Thứ Bảy, 28 tháng 1, 2017

Nobody talking about the Full Autonomy hint from Elon at Code Conference?! part 1

  • Jun 2, 2016
    flamingoezz
  • Jun 2, 2016
    McHoffa
    Two things he said:

    1. At the end of the year they will make a big announcement in regards to autonomous driving and the Model 3
    2. One of the people asking him questions asked about full autonomy, and he said that the technology will be in place in less than two years (though regulations will prevent that from being activated for a while after that)

    Also, when asked if the Model 3 would include autonomous driving at the $35k base price, he said he couldn't answer that and referenced the event they will hold later this year.
    �
  • Jun 2, 2016
    geoffreak
    I'm definitely one of the people who thought the Model 3 would likely have fully autonomous hardware (or at least level 4). I was starting to go back on this and think it was unlikely until Elon said that. I think it's very clear that we'll see the S and X updated right around the part 2 reveal sometime later this fall.

    There is still a possibility that level 4 or 5 autonomy won't be until APv3, but this would be weird to only be released a year after the Model 3 release (on par with Elon's 2 years quote). The other, more likely possibility is that the software will take an additional year or two but the hardware is ready now. Given that Tesla has admitted to testing future AP software in a dormant manner, so they may need a year or two of tests.
    �
  • Jun 2, 2016
    model3fan
    I'm just stoked there will be a 'big' model 3 announcement end of 2016. Wasn't expecting anything official until close to production.
    �
  • Jun 2, 2016
    hockeythug
    Posted this in the investing section. Also fits into Musk's comments about what will separate the Model 3 from the S and X.

    �
  • Jun 2, 2016
    GoTslaGo
    He did comment in the shareholders meeting that S/X will be technology leaders and 3 will get "trickle down" technology.

    I suspect base model 3 will have AP 1.0 hardware (what S owners have now) with option to get software (pay extra for SW like current model S). Future S/X will probably get AP 2.0 and future autonomy packages (hardware included, not sure if SW will be).

    Suspect it will be a full hardware/software option (AP 2.0 and/or future autonomy) for Model 3 for purchase.

    Edit:

    Yes, exactly as @hockeythug said too!
    �
  • Jun 2, 2016
    geoffreak
    I would have to disagree with you. Elon has very clearly stated they will not hold anything back from the 3 [citation needed]. Elon has also made it clear he wants this car to be as safe as possible, and that he views AP as much safer than a human (and presumably better versions even more so). Given this information, I find it very hard to believe that the 3 won't have APv2. Also consider that APv1 wasn't designed to be produced cheaply, whereas APv2 will be.

    The advantage that the Model S/X have is that they get it first... by the 9 months or so that it takes to produce the 3. Anything they don't put into the 3 at launch won't arrive for at least a couple of years and they intend to stay ahead of their competitors when it comes to autonomy.
    �
  • Jun 2, 2016
    McHoffa
    I just don't think it would be a big announcement that the 3 would be getting hardware from 2014 in their cars. I strongly suspect the S/X get 2.0 hardware later this year and activated next fall, and the 3 gets it as well to be activated later. I also suspect that given his comments and lack of comments when asked if the $35k base model would be fully autonomous that it will in fact get it. I think as far as the trickle down features, we're talking more about other cool features. With so many other companies going to autonomy, including the Chevy Bolt, I think autopilot will become a standard feature. I mean, I highly doubt he'll allow Chevy and Ford to outdo Tesla.
    �
  • Jun 2, 2016
    GoTslaGo
    ^^^
    You definitely have a point. The more I thought about it, it would be logistically painful/stupid to install a separate AP 1.0 in model 3 when everything else (S/X) will be going to AP 2.0. Software will be the issue.
    �
  • Jun 2, 2016
    DrivingTheFuture
    Is the Bolt really going to have autonomous features besides lane keep assist? Just wondering if you heard something and from where... from what I learned in a recent motor trend video the closest lane keep assist with other manufacturers cars comes to teslas AP is that it just ping pongs back and forth between lane markers (which would have me dry heaving in seconds :) )
    �
  • Jun 2, 2016
    ModelNforNerd
    Audi and Mercedes have recently pulled off "trickle UP" effects with their lines. the new CLA and the A3 received tech upgrades that hadn't been available in other models, and then those models picked them up as part of the refresh cycle.

    Tesla might not do anything similar, since they don't do traditional refresh cycles anyway.


    But there are at least 2 recent examples of it being done by other automakers.
    �
  • Jun 2, 2016
    electracity
    Level 5? Has Musk announced teleportation?
    �
  • Jun 2, 2016
    ModelNforNerd

    Level 4= controls are present and a human can drive at any time, but does not need to.

    Level 5=no controls, no expectation of human control at any time.
    �
  • Jun 2, 2016
    geoffreak
    Lane keep assists (yes, the ping-pong variety) are available in cars in the mid-size market and maybe in the compact market too. I would see it as a competitive disadvantage if Chevy left off such a system in the Bolt.

