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

”This Test Shows Why Tesla Autopilot Crashes Keep Happening”

This site may earn commission on affiliate links.
I'd like to see the same test with people behind the wheel. I doubt if people would do any better, and very probably a lot worse. (Want to see someone turn hard and flip a car?)

We didn't use to have electronic warfare suites. We did not hit cars in that situation. Go figure.
That same situation occurs constantly at stop lights or when the freeway backs up.
See hundreds of chain collisions each day?
 
Some Youtube clips has Teslas beeping warnings when the two cars ahead of it were closing on each other. Why didn't this happen in this test?

This is a different situation. AP is very good at detecting and registering vehicles that it was tracking that were once moving and now does not move. All these accidents are a different scenario where the vehicle hits a stationary object (fire truck, highway lane divider, etc.,). The situation is exasperated when the leading vehicle hides view of the stationary object and then gets out of the way leaving very little time to react. This is what happened in the well publicized accident in China.
 
  • Informative
Reactions: 12Pack
We didn't use to have electronic warfare suites. We did not hit cars in that situation. Go figure.
That same situation occurs constantly at stop lights or when the freeway backs up.
See hundreds of chain collisions each day?

There are actually no shortage of accidents in those situations. Plenty of people in these very forums have been rear ended in probably far more mundane circumstances than that! It's a bit like saying people didn't crash before ABS braking was invented or that people didn't skid off the road before ESP. Of course they did!
 
Safe FSD is still a long way off IMO. Too many edge cases and unpredictable human drivers around for it to guarantee safety. AP2 is a million miles away from being FSD. It can follow simple lines and avoid hitting objects in simplistic scenarios, but that's as far as it goes. It struggles with stationary objects, it struggles to react to vehicles cutting in front of you at the last moment, etc, etc, etc. It's just a fancy cruise control with lane assist and the rest is a pipe dream for the foreseeable future.

Exactly. If we are so close to FSD then why has Elon not been able to pull off his long ago promised and much hyped coast coast-to-coast FSD drive? I suspect we are at least 10-15 years from full-blown FSD for the masses, maybe more.
 
Exactly. If we are so close to FSD then why has Elon not been able to pull off his long ago promised and much hyped coast coast-to-coast FSD drive? I suspect we are at least 10-15 years from full-blown FSD for the masses, maybe more.

Yeah I think FSD is one of the most over-hyped concepts in the motoring world at present. All this talk of FSD (not just Tesla either) you'd think it would be available in the next 12-24 months, but it's just not going to happen that quickly. My guess is that specific features of FSD might appear soon enough (e.g. basic stop light detection) but that's about all. Even true auto lane-changing seems unlikely unless they can dramatically improve vehicle detection on rearward approach. Hands-off driving is not happening anytime soon!
 
There are actually no shortage of accidents in those situations. Plenty of people in these very forums have been rear ended in probably far more mundane circumstances than that! It's a bit like saying people didn't crash before ABS braking was invented or that people didn't skid off the road before ESP. Of course they did!

If a lane change exposing a stopped car causes an impact, you would never get anywhere in a car. It is too common of an event. Yes, people hit cars in front of them. Inattention or lack of adequate following distance causes it. An AEB system that cannot work in that situation is nearly pointless.
 
If a lane change exposing a stopped car causes an impact, you would never get anywhere in a car. It is too common of an event. Yes, people hit cars in front of them. Inattention or lack of adequate following distance causes it. An AEB system that cannot work in that situation is nearly pointless.

Well none of the AEB systems in any of the cars tested (Tesla, Merc, BMW) passed this particular test. The Tesla did actually give an AEB warning but it needed a human to steer the car. The key to this test is that the lead car swerved at the last possible moment to reveal the stationary car and then the AEB didn't have enough time to react and stop the car (in this case AEB slowed by only 5 mph before impact). The warning and AEB activation might have been enough to prompt an alert driver to steer around the car, but really the point is that the driver should have seen the stationary traffic ahead and probably would have done in everyday driving. What you cannot do with these systems is take your attention completely off what is happening in the road ahead, which was the point of the report.

You say AEB is pointless if it can't deal with this specific situation, but there are crashes every single day where people just go into the back of slowing or stationary cars without the complexity of a car in front swerving at the last second. AEB is very effective in those even more common situations, so not actually pointless at all. It's just limited in its ability to deal with EVERY braking scenario.
 
I've seen this video and in real life under the same motorway conditions my AP1 car has slowed down to a stop no problem. A card board car is not going to give the same radar signature as a real metal car!

The "cardboard" car theory has been discussed on the UK owners group and apparently it does give a realistic signature. Tesla even clearly warn about this exact scenario in the user manual, so it is a real issue. But not just a Tesla specific issue, most other systems have the same limitations.
 
So much depends on the simple point of having seen the object in question in motion or not...

We know that Tesla does try to track the car in front of the car you are following when possible and if it sees that car moving then come to a halt then it has a good chance of reacting correctly if the car you are following moves to expose the stopped, but previously moving object.

It is when the car in front moves to expose an object that your car has never seen moving that seems to be the 'flaw' in pretty much all these systems for driving at motorway speeds...
 
  • Like
Reactions: WannabeOwner
I think the problem is compounded because "it sometimes works", so people wills say "My car HAS stopped for a parked car / stationary object". Its just that the system (as described in the manual ... assuming all owners have read it ...) is not capable of doing that in all cases, and in particular not at higher speeds.

For me its similar to AutoPilot. because it works flawlessly 99.9xxx% of the time it is easy for drivers to get complacent / assume it is doing a perfect job ... good enough in fact that they then start Texting / Nodding off / climbing into the passenger seat :eek:

I'm always, without exception, one-hand-on-wheel (and never hands-on-lap) and right-foot-over-accelerator (to counteract AP phantom-brake, particularly for when there is traffic close behind me). The need to override AP with emergency brake does exist, but I have never had to do it (in tens-of-thousands of AP motorway miles) [and probably override with "land-change/swerve" is more likely than emergency brake]

I would urge anyone using AP not to become complacent, and to have road-awareness-rules that they stick to :)
 
Phantom braking was my biggest safety issue with AP2, but the latest update seems to have stopped it happening completely. I was often getting unwanted sudden braking in certain places and with very dark shadows on the road, but not anymore. Now I often get the infamous steering wheel twitch when passing junctions on the left instead! Hope they fix that soon.
 
If that were true, radar reflectors could be made from cardboard, instead of metal. I would be interested to see evidence that some cardboard on a wooden frame will give anything like the return that is received from one or more tonnes of metal.

It wasn't just cardboard, I believe there was some metal foil included to make it reflect a realistic signal. I also believe they tested the dummy car with the car's sensors to make sure it was being picked up. It may not be a perfect representation of a real car, but it shows up the limitations exactly as described in the Tesla user manual. If it wasn't a problem, there would be no such warning in the Tesla manual. But feel free to make your own test with a real car and let us know the outcome!