No, a Tesla didn't predict an accident and brake for it

Topic: 

You may have seen a lot of press around a dashcam video of a car accident in the Netherlands. It shows a Tesla in AutoPilot hitting the brakes around 1.4 seconds before a red car crashes hard into a black SUV that isn't visible from the viewpoint of the dashcam. Many press have reported that the Tesla predicted that the two cars would hit, and because of the imminent accident, it hit the brakes to protect its occupants. (The articles most assuredly were not saying the Tesla predicted the accident that never happened had the Tesla failed to brake, they are talking about predicting the dramatic crash shown in the video.)

The accident is brutal but apparently nobody was hurt.

The press speculation is incorrect. It got some fuel because Elon Musk himself retweeted the report linked to, but Telsa has in fact confirmed the alternate and more probable story which does not involve any prediction of the future accident. In fact, the red car plays little to no role in what took place.

Tesla's autopilot uses radar as a key sensor. One great thing about radar is that it tells you how fast every radar target is going, as well as how far away it is. Radar for cars doesn't tell you very accurately where the target is (roughly it can tell you what lane a target is in.) Radar beams bounce off many things, including the road. That means a radar beam can bounce off the road under a car that is in front of you, and then hit a car in front of it, even if you can't see the car. Because the radar tells you "I see something in your lane 40m ahead going 20mph and something else 30m ahead going 60mph" you know it's two different things.

The Tesla radar saw just that -- the black SUV was hitting the brakes (possibly for a dirt patch that appears to show on the video) and the red car wasn't. Regardless of the red car being there, the autopilot knew that if any other car ahead was braking hard, it should also brake hard, and it did. Yes, it's possible that it could also calculate that the red car, if it keeps going, will hit the black car, but that's not entirely relevant -- it's clear that the Tesla should stop, regardless of what the red car is going to do. Tesla reported in their blog about how they were doing more with the radar, including tracking hidden vehicles with it. The ability of automotive radar to do this has been known for some time, and I have always presumed most teams have taken advantage of it. You don't always get returns from hidden cars, but it's worth using them if you do.

In the future, we will see robocar systems predicting accidents, but I am not aware of this being announced by any team. All robocars are tracking all objects ahead of them, for position and velocity, and they are extrapolating their velocity and predicting where they will go. Those predictions would also include detecting that vehicles might hit (if they continue their current course) and also if they could not avoid hitting at a certain point. If an imminent accident is predicted, it would make sense to know that and also react to it in advance. A car might even be able to predict a bit of what will happen after the accident, though that is chaotic.

A system like that would outperform the autopilot or any automatic emergency braking system. Presently, those systems largely track objects in their lane. They don't brake because cars are stopped in adjacent lanes, because that would mean they could not work in traffic jams or carpool lanes when there are lanes at different speeds, and they could not deal with stalled cars on the side of the road.

However, if you saw what the Tesla saw from the lane to the right, it would still be a very smart thing to brake. Tesla has not commented on this, but I presume its system would not have braked if it had been in that lane, at least not braked before the accident. It might brake because other cars like the red car immediately moved into the right lane.

What the autopilot did is execute what is called AEB or automatic emergency braking -- sometimes also called FCA or forward collision avoidance. This is a common feature on many cars, and in fact car makers and NHTSA have already agreed it will be standard equipment in all cars early in the 2020s. It is useful, and the Tesla did it very well (not all AEBs will react to hidden vehicles) but it is not something new or special to the Tesla.

Comments

I was expected your comment, your eyes and ears are everywhere. I agree with you, but anyway the autopilot discovered that something could happens. This is the important point. The black car was hitting the brakes, the red car did not react in time (man mistake, to fast for the distance between cars or to slow or inadequate reaction). So autopilot brake. May be the man behind autopilot,if it would have not autopilot, would react late and kicked the other car too. This is the important fact. Really autopilot avoid an accident that a man (red car) did not avoid. Even as you wrote, autopilot brake, did not change lines that could take to another accident. Robocars mixed with cars will not avoid accidents, just make less, so better go fullrobocars ASAP.

The point is that while the Tesla did the right thing and a good thing, this is just standard "automatic emergency braking" -- a feature which has been present in many car models for many years. So long in fact that NHTSA and automakers declared it will become standard on all vehicles sold early next decade.

People were writing this up like some sort of miracle of AI, with the Tesla knowing two cars were going to crash and braking because of that predicted crash. That is not the case. It was just braking because the radar saw the black SUV braking hard. Note the Tesla has an independent AEB system on top of the autopilot, but the autopilot also contains a braking system and it was the autopilot that engaged the brakes.

Many AEBs out there are not radar based and don't trigger on hidden vehicles. In particular camera based AEBs like the MobilEye do not do AEB on vehicles not visible to the camera.

SO, autopilot is nearer or likes more a robocar, than what you call AEB. Which will be better,in the next decade as you say, robocars or more cars with AEB?. Robocars will be cheaper than today cars with some sensors, and some robot capabilities added. Better push for full robocars, than continue adding robot capabilities (and price increase) to actual cars.

Most car companies, for decades now, have been working on ADAS and planning more ADAS, such as AEB. AEB (and even FCW) has been found to be so useful that it is planned to be required standard equipment in all new cars in about 5 years.

At the same time, many teams (including many at car companies) are working on robocars. This is also good. Robocars will also brake when obstacles appear before them, including other breaking vehicles. You could call it AEB but really it is part of the whole system.

The Tesla autopilot is a hybrid. It is derived from ADAS tools (particularly the MobilEye in the versions up until now.) It hopes to become a robocar. There is debate if making a robocar from incremental ADAS is the right approach or if a completely independent path is right. Even in the latter case, you certainly learn things from doing ADAS, and Tesla has been doing that.

It became very clear last year with the fatality that Tesla's system is still mostly ADAS tools, combining the techniques of AEB, Lanekeeping and ACC. But they want to make it more.

Like you, I find the article title and the subsequent misunderstandings cringeworthy. But I don't know if the capability to avoid an accident like this is as routine as you seem to be implying. There's an impressive lack of youtube dash cam video to back up the assertion that it's a common capability of AEB equipped cars. And the AEB test video that I've seen suggests to me that the large majority of AEB systems on the road can not pull off this trick of avoiding a hidden car accident with a large safety margin. Most of them don't seem to even avoid collisions when used at highway speed; they just lessen the severity.

Yes, there are varying levels of performance. The camera based ones (MobilEye and cousins) can't see a hidden car of course, only radar sees that. Again, nobody sees or predicts an accident like this that I know of. Most early AEBs and ACCs where radar but now it's cheaper with a camera and you get other functions from the camera (like lanekeep) so people have been going that way.

In this case, the Tesla was fairly far behind. It signaled the 3 beeps and started braking but is also asking the driver to be ready to brake, which I suspect they did.

that seems to be something good. A man driving car, could have been to near to stop.

I must correct my earlier impression. The Tesla was following the red car by just over a second, which is actually somewhat close. AEB did a good job for the owner here.

Add new comment