-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
EVNow
Well-Known Member
You are in the wrong profession. You should be an urban planner in charge of traffic signs
I think you're right - but then it doesn't explain why it ignores some other stop signs, which it can see visually and shows up on visualization.
EVNow
Well-Known Member
May be those are in the map and properly assigned by the car to the other road.
With my car, I've problem with one sign where the other road is joining the current road at a sharp angle (may be 20 degrees). Doesn't happen all the time.
Sorry - I meant stop signs for the car's road. There is a video running around here where the Tesla is driving up a curved road and at the top of the curve there is an intersection with a stop sign, which the Tesla sees in visualization, but totally ignores and just starts to plow through the stop. That's what made me wonder if map data didn't have that stop sign, and even though the car visualized it, it ignored it. Not sure.
AlanSubie4Life
Efficiency Obsessed Member
Might be last one before the update to 10.69.2.1.
Or maybe he’ll get a chance to do some busy traffic ULT on Monday. He has not really had a chance to test ULT with 10.69.2 yet (just did a minimal traffic test).
Lots of problems, tons of disengagements:
1) Creeping behavior does not work well (inconsistent).
2) Vehicle road positioning too far left at stop signs.
3) Runs red light.
4) Still stopping on narrow roads rather than just driving to the right and continuing (easy). At least it pulls to the right!
5) Attempts to pass cars that are not parked.
6) Poor interaction with pedestrians.
Or maybe he’ll get a chance to do some busy traffic ULT on Monday. He has not really had a chance to test ULT with 10.69.2 yet (just did a minimal traffic test).
Lots of problems, tons of disengagements:
1) Creeping behavior does not work well (inconsistent).
2) Vehicle road positioning too far left at stop signs.
3) Runs red light.
4) Still stopping on narrow roads rather than just driving to the right and continuing (easy). At least it pulls to the right!
5) Attempts to pass cars that are not parked.
6) Poor interaction with pedestrians.
Last edited:
AlanSubie4Life
Efficiency Obsessed Member
It was an easy utility drive at 45 mph with no turns, but beta seemed to visualize the environment well and stayed engaged the entire time until it switched to AP on the interstate. It would not do NOA, as expected.
AlanSubie4Life
Efficiency Obsessed Member
EVNow
Well-Known Member
My experience is that FSD usually stays engaged. But won’t engage if it’s raining.
sleepydoc
Well-Known Member
There seems to be a bit of a hysteresis with rain and adverse weather; the threshold to not engage is lower than the threshold to disengage.
Chuck posted that himself?! What I see is that he paid absolutely no attention to the road. FSD Beta may be getting better but if this is any indication of what to expect when using it - drivers are getting much worse.
AlanSubie4Life
Efficiency Obsessed Member
I mean it is very distracting in heavy rain and he was focused on other things, talking, etc.
He should have noticed, but it’s not clear to me that he didn’t, in spite of what he said.
The thing that made him go “whoa” was FSD jamming on the brakes for absolutely no reason - the car was going slower than the car cutting in front. It’s even worse when you think about how FSD had been tracking this car the whole time, and still had to suddenly react, rather than blending in a reaction.
It was a poorly executed and risky move by FSD Beta. The sort of thing that can get you rear ended in rainy wet conditions.
Look how the car just sits stationary in the middle of the road for a second or two, as though dazed, rather than just moving on. It’s bizarre, and dangerous for your rear bumper. People expect you to be moving if there is no one in front of you, and occasionally they won’t notice you aren’t!
Last edited:
But he said it was a good move by FSD Beta. Firstly he says "#FSDBeta has eyes on the back of its head" in his congratulation tweet to Elon.
Then to the followup question "Do you think beta overreacted or was the braking appropriate and timely. It’s very hard to tell from the video", Chuck says "In this case it was more timely than me. It did a good save, with the way that car cut me off at the same time my mindset was in acceleration around the stopped vehicle.."
Just watch his gaze while he's driving in the longer Youtube video, he's spending a lot of the time looking to the right - at the screen. You can see when he looks ahead and to the left, but a lot of the time he's watching the screen with his hand on his chin.
