Forums

When is Full Self-Driving Capability to be delivered?

When is Full Self-Driving Capability to be delivered?

I ordered Full Self-Driving Capability earlier this year for my Model 3 with some expectation it would be delivered by December 2019. Here we are in December and still no mention of setting up an appointment to install it or even a software upgrade to use the current computer and new logic.

Who knows the schedule or latest promise?

rob | December 8, 2019

FSD complete will probably be defined by accountants not Engineers. So I expect end of year.

jerrykham | December 9, 2019

It depends on your, and Tesla's, definition of FSD. What most people would call FSD - full autonomy where you get in the car and tell it where you want to go and it takes you there without any intervention at all - is likely years away. This isn't just a technical issue, it is a legal issue, an insurance issue, etc. Local jurisdictions have to allow it, insurers need to have a way to handle fault, legislatures have to determine how fault works, etc. Even the technical stuff is going to take some time to fully work out.

That all said, some of the features that make up Tesla's revamped definition of FSD will be ready sooner. For example they have talked about upgrades to the current AP that will begin to stop for stop signs. That is supposed to be coming fairly soon. Other features are also supposedly decoupled from the full on FSD as well and will ship as they are ready.

Xerogas | December 9, 2019

@jerrycallahan66: nobody ever said anything about it being delivered. You might be referring to Elon’s statement about the software being feature complete by end of year, but that is a software development term that does not imply delivery to customers. It will require vast amounts of in-house testing, QA, and of course regulatory approval before we customers get our hands on it.

janendan | December 16, 2019

Has anyone who ordered FSD prior to 2019 received the hardware upgrade( Chip )? Ordered M3 June 2018, picked up August 2018.

TeslaTap.com | December 17, 2019

Also, Tesla has been very clear not to promise any dates for FSD. There are plans, desires, and hopes. No promises.

My expectation is FSD in some limited form (i.e. drive home to work and back with local and freeway) is about a year away. This will have limitations - such as geofenced areas where it doesn't work and it will have nags like today since regulatory approval is unlikely for a while longer. Regulatory agencies haven't even specified what is required or testing to allow FSD.

andy.connor.e | December 17, 2019

I dont think there is a release date for FSD. I think its just going to get better and better over time, with more and more features becoming available. Even when it reaches the point where its considered true fully autonomous self driving, i dont think they are going to claim it to be released, because it will be an eternal ongoing improvement process even at that point.

Dramsey | December 17, 2019

I'm still waiting for:

"Coming later this year: <--- Presumably 2019

Recognize and respond to traffic lights and stop signs.
Automatic driving on city streets."

But I've been waiting since November, 2016, when I paid for FSD.

rm760 | December 18, 2019

My MY 2018 MS is in today getting the HW3 FSD computer. I had originally made an appointment for the ranger to come to my home and complete the service, then asked to have the vehicle brought to the service center so as to have senior staff available should an issue arise. I received a 100 Uber credit to sufice transport to and from the service center. Make your appointment for FSD HW3 update.

hcwhy | December 18, 2019

I hope it gets here before my drivers license leaves.

billtphotoman | December 20, 2019

I think we will know true Level 4/5 autonomy is getting close (<2 years) when 3 things happen:
1) Tesla upgrades the side cameras so see more than 60 meters to detect cross traffic and to deal with rain and sun. Right now they can only "see" a vehicle moving 40 MPH (about 20 meters / second) about 3 seconds away which doesn't make turning right at a red light or an unprotected turn on a cross street practical. And if the AP gets it wrong and enters an intersection the driver basically has zero time to react.
2) I believe in efficient markets and once Tesla gets within 2-3 years of "robotaxi" level autonomy values of used Teslas will spike. Once the possibility of earning $100K (or even $10K) per year from a $50K investment becomes a reality investors will be all over this. If it was even 2 years away they would be buying up Tesla cars with FSD and storing them in warehouses.
3) The FUDsters will go into denial mode.
As for when the above will happen, I don't think anyone knows. I am really hoping when I buy my next Tesla in 2028 or so that Level 4/5 will have been out there and proven for a while. My wife and I don't have kids so at some point we will be counting on our car to drive us around.

