Telsa is only a level 2 self driving car, but it has been sold as something way more advanced (Telsa has since pulled back some of the hype). A lot of people don't understand that there are even levels at all.
A lot of people don't understand that there are even levels at all.
People shouldn't really have to. The phrase "self driving car" is a binary yes/no in common, regular English. I don't know if this is Tesla's fault or some committee's fault (SAE?), but it's really terrible marketing to call things "self driving cars" when they can't drive by themselves.
Cars shouldn't be allowed to be marketed self-driving, or autonomous, etc., until a driver isn't required.
I'm certain the colloquial use will remain, but "doesn't require a driver" should really be the one to get it's own name. Fully Autonomous Vehicle or Driverless Autonomous Vehicle.
"Self-driving car" is just woefully inadequate to describe the capabilities it encompasses. A century old Model T with a brick on the accelerator could be accurately described as a "self-driving car". It's not very good at driving, but neither are drunk people and we still say that they're driving.
Likewise, I would argue that a car that can maintain speed and steer itself to stay in it's lane is conceptually closer to a "self-driving car" than it is a "fully manual car".
> Cars shouldn't be allowed to be marketed self-driving, or autonomous, etc., until a driver isn't required.
That's going to be a licensing thing. Frankly it probably is time for the government to step in and create some kind of licensing for cars that indicates they're fully autonomous. Even if they aren't going to issue any at the moment, it would be good to set the direction that only cars with whatever government seal are actually able to operate without a driver. Then it'd be a little harder for companies to play word games.
> People shouldn't really have to. The phrase "self driving car" is a binary yes/no in common, regular English. I don't know if this is Tesla's fault or some committee's fault (SAE?), but it's really terrible marketing to call things "self driving cars" when they can't drive by themselves.
It's Tesla's, they're the ones calling their system "autopilot" (conjuring the idea of the car driving itself in the 99.99% of the population which has never flown a plane with an autopilot and has no idea how limited they can be) and literally selling a "Full Self-Driving Capability".
Other companies have generally been careful to bill their tech as assistive (Volvo's automation system is even carefully called "Pilot Assist").
And SAE is levels of automation, describing how little / how much automation there is in the driving system, it doesn't intrinsically imply self-driving capability, only that some aspects of driving can conditionally be handled automatically. SAE levels 1 and 2 are specifically worded in terms of assistance as well, here's SAE2:
> The driving mode-specific execution by one or more driver assistance systems of both steering and acceleration/deceleration using information about the driving environment and with the expectation that the human driver performs all remaining aspects of the dynamic driving task
Where I live I have seen Tesla owners riding in the passenger seat (driver's seat empty).
One of these douchenozzles has a sticker that says 'I'm probably not driving'. This particular tesla has whatever 'drive like an asshole' package where the car is accelerating unsafely, weaving between cars/lanes... as this lady is reading a book.
I followed it for a while just out of curiosity. Once it got to a shopping center with dog-legs to turn left, the tesla lost its shit. It began for no reason to ride the curb, and I mean ride it. Brand new tesla just ate up both its driver's side rims to the point sparks could be seen. The lady reached over and did something that did not work, and it just kept going until she slid over and regained control.
If tesla can't even get the simplest of shit right how tf can they claim self-driving? And does it matter that if this person kills someone they are at fault? No one would have died if Tesla hadn't lied about their capabilities.
How many people must die before they are stopped from doing this shit? Tesla says they aren't misleading anyone, but then why do so many of their customers drive hands-free?
There is no such thing as a "level 2 self driving car". Tesla does not have any form of a "self driving car". It has "level 2 automation" just like basically every other large manufacturer (Honda has some production cars with "level 3 automation").
In a safety culture oriented organization drivers would have to receive training before operating a machine like this. Some please explain why exactly isn't that the case with Level 2 self driving cars? Drivers clearly don't understand the limitations of the system.
You have to pass a driving test at age 16 (in US), after that you never need to test again (sometimes eye checks when you are old), cause um... nothing changes about driving or cars or rules in the next 64+ years it would seem.
