Software Update 2018.18.2 301aeee

[chillaban]

As most of us know 2018.10.4 was a HUGH improvement on detecting stopped cars and coming to a safe stop if driving < 50mph. Again, these updates "can" be pretty important so why not allow us to be on the current version.

But that's where it goes both ways. Many people who drive near there actually said it was a 10.4 regression that it was more likely to identify the gore lane as a driving lane. Previous versions didn't make that mistake.

Similarly, builds newer than 2018.12 seem to have more phantom braking events compared to earlier builds.


So yes, maybe some updates improve safety and that's an argument for being on the latest. But other times, arguably newer firmware can have bugs that regress safety in edge cases too. And that's probably one reason that Tesla is really slow at trickling out new updates with behavior changes.

 

[daktari]

How about the Model 3 brake failure? Anyone tested if one update suddenly introduces that? Or maybe we AP2 cars have it already?

 

[d21mike]

But that's where it goes both ways. Many people who drive near there actually said it was a 10.4 regression that it was more likely to identify the gore lane as a driving lane. Previous versions didn't make that mistake.

Similarly, builds newer than 2018.12 seem to have more phantom braking events compared to earlier builds.


So yes, maybe some updates improve safety and that's an argument for being on the latest. But other times, arguably newer firmware can have bugs that regress safety in edge cases too. And that's probably one reason that Tesla is really slow at trickling out new updates with behavior changes.

Good points. My experience was the car now "centers" in a wide lane which would not help in this case. But neither would being LEFT in the lane. But being RIGHT in the lane may have helped. Of course coming to a STOP would have really helped. And that was the biggest change for me in 10.4. But again I think that is mostly when traveling < 50mph which this car was not doing. And of course it was not a stopped car in his lane.

 

[chillaban]

Good points. My experience was the car now "centers" in a wide lane which would not help in this case. But neither would being LEFT in the lane. But being RIGHT in the lane may have helped. Of course coming to a STOP would have really helped. And that was the biggest change for me in 10.4. But again I think that is mostly when traveling < 50mph which this car was not doing. And of course it was not a stopped car in his lane.

I totally agree with your observations — I've found .10.4 to be better in basically 99.9% of cases compared to before, by a wide margin. I did call out a couple intersections that .10.4 could no longer get through without a sudden unwanted lane change, but that was "fixed" it seems in .12.

Personally I've been lucky enough to not see any overpass braking, but others don't seem as lucky.


And as someone who's on .18, it already feels old compared to .18.2 and .20. Totally agree that I want a way to force an update without making a service center feel like they're doing me a huge favor…. But I would be wary about whether or not it's good for safety for people to be able to get the latest version and bypass Tesla's tiered rollouts.


If anything, IMO they should have a "fast/slow" option like Windows Insider Builds that users can opt into. One thing that I've noticed in my line of work is that everyone posting on forums complains about too few updates / bugs not getting fixed quickly enough / not getting offered the latest. But the majority of the customer population is not represented by forum posts, and those customers tend to be complaining about the exact opposite: Too many / too frequent updates, updates that don't seem to have major feature changes, etc.

 

[S100D]

You really just suggested people call Tesla drivers DICs? Some, sure, but not all.

Croman, my Driver's In Command was a play on words trying to say that if you are on autopilot and you have an accident, you were the Driver In Command and it is 100% your fault. There are too many accidents where drivers take their eyes off the road and rely 100% on the AP to keep them from having an accident. As long as this continues, we will have drivers running into fire trucks while texting, running into guard rails in construction zones, expecting the autopilot to handle all situations.

I am currently in Colorado and by the time I return to Texas I will have used the autopilot close to 2,000 miles. It is a great tool to reduce driver fatigue but I watch it like a hawk and disengage it for construction zones, and other similar times where it has scared the crap out of me. I am also a pilot that has close to 4,000 hours of Pilot In Command time and understand whose fault it is if an accident happens.

Like all of us, I am waiting for all of the promised features that were due in 3 to 6 months starting in October to December of 2016. Hopefully we will get that next big release in 3 to 6 months that will give us new features that will make the Tesla safer on Autopilot.

