Developing a Conceptual Architecture for a Generalized Agent-based Modeling Environment (GAME)
Abstract
Agent-based technology is a relatively new, but rapidly proliferating decision technology. The relative immaturity of ABM software often requires significant programmer expertise in model representation and implementation. This limits potential users and their ability to utilize the software. We develop a high level conceptual architecture for an agent-based modeling environment which overcomes this limitation. This thesis defines a taxonomy of agents based on commonly accepted agent characteristics, reviews six of the most popular software platforms for agent-based model development, and maps their relationship to the taxonomy. Past modeling advances in the operations research and management science (OR/MS) domains indicate that a more generalized environment is possible. A conceptual architecture for a generalized agentbased modeling environment (GAME) based upon design principles from OR/MS systems was created that would overcome some, if not all, of these obstacles. The GAME architecture incorporates higher-level model representations separate from solver code, a library of transformation procedures, reusable model libraries and a robust language or equivalent interface for specifying experimental design procedures. Rapid technology development would allow for agent-based modeling software that subsequently benefits a much wider range of stakeholders than is currently the case. Finally, embedding GAME in an even higher-level integrated decision technology environment (IDTE) would facilitate the integration of computational and analytical modeling.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2008
- Accession Number
- ADA479981
Entities
People
- Christian T. Nguyen
Organizations
- Naval Postgraduate School