This problem is actually pretty obvious to fix and you don't even need a neural net - it can definitely be done "old school" computer vision.
Simply detect the presence of "flashing emergency lights" in the oncoming lane and disable autopilot when present. No object detector needed. The signal is so strong it's literally flashing extremely brightly in a regular, predictable pattern - any vision grad student should be able to figure this out.
Could there be false positives? Yep, but very few things will flash quite like that, and most half baked vision engineers can do this. The worst case is literally simply the driver taking over on occasion at night (maybe near strip clubs? lol)
I don't think the problem is as simple as that. Sure flashy lights with reflectors confuse the current model sure, but can we consider that an edge-case and work from there? Also LED's flash too in a regular pattern too. I don't know though, seems like something that would have been accounted for as it's a pretty important part of driving awareness...maybe they didn't think people would want to stay on "full auto" when a police/Emergency vehicle is in the vicinity since you might have to do some sort of maneuver to get out of it's way. So yeah, I'll say it's an edge case for now.
Disclaimer: I'm an idiot so take my view with a boeing 747 load of salt
Whilst I agree with your point, I'd also like to see an additional strategy employed.
How about emergency vehicles broadcast on 5Ghz their intended route (for the next 300 meters) or just that they are blue light in the area.
This would only be active in a blue light situation. Manufacturers can then add a detection and warn the driver.
A while back I heard about car to car communications, but I heard next to nothing these days, this is a great use case if you ask me. If someone has more details of recent developments, I'd love to read more.
> This problem is actually pretty obvious to fix and you don't even need a neural net
So can be controlled collisions as such.
Simplest FMCW radars are more than enough for every emergency braking system on the market.
It's not uncommon to even use them in parallel with some more "brainy" radar imagers like on Mercedes cars. That is to make a collision preventable in case the main imaging computer hangs, or crashes.
The solution of disabling autopilot in the presence of flashing emergency lights seems so incredibly obvious to me, that I can only assume I'm missing some glaring reason for it to not have been the case since day one.
There aren't separate stats for stationary firetrucks, but looking at all collisions with emergency vehicles (including at high speed), it's in the order of magnitude of 15,000 collisions with more than 3 trillion miles driven.
Tesla is so many orders of magnitude worse in this area than human drivers that it will be laughable for them to try to resort to Musk's usual lie -- that Teslas are better drivers than humans.
Sure, but not counting those makes Tesla look even worse in the comparison. Even in the very best comparison you can come up with for Tesla, they are still many magnitudes worse than other cars.
Does not matter, because it is not a systematic failure.
But when a piece of software deployed by an OEM has a problem, then it is a systematic failure where all cars of that OEM are having the very same problem and then the OEM has to do something.
Yeah exactly, a failure that is different than a human failure is going to be noticed, even if it's not more common.
It's probably hard to intuit, and no one explicitly thinks about this (except AI engineers and researchers), but emergency vehicles are parked in a way that is "obvious" to human drivers.
Every situation is different -- how much shoulder there is, etc. What the visibility is. The driver unconsciously makes a judgement call to be visible (to humans). There is no way they can judge how an AI is going to view them (which changes from time to time, and across AIs).
So the only solution is for the AI to err in the same way that humans do, or not at all.
Remember when there was the case of unexpected speeding leading to accident by one OEM?
It doesn't matter the question of "Oh, what was the accident rate when people where unexpectedly speeding?"
What matters was that single type of car by one OEM. If that OEM has some type of problem with its cars, then its a systematic failure which needs to be investigated and resolved by that OEM.
This belongs in the straw man Hall of Fame. The stuck accelerator is the problem and crashes are the terrible symptom. You would compare how often Toyota's accelerators stuck compared to other manufacturers in bringing a class action suit, would you not?
Similarly, if the AI driver crashes in that situation in a similar (or even less) rate to the base rate (manual driving), you would likely not reasonably bring a suit.
Either way you should arrive to improve the AI, but the base rate does matter, and telling someone their question does not matter because you personally don't care about it is bullshit.
