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

Fatal autopilot crash, NHTSA investigating...

This site may earn commission on affiliate links.
I would not recommend using autopilot if you don't have a special skill to be ready to take over the system at any time.
Not sure that's a special skill. And besides that's exactly what you agree to when you enable autopilot.

You ARE supposed to be able to take over at any time and you ARE supposed to be paying attention with your hand on the wheel.
 
I


a5AIwl3.jpg


I would not recommend using autopilot if you don't have a special skill to be ready to take over the system at any time.
could you explain what special skill is needed to be alert and to be ready to assume control instantaneously?
 
  • Like
Reactions: X Fan and cdub
I just got Model X with enhanced autopilot for 2 weeks and so far, I am quite pleased as a test pilot.

It still has problem of detecting tall truck from time to time.

I was on autopilot at 35mph approaching a stopped tall truck but I had to manually take over at the very last minute. I took a picture and the dashboard shows that the system does not register the very big, tall, stopped obstacle in front:


a5AIwl3.jpg


I would not recommend using autopilot if you don't have a special skill to be ready to take over the system at any time.


Am I to understand it right, that the TACC would have hit that truck you were following?

How does it do on motorcyclists?
 
All I can say is wow.

I was on TACC and AutoSteer at 35 mph. As MP3Mike said, it is a well known scenario as the truck was already there stationary at the red light.

The system seems to be quite reliable to consistently recognize a stationary vehicle and brake to a stop for the past 2 weeks until now.

The only difference here is: It's not a car in front but a truck, a big tall truck.

It's counter intuitive to me that a big tall truck is harder to be detected than a smaller car.

It is a special skill because you've got to have the discipline and not to let your guard down. It's been stopping for stationary vehicles reliably 99% and I started to get used to it and trusted it then suddenly this 1%!
 
  • Informative
Reactions: EVie'sDad
Most interesting is the Tesla Log:

View attachment 213082

and then

View attachment 213083

and various faults at the time of the crash:

View attachment 213086

View attachment 213087

Some other interesting logs showing what is being tracked:

View attachment 213085


View attachment 213084

other pages indicate pedal positions and cruise speed setting etc.


It was interesting to see that his MS at 13:30 noted lack of hands on the wheel and the collision notations appear at 13:36 (from your two inserts of log text there was a break in time so curious if he put his hands back on or not). If I have the time I'd like to look through the report and check the logs out further. As interesting as the logs are all this is still very sad reading as you know the driver was going to die at some point when the log was being generated.
 
...Autopilot does not mean it drives itself...

I have no argument there.

I am reporting the progress of the situation.

Elon musk said that he fine tuned the radar configuration so that the Florida's accident could be avoided.

As MP3Mike mentioned, Autopilot has a well known limitation that it is difficult to stop for an existing stationary vehicle in front.

For the past two weeks, that stopping for an existing stationary vehicle in front is very reliable in my case as long as the ones in front are not tall and hollowed (The truck bed is tall and under it is empty except for a skinny frame as bumper and then the tires further in front.)

Thus, there has been progress but I am not sure if it's good enough to reliably brake for a tall truck just yet.

Of course, Tesla is still working on it so hopefully the Florida's scenario could be dealt with the technology soon.
 
my car with AP1 sees slows and stops for vehicles that are stopped, what makes you say that the TACC doesn't handle that situation?

You are getting lucky. AP2 might fail 1 of 100 cars, with AP1 it's like 1 in 5000 cars.

I had two instances where AP1 wanted to accelerate into or refuse to start braking for a stationary car -- most commonly a hatchback like a Prius with a bike rack and bikes obstructing the license plate / brake lights.
 
  • Informative
Reactions: croman and Tam
I was not even aware that any Adaptive Cruise Control system would not normally stop when the cars in front are stopped. I'm stumped.
I'd recommend reading the owner's manual and cross referencing with other cars. All radar based ACC cars carry a similar warning about encountering stopped cars.

(I would find the paragraph for you if I weren't on my phone...)
 
  • Like
Reactions: Tam
I was not even aware that any Adaptive Cruise Control system would not normally stop when the cars in front are stopped. I'm stumped.

The manual is very clear with its first sentence below:

3mfm58v.jpg


In the old time, we had conventional cruise.

Then, we got conventional cruise with collision warnings. Those warnings were too annoying and too unreliable so people turned them off after paying good money for the feature.

We then got smart cruise that could speed up and slow down with the lead car in front but it could not brake for a full stop. Human still had to brake for full stop to avoid a frontal collision.

It then took a very long time from that point for the system to be able to brake to a full stop when a lead car would brake for a stop.

But to brake a full stop for an already stationary vehicle waiting for you in front has been a challenge for Tesla and all car manufacturers with radar technology as well.

It's a well known limitation of radar technology.

Thus, Google introduced LIDAR for that purpose.

Tesla insists radar technology is just as good as LIDAR.

Thus, it is where we are after all the hype of self-driving technology that could detect and avoid obstacles.
 
  • Informative
Reactions: anonim1979
It's a well known limitation of radar technology.

Thus, Google introduced LIDAR for that purpose.

Tesla insists radar technology is just as good as LIDAR.

Thus, it is where we are after all the hype of self-driving technology that could detect and avoid obstacles.
This is a limitation of older low resolution radar which uses moving targets to reject clutter. This is not an inherent limitation of radar. Radar can detect stationary objects just fine (it just depends on if you reject that as clutter).
Most new automotive radar systems advertise detection of slow/stationary objects.
Radar Systems | Qamcom Research & Technology
Delphi Electronically Scanning Radar

Google used a high resolution LIDAR, but if you go back to the low resolution LIDARs used in initial cruise control systems they have the same problem, for example the Dynamic Laser Cruise Control used in the 2000-2003 Lexus LS430 (the first laser ACC introduced in the US).

Here is the warning for that LIDAR system which is the same as radar systems:
"Under certain conditions where the vehicle in front slows drastically, or is stopped, the dynamic laser cruise control will neither warn you nor decelerate. The driver must depress the brake pedal to slow down, ensuring collision avoidance or that sufficient vehicle–to–vehicle distance is maintained."
http://shiprockhome.net/My_Homepage_Files/Download/2003 Lexus LS Owner's Manual.pdf
 
  • Informative
  • Like
Reactions: drklain and Tam
Maybe, some of the problem with not reacting to stopped vehicles, is the propensity for false positives, created by overhead signs and overpasses. False positives could be as bad as stationary non-detection.
Right: there are a lot of stopped objects, eg pot-holes, signs, fence posts, concrete blocks, etc. Whereas, previously detected moving objects which subsequently stop are ok.