How AI and robotics are revolutionising ocean conservation

Dr Scarlett Raine, QUT Centre for Robotics, QUT School of Electrical Engineering and Robotics

Imagine descending beneath the ocean’s surface. The light dims, the water cools, and a vast world unfurls – alive with alien-like shapes, vivid bursts of colour, and a buzz of marine life going about its day. A turtle glides by, weaving through its coral playground. This place is intricate, ancient, and brimming with biodiversity. This is a coral reef.

These vibrant living structures, built by tiny coral polyps over millennia, are breathtaking not only for their beauty but for the abundance of life they sustain. The Great Barrier Reef supports 25 per cent of all known marine life.  But underlying these complex ecosystems is a real fragility.  Scientists estimate that up to 90 per cent of the world’s coral reefs could disappear by 2050. Already, our Great Barrier Reef and Ningaloo Reef have experienced unprecedented levels of bleaching in recent seasons.

We are likely the last generation to witness reefs like this in our lifetimes.

But we have the chance to still change that trajectory.

By harnessing the power of AI, computer vision and robotics, we can buy time and improve the resilience of marine ecosystems. These technologies can revolutionise our fight to protect and restore the ocean.

Traditionally, monitoring coral reefs is performed manually by divers or snorkellers. One common method, the manta-tow technique, involves marine scientists being towed behind a boat and visually identifying and recording species of interest.  This approach is limited by dive times, range, safety, and the availability of trained experts. In the context of massive reef ecosystems like the Great Barrier Reef – so vast it’s visible from space – we need scalable, high-tech solutions.

That's where underwater robotics comes in. We can now deploy robotic vehicles or towed camera systems to collect hundreds of thousands of underwater images in a single mission.

But this brings a new challenge: how do we efficiently process the terabytes of imagery collected?

To address this bottleneck, we developed computer vision and AI algorithms to automatically analyse underwater images. Models can automatically identify species, track changes in ecosystem health, and detect early signs of bleaching or degradation.

Further, our AI can reduce the time and effort of labelling these images by up to 60 fold – this is a reduction in annotation time from one hour per image to just one minute.

To do this, we leverage large foundation models, similar to ChatGPT, to provide general knowledge of concepts – like the difference between sand and seagrass, while leaving the more challenging concepts, such as the difference between two species of seagrass, up to the marine scientists.  This reduction in annotation effort frees up the time of marine experts to perform other critical and specialised tasks.

The power of this technology goes far beyond the data. The insights generated by these AI systems can be used to tell how healthy ecosystems are, create maps of where different species occur, or calculate how much reef area has suffered from coral bleaching.

It can be used by custodians to estimate how much blue carbon has been sequestered by a seagrass meadow, or by governments to inform and improve fisheries management or establish marine park areas.

But monitoring and policy action might not be enough to save our reef ecosystems.  Through the Reef Restoration and Adaptation Program, QUT and the Australian Institute of Marine Science are working on innovative reef restoration efforts.

This program aims to re-seed degraded sections of the reef with devices equipped with temperature-resilient baby corals. Divers have already planted thousands of these tiny corals, but given the scale of the Great Barrier Reef, we need to go bigger.

Through the development of the Reef Guidance System, we’re working toward the goal of deploying three million coral devices in the coming years.

This system comprises a camera and computer mounted on a boat, which runs AI models in real-time. These models analyse the seafloor and visually determine where to dispense the baby corals to give them the best conditions and chance of survival. It simultaneously records GPS data so we can revisit these sites and track the long-term growth of the corals.

This represents a step change in the scale of the reef restoration effort. What was once a human-scaled effort can now become a fleet-scale initiative, guided by AI and powered by data.

AI is rapidly becoming one of the ocean’s most powerful allies. These tools amplify the efforts and expertise of marine scientists and conservationists: AI and robotics allow us to act faster and more efficiently, cover more ground, and base our strategies in large-scale, long-term data.

As the theme for World Oceans Day reminds us, we need to catalyse action for our ocean and climate. That means moving beyond awareness and towards scalable, efficient solutions. It means leveraging AI and robotic technologies for good, and not just to observe the damage as it occurs, but to actively protect what’s left, and to restore and rebuild what’s been lost.