FSD v12.x (end to end AI)

[FSDtester#1]

Thanks. But then that makes me wonder why it did, in a couple of places.

Those are where Oscar the Grouch and family live. VRU's...

 

[SidetrackedSue]

Thanks. But then that makes me wonder why it did, in a couple of places.

Any chance it was highways and V11 was in control? Or going really slowly and the parking program was doing the visualizations (I don't use that so don't know if it draws such things while searching for a parking spot.)

 

[enemji]

TESLA PANIC MODE

Have you seen one? I experienced one today. I was on the tollway at rush hour, on the second lane from right. Up ahead was my exit, and the traffic was crawling ~15mph. In a quarter mile I was expecting FSD to switch to the rightmost lane to take the exit. I was just staring ahead when the steering started twitching left to right, and right to left 5 degrees. I was shocked. Then I looked at the right lane(the exit lane) and there were two cars that were literally fighting to be the one ahead in the lane side by side in the same lane. As soon as they figured out their sh$t, the panic attack stopped.

Holy Moly!

 

[Tronguy]

TESLA PANIC MODE

Have you seen one? I experienced one today. I was on the tollway at rush hour, on the second lane from right. Up ahead was my exit, and the traffic was crawling ~15mph. In a quarter mile I was expecting FSD to switch to the rightmost lane to take the exit. I was just staring ahead when the steering started twitching left to right, and right to left 5 degrees. I was shocked. Then I looked at the right lane(the exit lane) and there were two cars that were literally fighting to be the one ahead in the lane side by side in the same lane. As soon as they figured out their sh$t, the panic attack stopped.

Holy Moly!

I've had a number, probably less than four, over the past few weeks on 12.x. Usually it's for no discernible reason, but I'll get the "Take Over Now!" with the red hands on the display. No strikes; just FSD losing its way.

I do think it tends to happen more often at toll booths; but I've gone through more toll booths where it doesn't do that than with ones that do, and the same ones, even. I've also seen it happen in traffic with other cars about.. but nobody else doing anything weird.

Figure that these are the kind of clips that get sent off to Tesla, but have no idea. Don't think I've been prompted for an explanation.

 

[AlanSubie4Life]

Flips out going through flashing red light intersections. Push the accelerator to get through. Red wheel. Stays engaged. Like it has no idea what is going on. Annoying. Feel like it may not have done so on 12.3, but don’t remember.

 

[BBTX]

So, I think the "head fake" maneuver for lane changes has been documented already: sometimes when it wants to change lanes in front of an oncoming car that's getting close, it noses over the line just a little and right back, apparently to gauge their reaction. If they back off, it goes and head and does a full lane change. If they don't, it waits for them to pass before trying again. It's a pretty slick move, really. Today I saw an odd variant of this, where it apparently head-faked 3 times in a row fairly rapidly, but never quite got the reaction it wanted. Then it let the other car go by and finished the lane change. As it was happening, I initially thought it was something more like the twitchy steering wheel panic sort of thing happening because it felt so odd, but it really was just a triple-head-fake, then lane change.

 

[AlanSubie4Life]

It's a pretty slick move, really.

As described, no.

 

[Twiglett]

First commute with 12.3.4 was pretty underwhelming.
It was so far over the left side lane marker that we were bouncing over the cats eyes. One time it was bouncing on the center lane markers with oncoming traffic, very concerning. Commute traffic meant that speed limits were never a concern, but lane select was appalling. It would move into the left lane only to come to a complete stop, while then middle lane we left continued to flow freely. General lane discipline and staying anywhere near central is pretty poor, much worse than 11.4.9. That is saying something considering the fetish that 11.4.9 has with the left lane.
I lost count of the number of times it signaled to move into the left lane while there was a car there already and it showed the flashing red warning. Other times it would signal with plenty of space then waffle on committing to making the move.
Can't test the speed limit detection on Hwy 95 and 290 in Texas to see how that is different, but on this performance I'm not expecting much.
Gotta say that I got to drive my basic AP Model 3 this past weekend - having access to TACC and basic AP was a breath of fresh air, so much less hassle. It had me wondering whether it was worth the hassle of FSD at all.

