we recorded both the combat results of the EINSTein simulations and the entity actions. These observed actions were programmed into JANUS. For each instance, we compared the combat effectiveness resulting from JANUS to those obtained from EINSTein. The paper presents the findings of our analysis and suggests further research areas for using ABM's.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 2002
- Accession Number
- ADA405462
Entities
People
- Randall R. Klingaman
- William B. Carlton
Organizations
- United States Military Academy