andy.connor.e | December 20, 2019

I cant wait for the concept of your car passively making money. Not sure how many people would want to hitch a ride in the Cybertruck, but if it could pay for its loan that would be PRETTY IDEAL IF YOU ASK ME

TeslaTap.com | December 20, 2019

@billtphotoman - No camera upgrade needed. Currently, only a small fraction of the camera is used by HW2 AP processor - something like 360 x 240, which is the limit of the processor. HW3 jumps the ability to process the camera's full-frame, or 1280 x 720, about 12 times the resolution. Not sure why some are hung up on 2016 specifications that are long out of date.

billtphotoman | December 20, 2019

@TeslaTap.com - I would like to see your source showing the side camera will be able to "see" past 60 meters with just a processor upgrade. Looking at the lens it seems angled downwards so at some point objects will be outside of its field of view and or out of focus. No amount of processing power can mitigate the missing data in this case. Plus, clearly Tesla themselves believe 3 cameras are required to see sufficiently far straight ahead. It is pretty simple to show with arithmetic that a car needs to be able to "see" nearly as far to the side as straight ahead to detect side traffic approaching, measure its speed and calculate whether it is safe or not to make at unprotected turn into cross traffic. As I have said before, I am sure Tesla is aware of all this and when L4 gets close will roll out an improved side camera system and some sort of wiper mechanism to keep the side camera clear. And before someone says "a driver can see out the side window in rain" the difference is the driver's eyes are a foot or more back from the window. Try putting you eye right up to the side window in rain to get some idea of what the camera would see.

jordanrichard | December 20, 2019

People need to not lose sight of the fact that it's your local legislators that will dictate when you get it.

hcwhy | December 20, 2019

....and the insurance companies.

PrescottRichard | December 20, 2019

That’s a good point about local laws. You know that laws will not be pro-active but reactive to the technical advances. That is - the tech will be done (or very near done) and THEN all the law makers will have to deal with it.

Lots of time before this becomes an everyday reality.

Also- interesting points about the side cameras and cross traffic. There is a trend to put in ‘roundabouts in the past few years here in AZ and that had me wondering about how a FSD car would be able to handle them. Seems like cameras looking for cross traffic would be best placed as far forward and rearward as possible (for backing).

billtphotoman | December 20, 2019

@jordanrichard "People need to not lose sight of the fact that it's your local legislators that will dictate when you get it." - This reminds me of when I worked in Telecom in the 80s (dating myself I know) and we had caller ID developed. Is was many years after that before the legislators allowed it. As long as L4/L5 is out there before I reach my late 60s (10 years) I am good. In late 60s one's driving accident rate really starts to climb so I really want technology to rescue me.

jordanrichard | December 21, 2019

Legislators make decision on what to support what to kill based on the constituents fears/emotions. We continue to hear reports of cars getting into accidents with AP engaged. With that constantly in the public’s mind and AP working correctly still freaks people out, what legislators is going want their name associated with a “robo-car” bill? Not one legislator is going want to be associated with allowing a bill that lets a type of car on the road that people fear might hit a kid. You can recite all the safety statics you want, but numbers don’t resonate with people’s emotions.

rxlawdude | December 21, 2019

The biggest question in my mind are the edge cases where an ethical decision must be made in a split second.

Think of the situation with a vehicle suddenly in the way of the self-driving car. Collision is unavoidable without swerving. The sidewalk with people is the only swerve route.

Who should the car prioritize to minimize harm to? The occupants of the SDC, the occupant(s) of the vehicle about to be struck, the people on the sidewalk? Does it matter if those people on the sidewalk are moms with babies in strollers or a gathering of octogenarians?

Can any algorithm properly assess all of these factors (or should it?)?

rxlawdude | December 21, 2019

Level 3 is attainable because the human is an override. Levels 4 and certainly 5 are another matter.

