VLM-driven UAV triage framework cuts operator workload in disaster response

Researchers Swapnil Saha and three colleagues from AIAA AVIATION 2026 have introduced a new framework that uses Vision-Language Models (VLMs) not merely as advisory tools, but as active coordination agents in disaster response with UAVs. Detailed in arXiv paper 2607.27597, the architecture embeds VLMs inside the human-UAV loop to mediate communication between operators, mission control logic, and actual UAV task execution. Built with Model-Based Systems Engineering (MBSE), it defines system roles via use case and block definition diagrams, then implements three key components—VLM Coordinator Agent, UAV Mission Control, and Task Allocator—within an integrated simulation and control environment.

A human-factors evaluation with seven participants showed the system reduced perceived workload across mental demand, effort, and frustration, while earning high ratings for AI trust and communication clarity. The framework aligns with Incident Command System (ICS) standards, meaning it can plug into existing emergency response structures. By going beyond decision support—where humans still must translate outputs into actions—this approach automates the coordination layer, letting operators speak naturally and let the VLM orchestrate multiple UAV assets. The work advances scalable human-autonomy teaming for high-stakes disaster response and has broader implications for aerospace autonomy and civil safety.