Attention-Tunable Safety Barriers for Human-Autonomy Teaming in Specialized Aviation Missions

Abstract

The integration of autonomous systems into civil and military aviation is shifting crew composition from purely human to collaborative human-autonomy teams. Introducing autonomy enables the implementation of automated safety mechanisms. In this letter, we propose a novel safety barrier that adjusts system behavior based on the operator's attention state. The aggressiveness of the adjustments is tunable, and we propose tuning it based on the measured attention of the human to avoid overly conservative or excessively aggressive behaviors of the system. Our proposed solution is tested in human subject experiments of an Intelligence, Surveillance, and Reconnaissance mission in an immersive flight simulator where attention is measured by tracking the operator's eye gaze. Our experimental results show that the proposed attention-tunable safety barrier improves mission performance compared to a standard control barrier function mechanism.