TMC is an independent, primarily volunteer organization that relies on ad revenue to cover its operating costs. Please consider whitelisting TMC on your ad blocker or making a Paypal contribution here: paypal.me/SupportTMC

Autopilot 2.0 Not Imminent Based On Production Model X Design Studio [Speculation]

Discussion in 'Model S' started by MarkS22, Nov 23, 2015.

  1. MarkS22

    MarkS22 Member

    Joined:
    Apr 6, 2015
    Messages:
    309
    Location:
    Morris County, NJ
    The Model X Design Studio for Production Models has been released. In it, they list the "Autopilot Convenience Features" which are identical to the current Model S. No stereo or tri-focal cameras, no additional sensors noted. I've believed Autopilot 2.0 hardware is further away than many of the recent speculations. Some of people were saying "by the end of the year." Can we put that to rest?

    Let's not forget Autopilot 1.0 hardware was announced with the P85D, only to be activated (Autosteer in Beta) a year later. A 70D configured today won't be out until mid-late 2016 and it's not be advertised with any new Autopilot abilities.

    I'm curious if anyone still thinks we're going to see a new hardware suite in the next couple months? Thoughts?
     
  2. robby

    robby Member

    Joined:
    Aug 25, 2014
    Messages:
    397
    Location:
    Andover, MA
    The speculation on this originated from Musk who said, in an earnings call, that Model X had autopilot hardware improvements. The Model X renderings on the subside reflected this in a stereoscopic camera, but the reveal and latest subsite images have walked it back. The most recent evidence we have agrees with you that there are no imminent changes.
     
  3. bwa

    bwa Member

    Joined:
    Dec 8, 2014
    Messages:
    314
    Location:
    Aptos, Ca
    My thoughts are often pie-in-sky and not as down to earth, and these are the pie type:

    1. I wonder how much engineering support specs the AP2.0 HW creators (that Israeli co?) have created -- what I mean is this: what would Tesla need to know in order to engineer their vehicles to be able to install AP2.0 HW? What cables would need to go where, what holes and glass and plastic where, etc..

    2. I wonder if #1 is done, if Tesla has that information.

    3. I wonder if #2 is done, if Tesla has started engineering.

    4. I wonder if #3 is done, if Tesla has finished engineering.

    5. I wonder if #4 is done, if Tesla has engineered an upgrade path into the chasis, such that AP2.0 HW could be installed after it has been released.

    6. I wonder if #5 is done, if Tesla has started to put it into the chasis of any vehicle they're producing.

    7. I wonder if #6 is done, if Tesla is even close to installing AP2.0 HW in any way, or has any desire to be HW upgradable.

    There may be some engineering reasons why HW upgrade would not be possible in the way that Tesla has approached it. Other than that, I'm just very curious of all of this kind of thing. To me it seems like if I were Tesla, I'd push the AP partner to give me future-upgrade physical and support specs so that I could engineer it into the cars coming off the line (all models) and offer it as an upgrade in the future. There's lots of engineering that needs to be done -- crash tests, etc.. -- so the sooner they do every step the better, in order to get it to market sooner.

    To be fair, #1 might not even be done; maybe they have ample Alpha testers with setups, but want to tune the positioning, layout and engineering details of AP2.0 HW in such a way they can't even tell what support, locations, views, angles, etc., it needs.
     
  4. MsElectric

    MsElectric Active Member

    Joined:
    Oct 11, 2014
    Messages:
    1,611
    Location:
    New York
    You do realize that the Model X photo on the Tesla Web site has a stereo front camera; right?

    And you also know that for the reveal they purposefully added a plastic piece to hide that second camera; right?

    I'd say they are withholding disclosing anything new in the Model X, including the stereo front camera, until they are done scheduling all the deliveries for this year.

    Autopilot 2.0 might very well be a while away but from the photo on the Model X information page on the Tesla Web site it appears that at the least the Model X has a stereo front camera. Also the ultrasonic sensors we know are more advanced and can see through the sheet metal. Within about a month we will know the full capabilities of the Model X sensors.
     
  5. luvnMyTS

    luvnMyTS Member

    Joined:
    Jan 6, 2015
    Messages:
    237
    Location:
    Los Angeles, CA
    I agree with the OP. Musk just released a statement yesterday that the current Auto Pilot version still has several "SOFTWARE" updates to come to increase its usefulness. If they were close to ready to release 2.0 hardware, unlikely they'd be working so hard on making 1.0 better.

