Proceedings of the 11th Computer Generated Forces (CGF) Conference, Orlando, FL (May 2002)
Through the development of increasingly sophisticated models and simulations to support a range of analysis and training functions, it has become evident that there exists a strong demand for realistic representation of the key human decision makers that populate simulated environments. Developing high-fidelity human behavior representations (HBRs) is not a trivial exercise, and is further complicated by the complexity of today’s simulation environments. The goal of this paper is to detail an approach taken towards developing a computational representation of cognitive functionality based on knowledge elicitation (KE) results. Specifically, we describe the development path from goal-directed task analysis (GDTA) methodologies that identify information requirements and decision structures through to the computational implementation of a model of the cognitive functions that define decision-making behavior. We apply our developing agent-based HBR architecture, SAMPLE, as the computational framework. Our discussion is based on a case study modeling a commercial Airline Operations Center (AOC) dispatcher who is responsible for flight planning, monitoring and potential on-line re-planning to avoid hazardous conditions. The KE effort identifies the specific hierarchy of tasks performed during dispatch operations, the information available to the dispatcher to support those tasks, and the analysis techniques applied by the human to perform them. Our goal is to translate this detailed task description into an agent-based dispatcher model consisting of a networked set of cognitive functions performing information processing, situation assessment and real-time decision-making in a manner that mirrors the human dispatcher’s approach. These individual cognitive representations are integrated within the SAMPLE architecture in a process-oriented manner through specific artificial intelligence technologies, namely fuzzy logic for information processing, Bayesian belief networks for situation assessment and expert systems for procedurally-driven decision-making.
For More Information
(Please include your name, address, organization, and the paper reference. Requests without this information will not be honored.)