Where is the evidence that replacing human drivers with robo-drivers advances safety on public roadways?
According to press reports, Volkswagen Group CEO Herbert Diess expects autonomous vehicles (AVs) to be ready for sale “between 2025 and 2030.” I read that headline and thought “Not again, haven’t we been here before?” To paraphrase Groucho Marx: “I have forecasts. If you don’t like them, I have others.”
Enough already. It’s time for automakers to get serious about safety for human-driven vehicles. Behind the hoopla, the speculation and what could politely be described as Tesla fan fiction, where is the evidence that replacing human drivers with robo-drivers advances safety on public roadways? Just show me definitive proof and I will go away.
I know, human drivers are responsible for more than a million traffic-related deaths globally each year. That is a lot of lives lost, with many more seriously injured. But humans collectively drive a mind-boggling distance, with many estimates of vehicle miles traveled on public roadways totaling over ten trillion per year. To put that number in context, Alpha Centauri, the nearest star system, is about 25 trillion miles away; the sun a mere 93 million.
So, I’m all done listening to unsubstantiated claims. I want facts backed up with hard data. I prepared this slide with some numbers from the World Health Organization (WHO), which I found very informative and I don’t think I’ve ever seen shared in public before.
When was the last time you heard an automaker or AV tech company talk about saving lives in India? How about Brazil? Or Nigeria? EE Times has summarized AV developments in China and highlighted China’s progress towards the deployment of vision-based driver monitoring systems (DMS).
Yes, China has a serious road safety problem and is tackling it with a two-pronged approach covering future technology (AVs) and existing proven safety technology (driver monitoring and assistance). Europe is following a similar strategy and at the end of 2019 the European Parliament updated the General Safety Regulation (GSR) for the type-approval requirements for motor vehicles.
The forthcoming changes to the GSR are summarized in this excellent graphic from Continental. The legislation is all for ADAS and driver monitoring, not AVs or “self-driving”.
Hoopla for me, foot-dragging for thee
AVs may offer the possibility to save lives sometime in the future, but people are dying on public roadways today and many automakers appear more interested in cheap headlines and vacuous forecasts than proactively installing proven safety technology. This is a point made recently by the Advocates for Auto and Highway Safety during the publication of the AV Tenets.
All this foot-dragging isn’t about a lack of money. Automakers have thrown many billions of dollars at the development of experimental AV technology, including: GM acquiring Cruise Automation, with a follow-on investment from Honda; Ford and Volkswageninvesting in Argo AI; Toyota investing in Uber ; and Hyundai investing in the Motional joint-venture with Aptiv.
Hyundai wasted no time in boasting that it will use Nvidia’s DRIVE platform for every Hyundai, Kia and Genesis model’s in-vehicle infotainment system, but has made no such commitment to the universal adoption of critical safety technology such as automatic emergency braking, lane departure warning and DMS. Hyundai’s actions make clear that infotainment is a higher priority than safety.
Similarly, Mercedes-Benz saw the greatest urgency in telling the world that it is working with Nvidia to build the “most advanced software-defined vehicles” while Volvo – previously the doyen of safety – was most concerned with announcing a Level 4 partnership with Waymo. As is always true in life, watch what is done as much as what is said.
In my opinion, the automaker that can best lay claim to the safety crown is BMW, with its proactive stance towards DMS. While GM received a lot of positive press coverage for its early adoption of vision-based DMS in Cadillac Super Cruise, it squandered its advantage by chasing headlines and focusing resources on the Cruise Origin AV.
Meanwhile BMW quietly moved into the clear leadership position for the adoption of advanced vision-based DMS, first with technology from Smart Eye and then Seeing Machines. Unlike just about every other automaker, BMW did not have to have its feet held to the fire by bodies such as Euro NCAP (New Car Assessment Program) before installing DMS and was quicker than most to understand the safety benefits of eye tracking and human factors science.
