Mercedes gets approval for turquoise automated driving lights
jalopnik.com
jalopnik.com
For red-green colourblind people (both types), this colour is at the neutral point where both available cones are saturated and the brain can't differentiate it from white light.
In their study paper they used a much more blue turquoise, decided it was confusing, and then moved it to a more green version (thus saturating the non-blue cone for dichromats) but did not reassess the new green-turquoise colour.
The net effect is that for something like 8% of men this light looks exactly the same as the white reversing lights and is placed exactly where reversing lights are.
to me the placement seems substantially different from the reversing light (which is usually more toward the middle of the back)
of course these pictures probably don't tell much about how it will actually look like in traffic, but some kind of pattern seems to be the way to go (that's recognizable from 2-3 cars afar in a traffic jam, for example if folks are curious WTF that particular car is doing, why is it stopped or going slow)
I don’t know if they have to do a test or something.
They should identify the color of a car or a person’s apparel fairly accurately .
How would a color blind cop catch say killer escaping in a teal car if they cannot identify that color ?
Most Physical disabilities sadly would be disqualifying in some jobs line first responders or police, such jobs that at times require them to assist the physically challenged
Imagine you’re in a parking lot and the car is in self driving. Is it in Reverse? I have no clue.
Also, aren’t all cars now required to have front on driving lights in most jurisdictions, separate to headlights?
If so, that's a good precedent for other jurisdictions. A longstanding gripe of mine has been that, in Canada at least, new cars are only required to have front lights illuminated in daytime, afaik. The practical effect is that it is a regular occurrence to have dangerous situations where cars with no rear lights illuminated are driving in dim / rainy conditions.
Distinguished from Daytime Running Lights.
Because otherwise this is just wrong.
That's the table stakes: legal liability. I don't want to hear any other automaker, Tesla or otherwise, crow about their "self-driving" tech if they don't take legal liability for it, because it proves beyond a shadow of a doubt that they don't have faith in their system. Kudos to Mercedes.
As for the colour, clearly Mercedes requested Turquoise for brand reasons. I wonder if there is a better colour to be adopted as a standard? External indicator lights of any kind seem like something that needs to be standardised before we end up with a whole mess of different colours on the roads causing confusion.
“The turquoise color for Level 3 driving was chosen by SAE, so other brands beyond just Mercedes will use it for their own cars.”I'm genuinely curious - what are you thinking when you say this?
I don't know what I'd do with that information. It shouldn't make me drive differently - e.g. not be aggressive to see if I can invoke reaction(s) in the autonomous system, and I should already be driving w/ due care.
I.E. If you're watching a TV show on the center screen it could be illegal if you don't have the system (not like that stops anyone today anyways -_-).
Like the article hints at, a true self driving car lets the driver behave in ways that would normally be reckless and illegal. Best to avoid ambiguity.
> if you see the teal lights, you know the car is using automated driving, so it’s less likely for other motorists to be shocked and concerned when they see the occupants looking away from the road, plus emergency services have a better understanding of what to do if medical assistance is needed. It will also be super helpful for law enforcement, so cops hopefully won’t pull you over for watching TikToks on your infotainment screen when it’s perfectly allowed by law.
I can’t clearly articulate what it was I was seeing, but doing it two or three times a day, I got very good at it.
As a pedestrian, I always try to make eye contact with drivers before crossing in front of them. It would be nice to know that the car is doing the driving.
Being able to distinguish between a mercedes driven by a mercedes driver and one driven by a mercedes automation is going to help predict behavior of the car. Of course, one needs to be ready for anything, but when I can predict behavior of other cars, it increases my planning window.
It's quite the opposite... if there's a tiny gap between two cars in the other lane, you usually don't squeeze in between them, becaus the rear driver in that lane could be distracted and wont brake and let you in in time, or might intentionally not widen the gap, to net let you squeeze inbetween. With automated cars, you can slowly swerve into the too-narrow gap, and the safety mechanisms on the self-driving car will automatically make room for you :) There are many dangerous maneuvers that agressive drivers can't do now, because there will be a crash if the other drivers don't notice them and brake in time, and the self-driving cars will always notice and make room for the either bad or intentionally agressive driver.
I'd be pretty surprised if those lights aren't already LEDs.
Halogen bulbs found in headlights produce a lot of light under that wavelength, and are going to be at a much higher intensity than turquoise indicator lights.
Tangential, but holy shit, this is great to know, and explains why I generally feel blinded every time someone with those horrible headlights drives by.
I wonder if there’s anything that can be done (to my windshield? to some glasses?) that can reduce the impact of this at night?
Also, halogen is an old lamp type that has been in use since the 1960s. Then towards the 1990s, there were changes in optics with "projector" housings for halogens, and then the shift to high-intensity discharge (HID aka xenon) lamps. The latest ones now are LED, of course.
Tesla could conceivably launch a level 3 system without approval by taking full responsibility for any accident that happens when the system is running even when the driver is inattentive. Even Tesla is not crazy enough to do that.
