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

Autopilot lane keeping still not available over 6 months after delivery

This site may earn commission on affiliate links.
... Standard equipment safety features are constantly monitoring stop signs, traffic signals and pedestrians, as well as for unintentional lane changes.

I have an issue with the above, boldfaced portion. Musk is on record saying that the Model S' current sensor suite is insufficient to detect certain soft objects, such as humans and pets, which is why Autopilot (in whatever form he's talking about) cannot be used in city situations at speeds between 20 MPH and 45 MPH. Musk said something along these lines during his interaction with NVidia CEO on stage a few months ago.

If that's the case, why is Tesla claiming that the system monitors for pedestrians? He already said current generation Autopilot sensors cannot do this.

In Elon's defense, I believe he specifically did say that the sensors *can* in fact detect soft objects, but can not do so at higher than a very slow speed (few MPH). Hence the gap between ~20 MPH and ~50MPH driving that he mentioned being not covered by the current sensor suite.

That said, however, I don't think the line about standard equipment safety features is, in the present, accurate at all. The car definitely does nothing in response to a traffic signal, stop sign, or pedestrians currently.
 
I don't remember the exact wording, but are you sure it just wasn't that it couldn't do it accurately enough to allow the vehicle to drive itself in city driving? I watched the conference so I know what you're talking about...but I don't remember the exact context.

My recollection is that Musk was talking about extending Autopilot functionality to city driving. I don't think he was talking about autonomous.
 
As a philosophical side thought, what on earth is the point of autopilot if not to reduce the amount of attention I need to give towards the road?

Interesting question. In my opinion, the goal of AutoPilot (as it was described at the 10/09/2014 event) is to reduce driver fatigue when driving:

1. long, boring distances
2. in stop-and-go traffic

These are the only two situations where I personally don't enjoy driving as much -- in general, I love driving and always volunteer to when others are involved.

The amount of attention that will need to be paid to the road when all AutoPilot software features are released (with the current hardware suite) is still the same as it AutoPilot weren't available.
However, in the two conditions described above, AutoPilot will greatly help in reducing the amount of small corrections by the driver that are needed to:

1. stay in a lane, by eliminating steering wheel input (Auto-steer should take care of that - main subject of this thread)
2. follow traffic, by eliminating foot pedal input (TACC already takes care of most such conditions)

The reduction of such inputs in turn reduces the amount of effort, and hence fatigue, a driver experiences over time. I can personally attest to that with #1 (when TACC came out).
At no point would I let the current hardware and future software of AutoPilot get me to be distracted and stop paying attention to the road.

Oh, and those two conditions I mentioned? For me, they would both be highway-only scenarios.
In no way would I expect AutoPilot to drive me through a city with traffic lights, stop signs, and pedestrians crossing the street. On-ramp to off-ramp only!
 
Interesting question. In my opinion, the goal of AutoPilot (as it was described at the 10/09/2014 event) is to reduce driver fatigue when driving:

1. long, boring distances
2. in stop-and-go traffic

These are the only two situations where I personally don't enjoy driving as much -- in general, I love driving and always volunteer to when others are involved.

The amount of attention that will need to be paid to the road when all AutoPilot software features are released (with the current hardware suite) is still the same as it AutoPilot weren't available.
However, in the two conditions described above, AutoPilot will greatly help in reducing the amount of small corrections by the driver that are needed to:

1. stay in a lane, by eliminating steering wheel input (Auto-steer should take care of that - main subject of this thread)
2. follow traffic, by eliminating foot pedal input (TACC already takes care of most such conditions)

The reduction of such inputs in turn reduces the amount of effort, and hence fatigue, a driver experiences over time. I can personally attest to that with #1 (when TACC came out).
At no point would I let the current hardware and future software of AutoPilot get me to be distracted and stop paying attention to the road.

Oh, and those two conditions I mentioned? For me, they would both be highway-only scenarios.
In no way would I expect AutoPilot to drive me through a city with traffic lights, stop signs, and pedestrians crossing the street. On-ramp to off-ramp only!

