Overview of the Agent-based Modeling and Behavior Representation (AMBR) Model Comparison Project

摘要

The Air Force Research Laboratory (AFRL) is sponsoring a series of human performance model comparisons under the Agent-based Modeling and Behavior Representation (AMBR) Model Comparison Project. The first comparison challenged the modelers to build human cognitive models of multiple task management and attention sharing, embodied in the behavior of an air traffic controller operating in a simplified ATC task. The second comparison involved modifying the human behavior representation (HBR) models and the simulation environment to operate as an HLA Federation. BBN Technologies, as the model comparison moderator, designed the experiments, provided the simplified ATC simulation environment in which the models run, and collected data on actual human operators performing the task. CHI Systems, Soar Technology, Carnegie Mellon University, and a team from AFRL's Logistics and Sustainment Division built the HBR models. This paper is the first of several presented as part of the AMBR Symposium at the 10th Annual Computer-Generated Forces and Behavior Representation Conference, and it is intended as an introduction to that set. The paper provides an overview of AFRL's rationale for investing in this line of research, makes explicit the AMBR Model Comparison's goals, describes the iterative model comparison structure chosen for accomplishing those goals, and concludes with an introduction to the first two rounds of the project.