    I think you're thinking of a different level system. Here's info about autonomous driving levels.
    �
  • Jun 2, 2016
    Adrien
    As Bernie Sanders would say, "This is HUUUUUUGGGE"

    So it's official that some reveal about Autopilot will be later this year. The question now is to what extent. I say that given the statements on the controls not being finalized, the lack of a instrument cluster, the feeling of "a space ship" I think all this leads to quite possibly the car being ready for Level 4 autonomy. I have been reviewing MobilEye's presentations and I believe the car will have at least 180* vision with three cameras on the EyeQ3 chip. Quite possibly the EyeQ4 chip but I'm not holding my breath.

    Either way, to me at least, it's all but confirmed some major autopilot advancement will be made for the reveal. Clearly sleeping in your moving car won't be legal for a while, but quite possibly the hardware could be in the car to do it
    �
  • Jun 2, 2016
    Omer
    I think a few of you are misunderstanding the value of full autonomy. Holding v2.0 back from Model 3 is based on it being a feature -it's not a feature.

    Full autonomy allows Tesla to turn on at any point a car sharing service with an established fleet of vehicles that have already been paid for. Most owners would be happy to earn extra money lending out their Tesla for an autonomous Uber. Autonomous Uber is the future of all transportation. Tesla's advantage over Uber/Google is those companies have to lease or buy the car fleet.

    The moment the hardware is available, Tesla will put it in every car because it's actually a future massive revenue driver to Tesla. If the hardware is $2k then they will assume that $2k investment will yield significant return so they actually don't need to get repayment for that investment up front when they sell the car.
    �
  • Jun 2, 2016
    ecarfan
    I listened to the entire Code conference interview a few hours ago. My recollection is that Musk said there would be a significant Tesla "event" before the end of this year. He did not specifically say it was a "Model 3 event".

    He said the topic of the even would be "the obvious".

    I can see only one rational way to interpret that phrase: Elon will announce some high level of autonomous driving capability on the S and X. It will be similar to the way that AP was introduced in September 2014: suddenly all cars being built have new hardware, there is an event where Elon announces what the new hardware is for (in that case it was the first version of AP) and that the software will be released soon (in that case it took about a year, certainly longer than he intended it to take).

    The event at the end of this year will, in my opinion, be an announcement of something like L4 autonomous driving, near the date of the event all S and X being built will have a new hardware suite of sensors, and the software will follow at some point in the future.

    I'm as certain as I can be that is what will happen.
    �
  • Jun 2, 2016
    Alketi
    I agree. I listened as well and I came to the same conclusion.

    He did make those statements in the context of "will the base Model 3 be autonomous", so I think there's still a chance that the event could also be part 2 of the Model 3 reveal. Time-wise, if they put "pencils down" in mid-July, then that would give them 4-5 months to build some functioning prototypes of the final car. It's quick, but possible.
    �
  • Jun 2, 2016
    ModelNforNerd

    We didn't realize how close to the "finished product" the mules from the Launch Event were until after the fact. So no....not impossible at all.....
    �
  • Jun 2, 2016
    CarlitoDoc
    Even if AP V.2.0 is an "expensive" option.....I will Gladly pay for it !!!!
    �
  • Jun 2, 2016
    ModelNforNerd

    They don't want to make it TOO expensive. Remember, if they're approaching this in a phased approach for regulatory purposes, they need DATA.

    They need to go to the NHTSA and say, APv1.0 prevented ____ accidents. APv2.0 prevented ____+X accidents.

    I know ALL AP hardware equipped cars will be sharing traffic and mapping data, but Tesla needs data with the vehicles actually IN CONTROL on the roads. They can't "price out" the beta testers. ;)
    �
  • Jun 2, 2016
    geoffreak
    @ecarfan I completely agree that Tesla will be launching APv2 for S/X at such an event. I would go a step farther than that and say the 3 would also be included for reveal part 2. It seems clear to me that the only thing missing from the 3 is autonomy and the interface for it (HUD or whatever). I think the only reason there is a part 2 is that the S/X wasn't ready to receive the new tech and they didn't want to Osborne themselves.
    �
  • Jun 2, 2016
    AZ Desert Driver
    I watched the event, and when the question was asked of Elon, he leaned back and rolled his eyes. From that body language, we are speculating AP2. Could be...but that is based on a head fake. I'm excited too, but .......
    �
  • Jun 2, 2016
    ModelNforNerd

    But even that is good news for us.

    It lets the S/X owners be the "technology leaders" he promised they would be at the Shareholders Meeting, especially if they quietly start rolling it out like they did with the refreshed S....and it still leaves enough lead time for him to say "yea, the hardware will be standard on the 3 as well".
    �
  • Jun 2, 2016
    electracity
    Full Autonomy is Level 4, which is the option to sit in the back seat.
    The current car is level 2.

    In Tesla's development cycle, Level 4 just doesn't appear as a new product. It is evolutionary. The current AP is bacteria, and level 4 is a bald eagle. (Or in keeping with Tesla's China First policy, I should probably say that level 4 is a giant panda).

    It does sound like Musk is gearing up to make some big claims about the potential of the new hardware.
    �
  • Jun 2, 2016
    Troy916
    I thought of part 2 reveal for the model 3 he said that too.
    �
  • Jun 2, 2016
    ModelNforNerd

    But APv2.0 could be HW for now....and a software update as regulatory hurdles are cleared and data is collected. To use your own analogy.....he's giving us the bald eagle, just keeping it caged until the time is right.
    �
  • Jun 2, 2016
    Kenypowa
    And this is relevant: Tesla Autopilot 2.0 hardware possibly spotted on Model S test mule.

