The worrying part if that's the case is that it's very rigid and doesn't account for temporary slowdowns deployed for construction, or even new speed limits that might be permanent but not in a DB yet.
Musk: All modes AP by end 2019
- Thread starter malcolm
- Start date
The worrying part if that's the case is that it's very rigid and doesn't account for temporary slowdowns deployed for construction, or even new speed limits that might be permanent but not in a DB yet.
jimmy_d
Deep Learning Dork
Yeah I agree that it seems simplistic to just not use speed signs. I don't know if they are ignoring them or not. There are places where my car adjusts the speed immediately when I pass a sign and there are also places where it doesn't. My sense is that AP1 was better about matching the signs, though I did see it fail occasionally. I agree with Shrimp's assessment that they don't seem to be as good as they should be on AP2.
On the construction zone sign thing - I really want to agree with you because pretty much all of the freeway lane holding failures I see around here are in construction zones with reduced speed, narrow lanes that are curving too tight for AS to manage. If it were to automatically adjust the speed down I think it probably wouldn't lose the lane. But there's this problem that the surrounding traffic NEVER slows for construction. Typically the 65 zone is dropped to 55 or 50 but the traffic goes straight through at the same 75MPH they were going before encountering the construction zone. So I feel like I'd probably end up cancelling TACC anyway if it suddenly dropped me from 70 to 55 while the surrounding traffic was still going 70 / 75.
Slowing down is the right thing to do, but it will often piss off the driver behind you. It kind of feels like a no win situation. Hopefully with the next major network revision the lane holding will improve enough that it won't lose the lane even when things get tight but it feels kind of like, at least for driving software, that slowing down is the correct answer even though my fellow Angelenos don't seem to like it.
What date did they start selling FSD? End of 2019 is more than 3 years, longer than some people's leases. Plus regulatory delays or course.
Yeah, the lane holding right now is just not impressive in a lot of curves. I haven't had it completely cross a line yet but I've had it hit the cat's eye or dots before because it's aligning against the outside line instead of inside for whatever reason.
buttershrimp
Click my signature to Go Mad Max Mode
Take a look at the second video, on autopilot on the Tesla website when you look at the model s, there is a section or two that don’t seem Fake especially with respect to pedestrians...
jimmy_d
Deep Learning Dork
Routing and AP maps are different things, AP maps just gives the road information the Tesla maps still do the horrible ass routing.
No, I think they could release it today, it’s really quite good. I don’t know why they haven’t quite honestly.
TaoJones
Beyond Driven
I look forward to any release that brings AP2 closer to AP1 and to the mythical parity that was promised, if the June timeframe is accurate, a year and a half previously. As noted, it's not fun to spend six figures to go backward, especially when the converse was explicitly put forth as fact.
With that said, I don't see the equivalency today between AP1 and AP2 whatsoever - despite diplomatic assertions by jimmyd and the shrimpmeister to the contrary. Broken is still broken and regressed functionality likewise, in other words. And that's beyond the accepted, and I think apt, characterization of "stable" versus "nervous" between the two. If it was just the same bag of tricks with one stable and one nervous, then sure you'd have equivalency - but that is not the case at all today.
As just one example, not being able to react correctly to, let alone recognize speed limit signs correctly, is a significant problem when AS decisions are then made for you based upon those database errors. The database being AP2's approach, while AP1 (MobilEye) simply read the signs correctly 99.x% of the time. As noted previously, I believe the correct answer is a combination of both approaches. Read the sign. If there's a big truck in the way, rely upon the database. Let's call that AP3 as a solution and it would be better than either AP1 or AP2 is today. The AP2 way today may seem benign to those unaware, until a) the wrong speed limits are used by AS (which happens frequently here in SoCal) and then b) all of a sudden you're either speeding or driving under the limit. Either case is less safe than what AP1 offered consistently and correctly. For that to have been the case for a year and almost a half now is... somewhere between mildly disturbing and appalling.
Anyway, again, if and when AP2 actually meets (dare I say exceeds) AP1, it will be a Great Day in the Neighborhood.
Tangentially, having had a chance to drive a bit more with 18.6.1, close quarters scenarios are indeed smoother, which is nice. Less stuttering when approaching stopped traffic - that sort of thing.
Lastly, I have to laugh at the occasional "but but but AP is not to be used except upon divided highways." As if that excuses the poor performance of AP2 relative to AP1, or of the state of AP in general - which by the way (as at least we know) is still is head and shoulders above anything any other manufacturer has on the road today that a person can actually buy (see SuperHumorousCruise, for example).
