Everything you need to watch your sky
AeroScope is not a roadmap deck, these capabilities are live on the platform right now, for anyone watching their own sky, a researcher, or a security analyst. Each one maps to the same mission: see what is above you, know what it is, catch what is deceptive, and stay ahead of it in real time. Deliberately absent: deep learning, satellite coverage, and anything resembling air-traffic control.
Your sky, rendered in 3D
A live instrument HUD over your position, a rotating radar dome, an animated altitude histogram, and a drag-to-explore situational globe with connection arcs to the contacts that matter. The app view is rendered client-side with three.js; the preview below is a live render in its own right, drawn at your display’s full pixel density rather than played back from a video.
Multi-source fusion
adsb.fi, adsb.lol and OpenSky, de-duplicated by ICAO 24-bit address, one airframe, not three dots. How fusion works →
MULTI-FEEDAttention scoring
Eight weighted rule-based factors summed into an explainable 0–100 rank. Always shows its working; never an accusation. See the score →
8 FACTORSConsensus anomaly detection
River, scikit-learn IsolationForest, a pykalman innovation test, OpenAP envelope checks, stumpy matrix profiles and PyOD, a flag fires only on ≥2-detector agreement. Detectors →
TORCH-FREESpoofing & integrity
A DO-260B-style 7-check plus a Kalman normalized-innovation test and altitude/speed/heading residuals expose self-inconsistent messages. Integrity →
DO-260BDrone / UAV heuristics
Remote-ID parsing plus behavioral signatures, low-and-slow, tight loiters, surface candidate uncrewed aircraft. Drone detection →
REMOTE-IDReal-time smooth map
Client-side dead-reckoning glides every aircraft at its true ground speed between fixes and corrects toward truth without snapping back. The pipeline →
DEAD-RECKONINGPattern-of-life
Automatic orbit, racetrack, loiter and grid-search classification with per-airframe behavioral baselines for deviation scoring. Baselines →
BEHAVIORALGeofencing & alerts
Circle and polygon zones with enter/exit events, emergency-squawk detection (7500/7600/7700) and observer-relative proximity thresholds. Geofence guide →
OBSERVER-RELATIVEOpen data & API
CSV, JSON and GeoJSON exports plus a documented REST/WebSocket API, reproducible analysis, not a black box. API docs →
CSV / JSON / GEOJSONWhat each capability actually rests on
| Capability | Implementation |
|---|---|
| Multi-source fusion | adsb.fi and adsb.lol polled and merged on the ICAO 24-bit address, with OpenSky as a fallback. No satellite feed. How fusion works. |
| Attention scoring | Eight weighted rule-based factors summed to 0 to 100, every contribution shown. The scoring method. |
| Integrity and spoof checks | DO-260B style NIC, NACp, NACv and SIL checks, a Kalman normalized-innovation test, and self-consistency residuals. Detection detail. |
| Consensus anomaly detection | River, scikit-learn IsolationForest, pykalman, OpenAP, stumpy and PyOD, flagging only on two-detector agreement. Baselines. |
| Drone heuristics | Remote ID parsing plus low-and-slow behavioral signatures. Drone detection. |
| Smooth live map | Client-side dead reckoning between fixes, positions refreshing about every seven seconds. |
| 3D airspace view | Radar dome, altitude histogram and situational globe rendered client-side with three.js. |
| Developer access | REST and WebSocket API, plus a public MCP endpoint for AI assistants. |
| Open data | CC-BY 4.0 anomaly benchmark, 38 documented columns. Research. |
Deliberately absent: deep learning, reinforcement learning, acoustic or radar sensing, and any claim of certification. See limitations.