Autonomous driving tech falters in latest AAA safety tests

[Shawn Knight]

In a nutshell: AAA recently conducted tests involving three vehicles equipped with an active driving assistance (Level 2) system: a 2021 Hyundai Santa Fe, a 2021 Subaru Forester and a 2020 Tesla Model 3, and the results aren't very encouraging as it relates to safety.

Each vehicle was able to detect a slow-moving dummy vehicle traveling in the same direction as the test vehicle. In each test (five runs per vehicle), the test vehicle applied the brakes and matched the speed of the stand-in to avoid an impact.

Similarly, all test vehicles were able to detect a simulated cyclist traveling in the same direction and slow their speed to prevent an impact.

When the dummy vehicle was placed in the same lane to simulate a head-on collision resulting from an impaired or distracted driver, the test vehicles impacted it each and every time.

The speed of the test vehicle and simulated oncoming vehicle was 25 mph and 15 mph, respectively – lower than what might occur in the real world. The Hyundai and Subaru made no attempts to apply the brakes and hit the target without reducing speed. The Tesla did apply the brakes in every test, but still managed to impact the dummy vehicle at an average speed of 2.3 mph.

What's more, the Subaru Forester failed to detect a simulated cyclist crossing the street in front of it. The Tesla and Hyundai were both able to brake to avoid impact in this test.

Greg Brannon, director of AAA's automotive engineering, said their testing demonstrates spotty performance is the norm rather than the exception.

AAA's advice to automakers is simple: listen to consumers and improve currently available systems before trying to focus on the future. "You can't sell consumers on the future if they don't trust the present," Brannon added.

Tesla did not reply to a request for comment from Reuters. A rep for Hyundai said they are "reviewing the findings in AAA's report as part of our ongoing commitment to customer safety." Subaru told the publication they are looking into the test to better understand AAA's methodology but didn't have a detailed response at this time.

Image credit: Riccardo

Permalink to story.

https://www.techspot.com/news/94572-autonomous-driving-tech-falters-latest-aaa-safety-tests.html

 

[hahahanoobs]

I just hope people that "fear" Windows 11 and updates aren't FSD beta testers!
Sorry, I had to!

FSD has a very long way to go. Save your money and upgrade your PC instead.

 

[Impudicus]

Human controlled vehicles are not very safe either.

 

[scavengerspc]

Tesla doesn't deserve customers. And the other two sucks at it, like they have at most things lately.

Not to mention that buying a car and letting it drive itself is like cooking a T-bone and having it blended into a smoothie.

 

[RudyBob]

I used to think it would never happen but now it should happen because the number of dufus drivers expands everyday. They are too busy doing something other than driving

 

[VitalyT]

Strike failed tests as unlucky. I'm feeling adventurous, could kick some puppies.

 

[PurpleYoda]

Bottom line is, we still have a few more years before Skynet takes over! Close but no cigar so lets celebrate! :-D We know what’s coming so we might just as well enjoy it while it lasts.. now, where did I park my bike..?

 

[Uncle Al]

From the results it appears that all of these cars will be great in a demolition derby, but not so much on the highways ......

 

[Mr Majestyk]

Not surprised, the complete BS hype over how quickly autonomous vehicles would be ready for prime time is a joke. The problems are enormous and just throwing more sensors and cameras at the problem is not the way to go.

 

[wiyosaya]

Sooner or later, probably the latter, they will figure it out. For now, IMO, its the race to market that is damning the technology.

Human controlled vehicles are not very safe either.

Absolutely! And that is the main reason that research and development should continue, IMO.

 

[CurrField]

These are human driven cars with assists, which is very different to a fully autonomous vehicle.
Driver assistance systems only intervene when they are 100% sure they have to, because you don't want your car to make emergency stops for no reason. This makes the assistence systems hesitent to intervene, and therefor also not good in different situations. Going from driver assistence to autonomous is going from taking action once every 100.000km to taking action 1000x per second, it is completely different. Then it is not about taking action only when it is 100% sure, but about fully driving a car in the best way possible. Hope this helps clear up that driver assists are not representative of autonomous vehicles, like this article suggests.

 

[AMN3S1AC]

So, if we eliminate the human factor, we get less accidents?

 

[godrilla]

So, if we eliminate the human factor, we get less accidents?

