Mountain search and rescue with an airborne relay.

The mission

A hiker is reported missing in the late afternoon. The probable area is a fan of ravines and forested slopes falling away from a ridge road — the kind of terrain where a ground team needs an hour per kilometre and the daylight is already spending itself. The rescue service deploys two aircraft from the trailhead: one flies the search pattern low along the ravines with a thermal camera; the other climbs and holds an orbit over the ridge.

The search pattern is dictated by where a person is likely to be — the drainage lines, the last known position, the terrain traps. It is emphatically not dictated by where the ground station’s radio can reach. That difference is the entire communications problem of mountain SAR.

What breaks

A ground station at the trailhead talks to the low aircraft over a direct radio link. The moment the aircraft descends behind the first spur, that link is gone — not degraded, gone, because VHF-and-up radio does not bend around rock. Cellular coverage in the same terrain is a patchwork that follows the villages, not the drainages. The classical answers all cost something the mission does not have: flying higher wastes the thermal camera’s resolution; moving the ground station takes time and often merely trades one shadow for another; adding transmitter power does not help at all, because the problem is geometry, not link budget.

There is a second, quieter failure. Even when the link only stutters, a conventional video-link architecture treats each dropout as a disconnection: streams reset, the operator re-establishes, and the detection that flashed across the screen during the gap was never recorded anywhere. In a search, a lost detection is not an inconvenience — it is potentially the outcome.

The architecture on this mission

Every platform runs the same software node on its companion computer: the ground station at the trailhead, the high orbiting aircraft, the low search aircraft, and — optionally — vehicle kits and team radios. Together they form one encrypted L3 mesh. No controller, no cloud, no configuration ceremony at the trailhead: nodes authenticate by key and join.

The high aircraft is the load-bearing move. From its orbit over the ridge it holds line of sight both to the trailhead and down into the ravines the ground station cannot see. The low aircraft’s traffic hops through it: two short links replacing one impossible long one. Short hops are not merely a workaround — a link a fraction of the length improves the signal geometry out of all proportion to anything transmitter power could buy, which is why the relay flies instead of the antenna growing.

Each leg is its own measured, encrypted link. The mesh floods link state continuously, so when the low aircraft rises back into direct view of the trailhead, traffic takes the direct path without anyone touching anything; when it sinks behind rock again, the relay path carries it. A lost node is detected and routed around in roughly a second at default settings. If the team carries cellular-capable kits, whatever patchwork coverage exists joins the bond as one more path — used when it helps, ignored when it is gone.

How the search unfolds

1. The callout. Late afternoon, a missing hiker, a fan of ravines. The team stages at the trailhead; both aircraft power up, their nodes authenticate by key and the mesh forms — no configuration ceremony while the daylight spends itself.
2. First legs. The relay aircraft climbs to its orbit over the ridge; the search aircraft drops into the first drainage on the direct link, every hop measured several times a second.
3. Behind the spur. The search pattern takes the low aircraft out of the trailhead’s sight. The direct link dies against the rock — and traffic is already flowing through the orbiting relay, two short hops replacing one impossible long one. The operator watching the thermal feed sees nothing change.
4. The gap. For a moment even the relay geometry breaks. The detection that flashes across the screen queues store-carry-forward on the aircraft and delivers, geotagged and in priority order, on the next contact — late instead of never.
5. Battery rotation. The relief relay joins the mesh on climb-out; the tired one lands. Routes re-form on the new topology in roughly a second, and the search aircraft never comes home for it.
6. The debrief. The per-sortie link record shows exactly where coverage held and where it did not — the data that places the next search’s relay orbit before anyone launches.

What each mechanism contributes

• Mesh routing with per-hop encryption — the relay geometry itself: every leg protected by its own session, the relay aircraft an authenticated member node, routes recomputed as the aircraft move. Shipped.
• Traffic classes — command and the aircraft’s position rank above video, so the moment a hop narrows, the stream sheds quality but control never contends for the thin pipe. Classification reads the traffic itself; the autopilot and camera are unmodified. Shipped, opt-in.
• Store-carry-forward — geotagged detections and events queue on the aircraft through any gap and deliver on the next contact, in priority order. The detection seen during a dropout arrives late instead of never. Shipped, opt-in.
• Continuous measurement — every link’s round-trip, loss and jitter are recorded throughout the sortie, so the debrief can show exactly where coverage held and where the next search should place the relay orbit. Shipped.

The honest boundary: Atlas contributes the network — the aircraft, their endurance, the search doctrine and the rescue itself are the team’s. Nothing here has been validated against deliberate interference in the field; what the mechanisms demonstrably survive is the terrain-driven loss, shadowing and dropout this mission is actually made of.

What a pilot should prove

• Continuous command and telemetry for the low aircraft across defined shadow zones where the direct link measurably fails.
• Detection delivery: events generated during induced link gaps arrive complete, in order of priority, after reconnection.
• Relay handover: the high aircraft is swapped mid-sortie (battery rotation) and routes re-form without operator action.
• A per-sortie link record from the built-in measurement, good enough to plan the next deployment’s relay orbit from.

Two aircraft, one valley, two days. The evaluation format describes how we structure exactly this kind of trial.