Will Navigate on Auto Pilot be safer than humans?

[fasteddie7]

Nav on autopilot, well, really any aspect of full self driving, will truly only be safe when human drivers are taken out of the equation. People drive aggressively and without logic, and the logical and passive nature of self driving vehicles will make the transition take longer than it should I fear. Heck, people zip past and cut me off with autopilot on driving 5 above the speed limit with a few car lengths between the next vehicle.

 

[pdamski]

I believe that the only way that FSD will work is in a future where all of the vehicles are communicating with one another through a National Mesh system. As long as humans are a part of the equation there will be the potential for accidents. If a basic premise of the algorithm is that no two objects can occupy the same space at the same time we should be pretty safe. My guess is that in 5-10 years there will be "Smart Lanes" built as part of our highway system that will be for smart tech only. Those lanes will expand as we transition away from ICE's. There will always be ICE fanatics and I have my love of them too. They will never go away but hopefully become a collectable rather than a daily driver.

 

[Gavyne]

Agreed, until all cars are autopilot capable, the roads won't be safe with human & autopilot mixed together. The problem isn't with machines and AI. The problem is with humans, we are the problems, that's one thing some people just hate to admit. But really, over 100 deaths each and every day are happening due to car accidents, and they are human errors. When Tesla's autopilot gets into an accident once in a blue moon, it makes headline news. Yet there are over 17,000 car crashes in the United States every single day. People are in such denial it's not even funny.

I agree they will need to build dedicated autopilot lanes until all cars have capable autopilot systems. Right now I don't trust the autopilot system as much not because of the system itself, but because of other human drivers. As is I am constantly having to make maneuvers to avoid accidents caused by bad/inconsiderate drivers.

 

[xav-]

I totally agree. Another point that I want to add is the car should avoid staying adjacent to those huge semi trucks at high speed for a long time.

The tough part for the software is to read the situation correctly (e.g there is a large truck in the adjacent lane)... but the very good thing is.. based on what we see on the screen: the software already do a good job at reading the situation.

Implementing something like we suggested would therefore be trivial... the hard part is already implemented.

 

[xav-]

I totally agree. Another point that I want to add is the car should avoid staying adjacent to those huge semi trucks at high speed for a long time.

The tough part for the software is to read the situation correctly (e.g there is a large truck in the adjacent lane)... but the very good thing is.. based on what we see on the screen: the software already do a good job at reading the situation.

Implementing something like we suggested would therefore be trivial... the hard part is already implemented.

 

[Msjulie]

Sometimes my car shows 2 complete tractor trailers to my right side if I am passing one - some tuning still needed on understanding I am passing the truck (singular)

 

[aronth5]

I've used NoA. Here in Boston, EAP's ability to change lanes is not sophisticated enough to handle the aggressive driving style. EAP is too conservative, which does make it safer in theory, but the overall experience is poor. The car often refuses to squeeze into a tighter spot in the adjacent lane, or it tries, detects a fast-approaching car, and swerves back. Also, the suggested lane changes are very short-sighted. It does a poor job deciding which lanes is actually faster, so you end up doing lots of lane changes for little gain.

In light traffic, NoA will do a lot better, but it's not much of an incremental improvement to the overall EAP experience. Yet.

Pretty much agree with everything you say. What has helped for me is to just disable the (disable/mild/average/mad max) setting and just let the car perform normal EAP until it gets close to the exit and then in most cases exists the highway by itself. Not perfect but reasonably ok. Disabling that feature reduces decision making and improves the experience. Just hoping that the overall NoA lane changing feature will improve enough in 2019 so that I will re-enable that part of the NoA functionality.
Will be interesting to see when the next release comes out that shows noticeable NoA improvement. The last few are most bug fixes and Summon for Canadian owners. I'm fine with that and would rather Tesla take more time to fix some of the larger problems reported by owners.

 

[insaneoctane]

Thoughts....
Don't assume parity between NOA and FSD because FSD will require hardware 3.0 which is 10x more capable than 2.5 hardware. Therefore capability of FSD could very conceivably be far greater than NOA.
This poll brings up statistics that I personally think are flawed. When Tesla claims EAP shows decreased accident rates:

• Tesla said it recorded one accident for every 3.34 million miles driven when the autopilot was engaged
• The most recent National Highway Traffic Safety Administration data shows one auto crash for every 492,000 miles driven in the U.S. without an autonomous assist.

On the surface this appears to indicate that EAP is 6x safer than regular cars. But I wonder ... Wouldn't most of you EAP users agree that you do not use EAP when your route is dangerous and challenging? EAP is primarily intended for divided highways. I know that I use EAP on the "easy" stuff and I navigate the hard stuff myself. So, by definition, wouldn't all miles driven with EAP engaged be the easier miles with naturally lower accidents? It just seems like the data really has to be fairly compared to be apples to apples.

