Machines learning: inside the University's centre where robots learn

The lab at the Centre for Automation and Robotic Engineering Science (CARES) is a little like a preschool, except it’s robots, not children, who are there to learn.

Undergrad to postdoctoral researchers working in the centre spend their days teaching robots not only what to learn – to inspect the hull of a boat, for example – but how. Just like a parent raising a baby, many of the lab’s researchers engage in deep learning projects that involve ‘feeding’ robots past experiences. They then reinforce desired behavioural responses with ‘rewards’, so the machines will one day be able to efficiently carry out complex tasks on their own.

The robots are trained to do tricky work. On a tour of the lab with robot learning team lead Dr Henry Williams, he points out machines being designed to prune trees, scan for mines or replicate the movement of human hands. Then there are the machines that are just plain cool: members of the team are working on a massive dancing jellyfish that spans ten metres, and they once created a scale replica of NASA’s Opportunity Mars rover (now displayed at Te Whatu Stardome).

“We have a lot of freedom in how we approach things,” says Henry. “We don’t have to worry too much about the risk of trying an idea; we throw something at the wall, see what happens, learn something from it and move on.”

Henry came to the University of Auckland around a decade ago, after completing his PhD in robotics at Victoria University of Wellington. CARES collaborates extensively on projects with other institutions and industry, and Henry initially began working on an agricultural robotics project.

He’s since worked on many others, including prototypes for a system that ultimately led to the development by industry collaborator Robotics Plus of Prospr, an autonomous vehicle platform used for a range of horticultural tasks and now sold in the US.

“Over the past ten years, I’ve spent a literal year of my life in an orchard or a vineyard with these platforms doing the testing and evaluations, either when it’s freezing cold or super-hot,” he says. “But those times are the most fun, because testing something in the lab is one thing, but seeing it happen in the real world is another. I still remember the first time I saw a robot I’d been working on reach up and deliberately grab a kiwifruit; after all the fixes and corrections and learnings, you finally see it at a point where it’s working.”

The practical nature of the research means many of those who have come through CARES now run their own companies or are working in cutting-edge startups.

Henry is embedded in the robotics community as chair of the New Zealand Robotics, Automation and Sensing network of researchers, engineers, technologists and students, and secretary of the Australian Robotics and Automation Association.

Robotics has always been a popular area of study – “because everybody loves robots” – but more students want to train in the area now than ever, he says. Advances in technologies like 3D printing and machine learning are also accelerating the possibilities of what robots can achieve.

“You never stop learning, which is a lot of fun. When you’ve solved one problem, there’s still another one to solve,” he says.

Caitlin Sykes