Collecting nodules while disturbing as little of the seafloor as possible.

Autonomous vehicles already do most deep-sea mapping. The new idea is to let them do the collecting too: survey a nodule field first, build a map of where the nodules are dense and where the sponges, corals and other animals live, then plan a route that takes the nodules and avoids the habitat. The vehicle then follows that route without a person driving.

The most-discussed example is Impossible Metals' Eureka series. Eureka II is a test vehicle; Eureka III is the full-scale concept. Both are untethered — no cable to the ship — and hover a short distance above the seabed on controlled buoyancy rather than driving on it. Robotic arms pick individual nodules one at a time. A camera and on-board computer vision look at each nodule first and, if there is visible life on it, the vehicle is meant to leave that one where it is. When the batteries run low, the robot surfaces, swaps packs on the vessel, and goes back down.

Tethered vehicles (the usual ROV) are powered and controlled through a cable, which is reliable but limits how many can work and how far they can roam. Untethered vehicles trade that cable for batteries and autonomy: freer to move, but limited by energy and by how well the software copes with the unexpected.

The Coanda effect is the tendency of a moving fluid to follow a nearby curved surface — it is why a stream of water from a tap bends around the back of a spoon. A Coanda collector head uses a jet of water directed along a curved hood. The flow hugs the hood, the pressure beneath it drops, and that pressure difference lifts the nodules off the sediment and into a duct.

Compared with the rakes, teeth and scrapers of earlier designs, the head does not have to dig. It rides just above the bed, and in principle takes nodules while leaving much of the mud behind. The jet strength and hood height can be tuned to the size of nodule and the softness of the sediment.

Every nodule has to get 4–5 km up to a ship. The conventional answer is a vertical riser: a long pipe with pumps, up which a slurry of nodules, sediment and water is lifted. On deck the nodules are separated and the leftover water — carrying fine sediment — is discharged back to sea. That discharge is the dewatering plume. Closed-loop riser designs pipe that return water back down instead of releasing it mid-water, which moves the plume rather than removing it.

A skip-lift system does away with the slurry altogether. The collector delivers nodules to a filling station on the seafloor, where they are loaded into sealed containers (skips). A winch on the vessel lifts each skip, empties it, and sends it back down. Because no sediment-laden water is pumped up, there is no dewatering plume to manage; the collector plume on the seafloor remains.

Older operations learned what they had done after the fact, from surveys weeks or months later. Newer plans put sensors on the collector, in the water column on moorings, and on landers downcurrent of the work. They measure turbidity (how cloudy the water is), noise, currents and water chemistry continuously and send the readings to a dashboard on the vessel and, potentially, to the regulator.

The point is not just to watch. Each sensor has a threshold written into the operating plan; if a plume drifts further than modelled or turbidity spikes, the system is supposed to slow or stop the collector until conditions are back within limits. Monitoring like this is a lease requirement under the federal rules for preliminary activities, not an optional extra.

Nodules are not spread evenly. Detailed survey data — from the same AUVs described above — can show which patches are dense enough to be worth collecting and which are not. Precision approaches plan 'harvest lanes' only through the dense patches, leave set-aside refuge areas untouched so animals can recolonise, and route around any mapped sensitive habitat such as sponge fields or seamount slopes.

The result, if it works, is a smaller total footprint for the same amount of material, and a landscape that still has untouched ground scattered through it rather than one continuous strip. It is as much a planning discipline as a machine.