October 28, 2021

I, Science

The science magazine of Imperial College

How robots could be used to help restore the natural beauty of coral reefs...

A group of scientists and engineers want to rebuild the world’s shattered coral reefs using “swarms” of intelligent robots. They say they already have the technology: artificial intelligence inspired by ants and bees to control robots that will have sight and dexterity underwater. Money, of course, is all that holds them back, so they’ve turned to “crowdfunding”: inviting donations from members of the public who want to support the plan. They call their project “Coralbots”.

The Coralbots team will use this little guy to test out their ideas.
The Coralbots team will use this little guy to test out their ideas.

The problem

Coral reefs around the world are in quite serious danger. The worst threat is global warming, which will hurt all reefs over the next few decades. But Coralbots aren’t intended to solve that problem. Instead, they may be a solution to natural, accidental and preventable damage to coral reefs, damage that undermines reef ecosystems, but which can be repaired by hand – human or otherwise.

Coral reefs are beset by hurricanes, ships running aground, destructive fishing methods, litter and aggressive species, among other things. A reef is built by the hard but brittle skeletons of coral colonies. Corals are small animals that can’t move, and they build their branching, plant-like colonies as they grow and multiply. Under impact, they fragment. The intricate reef environment falls to pieces, and corals that sink to the sea floor usually die. If they’re buried by litter or overwhelmed by predators, they can’t run or hide to escape them.

This matters because corals are the foundations of reef ecosystems, some of the most heavily and diversely populated places in the sea. They cover a thousandth of the total ocean area, but are home to a quarter of its species. Broken reefs can’t support as much variety or volume of life as healthy ones, and they are at more risk of death and decline by other causes, such as coral bleaching, disease and pollution. And reefs support hundreds of millions of people with seafood and tourism. Coastal people and international fish stocks depend on them. They are estimated to be worth hundreds of billions of dollars to the world economy.

Colourful tropical reefs like Australia’s Great Barrier Reef rival rainforests for their biodiversity. They bring life to what would otherwise be infertile water by capturing the sun’s energy, like plants. Some are food to other organisms, many of which people catch. Shallow-water reefs are major tourist destinations, from which coastal people make a living. And they protect coastlines from waves and rough weather.

Thriving deep-water reefs in colder waters, such as off Scotland, which hardly anyone knows about, are also important to ecosystems and industry. Like less colourful reefs, their stony, bushy colonies provide habitats to many other sea creatures. Some of these are food for the fish that we catch commercially – fish stocks that support and feed people who don’t know the reefs exist.

A fish on a reef 684m deep and 300 miles west of Scotland. Credit: Heriot-Watt University.
A fish on a reef 684m deep and 300 miles west of Scotland. The white coral is Lophelia pertusa, found throughout the world’s deep oceans. Credit: Heriot-Watt University.

Yet it’s fishing that is the biggest threat to deep-water corals now. Animals are caught from near the seabed by bottom trawling, which involves dragging a giant net over the seabed. The nets break corals apart, drag them along the bottom, and pull them up to the surface. Trawlers can bring up whole colonies from hundreds of metres down. Since oceanographers began regularly using underwater vehicles with cameras to survey the deep, it’s become clear that bottom trawling has devastated deep-water reefs.

It’s not the only important cause of damage, though. Storms regularly decimate shallow-water reefs. Litter is a serious problem, particularly for the animals that live on reefs, which are killed or injured when they are entangled by it or eat it. Population explosions of aggressive species, like the Crown-of-Thorns starfish, can cripple reefs, rapidly killing almost all the coral. They are thought to become a problem more often when fishing has removed their natural predators.

Damaged reefs take a long time to get back to health. Tropical reefs are relatively fast, recovering in a few years or decades. But cold water reefs can take several decades or even centuries to recover. For our purposes, that means that damage is essentially permanent. Some reefs won’t recover at all.

The Nessie 4 robot, which could become a Coralbot.
A Nessie 4 robot, which could become a Coralbot.

The project

All this explains why, when Lea-Anne Henry had a one-minute speed date to talk with David Corne, she asked: “Can we use robots to fix coral reefs?” Both are scientists at Heriot-Watt University in Edinburgh, Scotland, and Corne and Henry were at a Crucible, an event that brings together researchers from diverse disciplines to foster collaboration and discuss how research affects society. Henry is an expert in deep water corals; Corne in artificial intelligence and machine learning. The pair quickly realised that their specialisms were complementary and the idea for Coralbots was born.

Many people are looking for ways to help battered reefs. Some volunteer-led projects involve scuba divers literally gluing corals back together. Coral nurseries are growing corals in safe water and “planting” them on struggling reefs in the wild. These efforts are effective but have serious drawbacks. Repairing or replanting reefs by hand requires a lot of time, and there is little money to go round. People also can’t dive anywhere near deep enough to help cold water reefs.

So Coralbots are to be a small group of underwater robots equipped with artificial intelligence, computer vision and tools to manipulate coral. They will be able to recognise living and dead coral fragments, and they will work cooperatively without human guidance to restore and rebuild reefs. They will behave as a “swarm”, like ants and bees, individually simple but in tandem producing finely-tuned, complex structures. This will make them effective, economic, and able to go where humans can’t. And if they are successful in reef restoration, they could be adapted to solve other problems, like marine litter and aggressive species.

At the end of the Crucible, Corne and Henry pitched their idea to a panel in competition with other interdisciplinary proposals. They won seed funding, and have used it to develop their ideas. The team has since grown. A member of the Crucible competition judging panel, Professor David Lane of Heriot-Watt, liked the idea enough that he joined the team. Two others complete the core of the group: D. Richard Blidberg of Northeastern University in Massachusetts, an underwater robotics engineer, and Neil Robertson of Heriot-Watt, who works on computer vision.

