Reef Robot

The crown-of-thorns starfish (COTS) has become the scourge of the Great Barrier Reef. Infestations of the coral-eating predators are responsible for an estimated 40 percent of the decline in the reef’s coral cover, and they’ve proven to be difficult to eradicate, due to their ability to regenerate, even when cut in half.

Recently, Australian researchers launched a trial of a new robot designed to hunt down and kill COTS autonomously. Known as COTSbot, the underwater droid uses advanced image recognition and machine learning to identify its target and deliver a lethal (only to the starfish) dose of bile salts via injection. The robot is designed to operate for 8-hour runs and can deliver up to 200 shots in a single mission. An acoustic sensor enables it to navigate within a meter of the sea floor, and a vision-guided articulating arm injects the poison as the bot hovers above. The plan is to deploy COTSbots throughout the reef as a type of “first responder” to eliminate the bulk of the menacing creatures, followed by divers a few days later.

Given that underwater visibility is highly variable with depth and changes in water clarity, COTSbot had to “learn” how to think for itself by capturing images of suspected COTS, which were verified by a human. The feedback was then incorporated into the robot’s memory to create a robust recognition system that boasts 99 percent accuracy.