It's a given that autopilot is current only for freeway (and maybe highway) driving. However, the webpage seems to imply that it will be aware of pedestrians, traffic lights, and stop signs while driving.

If the system is so bad that I have to pay as much attention to the road as if I were driving, then I will find it pretty pointless, and be a little upset. Such a system would be more dangerous to me. I would find it far more fatiguing to have to babysit the autopilot system, than to just remain in control of steering. At least it gives me something to do on long stretches of freeway.


I am hoping for a system that can let me dink around on my phone, or browse this forum while cruising along the freeway at 60 mph. If visibility becomes limited, weather becomes bad, or construction is sensed, I would hope that it alerts me so I can step in.
 
That clearly isn't talking about what autopilot CAN do, but is a statement to get Tesla out of legal necessity. Yeah, if the car hits something it is ultimately your fault. However, even with the most pessimistic of views, it sounds like we can expect the car to:

Know a stop sign is approaching.
Know the status of a traffic light.
Know that a pedestrian is in the road.

That is very different from what you are claiming. I'll post it again in case you forgot. Emphasis mine.

These situations are where you would be required to be paying attention, the ones where "auto-pilot" and not "autonomous" driving comes into play.


I am currently "required" to pay attention to stop signs, traffic lights, and pedestrians in the road. The car can not see them, which is contrary to what Tesla's own website admits to.

As a philosophical side thought, what on earth is the point of autopilot if not to reduce the amount of attention I need to give towards the road?

The point is to reduce fatigue on long and boring highway stretches.
 
In Elon's defense, I believe he specifically did say that the sensors *can* in fact detect soft objects, but can not do so at higher than a very slow speed (few MPH). Hence the gap between ~20 MPH and ~50MPH driving that he mentioned being not covered by the current sensor suite.

He may have said that at some point, but that's not what he said at the D announcement:


I just watched the D announcement again, from start to finish: Tesla Unveils Dual Motor and Autopilot - YouTube
He also says very clearly that the sonar sensors work at all speeds--0 to 155 MPH, so that things like the blind spot detection will work at all speeds. (He says this at about 8:05 on the video.)

Edit: Start watching the video at 7:35 if you want to hear the entire portion on the "ultrasonic sonar." He also talks about the system "attempting to make the smart move, based on the ultrasonics."

 
Last edited by a moderator:
If we don't even have full autopilot with the current sensor suite on the S right now, I'm curious why people think that the X will have any software support for a more advanced sensor suite on launch. If anything, it will have yet more underutilized hardware for another year.
 
Nah, I've had 2 cars that auto park fine and they have had only reverse and front sensors max, less than the MS as mines on side even. Range Rover, etc. they do fine without any "curb" specific sensors.
If you are referring to Range Rover without the need for curb sensor for auto parking, it could be that Range Rover rides much higher than Tesla. When I said curb sensor, I am referring to curb height sensor.
 
There is no way it's going to be able to autopilot for stop signs etc. For example if it comes to a 4 way stop it would have to determine if it should go next etc. At a regular intersection it would have to watch for cars coming from the side. Trying to merge into traffic, nope not going to happen.

I always thought it's an on-ramp to off-ramp type of thing. Basically ACC with lane keeping. I'll be happy with that!

I don't trust it to read speed signs. It makes mistakes sometimes on the roads around here. I hope if it does try to adapt to the speed limit signs I'll be able to turn that feature off.
 
Last edited:
I am hoping for a system that can let me dink around on my phone, or browse this forum while cruising along the freeway at 60 mph. If visibility becomes limited, weather becomes bad, or construction is sensed, I would hope that it alerts me so I can step in.

You will not find this on any production car, available for legal use, in the next 5 or so years--maybe 10.

You COULD do this when auto steer comes out, but if you did your lifespan would be significantly shorter, and your life would end abruptly.
 
You will not find this on any production car, available for legal use, in the next 5 or so years--maybe 10.

