Some days walking into Steven Waslander’s lab feels like you’re entering the set of a science fiction movie. Drones hover and dance in front of a group of students, while wheeled vehicles scoot across the polished concrete floors.
The mechanical and mechatronics engineering professor heads the University’s Waterloo Autonomous Vehicles Laboratory (WAVELab) where, among other things, he and his team envision a day when quadrotor helicopters can function without human operators.
Today, you’ll find autonomous quadrotors used for precision farming and remote surveillance. As they become more advanced, Waslander predicts they’ll play a role in everything from inspecting offshore oils rigs to assisting wilderness search and rescue efforts.
Before that happens, however, several stumbling blocks need to be overcome. As Waslander explains, all the problems facing autonomous ground vehicles are multiplied in the air, where you can move in any direction.
First there’s the perception problem. GPS technology gives you a resolution of about three metres. That’s fine when you’re flying 20 metres above an open cornfield. However, to check for wear and tear on oil rigs or cracks on the underside of bridges, quadrotors need to get in close — and that demands precision.
To solve the problem, the WAVELab team built a system that uses several cameras with 185-degree fields of vision to map out anchor points in the environment. Triangulating from these anchors makes it possible to track vehicle movement down to the centimetre.
The next phase of their research focuses on processing image data to create a full 3D reconstruction of the environment. “Then we can completely eliminate the need for a human operator,” Waslander says.
After that comes the challenge of detecting moving objects. “You certainly don’t want to land where the kids are playing,” he explains.
Waslander’s team is also looking at ways to extend quadrotor flight time. The more power you draw to perform complex navigational algorithms, the less power you have to actually propel the quadrotor. And with an average flight time of just 15 minutes, every second counts.
Autonomous flying vehicles improved
Finally, they’re developing better models of how quadrotors perform at high speeds, where aerodynamic effects are less predictable.
According to Waslander, Waterloo provides the perfect setting for his research. “The thing that drew me to Waterloo was the access to top students in Canada and the mechatronics program itself,” he says. “There’s also no other place in Canada that has the same sort of concentration in robotic startup companies.”
That gives WAVELab the opportunity to collaborate with world-leading firms like Aeryon Labs and Clearpath Robotics — startups started and staffed by Waslander’s former graduate students. Waslander benefits from access to gigabytes of flight data from the alumni-founded companies — far more than he could ever collect on his own — while the companies are able to turn to WAVELab for the industry-transforming research.
In the next three to five years, Waslander expects to see breakthroughs that will put far more autonomous quadrotors in the air. So if you’ve been hearing a lot of buzz about drones lately, just wait. “It’s only the beginning,” he predicts.