Graph Neural Swarm Control: Reliability Guarantees for Disaster-Response Drones

We introduce Graph Neural Swarm Control (GNSC), a decentralized controller that combines graph message passing, distributed consensus, and runtime safety shields to coordinate disaster-response drones. Using synthetic, stress-tested swarms across earthquake, flood, and wildfire scenarios, we show that GNSC achieves higher mission success with bounded decision latency and fewer collisions under packet loss, GPS drift, and wind gusts. We present an analytical reliability framework and report trade-offs among latency, coverage, and safety.