Firmware 8.0

[legoman]

Here's a hypothesis regarding everyone's experience with Autosteer on 8.0...

We know that the system was re-engineered to start using the radar as a primary input source. What if that just means that the fleet learning has to re-learn all of the map tiles because the dataset is fundamentally different? I imagine with 7.1, the AP tiles covered just about every square inch of road, so most users had a steady drive. If that "reset" with 8.0, it will take a couple of months for AP to re-learn where the lanes are on hill tops and poorly marked pavement.

I've only had my car since March, but for the original AP owners, does the behavior people are reporting with 8.0 sound similar to what was experienced when Autosteer was first introduced?

 

[Trailblazer]

540 mile round trip this past weekend in South Carolina - 90% on AP. I noticed that 8.0 AP was definitely smoother with respect to speeding up and slowing down - other than that it seemed pretty much the same as it was under 7.1. I got the white pulsing nag 4 or 5 times (my hand was too light on the wheel). Never got the beeping nag. No sudden veering out of the lane. But I have a habit of applying leftward pressure on the wheel when I am passing a truck on the left (only disengaged AP once when doing this on this trip) - so maybe that's why I don't see sudden veering. In general, I'm a fan of the 8.0 update (I don't use USB sticks for music).

 

[rxlawdude]

There are a couple Ss and Xs reporting a 2.38.19 version, which is apparently next out of the bucket. Not sure what's different about it, other than showing up on a couple Xs before anything else.

Yes, but those reports are a few days old. Pretty much nothing being pushed out since the weekend began.

 

[MikeKFE]

My Tesla is in the shop getting the AC worked on. Got notification that I was updated to 2.38.19. Any ideas what's in that update?

Note: S60D that was previously on .108

 

[rxlawdude]

My Tesla is in the shop getting the AC worked on. Got notification that I was updated to 2.38.19. Any ideas what's in that update?

No idea, but you're the first poster here to have it. There have to be some bug fixes, but which and how many is unknown.

 

[MikeKFE]

No idea, but you're the first poster here to have it. There have to be some bug fixes, but which and how many is unknown.

Dang. First S on TeslaFi and EV-FW. So this is what it's like to be famous?

 

[rxlawdude]

Dang. First S on TeslaFi and EV-FW. So this is what it's like to be famous?

No, just Almost Famous.

 

[MP3Mike]

My Tesla is in the shop getting the AC worked on. Got notification that I was updated to 2.38.19. Any ideas what's in that update?

Note: S60D that was previously on .108

My understanding is that is the most recent 8.0 update. So you may be the first person posting on here that got it. When do you get to pick it up? (So you can figure out what has changed and let us know.)

 

[MikeKFE]

My understanding is that is the most recent 8.0 update. So you may be the first person posting on here that got it. When do you get to pick it up? (So you can figure out what has changed and let us know.)

No clue. My AC started going in and out on Friday. This was in addition to AP being temporarily unavailable for the past month (before and after 8). Could be tomorrow. Could be next month. No clue.

 

[Wuggado?]

Here's a hypothesis regarding everyone's experience with Autosteer on 8.0...

We know that the system was re-engineered to start using the radar as a primary input source. What if that just means that the fleet learning has to re-learn all of the map tiles because the dataset is fundamentally different? I imagine with 7.1, the AP tiles covered just about every square inch of road, so most users had a steady drive. If that "reset" with 8.0, it will take a couple of months for AP to re-learn where the lanes are on hill tops and poorly marked pavement.

I've only had my car since March, but for the original AP owners, does the behavior people are reporting with 8.0 sound similar to what was experienced when Autosteer was first introduced?

I floated the same hypothesis (post #2733) a few days back, but no one responded to it. It certainly seems like it to me, and makes perfect sense to me as well. If you're gathering metrics for a particular system, and you make a significant change to it, it seems quite likely that the new system would need at least a few metrics you never gathered. Given that, you can either somehow try to deduce what the metric would have been, or start gathering all over again.

From my perspective, I've been steering my car all by my lonesome for about 40 years now and have somehow survived in spite of that fact. I personally find the MS a lot of fun to drive, so not completely relinquishing driving responsibilities to the car is not even remotely a hardship for me. I have used AP a bit over the last few months, but never felt like keeping my hands on the wheel was a burden then, nor do I feel so now. Like most other people here, I have felt a regression in AP since 8.0 came along, but have discovered a simple workaround for it. It's called doing the driving yourself while waiting for the technology to get (back?) to a solid place.

Hopefully fleet learning takes place even while AP is not engaged. Does anyone know whether it does?

 

[ICUDoc]

Bug#4. This was around in the 7.1 days, or so I've read here. I was standing at a red light with AP engaged. The car in front of me decided to change lanes to the right (less of a line), my car started to follow him slightly (obviously) and then realized it's a different lane and then decided to ACCELERATE into the standing car in front of me. I slammed the brakes. Hard...

Hardly a bug- for crying out loud, when did Tesla say "Use AP on a road with red lights"?

 

[GlmnAlyAirCar]

Here's a hypothesis regarding everyone's experience with Autosteer on 8.0...

We know that the system was re-engineered to start using the radar as a primary input source. What if that just means that the fleet learning has to re-learn all of the map tiles because the dataset is fundamentally different? I imagine with 7.1, the AP tiles covered just about every square inch of road, so most users had a steady drive. If that "reset" with 8.0, it will take a couple of months for AP to re-learn where the lanes are on hill tops and poorly marked pavement.