For you see, at least here in TrafficLand aka SoCal, we *use* *our* highways. We even have stopped traffic on them with dystopian regularity. Why, some divided highways even have... traffic control devices. And what a picnic those are at around 6pm on a daily basis.
Point being that when viewed from a certain perspective, a lot of this is just noise. The delay, versus what was promised, regardless of version, is now at least as large a problem as feature discussions. Which is why this pending version is so potentially exciting.
I'm a little surprised, however, not that Elektrek is never wrong, but we've been led to believe that the only new feature for AP2 is going to be, within the context of this release, adjacent vehicle icons in the IC - something AP1 has had for years. Is there anything *new* beyond that, or is this release simply and finally a validation of AP2 versus AP1 for *existing AP2 functionality*?
In other words, no additional cameras "activated" (I hesitate to type enabled), no speed limit sign reaction *or* recognition, and no traffic control device reaction or recognition.
It is that last feature - stop sign reaction, such as what we were led to believe was to be in hand by the end of December 2016 or "soon" thereafter (see video), first mentioned in October 2014, that I've patiently (*chuckle*) been waiting for. While clearly FSD is much more than that single feature, call me crazy but I think that single feature will not only demonstrably increase safety but will send the share price soaring.
Last edited:
Brando
Active Member
That is fine for buyers today (I did NOT buy it on our 12-22-17 X75D) but because of promises and videos and Tesla CEO set expectations I DID buy (pre-pay for FSD) on my early '17 S100D. Hard learned lesson.[/QUOTE]
oops, sorry - color me clueless
hill
Active Member
I sure miss AP1's being able to Auto change lanes on th Tahoe Lanee local streets.
.
.
buttershrimp
Click my signature to Go Mad Max Mode
Well, I definitely think you can say that I have made those assertions, but I think jimmy provided a "smart as balls" explanation as to why there are differences... heck it was pretty damn unbiased too. Understanding the real challenges is rewarding for me even though you are right that regressed functionality is regressed functionality at the end of the day. The problem I have is that AP2 is better in certain ways, especially on roads I travel in Austin.
But this last weekend for me was jolt of reality of how two particular items can negatively affect an autopilot experience.... the first is speed limit reading and the second is inability to execute a lane change at a given time... the third, which is almost acceptable is the ability to stay centered or hug the interior of the turn.
@TaoJones, one thing you absolutely captured in your post, is how annoying it is when driving along a highway, passing through a mislabeled data provided speed limit can cause your car to change speeds is problematic at this stage. I thought the phantom speed changes were annoying but I understood it was a thing they were working on when it occurred several months ago.... but several months later, especially with my recent re-experience with AP1 this weekend, I think pressure is building and this feature and lane change limitations is really limiting the use of AP2. And I use the heck out of it. Also, keep this in mind, AP2 took me farther on the road than AP1 could in a variety of scenarios.... the problem? AP1 felt more stable because of these two simple features. We can talk about AP2 lagging in other areas, but ask yourself this, if this situation were reversed, AP2 could do the speed limit and on-demand lane changes, how many folks would think the lane keeping is worse in AP1? Both issues need to be addressed quickly with AP2 in my opinion, and I think they probably will soon (gulp, I hope)
Auto windshield wipers now work awesome in all forms of precipitation for me so clearly Tesla knows how to totally reimagine industry standards but holy God has Ap2 been a real struggle and Tesla's silence has been infuriating. I have no idea what is going on behind the scenes from one update to another. Some can hold lanes others are dangerous. I was excited with .5 existing not silence now for 5 weeks is making me thing something is not quite right despite @BigD0g assuredness.
hill
Active Member
it's likely others feel the same satisfaction with the rain sensing wiper feature.
my dad was satisfied with his rain sensing wipers too, on his cadillac .... over 20 years ago. I was satisfied with the feature on our (⅓ the Tesla cost) Lexus, about 13 years ago. Just sayin'
it's likely others feel the same satisfaction with the rain sensing wiper feature.
my dad was satisfied with his rain sensing wipers too, on his cadillac .... over 20 years ago. I was satisfied with the feature on our Lexus, about 13 years ago. Just sayin'
Did those wipers also have a mind of their own or were they just plain jane vanilla sensors with no actual ability to perceive and make decisions? Yea, I'm just sayin'...
hill
Active Member
I'm not sure about my dads caddy, but the lexus? Yes ... super light mist would just get an occasional wipe, & the heavier thw rain got - the more frequent & faster the wipers would go. Nothing wrong with being impressed - but it's old Tech. Just sayin
The result is not necessarily the accomplishment.
People had been crossing the Atlantic before Lindbergh or the Concord.
Canuck
Well-Known Member
It's kind of sad when we praise the wipers as an accomplishment.
Similar threads