That's funny. Are you the CEO of an auto company (serious question)? I ask, because at the shareholder meeting Elon said it would be 1-3 years before you could basically sleep in the car (technology wise, not legal wise). I would think if anyone had a good idea about the timeline it would be him, but I don't actually know you. Google, Uber etc. all think driver-less cars will be here by 2020.

I fully realize I will remain responsible for what my vehicle does until the laws catch up. I'm just looking for a car that drives as good as me on the freeway.

I'm not asking to sleep in the back. I am just asking for less attention than I already put in. Maybe I need to look up every 15 seconds or so. Does a pilot constantly look around and check everything constantly as if he is flying the plane when it is on autopilot? Probably not. They relax and let it do their thing, knowing that if things change the system will let them know and they will take over ASAP. FYI, this is only slightly less engagement than the average "texting while driving" teenage girl drives even without autopilot.

I am imagining a system, that on the freeway, would require you to be present, but not really all that alert (assuming good weather, not a lot of merging in your lane, etc.). Seems like this should easily be possible with the current hardware.

"Less fatigue" by having to turn the steering wheel 75% less doesn't even sound worth it to me. In fact, it sounds irresponsible to release something like that to the public. I already have trouble paying attention in stop and go traffic with the current TACC. If I have to steer significantly less, but still pay attention, that sounds worse than just being engaged with the road.

FWIW, I say this having been in a 65 mph accident in a Tesla, and hopping out without so much as whiplash. So, we aren't talking about a super high bar here. If a deer jumps in front of the car, I don't expect not to crash, just like I probably would crash even if I were in full control.
 
Since I'm a believer that a model S refresh is coming withe the model X reveal, I think the next software update is being withheld for the release of these vehicles. The side cameras likely will be part of a dash change. Both the model X and S will presumably share instrument panel changes. These changes require software changes.

All the changes are interrelated. Once lane keeping got pushed close to the X reveal, it may have been "captured" by this release process.

The more I think about this, the more I agree with you. Perhaps Model S 2.0 is the "something else" that we will see during the X reveal that Elon has been hinting at for months. If you think about it from Tesla's perspective, it makes sense. Tesla tour guides in Fremont are telling those on tour that S and X will be assembled on the same lines, so no special line for X just by itself. This would imply a lot of shared components between the two vehicles.

Model X will have a much nicer interior for sure, along with improved and more updated technology. Without updating the Model S to the same, who in their right mind would buy an S over an X? One will look dated while the other will be brand spanking new with all of the latest Tegra processors, sensors, LED headlamps, and redesigned interior. Tesla has to keep S up-to-date with X if it expects to continue selling the S.

Count me in. I think we will see a significant refresh of the Model S during the X reveal.

Yes, they have to, don't they? They kill S sales if it less up to date than the X. An S refresh is not something that they would just like to do. They probably have to sync the two cars or they create S "fence sitters".

Count me in too. As much as it will pain me, I think Model S 2.0 will be released at the same as the X. It fits.

** I'm declaring two new terms: MS1 = Model S 1.0, includes non-autopilot cars and autopilot 1.0 cars. MS2 = Short for Model S 2.0 with 2.0 sensor suite.

1/ It could be that the major X delays were in fact to get MS2 ready. Perhaps Tesla realized (late) that the X had become so much better than the S that it would cannibalize S sales, and their only choice was to pull back on the X, or bring the S forward.

2/ Software resources have been reallocated/prioritized to work on the first version of the X and MS2 software (probably 6.0). There is likely only a small number of people working on autopilot 1.0. I'll bet MS1 will get the 6.0 (or whatever the base is) software within a few weeks of X first delivery.

3/ MS2 and X will have feature parity with 1.0 autopilot at launch. MS2/X will get additional features later via OTA.

Damn it.. I had managed to stay out of these prediction/guess threads. So much for that.
 
That's funny. Are you the CEO of an auto company (serious question)?

Nope.

