Blog NTSB Releases Preliminary Report on Autopilot Crash

The U.S. National Transportation Safety Board issued a preliminary report on a fatal March 23 crash involving a Tesla Model X using Autopilot near Mountain View, Calif.

Investigators leveraged data pulled from the car’s computer that shows the driver’s hands were on the steering wheel for just 34 seconds during the minute before impact.

Data also showed that the Model X sped up to 71 miles per hour just before hitting a highway barrier. Tesla issued a release in March that included most of the info in the report. Tesla said “the driver had received several visual and one audible hands-on warning earlier in the drive” and the driver had about five seconds and 150 meters of unobstructed view of the concrete…but the vehicle logs show that no action was taken.”

The NTSB report said the crash remains under investigation, with the intent of issuing safety recommendations to prevent similar crashes. No pre-crash braking or evasive steering movement was detected, according to the report.

“Tesla Autopilot does not prevent all accidents — such a standard would be impossible — but it makes them much less likely to occur,” Tesla wrote in its March post. “It unequivocally makes the world safer for the vehicle occupants, pedestrians and cyclists.”

Read the full report here.

 

[Tiger]

If the driver did not respond to autopilot request to grab the wheel in so and so seconds, why didn't autopilot slow down and take extra caution (e.g., slowing down speed as preventive measure + automatic braking to avoid impact to obstacles at any speed) assuming the driver has fallen asleep or has a medical urgency?

 

[MrAustraliaTax]

If the driver did not respond to autopilot request to grab the wheel in so and so seconds

The last detected driver input was 6 seconds prior to impact.

 

[Tam]

...why didn't autopilot...

Human can figure out how to bypass computer's safety protocol.

For example, during the test drive of stricter GM system of driver monitoring by camera:

"One journalist I was riding with was able to communicate with his editor, log in to Facebook, and then start a Facebook Live session of him driving hands-free. While I held the camera for the majority of the time, the fact he was able to coordinate that many moving parts while ostensibly driving a car is either impressive or incredibly stupid."

 

[dhanson865]

If the driver to you left is going 60 and the driver to your right is going 60 and the driver in front of you is going 62 and you have TACC set to 71 and you don't pay attention to the speed change when the driver in front moves left or right:

• You probably had TACC set at too high of a speed for the conditions
• You are responsible for the safety of the car speeding up

I have no idea what he was doing. But he sure wasn't paying attention. A medical issue would explain it, but if not a health issue then I can only blame in this order:

1. The driver had primary responsibility
2. Caltrans had secondary responsibility (lane markings, barrier reset, signs left out that didn't apply or if they did are very confusing).

In no way does this incident make me feel unsafe driving a car in TN. It might make me hesitant to drive in CA. If I wouldn't feel comfortable taking that route without Tesla Autopilot features involved, why should we expect Tesla Autopilot features to handle the numerous irregularities on that road?

 

[Mark_T]

What we don't know:
- what the driver was doing (if anything) in those last 6 seconds.

We do know that he was not making steering inputs to avoid the obstacle, nor was he braking during those 6 seconds.

That much is confirmed in the report.

 

[mongo]

I've been thinking about this a little more, and perhaps it's time to turn this around on its head so as to not make it seem anti-AP or anti-Tesla.

What we know:
- the driver DEFINITELY DID have his hands on the wheel 6 seconds before impact.

What we don't know:
- what the driver was doing (if anything) in those last 6 seconds.

All we can say is the car sensed torque on the wheel.
I can steer with my knees...

 

[AdamAnt]

Would be interesting to know whether there was a vehicle in front of the crashed Model X that was going 60-70 mph, that then moved to the right lane to avoid the barrier. After which, the model X then resumed cruising speed to 71 mph and hit the barrier. This may explain why the driver all of a sudden grabbed the wheel in the remaining six seconds... "Car in front veers right to avoid already collapsed barrier, driver of Model X reacts to car in front swerving, grabs wheel at last moment, but at 70 mph there is little to no time to react."

 

[MrBoylan]

This report points out a fatal flaw with Autopilot. It can see a car in front of you only if the closing speed is under 30 mph. If not, the system is blind to anything in front of it. That includes street sweepers in China, tractor trailers in Florida and firetrucks in California and Utah, a car at a stop light in Arizona, and a concrete barrier in Mountain View.

Tesla needs to come clean what is causing these accidents so that people know what to be watching for. Guess this is why the IIHS only tests AEB systems (along with Autopilot) at 12 and 25 mph. They don't work at faster closure rates.

Disagree. I've only put a few hundred miles on my Model 3 with Autopilot engaged, but it definitely sees vehicles and obstructions at high speeds. When someone in front of me slows down on the highway, or cuts into my lane, Autopilot adjusts my speed and keeps me at a safe distance based on what I've set. As others have said, when the "obstruction" (e.g., car in front of you) moves aside, Autopilot will adjust your speed (which normally means accelerating) based on your set speed and follow distance. There are cases where obstructions are not seen (like a firetruck and this specific highway barrier) but I don't think there is any data yet that explains why this is happening. But it's definitely not based on (or not solely based on) the speed you are travelling.

 

[mongo]

This may explain why the driver all of a sudden grabbed the wheel in the remaining six seconds...

There was no last second grab.