    I doubt the new hardware is present already, but they may be present on the new S and X a couple weeks before the part 2 unveil. It's like how they snuck the 75 KwH battery and didn't announce it until a few weeks later.
    �
  • Jun 2, 2016
    ModelNforNerd

    Yes, they have a history of this. People ordered Model S's and saw their delivery date slip a month with no explanation....then the refresh was announced, and they were told they were getting the updated version.

    So this scenario wouldn't surprise me at all, and shouldn't surprise anyone who has been following Tesla.
    �
  • Jun 2, 2016
    McHoffa
    He specifically said that in relation to the Model 3 though.

    Walt: "That car will have autonomous for $35k?"

    Elon: "Umm, I have a, uhh... I'm going to hold another Tesla event, maybe at the end of the year, to talk more about that... It will be really big news if I start that here... let me just say, we're going to do the obvious thing... it's really obvious"
    �
  • Jun 2, 2016
    McHoffa
    obviously we don't know if this car is Level 1, 2, or 3 autonomous, but it's something
    Why GM's Autonomous Chevrolet Bolt is Running Around San Francisco
    �
  • Jun 2, 2016
    thebigbobowski
    Last I read, there will be NO autopilot software on the Bolt:
    Exclusive: Inside The Chevrolet Bolt With Its Chief Engineer - New Details

    I specifically asked if the Bolt would have adaptive cruise control, lane departure assist, self parking, automatic braking or any autonomous driving features and I was a little surprised to hear it won�t, not even as optional equipment.
    �
  • Jun 2, 2016
    McHoffa
    Interesting. That was 5 months ago. The one seen actually driving around with autonomous features was in May.

    also, there is this
    GM and Lyft to launch autonomous Chevy Bolts in the next year
    �
  • Jun 2, 2016
    zenmaster
    If the driver must take over when the car gets confused, how is that considered "fully autonomous"? The U.S. DOT regs say "Level 5" is fully autonomous, not level 4.
    �
  • Jun 2, 2016
    favo
    Towards the end of the year or beginning of next year, I expect Elon will do a presentation on AP 2.0, including saying that the last several weeks of Model S/X coming of the line have the new hardware (which TMC will already know). I will be surprised if he doesn't also say that Model 3 will get the same hardware (maybe as an option-but more likely hardware standard, safety software features standard, convenience software features an option). The software will roll out to Model S/X after a couple months, so those folks will be on the cutting edge. Model 3 will come out 6-12 months later, so AP 2.0 will have essentially trickled down from S/X.

    AP 1.0 has too many caveats and corner cases that aren't handled. Model 3 owners are likely to be somewhat less tolerant of "beta" features. There would likely be accidents as a result of people not RTFM. The negative publicity from this would far outweigh the cost of the AP 2.0 hardware. I can't see Tesla holding back on the AP features for Model 3. I have no problem seeing them release new software versions on Model S/X first, but those features would trickle down to Model 3 before too long.
    �
  • Jun 2, 2016
    favo
    While I agree that Uber and similar services (maybe a Tesla version) will jump all over autonomous vehicles, and that will significantly change the landscape of transportation, I think you are way overstating how many Model 3 owners would pimp out their cars for Uber. I would be shocked if it exceeds 10%. People willing to plunk down $35K - $60K for a car don't want to let strangers have their way with it totally unsupervised. I barely even want to give rides to my family and friends with their dirty feet, greasy fingers, and tendency to bump into the outside of the car :eek: (but I will).
    �
  • Jun 2, 2016
    zenmaster
    Typically software is released on all compatible platforms as it has nothing to do with production costs.
    �
  • Jun 2, 2016
    McHoffa
    For the returns you'd receive pimping your car out, it wouldn't be worth the hassle. People are already rude and disrespectful to Uber drivers, but can you imagine the things they might do in a car with no driver to see them??
    �
  • Jun 2, 2016
    CarlitoDoc
    Mmmmmmm, No. Not likely, but that is just me.
    �
  • Jun 2, 2016
    grommet
    Elon has stated, not specifically for Tesla, that he expects the full autonomy technology to be available in 2 years... and then another year for regulatory issues.

    The next-generation Mobileye EyeQ4 SOC will hit production around 2018; we can assume their technology will continue to be used as they add features... since Tesla currently uses EyeQ3 on Model S & Model X for the basic Auto Pilot functions it has now.

    As such, don't expect the far more affordable & easy to manufacture ("keep it simple") Model 3 to be fully self-driving in V1.

    FYI: I live in the world of very expensive Google Self-Driving prototype cars here... there isn't a day I don't see at least 3. :)
    �
  • Jun 2, 2016
    Alketi
    I think your conclusions are probably correct. Level 4 autonomy is looking like it will become available sometime in 2018.

    So, I suspect that the end of year announcement will be that the S/X will begin shipping with new sensors that allows Level 3 autonomy (essentially robust, full highway control), and that the Model 3 will ship with the same hardware. We'll see. Can't wait.
    �
  • Jun 2, 2016
    ModelNforNerd
    Just to pump the brakes on the Uber thing....

