SUBSCRIBE TO OUR BLOG
An older generation may recall riding in cars with their parents accompanied by huge tomes of confusing data, with names like Thomas Guide, Rand McNally, or AA Glovebox Atlas. Those unwieldy books were collections of printed maps made for every conceivable area of the road-navigable world, each sold separately. Despite no refresh rate and the occasional need to stop to ask for directions, they got drivers where they were going.
Though ours was a digital solution, the flat map approach is where HERE started in 1985. Maps were a two-dimensional system of links and nodes. The route provided to a driver was determined simply by where you could drive, and where you could not drive. We layered in a number of Points Of Interest (POI), and bang – we had a digital solution to create a custom map from point A to point B.
Over time, we added more information to that system. Most meaningfully, we added data that accurately captured the curvature and slope of the road. Then we added GPS to the mix, so that drivers could see where they were in relation to where they were going. The combination of all these things are what we have come to refer to as the Standard Definition Map.
In 2013, Mercedes asked us to help create the Bertha Benz project, and it made us rethink everything. At the time, our maps were made for humans. As humans, we interpret the map data to make the best decisions. For a computer to make decisions - what’s more for a computer to make safe decisions - we needed to provide a lot more information.
We named our solution 'HD Live Map', named so for the bigger data, higher accuracy, greater precision and freshness. HD Live Map is comprised of three maps layered in levels of detail.
The first level is not so dissimilar from the original Standard Definition maps – a highly accurate road map based on curves, elevation and coordinates. That road-level data stays at the core of the maps our cars use, but it’s not yet enough to meet the HD Live Map standard.
The second level of the HD Live Map is the HD Lane Model. To make the right decisions, an autonomous vehicle needs to know the type of lanes make up the road you’re on. Is the furthest left lane a shoulder lane or a passing lane? Is the right lane an HOV lane, or is it an express lane that changes direction depending on what time of day it is?
What is the width of the lane? Are the lane markers striped or solid? Are they yellow or white? What is the average speed one lane versus another? All of this data is being continually collected and updated. The lane model provides a highly accurate view of the road, and it’s a unique technology that enables cars to drive themselves safely and intelligently.
Now that we have an accurate look at the roads, and we know all lanes, we’re at the third element of the HD Live Map: the HD Localization Model. This is the layer that works with the environment around the car, and unlocks the ability for a car to dependably drive itself. What may surprise you, is that it does the job without depending on GPS.
That might be sufficient when a human driver is at the wheel, but if your car is driving itself, even a 15-foot location discrepancy can be quite dangerous. The computer needs a way to orient itself. Signs, poles, guardrails, walls, and barriers don’t exist in a standard definition map, but all of those elements exist in the HD Localization Model.
With this added layer, a vehicle continually checks its orientation, which enables it to stay in its lane, stop in the right place at intersections, and make turns accurate within 20 centimeters.
Mix these three layers together, and you have the HD Live Map. It’s a perpetually updating system that enables self-driving vehicles to navigate with incredible accuracy – quite a long way from having to stop and ask for directions.
In the next part of this series, where we’ll discuss how HD Live Map collects and processes the data of the road.