On Tuesday at UNSW, 22 teams from universities around Australia and New Zealand gathered to pit purpose-built robots against each other in a battle of intelligent design. A crowd gathered in the auditorium where two intricate looking obstacle courses had been set up for the robots to navigate, cheering whenever one team’s creation pulled off a perfect run. This is the National Instruments Autonomous Robotics Competition, and the entrants are the roboticists of Australia’s future.
The task seems simple, at first glance. The robot starts in a taped off square, drives to another square (the ‘loading bay’, in keeping with this year’s theme of ‘Transport and Rollout’), accepts a load from the person waiting there, drives off through a series of obstacles to a drop-off bay, and then navigates to its final square where the team are waiting to congratulate it. That is, if it makes it all the way. Points are allocated based on the robot’s performance, with consideration of factors like speed and which drop off bay they choose to navigate to — the bots are given the choice of an easy entry with no obstacles, a bay with a few speed humps out the front and the most difficult, and most rewarding, choice included a challengingly narrow gap to navigate.
I know I would have very little luck guiding a remote controlled machine through such a course — but that isn’t the challenge here. The teams don’t have any input on their robots at all once the race has started — they are built and programmed to be entirely autonomous.
It’s not a simple task for teams to make it to the live finals, either. 27 teams applied for the competition in March and only 23 got to the live competition earlier this week, with one team having to drop out at the last minute. In the six months since the building began, teams have had to pass five different milestones in order to qualify for the competition, ranging from simple software competency training, all the way up to being able to program a robot to navigate an obstacle course similar to the one they would be traversing on the day.
To make things just a little more difficult, the competitors weren’t actually shown that course until the day before the finals, meaning the teams had to program their robots to adapt to that particular course within the space of just 24 hours. “We only had one metre square to practice on,” says Manukau Institute of Technology entrant Jayme Salmon. “The material for the track is different, so there’s nothing really similar to here.” How did that turn out for their bot? “I got stuck in the loading bay,” Jayme laughs, but he doesn’t sound too disheartened. “But it’s been a good experience.” All the qualifying entrants get to keep their NI supplied kits, after all, and making it this far is a victory in itself.
Watching the qualifiers round up before the lunchbreak, I witnessed a number of successes and a couple of disappointments — with the crowd reacting as enthusiastically as if they were watching sports. One of the bots got stuck on its way to the first square, never quite managing to collect its assigned load. A bright, neon green robot aced its first couple of heats, but in its last round in the arena it failed to move once the load was given to it. “They hit the off switch with the block,” the announcer tells the crowd once the round is over, to a mixture of laughter and pained groaning.
No two teams have taken the same approach to this challenge. One robot is thick and perfectly circular, another deposits its load a little like a dump truck, while yet others have crane-like extensions to lift the block from above. A few of them look like something I might have built out of LEGO as a child — but they all have some serious technology behind them, some parts supplied (along with software) by National Instruments, and others sourced by the teams themselves.
The team from the University of South Australia took out first place in the end, with Swinburne University and Victoria University of Wellington coming behind them in 2nd and 3rd respectively. So what does it take to build a winning robot? “We need to get a laser,” Jayme from Manukau has already decided, watching the more successful teams. He’s referring of course to LIDAR technology, which most of the teams are already using to help their robots know exactly where they are on the course. This is the same kind of technology used for autonomous cars, and is even similar to how household robots like the Roomba navigates its way around the house.
And the kind of technologies these university students have been asked to develop today may well find their way into our lives tomorrow. “This kind of technology has enormous practical applications for indoor robots and household robots,” says UNSW team member Fred Westling, “because they always know where they are, they know where they want to go thanks to their path planning. It’s also applicable for outdoor robotics if you use difference sensors.” The competition bots were built essentially as an extra-curricular activity by these busy students — if this is what they can do with two hours a week, just imagine what they will do once they’re working on these sorts of projects full time.