By Brian Caldwell on Nov 09, 2018 09:01 am
Members of the Autonomoose research team gather for a photo with their pride and joy, a Lincoln MKZ hybrid sedan that they are working to make self-driving.
Researchers at the University of Waterloo reached an important milestone this summer by logging their 100th kilometre on public roads in a self-driving car.
Recorded in an industrial area of Waterloo, it was the culmination of almost two years of work since the research team won approval from the Ontario government to do on-road testing in an autonomous vehicle pilot program.
“We’re pulling this off and 100 kilometres is symbolic of that,” says Krzysztof Czarnecki, co-lead of more than 60 professors, engineers, researchers and graduate students working on the project. “It’s a number you can’t reach without having a robust system.”
The research team also showed off its progress in October by giving demonstration rides to more than 100 guests from the media, the University and industry partners at a test track in Waterloo.
Michael Smart, project software lead, explains what is happening as Autonomoose stops for a vehicle in its lane during a day of demonstrations in October.
Autonomoose, the modified Lincoln MKZ hybrid sedan at the heart of the initiative, completed several complex manoeuvres on its own after the driver behind the wheel turned over control to its self-driving systems.
That included identifying and avoiding hay bales set up as an obstacle, stopping for a parked car in its lane and negotiating a left turn at an intersection after yielding to an oncoming vehicle.
To make such manoeuvres possible, Autonomoose is equipped with nine cameras, LiDAR and radar scanners, and other sensory devices that gather data for processing by sophisticated on-board computer software.
The car travels in autonomous mode on public roads at a top speed of 35 km/h, negotiating intersections and light traffic for up to two kilometres at a time without intervention by a driver seated behind the wheel.
“This is an important stepping stone,” says Danson Garcia, the lead hardware researcher. “We can definitely see this is doable and that our research is moving forward.”
Next steps include reducing the number of human interventions, or takeovers, improving perception of other vehicles and pedestrians, and refining the artificial intelligence required for the car’s autonomy software to make complex decisions and manoeuvers on its own.
The project, the most advanced university-based program of its kind in Canada, also involves research into the unique challenges of autonomous driving in winter weather.
A longer-term goal involves developing the high-definition 3D maps and capabilities required for Autonomoose to drive itself to campus through signaled intersections in all kinds of traffic from the test track, which is located several kilometres away.
Despite their progress to date, one of the key lessons learned by researchers is how incredibly challenging autonomous driving is due to the sheer number of variables and possibilities in play.
“We always have to plan for unexpected issues,” says Czarnecki, a professor of electrical and computer engineering at Waterloo. “Everything takes longer than anticipated.
“There is still a lot of tuning needed and development to follow up on, but for us this is a huge achievement. When we did our first road test, the system was really just a fundamental subset of what we have now.”