Firmware 9 in August will start rolling out full self-driving features!!!

[X Fan]

Yeah, I drive on roads like that with AP all the time here in Texas. AP doesn’t know or care what road you are on as long as it can see the lanes it will let you engage... Maybe you are just referring to lane changes? That doesn’t work on local roads yet but that doesn’t mean AP isn’t working. If it is keeping you in your lane and maintaining speed/distance with cars ahead of you, it is working...

"If it is keeping you in your lane and maintaining speed/distance with cars ahead of you, it is working..."

lol: I feel like I just fell off the turnip truck and landed in Tesla's twilight zone (I've been using and driving an X with AP1 for over 2 years to great success). I'm really surprised that although AP2 has been out for quite some time, it doesn't come close to Tesla's wonderful AP1 version (that I used every day in Naples & on that road it worked flawlessly) for local driving.

Since almost every auto mfg on this planet can offer a maintain speed/distance, Tesla's AP2 limitation makes the product just not special.... you really can't drive effectively around town that way.

 

[X Fan]

Is AP all about lane changing only for you? Rather if you can’t lane change you consider that as ‘AP not working’?

You said AP is not working, but turns out only lane changing is disabled. Which is exactly how it is everywhere in roads that are not categorized as divided highways. So this type of road you can’t lane change yet in AP2 even in US.

"Is AP all about lane changing only for you?"

Yes, since most of my driving time is around Naples with terrific 6 lane divided roads.....it's a convenience that I've used for over two years with my other Tesla. On the freeways it will be beneficial but I had that already with my first X.

 

[daktari]

I would say AP2 on multi lane road with traffic, without auto lane change possible, is not a relief. To much plinging and re-engaging, breaking the "driving flow". I can't stand going as slow as the slowest though and need to keep my speed.
Better to use only TACC then.

 

[ChrML]

I would say AP2 on multi lane road with traffic, without auto lane change possible, is not a relief. To much plinging and re-engaging, breaking the "driving flow". I can't stand going as slow as the slowest though and need to keep my speed.
Better to use only TACC then.

Agreed. Just too bad TACC brakes hard once slower traffic merges into your lane, and waits several seconds after traffic ahead completely leaves your lane until it accelerates.

 

[emmz0r]

Agreed. Just too bad TACC brakes hard once slower traffic merges into your lane, and waits several seconds after traffic ahead completely leaves your lane until it accelerates.

VW's TACC is kinda like that too. Should be an "aggressiveness" slider of some sort.

 

[Bladerskb]

You’re both wrong, both parts are difficult and remain unsolved. If he was right, there’d be no simulation competitions and if you were right, nobody would bother with ImageNet.

What does a proprietary Mobileye solving vision sensing for sdc have to do with a research competition that has nothing to do with self driving cars?

 

[MarcusMaximus]

What does a proprietary Mobileye solving vision sensing for sdc have to do with a research competition that has nothing to do with self driving cars?

If you’ve really solved vision, then you’d be able to solve, well, vision. ImageNet is a much easier form of what a car would have to do to really, 100%, solve it. Now, maybe they’ve just never entered the competition and the likes of Google, Facebook, MIT, etc have just been wasting their time for the last several years struggling with the beginning phase of a problem that was already solved. But probably not.

 

[tontod]

VW's TACC is kinda like that too. Should be an "aggressiveness" slider of some sort.

Maybe there could be a "Mad Max" mode.

 

[SlicedBr3ad]

With delays to .24 and .26 on model 3, these delays are playing into my 3-6 months delay on v9 firmware release. I think worst case for me is if s/x get v9 and m3 gets the delay

 

[Bladerskb]

If you’ve really solved vision, then you’d be able to solve, well, vision. ImageNet is a much easier form of what a car would have to do to really, 100%, solve it. Now, maybe they’ve just never entered the competition and the likes of Google, Facebook, MIT, etc have just been wasting their time for the last several years struggling with the beginning phase of a problem that was already solved. But probably not.

Actually no, its a much harder. Imagenet challenges are all geared towards general intelligence and they are definitely not struggling.

Microsoft, Google Beat Humans at Image Recognition (2015)
https://www.eetimes.com/document.asp?doc_id=1325712

In 2017, 29 of 38 competing teams got less than 5% wrong.[16] In 2017 ImageNet stated it would roll out a new, much more difficult, challenge in 2018 that involves classifying 3D objects using natural language.The applications of progress in this area would range from robotic navigation to augmented reality

Imagenet HAS NOTHING to do with sdc vision. A sdc is not trying to classify and accurately label 120 breeds of dogs into their actual category. Its not trying to distinguish a Anemone flower from a Aster flower.

