Why autonomous vehicles won't be taking over Ontario's city streets in 2022

The drive to put autonomous vehicles on the market has sped up over the last decade but that doesn't mean self-driving cars will be truly road-ready anytime soon.

"We are way far from that level where a machine can drive like humans in any conditions," said Amir Khajepour, a professor of mechanical and mechatronics engineering at the University of Waterloo.

"Whether that will happen in my lifetime, I'm not sure."

In 2016, Ontario launched its Automated Vehicle Pilot Program. It recently expanded guidelines to allow some self-driving vehicles to be tested on certain roads without the help of someone behind the wheel. 

The changes allow for a passenger in the vehicle or a remote operator, but there are still big hurdles to overcome.

The biggest, according to Khajepour, is making machines that mirror brain functioning when we drive, and that is far from easy. Add in debates about insurance and ethics, and it's clear there's a lot more to the industry than just technology.

"When we drive, we draw from many other experiences we have in our lives," said Khajepour.

Programming robots to react to real world situations has been a challenge and is inherently linked to safety, says Steve Waslander, an associate professor at the University of Toronto Institute for Aerospace Studies specializing in AI perception.

"Where they still seem to be blocked is in … all the rare events, the weird situations, the infinite variety of the human world," he said. 

One example of that, Waslander says, is how self-driving cars can have a difficult time recognizing children in Halloween costumes. In an effort to improve that, Google's self-driving car project, a subsidiary called Waymo, recently had its employees' kids parade in front of their cars in their Halloween costumes so its sensors could collect more information.

Christmas lights can also present a challenge as can winter weather, Waslander says.

On top of that, autonomous cars aren't yet equipped to handle driving like humans, who give sometimes subtle cues to other drivers to signal their intention such as inching forward at an intersection to make a right turn.

"It's amazing all the things we do in parallel as we're moving through the environment, and to get to that point has been a much bigger challenge than we expected," said Waslander.

But technology and software aren't the only considerations for the sector; there's a whole "ecosystem" of factors to take into account, as Khajepour puts it. Engineers have the added job of debating the ethics of creating artificial intelligence and working out who's responsible if an autonomous vehicle crashes. Although rare so far, it does happen.