SamO | December 21, 2019

California already permits level 4 autonomy if tesla were simply to apply.

jordanrichard | December 21, 2019

What technically is Level 4?

Also, are John and Jayne Q Public allowed to have said Level 4 cars on the road. Just because a state gives permission to Google, Apple, Delphi, Tesla to have Level 4 cars on the roads as part of a testing program, doesn’t mean you and I can buy such a car.

TeslaTap.com | December 21, 2019

billtphotoman "@TeslaTap.com - I would like to see your source showing the side camera will be able to "see" past 60 meters with just a processor upgrade. Looking at the lens it seems angled downwards so at some point objects will be outside of its field of view and or out of focus."

I actually think it was quite silly for Tesla to state a range in feet/meters for each camera. It was clearly made by marketing, not engineering. For example, one reason there are 3 cameras in the front it to read signs and detect animals, objects, and pedestrians from a fair distance - items far smaller in detail than a vehicle in front or to the sides. None of these are issues to the sides.

The side cameras are pointed outwards and can see distances without any issue. In fact, some people want a 360-degree parking view, but the cameras do not capture the ground like some other automakers' systems. Of course, those other systems can never be used for Autopilot style features either. The cameras have to be pointed and focused where the desired action is.

I don't have any inside knowledge but have been reading/watching just about every video and talk Tesla has made publicly on Autopilot. Most non-technical people don't realize how little resolution is needed by the neural nets to make proper decisions. The cameras offer far more resolution that is used by the car today in HW2. Tesla has stated publically that HW3 can process the full-frame, which isn't possible today with HW2. This is for every camera - front and sides. It's sort of incredible how well it works today with HW2. Using the full resolution, I'm confident it will have no problems seeing moving vehicles on the sides at far greater distances than HW2.

There also was some video captured from the side cameras (not the rear-facing cameras used for the dashcam). The links to one of the videos I've seen are now dead, so I can't point you to a source right now.

Nothing wrong if you feel you have to believe in the marketing numbers from 3 years ago and never expect FSD to work. You have a few others with the same belief, and I'd recommend you not buy FSD until it is demonstrated to work. I think I have a clear view of the technology and don't see the same issues you do (and bought FSD). Now it could easily be another year before we actually get some form of FSD as the regulatory and legal issues still need to be sorted out. They are almost as complex as FSD is!

billtphotoman | December 22, 2019

@TeslaTap.com "never expect FSD to work" - I am _not_ in that camp and fully expect my _next_ Tesla will be driving me around (offer L4 at least and probably L5) in 2028 or so. The $3K I would have spent on FSD is in my investment account earning income towards my next Tesla and I am enjoying EAP while I wait. I am not implying I think L4 is 9 years out but I do think at some point before then it will be well baked (IE technology is ready, corner cases worked out, legal/regulatory issues worked out). As you said, we will see and I would love to be shocked by it rolling out in the next 2 years.

andy.connor.e | December 22, 2019

@rxlawdude

I am sure there is algorithm that can be developed that can weigh the injury severity based on crash test reports and probability of injury vs hitting a pedestrian straight on. The best maneuver would probably to have one of the two cars strike a wall or something that would not harm another vehicle or person. But a computer would be much more capable of a calculation with a second of time than a human with probably somewhere around 200ms of response delay, and more likely to swerve out of the way of the moving object and not have time to consider the pedestrians.

TeslaTap.com | December 22, 2019

@rxlawdude - "Who should the car prioritize to minimize harm to?"

It's an interesting idea, but I'm not sure it entirely matters. Humans rarely do any prioritizing. It's easy to set up some scenarios where you have unlimited time and thought and quite another when you are in the middle of an accident and have milliseconds to make a choice. Often people make what most would consider the wrong choice, using hindsight. FSD will never be perfect, just like humans, but it should be a lot safer in general. There will always be some scenarios where it causes harm - particularly if there are no options to avoid a crash.