    Also, Musk just sent out a message seeking qualified computer techs that can design autonomous driving software. That, more than anything tells me, they don't have version 2.0 anywhere near figured out yet. More so tells me they still have no clue and are on the hunt for individuals that might help them accomplish it. That being the case, I'd say we're at least two YEARS away from version 2.0. The good news is, version 1.0 will continue to get better.

    As we've already seen with the P85D promise for more power, once Musk comes up with a new package (P90DL), innovations for the prior models go out the window. When the P85D was released, it was promised that software updates would be ongoing to improve its performance. Instead, he decided to release the P90DL and tell everyone with a P85D, if you want more power now, you'll have to pay at least $5,000 more for it. Not part of the deal that was promised. I'd be pissed if I owned a P85D. I see the same happening eventually for Auto Pilot 1.0. Once 2.0 is ready, 1.0 cars will have what they have and Musk will likely tell 1.0 owners, if you want previously promised features, you'll have to spend $10,000 to upgrade the hardware on your car or just buy a new car.

    Fortunately, I think that's a couple of years away, so hopefully that means he keeps true to his promise for a car with 1.0 that will go park itself, return to pick you up, learn to read stop signs and traffic signals, etc, etc.....
     
  6. MarkS22

    MarkS22 Member

    Joined:
    Apr 6, 2015
    Messages:
    309
    Location:
    Morris County, NJ
    I believe Mobileye would already know the exact camera setup required for their multiple EyeQ3 and EyeQ4 setups. Nissan already has a test car with multiple EyeQ3s running 8 cameras (with a tri-focal up front), much like the one Tesla will likely use for autonomy. At least, based on their current path of using video for their primary sensors. So, it's possible the cabling harnesses are being put into place if they're looking to allow for an upgrade path. I don't think they'd do this, but it's technically possible.

    - - - Updated - - -

    I haven't seen anything indicating that the ultrasonic sensor is anything more than a door/ceiling sensor. From my knowledge of sensors, going through metal (and paint and clear-coat) would have a negative effect on range. I think it's good enough for a stationary ceiling from a couple feet away... and it's also a nice technical achievement. But I don't think the Falcon Wing sensor will be useful for more accurate or longer-distance Autopilot.

    - - - Updated - - -

    I think the X was supposed to introduce more new functionality, like dual motors and the original Autopilot hardware. I think the intention was that the X would be ready at the same time as the P85D. I wouldn't be surprised if stereoscopic (or multiple focal length) cameras were part of the design at some point. It looks like the next level would be the tri-focal camera cluster. Two cameras that close to one another aren't going to add substantial 3D information. A single camera is already getting a lot of 3D information from the forward motion of the car, given powerful enough software and hardware. If you think about it, the single camera is taking a photo once every, say, a standard 1/30th of a second. In that time at highway speeds, the camera has moved roughly three feet. By comparing the two frames, you can reconstruct 3D depth in the same way match-moving software (like Syntheyes or Mocha) does in the video production/special effects world. I believe two (or three) focal lengths would be more valuable than stereoscopic at this point.
     
  7. Canuck

    Canuck Active Member

    Joined:
    Nov 30, 2013
    Messages:
    3,949
    Location:
    South Surrey, BC
    It's probably safe to put to rest by the end of the year since that's only a little over a month away.

    Autopilot 2.0 will not be advertised before it is released. That would be nuts on Tesla's part. The cancelled orders would be too much to handle. They did it with the P85D only because there were no prior orders of that car and people were getting cars with autopilot hardware even before the P85D was announced (late version P85+ vehicles have it). So if history is any indication, there will be no announcement and cars will simply start to roll off the line with it. That's what Tesla did with parking sensors, folding mirrors, seats, etc. etc.

    Mercedes is already using the slogan "more sensors - more protection" as a shot against Tesla. The Youtube videos of Tesla AP in intersections (where 1.0 should never be used because the lack of cameras and sensors make it problematic -- the instructions are clear not to use it in intersections -- which limits its use severely), the problems with freeway off ramps, etc. have lead me to believe that Tesla (and Musk in particular) want to limit 1.0 hardware as fast as it can. The hardware is cheap and the limitations of 1.0 are now patently clear.

