Agent-based Approaches to Dynamic Team Simulation; Technical note, 1 Jun-31 Dec 2007

摘要

The Office of Naval Research (ONR) has recently funded the Personnel Integration of Selection, Classification, Evaluations, and Surveys (PISCES) effort. One of the products of the PISCES effort involves creating a virtual environment to enable team simulations. This report defines the capabilities required to develop Test Simulator (TESTOR), an experimental agent-based virtual simulation for a distributed team. These capabilities are organized around the technologies of agent-based approaches, simulation, and optimization relevant to team selection and performance. The report identifies supporting capabilities necessary for specifying and capturing team performance metrics in an experimental virtual simulation environment. Four applications of the model consistent with PISCES objectives are considered for the simulation: prediction of team performance, team selection, individual diagnostic assessment of teamwork, and assessment of teamwork. The report is organized into six sections. Section 1 characterizes current knowledge of teamwork and the factors that would need to be incorporated in a comprehensive simulation of team behavior. Section 2 reviews agent-based models of teamwork describing work involving both teamwork approaches to the design of multiagent systems and agent-based representation of human behavior. Section 3 examines the advantages and disadvantages of agent-based modeling in the context of the complexity and richness of human teams and explores possible methods for overcoming the difficulties. Section 4 discusses issues related to predicting team performance from simulation. Section 5 discusses advantages and disadvantages of conventional optimization and agent-based approaches to the team selection problem. Section 6 explores the problems and possibilities of using virtual team simulation for diagnostic assessment of an individual Sailor's teamwork behaviors and the extension of automatic assessment to human teams.