← PortfolioMachines that move

Ulysses

Dual-use autonomous underwater vehicles

Team
Akhil VoorakkaraCEO
Will O'BrienCCO
Colm O'BrienChief Engineer
Jamie WedderburnCTO
Founded
2023
Invested
2025
Links
The problem

How do you plant a meadow at the bottom of the sea?

Click the seabed to drop an acoustic beacon. Each time the vehicle passes one, its guess snaps back to where it really is.A side-on cut through the sea. The vehicle maps the seabed with sonar as it goes, and drills seed into the sediment, where it stays and grows. Seed scattered from the surface drifts away on the current. The dashed outline is where the vehicle thinks it is: that guess drifts further the longer it goes, and snaps back each time it passes an acoustic beacon.An illustration, not real data.
How it works underwater

An autonomous underwater vehicle, or , is a robot that goes about its work underwater without anyone steering it. Its older cousin, the remotely operated vehicle, is driven and powered from the surface through a cable. AUVs carry sonars and depth sensors, and survey companies send them to map the seafloor before a pipeline goes in, then back again to inspect it.

The first thing a robot loses when it dives is GPS, because radio barely gets into water. From then on it is guessing, carefully. An inertial system works out position from measured motion, a clocks speed over the seabed, and a pressure sensor gives depth. If someone has laid acoustic on the seabed, those can pin the guess down now and then. The name for all this is .

Seagrasses are flowering plants that pollinate underwater, and their meadows hold more than 10% of the ocean's carbon storage. More than a fifth of the world's seagrass has been lost in the last century and a half.

Further reading Autonomous underwater vehicle (Wikipedia)Seagrass (Wikipedia)

Why it is hard
  1. i.

    Errors that compound

    Inertial navigation has a flaw baked in. Tiny errors in measured acceleration and rotation add up into ever larger errors in position, and even the best accelerometers would drift 50 metres within 17 minutes. Picture crossing a field with your eyes shut, counting steps. Ten paces, fine. A thousand, and you're in the next field. So the robot needs outside corrections, and acoustic positioning only works if someone has set out beacons first.

  2. ii.

    Shouting through water

    Underwater, robots talk over , which is to say with sound. Sound travels at about 1,500 metres per second, carries very little data, and echoes send each message down several paths at once. Next to radio on land, the data rates are tiny. Engineers rank acoustic messaging among the most difficult communication media in use, which from engineers is strong language.

  3. iii.

    Murky, moving ground

    The places that need seagrass back are rough places to work. On the Great Barrier Reef that can mean murky water, silty seabeds and strong currents. The seabed doesn't hold still either: in one trial, keeping seeds at the right depth got harder as the sediment changed underneath the robot.

  4. iv.

    Seeds that wash away

    Most seeds never make it. In a Dutch Wadden Sea programme, seeds sown from buoys lost more than 99.9% before any seedlings took hold, very likely washed off by waves and currents. Burrowing worms bury seeds and seedlings too. Push a seed about 2 cm into the sediment, though, and it gets some protection from predators.

Further reading Inertial navigation system (Wikipedia)Underwater acoustic positioning system (Wikipedia)Autonomous underwater vehicle (Wikipedia)Underwater acoustic communication (Wikipedia)The robot that could lead a new era in seagrass restoration on the Great Barrier Reef (Great Barrier Reef Foundation)Adaptive intertidal seed-based seagrass restoration in the Dutch Wadden Sea (PLOS ONE)Refining the use of seed balls to support seagrass restoration (PubMed Central)

What Ulysses is after

Today seagrass is mostly planted by hand. Collecting and spreading seed by hand is slow, labour-intensive and costly, and it takes lots of people placing individual on a meadow. Done well, current methods restore about five hectares a year.

Ulysses wants a robot that does five hectares a day, with attachments to harvest seed, plant it and keep an eye on it afterwards. The same vehicles are meant for everything from ecological restoration to subsea surveillance.

Further reading The robot that could lead a new era in seagrass restoration on the Great Barrier Reef (Great Barrier Reef Foundation)Stewarding the Oceans (Ulysses)

How they go at it
  1. Step 1: Cheap, and lots of them

    The vehicles are built to be "up to 50x cheaper" than legacy ones, and they talk to each other, so they can work as a swarm. The Mako, Ulysses's AUV, is rated for 72 hours of endurance and 5,000 ft of depth.

  2. Step 2: Parts that swap

    The Mako's modules and swap like Lego, so the same vehicle can carry tools for fine work, from repairing infrastructure to planting seeds.

  3. Step 3: Seeds into the sediment

    For seagrass, the Mako maps the seafloor, then uses small robotic drills to put seeds straight into the sediment. In a five-day field trial at Gladstone the drills set seeds at the right depth, and the vehicle kept working in fast, murky water. The hard part turned out to be keeping seed flowing as supplies ran low.

  4. Step 4: Nobody at the dock

    Ulysses is also building Kraken, which launches, recovers and recharges vehicles, so a mission can run from start to finish without people.

Further reading Stewarding the Oceans (Ulysses)The robot that could lead a new era in seagrass restoration on the Great Barrier Reef (Great Barrier Reef Foundation)

Still open
  • What counts as success?

    Most restoration projects report habitat numbers like area or shoot density. What we actually want from a meadow is clearer water, fish and stored carbon, and measuring that is a major unsolved problem. Plenty of failed efforts never get written up at all.

  • Why do seedlings live or die where they do?

    In some studies few seedlings came up even at sites that habitat models rated suitable. Yet across the world, seedling survival predicts population growth, whatever the species, method or environment. Researchers say the disagreement needs more work.

  • How long does the carbon stay put?

    In Virginia's coastal lagoons, more than 70 million eelgrass seeds broadcast over two decades grew into 3,612 hectares of meadow, and within that time those meadows were burying carbon as fast as undisturbed ones. The carbon can stay down for decades to centuries, as long as the meadow doesn't go back to bare seabed.

Further reading Restoration of seagrass habitat leads to rapid recovery of coastal ecosystem services (Science Advances (via PubMed Central))Refining the use of seed balls to support seagrass restoration (PubMed Central)

About Ulysses

Ulysses builds low-cost autonomous underwater vehicles that operate across surface and subsea domains. The company started in ocean restoration—their robots restore seagrass 100x faster than manual planting at a fraction of the cost—and has since expanded into defense applications that leverage the same core autonomy platform.

The company has completed US Navy trials and attracted interest from US Customs and Border Protection. The founding team previously built championship Formula 1 cars, satellites, drones, and self-driving cars.

Words used here
AUV
An autonomous underwater vehicle: a robot that follows a mission underwater without a person steering it.
dead reckoning
Working out where you are from a known starting point plus measured speed, heading and elapsed time.
Doppler velocity log
A sonar that bounces sound off the seafloor to measure how fast the vehicle is moving over it.
transponders
Acoustic beacons that answer a vehicle's ping, letting it work out its range to known points.
acoustic modems
Devices that send data as pulses of sound through water, since radio does not carry.
payloads
The swappable equipment a vehicle carries for a job, such as a sonar or a seed-planting tool.
seedballs
Seeds packed into small balls of clay or sediment so they sink and stay put.
Sources