Uber restarts modified self driving testing in Toronto focused on AI
TORONTO—Uber Technologies Inc. says its self-driving vehicles have returned to the streets of Toronto in a modified program after it halted testing earlier this year when one of its autonomous vehicles struck and killed a pedestrian in Arizona.
The company said Thursday that the limited relaunch of a few vehicles in the Toronto area will have a driver in control at all times and include a second operator as a precaution. The Arizona crash in March happened while an operator was streaming a television show and the vehicle was operating autonomously.
Raquel Urtasun, chief scientist at Uber Advanced Technologies Group in Toronto, said in an interview that since the incident Uber has reviewed safety measures company-wide, including at the research group.
“Since the tragic incident, ATG has been working, has really been looking at all of our processes and the way we develop the technology.”
She said safety comes first and the company has looked to emphasize more simulations in autonomous development.
“We have revamped all our offline testing such that we minimize the need for tests on the roads, and instead we leverage all of our data collection processes and we can test and create simulations so that we minimize the risks until the technology is ready.”
The Toronto testing will focus on development and data collection for the company’s artificial intelligence systems including map building, more-so than testing the real-world self-driving capabilities of the vehicles, said Urtasun.
Self-driving vehicles generally require high-definition road maps to operate and supplement the various on-board sensors. The company says it’s working to build technology that will allow map building in real-time to speed up the mapping process.
Uber says it has also restarted testing in Pittsburgh, where vehicles will be operated in a fully autonomous mode with two safety operators, and in San Francisco where drivers will control the car similarly to the Toronto program.
The company says it also plans to move its ATG research and development hub a couple kilometres west from its current base in Toronto’s MaRS Discovery District tech hub as it plans to double the size of its AI research team to 100 next year.
The growth of the AI research team is on top of plans announced in September to invest upwards of $200 million and hire 200 engineers in its first engineering facility in Canada.
The Advanced Technologies Group will relocate to a mixed-use development under construction by RioCan called the Bathurst College Centre.
Urtasun said the company hasn’t had problems with recruitment yet despite strong demand in the Toronto technology sector.
“This is one of the fundamental problems that, you know, is going to influence the way we live in this century. So everybody’s very excited to be working in self-driving.”