Layer by layer, gate by gate, without dates we cannot keep. The evidence page is authoritative; where this page disagrees with it, the evidence page is correct.
Last reviewed .
Where the programme is
CarpatDome programme status by milestone, as of 15 September 2026
Milestone
State
Software demonstrator (v1.1, extended 7 September 2026)
Complete — node runtime, acoustic layer, integration port
Physical receiver node
Components received — assembly not started
Acoustic layer (RF-silent drones)
Software only — synthetic audio, no field capture
Multi-node TDoA localisation
Experimental — end-to-end in software, synthetic IQ only
Outdoor field experiment
Not started
Measured detection performance
None claimed
Operational deployment
Not claimed
Software, layer by layer
Everything below runs end-to-end in software on synthetic data or captured samples. 192 of 194 automated tests pass; the two excluded need a physical radio and a running broker.
Node sensors
One process per sensor: Remote ID decoding on Wi-Fi and Bluetooth Long Range, sub-GHz RF energy scanning and classification, an acoustic listener for RF-silent drones, and time-stamped IQ capture for TDoA. Scripted provisioning; node health reports timing state. Resilience under GNSS degradation is a design goal, not a validated capability.
4 LOGICAL NODES · RECEIVE-ONLY SENSING CHAIN.
C2 and fusion
Every source fuses into one contact per drone with a confidence score. TDoA cross-correlation and multilateration run end-to-end on synthetic IQ — experimental, not a validated capability. A live radar map replays 14 scripted engagement scenarios; the engagement ladder is a C2 simulation and commands no hardware.
14 SCENARIOS · NO EFFECTOR.
Integration port
Partner sensor detections come in and provider-neutral cues go out over the same versioned JSON contract — radar, EO/IR or an effector connect here. None of them exist in CarpatDome; the boundary does, and it is documented for an integrator.
MQTT · VERSIONED JSON · REST · WEBSOCKET.
Production tier
Authenticated transport, role separation, an audit log and continuous integration are scoped as a separate tier. Not built. The demonstrator runs on an isolated network and says so.
SCOPED, NOT BUILT.
Hardware phase
A node is built from COTS classes, not bespoke parts: sub-GHz SDR receivers, a dual-band Wi-Fi adapter in monitor mode, a Bluetooth Long Range receiver, a microphone array and a GNSS timing reference. A field tier adds timing resilience for GNSS-degraded operation — a design goal, not a validated capability. Components received 15 September 2026; assembly not started.
M1First physical nodes assembled and provisioned from script.
M3First outdoor Remote ID decode and first acoustic capture against a known airframe.
M4TDoA against ground truth → Field Note #001, figures achieved or plainly not achieved.
No dates. Each gate is published on the evidence page when it is passed, not when it is planned.
Supply chain
Phase A validates the concept on commodity COTS hardware, because that is the fastest honest path to a measurement. Phase B replaces the reference node with an EU-sourced, supply-chain-controlled build: component origin for RF front-end, SDR, GNSS timing, MCU and antenna is a tracked deliverable, not an afterthought.
Where we are listed
A listing is a matchmaking step, not a project. These are the places a coordinator can find us today: