Software Update 2018.10.4

[DayDrinker]

Got 10.4 and got my first MCU reboot while on the freeway....boo

 

[cbdream99]

I agreed the aggressive braking is something that should be improve, if I am not in AP then I only needs to ease off the pedal instead of braking.

 

[terryna]

Hope it’s a general release

 

[cbdream99]

Here's a video this morning with AP2 on the way to work, look at the point after it crosses a big junction with limited markings, it swerved to the right a little bit and then corrected itself, but it's pretty cool.

 

[Sawyer8888]

The "juddering" has been discussed at length here: Hard braking with TACC in stop-n-go traffic

My brake pedal just disengaged autopilot automatically without warning while I was in stop-and-go traffic. This was likely due to the stuttering/juddering. I think that's a first. I'm sure this was mentioned in that other thread that you're referencing.

 

[TaoJones]

Forgot one:

The AP2 lag to engage AS after TACC is still problematic.

With AP1, you could engage TACC and then AS with successive rapid pulls of the stalk.

With AP2 since its advent and to this day, especially at lower speed, if you don’t wait between pulls, the attempt to establish AS *cancels TACC* - and that is a PROBLEM since that also cancels your distance buffer maintenance between you and the car in front of you.

I am not going to suggest that anyone “try this”, but if you have already experienced it, it ain’t fun.

Maybe someday this will get addressed during AP2’s aspirational journey toward actual parity with AP1.

 

[sevenout]

My brake pedal just disengaged autopilot automatically without warning while I was in stop-and-go traffic. This was likely due to the stuttering/juddering. I think that's a first. I'm sure this was mentioned in that other thread that you're referencing.

First I've heard of anyone having AP disengage due to hard braking. Don't think this was ever mentioned in the other thread.

 

[SSonnentag]

Just to be clear, the car pressed the brake pedal, not you, right?

 

[SSonnentag]

I'm trying to understand this. Are you saying you pull once to engage TACC and then within a second or so double-pull to engage AP and TACC is disabled leaving you with auto-steering only?

Forgot one:

The AP2 lag to engage AS after TACC is still problematic.

With AP1, you could engage TACC and then AS with successive rapid pulls of the stalk.

With AP2 since its advent and to this day, especially at lower speed, if you don’t wait between pulls, the attempt to establish AS *cancels TACC* - and that is a PROBLEM since that also cancels your distance buffer maintenance between you and the car in front of you.

I am not going to suggest that anyone “try this”, but if you have already experienced it, it ain’t fun.

Maybe someday this will get addressed during AP2’s aspirational journey toward actual parity with AP1.

 

[mongo]

First I've heard of anyone having AP disengage due to hard braking. Don't think this was ever mentioned in the other thread.

@Sawyer8888
If the braking bounces the brake pedal enough, it could trip the brake pedal switch and make the car think the driver is pressing the brake. Switch may be slightly out of adjustment making it more sensitive.

 

[headcase]

Yep. This 10.4 update is a step forward and a step back, like so many other updates - well-received or otherwise.


Step forward:

Continuation of 6.1’s longer forward view (recognition *and* reaction), especially with regard to stopped traffic.

Seemingly more responsive sensor “balloons” to each side.

Less nervous overall, but still not as or more confident than AP1.

Better management of intersections *without medians* and with elevation changes.


Step back:

Car still heads for medians when crossing intersections with any kind of elevation change - intervention required to avoid impact.

Phantom braking events with nothing ahead and vehicles behind - intervention required to avoid getting rear-ended.

More conservative/timid when resuming after slowing down.

Limited lane changing on divided highways relative to previous versions.

No appreciable difference in HOV lane travel which is to say car still routinely crosses lane markers directly adjacent concrete barriers. AP1 rarely to never did this.

All in all, 10.4 is, so far, one of the better AP2 updates.

Hopefully, this also sets the stage for the “big” update due any month now. Which if I had to guess, would be within the next month - after the current EOQ merriment passes.