 

[Cheburashka]

I tried the new FSD again. It's trash. IMO v11 was way better.

It drove me onto the shoulder and thought it was a lane, consistently goes over the lines on curves, and it drove in a bike lane for way too long. It might be more confident, but it's not any better.

 

[hobertg]

Ok, I have had FSD for over a year. This past week was the first time this has ever happened and I had it happen again today and wanted to ask about it.

I was on a straight two lane road, no other cars are close and I get a warning noise for a 4-5 seconds. A little panic from me for a second like "what was going on" and then it just stops.
I get the warning noise with no visual indicators (no blue flashing, no "take over now w/ hands on the wheel", no red car in front of mine for being too close, etc) and wanted to ask if anyone else has had this happen on 12.3.4??

Very strange as I am always watching the road and car behavior closely and every warning is obvious and generally makes sense with it being overly cautious. This recent event now happening twice with no reason for a warning noise is VERY unsettling in the moment with no indicators.

Thoughts?

 

[OxBrew]

Ok, I have had FSD for over a year. This past week was the first time this has ever happened and I had it happen again today and wanted to ask about it.

I was on a straight two lane road, no other cars are close and I get a warning noise for a 4-5 seconds. A little panic from me for a second like "what was going on" and then it just stops.
I get the warning noise with no visual indicators (no blue flashing, no "take over now w/ hands on the wheel", no red car in front of mine for being too close, etc) and wanted to ask if anyone else has had this happen on 12.3.4??

Very strange as I am always watching the road and car behavior closely and every warning is obvious and generally makes sense with it being overly cautious. This recent event now happening twice with no reason for a warning noise is VERY unsettling in the moment with no indicators.

Thoughts?

I have this happen all the time at one particular spot. Can't decide which lane to be in at a UPL, just gives up, no warning, car coasting slow.

 

[Dewg]

"Human Reaction Time in Emergency Situations​

October 1, 2021

Reaction Time For Simple Tasks:​

Reaction time is defined simply as the time between a stimulus and a response. Human reaction time is sometimes quoted as a constant number: 0.2 seconds. While 0.2 seconds may be the average measured for simple tasks, reaction time is actually more complex.

Reaction Time For Complex Tasks:​

For more complex tasks such as emergency braking, human reaction time has been studied and measured as three different phases: the time to perceive or sense a danger or hazard (perception phase), the time to make a response decision (decision phase), and the time to respond (response phase). The response phase (i.e. braking) is further complicated by the physical response (i.e. apply the brakes with the foot) and the system response (i.e. the time the vehicle’s braking system requires to actually apply braking force to the wheels). Under ideal driving conditions, the entire human perception reaction time for braking has been measured to be approximately 1.5 seconds (R. Limpert)."

Note that is ideal conditions, it can be worse if you are distracted, tired, drunk, hungry, dehydrated, etc

I had no idea humans were that slow. 1.5 seconds is an eternity.

 

[AlanSubie4Life]

I had no idea humans were that slow. 1.5 seconds is an eternity.

Often humans are braking before any braking occurs (negative reaction time).

You’re thinking 1.5 seconds is reaction time somehow (what is being discussed).

Fortunately the perception cycle which may last 1.5 seconds often starts well before anything happens, so in many situations this leads to negative reaction time.

In worst-case situations of sudden hazards the 1.5 seconds might happen.

Compare to FSD which seems to have extremely limited ability to perceive and just has reaction, which seems to be clocked currently at 0.35s.

Check the possum video, for example (frame where timing began linked). I did not time that but in my bleary eyed (I have night vision problems) and distracted state you can see my reaction time was similar to FSD. It took 1 second for me to react (watch the brake pedal) in my compromised vision-impaired state, which is coincidentally the same as FSD (observe regen bar - note the path planner did not fully reflect the braking of the car in this case). (I would beat it handily in the daytime, assuming I saw the hazard.) And that is kind of a “sudden hazard” case.

