v9.0 speculation

[ucmndd]

From the description, it at least sounds like level 3 autonomy on freeway.

Automatically changing lanes and handling freeway interchanges certainly won't be L3 the way Tesla implements it (demanding full driver attention at all times with nags every 20 seconds if you don't). L3 implies the car is the responsible party, not the driver. Not a chance that's happening any time soon.

 

[MarcusMaximus]

Automatically changing lanes and handling freeway interchanges certainly won't be L3 the way Tesla implements it (demanding full driver attention at all times with nags every 20 seconds if you don't). L3 implies the car is the responsible party, not the driver. Not a chance that's happening any time soon.

As I said, they specified the handover to the driver would be more graceful, rather than ever requiring them to take immediate control. If it's as described, then that's level 3 autonomy(the spec gives no requirement about legal responsibility over the actions of the car). But, as I said, that all depends on that "if".

 

[conman]

But, as I said, that all depends on that "if".

And 'when'.

 

[TaoJones]

I do not see freeway to freeway happening in September (let's just say this Fall/Winter) except for perhaps the most vanilla use cases.

Which means that when such an exercise is attempted and then fails with the red hands of death or some other abrupt "return of control to the driver", a whole lot of rerouting is going to happen after missing whichever lane is needed at the time (due to encroaching traffic).

Again, vanilla cases fine - think freeway interchanges in Arizona - much uniformity, nice long well-planned ramps/grade changes. For example, I-10 West to I-8 (which would be westbound from there toward San Diego) Easy peasy. Gradual curve and off you go. I'd expect few or no problems with that.

Now let's have a look at SoCal. Just off the top of my head, how about the 405 South to, oh, I don't know, the 110 South. At rush hour. And you're in the HOV lane. Even if you're directed out of the HOV lane when legal, you've got to negotiate 4 lanes of traffic *and* the correct lane for the 110 South versus the 110 North. Good luck with that without a whole lot more than is going to be available in September. Maybe it'll work at 2 in the morning.

This is part of why I'd be just fine with a v9.0 that focused upon the broken UI, the eradication of phantom braking, and fixing the very broken speed limit database which affects AS and puts the driver at risk, just to name three. Call it the "finally fixed what's broke after 2+ years" release. Having yet another "feature" that works sometimes maybe is far less interesting or helpful than bringing AP2 up to parity with AP1.

This is also why I'll vote yes for the privatization. You don't see SpaceX not fixing things because of public pressure for the next big Sexy Thang. We need that mindset at Tesla. In other words, it's not either or - it's and. Quality matters. The engineers are more than capable of cleaning up rushed releases. Let them work to deliver a higher quality product. Everybody wins.

 

[MarkS22]

I'm going to throw my hat in and predict one of the first FSD features will be a form of "Traffic Assist."

It will only work on white-listed divided highways with no traffic lights. And only under 30MPH when the vehicle can see at least three other vehicles.

Under these conditions, it will allow hands (and possibly eyes-free) driving. For me, it makes sense because:

(1) This will eliminate the most frustrating, boring part of driving.
(2) It's a particularly safe time to remove your eyes. When you're on a divided road designed for 55+MPH, going roughly half that, there is a much less chance of damage or injury. The likelihood of pedestrians is minimized.
(3) Based on my current experience with Autopilot 1 and 2, it's what it does best. Under slow traffic conditions, it's already pretty flawless with a single camera.
(4) This sidesteps/delays concerns of FSD taking away the "fun of driving."
(5) It's a natural progression between where we're at and hands free on-ramp to off-ramp at all speeds.
(6) Most importantly to Tesla, it will sell FSD. Stop sign or traffic recognition is nice, but people won't buy the car (or FSD) just for that. If you take away the worst part of driving and turn that time into productivity or entertainment time, it's a game changer.

 

[S4WRXTTCS]

I'm going to throw my hat in and predict one of the first FSD features will be a form of "Traffic Assist."

It will only work on white-listed divided highways with no traffic lights. And only under 30MPH when the vehicle can see at least three other vehicles.

Under these conditions, it will allow hands (and possibly eyes-free) driving. For me, it makes sense because:

(1) This will eliminate the most frustrating, boring part of driving.
(2) It's a particularly safe time to remove your eyes. When you're on a divided road designed for 55+MPH, going roughly half that, there is a much less chance of damage or injury. The likelihood of pedestrians is minimized.
(3) Based on my current experience with Autopilot 1 and 2, it's what it does best. Under slow traffic conditions, it's already pretty flawless with a single camera.
(4) This sidesteps/delays concerns of FSD taking away the "fun of driving."
(5) It's a natural progression between where we're at and hands free on-ramp to off-ramp at all speeds.
(6) Most importantly to Tesla, it will sell FSD. Stop sign or traffic recognition is nice, but people won't buy the car (or FSD) just for that. If you take away the worst part of driving and turn that time into productivity or entertainment time, it's a game changer.

That's exactly what the Audi A8 L3 traffic assist does.