If a human gets into an accident the insurance is sued, what happens if the ai gets into an accident does the plaintiff go after all parties including the manufacturer? Can all the sensors, radars, cameras, algorithms prevent any blind spot anomalies?
I personally believe that a hybrid approach will be best where the ai is used as a tool to improve an end user ( human) to mitigate accidents.
I drive a q7 with radar front detection and when using cruse control the ai hugs the road better than I can personally at certain situations, but other times its cumbersome like when switching lanes and this is why I believe that hybrid approach to autonomous driving is best for the end commuter in terms of safety.

 

[dukemeister]

Another fake article. These vehicles *are not autonomous*. The driver must avoid the head on collision. Duh.

 

[godrilla]

Another fake article. These vehicles *are not autonomous*. The driver must avoid the head on collision. Duh.

Most Ai already brake to mitigate impact my q7 in cruse control will literally brake all the way to zero and accelerate to whatever set point the user wants and you can set the distance between vehicles. The q7 ai is considered infancy in comparison to Teslas fully autonomous driving. Also my the q7 will tighten the seat belts to mitigate damage from impact. I believe the BMW version is slightly more advanced than Audi's. The question is would you trust an ai at this stage with your life? Also not necessarily the car you are driving it could be in front of you on the side or driving behind you. Can an ai adjust for Reckless speedsters that swerve between lanes at a mm precision?

 

[Dd663]

I look forward to the day when the majority of vehicles are autonomous, because everyone will be a lot safer that way.

Unfortunately, getting there is going to be hard. One of the problems is perhaps inevitable: when some fatal accident occurs involving a fully autonomous vehicle, news organizations will jump on it and spread the story far and wide for those sweet, sweet clicks. Then average Joe will read about it and conclude that autonomous vehicles are unsafe, forgetting about the hundreds of fatal accidents involving all human drivers that have occurred in the same time frame.

It won't be about having 100% safety, with zero fatal accidents, because that's impossible. But if the number of fatal accidents decreases because of autonomous vehicles, then it's worth it.

Humans get angry, sleepy, drunk, careless, and distracted while driving, all things that significantly increase the likelihood of accidents. Autonomous vehicles can't get angry, sleepy, drunk, careless, or distracted.

 

[godrilla]

I look forward to the day when the majority of vehicles are autonomous, because everyone will be a lot safer that way.

Unfortunately, getting there is going to be hard. One of the problems is perhaps inevitable: when some fatal accident occurs involving a fully autonomous vehicle, news organizations will jump on it and spread the story far and wide for those sweet, sweet clicks. Then average Joe will read about it and conclude that autonomous vehicles are unsafe, forgetting about the hundreds of fatal accidents involving all human drivers that have occurred in the same time frame.

It won't be about having 100% safety, with zero fatal accidents, because that's impossible. But if the number of fatal accidents decreases because of autonomous vehicles, then it's worth it.

Humans get angry, sleepy, drunk, careless, and distracted while driving, all things that significantly increase the likelihood of accidents. Autonomous vehicles can't get angry, sleepy, drunk, careless, or distracted.

Some good points, but what about hacking vulnerability, bad weather conditions ( snow flooding, heavy rain, ) affecting visibility and sensors. Just because the problems didn't occur yet doesn't mean the summation of human intelligence can foresee every potential obstacle anomaly and patch an algorithm into a plateau ideal state of which you speak. It will inevitably take trail and error, hefty investment, time, Government policies, big insurance policy, market acceptance ( $12k premium for Tesla) etc etc to get to Ideal place.
That's why I am saying a hybrid approach seems more practical. The macroeconomics of the transportation business is probably somewhere in the multi trillions. While Autonomous vehicles will disrupt the transportation industry with new opportunities and also killing millions of jobs, I believe it will be decades before we see a significant impact in our daily lives.

 

[Dd663]

Some good points, but what about hacking vulnerability, bad weather conditions ( snow flooding, heavy rain, ) affecting visibility and sensors.

Indeed, malicious hacking, especially in a hypothetical centralized system that all cars are connected to, would be a concern. Security would have to be very robust. And yeah, bad weather would have to be accounted for as well, though, to be fair, it affects our visibility and sensors too!

I believe it will be decades before we see a significant impact in our daily lives.

Agreed, it's a long way off still.

 

[Hodor]

Until the AI concludes that running over pedestrians and killing the passengers actually reduces local CO2 footprint. You just set it to ECO mode and it will do the rest.