Airport airside operations.

The mission

Between the last departure and the next arrival there is a window — sometimes twenty minutes, sometimes four — and in it the airfield does its housekeeping: runway surface inspection, lighting checks, FOD sweeps, bird control along the grass strips. A drone does this work faster than a vehicle and sees more, which is why airports are adopting them. But it does the work inside the most procedurally unforgiving airspace there is. The operating approval says precisely what happens when the command link drops: the aircraft aborts, and until it is confirmed down and clear, the runway it was inspecting is not available. A communications hiccup does not cost a video frame here; it costs runway minutes, and runway minutes are the most expensive minutes the operator owns.

What breaks

The airfield looks like easy radio and is not. It is a kilometres-wide sheet of open ground ringed by hangars, fuel farms and terminals full of metal; the RF environment is dense with radar, navigation aids and every operational system the airport runs, and new emitters are coordinated conservatively for exactly that reason. The airport’s own operational network — Wi-Fi and fixed infrastructure built for vehicles and ground crews — covers the aprons well and the far ends of the runway unevenly. Public cellular covers the airfield the way it covers any large open site: adequately, except where it does not, and with cells that load up with every terminal full of passengers.

So the single-link options are all uncomfortable. The operational network alone has soft ends exactly where the inspection goes. One cellular modem alone hands the runway’s availability to a public network’s busy hour. A dedicated new radio link means spectrum coordination inside the most carefully managed RF environment in civilian use. Each is a single point of failure attached to an abort procedure.

The architecture on this mission

The aircraft’s node bonds the two networks the airfield already has — the operational network and one or two cellular carriers — into a single encrypted tunnel to the airside operations workstation. The command channel is duplicated across both networks for the entire sortie: not failed over, duplicated, so that a soft patch at the runway end or a loaded cell during the morning bank is a non-event rather than a race against the link-loss timer. The two networks share no masts, no backhaul and no maintenance windows, which is precisely what makes their duplication meaningful. Nothing new is emitted that the airport must coordinate: both networks already exist, and Atlas is software riding inside them.

The class system encodes the procedures’ priorities directly: the command-and-abort channel holds a reserved floor that inspection video can never crowd out, video sheds quality before continuity, and the full-resolution survey imagery drains as bulk after the aircraft is down. Every link is probed several times a second and every state change is timestamped into the event journal — which is where this mission’s second deliverable lives. An airside drone programme answers to a safety office, and a safety office runs on evidence: the journal gives it a per-sortie, per-network record of what the command channel actually did, in a form a review can cite.

How the inspection window unfolds

1. Before the window. The crew stages at the holding point; the node brings up the operational network and cellular and bonds them. The dashboard shows both links green and measured before ops confirms the window is open.
2. On the centreline. The drone runs the surface inspection at speed, video streaming on whichever legs currently carry best, command duplicated on both. The tower’s clearance clock is running; nobody is thinking about radio.
3. The soft patch. At the far threshold the operational network thins, as it always does. Traffic re-weights toward cellular within a second — the journal logs the seam; the sortie does not notice it.
4. The loaded cell. A morning arrivals bank fills the nearest cell and its uplink sags. The bond leans back onto the operational network; the abort timer is never approached, because the abort channel never depended on either network alone.
5. Window closes. The aircraft recovers at the holding point. Runway handed back on schedule; the full-resolution imagery drains as bulk traffic while the next movement lands.
6. The safety file. The sortie’s journal — both networks, every seam, every re-weight, worst measured interval — goes into the programme’s evidence pack. Fifty sorties later, that pack is the argument for more windows, not fewer.

What each mechanism contributes

• Dual-network bonding — the operational network and cellular as one tunnel with one address; no new emitters to coordinate. Shipped.
• Command duplication — the abort channel on both networks at once, so no single network’s bad minute starts the link-loss clock. Shipped.
• Class floors and ordered degradation — command reserved, video elastic, survey imagery as bulk after landing. Shipped, opt-in.
• The event journal — a per-sortie, per-network record of the command channel’s real behaviour, built for the safety office’s evidence pack. Shipped.

The honest boundary: Atlas does not change the rules — the link-loss procedure, the operating approval, the coordination with the tower and the decision to fly remain the airport’s, and if both networks genuinely fail at once, the abort happens exactly as the procedure says. What Atlas changes is how often that moment can occur, and what the safety office can prove about every sortie either way. The drone, its airworthiness and the airside safety case are the operator’s.

What a pilot should prove

• Command-channel continuity across the full inspection route, bonded versus each network alone, from the built-in measurement.
• Zero link-loss aborts across the trial period attributable to seams a single-network baseline would have aborted on.
• Ordered degradation demonstrated: video shedding quality through the worst measured interval while the command floor held.
• A per-sortie journal accepted by the safety office as evidence — the artefact that turns a trial into a programme.

One airfield, one aircraft, one month of inspection windows. The evaluation format covers the structure.