Dr Nicole Robinson and Dr Jürgen Leitner founded LYRO Robotics to solve an interesting agricultural robotics challenge – delicately yet firmly grasping and packing avocados.
This story featured in Issue 15 of Forbes Australia. Tap here to secure your copy.
When Dr Nicole Robinson was growing up in Melbourne in the early 2000s, she was fascinated by technology, robotics, and humanoids. “I was always that kid tinkering with technology and wanting to pull it apart and see what was connected to what and how it worked,” says Robinson, sitting in a café in an industrial area of Melbourne’s South-Eastern suburbs.
“When I was going through middle school, it wasn’t even clear that you could go straight into robotics. There wasn’t a very direct pathway. When I told my family I wanted a career in that field, it was no surprise to them. If I had told them I wasn’t going into technology, that would have been a bigger surprise,” she says
The Brisbane-based CEO and co-founder of LYRO Robotics is in her hometown, Melbourne, to oversee a research project. Her co-founder, Austrian-born roboticist Dr Jurgen ‘Juxi’ Leitner, splits his time between the Sunshine State, where LYRO is headquartered, and Victoria, where he is a tenured professor in robotics, space, and entrepreneurship at Monash University.
“We wanted to do something meaningful and impactful – robots for good, something Australia does well as our first use case. Well, we grow a phenomenal amount of food in Australia.”
Dr Nicole Robinson
The co-founders met in 2017 while Leitner led Queensland’s Australian Centre of Excellence on Robotic Vision. Robinson was a doctoral student studying human-robot interaction at the Queensland University of Technology and won the Amazon Robotics Competition in Japan.
The two chatted after Robinson presented her robotics solution on-stage and discussed the viability of scaling the idea into a company. In 2019, LYRO was born. The industry in which the robotics solution would be applied was where the most pressing need was.
“We wanted to do something meaningful and impactful – robots for good, something Australia does well as our first use case. Well, we grow a phenomenal amount of food in Australia. We have wide fields, we have lots [of] specialty crop opportunities, and we export so much of the food that we grow. If we had more labour, we could export even more, right?” says Robinson.
The ultimate Tetris puzzle
The ultimate test case for the efficacy of robots handling food items is an avocado, according to Robinson and Leitner. Being relatively soft on the inside, an ‘avo’ needs to be handled delicately to avoid damaging or bruising its internal structure.
Avocados come in different sizes and at variable stages of ripeness, so key to creating a robot that can effectively pack avocadoes is the human concept of hand-eye coordination – assessing the item to be picked up and transmitting that information to the hand or claw, so that it will neither drop it nor exert more force than necessary.
No two pieces of fruit and vegetable are the same. High-variation products are gentle and need to be handled with precision and care. They need to be pattern packed to fit multiple pieces of fruit together – it’s the ultimate Tetris problem.
The double-edged sword of Covid
Early in 2020, the world started shutting down, and supply chain disruption exacerbated the need for robots instead of humans. “The understanding of what we were doing was very clear, which is we can provide machinery to help support when there are difficulties with having staff on site due to health and safety concerns.”
LYRO had created a prototype, but it was far from the end product that exists today. “It was 3D printed; lots of pieces of duct tape and sticky tape and then zip-tied together just to show the concept and the farmer – this is what the system does,” says Leitner.
He laughs about the pre-COVID-19 prototype and the challenges the founders overcame to create a functional commercial product.“Transporting something in a car in 30-degree heat in Queensland, the 3D printed item would warp, so we had to have many spare parts with us.”
By the end of 2022, there had been a shift in a potential client’s appreciation of the problem LYRO was solving and the stability of the robot the company produced.
“Clients now understood that any disruptions to the labour force can be detrimental to supply chain certainty and efficiency, so I think COVID-19 accelerated conversations with key partners in that sense,” Leitner says.
On the hardware front, LYRO had made it past the duct tape stage, out of the laboratory and onto the farm. “It was very exciting to get our first commercial robot out into a packing house, have it run by itself and to see farmers getting excited and calling in other farmers to say ‘come over, I have a robot in my shed’,” he says.
The tech behind the robot
A part of what has enabled LYRO to attract clients like Costa Group – Australia’s largest horticultural company – are the rapid advancements in technology that have taken place in the last two decades. “You couldn’t build robots like this five years ago,” says Leitner.
As the leading grower, packer and marketer of fresh fruit & vegetables, Costa specialises in citrus, berries, mushrooms, glasshouse tomatoes and avocados. “The robot needs to have the smarts and the brains to handle the variability of fruit, and at the same time to provide the flexibility to hold on to the fruit while moving fast.
“There are three or four cameras on the robot at every point. Over the last 20 years, cameras have become a lot cheaper. We can now build our robot for a tenth of the price it would have cost,” says Robinson.
Additionally, advancements in AI have enabled the robot to learn from vast amounts of data taken from those cameras. A paper on deep neural networks, published in 2012 by the University of Toronto’s Ilya Sutskever and Geoffrey E. Hinton, changed the game for robotics.
“It was early days, and we weren’t sure how deep neural networks would translate into robotics,” says Leitner, who was working as an AI and robotics researcher in Switzerland at the time.
Thirteen years later, the advancement in deep neural networks has rapidly advanced what LYRO robots are capable of.
“Rather than just a bunch of PhD students taking pictures of fruit at their desks – which people did in the 80s and 90s. Now, our robot takes images through the multiple cameras in the robotic systems. We can collect a lot more data and have data to train these models with more complex data sets,” says Leitner.
Scaling internationally
Given the significant size and weight of a LYRO robot, the founders say that part of the company’s growth strategy is to license its software.
“We probably have one of the world’s largest data sets of fresh produce. A camera on a LYRO robot takes 100 frames per second, and it sits there for eight hours a day – you collect a lot of images of avocados,” says Leitner.
That data is a part of the intellectual property ‘treasure trove’ that LYRO sits on. “Images and data are a key differentiator – we now have about two and a half million items packed – that’s obviously a large data set of packed items,” he says.
There are currently a dozen robots with the LYRO stack running, and Robinson says by the end of 2024, there may be hundreds using LYRO-licensed software. In the future, the 35-year-old CEO plans to scale internationally.
“The bold vision is to have robotics software developed in Australia, operating worldwide so we can scale it up, ship it out, and have it plug and play to other companies and robotics providers integrators.”
From avocadoes to the moon
Within 10 years, Robinson wants LYRO to be a household name.
“We want to be able to service and support a variety of clients in different areas. The technology is broadly applicable, whether it’s manufacturing, agriculture, recycling or even mining.
“It could be any interesting task that automation can support – this is bigger than just fruits and vegetables,” she says.
