Predictions on when we will see AP 3.5 or 4.0 hardware

[Reeler]

It has been almost a year since Tesla released their first version of silicon. Typically, development cycles are about a year in frequency. I have a Model 3 and paid for FSD, but only have the AP 2.5 hardware. I think waiting for a couple more design iterations on their hardware makes sense before having Tesla perform the recall on my hardware.

I think there have been updated cameras and expect a new turn on the processing chips this year. I think that would make it AP 3.5. I bet with the Model Y release, we will see more redundancy that would be required to get regulatory approval and will call that AP 4.0.

What do you think?

 

[diplomat33]

It has been almost a year since Tesla released their first version of silicon. Typically, development cycles are about a year in frequency. I have a Model 3 and paid for FSD, but only have the AP 2.5 hardware. I think waiting for a couple more design iterations on their hardware makes sense before having Tesla perform the recall on my hardware.

I think there have been updated cameras and expect a new turn on the processing chips this year. I think that would make it AP 3.5. I bet with the Model Y release, we will see more redundancy that would be required to get regulatory approval and will call that AP 4.0.

What do you think?

I am a firm believer that Tesla will need to add more sensors at some point, including lidar, in order to achieve safe autonomous driving where driver supervision can be removed. But Tesla does seem to change hardware incrementally and only on a strictly as-needed basis. So, I suspect that Tesla will wait to see how AP3 is working once "feature complete" is released to the public. Tesla will study what areas are still lacking in the hardware or software and make the bare minimum of changes to address the issues.

I could see the Model Y having "AP3.5" that just has better resolution cameras or something minor, not a fundamental change to the sensor layout. But I think 2-3 years from now, we will probably see "AP4" with the new AP4 chip and (I hope) additional sensors like lidar.

 

[mikes_fsd]

The day Tesla adds LiDAR hardware to their vehicles is the day I start questioning their advantage in software in general.

My prediction, we will never see LiDAR on any Tesla vehicles because it is completely unnecessary.
Now, I can see the camera sensor upgrade and the housing (to help with heavy rain and ice/snow) and probably newer versions of radar and ultrasonic sensors, but not LiDAR.

 

[Kmartyn]

The day Tesla adds LiDAR hardware to their vehicles is the day I start questioning their advantage in software in general.

My prediction, we will never see LiDAR on any Tesla vehicles because it is completely unnecessary.
Now, I can see the camera sensor upgrade and the housing (to help with heavy rain and ice/snow) and probably newer versions of radar and ultrasonic sensors, but not LiDAR.

I agree, no LIDAR, but better cameras and radar (microwave). Constant improvement from camera and radar vendors. Ultrasonics probably not, no significant improvements coming in that simple technology.

 

[destructure00]

I think waiting for a couple more design iterations on their hardware makes sense before having Tesla perform the recall on my hardware.

What makes you think this is an option?

 

[mikes_fsd]

Ultrasonics probably not, no significant improvements coming in that simple technology.

Only reason I added Ultrasonic sensors is because they did upgrade them right after I took delivery of my Model X in 9/2016, with the release of AP hardware 2.0. So, still a possibility in the future.

 

[Reeler]

What makes you think this is an option?

I am sure there is some flexibility for snow birds and others that cannot get in right away for the upgrade.

I personally think there will be new silicon around the year anniversary. Almost never is the first attempt very good. At least bug fixes in the semiconductor.

I see no good reason to upgrade right now as the new hardware does very little. I will delay until there is a reason to do the update.

 

[diplomat33]

My prediction, we will never see LiDAR on any Tesla vehicles because it is completely unnecessary.

LIDAR is not completely unnecessary. Every autonomous car (other than Tesla) uses it because it provides useful information.

 

[DanCar]

LIDAR is not completely unnecessary. Every autonomous car (other than Tesla) uses it because it provides useful information.

Is anyone other than Tesla trying to do autonomous cars for the masses? There is a lot of taxi service out there but the add on hardware is described as costing more than $100K.
https://ww.electrek.co/2020/01/10/i...-is-a-platform-for-its-self-driving-efforts/#

He pegged the current cost of an autonomous system (above and beyond the vehicle platform) at $100,000 or more, depending on what sensors are used. But within about 10 years, it will be about $10,000, with continual cost reductions in the next few years.

 

[diplomat33]

Is anyone other than Tesla trying to do autonomous cars for the masses? There is a lot of taxi service out there but the add on hardware is described as costing more than $100K.

Yes, Mobileye is trying to do autonomous cars for the masses.

