Wiki Tracking FSD Feature Complete

[Wiki]

Updating this after a long time as V12 is gaining attention.

Previously ....

Background : In the Ark podcast Musk said they are planning to be feature complete by the end of the this year (2019).

Feature complete is a specific software term that means set of features that are planned for a particular project / release have been coded & tested to be working to meet the acceptance criteria defined for those features. It doesn't mean the quality bar for general release has been met. Usually more and wider testing is needed before the release quality bars are met.

In the AI podcast with Lex Fridman, musk agreed with Lex when he said FSD is basically
- NOA on freeways +
- NOA on city streets

Since we don't know what features Tesla is planning to complete by end of the year, I'm listing here what I think is needed for FSD feature complete. The list here involves 3 parts
- what I think a typical drive in the US might involve
- features NHTA has recommended
- additional features Waymo tests

I'm only looking for feature parity with Waymo, but in all US cities - because IMO it will have a big impact on SP.

Why this is important : I think feature parity with Waymo (but not geofenced) should be the near term target, rather than robotaxis. It is something achievable - and probably within a year or two. City NOA is also a great way to increase ASP / Margin. I think if Tesla can release to the entire fleet of 200k+ cars in the US features that Waymo can only demonstrate in a handful of urban areas, it will have a great impact on SP. Afterall Waymo is privately valued at nearly 5x TSLA.

As of 7/12/19 this is what Tesla website lists as part of FSD.

Autopilot Included

• Enables your car to steer, accelerate and brake automatically for other vehicles and pedestrians within its lane. (2.1, 2.3)

Full Self-Driving Capability

• Navigate on Autopilot: automatic driving from highway on-ramp to off-ramp including interchanges and overtaking slower cars. (9)
• Auto Lane Change: automatic lane changes while driving on the highway. (3)
• Autopark: both parallel and perpendicular spaces. (10.1, 10.2)
• Summon: your parked car will come find you anywhere in a parking lot. Really. (1)

Coming later this year:

• Recognize and respond to traffic lights and stop signs. (5, 4.1)
• Automatic driving on city streets. (2, 3, 6, 7) <-- This is the ambiguous one, which could mean just a few basic features or full set of features mentioned here.

Also, here is the explanation for FSD in the website.

Full Self-Driving Capability : Your Tesla will figure out the optimal route, navigate urban streets (even without lane markings), manage complex intersections with traffic lights, stop signs and roundabouts, and handle densely packed freeways with cars moving at high speed. When you arrive at your destination, simply step out at the entrance and your car will enter park seek mode, automatically search for a spot and park itself. A tap on your phone summons it back to you. (emphasis added)

The list is updated last on : 4/20/24

FSDb 11/20 : The last time I updated this in November '20. I've kept this for reference to see how many new features have been added in 3+ years !

Mod: this post is now a wiki at @EVNow's request. Anyone can edit it, but be good or else. --ggr.

 

[BrianZ]

I don't think parking is done. Autopark is nearly useless for me. It almost never activates at typical parking spaces. But that should be relatively simple to do for Tesla if they put in the work.

 

[anthonyj]

I don't think parking is done. Autopark is nearly useless for me. It almost never activates at typical parking spaces. But that should be relatively simple to do for Tesla if they put in the work.

Don’t worry. Feature complete FSD is coming this year (few months) and driverless robotaxis next year. Your car might be able to auto park in 2022

 

[RDoc]

I'd think that FSD would also avoid hitting cars, pedestrians, cycles, dogs, etc.

 

[EVNow]

I'd think that FSD would also avoid hitting cars, pedestrians, cycles, dogs, etc.

Yes - obstruction avoidance in general. I'd think of that as a basic platform / safety feature that applies to all of the above features.

 

[mblakele]

Let's keep in mind that "feature complete" is a term of art in software development, and may not mean what you think it means. Tesla may have their own definition, but this is a decent starting point.

Feature complete - Wikipedia

A feature complete version of a piece of software has all of its planned or primary features implemented but is not yet final due to bugs, performance or stability issues.[1] This occurs at the end of alpha testing of development.

Usually a feature complete software still has to undergo beta testing and bug fixing, as well as performance or stability enhancing before it can go to release candidate, and finally gold status.​

 

[EVNow]

I don't think parking is done. Autopark is nearly useless for me. It almost never activates at typical parking spaces. But that should be relatively simple to do for Tesla if they put in the work.

