Robotic fish supported by machine learning are being developed to help rid WA waterways of invasive species, a project which illustrates the impact of the Forrest Fellowship program.
Robotic fish supported by machine learning are being developed to help rid WA waterways of invasive species, a project which illustrates the impact of the Forrest Fellowship program.
Nowhere in the world is a better place to research biodiversity than Australia, according to University of Western Australia academic Giovanni Polverino.
Dr Polverino won a Forrest Fellowship in 2017 and is studying the use of robotics to support native fish species to survive against invaders.
The fellowships were funded by Andrew Forrest’s Minderoo Foundation.
“The biodiversity of Australia is so unique, preserving it is important,” Dr Polverino said.
“The issue of invasive species here is bigger than anywhere else.
“If you lose your fish or your marsupials you won’t find them anywhere else.
“Australia has always been this happy island of biodiversity, everybody that’s keen on working on it wants to move here.”
He said he was keen to come to the University of Western Australia to learn from experts in reproduction.
“The fellowship made it happen, without that there was no way,” Dr Polverino said.
Working in collaboration with engineers at New York University, Dr Polverino is using robotic fish to simulate the traditional predators of invasive fish species that have taken over ecosystems in WA.
The robots are trained to act as predators for invasive species but not for local species.
“The process is going quicker than I thought,” he said.
“We’ve published our first study that demonstrated the robot had a fantastic effect.
“We can build a biologically inspired predator that not only can scare them but also from the stress … they lose weight, they become very skinny and they die faster than normally.”
Invasive species were animals that were present in an environment where they do not belong, introduced by humans, Dr Polverino said.
“I work with fish that have been imported from North America all over the world, including Australia,” he said.
“(Importers thought) that these fish that are supposed to (be) eating larvae of mosquitoes, would actually be a good solution to have less mosquitoes around.
“The truth is that these fish eat mosquito larvae but also everything else.”
The result is a major impact on the population of tadpoles of local frogs, other amphibians and native fish.
Preliminary results from a second study, which is yet to be published, are also promising.
They show the robots can discriminate between species using a form of machine learning known as computer vision, a technique where computers are trained to recognise patterns or concepts in images.
The work also shows that stressed invasive fish species struggle to reproduce and the population drops over time.
“We could actually collapse their population,” Dr Polverino said.
The next stage of experiments, planned for next year, will be out of the lab and into the wild, with a small body of clear water to be selected.
Gradually, the collaborators will move to complicated and murky ecosystems such as the Swan River.
Computer vision is a growing field, applied in autonomous vehicles for the resources industry, in security and intelligence applications, and medicine, among many disciplines.
To train the system requires large numbers of accurately labelled images, and a bit of processing power.
“To get them (images) is easy,” Dr Polverino said.
“To process in real time is not.
“The robot needs to figure out who is whom and react in real time.”
It’s much easier to analyse images from a clear aquarium than a murky river, he said.
To help improve the speed of processing the images, the fish transmit to a nearby computer to do all the work.
This is where the NYU team comes into its own, with the engineers providing computational knowhow.
Dr Polverino has worked with them before.
Before coming to Australia, Dr Polverino undertook a PhD in Germany, specialising in behavioural ecology.
“Before my PhD in Germany I spent almost three years at NYU working in this lab, I was an undergraduate but I was the only biologist,” he said.
“They were looking for somebody that could help to transfer the biology into a robotic tool.”
In Perth, there are other researchers through the fellowship program making an impact.
Four post-doctoral researchers will come to Perth on the fellowships next year, including Jessica Buck, a Kamilaroi woman who studied at Oxford University.
Dr Buck will study treatment of brain tumours in children.
