How SAFE-UP research supports Euro NCAP’s future vehicle safety assessment
In an interview with the SAFE-UP project, Adriano Palao Bernal, technical manager for ADAS and automated driving at Euro NCAP, highlights the programme’s vision for future vehicle safety in Europe and how SAFE-UP’s innovations support their work.
Q: In which way do SAFE-UP’s outcomes relate to the 2030 Euro NCAP roadmap?
A: SAFE-UP’s project deliverables are promising. To achieve the goal of shifting towards scenario-based assessment and tackling ADAS (Advanced driving assistance systems) robustness in real-life situations, it is essential for Euro NCAP to rely on solid facts and data, ensuring we understand aspects such as:
a. Accidentology: addressing relevant scenarios based on magnitude (number of killed or severely injured) and selecting the appropriate parameters after understanding the most common accident types (e.g., delta speed, overlap/impact point, occlusions…).
b. Testability: implications of introducing additional variables in a scenario, and the feasibility of harmonising the procedure across labs.
c. Awarding real performance: fact-based understanding of which variations are a challenge to ADAS perception in the real-world – eventually closing the gap between true positive ratio in current track scenarios (which is high on average) and real-world scenarios (low on average). This will help to differentiate the truly robust systems from the ones that are not so robust.
d. Assessment: as we move towards a scenario grid which is exponentially growing in size, it is necessary to look at alternative verification methods, such as spot testing and virtual validation. For the latter, we need to understand to what extent we can trust the results.
Q: Which of the technologies being addressed in SAFE-UP do you think have the most potential?
A: Demo 2 on object detection in adverse weather conditions has generated a lot of interest at Euro NCAP. However, including it in the protocols and eventually putting in place a harmonised verification ecosystem still looks like a bridge too far. We first need to determine to what extent adverse weather conditions are actually a problem (or not) for ADAS perception, and then look into a business case for testability.
Q: Regarding V2X communication, will (safety-critical) scenarios be considered where connectivity has a key role in avoiding or warning the driver? Or will V2X only be considered as additional input for the driver?
A: Some of the high latency applications available on the market can already deliver a substantial safety benefit (e.g., local hazards such as traffic jam ahead, wrong-way driver…). These are clearly the low-hanging fruit, which is why they have already been rewarded in Euro NCAP’s assessment.
For the middle-range latency applications where high position accuracy is not crucial, what looks promising and feasible in the mid-term is to inform the driver or create awareness of a potential threat ahead.
Finally, the low latency applications are the ones that can enable trustworthy warnings and eventually system interventions, but these remain left for the long term.
Q: The V2X demonstrator being addressed in SAFE-UP does not cover some of the challenges of this technology, such as positioning accuracy and standard communication interfaces, topics which for Car-to-Car may have lower complexity. What is Euro NCAP’s expectation with this technology when it comes to VRU interaction?
A: As mentioned in my previous answer, some use cases involving VRU interaction can be enabled with middle-range latency applications where high position accuracy is not crucial, by informing the driver or providing increased awareness (e.g., a bicycle approaching an occluded intersection as a potential hazard ahead). Euro NCAP is therefore open to looking at those use cases as well.
Q: Euro NCAP protocols have a long-standing history with EU projects (ASpECCS – 2014, PRosPECT – 2020, SAFE-UP – 20XX). What does Euro NCAP consider as the next big challenge for VRU scenarios that requires further research?
A: As mentioned in point c) of question 1 above, accident data suggests that there seems to be a substantial gap between the true positive ratio we see in the cars we test and, for example, the AEB not being triggered in these accidents. That’s why we need to understand which aspects are affecting the ADAS perception across the board. We will need to deep dive into how variability in environment, scenery, vehicle motion before the collision, etc. affects perception performance.
Q: For safety applications, do you think that cloud-based systems are enough to inform the driver about possible hazards ahead? Or do you think it’s necessary to include V2V systems with direct communications to trigger AEB systems?
a. In the case of V2V direct communications for AEB systems, how do you think cybersecurity and Functional Safety (FuSa) can be addressed?
A: Indeed, see questions 3 and 4. We see enormous potential for dynamic information enabled by crowdsourced data – mainly for Speed Assist Systems, but valuable for collision avoidance as well.
Euro NCAP has recently started exploring how to address the cybersecurity and data access fields, so it is too early at this stage for us to make a statement here. As for FuSa, it is for now out of Euro NCAP’s scope as we understand that ISO compliance is the OEM’s ultimate responsibility. Anecdotally, during the recent roadmap consultation, we learned that the OEMs reluctant to adopt V2X are indeed rejecting the technology, as they claim it’s impossible to comply with FuSa, as they do with the vehicle sensors.
Q: When do you foresee eScooters being included in the rating? Do you think there will be any technological issues for vehicle manufacturers with this new requirement?
A: As soon as Euro NCAP understands and confirms the magnitude and typology of accidents involving eScooters, it should be a straightforward addition to the protocol, as we could make use of (relevant) existing scenarios involving other targets (e.g., a bicycle), eventually adjusting the parameters like target position, impact point, speed profile, etc.
Q: In the 2030 road map: how are you approaching AD systems? What will be considered? Is there a specific timeline for ‘Automated Driving (L3/L4)’ being addressed by Euro NCAP? What are the foreseen challenges?
A: For a few years, we have run an assessment on Assisted Driving, with a scope limited to the highway driving domain. Next, we are looking into an evolution of this assessment by expanding the driving domain to urban and/or inter-urban. Following a scenario-based approach, we will have to include related elements associated with these driving domains, like VRUs travelling longitudinally, and an assessment of the vehicle strategies to tackle these.
The upcoming DCAS regulation “hands-off” assisted driving could be a game changer and Euro NCAP will have to keep a close eye on the developments of such technologies, eventually developing a framework to assess their safe implementation.
All in all, regarding Automated Driving, this is an area that Euro NCAP is targeting for the mid/long-term, and the working group is set to kick off later this year.
Q: What will be the approach for Driver engagement (Human Factors) in advanced levels of AV?
A: As mentioned above, more advanced levels of AD are being targeted for the mid/long term. What I can say is that one of the pillars of Euro NCAP’s Assisted Driving assessment is “Assistance Competence”, which includes provisions to evaluate the level of driver engagement. We are looking into possibilities to assess this aspect in a more practical way.