    USAA sent me an addendum to this year's coverage....they won't cover ANYTHING if you're ride-sharing.

    Not sure about you guys, but I'm not losing insurance coverage on my $70,000 car because I wanted to make a few bucks shuffling drunk people home.
    �
  • Jun 2, 2016
    McHoffa
  • Jun 2, 2016
    ucmndd
    The pace at which autonomous driving tech is advancing is astounding. It's also fun to watch the progress. My office is across the fence from one of google's main testing facilities. They are apparently teaching their fleet to be aware of and respond to emergency vehicles at the moment, as they've been chasing the little pod cars around with ambulances and police cars for a few weeks now.

    Very excited to see Tesla's progress.
    �
  • Jun 2, 2016
    zenmaster
    No, this is won't be the case in the near term of course. I personally totally believe that there will be huge advances made compared to what we've seen. And these advances will go a long way toward making it seem to many like full autonomy is immanent, improving public perception and confidence of the tech. However, in the near future no car will be able to meet that standard. For example, to be able to properly negotiate many types of completely human-navigable hazards, or understand police or road worker's or emergency personnel's directions, or figure out how to deal appropriately with pedestrian crowded areas, or function adequately in many weather conditions, or handle blind driveways. Therefore, what Musk is relying on is a certain reaction to a carrot on a stick approach, whereby a some degree of demonstrable progress sets expectations. There will be implicit forgiveness and indulgence for any shortcomings to that standard, because "good enough" or "close enough" for most people's needs, most of the time, will simply suffice for actual full autonomy.
    �
  • Jun 2, 2016
    GoTslaGo
    Batons out? :D

    That must of been fun to watch!
    �
  • Jun 2, 2016
    Alketi
    I like that Google is testing all possible scenarios.

    Follow that car!
    [?IMG]
    �
  • Jun 2, 2016
    Clomer
    The NHTSA has defined 4 levels of autonomy, not 5.

    Level 1 is where just speed is controlled, as in a TACC system.
    Level 2 is where both speed and steering are controlled. The expectation at level 2 is that the driver remain alert and able to take over at any time. Tesla's current autopilot system is level 2.
    Level 3 is where the system is mostly able to drive from point A to point B without input from the driver, and in fact is successful the majority of the time. If it encounters a situation it cannot handle, it is able to pull over and/or safely stop as it alerts the driver that human input is needed. The expectation is that the driver has time to go from complete disengagement to assessing the situation before taking action. This means the driver can watch TV, read a book, etc., and doesn't have to maintain full alertness about the surroundings.
    Level 4 is the holy grail where the only input that is needed or expected is the destination. The car can fully drive itself, and can even be unoccupied.

    Some people have also described a level 5 where the car doesn't have manual controls (ie, no steering wheel or pedals). Such a car cannot be manually driven, but in terms of the NHTSA's definitions would still be a level 4 autonomous vehicle.
    �
  • Jun 3, 2016
    zenmaster
    Top 10 Autonomous Car Facts: When Will Self-Driving Cars Arrive, What's Holding Them Up?


    Basically, SAE level 4 "high-automation" is what Tesla considers fully autonomous. This doesn't mean the car can always operate without a driver, rather it means that it can perform well enough to keep passengers safe in most situations. That is, it will not proceed when road or traffic conditions are confusing and possibly alert driver to take action. That makes sense, but level 4 is not fully autonomous in the actual sense as it still needs someone to understand conditions like broken down vehicles, construction, non functioning traffic signals, many types of detour signs, etc.
    �
  • Jun 3, 2016
    sandpiper
    I will go on record that I will consume a hat (my choice of hat, seasonings, method of preparation and side dishes) if the Model 3, S or any car is Level 4 in 2018. The lawyers haven't even opened their briefcases yet. The autopilot in my S is a nice feature but it's really just a very advanced cruise control. It fails regularly in all manner of situations and it's nowhere near the point where you can trust it without continuous monitoring. To go from here to level 4 in two years is just not realistic - unless Elon has managed to find a crack in the matrix and has purchased technology from the real world.
    �
  • Jun 3, 2016
    JeffK
    I don't know.
    Teslas on the road are currently level 2 (maybe 2.5), Google has been working in the level 3 space for a long while now. It's not that big of a leap to cover 99.99% of all driving situations. That fraction of a percent left over will be unavoidable accidents, but it'd be statistically far better than humans driving.

    Just because you shouldn't have to control the car with level 4 autonomy doesn't mean you can't take over if you think a situation arises that it cannot handle like driving around an elephant standing on one foot supporting a dancing monkey with a hat, but day to day operations should be relatively easy.
    �
  • Jun 3, 2016
    Alketi
    I think you described why Google's car is technically level 4 autonomous. It's end-to-end autonomous, from driveway to destination, with a professional driver only on standby in case something goes wrong.

