Tesla in self-drive mode slams into police car in Orange County
ktla.com
ktla.com
I'd be willing to bet that Elon's insistence on only using cameras is combining with the flashing lights to effectively make objects that appear in some frames of video, and not in others, as the camera recovers from the flash... and this makes it appear as noise, and below a certain threshold, ignored.
Or, it could be the reflective decals... who knows.
On a different note, I like how Mercedes changes the color of external lights when in self-driving mode to make others aware. A bit like outsourcing the responsibility externally - but maybe this is the way forward until the tech is reliable/better than people.
It would be useful to have a universal outside light that informs drivers behind you that the reason you're keeping a safe distance ahead is that the car is on adaptive cruise control. Some people feel like 2 seconds distance is a reason to try crazy stunts to get around you. And then once the road is clear drive slower in front of you than the speed you had set...
When was their vision system not glitching? A year ago it was seeing the moon on the sky as a perpetual yellow traffic light.
Sometimes it might be difficult to tell, but in the case of moon -> traffic light it seems clear.
Maybe I'm wrong about the terminology.
it's like saying transistor and cpu is a cousin...
Which is why suitability decisions should not be made based on anecdotes.
Nobody proposed this. My Subaru can lane keep even in whiteout conditions (if there is a car in front of me) due to radar. I can’t do that. Tesla’s can’t do that.
> it will occasionally do something as stupid looking as "drive into a police car with flashing lights on"
Two sensors operating on different wavelengths can add more information than a single doubling of the first sensor.
Tesla took out the non-visible detectors for cost savings alone. I don't get anyone who every believed otherwise.
https://www.subaru.com/eyesight.html
IME it works pretty well but cuts out in heavy precipitation because the cameras don't get a clear image.
Second, radar wouldn't be better here. Afaik, all of the radar assisted cruise control / emergency braking systems ignore stationary objects when travelling above a certain speed. Using radar, it's difficult to determine the height or shape of a potential obstruction, so a car on the road looks a lot like a metal sign above the roadway.
Lidar is much better, but cost prohibitive to put onto vehicles for driver assistance. A camera system could possibly detect stationary objects of concern, but apparently doesn't.
This is false. Radar assisted cruise control definitely do not ignore stationary objects above a certain speed; indeed, detecting stationary objects is a fundamental part of how emergency braking systems work...
Other automakers are simply better at determining what objects are in the path of travel than Tesla, because they actually test their systems before they release them to the public.
> Sam Abuelsamid, an industry analyst at Navigant and former automotive engineer, tells Ars that it's "pretty much universal" that "vehicles are programmed to ignore stationary objects at higher speeds."
Chevrolet says [2]
> At speeds between 5 and 50 mph, Automatic Emergency Braking (AEB) can help you avoid or reduce the severity of a collision† with a detected vehicle you’re following using camera technology. It can automatically provide hard emergency braking or enhance the driver’s hard braking.
They don't specifically say it won't detect stationary objects on this page, but a vehicle you're following will have been in motion, or you're not following it.
Consumer reports describes several types of AEB [3], none of which mention stationary objects.
If you've got a reference that suggests that driver assistance features stop for stationary objects outside of parking speeds, please provide a link.
[1] https://arstechnica.com/cars/2018/06/why-emergency-braking-s...
[2] https://www.chevrolet.com/support/vehicle/driving-safety/bra...
[3] https://www.consumerreports.org/cars/car-safety/automatic-em...
He went so far as saying that a can on the road was indistinguishable from a highway underpass and then went on about how they would white list every underpass to avoid it triggering
But this is Elon, so I have no faith in its accuracy, but don't be surprised when people believe that the CEO is describing how their car works
Why, actually? Luminar sells LiDARs (that they market as intended for vehicles) for $1000. Tesla sells FSD Beta for $12000.
Do you believe a much more expensive LiDAR (than Luminar) is needed?
Is is an wavelength/antenna sizing problem? Surely a radar could tell how tall/how high above ground an object is in principle.
