In self-driving age, digital maps are crucial

Five years ago, the most sophisticated in-car navigation system displayed roads and street names, using GPS to locate a vehicle within a few yards and provide mostly reliable turn-by-turn driving directions.

To update the map on road changes, owners had to go to their dealership for a new disc with the latest data and upload it to the car. Maybe owners used their onboard nav systems; maybe they didn't.

For the coming generation of self-driving vehicles, digital maps are critically important. Mapping is no longer a convenience feature, but must be precise enough to safely guide vehicles when no human is behind the wheel.

To address the increased need for precision, traditional digital mapmakers such as Here and TomTom International, and newcomers such as DeepMap and Civil Maps, are using new developments in cloud storage and high-powered sensors to generate maps that can pinpoint a vehicle and its surroundings within inches.

The industry has not settled on a path for mapping technology, and emerging suppliers are moving in different directions.

Through a partnership with Bosch, TomTom is constructing live maps via radar sensors that surround research vehicles. Eventually, the maps will be constantly updated from data gathered by cars using Bosch radar systems installed as part of automated driving functions.

DeepMap is forgoing research vehicles, developing software that will translate data picked up by passenger vehicle sensors into a detailed map.

Here, a mapmaker jointly owned by BMW, Daimler and Audi, is building its product based on a roadway mapping system developed over the past decade.

Here introduced its HD Live Map concept in January 2016 as a system intended to provide autonomous vehicles with detailed and updated information on their surroundings. The map is generated from data collected from Here's global fleet of 500 research vehicles. The data created a "base layer" of road information, with real-time information continually added from sensors on other vehicles.

"The HD map bridges what the vehicle sees to what it should do," said Sanjay Sood, vice president of highly automated driving at Here. "It provides a detailed view of the roadway, helping the vehicle position itself on the road."

Navigation became a generic vehicle feature thanks to the proliferation of Google Maps and Apple Maps on smartphones, Sood said. But the heightened importance of maps to autonomous vehicles became apparent in 2013, when Daimler performed its first self-driving demonstration. That experiment had an autonomous vehicle follow the historic route of the first recorded long-distance drive in an automobile in 1888 — a 112-mile round trip in southwest Germany by Bertha Benz, wife and business partner of German auto pioneer Karl Benz.

The modern-day project team quickly saw that more technology was needed.

"They realized they needed some sort of high-precision map," Sood said, adding that the car would be unable to center itself in the road or avoid nearby obstacles if it were to rely on the accuracy of a GPS-based system.

High-definition maps such as the system Here is developing supplement the information the car takes in from its sensors. By generating "localization points," the sensors enable an autonomous vehicle to place itself on its map precisely within inches. It can use the resulting map to ascertain where it is on the map, anticipate upcoming obstacles or prepare for speed limit changes before its sensors detect the physical road signs.

Legacy navigation systems can't handle such precision, Sood said, because of how long it takes to update the information and the amount of onboard data required to keep it all straight. But the earlier systems have provided a base.

To usher the world into self-driving vehicles, Here is reducing map science to one URL.

Here's fleet of mapping vehicles, which was originally launched to generate the company's in-car navigation system, is constantly driving local roads, covering 200 countries and more than 31 million miles of roadway. Their flow of information is used to make a "skeleton map," marking the direction and bank of the road, lane markings and other basic information.

On top of that map layer, data collected from passenger or commercial vehicles with camera, radar and lidar sensors provide real-time updates to the map as they drive. One car passing by a construction zone can record it and upload it, updating the map for all vehicles on the route behind it.

This form of data collection, which is expected to improve as more vehicles with sensors for advanced driver assistance systems hit the road, allows the map to update in real time.

Sood said Here can collect data from vehicles made by any automaker, as long as owners agree to share their data with the automaker. Here is working with automakers other than the supplier's joint owners to create its high-definition map. He declined to reveal those companies.

All of the data for the Here map are stored in the cloud — a network of computers accessible by vehicles via wireless connection. To gain access to the map, automakers must program the car with Here's software, which gives the vehicle access to the data via a URL.

To save space and computing power, Here will give vehicles slices of map data at a given time, enough to provide information about the surrounding area and the immediate driving route.

In the near term, Here's cloud technology could benefit drivers by adding real-time data to in-car navigation systems — a feature analysts say could become more common as cars become more connected.

"With more map layers, car buyers are going to get more driver-centric data," said Colin Bird, a senior analyst at IHS Markit. "Maps can provide real-time traffic updates, updates for parking spots and predictive analytics."

Here's HD Live Map won't be online until the first Level 4 autonomous vehicles — representing cars that require no human interaction in defined conditions — are on the road, which some automakers say could be as early as three to four years. Sood said the first iterations of the map could include a screen inside the vehicle to let passengers see what the car sees, to provide a level of comfort and trust. But because there is no real need for such screens, automakers will phase them out, he said.

Once sensor-laden autonomous vehicles become a more common sight on public roads, Here's fleet of research vehicles also will no longer be necessary.

Fleet vehicles have become a point of contention among high-definition map suppliers. While Here is using its fleet until more advanced passenger cars populate the roads, TomTom's radar-based maps rely on vehicles with Bosch sensors, requiring a specially operated fleet.

DeepMap CEO James Wu, who also worked on mapping at Google and Apple, believes the amount of data already being collected is enough to render fleet vehicles an unnecessary and costly operating expense.

"We let the OEMs, who own the cars, do the data collection," Wu told Automotive News. "We don't need 300 cars to scan the world. We can never keep up with the data updates, and the OEMs have all the data."

Sood counters that such an approach works only in high-traffic areas such as large cities. Maps still need fleets to create a more complete picture, he said.

"You need hundreds of thousands of vehicles to create a map that's global," he said. "OEMs want a map that fits the globe."

Along with effectively using its fleet, the key to success for Here will be data that is mutually agreed on and standardized. If every vehicle is contributing data from its sensors, and every vehicle has access to that data, the supplier can reach a complete and accurate mapping system.

"No one player can have sufficient volume to create this kind of content," Sood said. "OEMs are likely going to be sharing this data for the betterment of everyone."