    FAQ � Google Self-Driving Car Project
    �
  • Jun 3, 2016
    JeffK
    They always claim level 4 but I've never seen high speed demos from Google. When you don't have highway speeds I'm not sure it's appropriate to claim level 4 if it can't actually get me to a destination.
    �
  • Jun 3, 2016
    tga
    Don't forget about the liability issues as well. Screw the $70k car. I'd be more concerned about being sued and losing my house, retirement fund, and having my paycheck garnished for eternity...
    �
  • Jun 3, 2016
    Alketi
    C'mon, highway is the easiest by far (no stop signs, intersections, turns, driveways, one-way streets). City driving is at least an order of magnitude more difficult. Plus, they still have their Lexus cars running at highway speeds.

    That's why my guess is Tesla's end-of-year announcement is highway level 3, it's the next logical step.
    �
  • Jun 3, 2016
    ModelNforNerd

    Exactly, that is all a side effect of your insurance company abandoning you for violating your policy. You'd be completely exposed to lawsuits.

    And even if you were to go about doing it the "right way"...how expensive is livery insurance? I'm not going to look it up, because I don't care....

    but if you just made the huge investment of the car, why put yourself further in the hole? Now you HAVE to use it as an Uber to justify the added expense. Tail wagging the dog......

    I'll just drive it myself and be happy. :)
    �
  • Jun 3, 2016
    ModelNforNerd

    Sorry, the current iteration of AutoPilot is already qualified as "Highway Level 3"

    U.S. Department of Transportation Releases Policy on Automated Vehicle Development | National Highway Traffic Safety Administration (NHTSA)
    �
  • Jun 3, 2016
    Alketi
  • Jun 3, 2016
    ModelNforNerd

    Limited Self-Driving Automation (Level 3): Vehicles at this level of automation enable the driver to cede full control of all safety-critical functions under certain traffic or environmental conditions and in those conditions to rely heavily on the vehicle to monitor for changes in those conditions requiring transition back to driver control. The driver is expected to be available for occasional control, but with sufficiently comfortable transition time. The Google car is an example of limited self-driving automation.




    Maybe level 2.5. it takes corners. either way, whatever level it's considered, it's the only one out there this advanced.

    I'm not counting the Google car, because it only goes 25mph and has a range shorter than a Nissan Leaf. It's a great experiment, maybe even a good limited range inner city people mover, but it's not nearly as practical.
    �
  • Jun 3, 2016
    Darryl
    During one of Elon talks last year he mentioned autonomous driving will come to Europe a long time before US. It is obvious the main reason is there is no state which has consistent road markings which is required to make it work. I am not saying there are not a few test beds but none are widespread.

    I have a Model X with the autopilot option. First Autopilot will only work if there is proper marking (lines) on the road which are clearly marked and not worn. It will work if there is a car in front of you but if the car drives into the ditch it will follow. It is not good on curves roads and you may end up off the road. Other than change lanes once you put the signal on it will not turn. It will not stop at a stop sign or yield at a yield sign or stop at a traffic light. It will in most cases go the correct speed as it can read the speed signs but only if they are black on white. If they are black on yellow it doesn't read those. It doesn't read lighted speed signs like at a school zone or curve. I can only use it a fraction of the time.

    I live on an island and the main state road going through the island is the minimum ten feet wide and there are no shoulders. Up until last month it did fine but with a software update it started driving five inches further to the right (this was to prevent head on collisions where the oncoming car was riding the line. This puts the right wheels right on the edge of the road. At times this puts it off the shoulder. If I relied on it I would be in the river.

    Based on my experience with the current capability of the X Autopilot and also based on the markings on our highways and roads, we are a long way off from having widespread rollout of autonomous driving. I am not saying the hardware is not there to do rudimentary autonomous driving but it isn't yet activated on the production cars.

    My car is in the shop for a list of 16 items. Several relate to autopilot and active avoidance. "Lane keep" and "blind spot" warning doesn't work at all. Even the service supervisor was unable to explain when we did a test drive how these two features work. I was recently in a new Volvo and when I would cross over the line the steering would vibrate and an audible signal would go off if there was a car in my blind-depot. Nothing happens on my Tesla even though it is discussed in the manual.

    The have already had myTesla 8 days and may get it back sometime next week. By then they will have some of the items corrected but I have already been told they will need the car 6-8 weeks (not days) more to correct the last of the problems. Since I originally was to take delivey of the car has spend 10 weeks at the service center. Add another 8 weeks that will make it 18 weeks in the first 6 months. I hope after this things get better.
    �
  • Jun 3, 2016
    ummgood
    I think Google is farther along than people give them credit for. I think they are close if not at level 4. Elon himself said the hardest part is 30-40mph in a populated area with pedestrians. Highway and high traffic areas are easy.

    Here is a cool video from SXSW google gave on the autonomous driving.
    �
  • Jun 3, 2016
    McHoffa
    Wow. That sounds terrible. Sorry that you've had such a bad experience. I can only hope things are better with the 3. Even a fraction of that would make the haters hate on Tesla and EVs in general even more.
    �
  • Jun 3, 2016
    sandpiper
    Uh no... the Google car is level 3, as per the document that you reference. Tesla is clearly level 2. You need to put hands on the wheel every 4-5 minutes on average even on the highway.

    Even Google is a long way from deployable level 4. Last I heard there has to be a driver intervention every 5,000 miles. When you look at the millions of cars on the road, that's a LOT of interventions every day. Large numbers demand a very high level of performance, and that last bit is going to get tougher and tougher.

