2017.28 c528869

[S100D]

I have a 2016 S90D with AP2 and I installed 2017.28 two days ago. I drove from Atlanta to Birmingham yesterday and had issues with auto lane change and reported it to Tesla. Auto lane change is a lot smoother than the previous 2 releases, but I had the following three issues:

• With the turn signal locked for a left lane change, with no cars close to me the car would not move out of the lane.
• With the turn signal locked for a right lane change, the car would move about half-way and then it would jerk the car back into the left lane.
• I also had a issue with the turn signal locked for a right lane change the car would not move out of the left lane

This was an intermittent issue, sometimes it would work but it failed more than it worked. In all cases there were good lane marking so this was not the cause of auto lane changes failing.

I also had an issue while on autopilot where the car was continually moving slightly to the left and right while in shallow turns.

 

[NerdUno]

I must wonder why there is such a huge discrepancy in experiences with each release. None of the issue above have I encountered, this is, in my experience, by far the best FW revision for AP2 yet.

- Bouncing around in the lanes (i have not experience, it has stayed centered and true)
- Following/distance acceleration and jerkiness (nope, not me, it feels fluid and natural just as I would have driven)
- Phantom braking... (been using it since upgraded about 2 days after start of roll out, not one phantom braking occurrence and I routinely pass through 8 overpasses on way to work every morning)

Is there a huge discrepancy in HW details? Calibrations?

Your point is well taken. I've wondered that myself. I think the answer lies in the types of roads encountered by different drivers. Our trips to Florida on perfectly marked, six-lane interstate highways make AP2 really shine. But driving on interstates around Charleston, SC makes me wonder if they will ever get this right. AP2 will get better in one release and then drop back to worse than awful in the next. Bottom line is I think the current iterations of AP2 are basically follow-the-dotted-line technology with little else. Good lines = Good behavior. Anything else and it's a crap shoot.

It's worth noting that you didn't hear a word about autopilot in all the trade rag test drives of the Model 3 this week, especially zipping around Mulholland Hwy in Topanga Canyon. Nice photo op but no need to send a Model 3 over the cliff with Tesla's lead designer during a test drive.

Exclusive: Tesla Model 3 First Drive Review - Motor Trend

 

[mickey6]

I have stopped using TACC and AP2 completely. Last thing I want is to hurt anyone on the road or in the car. I also refuse to be the use case / guinea pig for Tesla for now as it's dangerous specially on local roads. Hopefully the next update will iron out the "about to kill me" AP maneuvers

 

[realvvk]

I finally allowed the car to install the upgrade and WOW, this version is clearly so much better. I have not tried it on the highway yet but it is rock steady on local roads. Way better than before, immediately noticeable!

 

[ToddS]

I have AP2 and have traveled some 6-700 highway miles yesterday. I thought this version was a huge improvement over the prior version. Changing lanes was very jerky before, but not now. Much smoother.

I did have a few times where it would not change lanes with a clear lane, but I attribute those to mostly user errors since I hit it weird or had just changed lanes and it had not reset yet. I did experience fantom braking once while on local roads, but it was very brief and corrected itself very quickly.

Overall, I think this was the "silky smooth" Musk was referring to a few weeks ago.

 

[sunnyvale]

I can't wait that Tesla upgrade and update AP software even more. That would be really great.
What I noticed though is if you drive on highway by yourself after using AP a lot is dangerous.

 

[TOBASH]

Seems like regeneration is less, so braking less when foot off accelerator...Me No Like!

Still has trouble with stupid deceleration for no reason when using adaptive cruise w/wo autodrive when cars are parked on the side of the road. Constantly retards speed as if it sees an obstacle. Very crappy and annoying.

 

[AnxietyRanger]

Deceleration issues remain, but I agree the steering of the new update is more graceful.

 

[Tompilot]

Just picked up my new blue x three days ago. A/P is working and pretty good but not like in an airplane where you can look away. Yesterday was out for a drive on the highway and a Mercades SUV was trailing me in my left blind spot. Took me a while to figure out that he was watching the car drive itself. I had to wave...with both hands.

 

[impliedshin]

I find the lane changing to be really smooth. However when going down the freeway it ping pongs between the lanes ever so slightly. Didn't notice it until one of the passengers pointed it out to me.