Someday my DMS test protocols will come
EE Times first wrote about Euro NCAP developing DMS test protocols in April 2018. Since then, nothing much has happened publicly and for DMS watchers like myself, these have been barren years. I have no idea when Euro NCAP will finally publish its DMS test protocols, but I would guess at some time in the next three months if the 2024 target is to be met.
While BMW has emerged as the automaker taking DMS most seriously, almost every other automaker looks to be following the nickel-and-dime route to safety and is waiting to see what is the cheapest possible DMS implementation that will still fulfill Euro NCAP’s requirements.
Based on my research, I would estimate that the cost difference between pointless tick-the-box vision-based DMS and how-is-that-even-possible-state-of-the-art DMS is probably in the order of about ten bucks per vehicle. That’s what a bunch of penny-pinching skinflints these automakers are – just keep that in mind the next time you read about an automaker throwing another couple of billion dollars at AVs.
Here are some of my thoughts for what Euro NCAP might be considering for DMS test protocols and some possible timelines for introduction.
Necessity of eye-gaze tracking
The facial parameters most often measured by a vision-based DMS for “raw data” gathering encompass head pose (which includes yaw, pitch and roll), mouth and eyelid opening, blinks (including velocity, duration and frequency), and pupil dilation (pupillometry).
However, the critical measurement is eye tracking, specifically measuring changes in the position, velocity and acceleration of the eye-gaze. Without dedicated test protocols for the accuracy and reliability of eye tracking, almost all of the safety benefits of vision-based DMS are sacrificed.
Driver state analysis
Building on “raw data” gathering is the creation of reliable, higher-level signals for driver state analysis, encompassing visual distraction, cognitive distraction, drowsiness, intoxication and impairment, among others.
Human factors science and behavioral research are critical components to the creation of a real-time assessment of how attentive the driver is to the road ahead and how engaged they are in the task of driving. The goal of driver state analysis is to deliver a reliable, real-time understanding of the driver’s cognitive state as it applies to accident risk.
The Euro NCAP protocol document “Assisted Driving – Highway Assist Systems v1.0” from October 2020 states on page 14:
Direct Driver Monitoring systems that monitor driver engagement and cognitive workload using cameras and/or other sensors to check that the driver has “eyes-on” and / or “brain-on.” Euro NCAP is developing test and assessment procedures for these Driver Monitoring systems.
Driver state analysis is thus highly likely to be part of Euro NCAP’s future roadmap for DMS and I would estimate possibly from 2026.
Integration with ADAS
An obvious way to improve road safety and to make human drivers into safer drivers would be to integrate the high-level DMS signals for driver state analysis with ADAS, to modify the longitudinal and lateral responses of the vehicle. The sensitivity and performance of the automatic emergency braking and lane-keep assistance systems would be varied in real time, according to observations of the driver’s attention state and engagement level.
The Euro NCAP 2025 Roadmap “In Pursuit of Vision Zero” states on page 7:
Euro NCAP envisages an incentive for driver monitoring systems that effectively detect impaired and distracted driving and give appropriate warning and take effective action e.g. initiating a safe evasive manoeuvre, limp home mode, increased increasing sensitivity of Electronic Stability Control, lane support, speed, etc. The assessment will evolve around how reliably and accurately the status of the driver is detected and what action the vehicle takes based on the information.
Integration with ADAS is thus highly likely to also form part of Euro NCAP’s future roadmap for DMS and I would estimate possibly from 2028.
Clap for NCAP
In an ideal world, all automakers would voluntarily install proven safety technology into every vehicle manufactured to make human drivers into safer drivers. Sadly, it appears that Euro NCAP – and by association the other regional NCAP bodies around the world – must continue to push for safety to be at the top of the automakers’ priority list.
After years of discussions, it is to be hoped that when the DMS test protocols are finally published, that the performance bar will have been set high. After all, how many lives does Euro NCAP want to save? Perhaps “In Pursuit of Vision Zero” gives us a heads-up of what is to come.