Reef restoration in Belize by Lisa Carne and the Fragments of Hope coral nursery
Reef restoration in Belize by Lisa Carne and the Fragments of Hope coral nursery.

Robots have obvious advantages for restoring reefs: they don’t need to breathe and aren’t crushed by the pressures of the deep sea. But the Coralbots’ gift, in Corne’s words, is that a high intelligence can emerge out of several individuals with limited intelligence. As the group explains on their website: “Nature shows us how groups or swarms of organisms achieve complex things just by following simple rules”.

Corne devises computer programs that allow a “swarm” of machines to create complex structures by following simple rules. He develops them in computer simulations using “evolutionary computing”. That mean making a set of similar programs designed to complete a certain simulated task. Every program is tested, and those that perform best are kept and modified to produce another set of similar programs. These also have a go at the task. The cycle is repeated many times, with the set of programs becoming more and more similar each time. Corne ends up with programs that are very effective at their task.

For the Coralbots, the programs will contain rules for things like recognising objects, dividing up tasks and choosing where to stick coral fragments onto broken skeletons. Working in groups, the robots will each specialise in one task. No robot will be in charge, but by carefully tuning how they decide things like the number, spacing and position of planted coral fragments, the rebuilt reef will have an intelligently designed, though partly random, architecture.

Doing a task with a swarm of robots is cheaper, easier and more efficient than doing it with one very intelligent robot, or a person. Also having multiple, similar robots makes breakdowns much less serious, as no one machine is essential. “It’s great engineering sense”, says Corne. Henry says that an average reef could be restored in several months using Coralbots with swarm intelligence.

Swarm intelligence is inspired by communal animals, like the eusocial insects: bees, ants, wasps and termites. These creatures build intricate nests with huge numbers of individuals and organise themselves efficiently to find food, raise young and defend their colony. They can create very large, complex structures with reliable and detailed features even though the individuals can’t plan or see the overall effect. According to Corne, these creatures don’t know why they do exactly what they do. No mastermind guides the activity of the masses. Their achievements are the result of behaviour finely tuned through evolution so that, when many individuals work in parallel, they produce just the right overall effect. “In the end, nature follows simple rules, and what it can produce is intriguing,” says Henry.

Weaver ants building a nest in Thailand.
Weaver ants building a nest in Thailand, via Sean Hoyland.

Reefs have certain overall features that are important to their health. For instance, the spacing and position of colonies, the mix of species, and the network of nooks and crannies formed by their skeletons are all significant. Henry compares the intricate structure of a reef to a cage. This is important for coral health, but also to the many other reef residents, whose lives play out in this natural labyrinth. Prey hide from predators, who wait to ambush prey. Animals lay their eggs where snatchers won’t venture after them.

Coralbots will restore reefs so that they have all these features and can quickly return to health and host a variety of species. They will also keep the practical things in mind: that corals need to face the current to feed, that some need to face the sun, and that when rebuilding a reef dead coral should go on the inside to prop up live coral on the outside.

The team aims for their first mission to be in Belize, restoring reefs that have been damaged by hurricanes and grounding ships. Belize is a small, English-speaking nation in Central America that benefits greatly from its wealth of reefs – it’s one of the very few countries that have actually banned bottom trawling in their waters. There, the Coralbots team will work with collaborator Lisa Carne, who runs a coral nursery.

Before they are ready for that, they need to fine-tune the artificial intelligence for restoring reefs, which includes training it to recognise and manipulate coral. They also need to assemble a robot that can manipulate coral in all the necessary ways, including attaching fragments to reefs. They say this will be hard, and expensive, particularly when it comes to the machines. But they are also absolutely sure they can do it.

The campaign

On their Indiegogo campaign page, the Coralbots team has made tentative promises about what they can achieve with different sums of money. With $80,000, they think they can develop the software, build a prototype and demo the first Coralbot in a public aquarium. They hope $140,000 will get them to Belize with two prototypes for the first mission. But their current target is $30,000, which they think will allow them to get the artificial swarm intelligence working.

Coralbots already attracted over $30,000 of donations on Kickstarter. Indiegogo and Kickstarter are both “crowdfunding” sites, on which people can pitch their ideas to the world at large, and anyone can donate in aid of making them happen. Campaigns on Kickstarter have to set a target and reach it to get any money. Coralbots aimed too high, setting a $107,000 goal and reaching only an admirable $33,871, so the 853 donors kept their cash. Money raised on Indiegogo is kept whether it hits the target or not. 211 people have now donated on Indiegogo, and the number is increasing steadily. (I’m number 211.)

I asked Corne why Coralbots was not relying on public or philanthropic funding, like most science and conservation projects. He said that they are pursuing other sources, but getting grants for interdisciplinary projects is difficult. The UK funding councils, for instance, are all set up along disciplinary lines. Many panels that review grant applications don’t have the relevant expertise to handle interdisciplinary concepts. And most research using underwater robotics is supported by oil companies, but Coralbots probably won’t appeal to them because it isn’t developing robots in a way that those companies can use.

“We absolutely know it’s possible,” says Corne. “We’re basically there. We have the expertise, hardware and software. We just need the cash to do it,” says Henry. They can’t guarantee progress over specific timescales, because they may run into mishaps or price rises on the robotics hardware. “But we are sure it can be done, even if it’s 10-20 years down the line, even if it’s not actually us doing it anymore,” asserts Corne. “Though we think we can do it much sooner than that”.

I for one wish them the best of luck.


Visit the Coralbots website, Facebook, Twitter and Indiegogo.