 

[ecarfan]

Every Tesla Auto Pilot Release elicits a wide range of user experiences. I believe that this is largely due to the fact that user reports are anecdotal and idiosyncratic. People are different and they describe their experiences very differently. What to one person is "ping ponging between the lane lines" is to another person "slight twitches of the wheel with the car remaining centered in the lane". One person may describe TACC responding to a car cutting in front of them as an "abrupt deceleration" and another person may call it "gradual slowling". I still chuckle over a post made several years ago where someone described the original Model S cruise control causing the car to "jerk" with each 1mph speed change adjustment. To me the movement of the car when the driver pushed the cruise stalk up was barely perceptible. it was nothing like a "jerk".

I must wonder why there is such a huge discrepancy in experiences with each release. None of the issue above have I encountered, this is, in my experience, by far the best FW revision for AP2 yet.

- Bouncing around in the lanes (i have not experience, it has stayed centered and true)
- Following/distance acceleration and jerkiness (nope, not me, it feels fluid and natural just as I would have driven)
- Phantom braking... (been using it since upgraded about 2 days after start of roll out, not one phantom braking occurrence and I routinely pass through 8 overpasses on way to work every morning)

Is there a huge discrepancy in HW details? Calibrations?

 

[AnxietyRanger]

@ecarfan While no doubt true, I believe different calibrations/alignments in cars as well as differing regional road conditions also cause genuine objective differences. Also time of day, weather...

Therefore the subjective total of these reports is IMO averaging out in a more objective portrayal than any single user reports alone.

 

[larmor]

Every Tesla Auto Pilot Release elicits a wide range of user experiences. I believe that this is largely due to the fact that user reports are anecdotal and idiosyncratic. People are different and they describe their experiences very differently. What to one person is "ping ponging between the lane lines" is to another person "slight twitches of the wheel with the car remaining centered in the lane". One person may describe TACC responding to a car cutting in front of them as an "abrupt deceleration" and another person may call it "gradual slowling". I still chuckle over a post made several years ago where someone described the original Model S cruise control causing the car to "jerk" with each 1mph speed change adjustment. To me the movement of the car when the driver pushed the cruise stalk up was barely perceptible. it was nothing like a "jerk".

Agreed about idiosyncratic reporting. However, there is another possibility, that your individual car is monitoring your driving style and learning to mimic you. If you don't have enough miles to mimic then the output of AP is based on few miles, and may be those drivers who have a lot of miles, have an easier adjustment to the update, as there is just more data on hand to parse. Just a thought...

 

[NerdUno]

The other piece we don't know is what Tesla engineers are doing behind the scenes between these releases. Based on the changes between 17.17.4 and the current release, it certainly appears they have started over on some of this code and some of the road identification data... again. I say that because drives on some highways that were perfectly smooth in 17.17.4 are now a mess. For those that don't think the latest release zig zags on certain roads, you're either driving with blinders on or perhaps need to be. Some highway driving is fine as it was with 17.17.4. Other roads are a train wreck waiting to happen... without the need for the train.

 

[daktari]

Agree with NerdUno. Seems since the 17.14.x versions, it identifies poor lanes and curbs less than before - for me at least.
I also was on a long trip in Denmark recently, and AP2 performed really really good. Then, coming up through Sweden and back to Norway, performance was poorer as roads got more twisty with more undulations and different lane markers. Back on my experimental stretch, it really is no improvment at all. Lots of variables judging AP2 performance.
New update out according to ev-fw it seems, hopefully an improvment this time - waiting for reports.

 

[RubberToe]

Every Tesla Auto Pilot Release elicits a wide range of user experiences. I believe that this is largely due to the fact that user reports are anecdotal and idiosyncratic. People are different and they describe their experiences very differently. What to one person is "ping ponging between the lane lines" is to another person "slight twitches of the wheel with the car remaining centered in the lane". One person may describe TACC responding to a car cutting in front of them as an "abrupt deceleration" and another person may call it "gradual slowling". I still chuckle over a post made several years ago where someone described the original Model S cruise control causing the car to "jerk" with each 1mph speed change adjustment. To me the movement of the car when the driver pushed the cruise stalk up was barely perceptible. it was nothing like a "jerk".

Yay, someone gets it! This exactly.

Everything is subjective. Please post videos of what is believed to be erroneous behavior. A video removes all subjectivity from the equation, and also any inherent personal bias.

Not to offend anyone, of course

RT