-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
2017.38 f87c64d5
- Thread starter Peteybabes
- Start date
-
- Tags
- 2017.38 f87c64d5 Model X
StefanSarzio
Member
Seemed to be much smoother after the update and the car does not weave much in the lane like it used to before the update. However, I'm not sure if that is coincidental (i.e., car learning) or because of the update.
Interestingly, it changed lanes and moved over from the middle lane to the fast lane this morning. I let it do it because there was no traffic, but it didn't look like an intentional lane change (no indicators or chimes) even though it looked like it had detected both lanes and the display showed it was crossing over. It then continued at the set speed on the first lane. Not sure what to make of it.
Also, I did get a lot of warnings about steering and rear seats after the update which Tesla troubleshot and concluded to be a problem with the software update (and I'm hoping that's the case and not the alternative of my car week old car being a lemon).
bnsfengineer
Active Member
Does you car change lanes rather aggressively? I'm not really used to it simply because I have never changed lanes so fast before and really try not to use unless I find wide open spaces which is rare in So.Ca.
marksvend
Member
I believe I am seeing improvements with AP2 after the update. Previously AP was unusable for most of my commute in the HOV lane, but for the past two days it has made far fewer mistakes and has stayed more centered between lane lines without ping ponging. There's still one curve where it wants to cross over the lane line to the right. Other than that, AP is actually usable for about half of my commute now.
Lane changes seem a little less aggressive now, too.
Lane changes seem a little less aggressive now, too.
GlmnAlyAirCar
Active Member
SteveMost
Member
Sir Guacamolaf
The good kind of fat
Of course, YMMV, and this was by no means scientific (n=1 and maybe 30 miles driven), but I took my MS (AP 2.5) for a spin today after updating to 17.38 and I perceived significant improvements on local roads and on a two-lane stretch of interstate. On local roads, I was able to get through multiple intersections where lane lines ended on one or both sides and drive past turn-off lanes where there were lane line discontinuities without AP getting confused. It just seemed like the system was much better at picking the correct lane line to follow (on either side) and not randomly trying to drive me off the road/into oncoming traffic. Also seemed to track much better with less ping pong on a somewhat windy two-lane highway which AP has had issues with in the past (including veering halfway into the right lane from the left at a certain point where there is a down slope around a bend to the left). Maybe wishful thinking/placebo effect, but the AP in 17.38 seems promising so far.
Agreed. The subject version , 2017.38 f87c64d5,
does give us a rather spirited lane change when initiating a pass on the left. Personally, I liked it.
We have owned an MS with AP! for 1.5 years and just traded it for an MX with AP2 as of last Tuesday. This morning, three days after delivery and 24 miles on the clock, we hit the road for the SW. About 80 miles into the trip, AP2 become available. Images of other vehicles started to appear in the IC at about 140 miles and then showed only when in our lane. In the Willamette Valley, I5 is relatively straight and AP2 worked well, centered but favoring the left side. Initial feeling was that AP2 was close to AP1 performance.
Going into the Siskiyous, Stage Road Pass and Sexton Pass especially, I5 changes character. The turns are tighter combined with ascents and descents, large semi trucks crawling up the accents, the turns come in with no straights in between, AP2 is not ready for these conditions.
. AP2 often times starts drifting across lane lines in corners, even though we had the speed set at 65 MPH and the speed limit was 65 mph, dry payment, mid afternoon.
.When in the inside lane and coming up on a large truck in the adjacent lane entering or in a turn, AP2 nearly always, hesitates just before overtaking the truck. This can be alarming especially when some hot dog comes up behind you and all of sudden your car hesitates.
.The above hesitation can also occur when entering a turn even though no other vehicle is nearby. Phantoms.
We traveled this route a number of times when had our MS and, while it would stray some with AP1, it is obvious that AP2 will require much more development.
Would I go back to AP1, no. AP2 has the potential and I look forward to contributing to its development through our purchase of the MX and whatever feedback Tesla receives from the anomalies that we experience while driving in AP2. But stay awake and keep hands on the wheel.
Interesting I started noticing these exact improvements after a time with 17.36 I'm awaiting 17.38
Took delivery of my MS today. Received the 2017.38 update immediately. Release notes: supercharger power as stated above, but also Auto High Beam. Maybe this was included in previous update for others.
Can't comment on AP improvements as the donuts at Tesla hadn't enabled it as they were supposed to and new vehicle in any event.
Can't comment on AP improvements as the donuts at Tesla hadn't enabled it as they were supposed to and new vehicle in any event.
Yeah both of those were in previous. Did it really not say the usual comment of ‘here are previous release notes blah blah’?
Tjhappel
Member
Received this evening, drove about 30 miles on freeway and I can say it was WAY better in the carpool lane which before for me was almost unusable. Much less ping ponging when the lanes widen or become faded. I did some driving in the middle lanes and overall it seemed much more confident in general, noticeable improvement with the lane on hud screen not jumping around as much, Stop and go will be interesting to see if it’s calmed down there also.
Good steady improvement, (coming from 17.34), seems like we are seeing light at the end of the tunnel with ap2. At least parity is here or very very near. I really can’t wait until they start to use more than the front camera!
Good steady improvement, (coming from 17.34), seems like we are seeing light at the end of the tunnel with ap2. At least parity is here or very very near. I really can’t wait until they start to use more than the front camera!
RE post #34, 2017.38 f87c64d5, Second day with this release. Total milage, 807. Maybe 80% with AP2 engaged coming down from Mt Shasta City to Coalinga on I-5. Similar experience as yesterday, car doesn't like overtaking trucks especially on right hand turns. There are plenty of turns when coming down the mountains into the Sacramento Valley to observe this car's manners.
. Takes good care of itself as long as traffic is light and relatively free of large trucks.