Contrast this with Subaru, who uses binocular view through two widely spaced cameras to build a depth map of the view in front of it. It doesn't know or care what is there, only that something is there it shouldn't hit. And so the emergency breaking will trigger for things like trains and stationary emergency vehicles that Tesla seems oblivious to.
Sometimes simpler really is better.
If Tesla claims next week "it wasn't self driving at the time of the crash" that doesn't tell you anything. Because they will claim that even if the car self drove right into the police car but disengaged self driving mode 0.1 seconds before the crash.
Their CEO is a pathological liar that will say anything to make the stock price go up.
Because I have not seen them.
Not 100% sure but it almost seemed that it was in self drive mode. Or, the guy with MIT sticker at the back was just a bleeping bad driver.
They are making things worse for everyone.
Robot taxi: https://youtu.be/5cFTlx68itk?t=121
They admitted to being on their phone. This is an easy choice
I’m a pilot and airplane owner.
This bullshit where companies are allowed to say “the driver must always be in control”, is just that. Either it is self driving, or it is not.
You cannnot make a product that is physiologically impossible to use, and then say that fault lies with anything other than the product.
This is before you get to it being labeled and advertised explicitly as literally “full self driving” which is objectively fraud.
1. The NTSB and FAA both found repeatedly, as have pretty much every other study - academic or otherwise - that people _cannot_ focus on watching a task without actually being involved in or doing the task. Any "self driving" model that requires the non-operator to be engaged in driving is simply not humanly possible. So the only reason for that stated rule is the manufacturers know these systems are unsafe, but nonetheless want to make themselves not be liable for their faulty products.
2. The only other "vehicle completely controls itself" system in practice is aircraft flight, where the pilots are necessarily engaged in all critical times, and for the remainder of the time all the systems either give sufficient warning before anything goes wrong (e.g. when autopilot disengages on an aircraft you have a substantial amount of time before impact - not the <=2s warning Tesla gives people), or the aircraft systems have specific extremely heavily trained signals for which there is only a single action and no immediate requirement to regain situational awareness (stall, acas, tcas, etc). The time these "self driving" systems provide to regain situational awareness and then take appropriate action is less than what we expect of trained pilots, yet it's somehow acceptable for random drivers.
The current self driving "you are expected to be in charge of the vehicle at all times" is entirely liability shifting: manufacturers are knowingly selling unsafe products, especially Tesla, they are lying about the capabilities of those products, and then they are saying it is the drivers fault when they fail.
What manufacturers are doing is no different from a manufacturer selling a car with an ABS system that fails, and saying "the driver is responsible for identifying ABS has failed and pumping the brakes if it does, if they fail to do so we do not accept liability for the ABS failing"
Especially in the terminal area there is no such thing as the autopilot flying the plane with no pilot involvement. There are heading, speed and altitude adjustments as input to the autopilot just like a car driver would be actively involved when driving with normal cruise control.
The difference in aircraft autopilots is that if you take 5 or 10 or even 30 seconds to notice an issue, you correct. Whereas for “self driving” you have 2 seconds or less, and are expected to have complete situational awareness for the entire time. [edit: to notice non critical problems, immediate response failures have significant repetitive training to cause immediate correct response - stall, tcas, mcas, etc]
As in, the defined “correct” way to use self driving cars is to have as much focus awareness and concentration as you use while actually driving, only without actually doing anything at all. Which again, is not something that any person is capable of doing. This is established, even in the context of pilots.
In the context of commercial pilots as I understand it, autopilot is often disengaged for the purpose of ensuring they’re engaged (in addition to standard elements of not getting rusty).
And again, if the autopilot fails in flight, pilots who are trained specifically to deal with the myriad issues and complexities of failures and autopilot disconnects, have literal orders of magnitude more time to react than untrained occupants of “self driving” cars.
Referring to autopilot not existing in terminal or during takeoff/landing (though I do understand ILS or similar can handle normal landing?). Those are times when the pilots are actively engaged in operating the aircraft (even with ILS).
