jeffro01
Active Member
Please don't use the word "fanboys". It immediately causes members to dismiss your comment as insulting and childish regardless of the quality or validity of your point...
Jeff
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Please don't use the word "fanboys". It immediately causes members to dismiss your comment as insulting and childish regardless of the quality or validity of your point...
Jeff
CameronB
Member
I had commented earlier that I had not yet travelled in the right hand lane to see how the AP2 on my car would/wouldn't handle off-ramps properly. This morning I travelled a new (for my car under AP) highway and stayed for almost the entire 20 minute trip in the right lane to see how AP handled the cars merging onto the highway and whether AP would get confused or 'dive' for the off-ramps.
Off-ramps when driving in far right lane:
Passed a total of 8 off ramps. Car remained on highway lane all 8 times and didn't dive (or even meander) towards the offramp. Perfect!
Better yet, in a few of those cases, just before the offramp approached, the lane I was in was sub-divided. By that, I mean my single right hand lane turned into an extra wide section briefly and then there was both a keep-straight-to-stay-on-the-highway lane and a dedicated turn lane. In those cases, the car was unphased and smoothly moved over to the left side of the now-splitting lane and remained on the highway. It didn't "fall for" the long 'exit only' lane even though that new exit only lane was technically still the far right lane. It did the right thing just as a human would do. Very impressive! A+
Handling cars coming onto the highway:
The first two times that cars were coming onto the highway, AP2 seemed to handle things OK and I didn't have to disengage nor did the merging drivers have to do anything special (no hard breaking or accelerating or anything out of the ordinary). Worked smoothly. However, on the third merge, the onramp was a very long one and the merging car was driving ahead of me by about 200 feet. He was just about at the end of his long ramp but my car didn't seem to see him about to enter my path and was still driving at maximum (set) speed and was racing quickly towards the "intersection point". This led to my first (and so far only) highway disengagement as our cars were getting quite close as the onramp was running out of room and my car was on a direct intercept course at the transition point and I didn't get the sense that AP was slowing down at all. So I hit the brakes and disengaged. Would AP have slammed on the brakes a second later as the on ramping car fully went into my lane? I'd guess yes...but it would have had to brake really hard since I was going 72 and there was precious little room. It would have been very uncomfortable at best - and dangerous at worse. Seems the car in this case had "blinders" on and was just seeing my lane and not using "peripheral vision" to see what was happening to my right with the soon-to-merge car. Mixed results. I'd give it a C- so far because of the above issue.
Off-ramps when driving in far right lane:
Passed a total of 8 off ramps. Car remained on highway lane all 8 times and didn't dive (or even meander) towards the offramp. Perfect!
Better yet, in a few of those cases, just before the offramp approached, the lane I was in was sub-divided. By that, I mean my single right hand lane turned into an extra wide section briefly and then there was both a keep-straight-to-stay-on-the-highway lane and a dedicated turn lane. In those cases, the car was unphased and smoothly moved over to the left side of the now-splitting lane and remained on the highway. It didn't "fall for" the long 'exit only' lane even though that new exit only lane was technically still the far right lane. It did the right thing just as a human would do. Very impressive! A+
Handling cars coming onto the highway:
The first two times that cars were coming onto the highway, AP2 seemed to handle things OK and I didn't have to disengage nor did the merging drivers have to do anything special (no hard breaking or accelerating or anything out of the ordinary). Worked smoothly. However, on the third merge, the onramp was a very long one and the merging car was driving ahead of me by about 200 feet. He was just about at the end of his long ramp but my car didn't seem to see him about to enter my path and was still driving at maximum (set) speed and was racing quickly towards the "intersection point". This led to my first (and so far only) highway disengagement as our cars were getting quite close as the onramp was running out of room and my car was on a direct intercept course at the transition point and I didn't get the sense that AP was slowing down at all. So I hit the brakes and disengaged. Would AP have slammed on the brakes a second later as the on ramping car fully went into my lane? I'd guess yes...but it would have had to brake really hard since I was going 72 and there was precious little room. It would have been very uncomfortable at best - and dangerous at worse. Seems the car in this case had "blinders" on and was just seeing my lane and not using "peripheral vision" to see what was happening to my right with the soon-to-merge car. Mixed results. I'd give it a C- so far because of the above issue.