During the 60 seconds prior to the crash, the driver’s hands were detected on the steering wheel on three separate occasions, for a total of 34 seconds; for the last 6 seconds prior to the crash, the vehicle did not detect the driver’s hands on the steering wheel

 

[cbdream99]

3 seconds traveling 70mph can travel near 100 feet, not enough to come to complete stop even with panic braking for the X, but should reduce impact and chance to steer left/right to avoid hitting the barrier directly, no one will ever know what exactly he was doing at the last moment .. on the other hand, just a swipe on the phone to look up something can easily be few seconds.

With all recent crashes, the same first question always come into mind is - 'What the driver is doing at that moment ? ". Autopilot and Tesla has a long way to go to improve the system ..traffic awareness, road sign and lane recognition, safety..... It's not unsafe to use in all conditions but it has limitations and must pay full attention at any moment ... Just a hour ago, AP didn't detect a semi flat bed with a mobile lifter attach to the end and keep on catching up the speed even the flat bed has stopped for red light.

 

[mongo]

seconds traveling 70mph can travel near 100 feet,

3 seconds at 70 MPH is 316 feet.
264 feet at 60 MPH.

 

[jkennebeck]

There are cases where obstructions are not seen (like a firetruck and this specific highway barrier) but I don't think there is any data yet that explains why this is happening. But it's definitely not based on (or not solely based on) the speed you are travelling.

Tesla autosteer/autopilot systems ignore stationary objects. That's the data. Pay attention carefully when using.

And a new question I just thought of, has this (or any of the recent crashes) divulged what the follow distance was set on before impact? In my experience in my daily commute, when traffic is heavy I am usually at 3 maybe 4, in good weather. I've been on highways at 1 and had my car get too uncomfortably close to cars in front of me that are jammed like sardines. One swerve or blown tire etc would probably mean i'd hit something. I'm curious to know what these folks used as their 1-7 follow settings.

 

[mongo]

Tesla autosteer/autopilot systems ignore stationary objects. That's the data. Pay attention carefully when using.

And a new question I just thought of, has this (or any of the recent crashes) divulged what the follow distance was set on before impact? In my experience in my daily commute, when traffic is heavy I am usually at 3 maybe 4, in good weather. I've been on highways at 1 and had my car get too uncomfortably close to cars in front of me that are jammed like sardines. One swerve or blown tire etc would probably mean i'd hit something. I'm curious to know what these folks used as their 1-7 follow settings.

An Update on Last Week’s Accident

In the moments before the collision, which occurred at 9:27 a.m. on Friday, March 23rd, Autopilot was engaged with the adaptive cruise control follow-distance set to minimum.

 

[jkennebeck]

yeah 70mph at rush hour traffic with a 1 follow distance is probably the time you want to be paying the MOST attention to your surroundings. i've been there and like i said earlier, i tend to pull the follow dist back a bit in rush hour traffic.

 

[deonb]

Apart from the acceleration bit.

My Ford F350 (which is as pedestrian a vehicle as it gets) will quite happily accelerate to 70 mph if it loses radar contact with the car in front of it. All vehicles with TACC will do that.

 

[cbdream99]

3 seconds at 70 MPH is 316 feet.
264 feet at 60 MPH.

Thanks for the correction, forgot to x 3 for the number of seconds

 

[Barklikeadog]

3 seconds at 70 MPH is 316 feet.
264 feet at 60 MPH.

Why is the assumption that the driver was doing something careless? Perhaps he used those final 3 seconds to check his side and rearview mirrors. Somethinf that we all do and take our eyes off the road ahead for a few seconds.

 

[mongo]

Why is the assumption that the driver was doing something careless? Perhaps he used those final 3 seconds to check his side and rearview mirrors. Somethinf that we all do and take our eyes off the road ahead for a few seconds.

???
I was just correcting the distance.

The time with clear view of the barrier was 4+ (4 from NTSB, about 5 from Tesla) seconds.

 

[Mark_T]

Why is the assumption that the driver was doing something careless? Perhaps he used those final 3 seconds to check his side and rearview mirrors. Somethinf that we all do and take our eyes off the road ahead for a few seconds.

No assumption that he was 'doing something careless' merely stating that he was not doing anything to react to the impeding accidet, no braking and no steering.

Also FWIW, the car was accelerating at the time which tends to bring my focus immediately to what is in front of me...

 

[benfar]

Tesla autosteer/autopilot systems ignore stationary objects. That's the data. Pay attention carefully when using.

And a new question I just thought of, has this (or any of the recent crashes) divulged what the follow distance was set on before impact? In my experience in my daily commute, when traffic is heavy I am usually at 3 maybe 4, in good weather. I've been on highways at 1 and had my car get too uncomfortably close to cars in front of me that are jammed like sardines. One swerve or blown tire etc would probably mean i'd hit something. I'm curious to know what these folks used as their 1-7 follow settings.

Data? No, here are data, from 2017 Tesla Model S that proves that the evaluated Tesla detected a stopped vehicle and completely stopped from initial speeds up to 25 miles per hour, the highest initial speed that IIHS tests. Now, we can't tell from this how well does it reacts at higher initial speeds, but the "ignore stationary objects" is proved false by these independent third-party tests.

What confuses people is the owners-manual language for Traffic Aware Cruise Control, which has cautions about its not handling stationary objects well -- but that isn't its job. Instead, that is the job of forward collision warning and automatic emergency braking, whose owners-manual description is "Automatic Emergency Braking operates only when driving between 5 mph (8 km/h) and 85 mph (140 km/h)." AEB is NOT designed to ignore stationary objects, so a reasonable question is what (didn't) happen in this accident, and whatever the answer, it was more complex than the radar or AEB not being able/designed to react to stationary objects.

Tesla