Autopilot lane keeping still not available over 6 months after delivery

[joer00]

I agree that TACC is fantastic even under heavy ! I should be always on especially in city traffic where it is much more likely to bump into somebody than on the highway ! I have it always on. What I am mssing and is EASY to do is the TACC following the speed signs. The car is already able to do this, one must only secure the cruise control stick to always be pulled to the front

 

[Andyw2100]

However, recently I've only had it cut out one time (huge bug splat in front of the front cam, lol) and it actually didn't do the full regen like I had previously. It gave the alert and 3-beep chime error warning and slowly ramped down speed until I pressed the pedal.

So, I think they're working on that too.

That's actually an excellent improvement that I did not know about. I don't think anyone has written about that on TMC before, or if they have, I missed it.

In the early days of TACC, I had it cut out in good weather, when the radar decided it was having issues, and each time it was pretty startling, going from 65 or 70 to full regen braking. So I view this as a very significant improvement to the TACC system. Thanks for sharing!

 

[Cyclone]

That's actually an excellent improvement that I did not know about. I don't think anyone has written about that on TMC before, or if they have, I missed it.

In the early days of TACC, I had it cut out in good weather, when the radar decided it was having issues, and each time it was pretty startling, going from 65 or 70 to full regen braking. So I view this as a very significant improvement to the TACC system. Thanks for sharing!

At least under .253, TACC has been much smoother than in .188. That said, I still experience regular "gun it" and speed up to "slam the brakes" moments. But definitely much better than before.

 

[thegruf]

do we get the lane change with a nudge of the indicator with this next build?
I am guessing not.

But cars learning from each other ..
how cool is that. Yet another Tesla usp that will leave traditional auto makers scatching their heads

Q: do all the Tesla's become a sentient collective being at some point lol

 

[Andyw2100]

At least under .253, TACC has been much smoother than in .188. That said, I still experience regular "gun it" and speed up to "slam the brakes" moments. But definitely much better than before.

There's no question TACC has been improving. I've seen that myself firsthand, and many people have written about it. Personally the improvement that I think has affected my driving the most is the fact that if someone passes my car and immediately moves back into my lane, unless they are really quite close to my car the TACC seems able to recognize that they are going faster than I am, and does not slow the Model S. With earlier TACC my Model S would slow significantly, in situations where the car cutting in was cutting in much further in front of me than where the TACC now doesn't slow the car at all.

It was the change wk057 wrote about that I said I had not seen described before--the behavior when TACC has to turn itself off, and how that no longer involves full and immediate regenerative braking, but rather a gradual decrease in speed.

 

[Cyclone]

There's no question TACC has been improving. I've seen that myself firsthand, and many people have written about it. Personally the improvement that I think has affected my driving the most is the fact that if someone passes my car and immediately moves back into my lane, unless they are really quite close to my car the TACC seems able to recognize that they are going faster than I am, and does not slow the Model S. With earlier TACC my Model S would slow significantly, in situations where the car cutting in was cutting in much further in front of me than where the TACC now doesn't slow the car at all.

It was the change wk057 wrote about that I said I had not seen described before--the behavior when TACC has to turn itself off, and how that no longer involves full and immediate regenerative braking, but rather a gradual decrease in speed.

Yup, I noticed that as well. Sadly, even though it goes blue-lock on the next car when the one in front of me leaves, if that car was slowing down/braking, I still get the speed up to slam the brakes. Today on the way to check on my car at Tesla, it sped up from 35 to 55 when someone move out in front of me and the person in front was braking to a stop because of traffic in front. I gave it some time thinking TACC was going to slow down, but the collision alert went off and I slammed my brakes. But that's ok. Its continuing to improve and I think we all will be ecstatic when it's released. I may even get to try it then on a loaner when my battery comes back from Freemont.