As others have noted, YMMV. A lot of the rest of us are equally mystified about statements like yours. I've driven hundreds of miles on AP2 on long trips without any disengagements at all. It's not absolutely perfect 100% of the time(as an L2 system, I don't expect it to be), but I've found it to be extremely good the vast majority of the time.
As a former CA resident and a motorcyclist, I always found the lane splitting thing to be both incredibly dangerous and even more incredibly obnoxious. Knowing how much Elon hates CA traffic, maybe he programmed AP to discourage lane splitting
⚡️ELECTROMAN⚡️
Village Idiot
Is 17.17.17 the silky version? Is it worse than the previous version? Should we not install it until further notice? Although full speed emergency braking might be more important than having good autopilot. I don't really find autopilot very useful, but I do use it just to see how it's progressing. I think it works pretty good but I don't trust it very much.
I have 17.17.17 also (I believe this is one version behind silky).
I think AP is impressive in general but I am surprised it has difficulty on what I would have thought would be very easy roads for it.
When I leave my house I typically have a long drive down a simple two-lane road. Very wide, well marked. One lane each direction. Should be simple, but my Model S wants to drive much closer to the middle of the road than in the middle of the lane. At some points it has touched the yellow line. There is plenty of asphalt. I don't know why it wants to ride so close to the middle. People on the other side of the road will swerve out of the way to avoid me. I don't want to drive like a bully, I want to be a courteous driver.
So I don't use it normally. I do try to train it but I do not know if I am actually helping.
Some roads it seems to handle fine. I am not sure why it has so much difficulty on this road.
I would think this road would be very easy. (Interestingly, I think it may handle the other side of the road better but still not great.)
So I wonder why it doesn't do a better job. Other bad training? Incorrect map data? Maybe the road is too wide and it hugs the left line.
One possible contributing factor is that for most of the road the center marking is wider than just the double-yellow paint: it has a rumble strip that itself is painted as wide yellow stripe plus another yellow stripe on either side. I am pretty sure AP sees the lane as the space between the center of each divider, not the unpainted asphalt. I am not sure if that is correct or not, but people usually seem to drive in the middle of the unpainted asphalt.
I also wonder if my trying to train it is helping (turning it on, applying pressure in the direction toward the center of the lane until it breaks).
I have heard stories of individual cars learning specific roads through training, but I don't know if this is actually true or just applied to AP1 or what.
*fingers crossed* hoping silky is better.
I think AP is impressive in general but I am surprised it has difficulty on what I would have thought would be very easy roads for it.
When I leave my house I typically have a long drive down a simple two-lane road. Very wide, well marked. One lane each direction. Should be simple, but my Model S wants to drive much closer to the middle of the road than in the middle of the lane. At some points it has touched the yellow line. There is plenty of asphalt. I don't know why it wants to ride so close to the middle. People on the other side of the road will swerve out of the way to avoid me. I don't want to drive like a bully, I want to be a courteous driver.
So I don't use it normally. I do try to train it but I do not know if I am actually helping.
Some roads it seems to handle fine. I am not sure why it has so much difficulty on this road.
I would think this road would be very easy. (Interestingly, I think it may handle the other side of the road better but still not great.)
So I wonder why it doesn't do a better job. Other bad training? Incorrect map data? Maybe the road is too wide and it hugs the left line.
One possible contributing factor is that for most of the road the center marking is wider than just the double-yellow paint: it has a rumble strip that itself is painted as wide yellow stripe plus another yellow stripe on either side. I am pretty sure AP sees the lane as the space between the center of each divider, not the unpainted asphalt. I am not sure if that is correct or not, but people usually seem to drive in the middle of the unpainted asphalt.
I also wonder if my trying to train it is helping (turning it on, applying pressure in the direction toward the center of the lane until it breaks).
I have heard stories of individual cars learning specific roads through training, but I don't know if this is actually true or just applied to AP1 or what.
*fingers crossed* hoping silky is better.
Last edited:
NerdUno
Member
whitex
Well-Known Member
I think the issue is that it may work well 99% of the time. Small consolation for those in the 1% group who end up getting getting screwed. There is a reason why Tesla has all the small print how the driver needs to keep eyes on the road and hands on the wheel at all times... FSD on the other hand will not happen any time soon, because it's supposed to drive without anyone inside (ride hailing) so Tesla would have to take the blame if it screws up (and it will, see this post on how a simple auto-lights can screw up in AP2, and I think that would be much easier to get right than FSD).