As a feature its definitely complete. That doesn't mean it works for everyone - this is where the March of 9s comes in.

 

[EVNow]

For those who haven't been following @verygreen on you tube, checkout some of the uploads. In debug mode AP/FSD gives a lot of data that con be informative on what they are tracking and where they are in terms of features.

For eg. this shows they are detecting stop lines.

This capture for example shows the data it shows for individual objects - and thus what they'd consider important to track & debug. You can also see the detected stop line.

And roadside objects here.

 

[EVNow]

Stopping at Red traffic light here.

 

[GeorgeSymonds]

Another way of looking at this is to compare it to AP1


- Pull out of parking - AP1 has this
- Drive recognizing the speed limit - AP1 has a better version as it doesn't rely only on maps, it can read road signage

- Change lane - done
- Obey street signs - done if we're talking speed limits
- Obey traffic signals - as above​

- Turn based on navigation details - no and won't
- Park (done) - done

I've increasingly thought that Tesla might be better focusing on driving 100% of the time on 60% of the roads rather than driving 60% of the time on 100% of the roads. By that I mean I would much prefer the car to be reliably left to drive on good quality freeways with low complexity and push for full autonomy on those conditions and take on the regulators to get permission to have the car responsible so I can work in the car and not need to concentrate, than have the car trying to drive in city areas and turn in junctions and stop at stop signs etc. The benefit for me on a 3 hour drive to let the car do ALL the driving for 2 hours of that journey would be amazing even if I had to drive the remaining hour without any aid and much better than having to pay attention for all 3 hours.

 

[EVNow]

Another way of looking at this is to compare it to AP1


- Pull out of parking - AP1 has this

- Drive recognizing the speed limit - AP1 has a better version as it doesn't rely only on maps, it can read road signage
​

AP1 can read road signage - first time I've heard of this. Even Mobileye doesn't claim to do it.

The Evolution of EyeQ - Mobileye

- Obey street signs - done if we're talking speed limits
- Obey traffic signals - as above

No - we are talking about stop signs. And traffic signals (red/yellow/green etc) This is only available on dev (see above youtube video).

​

- Turn based on navigation details - no and won't

What do you mean - won't ? Musk has already talked about this being available in dev and we also saw the demo. Afterall Cruise & Waymo already do this.

I've increasingly thought that Tesla might be better focusing on driving 100% of the time on 60% of the roads rather than driving 60% of the time on 100% of the roads. By that I mean I would much prefer the car to be reliably left to drive on good quality freeways with low complexity and push for full autonomy on those conditions and take on the regulators to get permission to have the car responsible so I can work in the car and not need to concentrate, than have the car trying to drive in city areas and turn in junctions and stop at stop signs etc. The benefit for me on a 3 hour drive to let the car do ALL the driving for 2 hours of that journey would be amazing even if I had to drive the remaining hour without any aid and much better than having to pay attention for all 3 hours.

From investment angle, I definitely want Tesla to concentrate and get the NOA on city streets. Then, they can concentrate on getting everything done better - being able to recognize objects better, for eg., will help in both city & freeways. ​

 

[ggr]

Mod: the first post in this thread is now a wiki at @EVNow's request. Anyone can edit it, but don't do anything silly. --ggr.

 

[EVNow]

Thanks a lot.

 

[jyalpert]

From investment angle, I definitely want Tesla to concentrate and get the NOA on city streets. Then, they can concentrate on getting everything done better - being able to recognize objects better, for eg., will help in both city & freeways.

I'm not sure I agree with this "fast and loose" strategy. Doesn't this leave them open to greater liability? Their new feature QA leaves something to be desired already . . .

 

[EVNow]

I'm not sure I agree with this "fast and loose" strategy. Doesn't this leave them open to greater liability? Their new feature QA leaves something to be desired already . . .

Tesla is currently saying the driver needs to be vigilant. Until they get to much higher level of accuracy, they won't tell drivers to sit back and relax.

From a liability POV, all this is still driver assistance feature.

But if we wait for feature A to get perfect before going to feature B, it will take forever to get to NOA on city streets. From investment angle, Tesla needs to hurry up to catch Waymo and Cruise who have city NOA working now in some geofenced fashion.