    Non experimental, actual production level 4 is at least a decade away.
    �
  • Jun 3, 2016
    electracity
    Reality Check:
    Summon getting squirrelier
    �
  • Jun 3, 2016
    sandpiper
    It's really only intended for highway use, and I find it works really well there. It's not intended for in-town. And Tesla has been pretty open about this: it's a "beta" product intended only for use on the highway and with the driver paying continuous attention. Tesla has done a lot better than any other manufacturer, but it's still all very very new.
    �
  • Jun 3, 2016
    gregincal
    Both you and that article are wrong. The NHTSA is very clear about level 4 autonomy. This is directly from their website:

    Full Self-Driving Automation (Level 4): The vehicle is designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip. Such a design anticipates that the driver will provide destination or navigation input, but is not expected to be available for control at any time during the trip. This includes both occupied and unoccupied vehicles.

    Level 5 is basically identical to level 4 except that there are no controls. I.e. In Level 4 the driver is never required to be in control at any time, in level 5 they don't even have an option.
    �
  • Jun 3, 2016
    zenmaster
    Merely performing "all safety-critical driving functions" is grossly insufficient to get one reliably from point A to B. It is sufficient to avoid accidents while making the attempt, however. And while level 5 subsumes level 4, it is not simply level 4 w/o controls as, of course, it'd need to function as well as a person in all conditions. This should be obvious. Again, Tesla means SAE level 4 "full autonomy" which necessarily requires driver input. In other words, it can not reliably function properly without a driver being present.
    �
  • Jun 3, 2016
    ecarfan
    That is not the only part of the NHTSA Level 4 description as quoted by @gregincal. I think it is clear from the full description that Level 4 autonomy means the only human input the vehicle requires to safely go from point A to point B is to be given the address of point B or a set of GPS coordinates. That is all. I call that "fully" autonomous driving.

    Based on what I have read (I have no inside information) the Google self-driving cars that are modified Prius and Lexus and that I occasionally see on the road in my part of the world are very close to Level 4 cars, or perhaps are at that level. I don't have enough information to confidently make that judgement.
    �
  • Jun 3, 2016
    Electric700
    Maybe they are moving up the first Model 3 deliveries to the end of this year! That would be amazing :cool:
    �
  • Jun 3, 2016
    gregincal
    SAE level 4 is not full autonomy. It is High Automation. If you are using SAE standards you need Level 5 for full autonomy. So full autonomy means SAE level 5 or NHTSA level 4. Using "full autonomy" with anything less is completely false. It's either autonomous or not. Fully autonomous by definition means that you don't need a human.
    �
  • Jun 3, 2016
    JeffK
    haha... NO

    Elon just meant Model 3 reveal part 2
    �
  • Jun 3, 2016
    grommet
    There are two Google self-driving cars... one is a small 2 passenger Google designed electric city car, and really doesn't have traditional controls. The other is a heavily modified Lexus RX 450h technology mule. There is a fleet of these driving around me everyday. They are fully autonomous: enter destination... and it'll drive you there safely. The car -- and all of it's redundant sensors and systems -- constantly makes (conservative) decisions as it drives. These are not consumer products. :) And Google isn't the only ones with similar R&D vehicles...

    Google Self-Driving Car Project
    �
  • Jun 3, 2016
    gregincal
  • Jun 3, 2016
    zenmaster
    Read the requirement you quoted again. It clearly says don't need a human for purpose of safety when navigating. But of course a human is required for successful and reliable navigation. Got it? See the difference? That is why Level 5 is actually the one that doesn't need a human *by definition*.
    �
  • Jun 3, 2016
    gregincal
    A fully autonomous vehicle is one that can be summoned from the other side of the country, as per Elon's promise (I.e. NTHSA level 4 or SAE level5). Nobody but you chooses to call SAE level 4 full autonomy.
    �
  • Jun 3, 2016
    flamingoezz
    I don't think anyone is interested in the bickering over what you two are calling fully autonomous. Lets speculate on what we'll see on the M3.

    I'd be happy if it could fully take over on the highway with no human interaction necessary. I can handle rides around town, but am hoping my tesla can handle the bulk of long trips without me having to hold the wheel and keep my eyes on the road. Is that a realistic expectation?
    �
  • Jun 3, 2016
    calisnow
    My AP 1.0 does that now. Yes I'll be upgrading to a new Tesla as soon as 2.0 is released but given the reliability of autopilot right now - yes, your expectations for the Model 3 are very realistic.
    �
  • Jun 3, 2016
    zenmaster
    That would be a realistic expectation for SAE Level 4 on a highway under most conditions.
    �
  • Jun 3, 2016
    electracity
    That's a good question. Will AP 2.0 drive to an endpoint entered in NAV?