Thanks for the bulleted summary, as this is one the first detailed comparisons I've read of AP2 performance with 10.4 vs. AP1. Sounds like AP2 still has a little bit more to go before parity with AP1 (although it sounds like its not too far away). Sudden and hard ghost braking on AP2 loaners always scares the hell out of my family, and basically a show stopper for me to use AP2.

 

[Sawyer8888]

Just to be clear, the car pressed the brake pedal, not you, right?

I asked myself the same. "Did I do that?" Although my feet were at rest and my right foot was closer to the accelerator. And my hands were not on the wheel. Definately strange.

 

[prash]

I'll give the auto headlights a try. I've basically left them deactivated since they finally made them available because they were useless. I was constantly flashing people with my highbeams or they were turning on and off so often it was annoying.

I see some improvement in auto highbeam with the new update. I had similar on and off flashing issue before.

 

[buttershrimp]

Forgot one:

The AP2 lag to engage AS after TACC is still problematic.

With AP1, you could engage TACC and then AS with successive rapid pulls of the stalk.

With AP2 since its advent and to this day, especially at lower speed, if you don’t wait between pulls, the attempt to establish AS *cancels TACC* - and that is a PROBLEM since that also cancels your distance buffer maintenance between you and the car in front of you.

I am not going to suggest that anyone “try this”, but if you have already experienced it, it ain’t fun.

Maybe someday this will get addressed during AP2’s aspirational journey toward actual parity with AP1.

The lane change improvement and the fixing of the searching on hills and crests has made AP2 clearly ahead for me now... AP2 clearly navigates roads that AP1 cannot but not vice versa. This is a week after I said I don't see how AP2 would surpass AP1 until it can recognize speed limits. Turns out that there was a way to surpass it.... hills and crests and something closer to AP1 lane change parity. I still don't like how AP2 is a bit harder to engage in a lane change on certain roads, but I think as soon as AP2 can read signs, I don't think it will lag in performance in any area, though the IC display would technically not be doing the same thing.... but performance would be the same. @TaoJones, can you explain the TACC thing for me? It's clear you use TACC much more than I do and I am having a hard time understanding the benefit.

 

[buttershrimp]

I just stumbled on this post. Definitely some truth to it since your experience is very similar to mine. There are areas it is behind AP1, BUT... here is the deal. AP1 loses the fight when I can navigate any road with rolling hills without a disengagement or drifting. Unless AP1 fixed this issue, it's now clearly behind AP2 performance for me, especially on Austin roads with hills. What you are describing appears to me to be fit and finish edges of autopilot.... these matter greatly, and AP1 still has an advantage in some areas.... no doubt. But it just got leaped by AP2 (at least for me) because I now have true hill control and ability to see stopped cars when approaching them at speed even on curved roads.

I do want to see video of what people are talking about with the HOV lane issue, I have noticed getting too close for comfort to concrete in construction zones with temporary lanes... this must be the scenario you are experiencing on a non-construction road. Interesting.

Yep. This 10.4 update is a step forward and a step back, like so many other updates - well-received or otherwise.


Step forward:

Continuation of 6.1’s longer forward view (recognition *and* reaction), especially with regard to stopped traffic.

Seemingly more responsive sensor “balloons” to each side.

Less nervous overall, but still not as or more confident than AP1.

Better management of intersections *without medians* and with elevation changes.


Step back:

Car still heads for medians when crossing intersections with any kind of elevation change - intervention required to avoid impact.

Phantom braking events with nothing ahead and vehicles behind - intervention required to avoid getting rear-ended.

More conservative/timid when resuming after slowing down.

Limited lane changing on divided highways relative to previous versions.

No appreciable difference in HOV lane travel which is to say car still routinely crosses lane markers directly adjacent concrete barriers. AP1 rarely to never did this.

All in all, 10.4 is, so far, one of the better AP2 updates.

Hopefully, this also sets the stage for the “big” update due any month now. Which if I had to guess, would be within the next month - after the current EOQ merriment passes.

