Research buoys with opportunistic backhaul.

The mission

An institute moors a line of instrumented buoys along a shipping approach: currents, temperature profiles, acoustic recordings, water chemistry. The science wants everything, at full resolution, continuously. The budget allows a satellite modem on each buoy and a per-byte tariff that would be exhausted in days if the buoys transmitted what they actually measure. Meanwhile, vessels pass the moorings constantly — the institute’s own research ship on its rotations, a ferry on its daily crossing, a patrol boat on its rounds. Each one sails within radio range of data that has nowhere to go.

What breaks

The failure is arithmetic before it is radio. A buoy’s full-resolution output, multiplied by a per-byte satellite tariff, multiplied by a network of buoys, is a number no research budget survives — so operators summarise brutally, transmit a sliver, and store the rest locally for a service visit that happens a few times a year. The cost of that compromise is scientific: the interesting event is in the part that stayed on the mooring, and by the time anyone reads it, the season has moved on. The service visits themselves are expensive ship days spent, in part, as a courier service for memory cards.

The classical alternatives are to pay — continuous satellite for every buoy — or to wait. Neither uses the one asset the sea provides for free: traffic. Ships pass the buoys anyway, on schedules the institute does not control but can observe.

The architecture on this mission

Each buoy runs a node on its existing controller, queueing everything it measures in priority order over whatever short-range radio it already carries. Any equipped vessel that enters range is discovered automatically and authenticated by key — no scheduling, no operator on either end. For the minutes the geometry holds, the buoy’s queue drains onto the vessel at local-link speed, highest priority first, and configuration updates travel the other way: new sampling rates, new firmware, new priorities, delivered on the same pass. Then the vessel sails on, carrying data it cannot read — every hop is encrypted end to end — and forwards it to the institute the moment its own links allow, or at the quay.

Priority order is what makes an imperfect pass still a good pass. The drain starts the moment the link scores usable and takes the most valuable data first, so a crossing cut short by geometry or weather has already delivered the readings that mattered most; the remainder simply waits for the next hull. The satellite modem does not disappear from the design — it is demoted. Under a quota, it carries only the classes marked urgent: the storm-surge reading, the instrument fault, the daily heartbeat. The bulk, which used to be unaffordable, rides the drive-by channel free.

How the season unfolds

1. Deployment. The buoys go onto their moorings with nodes configured: keys for the fleet, priority rules for the instruments, a satellite quota for the urgent classes. Nothing else about the buoy hardware changes.
2. Ordinary weeks. Queues build in priority order. Heartbeats and flagged readings go by satellite under quota — a trickle the budget barely notices. The full-resolution record stays on the mooring, waiting for a hull.
3. The pass. The ferry’s daily crossing brings it within range of buoy three. Discovery, key authentication, drain — minutes of local-link speed, highest priority first — and the new sampling configuration delivered outbound. Nobody on the bridge is involved.
4. The interrupted pass. A patrol boat clips the edge of buoy five’s range and the geometry fails early. The most valuable data is already aboard; the queue holds the rest, intact, for the next vessel. Nothing is lost, nothing re-sent from scratch.
5. Delivery. Each carrying vessel forwards its cargo to the institute as its own connectivity allows — at sea over its links, or at the quay. The carried data arrives complete and was unreadable in transit.
6. The debrief. The journal shows what each pass drained from which buoy, and how the drained volume compares with what satellite would have cost. That record tells the institute which routes are worth equipping next, and which moorings should move closer to one.

What each mechanism contributes

• Dynamic peer discovery — vessel and buoy find and authenticate each other as range opens; no schedule, no operator, no standing session. Shipped.
• Store-carry-forward — data queued in priority order, drained while the geometry holds, physically carried, delivered onward. Shipped.
• Metered-link quotas — the satellite modem spends deliberately on urgent classes and never drains silently. Shipped.
• End-to-end encryption — the ferry carries the institute’s data without being able to read a byte of it. Shipped.
• External to Atlas — the vessels’ schedules, the buoys’ power budgets and the short-range radio fit-out. Atlas rides them; it does not provide them.

The honest boundary: the drive-by channel exists only when a vessel actually passes — Atlas does not schedule ships. Which routes pass which moorings, how often a queue gets a hull, and whether the buoy’s power budget sustains its radio are the institute’s decisions and the sea’s timetable, which is why this mission carries a Pattern badge rather than a Capability one. Atlas’s own share — the discovery, the queue, the carry and the quota — is shipped product.

What a pilot should prove

• Volume drained per pass on a real route, against what the same volume would have cost by satellite.
• An interrupted pass losing nothing: highest-priority data delivered, the remainder intact for the next vessel — from the journal.
• Urgent-reading latency by satellite under quota, running alongside the drive-by bulk channel for a full period.
• End-to-end integrity: data carried by a third-party vessel arrives complete, and demonstrably unreadable in transit.

One buoy line, one regular route, one segment of a season. The evaluation format covers the structure.