How do self-driving cars navigate with limited road lane markings?

What is swarm data?

Let’s think about crowd-sourcing – gathering information from many individuals and different sources, then combining and sharing it. Swarm data, in the automotive world, works in a similar way, only that information is gathered by and shared between multiple vehicles. As a prerequisite, drivers must consent to their car’s data being shared. Such information may include data not only about the individual vehicles themselves, but also their surroundings, the road conditions, the precise location of road infrastructure – including signs and lane markings – and even the line they drove on specific roads.

As a vehicle drives down a road, it gathers data with its various sensors and sends it to the cloud where it’s made available to other vehicles. Then, as more vehicles drive on the same road and record the same environment, they continuously verify the accuracy of the information. If the conditions and infrastructure in the environment change, vehicles detect those changes and update the data in the cloud for all cars on the road. In this way, millions of vehicles can communicate with each other, passing on information about a vast road network – even for cars that have never driven these routes before. They can easily follow their path and keep to their lane whenever their sensors are unable to detect clear road markings.

The foundation for automated driving functions

Swam data forms the foundation of very precise, highly automated driving (HAD) maps. These are more advanced than standard-definition and high-definition maps, which contain information like the street network and lanes, and are used for the purpose of orientation. HAD maps, by contrast, allow the self-driving car to calculate its exact location in relation to everything in its immediate environment. This is called localization. HAD maps tell the vehicle what the road geometry and surface topography look like – not just to a few meters, but with the absolute precision needed for automated driving functions.

The self-driving car therefore doesn’t always need clear road lane markings to use as a guide. Instead, it can base its trajectory on the HAD maps created by swarm data from other vehicles that have already used a specific road. Automated driving functions like Travel Assist can use just one identified road lane marking to keep the vehicle where it should be. This means that automated driving is also available on country roads without a central lane marking, for example. As the amount of swarm data gathered increases, so too does the accuracy and precision of HAD maps, broadening the availability of this technology and allowing self-driving vehicles to continue on their journeys.