 

[Vger]

With regard to whether camera optics and the current suite of sensors can or will do better than a human - the problem is hard, and there is plenty of work to be done. However, whenever I see the argument that machines will be unable to exceed the human capacity for (x), I see shortsightedness. I also see someone who has a lot more respect for the human sensor suite (and intelligence!) than I do.

I'm bullish on technology beating the pants off of human drivers. Maybe not very soon, but Tesla soon.

As a neuroscientist by training, I see this as quite uninformed. The human "sensor suite" is quite near the limits of the physics involved. The human visual system (dark adapted) can detect a handful of photons, and yet has a dynamic range exceeding 7 orders of magnitude (10,000,000:1). It would take a digital camera with more than 500 megapixels to equal the spatial resolution of the human eye/brain, full-field (see here for a decent summary).

I am not a luddite about autopilot at all, but some technologists are simply quite under-appreciative of the neurophysiology we are endowed with.

What the computers have going for them (so far) is absence of distraction. A totally focused expert human driver will take some time to replicate. Consider that at any given time, only a handful of humans in the world are competent to compete at the highest level of motorsport— Formula One.

 

[brianman]

I think this is the making of a shareholder proposal ...

Some bullet points that may make sense:

1. State clearly which features are actually available *today*. No features that are not yet available shown anywhere on the order page unless explicitly and very obviously stated as a not-yet-active/available feature and that the buyer would be prepaying for it's future delivery.
2. State actual combined output horsepower for each model along with or in lieu of the less than useful "motor" power numbers.
3. The first and most prominent price shown for the vehicle is the amount a buyer would write a check to Tesla for. After/below that price any savings from gas, tax credits, etc can be shown.
4. No present tense verbiage for any features that are not available at the time it is displayed on the website or elsewhere.
5. Any upcoming features, after clearly being described as not yet available or will not be available immediately upon delivery, will have a clear price for the prepaid-purchase of that future feature/item and a final set date by which Tesla will guarantee that item will be made available or otherwise delivered to the customer in it's entirety.
   • If the feature/option is not delivered/available by that date then the customer is entitled to request and immediately receive a no-questions-asked full refund of the feature's prepaid price any time after that date and up until the feature is actually available.
   • The customer is allowed the option to, at a later time once the feature is actually available, purchase the feature from Tesla for the original price plus 5% of refunded amount. Or at their discretion the customer can continue to wait indefinitely.
6. Warranty miles and time elapsed on parts specific to pre-paid features that are not otherwise able utilized by the buyer until feature delivery (re: autopilot steering control) will not start until the date the feature is made available.

These items are not at the right level for a shareholder proposal.

While I tend to agree, the 1st, 3rd, 4th, and 6th bullets seem like obvious ethical things for any company to do. Would you not agree?

 

[wk057]

While I tend to agree, the 1st, 3rd, 4th, and 6th bullets seem like obvious ethical things for any company to do. Would you not agree?

I'd be fine with a proposal that was just 1,3,4 and 6.

 

[brianman]

Something that seems like a good idea to me is the following.

Tesla has about 30,000-40,000 cars on the road with Autopilot hardware. What if, during this time, they were able to create a map of all the highways that people are on (which would include most of the highways). The map would tell them where Autopilot is okay to use, and where it is not. Then, when you put a destination into your NAV, it gives you a route overview on where you can use Autopilot, and where you cannot. As you approach an area where you cannot use Autopilot, it gives a countdown at 2 mins, 1 min and 5 seconds until you take over.

What do you think?

I've heard a variety of proposals along these lines. What I thought you were going to say (on my first skim) you didn't quite say... so I'll say it:

When choosing to navigate from A to B, in theory the NAV could offer you the (A) fastest route, (B) shortest route, or (C) [if available] pure autopilot route. I wish Tesla offered the (A)/(B) choice for the whole fleet already, but (C) would definitely be a nice compelling way to address the "I want AP the whole way."

 

[Andyw2100]

When choosing to navigate from A to B, in theory the NAV could offer you the (A) fastest route, (B) shortest route, or (C) [if available] pure autopilot route. I wish Tesla offered the (A)/(B) choice for the whole fleet already, but (C) would definitely be a nice compelling way to address the "I want AP the whole way."

