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

Drunken TACC

This site may earn commission on affiliate links.
Does anyone else feel that the acceleration/deceleration gain for TACC in traffic is way too large? The biggest disappointment with the car is that TACC, in even moderate traffic, feels like a drunken kid, hurry up followed by throw the humans forward with aggressive regen, repeat, over and over until I am nauseated. And I wonder what the people behind me think.

I wonder if Tesla will ever fix this. v10 was not an improvement from v9 in this regard. Configurable gain on the PID algorithm here would be a nice feature.
 
If you're talking about AP accelerating from a stop to follow the car ahead of you, I hear you. Prior to V10, it was way too slow to accelerate, and it fell behind traffic by several car lengths. They may have over-corrected with V10.
  • When the car ahead launches off quickly, AP now follows with a little bit of whiplash action.
  • When the car ahead launches slowly, I find the new "gain" just about perfect.
  • When there is a semi ahead, it almost makes you sick. With all the gears in their transmission, each only adds a few mph, with a long pause in between. The new gains cause whiplash followed by braking in each of the truck's gears.
Hopefully they'll moderate these settings soon.
 
In Los Angeles traffic, creeping along around 5 to 20mph I find the whiplash intolerable. Which makes TACC pretty useless. Which is disappointing because knowing I wasn't going to spring for FSD, and living in LA, TACC was a huge selling point.

My wife's Honda Accord handles cruise control much better than my Tesla. Granted it's only partially adaptive, but it's never had a phantom breaking incident, which TACC seems to want to do on average every 50 or so miles, and it's deceleration and acceleration are so seamless as to be unnoticeable. I often just won't notice when it's gone from 70 to 60 while I almost always notice, in a negative way, in the Tesla.
 
While room for improvement I think that a lot of the "drunken/motion" problems are somewhat inherent. You are paying full attention and making a decision to accelerate or brake based on what you would do. If the car deviates this confuses your vestibular causing this. Until we fully become passive passengers i think this ill be problem.
 
I haven't tried it, but I was just wondering out loud, if you set TACC with chill and low regen, would the car accel/decel more gently?
I am not sure about the regen factor (I haven't tested that), but Chill mode makes a HUGE difference. On my traffic intensive daily commute, I try to always remember to turn on Chill (but when non-TACC driving, it is off)
 
  • Informative
Reactions: Potpourri
Does anyone else feel that the acceleration/deceleration gain for TACC in traffic is way too large? The biggest disappointment with the car is that TACC, in even moderate traffic, feels like a drunken kid, hurry up followed by throw the humans forward with aggressive regen, repeat, over and over until I am nauseated. And I wonder what the people behind me think.

I wonder if Tesla will ever fix this. v10 was not an improvement from v9 in this regard. Configurable gain on the PID algorithm here would be a nice feature.

Yeah, I noticed that and posted on another forum. I even posted a video on it. This is the car accelerating from clearing some traffic. Look at the surging in the accleration bar. No reason for that whatsoever. The car in front of me is pulling away faster than I am accelerating.




If you're talking about AP accelerating from a stop to follow the car ahead of you, I hear you. Prior to V10, it was way too slow to accelerate, and it fell behind traffic by several car lengths. They may have over-corrected with V10.
  • When the car ahead launches off quickly, AP now follows with a little bit of whiplash action.
  • When the car ahead launches slowly, I find the new "gain" just about perfect.
  • When there is a semi ahead, it almost makes you sick. With all the gears in their transmission, each only adds a few mph, with a long pause in between. The new gains cause whiplash followed by braking in each of the truck's gears.
Hopefully they'll moderate these settings soon.

I don't think that this is the point of discussion.

While room for improvement I think that a lot of the "drunken/motion" problems are somewhat inherent. You are paying full attention and making a decision to accelerate or brake based on what you would do. If the car deviates this confuses your vestibular causing this. Until we fully become passive passengers i think this ill be problem.

No, it's not. As another poster said above, other cars have smoother systems. I had a loaner Porsche Cayenne and it was WAY smoother braking and accelerating than my Model 3. Also, this seems to have gotten worse on version 10 for some reason.
 
Exactly, the PID tuning is horrendous. They should drop the D coefficient almost entirely, or at least limit it to only affect drivetrain friction differences, or use a secondary function to determine D based on available traction over the past few seconds/minutes, etc.

