How a Robot Glider Avoids Hitting Sea Ice | Under-Ice Gliders Explained

A robot glider that has spent two days under Antarctic sea ice has to come up for GPS. But if there's ice overhead, it can get stuck or damaged, and nobody is there to tell it.

Researchers at the National Oceanography Centre and the British Antarctic Survey gave their Slocum gliders a second computer, a "backseat driver," that watches for signs of ice and decides when it's safe to come up. In the Weddell Sea, their gliders ran 16 missions, 6 of them under the ice, and the team reports that every one ended with the glider coming up without hitting ice.

In this independent explainer, I walk through how it works and what the paper shows.

In this video

• Why the polar oceans matter, and why they're so rarely measured
• How a glider flies, and how it estimates where it is between GPS fixes
• Why that routine breaks down under ice
• The backseat driver: a second computer that watches the sensors and can change the plan, while the glider's own computer keeps flying it
• Four signs of ice: water near freezing, an upward-looking sonar, a satellite ice map, and GPS itself. The glider creeps up tail first and only surfaces if a signal comes through
• The tank test, where the sonar couldn't reliably detect 10 cm ice slabs, and why the sonar alone isn't enough
• Levels of urgency: later on, the glider needs fewer signs, but the tests get stricter
• A false alarm off southwest England: out-of-date sonar data kept a glider down, and the other levels still brought it up
• Mission 9: the temperature and sonar said clear, but no GPS signal came on two tries. It got its fix that afternoon
• Mission 12: the sonar saw ice and the glider turned back down at about 14 m, and after about 46 hours its position estimate was about 8 km off
• The limits: missions of about two days, a sonar cutoff of 5 m, more than double the power draw on a test dive, and trials in early summer, while the temperature check works best in winter

For engineers

The backseat computer sits in the Slocum's science bay and can modify the frontseat's mission in real time. The only added sensor is an upward-looking altimeter at 27°. Ice sensing uses the median temperature over 20–50 m against a −1.78 °C threshold, ice draft from CTD depth minus altimeter range (5 m minimum detectable draft), a pilot-drawn ice-edge boundary from satellite products, and GPS connectivity: pitch battery forward and slight positive buoyancy, so the tail antenna surfaces first without inflating the air bladder.

Five urgency levels combine these with surfacing times; the highest is a leak. Under the ice the glider navigated by dead reckoning, so the ice-edge boundary wasn't used to control surfacing. Table 6: 0.081–0.093 A average current without the backseat system vs 0.236–0.247 A with it, diving or climbing without the altimeter.

The paper

"Enabling Under Ice Glider Operations: A Backseat Driver Approach"
Yaomei Wang, Benjamin Allsup, Georgios Salavasidis, Stephen Woodward, Adeniyi Adenaya, Mike Smart, Alexis Johnson, Felipe Marques Dos Santos, Jake Ludgate, Daniel Matterson, Alvaro Lorenzo, J. Alexander Brearley, Alexander Brian Phillips
National Oceanography Centre, Southampton, UK, and British Antarctic Survey, Cambridge, UK
Journal of Field Robotics 43(5), 3292–3311 (2026), doi:10.1002/rob.70193
Open access (CC BY 4.0), March 2026

Read the paper

Supported by the BIOPOLE National Capability Multicentre Round 2 funding from the Natural Environment Research Council (Grant No. NE/W004933/1) and by the European Union's Horizon 2020 research and innovation program (Grant No. 821001, SO_CHIC).

The paper goes much deeper: the ice-draft equation, the tank data, the full urgency logic and the mission maps. If you work on polar gliders or under-ice autonomy, it's worth reading in full.

This is an independent explainer. I'm not affiliated with or endorsed by the authors, the National Oceanography Centre or the British Antarctic Survey, and the research is all theirs. The animations are my own illustrations of what the paper describes, not footage from the missions; drawings marked schematic are illustrations. Any mistakes in the explanation are mine.

I'm Bora Celik from Piccard. I explain new research in ocean science and robotics.