The base rate of results doesn’t necessarily matter when you are talking about two different causes. For a stuck accelerator, you are comparing the same cause (stuck accelerator) leading to the same result (fast acceleration leading to dangerous driving conditions). You can’t compare the base rate of fast acceleration leading to dangerous driving conditions, that can happen for many reasons, including intentional driver choice, and make any meaningful conclusion. Because the base rate of an equipment failure (stuck accelerator) leading to fast acceleration, and the base rate of individual driver choice leading to fast acceleration aren’t the same thing.
It is the same here. Comparing individual human choices and errors leading to striking emergency vehicles with an equipment failure (Tesla’s detection and control software) leading to striking emergency vehicles is not a comparison and it is no more an excuse to not hold Tesla accountable than it would be to not hold Toyota accountable because their accelerators stick less often than a teenager decided to mash the gas on his new car and wrecks it.
This kind of thinking is how we got here in the first place. Edge cases that no one thought of, that no one wanted to deal with. Treating driving software like a web application that we can just fix later when the bug (i.e. a number of deaths) happen. No big deal, bro. Move fast and break things, right? We're disrupting the auto industry!
This is a good question. I couldn't find statistics for striking parked vehicles, but apparently there's something like 20000 accidents a year involving Emergency vehicles. [0]
I wasn't sure of your comment was being sarcastic or not.
In any case the article clearly suggests that it's a systemic issue due to using too few sensor modalities in conjunction with too little training data. Hence it's not restricted to firetrucks, but many new, unlearned situations.
As a human I find dealing with going around emergency vehicles at night is a bit troublesome. The emergency lights are overwhelming and obscure your vision from other things, like some worker near the vehicle. Also the human distraction that is target fixation can sink your attention on the emergency scene and distract you from checking your lane to see if it's clear. I don't think this situation is limited to tesla...
Is it possible that you are looking at the “wrong” Alice of data? These type of accidents are for sure very uncommon compared to any other accidents just by the ratio of vehicles on the road.
I wonder if you could look at accidents that involved a stationary police vehicle. I’d be curious if it’s still very uncommon.
Anyway, I heard that so many times from cops directly. Squad cars in the night are like magnets for drunk drivers.
> I wonder if you could look at accidents that involved a stationary police vehicle. I’d be curious if it’s still very uncommon.
Accidents involving any emergency vehicle traveling at any speed (including very fast through traffic) are less than 25,000/year in the US.
As I mentioned in another comment, humans drive more than 3 trillion miles per year in the US, so the rate at which Teslas are hitting stationary police vehicles even if they have only ever done it once would be much higher than the rate at which humans do.
If you watch a squad car pulled over you will notice they are at an angle so that the driver and get out of the far side of the car. That way when someone hits the car, the car protects them from the other car.
That's besides the point though. There's still a human that's able to react to new situations (albeit very slowly) on the road. Teslas apparently can't do that very well. The question is: how do we fix this? And is this potential for regulation? For me personally, this is one of my biggest hangups with these new technologies: they're black boxes. We don't know how they work, and they change all the time. Companies can get most of the general cases correct, but as for the panoply of edge cases, how do we approach them? How should insurance companies approach them? Whose at fault here, who pays who?
There's just so many unanswered questions with automated driving that a lot of people don't really want to solve, they just want the technology out there and we'll worry about the edge cases later...as in, we'll worry about it when people start dying. "Move fast and break things", that retarded shit. I hate it. This SV frat programmer mentality is now bleeding over into automobiles, an industry that does indeed need to be shaken up, but not when it's at the expense of lives.
Treating driving software like it's a fucking web application that can be fixed later is malicious negligence.
Much like elevators, computers are fundamentally better at controlling the system than humans. I assume you ride in computer controlled elevators? Do you understand that they were at one point controlled by humans manually? Did you know that there was a transition period where the computer control elevator was not as safe as it is now? In 50 years the idea of letting a 85 year old or a person walking directly out of a bar control a four thousand pound machine moving at 60 miles an hour will seem insane.
