The backend runs a simulation engine that generates realistic drone activity on a 1-second tick cycle:
- Drone spawning — Up to 3 virtual drones are spawned at random positions around Syracuse, NY. Each is assigned a real-world drone profile (DJI Mavic 3, Autel EVO II, Skydio 2+, etc.) with accurate RF characteristics.
- Flight paths — Drones follow random-walk trajectories with heading drift, altitude variation, and speed changes to simulate realistic flight behavior.
- RF detection — Four simulated ground sensors calculate signal strength (RSSI) using a free-space path loss model based on distance to each drone. Sensors only detect drones within their range.
- Classification — Detected RF signatures (frequency, bandwidth, protocol) are matched against a library of known drone profiles to identify the drone type, with a 5% misclassification rate for realism.
- Threat scoring — Each detection is scored 0–100 based on proximity to the protected zone, loiter time, altitude, and speed pattern. Scores map to low / medium / high / critical threat levels.
- Geofence alerts — When a drone enters the protected zone (red circle on the map), an alert is pushed to the dashboard in real-time.
In a production deployment, the simulation engine would be replaced with real SDR (Software Defined Radio) hardware on Raspberry Pi units, feeding live RF data into the same classification and threat assessment pipeline.