Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
Maybe. Need more examples.

Reaction time still seems to be one second, which dramatically increases risk of rear-end collision. This sort of situation was the only time I have been in a collision (I got rear ended but it was my fault, though that was not the way insurance saw it). 13:03 light turns, 13:04 slowing begins.

Much slower than a human.

Are you using the time from light going yellow to deceleration to extrapolate that the car would rear end a lead car?
The car didn't need to hit the brakes instantaneously and doing so would increase its chances of being rear ended. It was going 21 and appeared to stop in plenty of time/ distance.
Would be nice to have 360 camera view in the vidro...
 
Was this a lucky slowing or did 12.1.2 predict a cut-in for a vehicle not wanting to get stuck behind a stopped car?

View attachment 1012298

The speed started dropping from 29mph before the white SUV even signaled or turned to cut in. Did 12.1.2 notice the 3 brake lights turn off indicating the SUV was not going to continue slowing down or perhaps general defensive behavior when adjacent lanes suddenly slow down?
Quite likely. Karpathy said years ago there were collecting samples of non-signalled cut ins.
 
The car didn't need to hit the brakes instantaneously and doing so would increase its chances of being rear ended
I agree one has to think about an instant reaction. But flying toward a light with no slowing and with reaction much slower than a human, then slamming the brakes, clearly increases risk. It’s exactly the scenario which results in rear ending because the person behind may think you have committed to go and unwisely try to follow through, possibly accelerating.

But change to 0.5 sec reaction time (slow end for an human) and you end up with 0.38g stop which is tolerable rather than one which exceeded 0.6g (14mph in 1 sec), and averaged 0.48g - a very abrupt stop.

It’s a big difference.

Omar specifically said the person behind expected him to go.

Obviously the correct decision was to go, too. (Certainly would have entered intersection on yellow as legally required. Could do more work with distances to determine if could have crossed intersection before turning red - I think possibly, but it doesn’t matter.)
 
Last edited:
Go tell that to Apple
 
  • Like
Reactions: spacecoin
Quite likely. Karpathy said years ago there were collecting samples of non-signalled cut ins.
Just so everyone that doesn't know here. @mongo Is one of the best contributors on TMC and has been for years. He speaks from tons of research and knowledge. When he posts here, take a minute to respond if you disagree.
In years of my experience, he is almost never wrong.
Nice to see you buddy!
 
I was referring to the Bionic chips used in the iphones as well as the Macbooks. they moved away from IBM & Intel and built their own.
 
But change to 0.5 sec reaction time (slow end for an human) and you end up with 0.38g stop which is tolerable rather than one which exceeded 0.6g (14mph in 1 sec), and averaged 0.48g - a very abrupt stop.
Actually g force could be even lower due to the extra distance.

Compare stopping from 21 mph in 31ft, vs 45ft starting at 21mph if you start slowing half second earlier.

So if you start stopping 0.5sec earlier you can actually take 0.5sec longer to stop. For a total of 3 seconds.

So at 0.31g you cover 45 feet in 3 seconds.

Vs. what happened which is (roughly;deceleration was not uniform) slowing from 21mph to 0 in 2 seconds over 31 feet which is 0.48g (as mentioned peak was higher which means stopping distance was probably less than 31 feet).

Starting the stop earlier is very helpful!

0.3g is much more tolerable than a best-case ~0.5g. (Which actually peaked at 0.6g.)
 
  • Informative
Reactions: APotatoGod
From Boston area and FSD is certainly much better now than when I first used FSD in 2021. But it still craps out way too much.
Two main problems I see.
  1. 400+ year old road layout. Lots of obstructed views and just very difficult intersections that challenge human drivers.
  2. Poor road markings/paintings to compared to what I see in Omar's and Chuck Cooks videos. Night and day worse in greater Boston.
I've used FSD in California, Florida, Texas too and FSD works much better mostly because of the road layouts and better markings.


Poor road markings is a problem here too, for all drivers (computer and human.) They fade away too fast between the new environmentally safer paint and the use of road salt.

But I bet Boston will be a HUGE problem for AV because of the tendency of the drivers there to stop to allow pedestrians to cross. I found that if I was standing still at a corner, checking a tourist map, I needed to stand with my back to the road or else the drivers would stop to let me cross!