Tuesday, January 5, 2016

GM, Mobileye Concocting Crowd-Sourced Mapping Data for Autonomous Cars

GM/Mobileye lane-mapping

Today’s semi-autonomous driving features are neat, but they only work on highways—and well-marked highways at that. Self-steering systems used by Tesla Motors, Mercedes-Benz, Infiniti, and others rely on painted lane markings, but that won’t be good enough for fully autonomous rides. General Motors has a plan, however, and is working with supplier Mobileye on a mapping project that might offer future self-driving cars a redundant backup to their visual understanding of painted lane markings.

The answer is found in crowd-sourced imaging, which can update a central database on highway markings; autonomous cars can then cross-check what their cameras are seeing against what other cars have “seen.” According to GM, gathering this data is as simple as using Mobileye mapping strategies and applying them to GM vehicles equipped with forward-facing cameras (which many GM cars already have) and a 4G LTE data connection. As the vehicles roam America’s roads, they upload visual lane data to a central database. The database, then, becomes a sort of real-time fact-checking resource for self-driving cars.





This is both clever and important, because today Tesla’s AutoPilot highway self-driving function relies solely on the car’s visual understanding of lane markings. If the markings aren’t visible enough, the car shuts off AutoPilot. That’s fine when a driver is still there to retake control, but what about the coming fully autonomous cars? Those cars will require both precision GPS and high-definition understanding of lane markings, traffic signals, and more to safely maneuver themselves about. According to Mobileye, its “Road Experience Management,” or REM, system can process road details at just 10 kb per kilometer, reducing the load on the vehicle’s data uplink, while providing “localization at an accuracy of about 10 cm”—far more precise than today’s GPS.

Given the number of GM vehicles on the road that are equipped with both 4G LTE and forward-facing cameras, the lane-sniffing potential of the company’s fleet is prodigious. That said, so far the technology remains in an exploratory phase. GM’s executive vice president of product, Mark Reuss, says “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,” adding that GM plans “to explore the integration of REM into existing GM program launches sometime later this year.”

2016 Consumer Electronics Show


from Car and Driver Blog http://blog.caranddriver.com/gm-mobileye-concocting-crowd-sourced-mapping-data-for-autonomous-cars/


No comments:

Post a Comment