Tesla placed dead last in self-driving race by Navigant, GM and Waymo top list

Discussion in 'In the News' started by simonalvarez0987, Jan 18, 2018.

  1. simonalvarez0987

    simonalvarez0987 Active Member

    Joined:
    Dec 21, 2017
    Messages:
    513
    #1 simonalvarez0987, Jan 18, 2018
    Last edited by a moderator: Jan 18, 2018
    [​IMG]

    [​IMG]

    According to a recently published study by Navigant research group, Tesla is currently dead last in the self-driving race, placing beside second-to-last Apple on the list of 19 companies. At the top of Navigant’s study were GM and Google’s Waymo, companies whose initiatives to develop and release autonomous vehicles to the public are ranked as being close to perfect.

    Navigant’s analysis points the blame to Tesla and its eventual split with Mobileye, which was involved in the development and release of the first generation Autopilot system. Since its separation from the Israeli-based tech company, Tesla has spent significant effort in developing its own in-house self-driving suite - Autopilot 2. So far, however, the Elon Musk-led firm has encountered challenge after challenge, with improvements to EAP and new features trickling down in a rather slow stream.

    GM, on the other hand, appears to have struck gold with its acquisition of Cruise, a driverless startup, two years ago. Ever since its acquisition, Cruise has been able to focus on developing and improving its self-driving systems using GM’s very own mass-market electric vehicle - the Chevy Bolt EV. Over the past couple of years, Cruise has made so much progress with its autonomous systems that the self-driving startup and GM’s engineers were confident enough to request the production of Chevy Bolt EV units that do not have steering wheels or pedals. The production of these special Bolt EVs is expected to begin next year, as noted in an Ars Technica report.

    Waymo, on the other hand, has always been at the forefront of self-driving technology. Since the beginning of the decade, Google has been investing vast amounts of resources in the development of self-driving driving technologies. Based on what Waymo’s autonomous minivans in the Phoenix area can do right now, it seems like Google’s self-driving efforts are also paying off in spades.

    Overall, it is easy to see how Navigant’s study ended up placing Tesla at the lowest spot in its rankings. The Silicon Valley-based electric car maker and energy firm, after all, is still catching up to the refinement and features of its Autopilot 1.0 software from years ago. Tesla’s approach to autonomous driving is also relatively different from Waymo and Cruise’s strategy, using Shadow Mode and its drivers to collect billions of miles real-world driving data from its fleet. While GM and Google might have refined their tech to a degree beyond what Tesla has accomplished so far with Enhanced Autopilot, both companies’ vehicles have mastered pre-programmed routes but seemingly without scale. Cruise and Waymo’s autonomous cars are only effective on areas that have been heavily tested and uploaded to their computers.

    Tesla, however, is doing something far more ambitious and arguably riskier on many levels. Instead of mastering self-driving that’s isolated to specific regions, the company is aiming to roll out autonomous features that would work on a global scale through AI-based Tesla Vision technology. Looking at it from this perspective, Waymo and Cruise will probably take far longer than Tesla when it comes to rolling out their self-driving vehicles on a larger scale.

    Leaderboard for Automated Driving Systems by Navigant

    1. GM
    2. Waymo
    3. Daimler-Bosch
    4. Ford
    5. Volkswagen (VW) Group
    6. BMW-Intel-FCA
    7. Aptiv
    8. Renault-Nissan Alliance
    9. Volvo-Autoliv-Ericsson-Zenuity
    10. PSA
    11. Jaguar Land Rover
    12. Toyota
    13. Navya
    14. Baidu-BAIC
    15. Hyundai Motor Group
    16. Honda
    17. Uber
    18. Apple
    19. Tesla

    Article: Tesla placed dead last in self-driving race by Navigant, GM and Waymo top list
     
  2. vdiv

    vdiv Member

    Joined:
    Dec 15, 2016
    Messages:
    75
    Yeah, sure, if they say so. Who are they again?
     
  3. PDX-Roadie

    PDX-Roadie New Member

    Joined:
    Jan 19, 2018
    Messages:
    1
    Location:
    Portland
    That list doesn’t seem to differientate between self-driving systems that are only capable of navigating around a restricted area, or a system that can adapt to anywhere a human can drive.

    Tesla has the right approach of relying on adaptive AI and cheap hardware, rather than expensive LIDAR and precision mapping. Time will tell if Tesla can actually create an AI that takes advantage of all the data they’ve accumulated from shadow mode.
     
    • Agree Agree x 1
  4. JordanWilliamson

    Joined:
    Nov 6, 2017
    Messages:
    85
    Location:
    Ontario, Canada
    Are they speaking on production versions? I've driven in a Honda, Ford and Telsa with their top end systems. Honda as far as I'm aware can stay in lane but can't change lanes, also can draft or maintain speed but not alter speed to increase on open road. Ford again can't change lanes and such you can't remove your hands from the steering wheel at all..

    I was blown away by the Tesla, as is everyone who drives in one. If all these companies are so far ahead why does it seem like only Tesla and GM are doing this stuff?
     