PS> Having driven AP1 across the US, I will tell you another thing - it works orders of magnitude better in areas with lots of Tesla's (therefore lots of collected data and/or hand-tailored AP data).
Exactly the intelligent use of AP
Or just use it intelligently as described above.
Really shows the wide variability between drivers and their ability to adapt and implement use of tools for driving and their expectations of how the tools will work.
Indeed, some people just don't get it. It's an odd psychology thing that some people work themselves up into a lather of upset and disappointment, while different people are delighted, by the exact same functionality.
I don't think we're experiencing the same functionality. That's my point. I am not mystified because I think people are delusional. I am mystified because I see some people having great performance (which I believe that they are) and others like myself are experiencing downright dangerous performance. Blame my local road markings if you want. I'm not in a "lather of upset", but I do think it's totally justified to be disappointed when AP consistently runs me into the shoulder or an adjacent lane (without a turn signal/lane change request), or attempts to drive in the double yellow turn median, or keeps going straight when the road bends, or randomly brakes in the middle of the freeway with an open road ahead (almost getting me rear ended).
TL;DR - I'm not discounting your positive experiences, so please don't discount my negative ones simply because they're different than yours.
Stop it. You're not allowed to criticize anything on the Tesla Motors Cult Forum.
The obvious answer is you must be shorting TSLA. Or at least that's what some on TMC will claim.
3Victoria
Active Member
I have wonder whether you need the sensors checked .. are they aligned properly? Also make bug reports ... can't fix what isn't measured.
Yes, I had them look at it last service visit (i'm having sunroof issues, it only opens 50% of the time). They said all my cameras/sensors are totally fine.
And yes. I file bug reports daily.
Ulmo
Active Member
Sorry; I've been under pressure lately, having to do hourly jobs, and I've been having to commute with some of the stupidest drivers on the same roads. I have been in my Tesla whose AP* system has been trying to crash my car 3 or 4 times due to its safety features taking over, and I imagine what it would be like to all of those other stupid drivers to be using the car when it started doing that to them. While usually I like to be specific and accurate, if I had to summarize, I wouldn't want those types of drivers in this type of car at this time. If you think you are an above-average driver in terms of understanding driving and reaction times (and apply your abilities to actual driving rather than being distracted), then in that case, sure, use AP all you want, but for anyone else, it's not really ready. Even for myself, I find it disconcerting to be ping-ponged between a truck and a concrete barrier by the car thinking it knows better than I do when I'mcalready in a good portion of lane.
Although I don't have AP*, I could imagine using AP if I had it, and looking at the reservoir for a fraction of a second while in AP while the car careens into an immovable object, looking back, and barely or even not having reaction time. That's where my "don't use it" comment came from, when applied to average driving ability in that type of system.
Speaking of long detail, here is a great video about AP:
* Note to those who are confused: AP from Tesla comes in 3 levels for AP HW2 cars. I have the lowest level. Tesla has always been bad at naming it. I'm not going to put 10 minutes of mental power into coming up with perfect names today; I'm busy. I already asked for this to be fixed and contributed time into it, and their marketers already made it worse.
Last edited:
⚡️ELECTROMAN⚡️
Village Idiot
What AP do you have? You said the lowest? Is that the version where you paid $2500 for AP1, but by the time they made your car it had AP2 hardware installed? But you also say you don't have AP*. Is that because it doesn't work well enough to call autopilot?
Thanks for posting Bjørns video. My car (EAP+FSD) drives quite similar with 17.17.17. Decent to good on straight, wide roads, potentialy fatal errors on winding roads. I have some more ping-pong on the highway.
As the vid showed, a lot of those errors would be uncomfortable if a car (or bus etc) came the opposite direction.
As the vid showed, a lot of those errors would be uncomfortable if a car (or bus etc) came the opposite direction.
It simple. The whiners focus on the edges cases and the 1% where it has difficultly. The rest of us simply enjoy AP the remaining situations where it drives amazingly well.
Stop and go traffic on highways or forever straight roads is 1% of my driving. In 99% of my driving it gets troublesome.
Similar threads
