Meet the Aquatic Drone Saving the Great Barrier Reef with Machine Learning and Computer Vision

The Queensland University of Technology has announced that its robotic hunter-killer aquatic drone will now double as a seaborne midwife to save the Great Barrier Reef.

Underwater drones are finding more applications for ocean conversation. One of the most famous underwater drones is the RangerBot from the Queensland University of Technology (QUT) in Australia. This is an underwater drone that, according to its developer, Professor Matthew Dunbabin, “is the world's first underwater robotic system designed specifically for coral reef environments, using only robot-vision for real-time navigation, obstacle avoidance and complex science missions.”

The device, which won the 2016 Google Impact Challenge People’s Choice prize, employs multiple thrusters for locomotion, as well as computer vision and machine learning for obstacle avoidance and real-time navigation. A surface-based human can operate the 15kg, 75cm long RangerBot via a simple tablet-based controller.

The RangerBot. Image used courtesy of Great Barrier Reef Foundation

In a collaboration funded by the Great Barrier Reef Foundation, QUT’s Professor Dunbabin and Professor Peter Harrison of Southern Cross University (SCU), RangerBot’s capabilities will be utilized to restore damaged areas of Australia’s Great Barrier Reef.

RangerBot's Computer Vision

RangerBot's vision system comes largely from the research of QUT's Dr. Feras Dayoub. His 2016 paper in IEEE's Proceedings of the International Conference on Robotics and Automation was titled "Place categorization and semantic mapping on a mobile robot"—pretty clearly of enormous importance to RangerBot's ability to interpret visual data.

His most recent publication is "A rapidly deployable classification system using visual data for the application of precision weed management", a paper that has obvious possible future applications for classifying species of coral.

In RangerBot's and COTSbot's case, computer vision has proved key for identifying crown-of-thorns starfish:

COTSbot's computer vision at work. Image used courtesy of QUT

RangerBot Responds to the Starfish Threat

The RangerBot derives from the earlier COTSbot, also developed at QUT, which was aimed at eliminating the lethal threat to coral reefs posed by crown-of-thorns starfish infestation (hence "COTS" for "crown-of-thorns starfish"). Previously, all efforts at control were through the efforts of human divers.

The comparison below from the Great Barrier Reef Foundation tells the story:

Human divers vs. RangerBot. Image source (modified): Great Barrier Reef Foundation

Divers are limited to three hours a day of underwater work, and, of course, the costs of equipping, protecting and deploying divers are quite high. As we can see from the very first line of the comparison, it is estimated that RangerBot can kill starfish over 28 km of coral reef in a single day, as opposed to the 1 km length of the coral reef that can be covered by a human diver.

Additionally, RangerBot can collect vast amounts of data pertaining to the condition of the coral reef.

A look at the electronics inside RangerBot’s predecessor, COTSBot. Screenshot used courtesy of the Australian Museum

But these tasks have not been enough to keep RangerBot in one role. Its new job is more oriented towards nurturing than assassinating starfish. 

Rebirth of a Coral Reef: RangerBot to LarvaBot

Regardless of what killed them, coral reefs can, essentially, be “replanted”. While RangerBot's so far been focused on preventative action and data-gathering, QUT recently announced that it will now put on a new hat where it will promote new reef growth.

Under Professor Harrison’s direction, hundreds of millions of coral spawn are collected and then developed into baby corals or larvae. This allow's Professor Dunbabin’s versatile RangerBot to become LarvaBot, a drone re-tasked with spreading Harrison’s larvae into depleted regions of the coral. And, because the health of the coral reef has already been assayed by previous sojourns of RangerBot, the areas that needs “rebirth” are already well known.

Possible Future Applications: Additional Threats to Coral Reefs

According to the Smithsonian Institute, coral reefs cover less than two percent of the ocean’s bottom but are a crucial factor in the life-cycle about one-quarter of all oceans species—including, presumably, the fish that humans eat. Yet, coral reefs are dying, posing a grave threat food supplies and more.

As described by the Great Barrier Reef Foundation, the greatest threats to reefs are:

• Climate change leading to ocean acidification
• Coastal development affecting habitat
• Illegal fishing and poaching

RangerBot has already tackled hunting crown-of-thorns starfish and now spreading new coral larvae to counteract reef loss. In the future, perhaps QUT will address these other issues, possibly with sensors to measure acidification or new algorithms to identify and track species of coral.

Whatever job RangerBot gets into next, it will certainly be worth watching.