Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Autonomous Car Progress

This site may earn commission on affiliate links.
Nvidia presentation on their FSD platform DRIVE:


It's only 9mn35s presentation. Definitely check it out!
Well they do have like 25 sensors, 12 cameras, 10 radar a lidar etc..gheez. No idea how Tesla is going to make just vision work with so few cameras...the obvious limitations they have now are so easy to see by anybody with Beta. Big mistake I think Tesla using only cameras, and so few.
 
Well they do have like 25 sensors, 12 cameras, 10 radar a lidar etc..gheez. No idea how Tesla is going to make just vision work with so few cameras...the obvious limitations they have now are so easy to see by anybody with Beta. Big mistake I think Tesla using only cameras, and so few.
Here's the camera suite data:
8 exterior camerasWide & tele visionSony IMX728, 8.3MPix
4 exterior camerasFisheye near visionSony IMX623, 3.0MPix
NVIDIA DRIVE Hyperion Developer Kit

Basically it's the same 8+4 layout that was discussed on the HW4 rumor threads. 8 being the same camera angles Tesla's existing 8 camera suite is using (Nvidia's demo shows 7 of those), main difference being instead of forward facing side cameras being in b-pillars, it's on the mirrors.

4 cameras being fisheye near field cameras likely used for parking.

If you look at the Nvidia fleet list at 0:39, XPeng is using Nvidia Drive also, so it makes complete sense now that it adopts exactly the same layout. So judging from this, we'll see a ton of different cars use the same layout.
Rumor: HW4 will have 13 cameras

There's talking the whole 13 camera (the extra camera being the cabin camera) rumor may just be someone mistaken about other EVs using this layout, not necessarily Tesla. Given Nvidia is going this route, that increases the chances of that.
 
Interesting. I find their simulation to be much lower graphics quality / less realism vs Tesla's.


Tesla:


LMAO now you are hyping up sim. I thought it was useless? lol there is absolutely no logic or objective reasoning when dealing with Tesla fanatics.

2019
Building a test facility is like building an FSD simulation: not representative of real life situations.
Today's presentation has cemented in my mind that Tesla is FAR ahead of the competition:
  1. Lidar is useless. Must solve vision and have it be the #1 controlling factor.
  2. Simulations are useless. You can't simulate what you don't know. Simulations are mostly for testing what you already have.
..like the simulation paradox. If you can simulate it (edge case identification), then you already know it.
 
  • Funny
Reactions: Doggydogworld
What we now need are people to take the exact same route in FSD beta and we can compare.

Better still, tag team with one taking Waymo and another following with FSD Beta (and the other way!). Let us see if I can get someone in S.F. interested in the idea.

I'm all for another Urban Challenge involving every ADS company except in the real world traffic. Most importantly no one in the driver's seat, maybe someone in the passenger seat with a single RED disengage button that would stop the car in its track.

Apparently someone listened to me ;)

 
  • Informative
Reactions: hgmichna
FSD Beta did not take the same route as the Waymo. So it is not exactly like what you wanted.
Just one drive doesn't tell us very much anyway. This test should be repeated 10 times.

We can already see though that Waymo's advantage of additional sensors and high-definition maps is questionable.

Obviously, requiring a human supervisor or not is a decisive difference for certain purposes. For a private car, as opposed to a robo-taxi, it still matters, but not quite so much, that a human driver is required for the occasional intervention.
 

I hope NHTSA will conclude that HD Maps should be required to avoid such accidents


This assumes the crash actually happened of course.

It wouldn't be the first time someone filed false allegations against Tesla with the NHTSA.

IIRC with the suspension reports for example like 37 out of 40 fake reports were from a single dude, who was not the owner of ANY Teslas.
 
Motional and Lyft plan to launch fully driverless robotaxi service in Las Vegas in 2023 using the Hyundai Ioniq 5 vehicle.

I missed this gem.

Robotaxi is a "reality now" because .... they plan to launch a service in one city in a year or two ?

I mean going by that definition "reality", Waymo already has had one for years now ....
 
  • Funny
Reactions: thewishmaster
We can already see though that Waymo's advantage of additional sensors and high-definition maps is questionable.

No, we can't sat that. In fact it is very clear that Waymo's additional sensors and HD maps do provide an advantage because they make the autonomous driving far more reliable and safer than Tesla's FSD Beta. We know this because Waymo does not make the mistakes that FSD Beta makes and is able to drive autonomously in Chandler without a driver. Also, Waymo has a safety disengagement rate of 1 per 30,000 miles whereas FSD Beta has a safety disengagement rate that is orders of magnitude less than that.

Having said that, it is true that driverless is not necessarily required. Tesla does not need to do driverless since they could be content with letting the driver be the back-up. But for a robotaxi service like what Waymo is doing, driverless is required because you need to cut the costs of the driver in order to make a profit. So Waymo definitely gets an advantage from the additional sensors and HD maps as it makes the autonomous driving reliable and safe enough for driverless.

I missed this gem.

Robotaxi is a "reality now" because .... they plan to launch a service in one city in a year or two ?

I mean going by that definition "reality", Waymo already has had one for years now ....

Well, I agree that calling it a reality a few years before it actually happens is a bit exaggerated. But I am just the messenger. I did not write the article. I just wanted people to know that Motional and Lyft are planning to deploy a robotaxi service in LV in a couple years since this is a thread on autonomous progress.
 
Last edited:
Well, I agree that calling it a reality a few years before it actually happens is a bit exaggerated. But I am just the messenger. I did not write the article. I just wanted people to know that Motional and Lyft are planning to deploy a robotaxi service in LV in a couple years since this is a thread on autonomous progress.
Not a comment on you - but a comment on the press release.
 
  • Like
Reactions: diplomat33