UAV Swarms for Real‑Time Wildfire Monitoring

Problem

Early wildfire detection requires high temporal and spatial resolution beyond satellites; coordination and sensing must scale over terrain and wind.

Aim

Design a swarm strategy and perception stack that minimizes time‑to‑first‑detection and maintains coverage during spread.

System

Sensing: thermal + RGB  ─► onboard detector (tiny models)
Planning: frontier coverage + consensus re‑tasking
Comms: mesh links for alerts + map sync
UI: web dashboard for boundaries, tracks, and UAV status

Experiments

• Simulated fires with varying wind/topography; compare patrol baselines vs. adaptive policies.
• KPIs: time‑to‑detection, coverage continuity, comms overhead, false‑alarm rate.

Findings

• ~40% faster first‑detection and more stable boundary updates under gusts.
• Detectors tuned for thermal imagery hold recall at low compute budgets.

Outputs

• Simulator scenarios, policy code, and dashboard; procedure for moving to field tests.