You are talking about the brake for if the cable breaks. I'm talking about the control system that prevents the doors opening while it is moving, opening in between floors, shutting down if there is a problem, etc.... https://en.wikipedia.org/wiki/Elevator#Manual_controls
Those doors have also safety systems. Formerly, the circuitry to open the door was closed only, when the cabin was in floor level. Opening and closing required constant push of the push button.
(Have some history in elevation control, because we made some standards in the industry.)
Yes. Talk to any police or first responder, they have training to cover this. For this and some other reasons is why police often approach stopped vehicles from the passenger side, further off the road and shielded from oncoming traffic by the vehicles.
> Talk to any police or first responder, they have training to cover this.
That has no bearing on how common these types of collisions are. The original issue is whether Teslas are more likely to crash into a parked emergency vehicle than a human.
The training has a better explanation: while "number of collisions per mile driven" is vanishingly small, "percentage of law-enforcement deaths as a result of collisions" is very high. Out of the ~1,200 LEOs killed between 2000 and 2008, about half were due to collisions[1].
That alone tells you how such collisions can be rare, but still a top-of-mind issue for law enforcement.
I guess that means sharks frequently bite humans[1], dogs regularly befriend cheetahs[2], and people often hit the jackpot when they play slot machines[3].
(YouTube search results are not useful data about national trends, and YouTube would be intensely boring if they were.)
> Colliding with parked emergency vehicles is extremely rare, almost non-existent.
Maybe, but the very first thing i was taught for how to work a MVA was that under no circumstances were you allowed to walk between parked emergency vehicles, because fenders are just a very blunt pair of shears at knee height that will happily amputate you when some dingus rear-ends the parked squad car.
It isn't, but humans generally recognize emergency vehicles and know to slow way down. They might not find the correct path, but at slow speeds accidents are not deadly, plus there is a lot more time to respond.
Note that I said generally recognize and not always. Humans are bad drivers and never should be in control of a car. (which is why I'm so in favor of mass transit)
FWIW those blinding LED light bars have ambient light sensors & night time dimming support, but the department has to enable it. If you feel they are as truly blinding, message your department and explain that as a driver they are so blinding you are concerned one day an officer will be hit by a car that could not see them through the glare.
In my state (Montana) there is a law that vehicles must move one lane away from a parked emergency vehicle (leaving an empty lane between you and the emergency vehicle). I wonder if Tesla's firmware understands it needs to do that?
While it might be interesting to know what it does in such cases, it shouldn't matter. You as a human should always take full control when encountering unusual activity on road. Relying on currently available driving aid systems to handle edge cases is just begging for trouble.
I’m curious too. There are several states with a variation of this law. It tends to be: move over if you can or slow down by 20 below the speed limit if you cannot.
What's equally laughable and gutsy to me is Tesla's assertion that they can achieve parity with cameras, what radar (and now lidar) have been doing for decades.
When Mr. Rajkumar details how a camera works (he did dumb it down for the layperson quite a bit), it seems almost laughable that a technology like that can even work in a real-world scenario like vehicular traffic.
I've always been of the mind that this technology can only work if you reduce the number of variables at play... that means removing the human element altogether (and even then, mother nature may have other plans), meaning self-driving cars should have their own roads. Pipe dream, I know.
> meaning self-driving cars should have their own roads.
Or retrofit our roads with Machine friendly signals? Signs that can ping information to the car, other cars pinging information to the car...Those sound like cheaper alternatives and the models can focus more on the difficult problems/edge cases like overtaking and "oh human like dog is crossing the street" on the freeway.
Musk spoke about this (JRE Podcast) and he compared the camera to eyes. He said that the only way we'll ever achieve full self driving is creating a system that is at least as good or better than human eyes.
That’s a fine goal, but let’s not then call a system that has nowhere near the acuity, resolution, or dynamic range, and is not hooked up to a (trained) human brain, “Full Self Driving”.
> if you reduce the number of variables at play... that means removing the human element altogether (and even then, mother nature may have other plans), meaning self-driving cars should have their own roads.
Why a pipe dream? It's one of the most sound solutions to the problem at the moment.
This is the model China went with its self-driving pilot projects for example.