    I say by February we start to see 2.0 hardware rolling off the lines and the announcement of it will come shortly thereafter to coincide with the Model 3 March unveiling. That way, cancellations, while inevitable, will be limited to a large degree.

    Of course, 1.0 will keep updating, just like all software in the car keeps updating. But there's no way, in my opinion, Tesla is going to fall further behind other automakers when it comes to something as simple as adding cheap sensors and cameras, especially when Tesla has it goal set as being the leader in this field.
     
  8. MarkS22

    MarkS22 Member

    Joined:
    Apr 6, 2015
    Messages:
    309
    Location:
    Morris County, NJ
    I agree that they would release it and make it immediately available (on both the X and the S). I think some people were holding out that the new features would launch with the production X.

    - - - Updated - - -

    Mercedes does have some safety advantages, like Pre-Sense Rear, but I haven't seen anything remotely close to Tesla's Autosteer (even on S-Class Mercedes and stereoscopic Subarus). In theory, Autopilot could do their own version of Pre-Sense Rear (and make lane changing safer) if they used the existing rear camera for safety functions.

    That said, I don't think this is a hardware fight. This is a software fight. There's still a substantial amount of work that can be done to improve Autopilot with the current sensors. And only Tesla currently has OTA updates. To your point about "simple" and "cheap" sensors/cameras, they will be a commodity. The company with the software and infrastructure to use the data is what matters.

    Personally, I'd rather they wait a little longer while the software bakes. Maybe add some safety functions, like Mercedes, but release new hardware closer to when the next level of autonomy is ready in software and regulations. Why? Because that's likely to be three years or more away. Cameras and processors will continue to improve dramatically. I wouldn't want a three year old camera array when higher resolution, higher dynamic range, better low-light cameras are available when the software is ready. Even multiple EyeQ3s aren't as good as the upcoming EyeQ4 (which has new features, like two additional vector processors).

    At this point, I'm most curious about whether they'll do a stop-gap (like a second wide angle front camera for intersections and integrate the rear camera) or jump to a full 8-camera system.
     
  9. Canuck

    Canuck Active Member

    Joined:
    Nov 30, 2013
    Messages:
    3,949
    Location:
    South Surrey, BC
    #9 Canuck, Nov 23, 2015
    Last edited: Nov 23, 2015
    But the current hardware leads to a dead end...

    So will this sensor suite be enough to enable autonomous driving eventually? In a word, no. “This is not the full autonomy sensor suite,” says Musk

    http://www.wired.com/2015/10/tesla-self-driving-over-air-update-live/#slide-1

    To me, that makes no sense and is a good way to alienate a ton of owners. We have the hardware now for full autonomous driving. But that's not the issue. Regulatory approval is three years away for that. But just going through an intersection is not possible with the current cameras and sensors regardless of software updates because the field of vision is too narrow. That can be solved tomorrow with new hardware. Plus, as a bonus, those vehicles will be ready for 2.0. That's also the best way to avoid the AP fiasco the first time round. Put 2.0 hardware on the cars but it can't be fully used for a couple of years. Then the vehicle that just missed it on the assembly line was sold two years ago. Problem solved!

    As stated in another thread on this issue, I am bias since I am waiting for 2.0 before upgrading because I don't want to get burned twice after missing out on AP hardware the first time round. So take what I say as my own uninformed opinion. No one really knows except the insiders at Tesla and they're not talking. It will only be time that will tell.
     
  10. veryannoyingname

    Joined:
    Apr 19, 2015
    Messages:
    35
    Location:
    United Kingdom
    I thought Tesla was working with Nvidia and Bosch for their autopilot systems. May be I am wrong but am basing my thoughts on these videos and the fact that elon and Nvidia ceo are often together and he also has several Tesla cars.

    See GTC 2015: NVIDIA DRIVE PX Self-Driving Car Computer and Deep Learning (part 8) - YouTube
    And How Deep Learning Will Enable Self-Driving Cars - YouTube
    And Opening Keynote at GTC 2015: Leaps in Visual Computing - YouTube
    Here is nvidia talking with elon GTC 2015: NVIDIA CEO Jen-Hsun Huang Interviews Tesla Motors CEO Elon Musk (part 9) - YouTube
    And they seem to be working with Bosch too as per this video Tesla Model S - Bosch Automated Driving - YouTube
    The Nvidia system can handle 12 cameras.
    There is another Bosch highway driving video on YouTube from several months ago in a Tesla
     
  11. MarkS22

    MarkS22 Member

    Joined:
    Apr 6, 2015
    Messages:
    309
    Location:
    Morris County, NJ
    Yeah, I'm curious which route they'll take. Will they put in an "autonomous capable suite" in as the next step, or simply add a second wide angle camera and solve incremental problems? My concern is how quickly cameras and processors are improving. Would a 1Q 2016 "Autopilot 2.0" truly be the best setup for a 2019 autonomous launch?
    Case in point, the 2.8 second 0-60mph for the P85D. I believe Tesla thought it could get the original P85D down to 2.8 seconds through software, but then realized they needed a better fuse.

