Experts say driver assistance systems are doing more for safety than self-driving tech

[zohaibahd]

In context: Traffic accidents are an unfortunate reality on America's roads. The National Highway Traffic Safety Administration estimates a staggering 41,000 fatalities occurred in 2023. Autonomous car makers have promised a future where self-driving vehicles virtually eliminate traffic-related deaths, but things are never that straightforward.

InsideEV spoke to experts who insist that the solution to improving road safety already exists in advanced driver assistance systems, such as automatic emergency braking, lane-keeping assistance, and blind spot monitoring. Data from the Insurance Institute for Highway Safety (IIHS) supports the assertion. Its recent study shows that crashes under these scenarios have declined in vehicles equipped with driver assistance technology.

The report cites automatic emergency braking (AEB) as an example. This system first uses radar and cameras to scan the road ahead. It then checks whether the driver reacts to potential hazards. If not, AEB automatically brakes, preventing a collision. The study concluded that AEB lowered rear-end crashes by a significant 50 percent.

Senior Researcher David Kidd explained that allowing more autonomy, such as that offered by Tesla's Autopilot and GM's Super Cruise, can have the opposite effect. These systems lull drivers into a sense of false security, leading them to become dangerously distracted and complacent over time.

While Tesla touts its miles-per-accident stats to highlight Autopilot's safety, high-profile fatal accidents involving self-driving cars have continued to occur, triggering federal probes. In one extreme incident, a Cruise driverless taxi ran over and stopped on a pedestrian hit by another vehicle. This accident prompted the California DMV to suspend GM Cruise's permit.

"We're not seeing any change in terms of safety," Kidd claims. "[The] only benefits or any potential benefits are convenience and comfort."

Meanwhile, Euro NCAP Technical Director Richard Schram says manufacturers selling self-driving tech in their cars hide the truth regarding their vehicle's capabilities.

However, completely dismissing automation may not be a wise move either. Most of the technology is relatively new. Like any other system, it requires clocking many more miles before coming to a logical conclusion on its safety impact.

Complicating matters is that the data at hand is confusing and contradictory. While the National Highway Traffic Safety Administration (NHTSA) acknowledges distracted driving risks, it hasn't definitively linked Level 2 systems to increased crashes.

Similarly, a recent Nature study suggests self-driving cars generally have lower accident rates than humans, except in certain conditions like dawn/dusk or when making turns. However, an separate analysis of NHTSA data found that crashes involving Level 2-equipped vehicles are rising. So, there doesn't seem to be any consensus on the safety of AVs.

Despite the debate around self-driving safety, the NHTSA is moving to make systems like AEB, which have a solid record of reducing accidents, mandatory for all non-commercial vehicles by 2029.

Permalink to story:

Experts say driver assistance systems are doing more for safety than self-driving tech

 

[bviktor]

Agreed. Bought a 2020 BMW 330e a couple of weeks ago, and driving on the motorway is already PURE BREEZE. I turn CC on, and it activates all the other stuff. From then it stays within lane, and accelerates / decelerates just the right amount when something happens in front of me. All I do is hold the steering wheel in my hands, and not much else.

And this is a 4 years old model that already misses not one but 2 facelifts (and thus software and tech upgrades as well) since then, without any of BMW's advanced self-driving cr@p from the 5 series, it only has some very basic assistance features like lane keeping, adaptive CC, and blind spot monitor. Yeah, it also also has that thing to figure out speed limits from road signs and/or the built-in satnav, but it's still very much a hit-or-miss endeavor unfortunately.

The city is another story though, obviously... that'll probably take decades, no matter what these big corps say. And I say that as someone who works for the self-driving subsidiary of one of the biggest carmakers. Don't hold your breath for some big breakthrough - it'll be more like a million iterative improvements.

 

[Theinsanegamer]

Driving is an extremely complex task full of random inputs and variables, EG the worst case scenario for a computer to take control. We only think of it as easy because humans are a lot more capable then people think they are.

 

[bviktor]

Driving is an extremely complex task full of random inputs and variables, EG the worst case scenario for a computer to take control. We only think of it as easy because humans are a lot more capable then people think they are.

More like it's much less well-defined and formalized than we think. So it's extremely difficult to develop algorithms for it.

 

[Dr Roboto]

If they don't try, they will never be able to make it work. I think that all of the real world testing they are doing is great. It is what is needed to push the tech. Sure it will still take decades to perfect, but you got to start somewhere. I do wish it was much more regulated though. They sure do seem to get by with a lot of shady stuff that is pretty dangerous.

