
Autonomous submarines are mapping the ocean floor: from sonar to AI
Only about a quarter of the seafloor has been mapped in detail. Autonomous vehicles from the Gulf Institute of Marine Sciences are working to fill that gap in the depths of the Aegean.
Table of contents
It is often said that we have more detailed maps of the surfaces of the Moon and Mars than of the floor of our own oceans. Indeed, the share of the seafloor mapped at high resolution is only about a quarter today. One of the most promising ways to close that gap is uncrewed, autonomous underwater vehicles. The torpedo-shaped vehicles developed by the Gulf Institute of Marine Sciences dive for hours along preset routes, scanning the seafloor with multibeam sonar. AI software on board recognizes obstacles, corrects the route in real time and flags interesting structures for a second look. The team is preparing to share the results of the first long missions it completed in the Aegean this autumn.
Seeing in the dark with sonar
Because light is absorbed quickly in seawater, the ocean is completely dark below a few hundred meters. That is why sound rather than light is used to map the seafloor. Multibeam sonars send out hundreds of sound beams that fan out beneath the vehicle and measure how long they take to bounce off the seafloor and return. Since sound travels through water at about 1,500 meters per second, that time is enough to calculate the depth of each point precisely. The result is a three-dimensional map of the seafloor.
Ship-mounted sonars can scan wide areas quickly, but the level of detail drops as depth increases. At a depth of a few thousand meters, a ship's sonar can only distinguish features tens of meters across. Autonomous underwater vehicles, by contrast, can glide a few tens of meters above the seafloor and take much higher-resolution measurements. This lets small-scale features such as shipwrecks, underwater landslides, fault traces and cold seeps make it into detailed maps.

Mapping the ocean floor means opening the unread pages of our planet.
Sarah Bennett
Machines that correct their own course
One of the biggest challenges for autonomous underwater vehicles is that satellite positioning signals cannot be received underwater. The vehicles therefore estimate their position using motion sensors, acoustic devices that measure their speed relative to the seafloor and, from time to time, acoustic links with the ship on the surface. Because small errors can add up on long dives, the software can correct its position by recognizing distinctive features in previously mapped areas. The method is quite similar to the techniques used by autonomous vehicles on land.
The AI software used in the Gulf Institute of Marine Sciences' vehicles processes sonar data in real time during the dive. The software can change course when it recognizes rock outcrops or abandoned fishing nets in its path. And when it notices an unexpected feature, it updates its own plan to pass over that area again, more slowly and at a lower altitude, at the end of the dive. Dr. Connor Eastman of the institute explains that this capability significantly reduces survey time and ship costs.
First maps from the depths of the Aegean
This autumn, the team completed a three-week field campaign in the central Aegean, in a basin whose deepest point exceeds 1,000 meters. With dives totaling more than 400 hours, two vehicles mapped an area of about 250 square kilometers at high resolution. Preliminary assessments show that the maps contain previously unknown traces of underwater landslides, pockmarks on the seafloor that may indicate gas seeps and scarps that may be linked to active faults. The data are now being studied in detail.
| Topic | AI & Robotics |
|---|---|
| Highlights | robotics, artificial intelligence, ecology |
| Reading time | 6 minutes |
| Written by | Sarah Bennett |
Key takeaways
- Only about a quarter of the seafloor has been mapped at high resolution.
- Autonomous underwater vehicles scan the seafloor up close with multibeam sonar.
- The first Aegean missions identified traces of underwater landslides and possible gas seeps.
These data are valuable for both earth scientists and marine ecologists. Underwater landslides can be important for assessing tsunami hazard, while the areas around gas seeps can host unique communities of life. Once the maps are complete, the team plans to share them as open data and contribute to international seafloor mapping efforts. According to Eastman, the goal is to light up the dark depths of the Aegean, even if piece by piece, and leave that knowledge to future generations.
What do you think about this post?



































































Leave a comment