    I can see Autopilot 2.0 more tightly tied to Navigation. It may drive level 3 when it knows the full route.
    �
  • Jun 3, 2016
    AB4EJ
    I tend to agree. I love the whole concept of full autonomy, but many breathless futurists seem to be misunderstanding how tall a technical challenge it is to achieve; to me it is further out than many have predicted. (Hope it will not be like controlled fusion, always 5 years away). Having said that, if we start implementing robust V2V and V2I (vehicle-to-vehicle and vehicle-to-infrastructure) communication (requires adoption of standards), it could be accelerated substantially; full autonomy would then depend on the rate at which we could convert most of the existing fleet to V2V/V2I standards.
    �
  • Jun 3, 2016
    Bimbels
    I'll be interested to see the capability of the next version of AP, but I find it hard to believe the Model 3 will have full autonomy. As of now I can't use AP at all if it's snowy/Icy because the sensors are blocked - I'm curious to see how they can rectify that.

    Then again they are smarter than me. ;)
    �
  • Jun 4, 2016
    brianman
    This phrasing is troubling. It's easy to misinterpret that identification+validation of "compatible platforms" is costless, which is not the case.
    �
  • Jun 4, 2016
    electracity
    That's a self-contradictory statement. If the car has to hand control back because it starts to rain, it is level 3.

    I listened to the interview, and I don't believe Musk is talking about level 4.
    �
  • Jun 4, 2016
    FirstSea
    I don't think the AP needs to be able to handle every imaginable event or road condition to qualify as autonomous. Not every human can drive in all conditions, I live in LA and everyday I see people fail at driving in perfectly marked and flat ground. I think it's preposterous to expect a machine to perform perfectly under all conditions, where the average human would most likely fail.

    Autonomous AP will perform flawlessly in cities or any well paved road, but it will be as helpless as any inexperienced human when it comes to storms, snow or any other extreme conditions.
    �
  • Jun 4, 2016
    zenmaster
    Not free, but it's their own hardware and automakers tend to re-use most of their software across their various models. e.g. iDrive, MMI, COMAND, SYNC. Once the systems are in place the software updates aren't dictated by identification or validation cost concerns of course. This is by design.
    �
  • Jun 4, 2016
    zenmaster
    I don't think anyone thinks that either.
    �
  • Jun 4, 2016
    Chewy3
    This is why we need more autonomous cars:D
    �
  • Jun 4, 2016
    grommet
  • Jun 4, 2016
    JeffK
    Engineering samples for the EyeQ4 were available Q4 of 2015. The plan for production was "early 2018" so I wonder if they can move it up a few months for the Model 3.
    �
  • Jun 4, 2016
    zenmaster
    Or it would be nice if somehow there was an upgrade option so one wouldn't have to buy a whole new car just for that tech.
    �
  • Jun 4, 2016
    JeffK
    I'd be quite ticked off if I get a model 3 and it was a month or two short of the EyeQ4 and an upgrade wasn't possible.
    �
  • Jun 4, 2016
    AZ Desert Driver
    As I understand the relationships: the sensor suite is not very expensive and will be rolled out on all tesla models starting "soon". Those sensors are selected based on a long standing relationship with Mobileye - so the algorithms will be EyeQ based. Those algorithms are constantly evolving - but always talking to the same sensors. Tesla will update software when the time is right to update - perhaps monthly. One will not need to buy a new car to get the newest tech until the sensor suite is obsolete.

    This comes from the various Mobileye and EM talks of late - and my tea-leaves of those talks. Any disagreement with my understanding?
    �
  • Jun 4, 2016
    JeffK
    Besides the trifocal cameras, ultrasonic sensors, and radar, unless lidar becomes super cheap there's probably not going to be much in the way of upgrades unless you have camera sensors with greater sensitivity in the dark.

    Processing however can have upgrades. Since they currently use five EyeQ3 chips, I wonder how many EyeQ4 chips will be needed for a more full featured autopilot.
    �
  • Jun 4, 2016
    zenmaster
    Within a particular processing framework, the algorithms are evolving. But the new chips likely impose different requirements on sensor type, number, capability, interfaces and their various mounting locations. So is the sensor suite indeed adequate across the processing-hardware releases? If not, can the sensor suite be appropriately upgraded? And can the processing unit itself be upgraded?
    �
  • Jun 4, 2016
    grommet
    The sensors are key... and yes, there will be significant evolution in the next 5+ years in both sensors, and the technology harnessing them. Sorry, folks... progress means you'll only see some of this tech in newer versions of the vehicles.
    �
  • Jun 4, 2016
    physicsfita
    Just to add something else to the mix here -- a couple of months ago, I was at an event at the MCity autonomous-vehicle testing ground at the University of Michigan where we got to go on a walking tour of the facility. My guide, who has worked on vehicle autonomy for quite a while said that the steps for a rule-making was going on to require vehicle-to-vehicle (v2v) communication to be installed in cars starting in 2018. She said that research showed that if even 10% of the fleet could talk to each other that that was as effective as any other sensor system, and it provided data that other sensors could not get (e.g., the car three cars ahead is doing a panic stop). Adding this capability could help current systems quite a bit.

