[supernova87a]

"Witnesses claim the man behind the wheel and his passenger were asleep with their seats fully reclined, as their Tesla travelled up to 150 km/h on the freeway near Ponoka..."

Given this and the man's other past incidents, the Tesla was only a minor contributing factor in the guy's global stupidity and danger that he inflicts on the world daily?

Edit: Btw, what would be the end accident scenario here? Presuming that the car wouldn't hit any other car on the road, would it lose control on a curve that was too tight for the speed, or fail at the end of the freeway or something? Would it alarm and then come to a gradual stop? Not a Tesla owner here.

[gambiting]

"Presuming that the car wouldn't hit any other car on the road" - it absolutely would though. The tech is good enough to avoid hitting another car that slows down in front of you, but if someone was broken down(and fully stopped) in your lane, the car will hit them at full speed. It might start breaking just before impact, but not fast enough to slow you down from 150km/h. No self driving or adaptive cruise tech can save you from this situation, because they are all trained to ignore fully stationary objects. I've also seen situations where road markings disappear on the road for whatever reason and Tesla goes from being 100% confident and in control to "take over the wheel NOW" in about 1 second flat. Best case scenario here is that the car would eventually realize that you aren't in control and come to a gradual stop - the problem there is that now he is fully stopped on an active highway and chances of getting plowed from the rear are astronomical.

I think the biggest problem with all of these self driving systems is that in cars they can require you to be fully 100% aware of your surroundings and take over in a split second. People compare it to plane autopilot, but I don't think that's quite right - in a plane, you will have at least a minute or two before you hit the ground, even if the plane takes a straight nose dive. In a car, you go from being fine to being headed for a head-on collision in no time at all.

[anonu]

This [1] was the famous 2016 crash where a Tesla broad-sided a basically stationary semi-truck. Are you suggesting Tesla hasn't solved this type of problem yet?

[1] https://www.reuters.com/article/us-tesla-crash/tesla-driver-...

[gambiting]

Yes, of course Tesla hasn't solved it yet, because no one has, and in fact it probably isn't possible to solve without some magic tech that doesn't exist yet - it might be possible to fix with LIDAR, but obviously Tesla doesn't have it.

All of these systems have to ignore stationary objects ahead of them, or the car would be emergency braking in too many daily situations to be usable.

Simple case, at 150km/h you are travelling at 41m/s. The stopping distance at that speed is roughly given as 130m[0]. So the car would need to see and recognize objects at least ~150m ahead to stop autonomously from that speed. That's just physically not possible, the cameras on the Tesla don't have enough resolution to do such a recognition. Instead, a radar-based distance measure is used - but again, even if the car detects that you are rapidly approaching "something" 150m ahead of you, that information is nearly useless. At that distance, you cannot differentiate between a stopped car 150m ahead of you, an overhead sign, or a large rock next to your lane which poses absolutely no danger whatsoever - it all gives the same signature. LIDAR doesn't have that range either. And then of course even if the car could reasonably detect that there is something ahead of you that you are absolutely 100% definitely heading towards, it has no idea if the road doesn't curve in such a way that you would avoid it. Famous case of adaptive cruise systems freaking out at bends, because according to the radar/image recognition you are CLEARLY heading for that telephone pole standing next to the road - but of course the road curves so you aren't actually going to hit it. Problems like that.

[0] https://www.random-science-tools.com/physics/stopping-distan...

[diggernet]

Sorry, coming in late, but gotta object here...

My Subaru has no problem detecting stationary objects. It uses stereo cameras, for which object detection and locating the object is a thoroughly solved problem, whether the object is moving or stationary. No magic needed. It just works. For example, there is a sharp turn near my home with prominent turn signs along the curve, which puts them directly in front of you as you approach the turn. Those stationary signs quite reliably set off my collision warning if I'm approaching them too fast.

Tesla has the problem because it uses radar for object detection. The radar can tell them the presence of an object, but cannot tell them the object's location. An overhead sign or overpass or whatever looks the same as an object in the road. So to avoid countless false positives, they need to ignore signals from stationary objects because they chose to use radar. It's a self-inflicted problem that other vehicles do not have.

[jessaustin]

Aren't there fully self-driving vehicles undergoing testing on normal city streets with stoplights? We could say that's a different mode than high-speed highway driving, but there are a number of locations where those modes blend into each other. Are the experimenters extra careful to avoid those locations? How long before the tech exists to address this issue?

[gambiting]

Sure there are, and yes, they are different sets of problems though. Have a look at the British Tesla Driver Youtube channel, some of his videos are eye opening. Basically the car is in full autopilot mode, approaches an intersection, correctly slows down, waits for its turn, starts moving.....and in the middle of the turn goes BEEP BEEP BEEP and disengages entirely because the road markings aren't there and it wasn't entirely sure where to go. And now of course you're in a moving vehicle that's heading for a collision with someone else and requires IMMEDIATE attention to continue. One could(and I'm sure will) argue that the system "shouldn't be used this way". But that's a moot point, if the system is there and lets you do this, then people will use it this way.

