Waterloo autonomous vehicle team is looking forward to longer public drives in 2018
Nikolas Stewart (BASc ’08)
Alumnus, Faculty of Engineering
> Autonomous Vehicles Program Manager
In the year 2030, you might get into a car for a long highway commute and sit back and listen to your smartphone reading you the morning emails. When you come to a complex part of your journey that requires you to take the wheel, your autonomous car will alert you well in advance.
Such a near-driverless future is coming.
Exactly when is hard to predict, but Nikolas Stewart, a mechanical engineering alumnus and now the autonomous vehicles program manager for the University of Waterloo’s WAVE and WISE labs, is helping to bring you that future, along with student teams, Faculty of Engineering researchers and industry partners.
The University of Waterloo received a revamped Lincoln MKZ in 2016 that researchers and students turned into an autonomous vehicle dubbed Autonomoose. Later that year, it became the first vehicle in Ontario to get the special licensing needed for a self-driving car to be tested on public roads.
The Waterloo team wrote software for what are essentially the car’s “eyes,” – sensors, cameras, high-accuracy GPS, LIDAR (optical remote sensing that uses lasers) and inertial measurement units for navigation and control. The team also created the vehicle controller (which drives the car), a localizer so the car can recognize where it is using the sensors, and multiple AI systems such as obstacle detection (pedestrians, vehicles, bikes), behavioural planning of what these obstacles might do, and route planning based on its environment and what the obstacle and behavioural planner are feeding it.
Most of 2017 was spent on testing and improving the software, and working with industry partners to get the vehicle ready for prime time. In 2018, the plan is to have the car on the road for a more complete test run.
“We have tested out individual scenarios on small sections of road, but we would like to do more extensive routes this year with more complex challenges,” says Stewart.
The test run will happen with a driver behind the wheel who can take over in a pinch, but having the autonomous vehicle out for longer drives with more complex scenarios will give the team additional information about how well it performs in real road conditions and a better sense of what still needs to be perfected in a driverless car.
The development involves numerous interesting challenges for student engineers. “One of the challenges we have been working on is getting the object detection to happen in the car in real time at a fast-enough rate,” Stewart says. “We have a very powerful object detection algorithm but it also requires a lot of power to run. This is a common problem in the autonomous driving field that we will be working on extensively in 2018”
In Canada, there is also the challenge of a lot of winter snow, and the University of Waterloo is in a good location to work on overcoming that problem.
Completely self-driving cars that will allow us to just fall asleep in the back seat is probably a much more distant future vision. But cars will become more autonomous in the coming decade, Stewart says. Already, there are cars on the market with fairly sophisticated cruise control and self-parking systems. In the future, with fast object detection, artificial intelligence systems and other technology, “we will have cars with a lot more autonomy, even though there will probably still be a driver behind the wheel,” he says.
Autonomous vehicles will ultimately change the way transportation is designed. It might lead, for example, to designated lanes for self-driving trucks, and change the way families commute and share their vehicles.
The transportation future “is definitely very interesting and exciting,” Stewart says.
