These Owners Believe Tesla Autopilot 3.0 Is Shipping Early W/Two Radars: Video


Are Tesla vehicles already shipping equipped with Autopilot 3.0?

YouTuber Erik Strait (DAErik) is wrapped up in all sorts of Tesla-related business. One of his tasks is tracking Tesla inventory. With the help of his friend David, he discovered the Tesla Autopilot 2.0 update ahead of its official launch. Well, it’s about time for the Silicon Valley automaker to announce another major update. According to Erik and David, they claim they have an early indication that Tesla Autopilot 3.0 (or really Autopilot 4.0) is on the way. Erik and David believe the feature may be shipping on some cars already.

The guys point out that when they discovered AP 2.0 a few years back, it was due to a new code (AP2) that appeared associated with inventory on Tesla’s site. When Autopilot 2.5 came along, Tesla coded it as AP3. Now, they’ve stumbled upon an AP4 code, which they believe may apply to the upcoming Autopilot 3.0 technology. The feature, they believe, will use additional hardware that will eventually work to provide Full Self-Driving features.

This timing would make sense since Tesla could assure that the new hardware and update is already present in vehicles shipped to Europe. Additionally, the company’s earnings call is this week, so a big announcement would be advantageous for sure.

Erik and David provide some diagrams and further insight in the video above. Check it out and then let us know your thoughts in the comment section below.

Video Description via DAErik on YouTube:

Tesla AP 3.0 Shipping Early? With 2 Radars?

Tesla Autopilot 3.0 might be shipping on cars right now! Why do we think that? Check out the proof.

A recent tweet also seems to point in the same direction, however, we have no way of substantiating the information at this point. We can only hope the good news comes out during the earnings call on January 30th.

Categories: Tesla, Videos

Tags: ,

Leave a Reply

36 Comments on "These Owners Believe Tesla Autopilot 3.0 Is Shipping Early W/Two Radars: Video"

newest oldest most voted

Go get em, Tesla.
Hopefully, this really will be level 4.
But, if so, I want to see it handle roundabouts, esp multiple lane ones.
While relatively slow, they are complex and are likely one of the harder manuevers for driving.

There are humans that can’t seem to figure out roundabouts, but really, I think it’ll be vastly easier for a computer to figure those out than a non-perpendicular intersection. Roundabouts are just tiny highways with a high density of on-ramps and off-ramps.

Non-perpendicular with stops? Not a big deal. Yields might cause issues, but stops should be easy.

It won’t be level 4, you need to wait for AP5

Yeah, Tesla needs to work on getting to Level 3. Well actually, in my opinion, Tesla should stop this fiddling around with tiny incremental improvements, and indeed work on achieving Level 4 autonomy. But clearly Tesla has other plans.

IMHO the current fiddling around with tiny improvements is like working on how to improve the sails on a sailing ship, when what’s needed is steam power and a propeller.

So I’m a lot more interested in what type of new radar is coming in Tesla cars. If it’s just a second low-res Doppler radar, then that won’t be anything but just another tiny incremental improvement. What Tesla needs to put in its cars is a phased array, so-called “high-res” radar system. So long as Tesla keeps relying on cameras as its primary sensors for a self-driving system, it’s never going to get to Level 4… and possibly not even to Level 3.

L3/L4 are dangerous. AP is barely L2. And FSD is probably meant as full L2.

Absurd. Nobody is going to accept L2 as L4. And you are under-rating the current ability of Tesla Autopilot+Autosteer.

From various articles I’ve read, the consensus seems pretty clear that Tesla Autopilot (along with some similar but less able systems from other auto makers) qualify as L2 autonomy.

And if L3/L4 are “dangerous”… well, so are human drivers.

We can continue to expect autonomous vehicles to increase in average safety, as the controlling software is improved and hardware upgrades (including wirelessly connected offboard supporting systems) are implemented.

The average human driver? Not so much.

I have given two-lane roundabouts a go. It did fail as an aggressive driver entered pointed right at me. The failure was a defensive stop so not too bad for the current version.

Yeah, that is the problem. Get a Tesla on the inside lanes, and have aggressive drivers on outer. I think APs will screw up.

Please stop endangering others with your recklessness.

Biggest question is if they will retrofit 2.0 and 2.5 vehicles immediately, or will they wait? And if they wait, then for how long.

Also, if this is true, then we might as well end the no additional HW myth. If they will eventually need to retrofit LIDAR, it’s not more work, than adding additional radar. At least solid state LIDAR.

Tesla has upgraded hardware before to make it suitable for higher levels of autonomy like higher power computers so if there ever was a no additional HW myth that has long since been busted. It’s going to be interesting to see if these new high power computers will enable Tesla can avoid the use of lidars which are very expensive and not easy to integrate in a visually acceptable way in the body of the car.

The rotating LIDAR isn’t, but solid state LIDAR, like on the A8 looks just like radar. And it can be put anywhere. There are even ideas to put them into the headlights, which might even be super easy to retrofit.

You are right, Tesla always said the computer might need retrofitting, but many people claimed the sensors would not need to be changed.

To me that is nonsensical, if you need better sensors, it doesn’t need to be harder to implement, than a better computer.