But, are you willing to throw your hat that it's going to be full L3? Or still L2?

 

[MarkS22]

That's exactly what the Audi A8 L3 traffic assist does.

But, are you willing to throw your hat that it's going to be full L3? Or still L2?

L2 definitely. L3 maybe.

Honestly, I’m 50/50 with it launching eyes-free but I’d suspect it will get there maybe 3-6 months later.

Also, I’d like to see V9 launch with a video player. If it offered a streaming video service like Netflix, that would be amazing and even safer than using your own devices: When you need to take back control, the screen can flash a warning. Plus yours peripheral will still be on the road. The player can automatically hold your spot and resume. (Bonus points for USB support to download content while on home WiFi.)

 

[J1mbo]

Also, I’d like to see V9 launch with a video player. If it offered a streaming video service like Netflix, that would be amazing and even safer than using your own devices: When you need to take back control, the screen can flash a warning. Plus yours peripheral will still be on the road. The player can automatically hold your spot and resume. (Bonus points for USB support to download content while on home WiFi.)

This is why they went landscape for Model 3.... nailed it!

 

[rnortman]

If you haven't seen it; v9.0 will not bring any differentiation of FSD from EAP; that won't happen until HW3:

Elon Musk on Twitter

 

[gt2690b]

Geofencing / homelink - homemode that leaves door handles out!

I second the speed increase volume thingy suggestion

 

[arcus]

The screenshots for the 14-day trial give a hint on what's coming: Tesla starts free over-the-air Autopilot trials to boost usage and sales of the system

 

[Bladerskb]

Not joking. He deleted the initial tweet saying 3 months and the tweeted again saying same timeline as the hardware. Link to the current one is here(obvs, can’t link to the deleted one):
Elon Musk on Twitter

LMFAO really? anyone have a screenshot?

 

[Bladerskb]

As I said, they specified the handover to the driver would be more graceful, rather than ever requiring them to take immediate control. If it's as described, then that's level 3 autonomy(the spec gives no requirement about legal responsibility over the actions of the car). But, as I said, that all depends on that "if".

That's handover when you get to your exit, which isn't safety related.

"Enhanced Autopilot is should still be considered a driver's assistance feature with the driver responsible for remaining in control of the car at all times."

 

[MarcusMaximus]

That's handover when you get to your exit, which isn't safety related.

"Enhanced Autopilot is should still be considered a driver's assistance feature with the driver responsible for remaining in control of the car at all times."

I’m referring to the description during the conference call, not on the website.

 

[Bladerskb]

What we know from what's been released is V9 will have the E in AP where it automatically changes lanes without the user telling it to.

I can't be the only one that finds it a bit at odds with being L2 in that it's making a decision that requires way more situational awareness than is typically the case for an L2 system.

The fact that it's making a decision for you, and executing it without input from you means that there should be some liability in it.

Or are they seriously expecting a human to intervene if it screws up? It's one thing to intervene with what's coming up in front of you, but it's entirely different with what's coming up behind you, and in your blindspot area.

It makes more sense if it's an L3 system, but I'm not sure there is even a regulatory path towards L3 in this country. There is so little that Audi decided against introducing their L3 (but, very limited L3) in this country for the foreseeable future. Instead they're going to introduce it in Germany.

I called it before. EAP will simply do easy lane changes. It won't do lane changes in dense congested traffic. where you are in lane 1 and need to get to lane 5 in 1,000 ft.

Again most people refuse to watch this video.
But this is 30 mins of Audi A7 from 2015.
This is similar what EAP/ AP 9 is gonna be (with nags and way less impressive.)
People who think Tesla is ahead because of AP 9 are naive. Other automakers have been doing it for years. They just don't do agile development/incremental updates.

Starts 3:45

Mind you this is using only one camera (mobileye eyeq3 and several radars.)

 

[diplomat33]

What we know from what's been released is V9 will have the E in AP where it automatically changes lanes without the user telling it to.

I can't be the only one that finds it a bit at odds with being L2 in that it's making a decision that requires way more situational awareness than is typically the case for an L2 system.

The fact that it's making a decision for you, and executing it without input from you means that there should be some liability in it.

Or are they seriously expecting a human to intervene if it screws up? It's one thing to intervene with what's coming up in front of you, but it's entirely different with what's coming up behind you, and in your blindspot area.

It makes more sense if it's an L3 system, but I'm not sure there is even a regulatory path towards L3 in this country. There is so little that Audi decided against introducing their L3 (but, very limited L3) in this country for the foreseeable future. Instead they're going to introduce it in Germany.