In a way I think FSD Beta DID help him here, yes it over-reacted and stopped too long, but Chuck wasn't looking ahead and the car in front of him needed to be noticed. If he wasn't watching the screen, the car to the right, and maybe the rear view mirror he would have seen it.
AlanSubie4Life
Efficiency Obsessed Member
Yes, it was very good. As commented elsewhere:
1) Creep behavior needs some work (at least twice Chuck commented he could see traffic slowing (he saw the oncoming car angle drop as the suspension compressed) due to the body language). The limit is good, but could be a little further back in this case. Just need to work on slowing the roll a little earlier rather than surging at 3-4mph. This is important for driver comfort and comfort of other drivers.
2) Crossing speeds could be increased. That would solve the "failure to go" issue on the last turn. And will also make for more comfortable, higher margin, crossings. Getting more than 2-2.5 seconds of margin is a good goal.
Making progress. That creep behavior is super important. The rest can be solved with the accelerator so less of an issue but would like to see it happen by default.
Another video, some sort of bug with regen bar at 9:09 when switching views, showing regen limited (never got tested). I've seen this for both power and regen, just randomly.
A couple of unnecessary brakes. For example 13:46 it takes two seconds to even respond to the yellow light, and jams on the brakes. Has enough regen it didn't need to respond right away, though. It jammed on and then completely released brakes.
Last edited:
Correct me if i'm misremembering, but i recall you complaining that 12mph was too slow and the car should be able to go up to 25?mph? If you get a chance, i'd love to see an example or video from you of accelerating up to 25mph and then stopping perfectly in the median for traffic on the other side. IMO that doesn't seem realistic at all.
AlanSubie4Life
Efficiency Obsessed Member
Yes, 12mph is too slow, and I said it should go faster. In the commit case, going up to 25mph seems fine. In the "waiting" case, I think 15-20mph would be fine and is completely realistic. There's no need to get up to 25mph in that case; it's possible to cross in about 5 seconds with a reasonably rapid increase to 15-20mph. (Again, right now, it's limited to about 12mph, and it attains that speed slowly, and it takes 6-7 seconds to clear the traffic lanes in that case.)
I'll see if I can get an example on my ULT, even though I don't want to stop in the median (don't really like the strategy in most cases, though Chuck's is fine due to the median width), as it's narrower. I can do it with minimal traffic. The requirement is that it be smooth and reassuring.
Here's another link to the earlier example with traffic, but I didn't stop. Though I did have to stop accelerating, to "thread in." So it's somewhere in between. It's about 4.5 seconds to cross here. Started at 2:48, finished crossing at around 2:52.5. I got up to about 17mph before accelerating again.
I could have crossed faster if I were committing, of course - because I was being cautious here (just as well because the BMW changed into the target lane as I went!). But anyway in the "stopping" case with an ample median width, 5 seconds to clear seems fine. That would be a 20% reduction. Just increase all speeds by 30%, peaking at 16mph.
Last edited:
Would your method trigger the excessive acceleration and braking metrics that insurance companies use in their monitoring? If so, then IMO FSD shouldn't do it.
AlanSubie4Life
Efficiency Obsessed Member
I don't think so (for the braking part). I think it should be possible to do without using the brakes. We'll see. [EDIT: Calculating with 0.3g it seems like it would be tough...but again, we'll see - with the right pose it could have a significant amount of time in the median to slow down, so might be fine.]
Definitely FSD is triggering the excessive braking metrics all the time right now, though! I agree it shouldn't do that!
For the acceleration it's harder to know because I don't know what the metric would be. For SS there is no such metric so not clear what the limit "should be." Strong acceleration is fine and often leads to better safety margins (any deficiencies in anticipatory capability would be reflected in strong braking).
I think the key is that it be reasonably smooth (not jerky, predictable increase in power).
Certainly for the "commit" case (the only one I care about because I don't want to stop in the median), the car could be more assertive, for sure, likely with no increase in peak acceleration.
Rough desired profile would be 1.5s to 15mph, turn wide and angle to allow slowdow runway in median, maintain 15mph, then slow in 2.5 seconds to 0mph in median (perhaps 1.5 seconds of this IN the median). I think that should do it and keep deceleration below 0.3g.
Last edited:
Similar threads