 

[Malbers]

Thoughts....
Don't assume parity between NOA and FSD because FSD will require hardware 3.0 which is 10x more capable than 2.5 hardware. Therefore capability of FSD could very conceivably be far greater than NOA.
This poll brings up statistics that I personally think are flawed. When Tesla claims EAP shows decreased accident rates:

• Tesla said it recorded one accident for every 3.34 million miles driven when the autopilot was engaged
• The most recent National Highway Traffic Safety Administration data shows one auto crash for every 492,000 miles driven in the U.S. without an autonomous assist.

On the surface this appears to indicate that EAP is 6x safer than regular cars. But I wonder ... Wouldn't most of you EAP users agree that you do not use EAP when your route is dangerous and challenging? EAP is primarily intended for divided highways. I know that I use EAP on the "easy" stuff and I navigate the hard stuff myself. So, by definition, wouldn't all miles driven with EAP engaged be the easier miles with naturally lower accidents? It just seems like the data really has to be fairly compared to be apples to apples.[/QUOTE

I think you make a good point, but you also have to take into account many non fatal accidents are never reported. That's what the experts say and if you want to confirm it for yourself ask friends and relatives if they've ever been in an accident that they didn't report. You might have to press them and ask even if it was a fender bender.

 

[RedModel3]

Its not the same but if EAP is never worked out FSD will never happen. And if you don’t think EAP and NoA is part of the roadmap to FSD I don’t know what to say. I’ve had dirty camera issues with NoA so why will that not affect FSD as well?

I think you misunderstood my meaning. I totally agree with you. I'm just trying to say that there are too many drivers who didn't RTFM, then complain about almost getting killed because they didn't think they had to stay engaged with the car.

 

[pdamski]

I think you misunderstood my meaning. I totally agree with you. I'm just trying to say that there are too many drivers who didn't RTFM, then complain about almost getting killed because they didn't think they had to stay engaged with the car.

There is a great deal of evidence to support this if one just looks at these forums. So many people using EAP on local streets, as did I when I first started driving the Model 3. I know better now.

 

[SSonnentag]

NoA is currently too annoying to use for automatic lane changes. It doesn't "see" far enough in advance to prevent unnecessary slowing vs maintaining speed and passing. There is also way too much delay between the driver commanding a lane change and the car actually making the change. Why can't the car always "know" when it is safe to make a lane change instead of waiting until the driver requests the change before "looking" to see if it is safe?

 

[Enginerd]

There is also way too much delay between the driver commanding a lane change and the car actually making the change.

Clearly they still have some tweaking to do, relative to the initial implementation. It's happened to me several times that when I confirm the lane change, it is perfectly OK. But by the time NOA hesitantly decides, then finally commits, it's not a good lane change anymore. I want to say that it generally fails safe (don't change lanes), but the delayed lane changes probably cancel out that benefit. I'm hoping that they're currently crunching all the feedback from (1) aborted lane changes, (2) suggested lane changes that the driver did not confirm, (3) highway exit inconsistencies, and (4) lack of proper on-ramp merging... and that the next version will be better.

 

[PessimiStick]

Someday this is going to cause a major accident. And I know Elon will come and blame that on inattention.. I suspect that the victim will sue Tesla.. and don’t you think that Tesla would deserve it? Phantom braking has been mentioned countless times for nearly 2 years now, still unresolved.

They could sue, sure. You can sue for anything. They would lose, however, since rear-ending someone is almost universally the person behind's fault. If you couldn't stop before hitting them, you were too close.

My car has also never phantom braked, in 3500 miles. It does brake hard on one stretch of road near my house where it thinks the speed limit goes to 35 when it doesn't (it's actually 50), but I can just disengage before that happens.

 

[SSonnentag]

In further testing of NoA this past weekend, the car wanted me to move left to pass a slower semi. I accepted the lane change, noting that I was only 1 mile from my exit. I knew there were three semi trucks stacked up, and that 1 mile was not enough time to pass given our slight difference in speeds, but wanted to see what the car would do. It turns out that the car will leave you stranded out beside a truck, with no way to merge, just as I suspected. I floored the car to get ahead and into the right lane. Just another anecdotal example of how NoA is not ready to move out of Beta.

 

[aronth5]

In further testing of NoA this past weekend, the car wanted me to move left to pass a slower semi. I accepted the lane change, noting that I was only 1 mile from my exit. I knew there were three semi trucks stacked up, and that 1 mile was not enough time to pass given our slight difference in speeds, but wanted to see what the car would do. It turns out that the car will leave you stranded out beside a truck, with no way to merge, just as I suspected. I floored the car to get ahead and into the right lane. Just another anecdotal example of how NoA is not ready to move out of Beta.

Which is why I disable that feature. Since you still have to look and confirm anyway why not do it on your terms so the car is ready to auto exit the highway. Much simpler and that feature works reasonably well.