An Empirical Study of Human-Robotic Teams with Three Levels of Autonomy

Abstract

Abstract – Three search-and-identify experiments were conducted using the RoboFlag testbed to investigate the performance of human-robotic teams with different levels of autonomy. The first set of experiments required human operators to manually control robot trajectories by setting waypoints. In the second set, operators controlled robots via five pre-programmed plays. The third set allowed operators to operate waypoint control and automated plays, and performance feedback was provided to the user in real-time and/or at the end of each trial. Five performance metrics were analyzed across all experiments: game time, idle time, tag events, target location uncertainty, and target identity uncertainty. Performance increased with respect to all metrics when automated plays were available, and the addition of performance feedback in the third set of experiments further improved game time and idle time. Human efficiency and mission effectiveness are discussed with respect to autonomy level, form of performance feedback, and mission configuration. I.