Actually anyone even slightly interested in automotive industry isn't surprised by this. Tesla's tech is all smoke and mirrors for years now
Is Tesla FSD ready to replace human drivers? It is not. Is it smoke and mirrors? No - there is a mountain of evidence that it is a very capable product in many situations.
When looking, the Odyssey just didn't have the price/value ratio, it really needs an update and refresh. The Sienna just doesn't exist.
This is especially true for their EVs built on the E-GMP platform [1]. You'd never guess it by looking at them, but the Hyundai IONIQ 5, Kia EV6, and Genesis GV60 are all built on that same platform, with different bodies and interiors, along with some different suspension tuning.
The driving assistance features do vary with the trim level. In the EV6, the Wind trim has a more basic system, the Wind with Technology Package has more, and the GT-Line gives you the works.
I think the high-performance GT has the same assistance features as the GT-Line, but the GT-Line has a nicer and more comfortable interior than the GT.
My AWD EV6 GT-Line with the optional fake vegan suede seats (instead of the standard fake vegan leather) is by far the best car I've ever had.
(These trim level names are for the US models; they vary a bit in other markets.)
The only complaint about its driver assistance is that it can be a bit twitchy compared with how smoothly I can drive manually when I'm paying 100% attention. But it is great to have in case I get distracted for a moment!
[1] https://www.hyundai.com/worldwide/en/brand-journal/ioniq/e-g...
My response was simply to point out that it is wrong to say it is smoke and mirrors (or I would be dead in a high speed car crash instead of writing this).
It is certainly misleading to call it Full Self Driving - no argument there at all. It's also wrong to call all of Tesla's tech for the last however many years fakery.
Anyway, that doesn't really have anything to do with the original post - all I was saying was that it is a functional product and that it's just incorrect to call it smoke and mirrors.
I don't think most people realize how erraticly they drive.
But five warnings in 30 minutes could just be 15 or 20 seconds of inattention. So more than 99% full attention. A driver assist that just smooths things out only 0.1% of the time could save the day a lot.
These lane departure systems are trash, in my experience. They can’t distinguish between real lane markings, cracks filled with sealant, old lane markings that have been painted over, etc.
The worst thing is when they take over your steering wheel for these phantom departures. I hate it, and always disable these systems in rental cars.
Tesla's self driving was always neat, but it made a joke of itself by announcing "fully self driving" capabilities when it was clear nobody had the tech there. (Even when generously assuming their camera-only sensor kit could be sufficient. It would be counterintuitive but not unthinkable.)
Which has pros and cons. Because these systems have a wider margin of error, they can operate in more situations. The con is that they make more mistakes so a driver must babysit them.
The problem is they take “fake it till you make it” too far in their marketing, by implying it is more than level 2.
https://www.youtube.com/watch?v=k7YVxLLIuGM
It's not exactly what you describe but it comes close.
Lidar, cameras, wireless communication with other vehicles and infrastructure, and with the internal sensors in the car.
More concerning would be adversarially-crafted signals. (Think a Wil-E-Coyote style fake-tunnel, but for lidar and and whatnot too.)
And most systems use an array of sensors like radar, laser, and vision too. It’s only musk who is hell bent on lying to his customers about machine vision and also lying about his intentions where the goal is cost savings and not because machine vision is better.
Still don't know how to think about this - how is it possible for a phone to have LiDAR (barely utilized) but it's too expensive to have them in such a critical setup as an autonomous vehicle.
Just like you can have a 50MP camera sensor in the phone: it works? Sure. Is it sufficient to capture the Moon (heh) or something 50-100 meters from you? Not at the quality and performance of professional equipment.
Sure, you can have a matrix of iPhone LIDARs but whould they works at sufficient speed, accuracy, at +40C, at -35?
All L3 systems only operate under known easier conditions. The self driving could turn off in unfavorable conditions .
My car’s lane centering switches off if there is rain or snow obstructing sensor .
The iPhone lidar has a range of 5 meters.
- Only in CA and NV
- Only below 40mph
- Only on specifically mapped freeways ("only available on freeways that have been mapped by Mercedes")
- There MUST be a vehicle in front of your car
- "reasonable road conditions with readable markings and lines"
- "clear weather"
- Daytime only ("Drive Pilot can’t be used at night or in the rain, and the headlights and wipers must be set to auto for it to work. ")
- Can't be in a construction zone
https://www.theverge.com/2023/9/27/23892154/mercedes-benz-dr...
The requirement to have a car in front of you is also very interesting.
> Only on specifically mapped freeways
Those two seem to contradict each other - there are no freeways in CA or NV where you can (safely) drive below 40mph
So... That rules out nearly all driving, probably usable for 0.1% of driving.
It is not solving a pain point for you and others I understand but for the BMW sedan target demographic I think has a good overlap to these conditions to their typical usage
It was not as well marketed/hyped has the Tesla of today, but you could buy an S-class that would do much of the work for you in traffic well before you could buy a model S.
Period. They are one of the companies that develops massively expensive flagship products that change the industry.