I've only had my car since March, but for the original AP owners, does the behavior people are reporting with 8.0 sound similar to what was experienced when Autosteer was first introduced?

Yes, I made a similar point a few days ago. Autosteer is behaving very similarly to how it did when 7.0 first came out.

 

[GlmnAlyAirCar]

Hardly a bug- for crying out loud, when did Tesla say "Use AP on a road with red lights"?

Actually the owner's manual, while not citing red lights explicitly, does advertise that TACC and autosteer can be engaged from a full stop as long as following another car. Other than stop-and-go traffic, where else would you encounter this scenario?

 

[msnow]

Yes, but those reports are a few days old. Pretty much nothing being pushed out since the weekend began.

Elon said this week in a tweet.

 

[Tech_Guy]

Hardly a bug- for crying out loud, when did Tesla say "Use AP on a road with red lights"?

exactly most people just click agree and don't bother reading the disclaimer so here it is... View media item 115507

 

[TaoJones]

exactly most people just click agree and don't bother reading the disclaimer so here it is... View media item 115507

And, as expected, that says nothing about divided highways with red lights or any other traffic control devices.

The larger point is that the scenario described has little to do with the red light and everything to do with stop and go traffic.

*shakes head*

Fix the problem. Don't get caught up in fanboism. So simple.

 

[Tech_Guy]

 

[CSFTN]

Here's a hypothesis regarding everyone's experience with Autosteer on 8.0...

We know that the system was re-engineered to start using the radar as a primary input source. What if that just means that the fleet learning has to re-learn all of the map tiles because the dataset is fundamentally different? I imagine with 7.1, the AP tiles covered just about every square inch of road, so most users had a steady drive. If that "reset" with 8.0, it will take a couple of months for AP to re-learn where the lanes are on hill tops and poorly marked pavement.

I've only had my car since March, but for the original AP owners, does the behavior people are reporting with 8.0 sound similar to what was experienced when Autosteer was first introduced?

Not sure the question can really be answered. Why? Tesla started installing AP hardware in August 2014 and as far as we end users knew, didn't turn it on for several months (I think it was 6 months). However, suggestions have been that in fact it was on all that time, but it was watching us and gathering data on the roads not available via GPS. So there was a period approx 6 months where the system was learning and mapping in the background, without being asked to pilot the vehicle. Now, we are using it while (maybe) it is re-mapping via radar and re-learning its fundamentals of piloting via 3D radar.

 

[BertL]

What do you EE guys think about the possibility that Tesla added a hardware abstraction layer to AP, in order to prepare for their separation from MobilEye?

Perhaps, although 8.0 has been delivered to all AP-enabled MS since their inception more than a year ago. It would mean Your speculation would imply Tesla anticipated this far before then to create some unique hardware and begin installing it on vehicles just-in-case. Tesla was/is a cash-poor start-up, so IMHO not likely to spend too much engineering time and product cost on "just in case hardware" in every AP enabled vehicle, or creating a more recent variation under the covers that causes them to deal with two different implementations across the fleet before there really is the next big thing. I suspect Tesla has instead found a way to stretch use of existing AP hardware via firmware only with the radar bias delivered in 8.0 and that will be further extended via fleet learning and distribution in 8.1. What some here on TMC are calling "AP 2.0" will be the next hardware generation at some future date.

So we don't get too sidetracked in this thread, take a look at other threads for more discussion on this, which includes speculation MX has or will have, an iteration between what we have on MS and what 2.0 may be... i.e. the MX front camera housing, various options and needs on both models for changes in rear/side camera and radar, etc...

 

[pdxgibby]

I just completed about 800 miles of driving (mostly highway + AP) this weekend,

The pulsating IC border was a fantastic addition for when my hands were not being detected on the wheel. Not once did I encounter an audible warning.

About half of the AP miles were on straight freeways with 70 mph speed limit. I did notice the slight crest of hill veering issue. The diving toward freeway exits has returned -- leads me to believe that the theory of re-learning is indeed happening. The remainder of the miles were on winding freeways with 65 mph speed limit.

In general, I think AP did an excellent job in lane keeping. I think I noticed that initiating a lane change may reset the 1-minute timer, because it's obvious you are being attentive if you are changing lanes. It could also be that torque is detected and it gave me the impression of 1-minute timer reset. Not sure here.

My biggest complaint is the TuneIn podcasts. I love having access to all these podcasts on my car, but the bugs make it unusable. I would say that podcasts were equally terrible in 7.1, IMHO, but the issues were slightly different.

1) ~80% of the time the podcast will not load for me and I get a "Loading Error." Yes I have LTE connectivity at the time. Does not appear any better when I'm on WiFi. If I repeatedly hit the reload symbol sometimes it will load. Sometimes I give up.
2) Podcast playback location is not preserved if you leave your car (say to supercharge, or to hit a restroom).
3) Scrubbing a podcast to a specific location does not work for me. I waited until the entire podcast was downloaded and then tried to scrub to the position I wanted...but it would re-download the podcast (failing repeatedly), never allowing me to scrub to the desired location.

In the end...just easier to use my iPhone over bluetooth to stream podcasts. What a shame to have this giant screen and potential for great interfaces, and in the end...the software is just as primitive as many other automobile interfaces.