Assuming the lane-keeping is going to be released with some sort of data base of where it will work and where it won't, it would be great if some day the Nav system were able do this. assuming that comes to pass, it would also be great if at that time the nav would show, on routes A and B, the portions of those routes where autopilot (lane-keeping--whatever we're calling it) is available, via a different color or something.

So, for example, if the fastest route is an hour faster than the pure autopilot route, and has only ten minutes of non-autopilot, it would probably be a no-brainer to take that route. This is information we'd want to have.

 

[ohmman]

As a neuroscientist by training, I see this as quite uninformed. The human "sensor suite" is quite near the limits of the physics involved. The human visual system (dark adapted) can detect a handful of photons, and yet has a dynamic range exceeding 7 orders of magnitude (10,000,000:1). It would take a digital camera with more than 500 megapixels to equal the spatial resolution of the human eye/brain, full-field (see here for a decent summary).

Uninformed, probably so (I'm only human, after all!). However, I cannot agree with your assessment that our brains are near the limits of physics unless you qualify the statement having to do with energy consumption or size or some other parameter. For instance, axons transmit at a maximum speed of 120m/s. This creates size constraints in the brain that do not exist with machines, which can transmit at some much higher fraction of the speed of light - 300,000,000m/s - (depending on the media, say 67% for fiber optics).

What the computers have going for them (so far) is absence of distraction. A totally focused expert human driver will take some time to replicate. Consider that at any given time, only a handful of humans in the world are competent to compete at the highest level of motorsport— Formula One.

The computers have more than absence of distraction, in that they have speed and the ability to incorporate more than two eyes (for instance). And, as mentioned here, radar, which we don't have.

We have a ways to go, no question. However, I'm obviously on the bullish side of technology here.

 

[stopcrazypp]

View attachment 89109

My August 3rd bet is still looking good!

Interesting quote. I was considering if the AI has learning ability perhaps it can get better than just relying on a base algorithm. However, there is also the concern that there would be inconsistent results among cars (meaning a car with better "training" will be better at some scenarios). It'll be interesting if Tesla can consolidate all learning together (doesn't have to be real time, but can be synchronized at every software update).

 

[Lerxt]

My suggestion is to have a person walking in front of all Teslas that are driving on autopilot, with a red flag. Then everyone will be safe.

 

[Electricfan]

Something that seems like a good idea to me is the following.

Tesla has about 30,000-40,000 cars on the road with Autopilot hardware. What if, during this time, they were able to create a map of all the highways that people are on (which would include most of the highways). The map would tell them where Autopilot is okay to use, and where it is not. Then, when you put a destination into your NAV, it gives you a route overview on where you can use Autopilot, and where you cannot. As you approach an area where you cannot use Autopilot, it gives a countdown at 2 mins, 1 min and 5 seconds until you take over.

What do you think?

I think Tesla has chosen not to do this. There are two schools of thought on auto-pilot implementation, according to the video below, which is by an expert at Mobileye, who Tesla is partnering with on this. One is "store and align", and the other is "sense and react". Tesla is doing the latter, because that is Mobileye's philosophy, which is what Tesla is using (its mounted in front of your rear-view mirror if you have an auto-pilot version of the Model S).

Sense and react is cheaper - based on a camera. Cameras are great now, and doing nothing but getting better. I was very impressed with the video and Mobileye, and very glad Tesla is partnering with them.

Check 14:39 into the video.

https://www.youtube.com/watch?v=kp3ik5f3-2c

Your suggestion is very smart, but requires hi-def maps and massive amounts of storage on the car, and updates to it.

This is my understanding, anyway, and I may be wrong... so, fwiw.

 

[gpetti]

Why? Everyone can slow a Tesla on TACC. Change lanes in front of the Tesla, slow down until the Tesla stops.