The sensors don't cost $100k anymore. Luminar has lidar sensors that cost $500 a piece.
This Lidar Is So Cheap It Could Make Self-Driving a Reality

I am also not suggesting that Tesla pack the kind of huge sensor suite that Waymo does. But at $500 per lidar sensor, I think Tesla could afford to add a couple of lidar sensors. I'm really just suggesting a front lidar (for the stopped car and barrier problem) and rear blind spot radar (to improve detection of fast moving cars in the blind spots). I don't think that would be too costly.

 

[DanCar]

Mobileye is trying to do autonomous cars for the masses.

By 2025. They failed their last predictions of self driving.

 

[diplomat33]

By 2025. They failed their last predictions.

Yeah, and how is Elon doing on his predictions?

 

[Reeler]

I think they may not need Lidar, but more robust and redundant sensors is probably needed. Probably will require additional wiring at some point which will be costly to snake all that through the car.

At some point this 'house of cards' promising FSD is going to come crashing down. There are folks who paid for FSD with AP 2.0 about 2.5 years ago that will soon have their cars coming off lease with nothing delivered. They should at least get that portion of their lease refunded.

 

[DanCar]

... Typically, development cycles are about a year in frequency. ...

Tesla's cycles are two years optimistically. Elon said HW4 in about two years. He said that during Tesla Autonomy day , April 2019. So the soonest we will see HW4 is 2021 April. Other companies like nvidia have one year development cycles because they have multiple teams working in parallel. Tesla doesn't have that.

 

[S4WRXTTCS]

My prediction

4th Quarter of 2020 the NHTSA starts to put the brakes on what NoA can do with HW3. This is essentially the house of cards coming down. But, Tesla will put the blame on regulators for being too restricting.

1st Quarter of 2021 Tesla introduces HW4 with frontal solid state lidar, downfacing cameras, and rear/corner radars. But, they don't allow people to upgrade. Instead they simply give a discount on a new car purchase for FSD owners.

Keep in mind that the Lidar world is overcrowded so Tesla will likely buy an existing Lidar company that is on the verge of folding.

 

[Reeler]

HW4 with frontal solid state LIDAR would be great. Adding vehicle-to-vehicle (V2V) communication would be great too. With better sensors and hive processing, I think FSD would have a better chance of working.

 

[DanCar]

My prediction

4th Quarter of 2020 the NHTSA starts to put the brakes on what NoA can do with HW3. ...

Zero chance of this. NHTSA won't do this because:

1. Current administration has laissez-fair attitude. In other words, it is up to buyer to decide if they want Tesla or not and if they want FSD or not.
   
2. NHTSA operates on proof. They need months of research before reaching conclusion. By that time a new update is available and they need to start the cycle again.

 

[S4WRXTTCS]

Zero chance of this. NHTSA won't do this because:

1. Current administration has laissez-fair attitude. In other words, it is up to buyer to decide if they want Tesla or not and if they want FSD or not.
   
2. NHTSA operates on proof. They need months of research before reaching conclusion. By that time a new update is available and they need to start the cycle again.

It depends on how we define "putting the brakes on it"

When the first couple fatality accidents happened with AP1 the NHTSA put a lot of pressure on Tesla to add nags. This didn't take a long time because Tesla really acted proactively where they worked with the NHTSA on a resolution that satisfied their concerns. Back then it felt like to me that the NHTSA was really the one that put the brakes on it, but some people felt it was the intense media attention.

I see something similar happening as the result of things that have already happened along with Tesla attempting to do things like city NoA.

I don't believe Tesla can achieve L2 City NoA in a way that's safe while not having a proper way to measure driver attention.

The European union has already cracked down on AP, and it remains to be seen how Tesla will resolve it.

In any case maybe a better way of putting it is "regulatory push back will cause Tesla to introduce a new sensor suite in early to mid 2021". Because you're entirely right about the NHTSA. They tend to be very corporate friendly.

 

[Daniel in SD]

Seems more likely that states like California will decide that city NoA is autonomous vehicle testing and force Tesla to comply with autonomous vehicle testing rules. I’m skeptical that Tesla will even attempt to release city NoA though.
I’m not sure why Tesla would bother releasing a new sensor suite until they have the software to actually enable autonomous driving. The current sensor suite is more than enough for L2. They really should add better driver monitoring but I guess that would be an admission that robotaxis are not coming this year.

 

[Az_Rael]

There are folks who paid for FSD with AP 2.0 about 2.5 years ago that will soon have their cars coming off lease with nothing delivered

Those cars are already coming off leases. Oct 2016 were the first deliveries I believe. Dunno what the FSD take rate was though for a lease.

Instead they simply give a discount on a new car purchase for FSD owners.

I doubt they even do this. They will try to rewrite history and talk about how everyone DID get FSD because NoA and Smart Summon, etc! Elon has already referred to NoA as highway FSD at least once before.