The issue with autopilot/self driving is not what happens while operating in an environment that requires active engagement by the operator (pilot or driver), nor what happens at points of engagement (start or end of travel, making adjustments to travel), it is that humans are physiologically incapable of maintaining engagement in an activity if they are not actually involved in the activity. Again, this is something that has been established in study after study, and applies to extremely well trained pilots responsible for hundreds of lives just as much as it does random driver.
These cars “self driving” systems require the occupant (they are not driving) of a car to have greater control and engagement, and faster reaction times, than a trained pilot, in a system that has even less cause for engagement.
You referenced how autopilot in an aircraft requires adjustments (which as far as I understand are not being made constantly, but no matter, let’s assume they are), “self driving” cars do not even have that. The argument “aircraft autopilot requires engagement from the pilot” does not make my point about “self driving” cars incorrect: it in fact further demonstrates how unsafe “self driving” cars are - pilots in this model do have actual engagement with the operation of the self flying plane, in addition to increased training, significantly larger time to react, and significantly better response options when no-delay responses are needed.
Self driving cars say the occupant is required to have greater engagement in driving the self driving car, while doing less, having less training, having less time to gain full situational awareness, having less time to respond when the system fails, and not having any clear response options or training for immediate-response-needed events.
I’m genuinely curious, and I’m sure that NTSB or FAA must have published some statistics on the amount of time pilots need to notice and respond correctly to different failure modes.
The inherent problem of constant attentiveness being extremely challenging means that most relevant authorities say a life guard should not be on duty for more than an hour at a time, and they should have at least 15 minutes break between those hour shifts.
Even with those minds are going to wander (or there will be distractions - people asking for directions, etc) but there are multiple failsafes: there's generally more than one life guard, being distracted for a few seconds is not a failure (a life guard is having to scan a significant area so they don't have 100% awareness of 100% of their zone at once, so there's always potentially significant delay between something going wrong for someone, and a life guard seeing it), and the time to catastrophic failure is measured in (contextually) significant amounts of time.
The issue with self driving cars, as they are currently set up, is that they say "the car will do everything" and it does, but they then say "however the driver is still in control of the vehicle so if a crash happens it was the drivers fault for not paying sufficient attention".
In the pilot case: there are periods of flight where the pilot is doing very little for extended periods, but those are all at altitude, and the time from "something went wrong" to "it is irrecoverable" (in non-aircraft failure modes) is remarkably large (at least to me - my mental model was always 'something went wrong, it's seconds to crashing' until I binged air crash documentaries and even if they're trying it takes a long time to go from cruising altitude to 0). There are also modes where the pilots must always react immediately, whether or not they were distracted, or if they were focused but on a completely different task, but those modes are all close to "this alert occurs->reflexively do a specific action before you even know why".
Attentiveness is a real problem for long haul train traffic and multiple accidents have occurred because of it (or the loss of it), and there are many things they've tried to do to prevent the exact same problem that self driving cars introduce, and they simply do not work. At least for trains you can in principle (those the US seemingly does not) use safety cut offs such that a train that is not responding correctly to signals is halted automatically regardless of the engineers and operators. What companies frequently try to do is add variations of dead controller switches (similar to the eye tracking in "self driving" cars), but for the same reason that attentiveness is an issue in multiple hours of no operation those switches get circumvented (brains don't like to focus on a single thing while not actually doing anything for hours, muscles don't like being constantly in a single stress point for hours).
1) FSD makes me a better driver. Without it, I'm (uselessly) more aggressive and on the gas because it feels good. With it, it's basically ego-less driving.
2) The headlines are dramatic, but what does the actual data say? Tesla has thousands of miles driven under "full self driving", is it actually more dangerous (in the quantity or severity of average or long-tail accidents)?
3) FSD is surely an improvement over impaired driving, however that is defined. Does this encourage impairment?
> It was unclear if the driver would be cited.
Oh, come on!
That said I hope there are repercussions for the driver
That said, given that over my lifetime, we seem to have legalized all sorts of fraud and removed as many consumer protections as possible, I can see why people get a bit confused.
Sigh
Extra full self driving?