The individual can always turn off TACC and resume control, but such a feature is going to cut down on deaths for people who fall asleep or pass out behind the wheel. Plus, I don't think a criminal running away from the police would use TACC anyways--who uses cruise control on a police chase?

Not me!

- - - Updated - - -

August 9th, I have a nice little spot prepared for the 1:18 scale Model S

August 14th, I'm already shopping for the right shade of classic blue model paint.

 

[wk057]

I saw this in another thread about the Model X and got a laugh about it.

I checked my Elon-Time calendar and it looks like we're only into july. I think it will come out on 49 July 2015. It also appears that first deliveries will be 51 Soonvember. AP will arrive sooner in Almosgust. Nothing planned for August.

I'll just throw out there that today is August 1st. This thread is about a month old. My P85D is over 7 months old with nearly 14k miles (less than expected since I've done some flying vs driving). Autopilot was announced nearly 10 months ago (296 days ago).

So far the only thing I have to show anyone who asks me something along the lines of, "What happened to the whole autopilot thing?" are a few question responses and now a couple of tweets from Elon. (I'll just note, because someone is going to say it, that TACC is not Autopilot).


That said, I am happy Elon finally gave some progress update on his own without being prodded for one.

I'm still finding it somewhat amusing how utterly terrible Tesla's internal communication must be for the primary source of info to be tweets from the CEO... For example, it's been two weeks since the whole "Ludicrous" announcement and my service center still has zero details on the officially announced retrofit (parts availability, labor cost, etc). Wonder how long it'll take them to get that situated. Seems like something they should have had in order with the service centers, parts ready to roll on day one. *shrugs*

 

[thegruf]

go have a beer mate, life's too short for this level of stress.

Bits of autpilot are here, more's coming soon, and sure it will take a time to roll out the ludicrous update.
and learn yer lesson - anything that's not released yet, is not released, and it will be released when it's done, and nobody knows when that is ... until it's done.

The car is good now, and it will only get better, enjoy it

 

[MileHighMotoring]

go have a beer mate, life's too short for this level of stress.

Bits of autpilot are here, more's coming soon, and sure it will take a time to roll out the ludicrous update.
and learn yer lesson - anything that's not released yet, is not released, and it will be released when it's done, and nobody knows when that is ... until it's done.

The car is good now, and it will only get better, enjoy it

I would say three things to you in response.

1) I like your attitude. The car is fantastic and will improve.

2) I disagree that we have any semblance of AutoPilot right now. Adaptive Cruise Control has been around for more than a decade and isn't a special "AutoPilot" feature. It's commonplace. Auto Parking, Lane Keeping, et cetera, would quality as piloting the car automatically.

3) I would hesitate speaking for others. If you're happy, that's wonderful. But it's not fair or reasonable to tell others that their expectations aren't reasonable, unless they are empirically mistaken on some facts or something.

 

[breser]

2) I disagree that we have any semblance of AutoPilot right now. Adaptive Cruise Control has been around for more than a decade and isn't a special "AutoPilot" feature. It's commonplace. Auto Parking, Lane Keeping, et cetera, would quality as piloting the car automatically.

Go look at the order form. Under Autopilot Convenience Features you get:

• Traffic-aware cruise control
• Lane keeping with automatic steering
• Self-parking
• Automatic high/low beam headlights

We currently have half of those features (marked in green). So it's entirely fair to say we have "bits of autopilot." If you want to say we don't have Autopilot because we don't have all those features fine.

I'll caution however that I think a lot of people are expecting a lot more from this functionality than it really will be capable of. I'm sure Tesla will provide us with the best implementation of automatic steering. But I'm also not expecting my car to become KITT. I'm predicting that when it does come there will be a new thread about how disappointed people are with auto steering and how it's not really Autopilot. Because for whatever reason people all have their own ideas of what Autopilot means.

The truth is that Autopilot is whatever Tesla says it is. Autopilot is a marketing term. Don't dig yourself a hole of disappointment by expecting too much of it.