Skip to main content

Overview

Agentic, multi-domain command and control, designed for counter-UAS.

DomeCommand is a command-and-control system that runs a super OODA loop: it fuses every sensor into one live picture, reasons over that picture with agents, plans against it with a solver, and holds every consequential action at an authority gate the commander set.

It is built for counter-UAS, and it does more than monitor. It commands aircraft over MAVLink and vendor protocols, so the same system that shows you a track can task something to do about it.

The DomeCommand console during a running engagement
The DomeCommand console during a running engagement
Defenders in flight on the left, tracks on the right with the sensors that contributed to each, and the current plan below.

:::note On the term "Super OODA loop" is Michael Raska's (RSIS, NTU). We use it because it describes what the system does: one operator runs one loop at a time, and DomeCommand runs every loop in parallel. :::

What sets it apart​

Open stackWorks with the standards the field already speaks: MAVLink, MQTT, SAPIENT, ASTERIX, MISB ST 0601, ADS-B, Remote ID, CoT. No proprietary bus, and no requirement to replace what you own.
Configurable and extensibleThe same software, reskinned for the mission. An airbase and a water treatment plant run identical code with different doctrine, different sensors and different action catalogues.
Advanced sensor fusionNo dependence on one sensor or one algorithm. Several techniques run together to produce the best available picture, and fusion branches on what was measured rather than on which vendor sent it.
AgenticType what you want in plain language and confirm it. Agents classify, assess intent, narrate what changed, and draft scenarios and rules for you to approve.
Runs at the edgeThe whole stack fits on a single-board computer at the last mile, close to the sensors, and keeps working when the uplink does not.
Simulation and trainingBuild scenarios with AI, rehearse before anything is fielded, and train staff against repeatable fights. Several simulation modes, including a game-engine world for camera imagery.

How it fits together​

EXTERNALINTEGRATIONSENGINEDECISION SUPPORTCOMMANDSENSORSRadar, EO/IR turrets, RF and SDRreceivers, acoustic arrays,Remote ID and ADS-B.AIRCRAFT & EFFECTORSMAVLink autopilots, DJI docks,interceptors, jammers.Commanded back, under gates.PARTNER SYSTEMSSensing pods, from their cloudor ours. Tracks out toTAK and CoT consumers.⇄PROTOCOL ADAPTERSOne adapter per protocol.Everything normalises toobservations, detections andtaskable assets.No proprietary bus.MAVLINKMQTTDJI CLOUDSAPIENTASTERIXMISB 0601ADS-BREMOTE IDSIM▶World modelTracks, assets, zones,one time-aware store.Fusion & trackerFiltered association,confidence, CPA and TTI.SolverOptimisation over assetsand threats, in parallel.Agent hierarchyCopilot, threat, sensor ops,planning, briefing.Scenario & simPlain-language authoring,attached environments.Runs & replayEvery run recorded,replayable, defensible.DISTRI RUNTIMEChannels scope what each agent sees.Connections gate what it may command.▶RULESOne sentence each, withlive fire counts.POSTURE MAPPINGEvery effect againstevery posture cell.GATESAuto, notify, confirm,two-person, withheld.ALERTS & SITREPSPlain language, tracedto the evidence.▶OPERATORSets posture.Approveseffects.Owns everycall thatcommits.RUNS ON ONE WORKSTATION · EVERY ARROW AUDITED
System architecture. Data flows left to right, from the sensors and aircraft on site, through the protocol adapters, into the engine, and out as decisions the operator approves.

Green is what DomeCommand runs deterministically. Amber is doctrine and the gates. The operator sits at the right-hand end of every arrow that commits anything.

What it does​

  • Fuses every sensor into one picture. Radar, RF, EO/IR, acoustic, Remote ID and ADS-B collapse into one track per object, each with a confidence you can take apart. See Sensor fusion.
  • Reasons over that picture. Agents classify, assess intent and narrate what changed, and they draft scenarios and rules on request. See Agentic C2.
  • Plans against it continuously. A constraint solver authors candidate plans and scores them on leakage, cost and coverage. See Solver configuration.
  • Gates every consequential action. Five gates per effect, set by posture, enforced in the engine rather than in a prompt. See Decision support.
  • Commands aircraft. Arm, take off, move to, follow a route, return to base, over MAVLink and vendor protocols. See Integrations.
  • Rehearses and trains. Author a fight, run it, change one thing, run it again. See Simulation.

Where to start​

If you want toRead
Connect your first sensorGetting started
Understand the loop end to endHow it works
See how the two reasoning engines fit togetherDecision support
Work out which sensors you needSensor types
See what this build already speaksIntegrations
Run a fight with no hardwareSimulation
Evaluate it as something to deployDeployment