General Motors is planning to use cameras on customer cars to build the maps that will help self-driving cars navigate, it announced this morning. - wired.com
The automaker is exploring new technology from Mobileye, an Israeli provider of visual processing chips and software. Mobileye’s tech can detect vehicles, pedestrians, and other obstacles, as well as road markings, signs, and traffic lights. It’s the stuff that powers popular features like lane departure warnings, and is installed on hundreds of thousands of GM cars.
GM’s plan is to pull that camera data, via its OnStar system, from customer cars to create highly detailed, constantly updated road maps. Those maps would allow an autonomous vehicle to know its location within about 10 centimeters. That’s a huge advantage over current GPS systems, which count their margins of error in meters—good enough for knowing what street you’re on, but not for navigating a robo-car through traffic.
Mapping is an increasingly important part of the push toward autonomous driving.
Mapping is an increasingly important part of the push toward autonomous driving. The more a car knows about an area, the more it can focus its sensors and computing power on temporary obstacles like cars, pedestrians, and cyclists. “Creating and updating maps using on-board camera technology supplies the missing link between on-board sensing and the requirement for full redundancy to enable safe autonomous driving,” says Amnon Shashua, Mobileye’s co-founder and CTO.
That’s why the consortium of BMW, Audi, and Mercedes recently bought Nokia’s mapping arm, HERE, for $3.1 billion. It’s why TomTom is still relevant. It’s why Google has a fleet of cars loaded with sensors scoping out all the roads its autonomous cars will later traverse.
Since the maps GM wants to make will be based on visual data, they are unlikely to be as detailed as those that HERE, TomTom, and Google are creating, which are based largely on LIDAR data. But GM has an immediate advantage by using technology that goes into its cars anyway: scale.
“GM is committed to bringing semi-autonomous and fully autonomous vehicles to our customers, and this technology will be a critical enabler to getting us there,” says Mark Reuss, GM’s head of product.
Right now, GM is testing the technology on five cars; the second phase would increase that up to 30. If that testing goes well, a company rep says, “this could move quickly,” adding that it could get the tech into its new cars later this year. |
