Modeling human and organizational behavior using a relation-centric multi-agent system design paradigm

Abstract

Today's modeling and simulation communities are being challenged to create rich, detailed models incorporating human decision-making and organizational behavior. Recent advances in distributed artificial intelligence and complex systems theory have demonstrated that such ill-defined problems can be effectively modeled with agent-based simulation techniques using multiple, autonomoous, adaptive entities. RELATE, a relation-centric design paradigm for multi-agent systems (MAS), is presented to assist developers incorporate MAS solutions into their simulations. RELATe focuses the designer on six key concepts of MAS simulations: relationships, environment, laws, agents, things, and effectors. A library of Java classes is presented which enables the user to rapidly prototype an agent-based simulation. This library utilizes the Java programming language to support cross-platform and web based designs. All Java classes and interfaces are fully documented using HTML Javadoc format. Two reference cases are provided that allow for easy code reuse and modification. Finally, an existing metworked DIS-Java-VRML simulation was modified to demonstrate the ability to utilize the RELATE library to add agents to existing applications. LCDR Kim Roddy focused on the development and refinement of the RELATE design paradigm, while LT Mike Dickson focused on the actual Java implementation. Joint work was conducted on all research and reference cases