Will Tesla be able to deliver FSD with HW3.0 and current Model 3 sensor suite, ever?

[SergeyUser]

The more I see how slow Tesla is at delivering new FSD features, the more I think it will never happen. Assuming current HW3.0 and sensor suite.

I mean - Ok, stop signs. Great. In 5 years development timeline? How long it will be before Tesla releases unprotected left turns on intersections? Another 5 years?

Bottom line - my prediction is Tesla will never be able to release FSD, even in USA, on HW3.0 and current sensor suite.

Now I guess 99% of the forum would disagree with me. I’m curious - is there anyone else who agrees with my prediction?

 

[ElectricIAC]

I’m sure there’s code (if faulty still) to extend range seen for objects or vehicles coming your direction from the sides already as evidenced by the HARD braking action my P3D initiated when spotting a car on the cloverleaf ramp merging.

 

[Mo City]

I agree with you but want to be wrong.

The slowness to this point doesn't have anything to do with it. Now that software is being written specifically for HW3 and the stop sign and stoplight detection feature has been turned on, the end game has begun.

Within 12-18 months, it should be clear to most everyone legit FSD either (1) ain't gonna happen or (2) is quite likely.

One thing everyone should understand: Tesla does not need Level 5 autonomy to wildly succeed going forward. That said, if robotaxis become a reality, Tesla will incinerate competitors even worse than I already expect.

 

[jmaddr]

I personally think “maybe”. Elon has banked his sensor suite on mimicking human eyesight (with simple radar added). I believe that could be flawed as our society, including our legal liability laws, have a different level of acceptance when it comes to financial and life risk. Human eyes, along with reaction times, result in a certain percentage of accidents per miles driven. Current AA and FSD reduce that rate by combining camera sight plus radar with better/faster/consistent reaction times. However, it is my opinion that AI superior reaction capabilities will not be enough when financial and life risks are considered. Our legal liability laws will need to reflect this somehow. Sensors will be required that are better than what humans can sense as well to drive the accident rate to an acceptable legal level. Inter-vehicle communication, vehicle to infrastructure (5G), lidar, radar, and cameras will be necessary to provide the AI the decision data to fully drive a car In all conditions. The combination of better than human sensing AND reaction times will result in the lowest financial and life risk and provide full FSD.

Now I said maybe as the liability laws could change. Further, it’s based on my (and the generally acceptable) definition of level 5 autonomous driving. A change in liability laws and the generally acceptable definition of L5 would change that outcome. We all have to agree on what FSD is to answer your question.

I do think the current sensor suite and level of decision making is incredible and should work for most situations, even enough for robo-taxi alike. Is that FSD? I think it could be....

 

[vjason]

AI driving model aside (I’ve found more than a few intersections and random lights that FSD can’t interpret correctly, yet) they need cameras that can stare straight into the sun and still work.

Driving east/west after sunrise/leading up to sunset drivers the car mad due to the sun nailing the cameras.

 

[CertLive]

I believe the acceptable level 5 definition is if you take a Tesla, remove its steering wheel and you can sit in the back watching TV whilst the car takes you from A to B. This requires the same or better than human senses to achieve. If you go by the route Tesla has currently taken it is not as if the sensor suite has stood still. HW2.5 to 3 upgrades have been done so there is a chance if you buy FSD your car will be given what is needed if it does not have it now. Tesla is building a package for FSD and waiting for legislation etc to catch up. I actually want to see the city driving addition work this year because there are so many obstacles to overcome and unexpected challenges delivering on this promise would probably convince me the car can do it.

 

[hellocar]

I doubt that true full self driving will happen within the 6 years or so that I plan to keep my Model 3. It's possible that some of the features might become useful before then (currently none of them are for me, other than driver-initiated lane change with turn signal stalk). The only reason to buy the $7K FSD option now is if (like me) you enjoy seeing them chip away at this problem and playing with the new features as they are released.