Until Tesla gives us Automatic Emergency Braking to a full stop like other car manufactures are doing, I will never trust autopilot. I believe that they could have done this years ago and they are not doing it due to a liability issue. They will have to do this to implement Full Self Driving and I cannot understand why we do not have it now.

For all of those I offended, I apologize.

 

[croman]

@S85D It was a joke. Sorry I came off like a dic(k). ;D

 

[S100D]

I really wasn't coming from the point of responsibility in a crash. I agree 100% it is the drivers fault.
My point was the guy was 7 versions back and SC seemed reluctant to give him the update. They said they would update him this time but it would be the last time. Hopefully they meant until he is 7 updates behind again?

But regarding updates in general. The guy in the X who lost his life because the X ran head first in to the concrete divider. I think we all agree it was the drivers fault. He should have been able to easily take over if not distracted. But lets say he was 1 or more updates behind a major improvement with AP that would have stopped the car and he would still be alive. Accidents happen all the time and people die all the time due to driver error. To me I would like to believe the AP might just save me from making that mistake where I got into a very serious accident due to a mistake on my part. So, being on the latest firmware version just might be the one that did that. As most of us know 2018.10.4 was a HUGH improvement on detecting stopped cars and coming to a safe stop if driving < 50mph. Again, these updates "can" be pretty important so why not allow us to be on the current version.

Mike, I agree with you regarding autopilot becoming safer and more reliable, but we all know that we cannot trust it 100% of the time. In my earlier response to Croman, I said that until Tesla does an automatic emergency braking to a full stop, I will not trust autopilot. The AEB needs to work at highway speeds and not 50 mph and lower. There will be circumstances were a deer jumps out in front of you or a car pulls in front of you, even with the best AEB, there is not enough time to stop before a collision happens.

In one of Elon's earlier tweets he said that at some future date we will be able to request software updates. That would not have helped the person that was 7 updates behind because the logs said that his car was rejecting updates. Unfortunately, Tesla is not doing a complete regression test on all of their software before they release it. They fix some issue and create new ones so moving to the latest release may introduce more problems than it fixes.

 

[S100D]

@S85D It was a joke. Sorry I came off like a dic(k). ;D

No problem Croman, I think you knew I was joking also but it may have offended other people that do not understand our humor.

 

[S100D]

But that's where it goes both ways. Many people who drive near there actually said it was a 10.4 regression that it was more likely to identify the gore lane as a driving lane. Previous versions didn't make that mistake.

Similarly, builds newer than 2018.12 seem to have more phantom braking events compared to earlier builds.

So yes, maybe some updates improve safety and that's an argument for being on the latest. But other times, arguably newer firmware can have bugs that regress safety in edge cases too. And that's probably one reason that Tesla is really slow at trickling out new updates with behavior changes.

I have an AP2 car that had issues with phantom braking on earlier releases. The car slammed on the brakes for over head signs, over passes, large signs on the side of the road, and slight inclines like going over a bridge. If you are having this problem, I recommend that you take your car to the SC and ask them to calibrate your radar. I had this done and I have had very few if any phantom braking incidents. I am currently on 2018.18.2 and do not remember having any since I installed it.

By the way, I told the SC that this was a safety issue and if I got rear ended because of phantom braking, I would hold Tesla responsible. I sent an e-mail to the SC service manager telling him this.

 

[luckyj]

Finally got to do one of my 250mi regular highway drives with this version. I came from 2018.12. Autosteer seems a little smoother. Lane change, when it works correctly, is good. But I'm getting abrupt aborts mid lane change every so often that I didn't get before. TACC seems to be reacting to coming up on slower cars in my lane a little better now. But I did get a couple of phantom braking events in areas I didn't before, wasn't "slam on the brakes" kind like in the past thank goodness. Following distance does seem greater for the same setting. But actually I think it's gone back to what it was at some point in the past and I think it's better. I get the feeling that TACC decisions are using a little more radar. Overall, seems like an improvement save for the lane change aborts.

 

[caligula666]

Same here. Tried 18.2 on longer distance and I got much more random breaking as well as the car breaking for a car that had already moved in a different lane (but with AP still treating it like if it was on mine), so I ended up having to “force” AP quite a bit by accelerating over AP to not look like a drunk driver ...

