Back to Top

©HERE Technologies


Extremely Accurate and Always up-to-date

From buildings to bridges – to make autonomous driving a reality, high-resolution maps need to take the roads’ fingerprint. But how will it be possible to scan all of the streets in the world? The market leader among map providers and ZF have a plan.

Text: Christoph Reifenrath , 14. December, 2017

Christoph Reifenrath has been a TV and print journalist for 35 years and is part of the ZF Copy Team. His favorite part of the job is delving into complex technical topics.

Where am I? Where is my destination? How can I get there quickly and safely? These are questions that even cavemen asked themselves over 20,000 years ago. This is also when humans created the first maps which included important waypoints. And the high-definition maps that will be necessary for autonomous driving, and as such, the future of mobility, use more or less the same guidance and positioning mechanisms as our ancestors did thousands of years ago –important points of interest along a route. The key difference to the maps from back then are the number and precise position of the included locations. Today’s digital, real-time maps include the road itself, but also lanes, the radius of curves, lane widths, street signs, bridges, inclines or declines, guardrails, trees, embankments, ditches, and buildings, as well as their distance from one another.

Who has the best maps?

These billions of data points are then used to create a machine-readable image of the road surface as well as the entire surrounding environment – and as such, a nearly unique “fingerprint” of every stretch of road. Map provider TomTom calls this “RoadDNA,” competitor Here speaks of an “HD Live Map.” “When it comes to the development of autonomous driving in China, in the company Baidu we have found an extremely knowledgeable provider which develops high-definition maps and the services built on them,” says Matthias Benz. The head of Sales & Customer Development is responsible for this partnership within ZF (see “Autonomous with Apollo” below).

160,000 mapping vehicles would be needed to create an updated daily scan of the global network of roads.

Cars with sensor technology scan their surroundings

The information needed to create these maps is currently being supplied by special vehicles operated across the globe by the map providers themselves. They are equipped with state-of-the-art radar, lidar, and camera technology as well as differential GPS, which significantly improves positioning accuracy. As the cars drive around, they scan and record their surroundings – in their entirety and with centimeter-level precision. So far, this has primarily taken place on US and European highways, but projects to scan and capture fast-growing urban areas are currently in the works.

An HD map as a positioning tool

Since high-precision maps are not made for people, but instead for machines, the traditional way of presenting a map is no longer important. After all, for the purposes of autonomous driving, HD maps are performing a completely new task. Since positioning based solely on GPS isn’t accurate enough, and cameras can’t always provide the information needed – like when it’s snowing, for example – the maps themselves will become a localization tool. For instance, by comparing the input from sensors with the “fingerprint” of the street stored in the map, the vehicle can locate itself with centimeter-level precision, and even do so completely without GPS, if need be.

Mass-produced vehicles help keep the maps up to date

But this will only work if map providers constantly keep their HD maps up-to-date. The answer could be connected, mass-produced vehicles that capture the relevant data with their sensors while they are out on the road. Today, millions of mass-market vehicles that, in principle, could be used for this purpose are already driving around. The plan is to take the sensor data from these mass-market vehicles and compare it to the HD map in the cloud, use the data to update the map (if necessary), and then sync it back down to the vehicles. The goal is to keep the size of the updates as small as possible when doing so. This is why the HD maps will be divided into individual tiles measuring a few square miles in size, and the number of map updates from mass-produced vehicles will be kept to a minimum. Defining sensor categories and a standardized interface format should ensure that this works smoothly across vehicles from different manufacturers. In 2015, map provider Here drew up an appropriate specification together with its partners, and unveiled the standard they developed one year later. For the purpose of collaborative development, Here launched the innovation platform “Sensoris,” which is currently supported by more than 20 different companies.

“In Baidu, we have found an extremely knowledgeable provider in China that develops high-definition maps.”

Matthias Benz, Head of Sales & Customer Development at ZF

Available parking spaces and warnings shown in real time

The first real-time services based on current data from the field built on the Sensoris data standard were presented only a year later, back in 2016. For example, they show available parking spaces on the street and provide more accurate information about traffic congestion. In addition, these real-time data services speed up the announcement of warnings, like if a broken-down vehicle is blocking a lane on the highway.

How the HD maps stay up-to-date

1. Vehicles equipped with the necessary sensors, GPS, and an internet connection report a change in traffic routing on a specific section of road.

2. The information is aggregated and its validity is verified in the cloud.

3. The system uses this information to automatically generate new map data for the section of road in question.

4. The update is transferred to all compatible vehicles.

Autonomous driving with Apollo

In September 2017, ZF and the Chinese web services company Baidu entered into a partnership . Baidu is a leading provider of high-definition maps in China. But the two companies’ collaboration goes far beyond just HD mapping. Both partners want to work together in the fields of artificial intelligence, Big Data, and cloud-based solutions to make significant advancements to autonomous driving in China. To achieve this goal, ZF is contributing ZF ProAI , the vehicle control system it developed with Nvidia, to the strategic cooperation. The basis for the partnership is “Project Apollo,” an open development platform launched by Baidu in April 2017, that can be used to rapidly build systems for autonomous driving.

In a nutshell: For autonomous driving to work, we need high-definition maps. These HD maps must always accurately reflect reality on the road, including guardrails, trees, ditches, and other objects in the environment. Using modern sensor technology, special vehicles operated by the major map providers are already scanning a large portion of the global road network – but still far too few to update the entire relevant traffic infrastructure once daily. Connected mass-produced vehicles could be the solution. ZF has partnered with Baidu, the leading provider of high-definition maps in China, to push autonomous driving forward.

Related Posts