I ask, because at the shareholder meeting Elon said it would be 1-3 years before you could basically sleep in the car (technology wise, not legal wise). I would think if anyone had a good idea about the timeline it would be him, but I don't actually know you.

Forget about me...it sounds like you don't know Elon. I think Elon is the genius of our generation, but if he has a flaw, he is unrealistically optimistic on his timelines. If you don't know this, you don't know Elon.

It's been almost a year since the autopilot announcement and there's still no auto-steering. What on earth makes you think cars will be autonomous in a year?

In all actuality, Google's cars have the capability for the driver to sleep. But that's in certain scenarios, and it's certainly not legal. And the cars have hundreds of thousands of dollars of sensor and computing hardware onboard.

Google, Uber etc. all think driver-less cars will be here by 2020.

Technically, they might be. But the sensor/hardware suite will make them prohibitively expensive, and they sure as heck won't be legal. I think they drastically underestimate the sluggishness of the legislative branch.

I fully realize I will remain responsible for what my vehicle does until the laws catch up. I'm just looking for a car that drives as good as me on the freeway.

Well yeah, we all want that. All I'm saying is that you're not going to find it for a good while.

I'm not asking to sleep in the back. I am just asking for less attention than I already put in. Maybe I need to look up every 15 seconds or so.

There's a flaw in this statement. Going 70 mph, you travel a third of a mile in 15 seconds. A whole boatload can happen in 15 seconds and a third of a mile. If there were a system in which you didn't have to look up more frequently than 15 seconds, then you wouldn't have to monitor it at all.

Does a pilot constantly look around and check everything constantly as if he is flying the plane when it is on autopilot? Probably not.

Good question. Former Air Force pilot here, so I can provide some insight on that. No, it's not constant monitoring 100% of the time when you're on autopilot, but depending on regime of flight (low altitude maneuvering vs. high altitude cruise) a diligent pilot is aware of potential landing areas, systems performance, traffic in the area, etc.

It's true that with cruise flight under air traffic control you can somewhat relax while on autopilot, but that's an entirely different animal. You've got radar monitoring traffic around you, radio communications, lots of space, and much less spatially dense traffic. Imagine flying a jet on autopilot when there are dozens of aircraft flying around you in formation, hundreds of aircraft flying the opposite direction, skirting right by your wingtip, crossing in front of you, and thousands of stationary objects floating in the sky that you can't hit or else you'll die. That's driving. And that's why the analogy doesn't apply.

They relax and let it do their thing, knowing that if things change the system will let them know and they will take over ASAP..

That's a common misconception regarding autopilot (on aircraft). I can find many examples of dead pilots who relied on their autopilot to tell them when something was wrong.

FYI, this is only slightly less engagement than the average "texting while driving" teenage girl drives even without autopilot.

Agreed. There are a lot of dead teenage girls who were texting too.

I am imagining a system, that on the freeway, would require you to be present, but not really all that alert (assuming good weather, not a lot of merging in your lane, etc.). Seems like this should easily be possible with the current hardware.

I guess you're not an engineer? If you were you'd realize this is a far more complex system than it seems.

"Less fatigue" by having to turn the steering wheel 75% less doesn't even sound worth it to me.
You won't have to touch the steering wheel--just monitor the car.

In fact, it sounds irresponsible to release something like that to the public. I already have trouble paying attention in stop and go traffic with the current TACC.

Please be careful. If you find you can't monitor the car with TACC engaged, please don't use it.

If I have to steer significantly less, but still pay attention, that sounds worse than just being engaged with the road.

Don't use autopilot then.

FWIW, I say this having been in a 65 mph accident in a Tesla, and hopping out without so much as whiplash.

That's a bit of a selfish view though. Not sure what being lucky in an accident has to do with anything. What about the car you might rear-end at 70 mph? Would they die?

So, we aren't talking about a super high bar here. If a deer jumps in front of the car, I don't expect not to crash, just like I probably would crash even if I were in full control.

Yeah we are. Autonomous driving (even for 15 seconds) where the driver need not monitor is extraordinarily difficult, as Elon is finding out.