How do we make sure that the human is a reliable fallback and can promptly, safely, and efficiently take back the driving task from automated driving? That’s the key question for over 30 stakeholders from academia and industry, now present their findings from Europe’s first comprehensive pilot test of automated driving on public roads. The result is believed to help speed up and harmonize the development of automated driving systems in the future.
The European research project L3Pilot, led by Volkswagen and co-funded by the European Commission, has run from 2017 to 2021 with stakeholders from the whole value chain: car manufacturers, suppliers, academia, research institutes, infrastructure and governmental agencies, user groups and the insurance sector. The four-year project successfully ended with performing its Final Event in conjunction with the ITS World Congress in Hamburg 2021 on October 11-15.
Present at the ITS World Congress are Chalmers researchers from Vehicle Safety at the department of Mechanics and Maritime Studies – Marco Dozza, Linda Pipkorn, Pierluigi Olleja, along with SAFER representative Erik Svanberg - to show-case their research findings, which once started with the quest to find out how to optimize safety in automated driving:
“We all want full automation, that is, a vehicle that pick us up and takes us places without us having to think about driving. But tech is not there yet and there will be a transition phase with partial automation. That means humans and vehicles need to help each other and take turns in the driving task. The most relevant scenario is when a vehicle needs help from the human to sort out a critical situation that may lead to a crash. In such case, the research question is “how do we make sure that the human is a reliable fallback and can promptly, safely, and efficiently take back the driving task?” In our research, we tackled this question by exposing drivers to critical situations, where they need to take over control, and see how they do it. In this way, we can design vehicles that help the driver to efficiently coming back to the driving task rather than setting unreasonable expectations on human beings,” says Marco Dozza, professor at Vehicle Safety at the department of Mechanics and Maritime Sciences at Chalmers.
Europe’s first comprehensive pilot test on public roads
The project is the first comprehensive pilot test of automated driving on public roads in Europe, which makes it unique in its kind. Fourteen partners focused on testing automated driving functions in normal motorway driving, traffic jams, urban driving and parking. The pilots, running from April 2019 until February 2021, involved six countries besides Sweden: Belgium, Germany, France, Italy, Luxemburg and the United Kingdom and included two cross-border activities between Germany and Luxemburg as well as Germany, Belgium and the United Kingdom.
The project equipped 70 vehicles and the test fleet comprised 13 different vehicle brands, from a passenger car to a SUV. More than 400,000 kilometers were driven on motorways including 200,000 kilometers in an automated mode and 200,000 km in a manual mode as a baseline for comparison of the user experience and evaluation of the impacts. More than 24,000 km were travelled in the automated mode in urban traffic. With the aim to put the focus on the user experience of automated driving functions, over 1,000 persons participated in piloting and complementary virtual environment tests.
“We’re proud about the high number of advanced studies, with a real vehicle on test track and public roads, that we managed to perform within this project, especially given the pandemic. All of these studies advanced our understanding of how drivers behave - how they act and where they look - when transitioning from automated driving to manual in response to take-over requests,” says Linda Pipkorn, PhD student at Vehicle Safety at the department of Mechanics and Maritime Studies at Chalmers.
Unique data collection to enhance safety in automated driving
One of the major achievements of L3Pilot is establishing a Code of Practice for the development of Automated Driving Functions (CoP-ADF). It provides comprehensive guidelines for supporting the design, development, verification and validation of automated driving technologies.
The four-year project has also involved a considerable collection of valuable data based on the research findings on how pilot participants reacted when going from automated to manual driving in real traffic scenarios. The data will in the next step enable virtual testing to further enhance safety in automated driving.
“We found out that, in real traffic, drivers are able to transition control from automation to manual in response to a take-over request. The transition should be considered as a process of actions - look to instrument cluster, putting hands on wheel, look forward, deactivate automation - that requires a certain amount of time: up to 10 s in real traffic. Our research also showed that, in real traffic, drivers’ visual attention towards the forward road return to similar levels as in manual driving 15 s after a take-over request. In response to take-over requests, drivers may look away from the road towards the instrument cluster rather than to the road. This means that, designing safe automated driving functions requires take-over requests to be issued in all situations that require driver input. In addition, it is important for the automated driving function to be responsible for safe driving at least up to the moment of the automation deactivation but preferably also some time after,” says Linda Pipkorn.
As a part of the L3 pilot project, PhD student Linda Pipkorn carried out a study on a public road in Gothenburg (E6) together with Volvo Cars, in Gothenburg (E6) aiming to find out how the drivers’ gaze behavior changed when going from driving with automation to driving manually again.
“It turned out that, paradoxically, a take-over request, i.e., the signal from the car that the driver needs to take control can contribute to the drivers looking away from the road rather than looking at the road, which from a traffic safety point of view is not optimal,” Linda explains.
Her work received the Honda Outstanding Student Paper Award at the 2021 Driving Assessment Conference, an achievement that Linda herself believes can be explained by the project’s unique design:
“I believe that an important factor is that our results are based on data collected on public roads, with a real car and a realistic human-machine interface, which is relatively rare in our research area as tests in a simulated environment are more common. Data collected in a realistic environment is important to be able to draw conclusions that are in line with how the systems will be used in real scenarios in the future,” says Linda Pipkorn.
L3Pilot is now believed to pave the way for scaled-up driving tests with automated series vehicles in real-life traffic. Together with 40 partners – OEM:s, automotive suppliers, research institutes, traffic engineering and deployment companies – Chalmers researchers have already started working on the project Hi-Drive with the main objective to extend the data collection across EU borders in variable traffic, weather and visibility conditions.
L3Pilot is an Innovation Action, co-funded by the European Union under the Horizon 2020 programme with the contract number 723051.
34 organizations have committed to scientifically test and assess the impact of automated driving systems on driver comfort, safety and traffic efficiency as part of the project.
Duration: 50 months, 1 September 2017 – 31 October 2021
Total cost: €68 million
EC contribution: €36 million
Coordinator: Volkswagen AG
- Automotive manufacturers: Volkswagen AG, AUDI AG, BMW Group, Stellantis | Centro Ricerche Fiat SCPA, Ford, Honda R&D Europe, Jaguar Land Rover, Mercedes-Benz AG, Adam Opel AG, Stellantis, Renault, Toyota Motor Europe, Volvo Car Corporation
- Suppliers: Aptiv, FEV GmbH, Veoneer Sweden
- Research: German Aerospace Center DLR; ika RWTH Aachen University; VTT Technical Research Centre of Finland; SAFER Vehicle and Traffic Safety Centre at Chalmers University of Technology; SNF – Centre for Applied Research at NHH; University of Leeds; Institute of Communication and Computer Systems ICCS; Würzburg Institute for Traffic Sciences WIVW; University of Genoa; TNO – Netherlands Organisation for Applied Scientific Research; WMG, University of Warwick; European Center for Information and Communication Technologies – EICT GmbH
- Authorities: Federal Highway Research Institute BASt; The Netherlands Vehicle Authority RDW User
- Groups: Federation Internationale de l’Automobile FIA Insurers: AZT Automotive GmbH, Swiss Reinsurance Company SMEs: ADAS Management Consulting
Text: Lovisa Håkansson