Again Mobileye has solved vision sensing for self driving cars

 

[Bladerskb]

Amon and Elon are both marketers and showmen. They both have real, non-fake products but both talk about capabilities that are far beyond what their non-fake products can currently do.

Not at all. Amon does not talk about research projects. He has said it time and time again anything he talks publically about has a production agreeement/contract from a manufacturer. Unlike Elon who simply makes predictions without a line of code being already written.

Elon Musk: In Two Years, Your Tesla Could 'Summon' Itself Cross-Country

Amon says stuff like Nissan planning to release a L3 highway system in early 2019 because that's what the OEM are telling him. He never gets out there and make fairy tale prediction and guarantees to hype up his products.


If eyeq4 didn't solve vision for self driving cars then why is a bunch of self driving companies using it as their main perception vision system? The latest being Baidu? On top of already bmw, audi, Nissan, Delphi and more...

You know they dont get monthly firmware updates like Tesla that ends up regressing the software.

 

[rnortman]

Again Mobileye has solved vision sensing for self driving cars

Source, other than Mobileye itself? I'm finding it hard to imagine what would qualify as a "source" for a claim like this. You can make this claim mean whatever you want it to.

If eyeq4 didn't solve vision for self driving cars then why is a bunch of self driving companies using it as their main perception vision system? The latest being Baidu? On top of already bmw, audi, Nissan, Delphi and more...

Mobileye is the leading supplier of vision systems for driver assistance systems (L2/L3). The companies working on L4/L5 systems (e.g., Waymo) do not use Mobileye.

This is a question of terminology, but to me when you say "self driving cars" I hear L4/L5, not L2/L3.

 

[BigD0g]

Based on this thread, there's going to be some really disappointed people in a few weeks.

 

[croman]

Based on this thread, there's going to be some really disappointed people in a few weeks.

Bearer of bad news, eh? Great... Tesla hype again meets reality.

 

[diplomat33]

As someone who now owns a Tesla (Model 3) and has used AP, I feel like I can add to this discussion from a bit more authority that I could not do before in my earlier posts.

I think V9 will be good. It won't be level 5 self-driving of course but I think it will gradually introduce some good stuff. From the hints, we've gotten, I think we will get cars on the display in our blind spots and I think we will get one of the EAP features like auto highway transition. If I am right, I'd be very happy with that. Having used AP, I think AP is quite good as a driver assistant and the 8 cameras have a lot of untapped potential. I look forward to what Tesla does with them. I do think we will eventually get some cool stuff. I am cautiously optimistic about V9. I do think that it will introduce some good stuff in the not too distant future.

 

[rnortman]

From the hints, we've gotten, I think we will get cars on the display in our blind spots and I think we will get one of the EAP features like auto highway transition.

Would either of those count as an FSD-only feature? He said something that differentiates FSD from EAP is coming. I think both of those would be EAP features. (Well, blind spot display isn't promised anywhere, but uninitiated lane change is an EAP promise and obviously the car needs to be checking blinds spots in order to do that... whether it displays to the driver or not.)

Having used AP, I think AP is quite good as a driver assistant and the 8 cameras have a lot of untapped potential.

Yes, AP is at least among the best driver assistance systems available today, if not the best. I think most people who bought the FSD option thought they were buying something more than a driver assistance system though... the real question is, will AP2/2.5 ever be anything other than a sophisticated driver assistance system?

 

[diplomat33]

Would either of those count as an FSD-only feature? He said something that differentiates FSD from EAP is coming. I think both of those would be EAP features. (Well, blind spot display isn't promised anywhere, but uninitiated lane change is an EAP promise and obviously the car needs to be checking blinds spots in order to do that... whether it displays to the driver or not.)

It might depend on how Tesla implements the features. For the first one, if it is just an interface change of adding little cars behind you the display as part of a better blind spot notification system, then I think it would EAP, not FSD. But if Tesla gives the car some type of driving functionality related to the side cameras then they might classify it as a FSD feature since it pertains to the car "self-driving" in a particular way. In terms of auto highway transition, that feature was marketed as a EAP feature so I think it would be iffy to suddenly make it a FSD one.

It is possible that Tesla will introduce a different feature altogether like traffic sign recognition. I could see that fitting into FSD.

Yes, AP is at least among the best driver assistance systems available today, if not the best. I think most people who bought the FSD option thought they were buying something more than a driver assistance system though... the real question is, will AP2/2.5 ever be anything other than a sophisticated driver assistance system?