    I think there are arguments for both cases. I know you want that upgrade, so I hope you're pleasantly surprised when the Model 3 is announced.

    - - - Updated - - -

    Based on Earnings Calls, the current Autopilot suite is powered by the EyeQ3 processor from Mobileye:

    "Tesla announced plans to provide all of its Model S cars with full ADAS functionality, which also includes our camera technology running on our EyeQ3 chip"
    Source: Mobileye's (MBLY) CEO Ziv Aviram On Q3 2014 Results - Earnings Call Transcript | Seeking Alpha

    An Nvidia Tegra powers the car's user interface. Bosch, I believe, did their video independently as a technology demo.

    I think the fact the CEO of Nvidia got a Founders X is very interesting. Could this be a hint of a switch from the EyeQ3 to the Drive PX in a future version? Sure. However, as of now, all signs seem to indicate the next generation will be multiple EyeQ3s or the EyeQ4 (depending on release date). In this video from the CEO of Mobileye, you'll see he addresses his perceived advantages over the PX, but you can read between the lines and see they're clearly competitors:
    The Future of Computer Vision and Automated Driving by Prof. Amnon Shashua - YouTube
     
  12. mn4az

    mn4az Member

    Joined:
    Aug 27, 2015
    Messages:
    151
    Location:
    USA
    Here’s my logic….. Which admittedly is feeding into a confirmation bias as I made the decision to wait for hardware to be refreshed.

    https://youtu.be/0UzVBTgHqSQ?t=839 - Amnon Shashua of Mobil Eye on March 23, 2015

    “By 2016 there’s going to be new launches by GM and Tesla as well”

    Supplier hints at next generation Autopilot hardware for Tesla as soon as this year | Electrek (9/9/2015)

    “Today we are already preparing with one of the OEM, a first vehicle based on 8 cameras, one radar and ultrasonic around the vehicle. So this is much wider implementation of the first introduction of semi-autonomous driving and the trifocal is going to be here as we planned, but additional 4 cameras around the vehicle and one camera looking back. The system will run on 5 EyeQ3 chips and all of them will be connected.”


    Mobile Eye’s Q3 Conference Call (11/3/2015).


    “As we said, these capabilities are already implemented and will be implemented in the future semi-autonomous launches, including in 2016 by two of our OEM customers (from the YouTube clip above this would be GM and Tesla). The Tesla auto pilot feature is currently using a mono camera sensor for performing the most important understanding of the scene the visual interpretation. Our multiple camera sensor configuration launches are planned to begin as early as next year”

    "So, currently, we are very happy with the introduction of Tesla with what we call Lane Keeping Assist, which is the best in class Lane Keeping Assist today. And we hear a lot of great feedbacks on this system. “

    “What we presented is Lane Keeping Assist system rather than auto pilot system. Auto pilot system is going to be presented next year, where is going to be 360 degrees coverage around the vehicle and is going to be multiple cameras with additional sensors. What we have today is just a mono camera looking forward. So it’s a very limited input that we have on the road. But importance of this launch is, Tesla is willing to push the envelope faster and more aggressively than any other OEM, and definitely this is a very important in step forward to introduce the beginning of semi-autonomous application that will start being launched next year."


    Elon Musk on Twitter:

    "Should mention that I will be interviewing people personally and Autopilot reports directly to me. This is a super high priority."

    My theory.....

    Just like the v1 hardware did the v2 hardware will just "show up" unannounced in cars (February/March?) with the same lane keeping assist software that Tesla has today. Over time the software will improve on top of the v1 and v2 hardware. However, software with the v2 hardware will eventually allow semi-autonomous driving and this is what Musk's "super high priority" tweet was all about.