Even though it seems unlikely, part of me is hoping to be proven wrong haha

 

[ElectricIAC]

As someone who took a chance on FSD, I’m hoping they can crack it.

The realist in me says: No way, don’t make that mistake again until they have a finished product. Tesla in and of itself though is the antithesis of that as the only constant is change.

 

[M3BlueGeorgia]

I doubt if the problem with delivery of complete FSD is the sensor suite.

I think the problem is the thousands of edge cases which have to be successfully handled.

Let's talk about speed limits as an example:
- We currently have speed limits from a massive database, which is mostly accurate for static speed limits
- Adding sign reading will allow the database to be updated and allow more accuracy, but still static
- Edge case where GPS doesn't determine correct speed limit because car isn't sure which road its on (ex: 45mph on slip road vs 65 mph on freeway)
- Edge case for construction speed limits, but a static sign
- Edge case for variable speed limits from signs (common in Europe, not so much in USA)
- Edge for reading time-variable speed limits
---- School zone: Restrict speed when sign has flashing light
---- School zone: Restrict speed "whenever school is in session", which requires accurately knowing all the school schedules
---- Construction zone: Restrict speed "when workers present"

I could go on, but even an apparently simple job of figuring the current speed limit for whatever piece of road the vehicle is on, rapidly gets you into some difficult edge cases.

 

[ElectricIAC]

I doubt if the problem with delivery of complete FSD is the sensor suite.

I think the problem is the thousands of edge cases which have to be successfully handled.

Let's talk about speed limits as an example:
- We currently have speed limits from a massive database, which is mostly accurate for static speed limits
- Adding sign reading will allow the database to be updated and allow more accuracy, but still static
- Edge case where GPS doesn't determine correct speed limit because car isn't sure which road its on (ex: 45mph on slip road vs 65 mph on freeway)
- Edge case for construction speed limits, but a static sign
- Edge case for variable speed limits from signs (common in Europe, not so much in USA)
- Edge for reading time-variable speed limits
---- School zone: Restrict speed when sign has flashing light
---- School zone: Restrict speed whenever school is in session, which requires accurately knowing all the school schedules
---- Construction zone: Restrict speed "when workers present"

I could go on, but even an apparently simple job of figuring the current speed limit for whatever piece of road the vehicle is on, rapidly gets you into some difficult edge cases.

This is desperately needed, as well as better inclement weather performance.

 

[RichAZ/CapeCod]

I own a Chevy Bolt (2019 Premier) at the moment but will be purchasing a MY in the Fall. I fully expect that one day FSD will be a reality for Tesla. It's just that this, the self-driving auto, is a beastly complex problem to solve. It will take a while. So, my MY will not have FSD (unless they permit me to move this pricey technology to another Tesla at some future date. Not likely it would appear.).

I wonder why Tesla doesn't break up self-driving into a robust true self-driving mode for limited access highways, and later come out with a more complex mode for all other roads, perhaps even cities?

Anyway, watching the progress of this company is never dull!

Rich

 

[SomeJoe7777]

I concur with the opinion that FSD is not going to happen any time soon, but for a different reason. The sensor suite has dozens or hundreds problems with edge cases, but that's not the real impediment.

1. The car relies on GPS to know where it is, and GPS is only accurate to around 30 feet / 10 meters. That's not accurate enough. The car does not know exactly where it is.

2. The car relies on the map to know where it's supposed to be, and I've personally seen hundreds of map errors where the roads on the map don't match the actual roads. Most of this is in construction areas, but construction areas are a reality -- they exist, and they're not infrequent. The maps don't get updated often enough, in some cases months or years. That's not good enough, you need the map accurately updated in a matter of hours. Therefore, the car doesn't know exactly where it is supposed to be.