My pedestrian case seems to have made the Dan O'Downer highlight reel. A dubious honor. I'm sad that he did not feature the Awesome Possum.

 

[Ben W]

So what's an accident that can be avoided by that eternally vigilant computer with 50 ms reaction times, but not 500 ms reaction times? (Assuming that FSD is that delayed)

A kid running out into the street from behind a parked car.

 

[AlanSubie4Life]

A kid running out into the street from behind a parked car.

I would argue than in about 60%-80% (arbitrary) of those cases, humans would see the child's feet from a great distance away (has to be for the angles to work), before the child emerges. And then would ensure there is no close call.

That long-range incredibly capable perception is something FSD would struggle to do currently and would have to make up for with superhuman reaction time. (And that superhuman reaction time would help with the edge cases not included in the 60%-80% or whatever it is.)

 

[mtndrew1]

My FSD test route on 12.3.4 continued to trend slightly better than v11 did, though last Friday’s test was during Coachella weekend which is like the rapture in Los Angeles. Traffic was light because half the population was migrating toward the desert.

Anyway, still doesn‘t pass the spouse test, still drifts over lane lines and Bott‘s dots, still sets off its own FCW alerts, yada yada yada.

 

[djon]

12.3.4 hit a curb today. Wide lane narrows at a bridge. Older FSDs would get close to this curb, but never hit it.

 

[swedge]

The v11 highway stack isn't that great. I had a scenario where fsd was changing into another lane when another car from the other side was getting into the same lane and if I hadn't disengaged, my car would have hit the other car which was another Tesla. Not sure if the other Tesla was using fsd or not.

In a similar situation, for me FSD saw the other car moving into the lane we were moving into, and swerved back into it's original lane. In your case, if the other guy was also on FSD, I bet FSD would have steered your both away, back to where you started. But trying that is something for a test track.... I suspect Tesla simulates such situations in their training data, rather than relying on us to supply enough real world examples. There are plenty of NHTSA reports of serious accident on which to base simulations for teaching FSD how to survive them.

I am not recommending to wait to see what will happen! Sometimes the car reacts before we do, and we get to see that scary is not exactly the same thing as unsafe.

On the other hand, ff you two had collided, and both Tesla's were on FSD, you could have wound up on the news! Sort of like all those Teslas crashing into each other in that silly SciFi movie where everything even vaguely imaginable goes wrong at once.

 

[Dewg]

Often humans are braking before any braking occurs (negative reaction time).

You’re thinking 1.5 seconds is reaction time somehow (what is being discussed).

Fortunately the perception cycle which may last 1.5 seconds often starts well before anything happens, so in many situations this leads to negative reaction time.

In worst-case situations of sudden hazards the 1.5 seconds might happen.

Compare to FSD which seems to have extremely limited ability to perceive and just has reaction, which seems to be clocked currently at 0.35s.

Check the possum video, for example (frame where timing began linked). I did not time that but in my bleary eyed (I have night vision problems) and distracted state you can see my reaction time was similar to FSD. It took 1 second for me to react (watch the brake pedal) in my compromised vision-impaired state, which is coincidentally the same as FSD (observe regen bar - note the path planner did not fully reflect the braking of the car in this case). (I would beat it handily in the daytime, assuming I saw the hazard.) And that is kind of a “sudden hazard” case.

My pedestrian case seems to have made the Dan O'Downer highlight reel. A dubious honor. I'm sad that he did not feature the Awesome Possum.

Ah, that's much better. I knew it wasn't 1.5 seconds.

 

[David Wang]

In a similar situation, for me FSD saw the other car moving into the lane we were moving into, and swerved back into it's original lane. In your case, if the other guy was also on FSD, I bet FSD would have steered your both away, back to where you started. But trying that is something for a test track.... I suspect Tesla simulates such situations in their training data, rather than relying on us to supply enough real world examples. There are plenty of NHTSA reports of serious accident on which to base simulations for teaching FSD how to survive them.

This was on the freeway and my understanding is that v11 is still rules based and not using any training video?