When ADAS Features Fail

June 8th, 2021

Comments Feed

I don’t quite know why I’ve become more attentive to learning about a car’s ability to protect its occupants in the event of a collision, along with its ability to avoid the collision altogether in the first place.  I expect it has a lot to do with having close family members who occasionally need to drive themselves places.  Advanced Driver-Assistance Systems (ADAS) are growing in popularity.  ADAS systems can help prevent accidents not only at speed, but also when parked as a stationary car.  ADAS features are designed with one purpose in mind and that is to increase driver and occupant safety.

ADAS features include things like automatic emergency braking, blind spot detection, collision warning systems, cross-traffic alert, forward and rear collision warning, lane departure warning, lane keeping assist, park assist, pedestrian detection and avoidance systems, cyclist detection and avoidance systems, road sign recognition, active radar cruise control… and the list goes on.  ADAS employs cameras and sensors to detect a potential collision or event and then proceed to activate systems of avoidance if necessary.  These are important safety features which help prevent accidents.

Research on insurance claims that was carried out by LexisNexis Risk Solutions showed that vehicles involved in incidents that had ADAS on-board exhibited a 27% reduction in the frequency of claims made for bodily injury.  The results also showed that vehicles that had ADAS on-board exhibited a 19% reduction in the frequency of claims made for property damage.  Obviously, this would suggest that the systems must be doing some good.

A study by the Insurance Institute for Highway Safety (IIHS) revealed that the crash involvement rate for vehicles with blind-spot monitoring was 14% lower than for the same vehicle without the equipment.  Researchers also stated that the study also suggested that if every vehicle sold in the US in 2015 was equipped with blind-spot monitoring, 50,000 crashes and 16,000 crash injuries might have been prevented.

At present, one of the big downsides of the ADAS features is that they are darn expensive.  Not only do they put the price of a new car up, they also make the car costlier to insure because if any of the systems gets damaged the insurance and repair bills are usually eye-watering.  Hopefully, ADAS features will come way down in price and become similar to standard computer software and technology which is, on the whole, a dime-a-dozen now.

The other thing is that I hope ADAS will function 100% of the time correctly as intended, because vehicles designed to be able to automatically brake for objects such as other cars, pedestrians, and cyclists, and to drive themselves inside highway lanes without driver input, is not an exact science.  A slightly frightening example of my concern here is when Volvo was demonstrating its pedestrian AEB technology to journalists in 2016.  Volvo used their V60 model in the demonstration, where it was travelling toward a dummy named Bob.  The V60 didn’t detect Bob being in the way, and so Bob was hit in what was a controlled environment.  An alert driver in the V60 may well have returned a better outcome.

Then shortly after, another Volvo V60 was demonstrating its collision detection and avoidance system where it was to avoid hitting a stationary truck.  The failure to detect and avoid the collision can be seen here: https://www.youtube.com/watch?v=aNi17YLnZpg

Again, an alert and competent driver could well have resulted in a better outcome, should this have happened in the real world.

In 2018, the IIHS took five new vehicles and tested them.  The Tesla Model 3, the Tesla Model S, the BMW 5 Series, the Mercedes E-Class and the Volvo S90 were the test vehicles.  Each vehicle’s AEB, adaptive cruise control and lane-keeping assist systems were tested.  Some of the problems IIHS encountered was that the AEB didn’t actually work in some vehicles in some circumstances.

In other tests, the IIHS observed: “The BMW 5-series steered toward or across the lane line regularly, requiring drivers to override the steering support to get it back on track.  Sometimes the car disengaged steering assistance on its own.  The car failed to stay in the lane on all 14 valid trials.  The Model S was also errant in the hill tests.”

Sadly, just a couple of years ago an autonomous Uber fitted with even more sensors than any standard ADAS-equipped road car killed a pedestrian at night in the US.  This happened while researchers and designers were conducting public testing.  What this suggests is that the ADAS technology is amazing and good enough to be placed into new cars.  However, it doesn’t mean ADAS will always work as intended, and it does point to the fact that drivers must still always be fully alert at the wheel.  If the driver is not fully alert, the outcome from these system fails can sometimes be way worse than if the driver was fractionally slower to manually override the systems detection time and action times.

I’ve heard of numerous occasions when vehicles have falsely detected situations.  A more common fail is when accident emergency braking (AEB) engaged on-board a car when it shouldn’t have, which meant that the AEB stopped the vehicle abruptly and unexpectedly on a clear road.  At the time, traffic is still coming up behind the vehicle.  Lane keep assist isn’t always that great either, and the results of a high-speed mishap on a main highway is tragic.

ANCAP is Australian’s big car-safety tester, and a recent representative suggested that AEB and lane-keeping assist technology, which is where the car will steer itself, was beginning to be put under the microscope.  This would test for how accurate the system actually is, and if it would actually do the opposite and steer the vehicles into a dangerous situation.  Testing ADAS features should take priority over just saying that the technology is available in the car at the time of crash testing, whereby the appropriate ADAS feature box is ticked and the job done.

ADAS mostly works for the better.  It does raise obvious safety problems, particularly when manufacturers have all the pressure to pack in as many ADAS features into their vehicles as possible for as little cost as possible to remain competitive on the sales front.  This pressure would suggest that these systems could be prone to potentially become unsafe.

With cars loaded with ADAS features, you could also say that drivers of these new vehicles might be tempted to hop on the mobile phone to check messages once they have activated the adaptive cruise control and lane-keep assist systems.  Essentially, it becomes easier to break the law; which takes us back to the point that we shouldn’t rely heavily on ADAS technology because it can fail to work.  We don’t often hear this preached at the car sales yard or on new-car adverts.

In Australia, features such as antilock brakes (ABS) and electronic stability control (ESC) are mandatory in new vehicles that are sold to the public.  These mandatory requirements are set to be pushed to the next level, where automatic emergency braking (AEB), adaptive cruise control and lane-keeping assist would have to be on-board any new vehicle being sold to the public.  Even alcohol detection devices may well be part of these standard requirements.  Europe is set to introduce some of these requirements over the next few years, and Australia is likely to follow the lead.  Newly imported European cars would end up with these features anyways, a win-win for us new-car buyers.

ADAS is good, but we still need to drive our cars.