If the car doesn't know either where it is or where it is supposed to be, it is impossible for it to do anywhere near full self driving, no matter how good the sensor suite or the algorithms are. You could have the best algorithm and sensor suite on the planet, but if the car thinks it's on the service road instead of the freeway's right-hand lane due to a map or GPS error, you're doomed. (And I've seen both cases, repeatedly).

I see no easy way to solve either issue. There are some hard ways to solve both issues, however, and Elon does like to do things the hard way, so who knows?

 

[Knightshade]

I concur with the opinion that FSD is not going to happen any time soon, but for a different reason. The sensor suite has dozens or hundreds problems with edge cases, but that's not the real impediment.

1. The car relies on GPS to know where it is, and GPS is only accurate to around 30 feet / 10 meters. That's not accurate enough. The car does not know exactly where it is.

Not sure where you get this idea?

GPS hasn't been that inaccurate in 20 years.


Discussion of Teslas GPS chip here-
What kind of GPS chip does Tesla use? : teslamotors


Actual specs on it here:
https://www.u-blox.com/sites/default/files/NEO-M8L_ProductSummary_(UBX-16000760).pdf


That's the GPS tesla uses. Accurate to 1.5 meters as it uses SBAS.


And Tesla has filed several patents in the last 2 years for even more accurate methods than that.

For example:
https://www.geospatialworld.net/news/tesla-develops-technology-for-more-perfect-gps-positioning/

I
2. The car relies on the map to know where it's supposed to be, and I've personally seen hundreds of map errors where the roads on the map don't match the actual roads.

Hence why the AP re-write currently underway to create a full 360 view/understanding of the surrounds is needed

 

[TexanEV]

Cities need to adapt to it as well. Make shitty intersections better, make traffic lights readable. Without help from cities, it won’t happen

 

[vickh]

another question is will Tesla upgrade current 3 FSD owners to current hardware if 4.0 etc are needed.

 

[Watts_Up]

I just wonder why trains, especially freight trains, are not yet autonomous? I think that would be the first step to validate FSD.

Trains are running on tracks, they cannot change lane by them self, and use section areas which are the ultimate geo-fence model.

 

[Uncle Paul]

I do not consider current technology to be precision FSD, but more like very capable cruise control.

Love to be able to take some of the stress off long highway drives and heavy stop and go traffic.

If I need to take over, from time to time, that certainly works for me.

Statistcs seem to show that current FSD use by knowlegeble drivers can significantly reduce the number of accidents per mile on the road.
To me that is a tremendous and kinda unexpected result from the current technology.

Perfection is the goal, but current reality is pretty good as well.

With lowered expectations comes greater satisfaction.

 

[Uncle Paul]

Trains are not automatous because it is expensive to implement, and there is significant push back by the train trade unions to not automate away jobs.

 

[dave999z]

Recycled from my post in another thread:

I fear Tesla is on the wrong track with FSD. 90% of the driving I (and many others) do is Home-->Work-->Home every day. I should be able to program everything about that route... where every stop sign is, where every stop light is, the rules for each intersection, location of the actual lights, my preferred lane, speed, etc. It will be great if FSD gets to the point of working in unfamiliar environment. But it would have SO MUCH more daily utility if I could program it to be a good self driver on my usual route.

 

[Uncle Paul]

Was interesting to note, when watching some old Star Wars movies, that all the fighters and ground transportation was still operated manually by pilots. They had technology to navigate over longer distances, but in challenging situations a pilot would always take over the controls. Going through the canyons, around trees, dodgeing boulders. Some of the best heros had the best flying/driving skills.

Even Hans Solo was at his best when he switched off the targeting computer and his hands be guided by "The Force". Also tended to take driving over from the Droids in difficult situations. Darth Vader was at his most competent and evil self when he got behind you, with his hands on the controls of his personal fighter.

The Mandalorian also kept very busy flipping switches and staying in control as he piloted his beat up interstellar junker.

Not unusual for a dramatic scene to show a pilot handing the controls over to Chewy while running to the back to personally handle one crisis or another

May the Force be with us...