-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
- Status
Personally I would accept a human inserting the charging cable after the car has self-driven into a stall as no controls touched.
It will be interesting to see if they can do it. I mean others have already done the 99% highway portion autonomously. Could Tesla's (with nVidia's help) FSD codebase extend to doing the rest reliably enough for a controlled (though uncontrolled) test...
It does not seem impossible to me, as Tesla could map and teach the car every relevant area in detail. Does not tell too much of real world performance, though.
Did Tesla ever complete that Roadster 3.0 roadtrip, by the way?
Even though I believe FSD with AP2 is possible in good weather and especially on a controlled geographical route, in honesty people have let so many past Tesla promises slide that the scepticim seems warranted how they would handle this one.
For example, would it be OK if Tesla does this on the 2.5 suite in Model 3 instead of 2.0...?
Many a Tesla promise has been moved to the next-generation product as they learn after the fact the limits of their current products...
On this note, the second to last update tried to do this for me (posted earlied on this forim) - have not repeated the conditions on the last:
"Had the first "it tried to ruin my car and maybe hurt someone" moment with the AP2 (17.26.76) today. Roadworks caused a freeway to turn and narrow a bit, with a solid white line on both sides of the lane, adjoined by another lane on the left (and beyond that a highway divider, behind which the opposing lanes) and by concrete blocks on the right. No problem with this for AP2, as the markings were clear and the weather was nice enough. Really, the road was clearly marked, the only problem was slightly narrower lanes and no room for error on the sides, so this was not the "road markings lead to a concrete divider" scenario at all.
Suddenly the road has a sort of diagonal, but partial (not fully across the lane) dark line going across the lane, covered in pitch/resin. Probably some remnant of old road markings or covering some cavity on the road or temporary changes made to the road during the work there. The actual roadmarkings on the side continue as usual in bright white, while this diagonal line twists/turns to one side, but again is dark and not a road marking by any standard. There are straight, solid, bright white lines marking the road on both sides of the lane. But this dark line does end under the left-side road marking, so it sort of starts extending from that.
It is sufficient for me to stiffen up expecting the worst - half thinking to myself I've spent too much time on TMC and am now dreading the unnecessary. Until now the only real problems with AP2 had been ghost brakes at overpasses (one bad, many unpleasant) and terrible undivided road performance. It had been pretty good on divided roads (which I was on), as far as following lane markings in good weather go...
But sure enough, AP2 starts turning alongside this sort of diagonal line! Straight towards those concrete blocks lining the lane on the right side! I mean - it stopped following beautiful, solid while lines on both sides of the lane and chose to start following a dark, partial twist in the middle of the lane into concrete obstacles. An absolutely insane decision from the EAP that just shows how limited its ability at the moment is... even if those road markings had not made sense, there was a huge wall of concrete right next to the lane it wanted to turn into, instead of that beautiful free air straight ahead.
It took a very strong twist of the already turning steering wheel on my part to cancel this movement. Everyone else in the car was stratled, though personally I was ready and was reacting, so it was more like adrenaline kicking in on my part.
Later on on the trip it got very confused again when overtaking and wanted to start following a car on the adjacent lane. At that point I just turned it off and drove the rest of the way. It stays off for now.
Other than that, staying within lane was slightly less ping-pongy for me. The overall feel is more aggressive than previous versions, I would not call it silky smooth."
You're speaking from watching YouTube videos, that have been posted for a reason.
I'm speaking from having driven hundreds (more likely thousands) of miles with autopilot.
I agree with the YouTube videos. Many times it performs admirably. However it utterly fails often enough for it to be barely useful.
Disagree, AP1 remains superior to AP2 in lane-keeping as well as being aware of cars in adjacent lanes.
Being aware? Or showing it on the UI?
In terms of cars next to me I can't imagine anything being worse than AP1. The side monitoring system with ultrasonics is pretty bad. I expect that fairly quickly will see AP2 far surpass AP1 in side monitoring as soon as a Tesla utilizes the side cameras.
One area I am a bit concerned with regarding AP2 is the speed of the IC. I can't be the only one that notices how delayed the response is between when an event happens versus when the IC shows it. I'm not aware of any improvements of anything IC related with HW2.
Well, AP2 is.
It does not employ its side cameras and it is not as good at keeping lane and using its ultrasonics...
Yet, it's going to keep getting better. They've had a year+ with the AP1 hardware, so I imagine this time next year it will be very solid, but there won't be any coast to coast FSD drive in December, that's just not happening IMHO.
NerdUno
Member
If you look at how (not) far EAP has come with AP2 in the last 10 months, it shouldn't take a rocket scientist like Elon to figure out we probably aren't going to see FSD magically appear in the remaining 5 months of this year. The cars are having a difficult enough time following a dotted line much less dealing with multiple cameras, maps, street signs, traffic signals, pedestrians, bicyclists, rain, snow, and 18-wheelers.
Disagree on the ultrasonics also @S4WRXTTCS - if anything my AP2 picks them up faster than my AP1. Seriously - is there anybody else on the forum who owns both AP1 and AP2 that I can have a dialogue with to compare direct recent notes on driving both cars with the latest firmware - not just youtube watchers and guys who own only one vs the other?
This is disappointing. The trust many people put in AP1 off highway was very ill-informed. Now it seems to have flipped to a fundamental lack of trust for many.
The upside is probably fewer decapitations while using AP. Few things ruin the day more than a relaxing drive ending in decapitation.
boonedocks
MS LR Blk/Blk 19”
I wonder if Elon's eggs on his face will be raw or scrambled.
It hasn't come far in 10 months? Agree it is probably not doing NY-LA this year - but what do you mean with the first part of your sentence?
Your analysis of AP2 does not seem accurate to many, nevermind AP1.
Perhaps you live in an area where AP2 works better than average.
Well, if there is a separate FSD code branch and Tesla teaches and maps the LA-NY route to that FSD branch, I could see them getting it done in good weather... as a demo.
Hmm - maybe I haven't seen it because I haven't been in rush hour slow with the AP2 - going into chicago in a few days - i'll get a dash cam and report back on what happens.
Well if they go down that route, that's kinda / sorta cheating, as Elon has said in calls / presentations since then it doesn't really have to be LA -> NY, he said the could say dynically change it along the way and decide to go from LA -> TX let's say. If they preprogram it for a route and conditions, it's going to be the same B.S. marketing ploy that the FSD on the website is about 265 disconnects.
I think all the major car automation projects would have a chance going coast to coast on a preplanned route. The question is whether Elon will announce in advance. Is this a one shot effort, or is he going to keep running cars until one is successful?
Regardless of the conditions, this will be a demo. Self drive demos aren't particularly meaningful. Same with AI demos. Doing this runs takes significant resources away from the effort to improve AP.
A coast to coast run will be good publicity, but most of his customer base won't care. He should get credit for what he delivers to the cars of AP user, nothing more or less
- Status
Similar threads
