The next step in the race for self-driving vehicles

By John Bendel, editor-at-large

What can a driverless truck do if the roadway is covered with snow? Pull over and wait for the spring thaw?

Not really, say the experts. High-definition digital maps and GPS accuracy within a few inches will keep it going. HD maps will know exactly where a lane is; GPS will keep the driverless truck in it – even in a whiteout blizzard. The GPS precision is already available. The HD maps are on the way, and in what seems like a big hurry.

It’s a new chapter in the developing story of self-driving vehicles. Semi-autonomous vehicles with drivers can do fine without HD maps. Totally driverless vehicles cannot, engineers say, and totally driverless is what the big money investors want. So even as tech and automotive giants tweak and test self-driving vehicles at a frantic pace all over the country, the same players find themselves in a new race – this one to develop those all-important HD maps.

A consortium of Daimler, BMW, and Audi recently acquired Here, a digital mapping company. Ford took a stake in Civil Maps, an HD mapping startup. General Motors bought Cruise Automation, a developer of driverless technologies; Cruise promptly jumped into HD map development. Intel bought Mobileye, an Israeli tech company that works extensively with HD map makers. Lesser names like Tom Tom, Bosch, Sanborn and NVIDIA are also in the race.

Except for that snow-covered road, HD mapping is less necessary on the highway than in the city with so many signals, signs and traffic patterns, engineers say. Virtually all the big players share a vision of the future in which Uber-like services replace both bus transit and individual car ownership. Of course those players include Uber itself. So most of the money is in cars because that’s where the biggest payoff is seen – at least for now.

On city streets, cameras, LIDAR (laser-based radar and short for light detection and ranging) and other sensing gear won’t provide all the data engineers believe driverless cars will need. HD maps will fill in the gap, providing knowledge of what lies ahead, they say. Because things change so quickly and so often, the maps will have to be updated constantly, and HD maps tend to have more than one layer of data. For example, Here’s HD map has three layers. One is for real-time data like traffic and weather. Another is for analytics that note and adjust to your driving habits. We’ll come back to that one.

Then, of course, there is a layer with roadway and lane location information accurate to within inches. So the three-layer map will do more than keep a vehicle in lane. It will adjust its route in real-time for construction, emergency street closings, parades, whatever.

Creating an HD map is a big job. A unit of Google began working on self-driving cars and HD maps in 2009. That unit was spun off as Waymo, a separate company under the same corporate umbrella. Waymo came to life in 2016 with HD maps for roads around Google’s Silicon Valley headquarters and for a handful of other cities, but that’s all. Given Google/Waymo’s huge investment in self-driving cars, why such a limited range of HD maps?

It’s because creating those maps ate up a large chunk of that huge investment. The company had to deploy cars equipped with high-definition cameras and LIDAR to note every curb, driveway, stop sign, fire plug, tree, and bump in the pavement. All that data came back to Google where technicians pored over it to tag every relevant item by hand.

That’s not necessary in 2016. Advanced algorithms now analyze and tag the data quickly and cheaply. Recently, NVIDIA, the Silicon Valley company best known for computer chips and video cards, introduced a system it claims can create HD maps in near real-time even as the data-gathering vehicle gathers the data. That speeds up the process, but as it turns out the process may never really end. Engineers say the maps will have to be updated continuously, so those data gathering vehicles will never be done. They will have to traverse every highway, street, and alley in a given area over and over to keep the maps accurate and current.

Experts say that gives a major boost to the automobile manufacturers like GM and Ford. They have the opportunity to deploy many more data gathering vehicles than, say, Waymo or Uber. In fact, according to the New York Times, the Here consortium that includes Daimler “is drawing on data scooped up by scanning systems that trucking companies have agreed to install on their vehicles.” However, when I asked if they are involved in the effort, Daimler responded “no comment.”

Given the complexity and fluid nature of HD maps, you probably won’t find them on a CD, DVD or thumb drive. HD maps will reside in the cloud and will communicate with your truck or car over an internet connection. According to Here, which already provides HD maps on the web, map data will allow your vehicle to “see” the road ahead.

That brings us back to that analytical layer on the Here map. With the ability to see and plan ahead, says Here, your truck or car will be able to adjust its driving style, “be it sporty and fast or relaxed and easy.” LL