Exclusive: How robots and AI are transforming Australian farming

By Nurfilzah Rohaidi

Robots enhanced with vision can help reduce manpower in agriculture, says Professor Peter Corke.

It’s 2017, and farming has come a long way from its subsistence roots. Drones, satellites, precision agriculture—and now fruit-picking robots?

That is but one application of robots combined with artificial intelligence, Professor Peter Corke, Director of the Australian Centre for Robotic Vision at Queensland University of Technology, tells GovInsider on the sidelines of the Commonwealth Science Conference held in Singapore last week.

The robots may be slower than humans at fruit-picking, but they can work 24 hours a day, Professor Corke points out. “If you just look at the cost of the robot doing the job or humans doing the job, we are now at the place where it gets interesting,” he shares.

Professor Corke shares with GovInsider how robots can alleviate lack of manpower in several industries, and the future of this technology when combined with artificial intelligence.

Robots to replace manpower


Australia faces the “very acute” problem of lack of manpower in farming and agriculture—which is where weeding and fruit-picking robots can come in, Professor Corke says. “Human fruit-pickers are hard to get,” he notes. Besides the fact that less and less Australians want to go into farming, Australia relies heavily on tourists who go on working holidays at farms—but “even that will probably dwindle”.

Professor Corke’s work currently focuses on two uses of robotic vision in agriculture: robots that can see and recognise weeds, and robots that can see and recognise fruit on vines. “We can use vision to recognise the weed based on the size, shape and colour. And if we see it there we can spray some herbicide or we can actually dig it out on the ground,” he explains.

Agricultural robots can reduce the costs of weed operations by 90%, he believes. These currently cost approximately US$1.14 billion a year, with a further US$1.9 billion a year in lost agricultural production, according to the Department of the Environment and Energy.

Robots enhanced with sight can also recognise and pick fruit, even if it is covered in leaves, Professor Corke says. This technology is on the bounds of “being economically viable”—meaning it is just as effective as humans are at doing the job, Professor Corke says.

The next frontier in robotics


When it comes to robotics, “vision is very definitely the next frontier,” Professor Corke says. This is important if the world wants to bring robots into industries and environments we don’t normally see them in, he explains.
 
“Vision is very definitely the next frontier.”
“Robots are blind; they cannot see and understand their world. If I’m a robot and I’m picking apples on a farm, I’ve got GPS to know where I am, but the GPS doesn’t tell me where the apples are.”

Robots work best in structured environments where everything is in a set place, Professor Corke explains, like in factories on production lines. “The rest of the world is not like that; that’s why we don’t see robots in houses, around hospitals, in farms, in forests,” he remarks.

This is where artificial intelligence (AI) comes in. Robotic vision is made possible by neural networks, a type of AI that mimics the human brain. Advances in AI and machine perception can go a long way towards robots that can effectively do more useful work in our world, Professor Corke says. “Ideally I’d like to create something that had the visual skills and hand-eye coordination that humans have,” he says.

That could be rather challenging, which is why robotics in agriculture is an ideal place to start—requiring tasks that are “quite a bit simpler but have got some economic value”.

In the shop, bank and hospital


Once robotic vision is advanced enough, there are several industries that can benefit, according to Professor Corke: mining, logistics, and healthcare, to name a few. He is also working on “social robots” that can greet you in shops, banks and government offices. “Robots that are competent, intelligent and can see, have got lots of advantages in many other sectors beside agriculture,” he says.
 
“Robots that are competent, intelligent and can see, have got lots of advantages in many other sectors beside agriculture."
Robotic vision could have the potential to be very useful in healthcare, as costs balloon and governments grapple with manpower crunches. “If you look at a hospital, it looks very chaotic and almost primitive compared to a manufacturing plant where it’s all very orderly and very fast,” Professor Corke notes. “Hospitals have got a large number of manual processes and they cost an enormous amount for society.”

As one solution, another area that Professor Corke is moving into is robotic surgery, he says. “We’re doing work now for orthopaedic operations, in particular, knee arthroscopy. The system that we envisage is a cooperation between the human surgeon and a couple of robots that are helping the surgeon do the procedures,” he explains.

Professor Corke predicts that as the technology progresses, hospitals can begin to automate processes outside of the surgical theatre as well. Soon, we could be seeing “robots to help patients get in and out of bed, help them go to the bathroom, help them have a shower, cook and deliver meals”. “That’s a huge opportunity for managing a hospital,” he adds.

With Professor Corke’s team and other roboticists plugging away in robotic vision research, one day we’ll be seeing robot fruit-pickers, weeders, miners, and even surgeons. What a sight that would be.

The Commonwealth Science Conference held on 13-16 June 2017 was jointly organised by The Royal Society and the National Research Foundation, Prime Minister's Office, Singapore.