"How long before the tech exists to address this issue?" I'm not sure if that's a problem with tech as such. We have fantastic cameras, yet famously Google's best image recognition algorithm just couple years ago would reply, with 100% confidence, that a sofa in a zebra print is in fact a Zebra, after all the stripes are there, it has 4 legs.....it must be a zebra.

So in my(personal) opinion, self driving will face the same challanges image recognition has faced - we will rapidly get 90% of it right, then the last 10% will be a massive pain to get right for decades if ever.

[Isinlor]

It's an active research field. E.g. from October 2020: Calibrating Deep Neural Networks using Focal Loss: https://arxiv.org/abs/2002.09437

> Miscalibration - a mismatch between a model's confidence and its correctness - of Deep Neural Networks (DNNs) makes their predictions hard to rely on. Ideally, we want networks to be accurate, calibrated and confident. We show that, as opposed to the standard cross-entropy loss, focal loss [Lin et. al., 2017] allows us to learn models that are already very well calibrated. When combined with temperature scaling, whilst preserving accuracy, it yields state-of-the-art calibrated models. We provide a thorough analysis of the factors causing miscalibration, and use the insights we glean from this to justify the empirically excellent performance of focal loss. To facilitate the use of focal loss in practice, we also provide a principled approach to automatically select the hyperparameter involved in the loss function. We perform extensive experiments on a variety of computer vision and NLP datasets, and with a wide variety of network architectures, and show that our approach achieves state-of-the-art calibration without compromising on accuracy in almost all cases.

Calibration will be practically solved in couple of years. Then a bit longer for addressing adversarial robustness.

[camjohnson26]

There have been multiple reports of Teslas on autopilot hitting stationary roadside emergency vehicles.

[fiddlerwoaroof]

My Tesla has regularly come to a full stop from freeway speeds in response to traffic ahead. This may not be 100% reliable, but it’s not true that “the system is trained to ignore stationary objects”.

It also beeps when it thinks I’m going to hit a stopped car and other things that demonstrate that this claim is false.

[gambiting]

The question was specifically about this situation as described in the article - the car moving at 150km/h. No, your Tesla wouldn't stop in time for a fully stopped car in that situation.

It does slow down for traffic, but I think many people don't realize something - it works with moving traffic, because then it can definitely recognize that you are approaching a car(or at least something that moves in the same direction you do), so it knows it has to slow down for it.

>>It also beeps when it thinks I’m going to hit a stopped car and other things that demonstrate that this claim is false.

Read up on it, I'm sure the upper limit for this function is when the delta speed is <50km/h. It won't work with a delta of 150km/h because it's physically not possible.

>> but it’s not true that “the system is trained to ignore stationary objects”

These are not my words, that's exactly what Tesla said after the "trailer across a highway" accident, saying that of course they have to ignore stopped objects otherwise the car would emergency brake for overhead signs since they reflect radar the same way a stationary car does - at large enough distance there is no difference.

[fiddlerwoaroof]

>> These are not my words, that's exactly what Tesla said after the "trailer across a highway" accident

>> Read up on it, I'm sure the upper limit for this function is when the delta speed is <50km/h. It won't work with a delta of 150km/h because it's physically not possible.

I think both of these have the same explanation: if you want to release a feature like this specced at a 50kph delta, you design for a safety factor of 2-3 (100-150) so that you can be confident that it’s safe at 50kph. The claim that “it’s physically impossible” doesn’t make sense to me: humans drive safely with such deltas using only “a video feed” and sound.

Anyways, I’m fairly certain I’ve come to a stop on autopilot from at least 60mph (100kph).

[gambiting]

>>The claim that “it’s physically impossible” doesn’t make sense to me: humans drive safely with such deltas using only “a video feed” and sound.

My logic is as follows - at 150km/h, you are covering 41m per second, and an approximate stopping distance from that speed is about 130m. Human eyes are much better at recognizing objects from a distance than computer based vision is, and Tesla is in fact relying on cameras for its forward object detection, plus a rudimentary distance-based radar. There is no chance(that's why I said "physically impossible") that whatever camera is mounted in the Tesla can reliably recognize an object(and tell that it's stopped!) at 130m. Of course the system needs to do the processing, make a decision, send a signal to the brake actuators and actually engage them. Let's be generous and add a full second to this - so to stop from that speed Tesla would need to recognize a car, identify it as a hazard, and make a critical "all brakes at maximum strength" decision from 170m away. There's no chance.

>>Anyways, I’m fairly certain I’ve come to a stop on autopilot from at least 60mph (100kph).

Ok, but there will be an upper limit to this, and I'd love to know what it is. I know that Deimler's solution only guarantees full avoidance at deltas up to 50km/h, and "reduced" impact at higher deltas - it just doesn't see far enough. Tesla's technology is fundamentally very similar, so I'd love to know what they consider as reasonable distance for full autonomous stop.

[vmception]

Tesla's have crashed and obliterated their owners, even right here in California on major highways on the way to their Silicon Valley office park.

Alarmed engineers taking the exact same route to reproduce the bug successfully.