Reef robocop
Coral-killing starfish are in the sights of a new robot army that’s being trialled on the Great Barrier Reef.
DOZENS OF UNDERSEA robots could soon be combing the northern Queensland waters, taking on a major threat to the Great Barrier Reef – coral-eating crown-of-thorns starfish (COTS) plagues.
Crown-of-thorns starf ish ( Acanthaster planci) outbreaks are a leading cause of coral loss on the Great Barrier Reef (GBR), and in 2015 Matthew Dunbabin and Feras Dayoub, roboticists from the Queensland University of Technology, tested a promising solution – an autonomous diving machine called a COTSbot (see opposite). These can locate and identify the rogue starf ish under water and kill them with toxic, but environmentally friendly, injections of bile salts.
Now trials are underway of a vastly improved version of the COTSbot called RangerBot, funded largely by a $750,000 grant from Google’s Impact Challenge program. And the early results look so good that plans are underway to substantially scale up the project.
The latest robot cruises at about 1.2m above the seabed, searching for the starf ish by processing up to 10 photos a second. “We’ve got hundreds of thousands of images of starf ish, and we’ve trained RangerBot to recognise only crown-ofthorns,” Matthew says. “Our detection system is about 99.4 per cent accurate. We have ver y strict thresholds. If the machine has any doubt, it won’t inject. Later we review the photos it doubted to further improve our system.”
RangerBot swims lower than the prototype COTSbot, taking it closer to its starf ish targets and improving the accuracy of its robot vision or ‘image-based automated inspection technology’.
Rate of processing was a key challenge. “It requires serious computation but we’re limited by the power we can carry on board – even a laptop’s too big,” Matthew says. “We’ve had to optimise software and hardware to run in real time.” Having identif ied a crown-ofthorns starf ish, the robot hovers about 90cm above it and injects the solution, which kills the starf ish within 24 hours. It injects for about a second then quickly retracts so it doesn’t get caught on coral.
The original COTSbot was powered by a small, soft drink bottle–sized air cylinder that lasted for about 100 injections. “With RangerBot, the hydraulics use an electric motor, with sea water driving the inf lation,” Matthew says, “so you don’t have to ref ill gas bottles.”
An even bigger advance was replacing COTSbot’s acoustic navigation sensors with a purely visual system This makes RangerBot seven times cheaper to build, and much smaller. RangerBot’s batteries are different, too. Matthew says, “With COTSbot we had to recharge batteries in situ, but the new RangerBot has removable battery pods. On a boat, you can take the battery out – without turning it off – and put another one in, which meets our goal of an eight-hour operational day.”
Another innovation responds to the crown-of-thorns’ lifestyle. “One of the big problems is there are not enough crown-of-thorns starf ish visible during the day,” Matthew says. “A lot of the time they’re hiding underneath corals where we can’t get them.” RangerBot’s inbuilt lighting means it can operate at night, when they come out to graze in the open.
Built to dive to 100m, RangerBot can travel about 14km in an eight-hour day (with a battery change halfway). Its new single-shot injection was a prior invention, made in 2014 by James Cook University researchers, to replace the 20 or so injections previously needed to kill a single starf ish.
That development a lone meant COTS-culling human divers, who are still the main means of control for the starf ish when they’re in plague proportions, became 250 per cent more efficient.
RangerBot’s potential to support human divers targeting crown-of-thorns is obvious, but its capabilities don’t stop at starf ish-zapping.
Matthew says coral-health monitoring, water sampling and f ish surveys are tasks RangerBot could also take on.