80s kids should know what I mean.
IYKYK
Yet, Telsa continues to market FSD. Where is Lina Khan on this?
I do not think that a single light can portray the complexity that we look for in the current situation.
I would guess that at some point cars will communicate directly with each other, and a light might appear on the dash to indicate that a vehicle is aware of your presence/intention. That way human drivers would know that it's safe to maneuver around the AVs whose paths he might cross. But it will probably be a decade before this sort of stuff is worked out, at the earliest.
See Johnny Cab except of course one inside to chat the passengers (turned toward the passengers then in chat mode, and deflated in privacy mode), and one on the hood to make pedestrians feel safe.
In this context it means basically light cyan. A much less marketable name.
This will backfire horribly. While there is some residual benefit to law enforcement, these lights will signal to human drivers and pedestrians that this vehicle can be bullied with impunity, and with no risk of angry retaliation. Human drivers will cut in front of them mercilessly. Pedestrians will step onto the road in front of them. Maybe not in classy affluent suburbs, but definitely in rough neighborhoods.
--
[0] I think it's obvious that anything less than two orders of magnitude (100X safer) is socially unsustainable. Think about it. 40 thousand people die on roads in the USA each year. Replace all drivers with an AI that's "only" 100X safer and we'll initially marvel at the 39k lives saved every year. But that will quickly be forgotten. What's left is the brutal carnage of 400 people mercilessly killed by robot cars annually...
A car at even 20 MPH has a tremendous amount of kinetic energy, and we are all well aware of the destructive consequences of rapid deceleration when large amounts of energy are involved, particularly when fleshy human bodies are involved. There exists no other system with anything close to a comparable level of risk where we do not require or at least expect a human to be actively monitoring the system and ready to take action in the event of a problem.
I'm not saying we should ban Level whatever auto magic AI driving. All I'm saying is that we should use the same standards of responsible operation that we apply to literally every other area of human activity. Why does driving get a free pass?
> Mercedes-Benz is the first automaker to gain approval to sell a Level 3 automated driving system in the United States; the first customer-owned S-Class and EQS sedans in California and Nevada equipped with the Drive Pilot system hit dealer lots this month.
If the vehicle has a driver's seat and you're sitting in it, you should be considered responsible for the car's actions. If you want to watch a movie, that's fine, don't sit in the driver's seat. If you want your car to drive your drunk ass home, that's fine, don't sit in the driver's seat.
do you really think there’s a future where people in ‘rough neighborhoods’ (this is very poorly hidden classism/racism) where they are jumping out in front of cars? there’s very little evidence that people in a mentally sane state, read not incapacitated, will self harm to spite technology.
i know living in the bay i’d mess with the self driving cars, but never by getting in the right of way and risking my damn life, only by signaling a left and forcing them to slow down behind me. i never once tried to get infront of the car at the last moment to see if i could be menaingfully maimed. i strongly suspect the desire to not be hurt will trump your concept of how ‘poors’ live
For starters, dash cam footage of how drivers and pedestrians are willing to interact with vehicles with human drivers. If that's what they're willing to do when it's an unpredictable human in control, I can only imagine what they'll do when it's an extremely predictable computer in control.
> people in ‘rough neighborhoods’ (this is very poorly hidden classism/racism)
I think it's racist to assume that rough areas implicitly means a different racial demographics. That might be the case in your country, but it isn't in mine.
In cases where there are different racial demographics in rough areas, the fact of this disparity is a fact. Your dispute is with reality, not the person who might dare to point at reality. It's my opinion that systemic improvement is made harder when we cloak reality in double-speak.
> i strongly suspect the desire to not be hurt
I'm not necessarily talking about jumping into the path of an autonomous car so that it's forced to perform an emergency brake. I'm talking about jaywalking without concern for the car having to slow down well ahead of you to maintain a safe distance.
about the dashcam footage, i think we’re seeing different things. it’s true, some of the behavior in the footage is similar to what’s normal in parts of asia. cultural habits can clash, and it’s rough to see. but i agree, it’s probably not deliberate, just different habits.
and you’re right, some drivers are just inconsiderate. but an autonomous car, with its 360-degree awareness, might handle things like unexpected merges better than we think. it’s a good point, technology might adapt to these situations more efficiently than we anticipate.
‘jaywalking’ is racist in origin, a term used to denote ‘jays’ who didn’t understand the rules of the road because the rules where so new culturally. it’s a fitting term, but poor narative, for the clashes you’ve identified in social expectations
This is false. The origin of the term is NOT racist. Please stop reflexively assigning racism to things, it dilutes and degrades actual racism.
https://www.merriam-webster.com/wordplay/why-is-it-called-ja...
Pedestrians SHOULD be able to just step onto the road on a crosswalk without having to worry if drivers will stop as they're legally required to.
The risk of angry retaliation is identical to what we have already. The driver of this car can just take manual control at any time and react as they would in a non-AV.
If what you're claiming is true, the behaviors you're describing would already be common with traditional cars.