    She also said that work was being done with cell phone companies to provide a similar communication capability to help cars better detect pedestrians and bicyclists.
    �
  • Jun 4, 2016
    Booga
    Level 3 would be great. Just make long distance and otherwise mundane driving even easier. I don't yet expect to be able to summon my car from across the country.
    �
  • Jun 5, 2016
    calisnow
    Eyeq3 is capable of full autonomy - it was architected to be able to have up to five EyeQ3's networked together, each handling its own camera. EyeQ4 will be able to do this on one chip. Mobileye's CTO has publicly stated that even EyeQ3 is only using about 10% of its processing capability today. You don't need to worry about missing EyeQ4 in your Model 3.
    �
  • Jun 5, 2016
    cronosx
    Elon ( and mobileye ) stated that the lidar is not necessary.. not now, and not in the future, and in a talk mobileye continue talking about "only with the camera we are doing all of it", telling something like "before you need the radar in front, now you can use only the camera", "before you needed this.. now with a camera with more angular view you don't need it" etc etc.
    Google uses lidar because he don't have estetic problem, and want to have all the possibility of the world, if in the end, when the software is done, they discover that they don't need lidar, so be it, but meanwhile they want to have all the data they can to be sure they don't hit something, and a person in the car, just to be sure. this doesn't means it's really needed, they are just being paranoid ( and rightly so, since they are using the full AP while it's not ready )
    �
  • Jun 5, 2016
    grommet
    Isn't marketing grand? Only 10%! Future proof! ;) As I linked above, consider reviewing Mobileye's Q1 call for a little more info on where their tech is heading... it's a long, complex road to fully redundant production-ready self-driving autonomy.

    Anyway, we're all looking forward any improvements Tesla makes to the early Auto Pilot tech they have now... especially this guy: Tesla Model S driver crashes into a van while on Autopilot [Video] :)
    �
  • Jun 5, 2016
    calisnow
    Uh dude - I've not only reviewed their call I've posted detailed notes on Amnon's CES presentation - we all know where they are headed - anybody on this board who has kept up on the reading anyway - you're welcome to review my posts if you need some help with that.

    Linking multiple EyeQ3's is not marketing nor does it have anything to do with pretending that EyeQ3 is future proof - it's how the SOC was designed. Tesla chose to implement a single camera EyeQ3 system at first instead of building a multiple camera one from the get go. EyeQ4 will be more efficient and cheaper at the same task and have some vector capabilities that EyeQ3 does not, but Tesla will be building a multiple camera, multiple EyeQ3 based system prior to the Model 3's release. Tesla is still doing their own sensor fusion for the time being. Personally I think Tesla will ditch Mobileye entirely eventually but that's another story.
    �
  • Jun 5, 2016
    AZ Desert Driver
    neophyte question - can one swap out an eyeq3 chip with an eyeq4 chip - like adding memory cards or an 8087 processor? Or is this much deeper surgery?
    �
  • Jun 5, 2016
    JeffK
    Tesla is already using 5 EyeQ3 chips on current autopilot.

    It was MobilEye that said if Lidar becomes cheap their systems will work just fine with it. Lidar would simply provide redundancy to the camera systems. Camera sensors are only a few dollars so cheap is a relative term.
    �
  • Jun 5, 2016
    xav-
    So when people refer to APv2 for the model S.. They really mean adding two cameras to the front of the car? i.e. The chip is already latest and will be for a while
    �
  • Jun 5, 2016
    zenmaster
  • Jun 5, 2016
    grommet
    No, current Auto Pilot does not use 5 EyeQ3 chips on Model S and Model X.
    �
  • Jun 5, 2016
    JeffK
    The CEO from MobilEye in interviews says otherwise in an interview from last year...
    �
  • Jun 5, 2016
    zenmaster
    JeffK, we know the new AP 2.0 system will use the 5 EyeQ3 chips (EyeQ4 can do it and by itself) when it comes out, but none of this hardware is put into Tesla's production cars yet - not even the new X.
    �
  • Jun 5, 2016
    JeffK
    He said "this year" (2015) but future tense, my guess is that both the new Model S and X have the five chips...
    �
  • Jun 5, 2016
    ecarfan
    "People" mean different things depending on which person is doing the speculating.

    What seems likely is that to significantly improve AP, what needs to be added to the current sensor suite is; rear facing radar, rear facing camera (current rear camera is not adequate, some believe), multiple forward cameras of at least two different focal lengths, and something for long range on each side. The current ultrasonic sensors in the bumpers only go out to 16 ft and are long latency compared to radar.
    �
  • Jun 5, 2016
    zenmaster
    The only reason I could imagine that Tesla would do that is if they could retrofit the required sensor hardware, which is not present even in the recently facelifted S. The only reason they would not have also added the required sensor hardware is if they had supply problems. But it stands to reason the most likely case is that AP 2.0 hardware isn't in the cars being built now. I believe this important feature is something they would have announced.
    �
  • Jun 5, 2016
    JeffK
    Which sensor hardware are you referring to? The cameras, the radar, or the ultrasonic sensors?
    �
  • Jun 6, 2016
    zenmaster
    Cameras.

    AP1: forward-looking camera, a radar, and 360 degree sonar sensors
    AP2: trifocal forward camera, 2 cameras per side, one rear camera, a radar, 360 degree sonar sensors.

    The second setup will work with 5 EyeQ3 chips or a single EyeQ4.
    �
    • Người đăng: vào lúc

    Không có nhận xét nào:

    Đăng nhận xét

    Đăng ký: Đăng Nhận xét (Atom)