    • Like Like x 1
  5. David MM

    David MM New Member

    Joined:
    Jan 19, 2018
    Messages:
    3
    Location:
    Europe
    Tesla has bet that nobody will succeed at making solid state lidar. Everybody else is convinced that solid state lidar will be possible. A year ago the odds moved against Tesla when Velodyne (the main company making LIDAR's) reported a breakthrough which would allow per system costs of around 50 dollar (given a production car). Whether they will or will not be able to deliver is anybody's guess, but they aren't the only player either.

    And beyond that, everybody is relying on self learning AI's, so not sure what your point is there. Precision mapping is useful especially as an addition in the same way that knowing an irritating intersection is useful to a human and precision mapping decreases the chances of misidentifications. Honestly, trying to defend Tesla's approach is just showing a complete lack of knowledge about self driving technology. I mean, I like Tesla as well, but there approach is simply incredibly limited. My hope is that they will succeed in a highway only solution (including getting from one highway to another), because that should be achievable with their hardware, but beyond that... :(
     
  6. imipsgh

    imipsgh New Member

    Joined:
    Jun 26, 2017
    Messages:
    12
    Location:
    US
    Aerospace Engineer-Member SAE AV Testing Task Force - (ORAD V&V)

    Navigant missed the most important criteria. Do they know most AV makers are using the wrong process to make this tech and are their embellish their capabilities?

    The leader may very well be Toyota. They appear to have always gotten the issue I mentioned. They themselves confirmed the one trillion miles.

    Waymo should be near the lead. Their recent massive paradigm shift away from public shadow driving means they unlike most can now actually create a full autonomous vehicle. Most of the others won\'t get remotely close. Most will go bankrupt trying. You simply cannot drive the one trillion miles, spend over $300B to do so and start creating casualties when the AI had to be trained on dangerous and actual accident scenarios. If Waymo has the right simulation and simulator tech, which most do not have in the AV and automotive industries, they create a top down scenario matrix for the AI and testing and put their vast funds to use they could be first. They do however have to make up what looks like 7 years doing it the wrong way.

    The only other AV maker/OEM combo who may be doing it right is Ford. Any others following the path I stated?

    I provide much more detail here

    Autonomous Levels 4 and 5 will never be reached without Simulation vs Public Shadow Driving for AI
    https://www.linkedin.com/pulse/autonomous-levels-4-5-never-reached-without-michael-dekort
     
  7. David MM

    David MM New Member

    Joined:
    Jan 19, 2018
    Messages:
    3
    Location:
    Europe
    Because a lot of companies are in agreement that Tesla's approach to self driving cars is reckless. I mean, sure, even Waymo pointed out that they considered the same thing, but the simple truth is that you can't rely on humans to be responsible in a car (do I need to show all the videos on youtube of people getting out of driver seats in Teslas?). The advantage that Tesla has is that they have so few cars on the road that chances of stuff going wrong are fairly low, but deadly crashes have still happened. Now it's easy to say it's the drivers fault for not paying enough attention, but as a UX designer I have to say that there is no way you can put a user in such a situation and say it's completely his fault. So yeah, Waymo early on realized that it's either the car having complete control or the user, trying to have the car do simple stuff and the user switch in complex situations is simply not a safe solution.

    Point is: A lot of companies either think they can not completely take over from the user (Waymo) or do not want lots of bad publicity on their hands (most companies don't have the cult like adoration Tesla has).

    I mean, honestly, if Tesla would be claiming they have the best adaptive cruis control nobody would be so irritated with Tesla, but it just isn't in the same ballpark as what the rest of the companies are doing.
     
  8. David MM

    David MM New Member

    Joined:
    Jan 19, 2018
    Messages:
    3
    Location:
    Europe
    First of all, Navigant isn't claiming they have a level of expertise where they can say that a certain process is right or wrong. They look at what the experts say, and the experts all agree that simulation is one of many things you do.

    And where did you get the idea from that Waymo only recently started doing simulations? Here is an article from the beginning of 2016:



    There has been no paradigma shift as far as I am aware.
     
  9. margiebartley67

    margiebartley67 New Member

    Joined:
    Dec 28, 2017
    Messages:
    2
    cost of LIDAR will come down with mass production. I think break-up with Mobileye will hurt Tesla more than Mobileye(Intel). Camera Vision has its own set of limitations (rain/snow/poor visibility) where as LIDARs limitation is it is expensive.

    I predict at the end, Tesla will have to incorporate LIDAR to get back on the top. Otherwise, current HW2.5 will become absolute with introduction of GM/Waymo\'s FSD cars.
     
  10. T-Will

    T-Will New Member

    Joined:
    Jul 31, 2017
    Messages:
    1
    Location:
    Rocklin, CA
    Was this list put together by a TSLA short? How can Tesla be behind Apple when Apple hasn’t even publicly stated where they are with self-driving development. 
     