However, what I don't like are the <bleeping bleeps> in the marketing department and the CEOs (yeah, I am talking about you Musk) that pushes this stuff like it will be ready tomorrow and it will change the world. Maybe that BS is needed to raise the money from VCs to pay for it all. Doesn't mean I have to like it.

Now that I read my own writing, you could pretty much say this about any big tech "revolutions": bitcoin, AI, etc.

 

[DanteSmith]

I can buy people to say anything, including creating studies to corroborate it.
it doesn't mean anything. The fact that it is so sparsely utilized, coupled with the occurrence of accidents everywhere, is more than enough to raise suspicion flags

 

[Theinsanegamer]

If they don't try, they will never be able to make it work. I think that all of the real world testing they are doing is great. It is what is needed to push the tech. Sure it will still take decades to perfect, but you got to start somewhere. I do wish it was much more regulated though. They sure do seem to get by with a lot of shady stuff that is pretty dangerous.

However, what I don't like are the <bleeping bleeps> in the marketing department and the CEOs (yeah, I am talking about you Musk) that pushes this stuff like it will be ready tomorrow and it will change the world. Maybe that BS is needed to raise the money from VCs to pay for it all. Doesn't mean I have to like it.

Now that I read my own writing, you could pretty much say this about any big tech "revolutions": bitcoin, AI, etc.

They've been trying for two decades already. The DARPA autonomous car rally was back in 2005. No matter how much money or tech you throw at a problem, there are two problems you cannot overcome, one is physics, and the other is on the fly thinking. The first is gonna block your sensors, your cameras, your LIDAR, whatever you use, and the second is something you'll never fir into a car brain.

And seriously you guys need to get an Elon body pillow and work through your emotions already. It;s getting embarrassing.

 

[yRaz]

They've been trying for two decades already. The DARPA autonomous car rally was back in 2005. No matter how much money or tech you throw at a problem, there are two problems you cannot overcome, one is physics, and the other is on the fly thinking. The first is gonna block your sensors, your cameras, your LIDAR, whatever you use, and the second is something you'll never fir into a car brain.

And seriously you guys need to get an Elon body pillow and work through your emotions already. It;s getting embarrassing.

The Musk issue is something I find hilarious. He was once the heralded by the left as a futurist. The left got too powerful and started having crazy ideas and Musk called them out on it. Now Musk is "intolerant" and "arrogant". While I can agree with him being arrogant, I really love his "try to cancel me, my net work is greater than the GDP of most countries" attitude. In a recent thread I guess that makes me a right winger despite me criticizing the right just as much as the left.

 

[trparky]

In a recent thread I guess that makes me a right winger despite me criticizing the right just as much as the left.

No, that just makes you a sane person. I have on multiple occasions called out both sides on their bullsh*t but in some circles on the Internet, but that's not good enough. Reddit is a perfect example of that.

 

[noblody]

They've been trying for two decades already. The DARPA autonomous car rally was back in 2005. No matter how much money or tech you throw at a problem, there are two problems you cannot overcome, one is physics, and the other is on the fly thinking. The first is gonna block your sensors, your cameras, your LIDAR, whatever you use, and the second is something you'll never fir into a car brain.

And seriously you guys need to get an Elon body pillow and work through your emotions already. It;s getting embarrassing.

I don't think there is a sensor issue. People mostly use their sight when driving, so the obstruction of the sensors is more of an issue for human drivers. An auto driver has more cameras, more perspectives and will ultimately face less obstruction. Also it can use sensors that use wavelengths that are not blocked by fog. The current safety mechanisms proved that they can enhance human driving. I believe the problem is with the processing of the data and that could be solved at some point.

 

[wezunblai]

"experts"

 

[p51d007]

If you have to "depend" or "rely" on some computer to do your driving, you have no business operating a vehicle.
Too many distracting do-dads in vehicles today as it is. Especially all of that touch screen nonsense.
Buttons, knobs, dials are safer because you don't have to take your eyes off the road to adjust the stereo, cooling, heating etc. Heck, I still prefer a manual transmission over automatic, which is getting harder to find. I like it more in "snow" weather because you can start out in 2nd and reduce the spinning of the tires.

 

[Northy]

I'm looking forward to when this technology finally matures. I'll happily sit in the back watching a film or reading a book whilst the machine takes me where I want to go. Sounds epic.