Drivers new to level 2 driving assistance had lower awareness of on-road events

I guess that's not true, it makes perfect sense that regulators don't do anything anymore in this crony capitalist world, but I guess it's disappointing.

The legal system needs to blame someone. If the system fails and it kills someone, whos fault is it? Currently its most likely the drivers fault because its their job to ensure everything is ok and brake manually if they see something.

But I don't know if humans can handle that well. It's safer to brake instead of swerve if an animal darts into the road. Of course a kid darting into the road is more serious and traumatic, but the typical driver may not be able to avoid a crash.

That's certainly my experience. I love L2 for night driving, it lets me spend more time scanning the ditches for animals.

But you are going to use L2 driving a highway or even just decently paved road that happens to be going through a forest.

Someone who uses this technology because they dislike driving will always pose a danger. Those who enjoy driving will be more diligent.

"Adaptive cruise" is quite a nice thing.

And this is on the front page of HN because that headline reinforces the common worldview here: that partial self-driving is more dangerous than none at all, and full autonomy is further away than manufacturers like to suggest.

"Drivers who missed every instance of the bear spent more time looking at various parts of the car's dashboard than any other group."

I hypothesize that, while some of that was surely trying to navigate a radio or air-conditioning setup they were unfamiliar with, much was likely spent monitoring the speed of the vehicle.

An simple HUD illumination reflection on the wind shield for the speed might help as a safety measure. Or possibly changing laws and enforcement so that police were focused on non-revenue generating actives and mostly about curbing any behavior that harms the safety of a steady and orderly traffic flow.

Note that the article also mentions that other studies have found increasing complacency with experience.

> Does it make them more likely to note and deal with unexpected risks or not? The article guesses the answer though the actual paper may not have.

Given that noticing a problem is a prerequisite of dealing with it, this is not an unreasonable supposition, though one should also wonder if their reaction times are slower. Note that even if that is so, noticing more things, or noticing them sooner, could more than compensate for this.

In my experience the software developer in me looks at every odd action or lack thereof in the frame of, why this and not that and how do I think it should have performed.

Now there is a big claim out there we will all be able to opt into the new FSD beta and I truly want to give it a whirl.

We recently rolled out capacitive steering wheels in several Mercedes-Benz models for that reason. It's pleasant.

Why is this a problem? My car is at a complete stop and is not legally allowed to proceed. What's the threat? Doing things while moving (even creeping along in stop/go traffic), sure, I'll agree, but stopped at a light?

Morons will use devices that clamp to the steering wheel to defeat this sensor. Actively turning the wheel is harder to spoof.

edit: looks like we went through this process already.

https://www.theverge.com/2018/6/19/17479316/tesla-autopilot-...

From the article:

“ ‘Our data suggest that Level 2 driving automation has the potential to improve a driver’s [situational awareness] once he or she is familiar with the technology, although it does not guarantee it,’ the report reads.

Per the IIHS study's results, nearly all of the experienced users of Level 2 systems were able to notice the bear during their drive, and more drivers in this group were able to correctly note the number of times the bear drove past them than anyone else. Meanwhile, over twice as many of the drivers who weren't experienced with Level 2 automation but used it for the test didn't remember the bear at all compared to either of the other groups in the test.”

One might conclude that we need to reduce this “learning period” with formal driver training for these systems, perhaps the way we certify people for drivers licenses or something.

No, that's not at all what the article says! It says the opposite

>"Our data suggest that Level 2 driving automation has the potential to improve a driver’s [situational awareness] once he or she is familiar with the technology, although it does not guarantee it," the report reads.

Beyond that, even if nobody noticed the teddy bear, there were no controls who were driving with no assistance features, so you wouldn't be able to make that conclusion anyways.

I've got an electric car with level 2 assistance since last year and although I didn't drive as much as expected due to Covid-19 and lockdowns I feel how driving has changed.

At first I was very "afraid" of the systems doing stupid things and probably killing me while doing that. I was very focussed on controlling the slight steering it does when getting near white lines (Germany here) or how it would brake when following another car.

As that fear passed, I became much more concious of my surroundings, paying more attention to what other drivers might do (especially on the Autobahn) and how I could improve my "driving style" to create a better flow of traffic.

Can't wait to experience higher levels of assistance systems. I'd LOVE to get into the car, position myself behind a truck with 90 km/h on the right lane and let the car do it's thing while I'm not watching what it does.

I think a better design than an audible alarm is to show the driver using quadrants or a diagram when an object is within the "close," "very close," or "near-collision" zone around a vehicle and beep once but only when the object passes into another zone to alert the driver to a change in state. The tones should change for each transition so you can learn the tone for "object moved from close to very close in right driver's-side zone" and distinguish that from "object moved from very close to close in right driver's-side zone."

Having to constantly pay attention to an alarm tone is extremely nerve-wracking. In medical devices this will often result in "alarm fatigue" where you spend so much time dealing with false alarms that you miss a real alarm.

https://medicalxpress.com/news/2011-04-gorilla-dont-front-ey...

Yeah, but planes aren't generally any better at this either. In almost everything except jet aircraft and especially jet airliners, autopilot can be as crude as "maintain this heading and altitude"; i.e. you still need to manage your speed (the engine throttle lever) yourself, including in response to changes of wind direction or intensity.

Then we get to airliners, which have flight management computers. You can give your flight plan to that, and they will even calculate optimal speeds along the way, which the autothrottle can maintain for you. This allows you to, for example, know when the optimum descent point from cruise is; too early and you need more fuel to maintain altitude when you get down to where you need to be, too late and you need to turn more to bleed off more altitude.

The more advanced flight management computers are also loaded with a complete navigation database that includes e.g. speed and altitude restrictions at each waypoint, so they know, for example, that you shouldn't descend below 12,000 feet or be doing more than 260 KIAS at <waypoint x>. This is basically what's on your charts, but it's still up to you to verify that you're complying with the chart, even if you let the flight computer handle it.

Then we get to the autopilot itself. The autopilot on a modern airliner doesn't mean the plane can "fly itself". They still need to be told what to do, and told correctly. If you take off from Paris and head south-east and tell it to maintain 8,000 feet, it /will/ fly you straight into the side of the Alps. The Ground Proximity Warning System will tell you that you're in imminent danger of smashing into a mountain, but the autopilot will not react to that. If you tell the autopilot to descend into the ground it will. If you tell the autothrottle to fly too fast you will damage your aircraft. If you're doing 280 KIAS in a 777 and you throw out flaps, the autothrottle will not slow the aircraft down, and you will damage them. If the autopilot is about to fly straight into another aircraft, the Traffic Collision Avoidance System will tell you that, but the autopilot will not react to that either.

Note that some of this doesn't apply to Airbus aircraft, as Airbus thinks they know better. Whether that's a good thing is up for debate; it did result in, for example, Sullenberger not being able to make an optimal ditching approach into the Hudson, because the Airbus thought his nose-up input would stall the aircraft (which it would have eventually, but he was only a few feet above the river anyway, so that's inconsequential). Then there was Air France 447, where the usually fastidious Airbus flight management computer would usually prevent the pilots from making input that would cause a stall, but because their speed sensors were not working, the computer didn't do that. The first officer held that plane in a constant stall for their entire 38,000 feet descent into the ocean, probably because they were relying on the system to not let them do that. The ground proximity stuff also doesn't apply to some fighter jets, which won't let a pilot aim the aircraft at the ground (because it could be because the pilot has passed out during a high-G manoeuver). Still, the vast majority of it applies.

The point is, these systems require the same kind of checks, instruction, and constant supervision that Level 2 autonomous driving systems do.