Tesla can leapfrog them by releasing NOA in any city (US) using just vision. This will prove their approach is working and we may see a big jump in SP.

 

[curry684]

Drive around a round about, waiting for any traffic coming from the left

This varies per country and even roundabout. The Netherlands has only been changed relatively recently from defaulting to "oncoming traffic has right of way" to rebuilding all roundabouts to "traffic on roundabout itself has right of way". It's been a mixed bag most of the time.

 

[GeorgeSymonds]

AP1 can read road signage - first time I've heard of this. Even Mobileye doesn't claim to do it.
​

We're talking about reading street signage of speed limits - did you really not know?

Does your AP1 read speed signs?​

 

[curry684]

My 2012 BMW did that even so I'm more surprised my 2019 Tesla doesn't.

 

[jyalpert]

Tesla is currently saying the driver needs to be vigilant. Until they get to much higher level of accuracy, they won't tell drivers to sit back and relax.

From a liability POV, all this is still driver assistance feature.

But if we wait for feature A to get perfect before going to feature B, it will take forever to get to NOA on city streets. From investment angle, Tesla needs to hurry up to catch Waymo and Cruise who have city NOA working now in some geofenced fashion.

Tesla can leapfrog them by releasing NOA in any city (US) using just vision. This will prove their approach is working and we may see a big jump in SP.

The investment you're referring to is "Tesla will be an autonomous taxi company soon," right?

I'm not sure what the path is from NOA working in a city to that, though.

This is where I'm just sort of confused as to what Tesla's autonomy strategy is. Cruise/Waymo is obvious - they are trying to create a solution that will drive 100% autonomously in limited conditions. Either the thing can navigate a city with no intervention, under some conditions (maybe: no fog, no snow, avoid certain types of roads, etc.), or it can't. Really great for focus, economics, "land and expand" network development.

And it always seemed to me that Tesla, as a consumer products manufacturer, seems to have basically made a laundry list of capabilities, then drawn a line somewhere in that list, and said "do everything above this line and the whole could be a self-driving car." Kind of like treating self driving as an emergent property of many point solutions acting together. Makes a lot of sense because they can pick off one feature at a time, launch it to the public, and call each one a driver's aid. Makes their drivers happy, and makes their cars more attractive for purchase. And this strategy was very clearly the development process laid out by Andrej at autonomy day. He explained a development process where they choose a capability (e.g. cut-ins), then painstakingly gather information, manually label it, and eventually determine how to deploy it. Each new feature takes a lot of time and effort to develop, one at a time. Mostly for open-highway driving, which is where car owners spend a lot of their time.

So NoA in cities seems like a small collection of point solution drivers' aids, but on city streets. What's the path from that to Level 4/5 robotaxi?

Which is why this pivot toward "this car is actually an autonomous taxi" feels kind of weird. One would assume that network effects for a taxi service are huge (as Uber found in China), so being first to market is extremely important. But Tesla doesn't seem to have been working in that direction, until this latest pivot.

In general I'm most confused about is what capability is actually required to achieve true autonomous driving, and where each company is positioned. Can you treat it as an emergent property of point solutions? If so, what list of capabilities do you need to achieve? Or do you need to take a more holistic view? I don't know. Anyone have a view?

 

[EVNow]

In general I'm most confused about is what capability is actually required to achieve true autonomous driving, and where each company is positioned. Can you treat it as an emergent property of point solutions? If so, what list of capabilities do you need to achieve? Or do you need to take a more holistic view? I don't know. Anyone have a view?

That is what this thread is about.

First start thinking about your daily drive and see all the major things the car needs to do to accomplish that drive. That's the feature list you need.

Ofcourse, everyone's feature list is different. But some are very common (like the ones I've listed) and some more uncommon ("edge cases"). What Tesla wants to do is to knock out scenarios few at a time. This is what he means by "March of the 9s". As they get deeper into edge cases and hit 6 9s (in terms of crashes per million miles), its better than average human.

The investment you're referring to is "Tesla will be an autonomous taxi company soon," right?

In particular I'm referring to the market taking Tesla seriously in the AV field. Currently they assign zero value to Tesla's FSD efforts. If Tesla can prove with NOA in the city that they are close to Waymo in capability, the market cap can double. Currently Waymo is privately valued at 4 times that of Tesla.