Understanding Situation Awareness in Multi-UAV Supervision: Effects of Environmental Complexity, Interface Design, and Human–AI Differences

Situation Awareness (SA) is paramount for successful multi-UAV supervision, requiring real-time integration of complex environmental dynamics and system states. However, most prior research has centered on system-oriented monitoring tasks, including tracking UAV positions and statuses, rather than supporting users in interpreting environmental information. To address this gap, we conducted a simulated wildfire monitoring study manipulating key environmental and interface factors and compared the performance of human operators and GPT-4o across object-specific and global queries. Human participants demonstrated significantly higher accuracy on object-specific tasks than on global queries, highlighting their strength in real-time perceptual processing and the challenges of memory-based reasoning while the reverse was true for GPT-4o. Across conditions, SA accuracy was lower at higher SA levels, under higher target density, and in the presence of event clustering. Interface support in the form of centralized layouts and retrospective access was effective in mitigating cognitive load and enhancing SA accuracy.