Firmware 9 in August will start rolling out full self-driving features!!!

[Bladerskb]

The regulatory approval must have come from somewhere else.

Or there wasn't one at all. which would be par for the course for things elon tweets.

 

[ucmndd]

The regulatory approval must have come from somewhere else.

REGULATORS!!!!!!!! Mount up!

Elon on Twitter: Approval secured

 

[HolyGrail]

@Bladerskb Have you ever really worked with ME's data?
If you had, you would probably know better.

Imho Teslas latest 3D vehicle pose estimation is about on par with the current ME offering, maybe even less yaw jittery.
Teslas lanes seem even a bit further reaching and more stable.
ME may be more advanced in international sign recognition though.
I guess that is what is currently keeping NoA from being rolled out in Europe.

Never forget teslas function developers have proven they can create astonishing functions with surprisingly sparse data (2016).
Imagine what they could do with this state of the art data?
And the data we have seen isn't even from the ingenious AKNET_V9 which surely needs HW 3.0 to run fast enough.

 

[emmz0r]

Regulatory approval = codeword for nearing beta release

 

[Bladerskb]

@Bladerskb Have you ever really worked with ME's data?
If you had, you would probably know better.

Imho Teslas latest 3D vehicle pose estimation is about on par with the current ME offering, maybe even less yaw jittery.
Teslas lanes seem even a bit further reaching and more stable.

You are joking right? They haven't even reached the accuracy of AP1's 5 years old EyeQ3

Navigate on Autopilot is Useless (2018.42.3)

 

[diplomat33]

You are joking right? They haven't even reached the accuracy of AP1's 5 years old EyeQ3

Navigate on Autopilot is Useless (2018.42.3)

2018.42 is old now. I am on 2018.50 and NOA is much better now.

 

[Bladerskb]

2018.42 is old now. I am on 2018.50 and NOA is much better now.

Sure its 2 months old. But the fact that they are still trying to catch up with a 5 year old tech doesn't sound alarming to you? Yet every tesla fan you talk to tell you how they are 3, 10 years ahead? don't you see the problem here?

Here's what @wk057 said about the AP2's output.

• Vehicle detection to the sides and behind your vehicle is complete garbage.
  • This is super obvious when sitting still with other still vehicles all around. You'll seem them "swimming" around the visualization, colliding with each other, with you, etc.
  • Also obvious when overtaking large vehicles. Almost every single semi truck, bus, or RV I passed ended up with a twin ghost visual on the screen.
  • Finally, vehicles to the side are regularly shown overlapping my own vehicle visual, despite them being firmly in their own lane.
  • Radar/vision fusion on AP2 appears to be significantly worse than AP1, with AP1 easily accurate for a few cm... AP2 easily worse than +/- 1m... very obvious when looking at the lead vehicle visualization.

The lane outputs are so bad and unstable that he couldn't even create a HD Map like he could easily do with AP1's eyeq3 super accurate outputs.

I've managed to make my AP1 S drive with near-zero interaction from my driveway to my old shop (6 miles, 6 stops/turns). I did this in a pretty hacky way, though, and probably not something I'd ever consider being usable as a real AP1 feature.

Basically I setup a database that monitored that route over dozens of human-driven passes and built up a pretty high resolution representation of the route using GPS data, reported lane positions at those coordinates, and other reported consistent factors with a particular location point. Through some clever software, I can use this data to augment the actions of AP1 on the route, including full stops and full turns. The result was surprisingly good, and I managed to do the route a few times with zero interaction. Most of the time, though, since AP1 can't see cross traffic, I'd have to disengage at intersections and reengage when clear.

The biggest problem with porting these mods to AP2 would be that the data produced by AP2 is super inconsistent. With AP1, I can drive a path at a constant speed, record the data, do it 10 more times, and compare the data of each pass only to find super minor variations in the information recorded about the path. With AP2, I logged some of this info and the same path run 10x might as well look like you were on another planet when each recording was made. So they definitely have some issues to work out to catch up to the quality of info produced by AP1.

@HolyGrail

 

[diplomat33]

Sure its 2 months old. But the fact that they are still trying to catch up with a 5 year old tech doesn't sound alarming to you? Yet every tesla fan you talk to tell you how they are 3, 10 years ahead? don't you see the problem here?

To be fair, I never had a AP1 car so I can't directly compare the two. But I am very happy with the performance of NOA and EAP in my current AP2.5 car. So for me, it does not feel like they are "catching up", it feels like a very solid ADAS system. Lane keeping is rock solid, auto lane changes are very smooth, exit taking is good etc..

 

[Bladerskb]

To be fair, I never had a AP1 car so I can't directly compare the two. But I am very happy with the performance of NOA and EAP in my current AP2.5 car. So for me, it does not feel like they are "catching up", it feels like a very solid ADAS system. Lane keeping is rock solid, auto lane changes are very smooth, exit taking is good etc..

I'm talking about the NN output, @wk057 sees both.

You just liked a post that speculated terribly that AP2 was better.

 

[am_dmd]

This is how good NOA is at this time. Only use it once in a while to check if it is worth turning it on and absolutely not worth it at all.
Recently had AP1 loaner and just TACC and lane changes alone is worth more than AP2+

 

[diplomat33]

This is how good NOA is at this time. Only use it once in a while to check if it is worth turning it on and absolutely not worth it at all.
Recently had AP1 loaner and just TACC and lane changes alone is worth more than AP2+

That's just a still picture as far as I can tell. Is it supposed to show something in particular?

