Readable roads for self-driving cars

Ma(r)king roads readable for self-driving cars

For safe and efficient deployment of self-driving vehicles to become a reality, we first need lane markings and road signs that can be read easily by cameras and sensors

Words | Richard Neumann, Swarco, Austria

CAV, V2X, C-ITS, CCAM – the acronyms differ, but they all have one thing in common. They all represent a connection between road infrastructure and vehicles that aims to improve traffic flow, reduce emissions, make traffic safer and make our journeys more convenient. Vehicle automation is often cited as one of the biggest disruptive technologies, both economically and environmentally. But for automation to happen, the surrounding infrastructure must be adapted to support efficiency and safety. Connected infrastructure is key to design and certification procedures as well as for testing environments. Connectivity between infrastructure and vehicles is also necessary to support operational conditions in the transition period, which could last for decades.

Apart from the automotive and C-ITS aspect of managing the traffic of self-driving cars, it is also essential to consider the road itself, and in particular its readiness to be read by driver assistance technologies.

“Lane markings are currently read by two kinds of sensors: camera and lidar,” says Harald Mosböck, Swarco’s vice president of road marking systems for Europe and the APMEA regions, and vice president of the European Union Road Federation (ERF). “Even the most advanced vehicles equipped with high-definition maps need sensors to capture information from lane markings and traffic signs, and to determine whether all the data fits together. Road markings and traffic signs still lack uniformity on roads. There are too many differences. And for as long as these things remain inconsistent, we will not be able to fully rely on sensors.”

Global initiative

On an international level, it is mainly the Japanese, American and European standardization bodies pushing to improve and standardize road markings. The latest edition of the North American Manual on Uniform Traffic Control Devices (MUTCD) is to contain a requirement for major roads that normal lines be 150mm (6in) wide and for markings to be more uniform. Just recently, a major standardization body proposed to the Federal Highway Administration (FHWA) that lane markings on the future highway network should not only have a standardized width, but there should be a general requirement for dotted lines at highway exits and a standardized gap width for interrupted lines. Right now, there is work in progress addressing the topic of better retroreflectivity, which is a very challenging subject. Another issue is how lane markings in roadwork zones can be improved to attain reliable and consistent machine readability for automated or assisted driving.

In 2019, the European Council revised its road safety management directive and charged the European

24

The number of Commission to different weather conditions in which form an expert group to work out how to make Swarco has tested road markings lane markings and traffic signs easily readable for vehicles and, of course, for humans as well. Initial results were expected by mid-2021 but, due to the COVID-19 pandemic, will not be available before spring 2022. “In general we can say the communication between carmakers, industry associations and Brussels on the expectations for the quality of road markings has already greatly improved,” says Mosböck. One recommendation pushed forward by the European Road Federation (ERF) is very similar to the approach in the USA. The formula was created based on what the human eye can see, with the goal to improve road safety. Interestingly, several studies show that what is good for the human eye also works very well for machine vision. The 150 x 150 formula recommends a line width of 150mm and a

retroreflectivity (RL) of >150 mcd/m²/ lx under dry conditions (RL > 35 mcd/m²/lx under wet conditions).

Machine vision

To find out more about the ability of cameras and lidar to read lane markings, Swarco carried out tests in the world’s biggest climate tunnel in Vienna, in partnership with automotive lighting technology company ZKW. They implemented eight different kinds of lane markings - from standard to high performance systems - and investigated how they were ‘seen’ by three different kinds of cameras and eight different kinds of lidar sensors. They also simulated 24 different weather conditions, including bright daylight, nighttime, wind, stormy weather, heavy rain and fog.

The tests confirmed that anything the cameras could read, could also be read by the human eye. However, the cameras could not read everything that the human eye could see. “One strong point of cameras is their identification of contrast,” says Mosböck. “However, they show a weakness when glare light is coming from vehicles in the oncoming lane.” When it comes to lidar sensors, some highly retroreflective road markings successfully increase readability and detectability. “Special pigmentation enables the lidar to read orange road markings very accurately,” Mosböck confirms. “This is important in light of ongoing discussions in the USA about

Road markings and traffic signs still lack uniformity on roads. There are too many differences. Improving uniformity is a prerequisite to increase the reliability of readings by different types of sensors

Harald Mosböck, vice president of road marking systems for Europe and the APMEA regions, Swarco, and vice president of the European Union Road Federation (ERF)

Above: Connectivity between infrastructure and vehicles is essential for the future of automation whether lane markings in roadwork zones should be orange in the future. The lidar is also not affected by glare. If you combine cameras and lidar in a car, you will have very good readability and reliability for assisted driving.”

There is of course a financial element to improving road markings. “Whenever you discuss higher performance lane markings with the automotive industry or road authorities, the first question is always: Who is going to pay for that?” says Mosböck. “My belief is that all over Europe we need better performing lane markings on our road networks. Once we have achieved a level that is suited to the human eye, it will work for automated vehicles as well. There is still a long way to go because road authorities lack sufficient funds and I doubt they will be able to finance high performance lane markings for all road networks. But I hope to see major progress in better road markings for human and machine vision on the long-distance arteries of the trans-European network in the near future.” ■