Anyway, the bottoms line is that the monkey-on-my-laps didn’t get much better at driving in the last few months.

 

[ChrML]

I really wasn't coming from the point of responsibility in a crash. I agree 100% it is the drivers fault.
My point was the guy was 7 versions back and SC seemed reluctant to give him the update. They said they would update him this time but it would be the last time. Hopefully they meant until he is 7 updates behind again?

But regarding updates in general. The guy in the X who lost his life because the X ran head first in to the concrete divider. I think we all agree it was the drivers fault. He should have been able to easily take over if not distracted. But lets say he was 1 or more updates behind a major improvement with AP that would have stopped the car and he would still be alive. Accidents happen all the time and people die all the time due to driver error. To me I would like to believe the AP might just save me from making that mistake where I got into a very serious accident due to a mistake on my part. So, being on the latest firmware version just might be the one that did that. As most of us know 2018.10.4 was a HUGH improvement on detecting stopped cars and coming to a safe stop if driving < 50mph. Again, these updates "can" be pretty important so why not allow us to be on the current version.

What if the new version had a quirk that causes a death, that would not have killed the driver in previous version?

I fully understand they want to roll out software incremental and careful. If you have just changed a lot of code, how would you choose to roll out after internal beta?

Press the red button and 200 000 cars are instantaneously on the new version, shame if it bricked them all causing 200 000 ranger outcalls to manually flash firmware.

Or press the blue button and try it on 5000 first, then 20 000 in a week and 50 000 in two weeks. Cancel at any time. Much better risk management, reducing consequences if something bad.

Altough what I don't understand is why they don't add this Request Latest firmware button immediately. Early adopters tend to be more forgiving, while those who don't nag for new firmware get a new slow ring version every few months a few weeks after the early adopters have tested the *sugar* out of it.

 

[daktari]

I can now confirm auto lane change suddenly became available on my 18.2 car, over a week after I updated. Still no speed sign recognition though.
How did they enable it? ADAS-tiles changed? Changed a switch OTA? Spooky?!

 

[chillaban]

I can now confirm auto lane change suddenly became available on my 18.2 car, over a week after I updated. Still no speed sign recognition though.
How did they enable it? ADAS-tiles changed? Changed a switch OTA? Spooky?!

Wow, that's fascinating! I didn't notice any changes this morning into work on 2018.18. I wonder if this is region specific.

 

[daktari]

Yes, maybe? I have not gotten the new NAV maps yet.

 

[SeBsZ]

A bunch of us including me on the NL forum also noticed lane change support on previously unavailable roads all over the Netherlands. I first noticed three days ago. There has definitely been an update of ADAS tiles, possibly in certain regions only.

 

[luckyj]

After doing a fair bit of local driving today, it seems like this version is frequently identifying multiple cars ahead of me in the IC. I've noticed several times the IC displayed and tracked three cars - the one I'm following and two in front of it - fairly accurately. Don't remember that with 2018.12.

 

[arubinst]

After doing a fair bit of local driving today, it seems like this version is frequently identifying multiple cars ahead of me in the IC. I've noticed several times the IC displayed and tracked three cars - the one I'm following and two in front of it - fairly accurately. Don't remember that with 2018.12.

This is AP2, right?

 

[MelaniainLA]

Tesla not only needs to let us request updates - they ALSO need to allow easier updates (and maps and nav) over LTE. Especially with Model 3 coming out, many of us live in apartments and garages with no WiFi. We are always multiple updates behind. Service hates pushing over LTE because they claim “data is too expensive”. So I have to bring in the cars and use MORE expensive service center labor. How does that make sense? Happy to even pay them a $5 per update Over LTE fee!

 

[ahkahn]

Tesla not only needs to let us request updates - they ALSO need to allow easier updates (and maps and nav) over LTE. Especially with Model 3 coming out, many of us live in apartments and garages with no WiFi. We are always multiple updates behind. Service hates pushing over LTE because they claim “data is too expensive”. So I have to bring in the cars and use MORE expensive service center labor. How does that make sense? Happy to even pay them a $5 per update Over LTE fee!

Perhaps they offer a "premium" LTE service with additional features (like LTE updates, etc) for an additional charge per month. Those of us who have adequate wifi at home and at work don't really need it...