The autonomy project team with the Blueye X3 underwater robot and docking stationNTNU

Subaquatic on-call robot guards critical deep-sea infrastructure

by · New Atlas

Ancient pharaohs had round-the-clock guards watching their tombs. Today, some of the world’s most critical infrastructure sits on the seabed with far less attention, mostly because adequate tech doesn't exist. Researchers at the Norwegian University of Science and Technology (NTNU) are now developing something closer to a permanent subsea guard: an autonomous underwater robot that can live at a docking station on the seabed, leave to inspect infrastructure, return on its own, recharge, and upload data without human intervention.

The technology is an autonomous underwater vehicle (AUV) paired with its own seabed docking station. The robot inspects nearby infrastructure on its own schedule, then finds its way back to the station, latches on, recharges inductively, and offloads its data, all without a ship or a crew anywhere nearby. The availability of the docking station means the robot can remain “on-call” underwater for very long periods.

The seabed has become host to a rapidly growing share of the world's critical infrastructure, such as gas pipelines, power cables, internet backbone fiber, oil and gas installations, and, increasingly, data centers. Some of these placements are simply down to necessity; undersea cables are the only practical way to link continents with high-bandwidth data, and pipelines often have no viable overland route.

Also, the ocean offers genuine advantages, such as stable temperatures, an abundance of undisputed land, and no neighbors complaining about noise. However, putting infrastructure on the seabed comes at a cost. It's remote and largely unwatched, a combination that makes it an easy target. The cutting of a Svalbard communications cable in January 2022 and a string of sabotage incidents involving cables, fiber, and pipelines across the Baltic Sea in 2023 and 2024, some of them classified as hybrid warfare, have highlighted this vulnerability.

The installations aren't maintenance-free either. While they are built to withstand harsh conditions, the infrastructure still needs regular inspection to catch damage and wear before they lead to failures that ripple through the economies that depend on it.

Right now, keeping watch over all these means sending out a large vessel, parking it above the site, and either lowering an underwater vehicle or sending divers down to look things over directly, sometimes paired with sonar scans run from the surface. This approach works, but it's slow, expensive, and labor-intensive, characteristics that severely hinder the kind of frequent, near-continuous monitoring that today's security climate increasingly demands.

An ideal situation would be Aquaman standing guard. In his absence, however, Professor Martin Ludvigsen and his team at NTNU's Department of Marine Technology have set out to build the next best thing: a "resident" Autonomous Underwater Vehicle (AUV) that parks itself on the seabed, works to its own schedule, and only needs to be retrieved for repairs or an upgrade.

The test rig was built around a modified Blueye X3, a roughly 10 kg (22 lb) underwater robot normally aimed at the consumer and light-commercial market. For the trials, the robot was kitted out with a camera, sonar, an induction charging coil, and a magnetic mooring system so it could physically latch onto its docking station. The docking station itself sits on the seabed, wired to shore for power and communications, and does the heavy lifting.

For navigation underwater, where GPS doesn't work, the AUV uses several tools. It combines Ultra Short Baseline (USBL) acoustic positioning with visual markers read by onboard machine vision to locate the docking station and line itself up for the final approach, switching to an optical modem for short-range data exchange once it's close. Once docked, it switches over to high-frequency broadband to dump its inspection data and starts charging inductively, eliminating the possibility of exposed electrical contacts in seawater.

The underwater robot uses a combination of acoustic (1) and visual navigation (2) before docking to charge and upload the collected dataNTNU

This docking step is the core of the whole idea. The robot is useful only if it can come home. If it fails to dock, it cannot recharge or offload data and may eventually need to be recovered by a ship, which defeats much of the point.

“Docking is absolutely critical. Without an operator, the consequences of errors are much greater. If the AUV is unable to find its way back to the docking station, it cannot be recovered,” says Ludvigsen.

During the trials, the AUV carried out inspection missions, found its way back to the docking station, connected to recharge, and uploaded the collected data without human intervention. So far, it has been deployed on two occasions, spending a total of four weeks in operational service. The system completed its inspection missions and achieved a 90% autonomous docking success rate.

Now, a 90% success rate sounds solid until you remember that in a real-world, untethered deployment, the other 10% could leave the robot stranded on the seafloor. NTNU says the system must reach 100% reliability, with new deployments aimed at closing that gap planned for next Northern spring.

The Blueye robot created this 3D image of the docking station at a depth of 90 m (295 ft)NTNU

There's still a fair distance between the current realities and a fully hands-off system. The current tests were run with a tether attached to the AUV for safety, a precaution the team plans to keep using until the autonomous docking is rock solid. The underwater navigation itself is also power-hungry, as the robot has to rely on inertial sensors and acoustic Doppler velocity logging to track its position, since satellite navigation is useless once it's submerged.

Then there's the problem of the landlords of the sea, fish. During testing, fish swimming past the camera occasionally confused the machine vision software trying to interpret what it was seeing, another quirk the team is still working through.

The researchers are candid that a commercial version is still some way off, but they're not pessimistic about it.

"The tests showed that this solution works and that the operations are repeatable," Ludvigsen said. "Resident underwater vehicles are a scalable and cost-effective solution for underwater monitoring."

Given how much of the world's economy is running through seabed pipes and cables most people never think about, a robot that's able to stand watch over them for months at a time, without a ship, a crew, or a coffee break, is exactly the kind of guard duty that's overdue.

The project is being run under NTNU's Vista CAROS center and the Safeguard research initiative, with funding from the Research Council of Norway and backing from the Norwegian Academy of Science and Letters and Equinor.

The team's preprint paper is available here.

Source: NTNU