 

[MorrisonHiker]

This is how good NOA is at this time. Only use it once in a while to check if it is worth turning it on and absolutely not worth it at all.
Recently had AP1 loaner and just TACC and lane changes alone is worth more than AP2+

I saw that prompt several times on a recent road trip but just ignored it. I usually saw it at highway on/off ramps.The car continued driving the selected route without problems. I think it was wanting me to get in the left lane in case there were cars entering the highway.

 

[rnortman]

That's just a still picture as far as I can tell. Is it supposed to show something in particular?

I believe the complaint is that it consistently asks you to get in the left lane "to follow route" when you have no exists for the next 50 miles and the right lane is perfectly good. Or in the case in the photo the middle lane.

I believe it wants you to be in the left lane because it dives into the exit and entrance ramps if it's in the right lane at that point, and this is embarrassing for Tesla and frightening and dangerous for the people on the road. So their "fix" is to ask you to be in the left lane through interchanges if you're not taking an exit. AAF.

 

[MikeKFE]

FWIW, I picked up a 2019 S on Friday. This is replacing my 2016 S. I installed 2018.48.12.1 Saturday morning.

I find the EAP to be much better than my AP1 car. MUCH better.

It seems like NAP still has a lot of work to be done on it. Just like others have noted before me, the lane change requests don't seem to make much sense. In more than one instance, I would change lanes to pass a car(s) on the interstate as per NAP's request. As I'm passing those cars, NAP would then tell me to go back into the far right lane to prepare for an exit that was as far as 10 miles away. It was encouraging me to do this despite the fact that there was literally no room for me to merge back over.

A few times, just for S&Gs, I would follow NAP's recommendation. I'd merge back to the right lane many many miles before my next required exit into what I knew to be traffic going slower than the speed limit. As soon as I merged right, NAP would then tell me to change into the left lane to pass the slow traffic. Again, I tested this multiple times.

So yeah, work to do.

I was impressed-ish with NAP's handling of interstate interchanges. That was pretty much flawless.

Overall, I am happy with the new EAP functionality compared to what I had w/ my AP1 setup.

 

[MorrisonHiker]

I believe the complaint is that it consistently asks you to get in the left lane "to follow route" when you have no exists for the next 50 miles and the right lane is perfectly good. Or in the case in the photo the middle lane.

I believe it wants you to be in the left lane because it dives into the exit and entrance ramps if it's in the right lane at that point, and this is embarrassing for Tesla and frightening and dangerous for the people on the road. So their "fix" is to ask you to be in the left lane through interchanges if you're not taking an exit. AAF.

I ignored it every time I saw it on an 1100 mile trip last month. It never dove for any exits. It did have that problem back in 2017, many versions ago.

 

[rnortman]

I ignored it every time I saw it on an 1100 mile trip last month. It never dove for any exits. It did have that problem back in 2017, many versions ago.

It dives for exits for me regularly if they are the sort where the lane widens without a dashed line delineating the exit lane. It is even more likely to dive into entrance ramps that are merging in, again if there is not a dashed line delineating them all the way until the acceleration lane disappears (which is very common).

 

[d21mike]

It dives for exits for me regularly if they are the sort where the lane widens without a dashed line delineating the exit lane. It is even more likely to dive into entrance ramps that are merging in, again if there is not a dashed line delineating them all the way until the acceleration lane disappears (which is very common).

For me the problem is that it always try's to "center" in the lane. This is almost always a mistake with AP. A standard lane width is 12'. I think it should first try to center starting with the left lane marking (US) and a 12' max lane when the lane marking to the right is lost (goes wide) until it has other information to go on. If it did that it would solve 95% of my problem with missing right lane markers.

 

[rnortman]

For me the problem is that it always try's to "center" in the lane. This is almost always a mistake with AP. A standard lane width is 12'. I think it should first try to center starting with the left lane marking (US) and a 12' max lane when the lane marking to the right is lost (goes wide) until it has other information to go on. If it did that it would solve 95% of my problem with missing right lane markers.

Except when you're in the left lane, it should do the opposite to avoid diving into left exits. As it currently stands it is more confused by left exits than right exits, presumably due to more training data for right exits, or perhaps the data about right exits in the map is better.

 

[d21mike]

Except when you're in the left lane, it should do the opposite to avoid diving into left exits. As it currently stands it is more confused by left exits than right exits, presumably due to more training data for right exits, or perhaps the data about right exits in the map is better.

Agreed it needs to do both. I do a lot of driving on surface streets where you loose the right lane marker and it need to pick up the curb and it decides to center between left lane marker and the curb and this can be a very large lane to center in. Really not sure why they have not solved this already. Seems like it would be one of the more easier problems to solve but I guess not.

 

[HolyGrail]

@Bladerskb
Please point us directly to the post where wk057 said that:

• Vehicle detection to the sides and behind your vehicle is complete garbage....

It would surely be interesting to read.
I couldn't find the original post googling some parts of it.
Couldn't pick it up from a quick glance over his recent posts either.