 

[fasteddie7]

I just stumbled on this post. Definitely some truth to it since your experience is very similar to mine. There are areas it is behind AP1, BUT... here is the deal. AP1 loses the fight when I can navigate any road with rolling hills without a disengagement or drifting. Unless AP1 fixed this issue, it's now clearly behind AP2 performance for me, especially on Austin roads with hills. What you are describing appears to me to be fit and finish edges of autopilot.... these matter greatly, and AP1 still has an advantage in some areas.... no doubt. But it just got leaped by AP2 (at least for me) because I now have true hill control and ability to see stopped cars when approaching them at speed even on curved roads.

I do want to see video of what people are talking about with the HOV lane issue, I have noticed getting too close for comfort to concrete in construction zones with temporary lanes... this must be the scenario you are experiencing on a non-construction road. Interesting.

Driving and AP1 and AP2 vehicle I can say they have been both improving together. On 6.1 ap1 started getting reliable on crests of hills, and not little hill bumps, but big mountains of Pennsylvania crests. 10.4 brought ap2 up to speed and ap1 got even better over hills and they are both pretty solid. There’s still a couple of sharp curves into crests of hills that ap1 is tackling because it slows into the curve and ap2 is failing because it barrels into the turn and slows down too late.

 

[ucmndd]

First time ever this morning I had autopilot disengage with the "take over immediately" message and refuse to start back up again for about 10 minutes due to "reduced front camera visibility" (not sure I've got the error message exactly right, but it's something like that).

To be fair, it was raining pretty hard. But I've never had that happen before, even in pretty significant downpours. Just an observation.

 

[animorph]

Forgot one:

The AP2 lag to engage AS after TACC is still problematic.

With AP1, you could engage TACC and then AS with successive rapid pulls of the stalk.

With AP2 since its advent and to this day, especially at lower speed, if you don’t wait between pulls, the attempt to establish AS *cancels TACC* - and that is a PROBLEM since that also cancels your distance buffer maintenance between you and the car in front of you.

I am not going to suggest that anyone “try this”, but if you have already experienced it, it ain’t fun.

Maybe someday this will get addressed during AP2’s aspirational journey toward actual parity with AP1.

I haven't seen this, but maybe my touch on the stalk is a little different than yours. I can do an initial two pulls to enable both TACC and AS. Or I can pull once for TACC and then pull two additional times for AC. Certainly I've done the later case with a long delay between TACC and AS. I'm pretty sure I've done it with just a few seconds between when I screw up and just start TACC.

 

[conman]

I haven't seen this, but maybe my touch on the stalk is a little different than yours. I can do an initial two pulls to enable both TACC and AS. Or I can pull once for TACC and then pull two additional times for AC. Certainly I've done the later case with a long delay between TACC and AS. I'm pretty sure I've done it with just a few seconds between when I screw up and just start TACC.

I find the same. If the steering wheel icon is there I can go straight to AS with two pulls every time without TACC being engaged. I'm not sure what he's saying there..

 

[buttershrimp]

Driving and AP1 and AP2 vehicle I can say they have been both improving together. On 6.1 ap1 started getting reliable on crests of hills, and not little hill bumps, but big mountains of Pennsylvania crests. 10.4 brought ap2 up to speed and ap1 got even better over hills and they are both pretty solid. There’s still a couple of sharp curves into crests of hills that ap1 is tackling because it slows into the curve and ap2 is failing because it barrels into the turn and slows down too late.

Are you on 10.4? I did a bunch of video with equal firmwares a week before 10.4 launched on ap2.... and ap1 was improving but on hills and crests but was definitely still a problem. Once 10.4 dropped, it not only improved, it fixed it. I have had zero issues with it. The fact that I can navigate more real roads than AP1 without it being awkward puts the capability performance ahead, even if there are certain aspects of AP1 that are still better, But as a whole, the success rate on difficult roads is better... can't wait for the next step. This road on AP1 still fails multiple times. Check the aP2 video out.