>I've always been of the mind that this technology can only work if you reduce the number of variables at play... that means removing the human element altogether (and even then, mother nature may have other plans), meaning self-driving cars should have their own roads. Pipe dream, I know.
That’s not the only way of reducing variables. I think the key is rethinking how self-driving cars can fit into a broader ecosystem of transportation, instead of narrowly thinking about self-driving cars as a one to one replacement for human driven cars.
Self-driving or not, cars are a terribly inefficient way to move people. 10 buses that are only a third full replaces 260 single commuter cars. A light rail running at 50% capacity carries more people per hour that 4 full lanes of highway [1]. Even if we build more and more roads, that just causes more traffic, not less [2]. The problem with public transport is the “first mile, last mile” problem. Unless you are in a dense well-designed environment, it is a huge pain getting from your house to the public transport, and from the public transport to your office/store/ballet recital. People don’t want to drive 3 miles and park their car at the light rail station, ride to the closest stop, take a bus 2.5 miles, and then walk the last half mile to the office. But if a self-driving car could pick you up at the curb and take you right to the lightrail, and another could pick you up from your stop and drop you right at the office, suddenly that changes the calculus. Why sit on the highway for 90 mins in your self-driving car, surrounded by 100,000 other people in self driving cars, when you could make the same trip in 40 mins using the light rail and a couple Uber-driverless.
In that situation, you can reduce variables by having self driving cars drive at lower speeds (in residential and urban areas) and by having a few dedicated lanes on strategic artery roads to move people to central locations for transport on bus/lightrail/train, much like we have dedicated lanes for buses in some places.
Once you have some basic infrastructure like that, then it makes more sense to reduce the massive amount of streets and parking wasting space in dense urban environments. If you want a car for your suburban life, fine, but you have to park it outside downtown and take the lightrail in and grab a little self-driving car to get where you want to go.
Almost certainly due to the fact that these vehicles tend to be surrounded by flashing lights in 'random' shapes and colors, and that is confusing the cameras and AI.
Flashing lights are common and important in the real world. If they confuse AI, then AI isn't ready. No excuses, either figure it out or don't claim you are ready.
It would help if Tesla was fined $1B for false advertising then... Even naming the whole thing "Autopilot" qualifies as punishable offence, in my opinion.
It is so insane to me how people keep putting the burden of understanding the capabilities and limits of a computer and engineering system on consumers, who aren’t engineering or computer experts, and not putting the burden on the company full of engineers and computer experts to properly and adequately inform the consumer. Or the burden of consumers, that are not legally trained in any way, of having a full understanding the implications of legal matters, and not the company with a staff of lawyers, in properly and adequately explaining it.
I can only think of three reasons: incompetence, arrogance or cost. I tend to think it's mostly arrogance in believing the self driving "problem" can be purely done with cameras and software. Cost as well as availability of parts became an issue this year as well.
Radar has too many issues detecting stopped objects. A parked car vs a manhole cover. Both Tesla and Comma.ai think solving self driving is purely a software problem and can be solved with just vision. For instance, humans, the only known driving implementation currently, drive with only vision. This videos goes over sensors briefly.
Radar has serious issues detecting stationary objects. Which is why early automatic cruse control systems had minimum speeds, they simply don’t work in stop and go traffic.
Lidar on the other hand has no problem with stationary objects, but cars can’t depend on it in all weather conditions. Which means full self driving cars need to be able to operate based on vision like we do. I would expect all level 5 systems to use Lidar especially if costs continue to drop, but it’s not as clear cut as you might think.
The advantage of computer integration is that self-driving vehicle can use all three technologies to make the best use of everything and compensate for any technology's failures.
Unless they remove them and rely on only one type of sensor, which is exactly what Tesla has done. It's like complaining that each sense has issues, then removing all but one. Multimodal integration is literally how we use our senses normally.
I am not convinced any self driving system is going to be dependent on RADAR in 20 years. Tesla was using RADAR and stopped because it just don’t really solve any specific problems well.
RADAR is great for collision detection and early breaking systems because drivers don’t depend on them. If they fail in an early breaking system noting worse happens. But self driving systems can’t accept failures like that. These systems would need to stop if their camera/Lidar system failed so what’s it actually useful for?
