Usually animal preservation is a passive effort, creating protected zones or taking other measures to protect plants and animals from humans. But scientists and students at the Polytechnic Institute in Massachusetts want to help protect coral reefs from an invasive species in a more aggressive fashion: They’re building a robot designed to autonomously hunt for and harvest lionfish threatening coral reefs.
Lionfish have threatened coral reefs off American and Caribbean coasts for years. The National Oceanic and Atmospheric Administration describes them as “flexible predators potentially capable of reducing the abundance of a wide variety of native reef-associated fishes.”
Native to the Indo-Pacific and Middle East, lionfish have distinctive features which make them prized aquarium pets. After likely being dumped into the Atlantic by owners who no longer value them, their eggs have the Gulf Stream southbound to allowing them to become vicious predators amidst shrimp, small crabs, Nassau grouper and yellowtail snapper, just to name a few species who have come under attack.
Their style of attack is unique to their newfound waters. Lionfish have hollow bones in their dorsal and pectoral fins which they inject with toxins. Whipping their fins towards a target, NOAA estimates that they’ve created a diet of around 40 species.
That’s where the robots come in.
“The goal is to be able to toss the robot over the side of a boat and have it go down to the reef, plot out a course, and begin its search,” says Craig Putnam, a senior instructor in computer science at WPI, in a press statement. “It needs to set up a search pattern and fly along the reef, and not run into it, while looking for the lionfish. The idea is that the robots could be part of the environmental solution.”
A fish-hunting robot has many complex requirements. When designing a robot to hunt one specific fish, getting the identification process right is crucial. The robotic hunter needs to be able to distinguish lionfish from other fish within the reef ecosystem to choose the right target. Ideally, coral reefs are busy and flourishing environments full of sight and sound. The robot needs to cut down a tremendous amount of noise to find its target.
That training comes from machine learning. The students at WPI showed their robot thousands of pictures of lionfish of different colors, taken from different angles and with varying lighting conditions, training it to recognize a lionfish with greater than 95 percent accuracy. The robot also got pictures of human divers in order to train it in what not to absolutely avoid shooting.
The robot will use a revolving carousel that WPI compares to the cylinder of a revolver, it will hold eight detachable spear tips. A motorized mechanism will thrust the spear’s tip into the fish body. When this mechanized shaft retracts, it will leave the spear tip within the fish’s body and the carousel will move on.
When operating with mechanized attack systems like this, buoyancy suffers. The WPI students working on the robot as a Major Qualifying Project (MQP) decided to compensate for this through a watertight, air-filled chamber that enlarges slightly after each spearing.