    I believe the MS/MX needs to have something special to go with the M3 reveal and/or at its FY15 results investor briefing. Having v2 hardware (8 cameras and additional EyeQ3 chips that would enable a more semi – autonomous driving experience) would/could be that "something special" that Musk/Tesla can market as APv2.

    The nextgen MobilEye product is the EyeQ4 and it’s not going to be available until 2018 – about the time the M3 shows up. I speculate Tesla putting the EyeQ4 it in all their cars (or having the option for it) at that time.
     
  13. peterspd

    peterspd peterspd

    Joined:
    Nov 1, 2013
    Messages:
    71
    Location:
    WA State
    Does it total 6 (existing one plus ... additional 4 camaras around the vehicle and one looking back or 8 cameras? What semi-autonomous can do vs full autonomous?

    “Today we are already preparing with one of the OEM, a first vehicle based on 8 cameras, one radar and ultrasonic around the vehicle. So this is much wider implementation of the first introduction of semi-autonomous driving and the trifocal is going to be here as we planned, but additional 4 cameras around the vehicle and one camera looking back. The system will run on 5 EyeQ3 chips and all of them will be connected.”


     
  14. MsElectric

    MsElectric Active Member

    Joined:
    Oct 11, 2014
    Messages:
    1,611
    Location:
    New York
    I wish Tesla would just figure out a way to make the Autopilot sensors and processing modular so you can upgrade to future versions. Basically have enclosures around the car for the 8 cameras along with the Ultrasonic and radar sensors and have them all feed data to a central Autopilot processing hub. Set it up so that each of those sensors are plug and play and can be updated. Imagine if they figure out a way to incorporate FLIR/Laser imaging along with an optical camera in the same space in the future? With such an approach you can update those sensors when they become available.

    This way you always have the option of paying to upgrade your Autopilot hardware the same way customers have done with the LTE upgrade.
     
  15. gizmoboy

    gizmoboy Member

    Joined:
    Jul 2, 2015
    Messages:
    611
    Location:
    Torrance, CA
    I'm sure they know how to do this, but it's not strategically important to them. Having features that drive existing owners to upgrade is more valuable to them than future-proofing existing cars.
     
  16. mn4az

    mn4az Member

    Joined:
    Aug 27, 2015
    Messages:
    151
    Location:
    USA
    I view it this way....

    Semi - autonomous: means the driver still must be in the seat with the steering wheel and able to engage as needed.
    Full - Autonomous: means the driver can get in the back seat and have the car drive the person to their destination
     
  17. brkaus

    brkaus Member

    Joined:
    Jul 8, 2014
    Messages:
    948
    I think upgradability is unlikely. New features sell new cars. But more importantly, it adds cost to the current version and ties the hands of the future versions.

    I understand the software and laws are likely behind. But I want some sort or 360 view to safely park in my excessively tight garage. I can wait for more auto pilot features.
     
  18. shokunin

    shokunin P85 & S40

    Joined:
    Feb 28, 2012
    Messages:
    960
    Location:
    Irvine, CA
    The way I look at it.. what can be done with the current hardware? What is the next evolution of Tesla's vision for auto-pilot and can it be done with the current hardware? I think the answer is "not much" more with the current hardware.. IMO, the next set of "lane keeping assist" features would be to use it on streets vs highway. I don't think the current camera is setup for identifying stop lights and stop signs when there is not a car in front of you.
     
  19. donv

    donv Member

    Joined:
    Jul 15, 2013
    Messages:
    567
    Location:
    Lake Jeanclia, OR
    I'm pretty sure it can identify stop lights and stop signs, actually. What it can't do is determine if it's safe to cross an intersection, since it can't see very far to the sides.

    It wouldn't surprise me if they brought out stop light/sign detection as a feature with current hardware, and had it go into a "hold" mode when stopped. That way, you have to check that it's clear on the sides and press on the accelerator pedal to get it going again.

     
  20. Canuck

    Canuck Active Member

    Joined:
    Nov 30, 2013
    Messages:
    3,949
    Location:
    South Surrey, BC
    Yes, that and it can't see far enough to the sides to even know it's an intersection which accounts for the vehicle reverting to following the vehicle in front, even changing lanes, when the road lines disappear in an intersection as seen in the Youtube videos.

    Hence the urgent need for just one cheap additional camera to solve this problem but that would make no sense. Make it 8 cameras and sensors total and when the regulators give approval it's ready for autonomous driving.
     

Share This Page