    • Like Like x 3
  11. ivo welch

    ivo welch New Member

    Joined:
    Jan 28, 2017
    Messages:
    7
    the most accurate and indisputable observation is the lack of progress---and possibly regress---during the last two years. no target date even for a solid incremental update, not for self-driving but just for assisted driving and use of all the cameras. my impression is that Tesla's AP efforts are faltering. I hope I am wrong.

    while other companies are progressing, Tesla seems to lose engineers and focus elsewhere (rain sensors?!). other aspects are more speculative. look, if prices come down, tesla could install a lidar unit later into its cars, too. the long-term details are less important.
     
    • Like Like x 1
  12. Darren

    Darren New Member

    Joined:
    Mar 23, 2017
    Messages:
    12
    Location:
    NYC
    I can't make anything of this study. While I will agree that AP2 is disappointing, I am unclear on how anyone can include carmakers in this ranking who don't have actual consumer products on the market.
     
    • Agree Agree x 2
  13. JordanWilliamson

    Joined:
    Nov 6, 2017
    Messages:
    85
    Location:
    Ontario, Canada
    This doesn't answer my question at all... If I go to a dealership today, can I test drive a car that has better autonomous driving than Tesla? Waymo isn't in any production cars, neither is Uber. Of the companies in market are any of them as good as Tesla? Or are they all just researching and developing technology that will/could be better than Tesla if it ever reaches market?

    As I said from driving in a few from this list Tesla was not even comparable. If this list is simply Tesla is last because they don't plan on LIDAR while everyone else "plans" whether they have it in production or not, it's silly. In the production market, Tesla is the best. If all they companies have better outlooks because they are planning on LIDAR, Tesla could just announce 3rd Gen hardware with LIDAR....
     
    • Agree Agree x 1
  14. Konablue

    Konablue New Member

    Joined:
    Jan 22, 2018
    Messages:
    1
    Location:
    Everett, Wa
    When your system is nothing but an oversized roof rack full of cameras and sensors, misc antennas and various electronic gizmos, you have nothing but a science experiment. Let's rate vehicles that you can buy today. The system has to almost entirely 'invisible' to the consumer. I don't plan on ever buying a car that has a safari roof rack full of crap that looks as if it's ready to go do some google street mapping.
     
    • Agree Agree x 2
  15. simonalvarez0987

    simonalvarez0987 Active Member

    Joined:
    Dec 21, 2017
    Messages:
    513
    Yes, the bias is palpable. As stated in this thread, Apple hasn't even publicly revealed any plans on Project Titan. The conclusions of the study almost completely discounts Tesla's progress (albeit slow) with AP2.
     
    • Like Like x 1
  16. Roy_H

    Roy_H Member

    Joined:
    Jan 12, 2018
    Messages:
    84
    Location:
    Ontario, Canada
    I am a long time Tesla fan, but…

    Every time I read about a Tesla smashing into a semi trailer, fire truck or telephone pole, it proves to me that the automatic detection and braking doesn’t work (this is standard, not auto pilot but as I understand it the technology is the same as used on AP). I blame Tesla’s over-reliance on radar and lack of binocular vision with cameras. If they used twin cameras forward and matched images for edge detection they could triangulate and determine distance. Very difficult to determine distance with one eye. Some have pointed out to me that the M3 has 3 forward facing cameras, but according to Tesla website they are wide-angle, normal, and narrow (long distance) and not used as binocular vision. Many criticize Tesla for not using the very expensive LIDAR system to range detect objects, yes this is a viable alternative, I just think binocular vision is a much cheaper and in some ways better option. Camera resolution is much better than LIDAR and so should lead to more accurate determination of objects, i.e. I doubt LIDAR could distinguish between a shopping bag blowing across the road and a cat running across, but AI and binocular vision could.

    Reply
     
  17. Michael Russo

    Michael Russo Moderator

    Joined:
    Dec 16, 2016
    Messages:
    1,734
    Location:
    Pau, France
    I am a long time Tesla fan, therefore

    Every time I read about a T≡SLA smashing into a semi trailer, fire truck or telephone pole, I am reminded of the fact that, currently AP2.5 (or or any previous AP from the Palo Alto carmaker) does not relieve the driver of the need to staying attentive!

    T≡SLA should be commended for putting thousands of cars on the road and accumulating massive knowledge & data which will ultimately contribute to progress. Yet anyone thinking that they are anywhere near full self driving (a.k.a. Levels 4 or 5) is misinformed. We all should remember the above warning though. Keep your eyes on the road and, mostly, hands on or very near the wheel!

    I agree with those who said that the comparison should only be made between production cars that one can actually purchase and drive today.
     
    • Agree Agree x 1
  18. Roy_H

    Roy_H Member

    Joined:
    Jan 12, 2018
    Messages:
    84
    Location:
    Ontario, Canada
    The author chose to compare up coming technologies and state of research and development. This is exactly the kind of information I like to read about. There are lots of articles on automotive reviews that compare existing safety systems on cars available today. Go read them, do not complain that this article is not one of those.
     

Share This Page