. When overtaking a truck, whether in a corner, approaching a corner or on a straight away at about 150 feet distant, if the truck tends toward the lane boundary, the car is very likely to hesitate as it tries to decide whether the truck is going to cross into the cars lane(my opinion). This deduction comes from observing the lane lines as depicted in the IC along with the image of the truck as opposed to the where the lane line and the truck actually are. In the IC, the truck appears to be touching or on the lane line while it may actually be a foot or less from the line. Note that image of the truck may not always be shown but does appear often enough to suggest there are two systems working together.
. When engaging AP2, TACC engages automatically. At my wife's urging, I turned off AP2 and we observed that the car still acted similarly when we chanced to overtake a large truck that was straying toward the lane boundary.
. At about 150 ft distant, there maybe some resolution problem when it comes to interpreting what is going on in an almost horizontal plane. Especially with two independently moving objects on that plane. I suppose information from at least two sources are being merged. Further, the lane lines are a projection which would become less accurate with distance.
. Another issue that probably related to the lane projection problem is manifest in how the car always seems late in understanding that it is entering a turn and consequently swings wide in turns. At what point does the car's system notice that the lane line is not straight? At one point today, the car drifted completely over the lane line at about 65 mph. We were in the center lane with no one around so I left the car in AP2 to see if would recover and move itself back to the center of the lane which it did do.
. Ping-ponging has been often mentioned. I do not recall experiencing this activity with our MS, AP1, but certainly have seen it in our MX on this trip. Again, the car must be hunting and correcting until it finds itself where it thinks it ought to be. It is evident in irregular striping and ramp areas. I also wonder if bad pavement such as the outside lane of I-5 south of Stockton can cause hunting/ping-ponging by shaking the car and seemingly presenting a shaky world view to the cameras.
. I do not know why the car sometimes just does not seem to recognize that the lane is turning and it keeps going straight. This happened a number of times over the last two days, all on I-5, lane lines present, dry pavement, plenty of daylight. This morning, I initially thought that the low angle of the sun might be part of the problem but by this afternoon I was pretty confident that sun angle has little to do with it.
. The only time images of other vehicles on the road show up is when they are in the car's lane except for when a vehicle in an adjacent lane is roughly 150 ft ahead and very close to the lane line. Vehicle images nearer but in the adjacent lanes do not appear in the IC untill they cut into the car's lane. Not just hang about near by. The sonar sensors will pick them up as fan lights. It would be interesting to know how the awareness systems interact and what the limits of resolution are.
. Takes good care of itself as long as traffic is light and relatively free of large trucks.
. When overtaking a truck, whether in a corner, approaching a corner or on a straight away at about 150 feet distant, if the truck tends toward the lane boundary, the car is very likely to hesitate as it tries to decide whether the truck is going to cross into the cars lane(my opinion). This deduction comes from observing the lane lines as depicted in the IC along with the image of the truck as opposed to the where the lane line and the truck actually are. In the IC, the truck appears to be touching or on the lane line while it may actually be a foot or less from the line. Note that image of the truck may not always be shown but does appear often enough to suggest there are two systems working together.
. When engaging AP2, TACC engages automatically. At my wife's urging, I turned off AP2 and we observed that the car still acted similarly when we chanced to overtake a large truck that was straying toward the lane boundary.
. At about 150 ft distant, there maybe some resolution problem when it comes to interpreting what is going on in an almost horizontal plane. Especially with two independently moving objects on that plane. I suppose information from at least two sources are being merged. Further, the lane lines are a projection which would become less accurate with distance.
. Another issue that probably related to the lane projection problem is manifest in how the car always seems late in understanding that it is entering a turn and consequently swings wide in turns. At what point does the car's system notice that the lane line is not straight? At one point today, the car drifted completely over the lane line at about 65 mph. We were in the center lane with no one around so I left the car in AP2 to see if would recover and move itself back to the center of the lane which it did do.
. Ping-ponging has been often mentioned. I do not recall experiencing this activity with our MS, AP1, but certainly have seen it in our MX on this trip. Again, the car must be hunting and correcting until it finds itself where it thinks it ought to be. It is evident in irregular striping and ramp areas. I also wonder if bad pavement such as the outside lane of I-5 south of Stockton can cause hunting/ping-ponging by shaking the car and seemingly presenting a shaky world view to the cameras.
. I do not know why the car sometimes just does not seem to recognize that the lane is turning and it keeps going straight. This happened a number of times over the last two days, all on I-5, lane lines present, dry pavement, plenty of daylight. This morning, I initially thought that the low angle of the sun might be part of the problem but by this afternoon I was pretty confident that sun angle has little to do with it.
. The only time images of other vehicles on the road show up is when they are in the car's lane except for when a vehicle in an adjacent lane is roughly 150 ft ahead and very close to the lane line. Vehicle images nearer but in the adjacent lanes do not appear in the IC untill they cut into the car's lane. Not just hang about near by. The sonar sensors will pick them up as fan lights. It would be interesting to know how the awareness systems interact and what the limits of resolution are.
Got around 200 km on 2017.38 f87 now. Improvment.
Dancing lane lines gone!
Does hunt a bit if it has troubles finding lane markers, should lock on car in front instead.
Now dead centre in lane where it previously had trouble.
Still troubles over crests and in bottom of bumps but not to far from AP1 in this regard. I do feel its vision has too short range atm.
On my test road performs as Ap1.
Dancing lane lines gone!
Does hunt a bit if it has troubles finding lane markers, should lock on car in front instead.
Now dead centre in lane where it previously had trouble.
Still troubles over crests and in bottom of bumps but not to far from AP1 in this regard. I do feel its vision has too short range atm.
On my test road performs as Ap1.
