Deel 2
MORE BY DRIVING TO THE FUTURE
Josh Switkes, CEO of Peloton Technology stands for a portrait at Peloton Technology headquarters on Wednesday, February 15, 2017 in Mountain View, Calif. Mountain View’s Peloton announces truck platooning partnership Ford's CEO, Mark Fields(middle with sunglasses), visits Ford's Silicon Valley research center to announce research on a self driving car planned for 2021 on Tuesday, August 16, 2016, in Palo Alto, Calif. Ford puts $1 billion in stealth artificial intelligence startup FILE - In this Thursday, Aug. 18, 2016, file photo, Matt Grigsby, senior program engineer at Otto, takes his hands off the steering wheel of a self-driving, big-rig truck during a demonstration on the highway, in San Francisco. Uber's self-driving startup Otto developed technology allowing big rigs to drive themselves. After taking millions of factory jobs, robots could be coming for a new class of worker: people who drive for a living. (AP Photo/Tony Avelar, File) Otto’s robot trucks face questions about legality in California The Waymo self-driving car is unveiled at Google's offices in San Francisco, Calif. on Tuesday, Dec. 13, 2016. Self-driving cars are getting smarter, reports show
“Basic navigation is not enough for automated cars,” said Ugur Demiryurek, associate director of the Integrated Media Center at the University of Southern California. “Robots need granular data on every curve, elevation and deviation. It has to be extremely precise, almost like brain surgery.”
Here will have 310,686 miles of HD maps by year end, covering all the major highways in North America and Western Europe. By 2020, its HD maps will cover 2.49 million miles.
Here employees worldwide drive some 400 mapping cars (Here called them “terrestrial capture units”) topped with 4-foot high rigs bristling with sensors to collect the HD data. The high level of precision comes from $70,000 lidar scanners. Lidar, commonly used in self-driving cars as well as mapping vehicles, is a type of laser radar.
For both regular and HD maps, Here collects data from municipalities to stay abreast of construction projects and other road changes. It also has a crowdsourced site called MapCreator that draws about 100,000 visitors a month who contribute updates on their neighborhoods. Here automatically extracts data from various sources such as trucking fleets and mobile phones.
Even before robot cars hit the road, Here is working on better information for human drivers. It is now pooling anonymized real-time data from hundreds of thousands of newer cars (2014 and later) from the three German carmakers that own it. If several cars turn on their windshield wipers, its system will know that there’s rain ahead, for instance. It will use that data for a new service called Lighthouse to give drivers dynamic updates on road conditions: traffic, hazards, parking and changing speed limits.
While the 105-person Berkeley office is just one small offshoot of a global company with a workforce topping 7,000, engineers and programmers in the East Bay location play key roles in creating the HD digital maps. Other Here workers get even more hands-on to assemble the rigs for the mapping cars at a garage on Gilman Avenue.
After being built in Berkeley, the rigs are shipped to Here locations worldwide to be mounted on cars for mapping runs. Each rig takes about a week to put together, with a day or so of test runs in Berkeley and San Francisco.
“These are some of the most well-mapped cities anywhere,” said Cyrus McGuire, production lead. “People see us and wave, thinking they’ll get to be in a map, but don’t realize we’re just testing.”
Besides Google and TomTom, USC’s Demiryurek sees Uber as a Here rival. Uber bought mapping comany enCarta and some of Microsoft’s Bing mapping technology, both in 2015. “Uber is in a position to collect data continuously, since they have drivers all over the world,” he said.
Uber hasn’t disclosed whether it will license the maps it is creating for its future self-driving taxis.
“The world changes,” Intel’s Davis said. “A city may rip up a street and the barriers move around as construction is under way. We think Here has the footprint, the technology and the approach to optimize the kind of maps automated cars will need.”
Here’s Sood put it even more simply.
“Building a map is hard,” he said. “Keeping it up to date is equally challenging.