Freescale vision chip makes self-driving cars a bit more ordinary
The company's partnership with software companies will yield an electronics package that can see pedestrians, tell if a driver is dozing off, and initiate emergency-stop decisions.
Self-driving cars today are exotic. But with a partnership between chipmaker Freescale and software companies Neusoft and Green Hills Software, the idea is becoming just another technology on the computing industry's parts list.
The partnerships will produce an automotive vision chip that can be used to help
understand the environment around them. That includes things like detecting pedestrians, traffic lights, collisions, drowsy drivers, and road lane markings.Those tasks initially are more the sort of thing that would help a driver in unusual circumstances rather than take over full time. But they're a significant step in the gradual shift toward the computer-controlled vehicles that Google, Volvo, and other companies are working on.
Self-driving-car technology is arriving in the marketplace not with a single dramatic change but instead with a series of developments that gradually augment human abilities. Fully autonomous driving, which lets a driver take a nap or read a book, is qualitatively different from systems that are designed to operate when a human driver still has control. But driver-assist technologies are an important proving ground for fully autonomous driving technology.
Freescale, which Motorola spun off in 2004, specializes in embedded computing devices that give smarts to things like medical devices, mobile network equipment, and the motion-detection sensors used in the Guitar Hero video game. It's long had an automotive electronics business, too.
Through the partnership, Freescale will supply a processor with 64 computing engines that can be used to analyze the digital data from cameras' image sensors. The vision-processing chip design is licensed from CogniVue, and Freescale is its sole supplier to the automotive market.
Green Hills will supply an operating system for the technology, and Neusoft has software for tasks such as recognizing objects.
The technology package is set to arrive on the market in 2015, with upgrades that add more features due in 2016 and 2017. For example, one new feature arriving in 2016 will be support for "top-view" cameras, which combine imagery from cameras around a car's periphery into a single image that seems to come from a virtual camera directly above the car looking down. That can be handy for backing up, maintaining a lane, and seeing if there's a car in a driver's blind spot.