IMO for level 4 LIDAR is the way to go. And since it can be done solid state, we will see pricing drop incredibly with production numbers as well as improvements in resolution. Digital cameras and LEDs were once super expensive, both are closely liked to what you need for solid state LIDAR, and they are super cheap now.

Solid state lidars are surprisingly inexpensive. An array of such lidar scanners pointed in various directions is already being used in some test vehicles.

What is “visually acceptable” is going to change depending on practical needs. That always happens as technology advances. Tesla is setting trends, not following them! For example, just look at the “shark’s nose” on the Model 3.

There is a lot of argument over whether lidar is actually best, given the way it degrades in heavy rain or snow. It may be that phased-array, so-called “high-res” radar would be better.

But either would be a vast improvement over trying to rely on software interpretation of video camera images!

How do humans do it?

Billions of years of evolutionary training, vastly superior brain compared to a computer and years of image recognition training as a child.

Sure, you could replicate that with 2 very good cameras in each direction, but then you will have to wait for much smarter neural nets.

On the other hand a much dumber computer can easily park a car with radar, much better than a human. So sensors matter.

(1) Humans have the advantage of a highly sophisticated visual cortex, the result of billions of years of evolution. Computers have… a few microprocessors.

(2) Humans don’t “do it” all that well. We can’t see in the dark, nor thru fog, and heavy rain or snow can be almost as bad as heavy fog. Not that I expect sensors in autonomous cars to handle heavy rain or snow much better than we do, but such sensors should be able to see in the dark or thru even pea-soup fog just fine.

I feel pretty confident in saying that Level 4 autonomy won’t be approved by regulators until they use either lidar (possibly long-wave lidar, which handles rain a bit better) or phased-array (so-called “high-res) radar as the primary sensor system.

Tesla’s insistence on relying on video cameras as the primary sensors will continue to hold back Tesla’s advancement toward Level 4 autonomy, until Tesla eventually gives that up. That’s my prediction, and of course not yet fact. But will Tesla allow itself to fall behind as other auto makers advance by using lidar or phased-array radar? Not for long, anyway!

Only people that bought FSD will get the retrofit for free. Others will likely be able to pay for it. But it’s gonna cost a pretty sum.

If humans can do it without radar, surely Tesla AI can do it too??

Radar sees thru fog and rain. Humans don’t at the same quality or level. Radar also sees two cars ahead. Humans can not when the 2nd car is smaller than the 1st car (ie. you can’t see the 1st car thru the 2nd cars window or the 2nd car is a van with no see thru).

But human can see a stopped fire engine/truck on the road most of the time. /S

…and if Tesla cars controlled by Autopilot+AutoSteer had lidar or phased-array radar, then they could “see” fire trucks stopped on the highway much more reliably.

Binocular vision works too and has working examples.

Yes, working examples in animals (including humans) with brains containing visual processing centers developed thru billions of years of evolution. Wetware is much, much more advanced and much more specialized than the tiny general-purpose microprocessors used in computers.

Current advances in autonomous driving have abandoned any attempt to use binocular vision, or at least that’s what I’ve read; there may be exceptions. I’ve seen some older self-driving vehicles (I recall one DARPA desert challenge) using binocular vision, but apparently more advanced systems use other methods of calculating distance from video images.

You are not a software engineer.

“If humans can do it without radar, surely Tesla AI can do it too??”

If human driving was safe, then we wouldn’t need self-driving cars. Cameras can’t see in the dark any better than the human eye can. Self-driving cars need a better sensor system than the human eye.

Umm, there are lots of cameras that can see quite well in the dark … and definitely better than the human eye.

To the broader point of additional, non-human capable, sensors that provide enhanced capabilities I agree. But it does reinforce the point that Tesla’s system – with enough compute power – theoretically could be as good as a human. The point people tend to miss is that we need better than human.

Yeah, infrared cameras. Those have their own limitations; you might try reading up on their requirements for cooling systems.

And do we really want software developers to have to come up with two sets of software, one for optical object recognition using visible light images, and the other using infrared images? You’ve just doubled the challenge, if so… plus now you need two sets of cameras.

“The point people tend to miss is that we need better than human.”

Indeeed! 🙂

Humans do it very poorly. That’s what there are ~30,000 fatalities on US roads per year.

Sigh. Tesla is not even Autopilot level 1.0 at this point. To be more, it has to take speed-restrictions into account. since that feature (on my X anyway) does not work 1/2 of the time and is buggy the other 50%, well….

Tesla’s Autopilot+AutoSteer is at least Level 2 autonomy, and I’d say 2+. Quit posting FÜD.

It’s amazing how many hardcore Tesla bashers pretend to own Tesla cars.

Yes we need two radars in order to avoid crashing when we are using the ridiculous UI V9 for model S and X

Another clue to why they stopped selling FSD. They knew they were going to need additional sensors and hardware and didn’t want to have to retrofit thousands of cars for free.

Yup. I’ve said the same in comments to previous articles on that subject.

I don’t care what technology Tesla uses; I just want my car to stop at an intersection where cars are already stopped. Scenario that plays out just about every day for me: I am in the right lane of a four-lane dixided highway on Autopilot following a car. We come to an intersection where another car is already stopped in our lane. The car in front of me moves to the left lane. Suddenly my car ACCELERATES toward the stopped car because my radar never saw it move and ignores it. If I dont take over I would probably crash into the stopped car.