Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Think we'll ever get follow distance of "1" again?

This site may earn commission on affiliate links.
AP1 would panic
Do you even Beta, bro?

(sorry, I had to haha)

I use the Beta for all of my driving, and the approximately half-second latency is painfully obvious in numerous scenarios.

It may not be across the board and may only affect certain NNs, but it’s definitely there. It’s most obvious when there are vehicles up ahead driving perpendicular to you or otherwise could be in your way. A lot of times I’ll know the car will be out of the way by the time we cross their path before the Beta realizes this leading to unnecessary braking. I’ve also seen it happen with traffic signals in rare cases (eg like in Chuck’s traffic light phantom braking video).
I don't have FSD, just EAP. But the ability to follow at a reasonable distance and smoothness isn't even related to EAP - it's really just how the car behaves while using TACC.

AP1 had a similar problem, aggressively slowing down if a car passes in front of you in the distance or if someone pulls out in front of you. This doesn't seem like a latency problem however. It is more that the car is braking out of an abundance of caution, assuming the car ahead does not get up to speed by the time you reach it, or that the crossing car is not out of your path by the time you reach it.

The engineers may also have introduced pauses in reaction, but the actual computer is not slow.
 
No. Vision is just not as accurate as radar for monitoring precise distance and speed differential so that’s why they can’t do “1”.

My car without radar behaves very poorly at adjusting to varying speeds of the lead car. And frequently gives false forward collision warnings when the speed differential is not nearly enough to cause a collision.
Radar-equipped cars, including mine, have been doing that for years.
 
How can you assess the latency of the AP computer when you are not familiar with the code that it executes? Do you have some detailed insight into Tesla's software?

I can see the latency at work simply be watching how long it takes for the car to react to various situations. For example, when a lead car slows and turns into a parking lot, my car continues to apply brakes for at least a half second after the lead car has completely departed the lane.

BTW, human reaction time of 0.25 sec is for unexpected, or sudden changes that are not predictable. When a human is performing an action where the situation is predictable, reaction times are much better. You can steer a car smoothly around corners because you continually assess the position of the car in response to your steering and can correct without a long reaction time. Same goes for the example I gave above where a car turns out of your lane. You anticipate the other car's turn so can stop braking and start speeding up the instant the lead car is clear of your lane.

Even so, 0.25 sec latency is far better than 0.5 sec in a driving situation, right?
Humans have a 0.2 sec reaction time in normal situations where you expect something to happen.

In traffic, a driver has an average reaction time of 1.0 second. This is much greater than the normal 0.2 second. This is because the situation to which he or she must react is unexpected. The extra 0.8 seconds are needed to figure out the best action to take on the new situation. (dutch source for this statistic)

There is a difference between perceiving something happening (i.e. photons hitting the retina are interpreted by the brain, 0.2 secs average) and being able to react to it (in traffic this means steering or pressing pedals). The latter is the "reaction time" in traffic.

AP1 and Tesla Vision AP are both vastly better than humans in this regard.

Radar equipped AP1 is probably lower latency right now, but Tesla Vision is improving (the Occupancy Network is built to reduce latency).

I am hopeful Tesla Vision will therefore in the future provide feature parity with Radar-AP1.

Another tidbit from Elons last Lex Fridman interview was regarding latency versus variation: latency is the amount of time between an occurance and the output of the FSD software. Variation is the difference in latency, and is a much bigger problem to solve for.

For example: if you have 500ms latency but no variation (in latency), things are pretty straightforward because the software can account for its fixed latency.

If you have 250ms of latency but with 250ms variation (meaning there are spikes of 0ms latency to 500ms latency) then it is much harder for the planner to calculate the optimal steering/control input.

Elon said they are more focussed on reducing variation and noise (signal that should be disgarded) than latency.

It all comes down to perfecting the perception problem, which Tesla still hasn't fully figured out yet. Once perception is perfect, the planning is pretty straightforward. (I'm not saying it's trivial, but it is trivial COMPARED to perception)

So since true FSD will require better perception than we have now, and Tesla is hell bent on achieving said FSD (and therefore better perception), I am hopeful a closer following distance than "2" is one of the future goals for Tesla Vision. (It just has to have this capability or else it shows that Tesla Vision is not solved yet).
 
  • Funny
Reactions: AlanSubie4Life
That's a planning problem, not a perception problem IMO. The forward collision software is set to "too safe".
I don't disagree, but was merely responding to the other poster, who seemed to be attributing the behavior to his car's lack of radar. Unfortunately, it can and does happen with radar-equipped cars as well. I've had my car for 4.5 years and only recently moved to Tesla Vision. I haven't noticed any increased frequency of the forward collision warnings. If anything, they might be occurring less frequently.

That said, while I agree that the warning still happens too frequently, at least it's only providing an audible warning and is not panic braking at inappropriate times. I'd rather it warn me a bit too frequently rather than not often enough.
 
I don't disagree, but was merely responding to the other poster, who seemed to be attributing the behavior to his car's lack of radar. Unfortunately, it can and does happen with radar-equipped cars as well. I've had my car for 4.5 years and only recently moved to Tesla Vision. I haven't noticed any increased frequency of the forward collision warnings. If anything, they might be occurring less frequently.

That said, while I agree that the warning still happens too frequently, at least it's only providing an audible warning and is not panic braking at inappropriate times. I'd rather it warn me a bit too frequently rather than not often enough.
precisely my experience as well on our 2017 mx
 
This is an easy problem to solve, as others have suggested. Just change the numbering so 1 is the current 2 and so on. Vast majority of people won't even notice.
I stated that it is a relative measurement and not an absolute. You CAN'T change how long a minute is but a moment is variable to your situation. So Follow distance is not a "minute" in distance but instead a "moment" and can be what Tesla defines it as.

Also it may not matter for too much longer anyway (or maybe it will as long as it is taking). Beta doesn't use follow distance so it likely won't in V11. [conjecture]And at some point the new Stack should replace legacy AP and that would remove follow distance from all cars.
 
Last edited: