Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Autopilot: Crashed onto a stopped Switzerland Van

This site may earn commission on affiliate links.
....Unfortunately, many Tesla drivers do not understand the capabilities and limitations of TACC and AP. I do not place all the blame for that on them. Tesla deserves much of the blame, in my opinion, because Tesla does not appear to spend enough time educating owners when they buy their car. ....

Exactly. I would go even further. To me it looks as if Tesla's marketing presents the AP as far more capable than it is, just to start presenting all sorts of disclaimers in the manual that tell you that the car actually doesn't do what they make the public believe what it can do.

May I point out what Tesla says both on the German version of their Swiss homepage as well as on the German homepage as regards AP?

They say:
Lenkautomatik mit verkehrsadaptiver Geschwindigkeitsanpassung
Aktive Sensoren in Kombination mit GPS und hochauflösenden Digitalkarten bilden ein Aggregat sich gegenseitig überwachender Systeme. Dank ihnen kann das Model S Autobahnen auch bei Kolonnenverkehr vollkommen autonom befahren - ohne von der Spur abzuweichen.

This means that thanks to its sensors etc. etc. the Model S will operate on the freeway even in heavy traffic "vollkommen autonom" = fully autonomous. (There isn't even an asterisk with a disclaimer.) Sorry, a vehicle that requires driver input isn't "fully autonomous" in my book, let alone a vehicle that will smash into a stationary vehicle that's sitting in your lane.
I don't know about Swiss law but under German law this looks like a clear case of deceptive advertising and both their customers as well as competitors could take legal action against Tesla.
 
  • Like
Reactions: Prefect
When your moving at any speed over a crawl, the ultrasonic sensors are useless at detecting anything in front of you. They only have a range of about 16 feet max. This is adequate to know if a car is in the next lane in your blind spot, but not for forward view. For forward sensing on the road, it uses the camera and the radar. The camera today only detects lane markers and reads the speed limit signs. The radar is the primary sensor for detecting cars in front of you.

The problem is the radar can detect things that are moving from a doppler signature. The radar detects the road surface by knowing how fast the car is going and anything moving at the exact opposite velocity as the car is considered road. For example if you're doing 60 mph, the road is going -60 mph as far as the radar is concerned. Anything that is not moving at -60 mph is assumed to be another car. The radar can tell the speed of that other object and if it is directly in front it sets it's speed to keep x distance from that object.

The radar does not see like a human. In fact the optical camera probably doesn't see like we do either, it's probably tuned to do its job and not see much else. The radar sees and works out everything based on its relative position in space and the differences in velocity. When it sees a car for the first time, it that car isn't moving, it sees road, not a car. The van wasn't camoflauged because of its color or shape, it was camouflaged because it wasn't moving the first time the car saw it. If the van started moving right after the car saw it, it would have figured out it was a vehicle and set a distance, but it being stationary, the radar couldn't detect it was not clear road.

To solve this problem, Tesla needs to add new sensors that aren't there. The current sensor suite is not capable of gathering the data needed. Doppler radar can only tell you within it's range and cone if there is something moving and how fast it's moving, that is the only data you get. It's fixed and doesn't move like aircraft radars that sweep a pattern. Other types of radar could tell you the angle of the object with respect to you, which you could use to figure out that the rear end of that vehicle is the wrong angle to be the road going up a hill, but again, that's a different type of sensor.

The schematic of the AP 2.0 system that has been floating around is a 3 camera system with an ultra wide angle, and two narrower views. With that system you can gather a lot more information about your environment including pedestrians crossing in front of the car, traffic lights, and detecting stopped cars in front of you.

Until your post, I did not understand the technical shortcomings to the AP and particularly AEB. Thank you for consolidating that information and making it understandable. Hopefully improvements will come sooner rather than later.

On a positive note, it appears that the accident in the video was at a relatively slow speed with minor impact. Was it stated anywhere what speed the Tesla was traveling as the lead car moved out of the way?
 
...this just TACC -- only the adaptive cruise control -- not the autopilot...

It is just like saying legendary soccer Pele won the match, NOT Brazilian National Team.

A team has many members and although it might be true that without one single player Pele, Brazilian National Team might have not won but saying one single member Pele won the match or Brazilian National Team won the match-- are both correct.
 
...TACC knows that the lane is blocked...

That is not what current design is. Eventually, the design will evolve but not right now.

Current design is to lock onto a leading car (Black Wagon) and follow it tightly whether there is any steering or not, whether there's an obstacle (Stopped Schubiger Van) inside the lane or not.

We need to take advantage of current capabilities and limitations and don't act according to future designs that are not available right now.
 
  • Like
Reactions: GSP
I think if he was using full autosteer with TACC (i.e., AP), his car would have slowed sooner, and perhaps stopped behind the truck.

What you are describing is a future design that are not available right now.

His car did slow down but not because the system took into account of the Stopped Schubiger Van, it slowed down because it locked onto the leading Black Wagon's speed.

Someday, Tesla AEB Automatic Emergency Braking might be redesigned to brake to a stop, but the manual currently is saying that its job is not to stop a car or to avoid a collision. Its current design or current job is to reduce the force of collision, and it is still a collision.
 
This means that thanks to its sensors etc. etc. the Model S will operate on the freeway even in heavy traffic "vollkommen autonom" = fully autonomous.
I can't read German. When I ask Google to translate "vollkommen autonom" it shows "completely autonomous", essentially exactly what you wrote.

I agree, that is not only inaccurate but dangerously deceptive language for Tesla to use. They need to rein in the marketing exaggeration. I have heard Elon talk about the parallels between existing autopilot functions on aircraft and the Tesla Autopilot function and it sounds reasonable. But the marketing language seems to have gotten ahead of reality in this case.
 
Yet in another comment he states "At least the ultrasonic sensors should have seen the object right ahead. They reach 20m if I recall correctly".

That shows that he does not fully understand how the current Tesla ultrasonic sensors work, since Tesla clearly states that they only have a a 16 ft (about 3 meter) range, and in any case TACC relies on the forward facing radar and not on the ultrasonic sensors to maintain distance and detect other vehicles in front of the car.

I agree that people have a lot to learn about what AP can and can't do. Just one small nit, 16 feet is about 5 meters.

In any case the ultrasonic sensors can't sense further than about one car length (give or take a bit). 20m would only be 4 car lengths, which is better, but definitely not enough for deciding to make an emergency stop at highway speeds.

Until your post, I did not understand the technical shortcomings to the AP and particularly AEB. Thank you for consolidating that information and making it understandable. Hopefully improvements will come sooner rather than later.

On a positive note, it appears that the accident in the video was at a relatively slow speed with minor impact. Was it stated anywhere what speed the Tesla was traveling as the lead car moved out of the way?

Your welcome, I didn't fully understand what was going on until I started looking into the patterns of how these accidents were happening and then what the limitations on the sensors were. Tesla doesn't explain it very well, but it is a complex problem to explain. People automatically assume that computers with sensors sense the world pretty much like we do. And because computers are so much faster than a human brain, they must always be better at making decisions in an emergency.

Humans are great at pruning information, much better than computers. Computers generally have to think through everything before making a decision, which is why chess programs can be very slow. A computer chess program will sit there running through moves that a human would discount so fast they wouldn't even register it consciously.

But automated machinery also see the world differently than we do and have different information we do. The three camera setup coming with the next generation of AP has wide angle and narrow cameras with more closely resembles human vision. We see close to 180 degrees (some people with better peripheral vision can see more than 180 degrees), but we only see clear detail in the middle of our vision. The peripheral vision is an early warning system that will alert us to something going on over "there", but we don't know what that is until we pivot our center vision to tune it in. We also have two eyes which gives us a 3D view of the world. A single camera misses depth details we pick up with the two eyes.

Once cars get the right suite of sensors and the software is programmed to handle all the scenarios drivers run into, autonomous driving will be possible, but the tech isn't there yet. I've heard even Google's test cars have problems. One was stuck at an intersection for a long time because other drivers weren't following the rules.

http://www.nytimes.com/2015/09/02/t...erless-cars-fault-its-other-drivers.html?_r=0
 
The camera today only detects lane markers and reads the speed limit signs. The radar is the primary sensor for detecting cars in front of you.

The camera also lets you know if the vehicle in front of you is a large truck, bike, or a car. I don't think it's entirely clear as to how much of a role it has in adaptive cruise. In my own experiences it doesn't seem to play any role, but others have said the car has slowed for shadows in certain lighting conditions. Something that wouldn't happen if it wasn't playing any role at all.

In any case the car as it's currently equipped can't detect stopped cars. Like you I expect that to change with the next generation of Autopilot that has three front cameras.
 
Having just driven drove an autopilot loaner for 4 days, I can easily see how you can come to have a lot of confidence in it, and be reliant on it, and expect it to do a lot more than it is actually capable of doing. This accident is a prime example of that. The driver expected AP to do more than it is capable of, and is at fault, but I also agree that Tesla may share some blame for overselling the current system. Then again, it's a pretty damn good system but still it's far from perfect and I don't think it's going to get any better until they add more hardware.

I don't know about Swiss law but under German law this looks like a clear case of deceptive advertising and both their customers as well as competitors could take legal action against Tesla.

I'm quite certain under Swiss law they could take legal action against Tesla too. The chances of success might be better under German law though which is probably what you mean.
 
Exactly. I would go even further. To me it looks as if Tesla's marketing presents the AP as far more capable than it is, just to start presenting all sorts of disclaimers in the manual that tell you that the car actually doesn't do what they make the public believe what it can do.

May I point out what Tesla says both on the German version of their Swiss homepage as well as on the German homepage as regards AP?

They say:
Lenkautomatik mit verkehrsadaptiver Geschwindigkeitsanpassung
Aktive Sensoren in Kombination mit GPS und hochauflösenden Digitalkarten bilden ein Aggregat sich gegenseitig überwachender Systeme. Dank ihnen kann das Model S Autobahnen auch bei Kolonnenverkehr vollkommen autonom befahren - ohne von der Spur abzuweichen.

This means that thanks to its sensors etc. etc. the Model S will operate on the freeway even in heavy traffic "vollkommen autonom" = fully autonomous. (There isn't even an asterisk with a disclaimer.) Sorry, a vehicle that requires driver input isn't "fully autonomous" in my book, let alone a vehicle that will smash into a stationary vehicle that's sitting in your lane.
I don't know about Swiss law but under German law this looks like a clear case of deceptive advertising and both their customers as well as competitors could take legal action against Tesla.
Same section in English says this, nothing about fully autonomous:
Automatic steering with traffic aware cruise control
Active sensors combined with GPS and high resolution digital maps comprise a mutually reinforcing system that pilots Model S along the highway, staying within your lane, even in stop and go traffic. Real time feedback from the Tesla fleet ensures the system is continually learning and improving upon itself.
Model S | Tesla Motors
 
..."vollkommen autonom" = fully autonomous...

Lawyers might argue "vollkommen autonom" = fully autonomous here is to be understood within the context of the paragraph: It can fully drive itself in certain conditions:

e3joxpl-imgur.gif


but not in other conditions.
 
From the driver's response, it appears that both driver and Tesla's log confirm both AP and TACC were on:


D7dImrY.jpg

Thanks, that's new info, previously he was convinced that only TACC was engaged. If true, the logs now tell a different story.

If that's really the case we will get lots of similar situations in the time ahead, until somebody dies. A system that works perfectly "most" of the time is too dangerous to release to the masses, Tesla will agree then. Or will be forced to deactivate the system by law makers around the world.

So enjoy AP as long as it lasts and be safe out there!
 
  • Like
Reactions: bhzmark
From the driver's response, it appears that both driver and Tesla's log confirm both AP and TACC were on:


D7dImrY.jpg

That's too bad and blows my theory. His first remarks and title of his video just stated adaptive cruise. He is still a little confused as tacc is a subset of ap.

I'm surprised that auto steering couldn't work with tacc to calculate: stopped obstacle in lane + can't steer around it = slam on brakes.
 
That's too bad and blows my theory. His first remarks and title of his video just stated adaptive cruise. He is still a little confused as tacc is a subset of ap.

I'm surprised that auto steering couldn't work with tacc to calculate: stopped obstacle in lane + can't steer around it = slam on brakes.

It doesn't detect stopped objects. Why would it steer around something it doesn't detect?

I agree the term "auto pilot" should not be used. However, if used as intended, it is a wonderful system and will reduce accidents.
As any tool, if not used as intended, the results can be poor if not worse.
 
It does detect stopped objcts. I stop behind stopped cars, even cars
It doesn't detect stopped objects. Why would it steer around something it doesn't detect?

Yes it does. And steers around cars parked on the side of the road on a curved lane without lines too.

Full AP (TACC plus Autosteer) should be able to determine obstacle in lane so stop behind obstacle. It does all the time at stoplights with people already waiting at light. I guess the issue here was that only part of the lane was blocked and that perhaps delayed or prevented an instruction to stop behind the obstacle.