Human Systems Integration Division @ NASA Ames

This paper reports on the development of agent models with human-like performance characteristics for agent-based modeling and simulation. In order to provide models of advanced airspace it will be necessary to investigate and delineate the factors and issues affecting performance in highly automated air transportation systems and integrate those into a specification of controller functionality for developing computational human performance models. The goal of our effort is to develop a computational human agent model that can be incorporated into agent-based simulation to evaluate the impact of new technologies, such as datalink technology, on the workload and situation awareness of air traffic controllers. Our approach combines a task analysis, which provides a functional description of the domain, with a human performance architecture, from which detailed performance predictions can be made. We are in the preliminary stage of developing agent model that can produce human-like behavior. Here we will focus on how a standard task analysis can be paired with a human performance model to generate behavior that approximates that of the human user. This paper will also outline the issues of integrating human performance models into large-scale distributed agent-based simulation. We will also discuss our progress to date.