Eureka II finds its ship and returns to deck in open-water test

The robot had to find its way home before anyone could call the mission complete. In footage from an offshore demonstration near Halifax, Nova Scotia, Impossible Metals’ Eureka II located its surface vessel, navigated toward it, aligned itself and attached with a SmartHook—all without an operator controlling it. The recovery vessel then brought the autonomous underwater vehicle back to the deck.

Open water makes that return harder than a controlled laboratory test. Waves, currents and the movement of the ship constantly change the target. Eureka II first used acoustic positioning to locate the vessel below the wave zone, then switched to computer vision for the final approach. AprilTag visual markers on its top plate helped the robot line itself up before the mechanical hook secured it.

The scale of the machine explains why recovery matters. Eureka II can operate at depths of up to 19,685 feet (6,000 meters), weighs approximately 8,710 pounds (3,950 kilograms) on land and carries a 14.4-kilowatt-hour (kWh) battery. Three sets of robotic arms are designed to collect mineral nodules from the seafloor.

Concretely, a self-recovering robot could be sent on subsea missions more often, with less waiting for crews to intervene. Impossible Metals says that could reduce labor, downtime and exposure to risk at sea, while making commercial operations easier to scale. Impossible Metals presents the achievement as relevant as interest in critical minerals, subsea infrastructure and ocean data grows, while environmental concerns become more important for ocean operations.

The boundary is clear: Halifax was an offshore demonstration, not evidence that Eureka II is already running routine missions without support. Impossible Metals CTO Jason Gilham called autonomous recovery a “gating function” for scaling, while executive chairman Steve Curnutte said the company is building toward greater autonomy and precision. For now, the test shows a robot completing a difficult handoff in real ocean conditions; the next question is how reliably that performance can be repeated in daily field operations.