By the fall, if all goes according to plan, a boxy van, une navette autonome, will make its way without a driver along a main strip of the Montreal suburb Terrebonne. This kind of project may be among the first steps of the revolution-to-come of self-driving vehicles.
Gently rolling along a preset course, at no more than 25 kilometres an hour, the pilot project is based on one in Lyon, France, and is in the process of getting provincial regulatory approval.
But the shuttle would seem far more sedate than the fleets of driverless cars people have imaged tooling around residential blocks or self-driving truck convoys thundering down the highway. This may be the more practical, immediate future of autonomous vehicles in Canada, according to experts.
And in the near term, it is a future in which driverless systems won't actually drive the cars. Instead, the driverless technology will be used to monitor drivers for potential errors.
More than just stepping-stones, these applications will show how an autonomous-mobility future might realistically come to be, said Steven Waslander, associate professor at the University of Waterloo and director of the Waterloo Autonomous Vehicles Laboratory.
On open roads at regular speeds, the technology is still far from adequate. An alert driver still needs to be at the wheel. Witness the driver killed in Florida last year in a self-driving Tesla. He was reportedly watching a film, distracted when the car collided with a tractor-trailor truck.
"The levels of reliability that humans can achieve – we're talking about one death every hundred million miles for drivers in North America – we're nowhere near that with autonomous cars," Dr. Waslander said.
The best driverless systems can run roughly 8,000 kilometres between times when they become confused and an alert driver has to intervene, Dr. Waslander said. Ironically, the danger will only worsen as the technology improves. Drivers will become more complacent, letting the cars drive themselves and not staying alert to that one rare instance when a driver suddenly has to take control. The technology has to become far better than humans for it to work.
"I don't think people are really going to accept autonomous cars until they are significantly better than human drivers," Dr. Waslander said.
Industry watchers see future funding for development in this sector in part from sources such as Innovation, Science and Economic Development Canada, with its recently announced $950-million supercluster funding initiative, likely to be partly earmarked for autonomous vehicles and all of its artificial-intelligence offshoots. Meanwhile, the University of Waterloo's research, for instance, has in part been coming from the federal Natural Sciences and Engineering Research Council of Canada's Research Tools and Instruments Grants Program. Expectations are that funding will only increase as the auto industry, insurers and policy makers continue to concentrate on driverless vehicles.
And as ambitious research continues, the near-term outcome will be the more practical applications, the gently rolling shuttle buses and systems that monitor drivers.
"It will monitor what you are doing in your car, and look at you and the road around you," Dr. Waslander said. "And whenever it's absolutely certain that you're making a mistake, it can buzz you, alert you, even take over and mitigate the effect."
But Canadian engineers and policy makers view autonomous technology, in some form or another, as the future of cars. Within 10 years, cars with artificial intelligence and accessing massive data systems to navigate may become the norm, especially as the next generation 5G mobile networks emerge.
"The important thing is that Canada's biggest impact is not in the traditional auto sector, but on the technology side. If you buy a new car today, about 5 per cent of the value is technology. By the mid-2020s, between 40 and 60 per cent of the value of the car will be technology. So, we've got companies involved in processing software, connectivity, cybersecurity," said Barrie Kirk, executive director of the Canadian Automated Vehicles Centre of Excellence, a consultancy and advocacy firm.
The hope for the Terrebonne shuttle is to use it for short, added extensions of the metro system. The project is a public-private partnership between Urbis, the public transit arm of the Les Moulins Regional County Municipality, the French company Navya developing the vehicle, and French-Quebecois passenger transportation operator Keolis Canada.
The front and rear look the same, and the shuttle will roll along automatically with a maximum of 15 passengers to the larger transit system. Although it will be restricted to 25 km/h, the shuttle's technology can go up to 40 km/h.
Although the pilot project would have just one shuttle, the hope is to have many in use eventually, "bringing more people to public transportation," said Marie Hélène Cloutier, vice-president of passenger experience, marketing and sales at Keolis Canada.
"We basically map the road that we want the shuttle to take. So, as an example, in Lyon, we have five stops," she said. Passengers select a stop by clicking on a screen. Keolis plans to have an attendant on board at all times during the project.
As simple as the ride may seem, it will rely on complicated mapping, as do all driverless vehicles. "That's a huge amount of data. And especially considering that conditions change so quickly, you actually need not just one map of every environment. You need hundreds. You need one at night, one in the daytime, one in snow, one in rain," Dr. Waslander said.
The next major step will be to manage that data, to eliminate information that is not needed in every map, such as structures which never change, and to focus on data that is needed at that moment for that car.
"You go to any autonomous driving conference, and there are four or five map companies that are showing what they have collected so far, and where they're going. It's pretty fascinating," Dr. Waslander said.