SkyMesh / Operations
Participant map
Real participants · satelliteSolid nodes: assigned room positions. Dashed nodes: unplaced participants in a staging row, not GPS locations. Scenario overlays never become live readings.
Join the mesh
Scan to turn your phone into a sensor node. Detection runs on the device — audio never leaves it.
Drone audio test
Play near a participant phone to exercise its real microphone detector.
Real DADS drone audio. Detection should trip within about a second. Volume up, speaker 10–30 cm from the phones.
Sensor confidence
Each node's on-device CRNN confidence, drawn from records that gossiped here. Dashed line is SkyMesh's 0.35 threshold.
Scan the QR code, allow microphone access, then admit the phone here.
Network topology
Adjacency is imposed here — every phone can physically reach every other. It stands in for the radio range of a real deployment, and the bridge preset is what makes partition-and-heal visible.
Scenario controls Simulation
nominalDemo only. Joined nodes receive a separate simulation alert, never a microphone reading. Impact and isolation cut mesh links. Simulation notices use the control channel and still arrive during those cuts.
Live to admitted nodes in this session. Acknowledgements confirm a participant saw the demo, not mesh delivery. Notices expire after 2 minutes.
Run a scenario to notify participants.
Scenario activity
0 eventsScenario events land here.
Sensor inspector
Select a sensor on the map to inspect its readings and connections.
link emulation
Network conditions, not detection. These stand in for radio links: latency and packet loss the phones would face in the field but never see on one WiFi. Cutting a link is how partition-and-heal is demonstrated.
no links yetSensor directory
| node | p | pos | neighbours |
|---|
Network simulation Synthetic
no microphones · no relayPlan an array against a drone that does not exist. Place synthetic sensors, draw an ingress a threat would really fly, and watch a timed mesh acquire it and gossip the contact outward — per-node replicas, 50 ms links, packet loss and all. Nothing here touches a participant phone or the relay.
- See what fraction of a route the array would actually hear, and where it is blind.
- Ask where the next sensor should go, for the area or for a specific ingress.
- Open any node and redraw the map from its replica, to watch two nodes disagree.