From my understanding, the problem is the software doesn't know if that radar object is a car or a sign or an overpass. So it assumes that moving objects are cars, and makes a good guess at stationary objects. Occasionally when it is wrong, there would be unusual braking for signs, or just plowing into stationary cars.
Detection is overstating things, current RADAR systems don’t have much in the way of resolution. So a manhole and an object in the road can be effectively indistinguishable in the raw sensor data.
RADAR systems are rapidly improving, but there’s a reason everyone shows Lidar data over RADAR.
But you are right on the point of there being no silver bullet sensing technology. All of them have their own very glaring weaknesses.
Radars are superior for simple collision avoidance for them being able to directly sense the velocity, but they are obviously useless for lane keeping.
Lidars are super unreliable from both internal factors, and the environment. Basically, a splat of dirt, and a collision.
Cameras are subject to all uncertainties of a CV system, and smog, smoke, and weather as well.
The main reason is cost. LIDAR has issues and limitations, but so does every imaging system. A LIDAR/camera system is far superior than a camera only system, and LIDAR does far more thing better than cameras, than cameras do better than LIDAR (judging distance, reduced computing effort, detecting reflective surfaces and surfaces with 2D images on them).[0] But LIDAR was running $70k per vehicle, which is just not feasible. But thanks to technological development and economies of scale, the price of LIDAR (and the size) has drop precipitously, and now can run below $1,000. [1][2]
Now in the long run, for true driverless cars, LIDAR will be rendered obsolete by sufficiently accurate AI interpretations of camera feeds. But in the medium term, I think they are key to filling in huge weaknesses camera only systems have, and freeing up computing power currently being used by trying to judge “depth perception” using cameras, to be freed up to better deal with other things cameras are actually good at like color/text recognition for reading traffic signs, facial recognition for determining if a pedestrian sees you, differentiating between hard and soft objects, etc.
My bet is on cost. It's the stated reason not to use lidar and it only makes sense to apply to radar (not knowing the cost of a radar sensor compared to camera - please share if you do).
Elon may not be a great engineer but he is a genius businessman in many aspects so that would also make sense.
This is quite subjective. My opinion is not very representative, but I for one find teslas horrifically bulky and ugly (and most other cars, in fact). They all look like gigantic broken eggshells. On the other hand, a lidar rig looks very cool.
IIRC, in this[1] video Andrej Karpathy mentions radar was mostly redundant, and in addition was occasionally harmful (sending conflicting signals vs. video), so they got rid of it to reduce cost and complexity.
Yeah, nothing makes sense except cost. Even for these kind of level 2 system, they should be forced to have camera + lidar + radar. There is no reason to not use as many data sources as possible. If it is to expensive, work on that. Batteries used to be more expensive too but they focused on that. Spend some of your stock on lidar/radar acquisitions.
System redundancy in safety-critical systems exists for a reason. Either their engineering team refused to see this or they were unable to make Product/Management/Regulatory people see it.
AEB is quite finicky and needs quite a fair bit of calibration and mechanical alignment, still like the article says can suffer from false positives.
But although LIDAR is still prohibitively expensive it is possible to design ADAS that can integrate stereo camera, RADAR and long range sonar, that would at least give you three sources of data for some margin.
> What wavelengths do other cars' radars operate at?
The spectrum used for AEB in most transceivers is 76 GHz to 81 GHz (so between about 3.9 and 3.7 mm) some older implementations also worked at 24GHz i think.
Compared to Teslas, they statistically do not. You may make an argument that they do compared to other makes of manually driven cars, but you'd have to adjust for all sorts of variables, and a YouTube search won't do that for you.
OK, can you provide some statistics with sources? How many police cars were hit on the side of the road last year? What is the breakdown of the Make/Model that hit them?
I wonder if this is something to do with temporal integration periods in the HDR or AGC processing in the camera, in which case it might be an algorithm developed by an camera imager supplier (for example someone like OmniVision) rather than Tesla themselves. Does anyone know which imagers Tesla uses in the cameras on their cars?