Agreed. It goes back to expectations. (And yeah, we can get into the whole debate of whether that's Tesla fault or not because Tesla released that FSD teaser that certainly gave people the impression that the car would be FSD). But If you look at a lot of the youtube videos that are critical of EAP, the criticism often boils down to EAP not being FSD. For example, "EAP did not react to a parked car!" or "EAP did not merge correctly at a busy 4 way intersection!", things that FSD will need to be able to handle but EAP does not. When you use EAP on a divided highway as intended, it is indeed an excellent driver assistant.

To your last point, I do think that AP2/2.5 will be capable of some self-driving (probably L3 but maybe a hint of L4). But I think EAP will remain a sophisticated driver assistant system in order to differentiate it from FSD. After all, if people could just buy EAP only and get some self-driving then why pay for FSD? There needs to be a big enough difference between EAP and FSD to make them worth it as separate packages. So I do think that as soon as Tesla feels that they have some features that start to get into FSD territory, they won't hesitate to make them part of the FSD package in order to separate the two packages and make FSD look more enticing. And, I think Tesla will do this by the end of this year. And, I think that folks who are expecting EAP to become self-driving without the FSD package are kidding themselves.

 

[croman]

@diplomat33 -- we can stop speculating. People like @BigD0g and @verygreen actually know what is going on and we'd rather have actual information rather than more useless speculation.

The fact you just got your car is great but that doesn't change Tesla's development or timeline (or lack thereof) for EAP features much less FSD capability.

Apparently we're all in for a disappointment come August. Best to acknowledge that and not turn to spin. Tesla needs to own up to the owners who bought EAP/FSD. We're only fed fairy tales by Elon Musk and I, for one, have had my fill. Its past time for Tesla to deliver what they've touted for nearly 2 years.

 

[Bladerskb]

Source, other than Mobileye itself? I'm finding it hard to imagine what would qualify as a "source" for a claim like this. You can make this claim mean whatever you want it to.



Mobileye is the leading supplier of vision systems for driver assistance systems (L2/L3). The companies working on L4/L5 systems (e.g., Waymo) do not use Mobileye.

This is a question of terminology, but to me when you say "self driving cars" I hear L4/L5, not L2/L3.

L3 is a self driving car according to SAE, its not a driver assistance system. SAE refers to it as an "automated driving system". It refers to L1/L2 as "driver assistance system".

Nissan, BMW, Delphi, Audi, Baidu and more are all working on L4 systems and all use Mobileye.
Also I wanted to add that L3 needs the same level of perception validation and accuracy as an L4 car because an L3 car needs to recognize every scenario that it needs to hand off control to the backup driver.


No one is realistically working on L5 right now. L5 is at-least 7-10 years out, so it shouldn't even be in discussion.

 

[d21mike]

I notice a lot of people use "divided highway" which can sometimes mean the same as "freeway = expressway" but not in all cases. Maybe if you added "limited access" it would help to clarify the difference. But there can be a major difference. We have many many "divided highways" where I live that are not freeways in the way Tesla defines them. See Enhanced Autopilot below.

I believe the major difference between EAP and FSD is where it is used. I.E. On a freeway or on surface streets. To do what is described for EAP for On-Ramp to Off-Ramp (Freeways) it will require a lot of "self driving" to accomplish that. I do not believe there is a difference between EAP and FSD while they are used on a freeway. Some think that if you only have EAP and on a freeway then you are still in full control but if you have FSD then you are not in full control but the car will be doing the exact same thing. I am not one of those. But, while not on a freeway then you will need FSD for pretty much everything like the car using Navigation to go from point A to point B and to stop at stop signs and red lights. To me surface street driving is a major difference between EAP and FSD. So, those New features in FSD that Elon is talking about will be all about surface street driving and not freeway driving. However, there will of course be more features added to EAP for the On-Ramp to Off-ramp to work but they would not be FSD Only features.

Enhanced Autopilot
Enhanced Autopilot adds these new capabilities to the Tesla Autopilot driving experience. Your Tesla will match speed to traffic conditions, keep within a lane, automatically change lanes without requiring driver input, transition from one freeway to another, exit the freeway when your destination is near, self-park when near a parking spot and be summoned to and from your garage.

Tesla’s Enhanced Autopilot software has begun rolling out and features will continue to be introduced as validation is completed, subject to regulatory approval. Every driver is responsible for remaining alert and active when using Autopilot, and must be prepared to take action at any time.