Mission library · Energy & infrastructure

Powerline and pipeline inspection beyond visual range.

A transmission corridor is hundreds of kilometres of exactly the terrain mobile networks serve worst. This mission is about flying it anyway: coverage holes as non-events, no imagery lost to dead zones, and a communications cost the inspection contract can actually carry.

Who flies it

Transmission and pipeline operators and their inspection contractors flying long corridors beyond visual range.

What breaks

Single-carrier coverage is a patchwork along the route; every dropout means hold, climb, return — and imagery quietly stops being streamed at all.

What Atlas contributes

Two or three carriers bonded with command duplicated across them; imagery queues through dead zones; per-SIM quotas keep the bill predictable.

Runs on

The aircraft’s companion computer and the crew vehicle’s laptop — both ends mobile, ordinary SIMs. One Linux binary, one configuration file.

The mission

A 400 kV line crosses a hundred and fifty kilometres of farmland, forest and low hills. The operator flies it in segments: the aircraft works the corridor with a zoom and thermal payload — insulators, clamps, vegetation encroachment — while the crew leapfrogs ahead by road, and the imagery feeds the asset-management system that decides which towers get a climber this year. The same shape of operation covers gas pipelines, rail catenary and canal embankments: long, thin, remote, repetitive. Repetitive is the point — the economics only work if a segment flies on schedule, every time.

What breaks

Corridors go where land is cheap, which is where coverage is thin. Along any long route the single-carrier link is a patchwork: strong near villages, absent in the forested stretches, flapping on the boundaries in between. Each dropout has a cost written in the operations manual — hold, climb, return — and an abort thirty kilometres from the crew vehicle is an hour of lost daylight. Meanwhile the survey itself produces gigabytes against a link of megabits, so operators quietly stop streaming imagery at all and discover at day’s end that the one anomaly worth a second pass is forty kilometres behind them. And a fleet of always-on modems, left unmanaged, produces a data bill that surprises someone every quarter.

The architecture on this mission

Corridor inspection coverage geometry An inspection aircraft over a transmission corridor holds command on one carrier while the other carrier's coverage hole passes; the crew vehicle is the mobile ground station. CARRIER B — NO COVERAGE C2 HOLDS ON CARRIER A Inspection aircraft working the corridor Crew vehicle mobile ground station
The scene: the aircraft is inside one carrier’s coverage hole (shaded) — that path dies at the boundary (crossed) while command holds on the other carrier. Imagery captured in the hole queues on board and drains when coverage returns.

The aircraft carries modems on two or three carriers, each an independently measured link, bonded into one encrypted tunnel with a fixed address. Command-and-control is duplicated across two carriers for the whole flight — coverage holes stop lining up with aborts because the carriers’ holes are in different places — and endpoint roaming carries the session across every cell handover and mid-flight re-addressing without a reconnect. The crew vehicle is the mobile ground station and runs the same node with its own bonded links, so both ends of the tunnel move and neither end’s network changes are events.

Inspection imagery rides as bulk-class traffic: streamed opportunistically where capacity exists, queued store-carry-forward through the dead stretches, drained automatically — priority first — the moment coverage returns. Nothing is lost to a dead zone; it is merely late. Priority findings the operator flags in flight travel as high-class traffic that gets through even on the thin segments. Per-SIM quotas cap each modem’s monthly spend, with the standing rule that a quota never takes down the last usable path.

How the survey day unfolds

  1. Segment brief. The crew vehicle parks at the day’s first pull-off. Aircraft and vehicle nodes come up, the bonded tunnel forms across two or three carriers, and the dashboard shows each carrier’s measured state against yesterday’s coverage map.
  2. Working the towers. The aircraft flies the line — insulators, clamps, vegetation — with command duplicated across two carriers the whole time. Imagery streams as bulk-class traffic wherever capacity exists.
  3. The forested stretch. The aircraft enters a known hole on one carrier. Command holds on the other without a switchover event; the imagery stream pauses and queues on board. Nobody holds, climbs or returns.
  4. The find. The operator flags a cracked insulator mid-flight. The flagged frames travel as high-class traffic through the thin segment immediately; the rest of the survey stays patient in the bulk queue.
  5. Coverage returns. The queue drains automatically, priority first, while the crew leapfrogs to the next pull-off. Endpoint roaming carries both ends’ sessions across every cell handover and re-addressing on the way.
  6. End of segment. The journal has quietly produced a per-carrier coverage map of the day’s kilometres, and the quota counters state the day’s data spend — the numbers that plan next season’s segments and modem mix.

What each mechanism contributes

  • Multi-carrier bonding with duplicated C2 — the abort-per-dropout economics inverted; the corridor’s patchwork becomes one usable fabric. Shipped.
  • Endpoint roaming, both ends — aircraft and crew vehicle both mobile, sessions bound to keys rather than addresses. Shipped.
  • Store-carry-forward for imagery — the survey product survives the coverage map instead of being censored by it. Shipped, opt-in.
  • Metered-link quotas — wire-byte accounting per SIM, predictable cost per segment. Shipped.
  • Continuous measurement — every flight quietly produces a coverage map of the corridor, per carrier, which plans the next season’s segments and modem mix. Shipped.

The honest boundary: Atlas contributes the link layer. The aircraft, the payloads, the flight permissions and the inspection analytics are the operator’s — and where a corridor stretch genuinely has no carrier at all, the answer is a relay node or a satellite link joining the bond, not magic.

What a pilot should prove

  • Abort statistics on a real corridor segment: the bond versus each single carrier, across enough flights to be boring.
  • Imagery completeness: one hundred percent of the survey product delivered, including everything captured inside dead zones.
  • A per-carrier coverage map of the corridor generated from the built-in measurement alone.
  • A month’s communications cost under quotas, including the duplicated command traffic, at a number per kilometre the contract can absorb.

The evaluation format describes how such a trial is structured.

Request a briefing

Flying a corridor that keeps aborting?

Bring the route and last season's dropout log — we will define what a bonded pilot on the same corridor should prove.