Actually, if you think about it, features like auto lane change without driver input probably don't have to be as sophisticated as you might think. For example, Musk mentioned that your car will be able to auto lane change into the left lane if it would allow you to move faster than the traffic in the right lane. It could just mean that the car will compare the speed of the car in front of you and the speed limit and check if there are any cars in the left lane or in your blind spots. If the speed difference is big enough and if the left lane is clear, the car makes the lane change. It might not be super smart in terms of analyzing everything in the environment, just making a simple calculus that if the car in front is slow and if a lane change is safe, then it makes the lane change. Likewise, the freeway transition or on-off ramp exit feature might be as simple as just looking at your nav and if you need to be in say the right most lane to take an exit, the car checks to see if a lane change is safe and then makes that lane change for you. Again, it might not be super analyzing the environment, just making a simple calculus that based on the upcoming nav instructions you need to be in the right most lane, so it checks to see if a lane change is safe and if yes, then it makes the lane change for you. So it might still be a L2 driver assist, where the driver needs to pay attention to the environment but where the car can perform some automated driving responsibilities like making a particular type of lane change in certain conditions.

 

[S4WRXTTCS]

I called it before. EAP will simply do easy lane changes. It won't do lane changes in dense congested traffic. where you are in lane 1 and need to get to lane 5 in 1,000 ft.

Again most people refuse to watch this video.
But this is 30 mins of Audi A7 from 2015.
This is similar what EAP/ AP 9 is gonna be (with nags and way less impressive.)
People who think Tesla is ahead because of AP 9 are naive. Other automakers have been doing it for years. They just don't do agile development/incremental updates.

Starts 3:45

Mind you this is using only one camera (mobileye eyeq3 and several radars.)

Even for simple lane changes you have to have situational awareness. Some examples:

are you merging into a gore point?
are you merging into a lane that's ending?
are you merging into a lane that has a stopped car or some other object?
are you merging while some other car is merging?
are you moving into lane where another car is fast approaching?

I'm also not sure the automatic lane changes will be limited to simple ones. Elon tweeted about having various modes where you could set how aggressive it was. I don't know if there is even such a thing as an easy lane change in California. In WA state lane changes are are pretty easy as drivers generally give you room, and typically respond nicely to the blinker.

How does EAP even thank the other driver when another car gives the Tesla room?

As to that video I'm not sure how it's relevant. Why are you comparing a pilot unit with something that's going to be released soon?

I'm not concerned about who's ahead, and who's behind. I'm not really brand loyal. I just like semi-autonomous features, and autonomous features. It's interesting technology without worrying too much about who is doing it.

Tesla gets talked about because it always feels like they're about to release something exciting, and they generally don't dumb down or limit things. They make the Lawyers sweat, and Elon gives them strokes.

Technology wise I have a hard time seeing a Tesla accomplishing safe lane changes without a rear facing radar. The A7 pilot looks like it has rear facing radar.

Now I certainly would still use it in simple situations like 2 or 3 lane roads without too much traffic.

 

[rnortman]

Even for simple lane changes you have to have situational awareness.

People hugely underestimate the difficulty of unsupervised, uninitiated auto lane change. It is one thing for the driver to request a lane change and for the car to execute it. The car currently does an OK job of that (certainly not perfect) and I imagine that it will get much much better when they start using the side cameras. But for the car to decide on its own and not seek confirmation of some sort from the human driver? That is a whole different ballgame. That's the car taking responsibility for what's happening. If the human initiates it or approves it, the human is responsible for the decision, even if the car executes the decision. When the car takes responsibility for the decision it has to be much, much more sure of itself.

Also I honestly don't know what the law and the insurance industry has to say about that...

 

[MarkS22]

Also I honestly don't know what the law and the insurance industry has to say about that...

That's the big question about all this. The instant you remove the human element, you basically move from the driver's car insurance to product liability insurance. If the vehicle gets into an accident because of an unassisted lane change error, it would be no different than releasing a defective tire. I suspect we'd have to get clarification from the insurance industry (and Tesla and our specific carriers) before using any of these features.

 

[S4WRXTTCS]

People hugely underestimate the difficulty of unsupervised, uninitiated auto lane change. It is one thing for the driver to request a lane change and for the car to execute it. The car currently does an OK job of that (certainly not perfect) and I imagine that it will get much much better when they start using the side cameras. But for the car to decide on its own and not seek confirmation of some sort from the human driver? That is a whole different ballgame. That's the car taking responsibility for what's happening. If the human initiates it or approves it, the human is responsible for the decision, even if the car executes the decision. When the car takes responsibility for the decision it has to be much, much more sure of itself.

Also I honestly don't know what the law and the insurance industry has to say about that...

That's exactly why I'm nervous if it's an L2 implementation as it looks like it's going to be.

That means the responsibility is on us, but we didn't initiate the lane change. Are we seriously supposed to overrule it?

Right now it's pretty easy as I'm either in TACC or AP. TACC allows me to look over my should while not worrying too much about the car in front of me suddenly slowing down. AP has an additional capacity to over rule me if I didn't see someone. So it's basically extra protection without any added risks.

But, if the car does it automatically then I don't really have the opportunity to check.

One possibility is the car will alert us slightly before it changes lanes so we can do a quick confirmation check before it actually does it. if we don't want it to we simply hold the steering wheel preventing it from moving over. But, I can see that leading to trusting it while not going through the normal checks.