I hope this shittiness is due to them allocating software resources to more pressing matters, and that, it will eventually clear up.

I'm sure they have strategies in place to handle all of this, but the application is so damn poor. I would love to help them work on this if I could. I wish there was a way for me to contribute by providing datalogs and comments. I would do this for free.

There should be a 3rd option (software upgrades) for qualified engineers (crowdsourced) to submit data/comments. Come on, Elon, let me help you help us!
 
Increasing the following distance setting also smooths it out considerably. If its got more leeway to get to the target following distance/time then it doesn't have to brake/accelerate so hard.

Unless you set it to the max (7). For some reason it feels like 7 might be at the limits of what it can detect at times, so it can be a bit jumpy as if it lost sight of the fact there was a vehicle in front until it happens to see it again. Not consistently, but it definitely occurs more at 7 than 6. Even then it's holding that position way too rigidly IMO.

Exactly, the PID tuning is horrendous. They should drop the D coefficient almost entirely, or at least limit it to only affect drivetrain friction differences, or use a secondary function to determine D based on available traction over the past few seconds/minutes, etc.

I hope this shittiness is due to them allocating software resources to more pressing matters, and that, it will eventually clear up.

I'm sure they have strategies in place to handle all of this, but the application is so damn poor. I would love to help them work on this if I could. I wish there was a way for me to contribute by providing datalogs and comments. I would do this for free.

There should be a 3rd option (software upgrades) for qualified engineers (crowdsourced) to submit data/comments. Come on, Elon, let me help you help us!

99.9% agree! I'm not sure if PID is what they're using, but it very much feels like a heavy D bias. It even sorta makes sense -- for responsiveness on takeoff from a red light, that's exactly what would be helpful in order to emulate human driving. But when it comes to highway speeds it's incredibly annoying. It did seem to get better with V10, but it's definitely not as smooth as nearly any other adaptive cruise system.
 
This is strange, I use it everyday and my commute is 135 miles each day. I have not noticed any sudden acceleration even if the car in front switched lane and the next car up is far away. I normally set my max speed at 75. Lets say I am following a car at 68... and he changed lane. My car would slowly go up from 68 to 69 to 70... at about 1mph per second. I don't think that's too rough.

And on the way back I get traffic... sometime stop and go. Even if a car cut in right in front of me, the car does not hit the brake...it just stop accelerating. I am not in chill mode either. I do have an AWD. I know from experience having a RWD loaner that RWD behave a lot differently in regen braking and acceleration in manual driving. Maybe that carries over to TACC....
 
  • Informative
Reactions: Potpourri
This is strange, I use it everyday and my commute is 135 miles each day. I have not noticed any sudden acceleration even if the car in front switched lane and the next car up is far away. I normally set my max speed at 75. Lets say I am following a car at 68... and he changed lane. My car would slowly go up from 68 to 69 to 70... at about 1mph per second. I don't think that's too rough.

And on the way back I get traffic... sometime stop and go. Even if a car cut in right in front of me, the car does not hit the brake...it just stop accelerating. I am not in chill mode either. I do have an AWD. I know from experience having a RWD loaner that RWD behave a lot differently in regen braking and acceleration in manual driving. Maybe that carries over to TACC....

To clarify the OP's wording, which was specifically "acceleration/deceleration gain", what they mean is not how hard it accelerates in one motion but how it can somewhat jerkily flip between acceleration, then regen/braking, then acceleration, then regen/braking, etc.

With V10, if you have people reliably travelling at a steady pace with a car that is easy for the Model 3 to detect (e.g. not a lifted truck), it is a lot smoother than it used to be.
 
  • Informative
Reactions: Potpourri
To clarify the OP's wording, which was specifically "acceleration/deceleration gain", what they mean is not how hard it accelerates in one motion but how it can somewhat jerkily flip between acceleration, then regen/braking, then acceleration, then regen/braking, etc.

With V10, if you have people reliably travelling at a steady pace with a car that is easy for the Model 3 to detect (e.g. not a lifted truck), it is a lot smoother than it used to be.

Yeah I am not seeing that either. To me, it drives really smooth unless there is an overpass. The only time I have the really bad jerky motion is when there are unexpected shadows like overpass or shadows from semi when sun is on the other side. Somehow the obstacle detection thinks there is something there for like half a sec. I am really sensitive to motion sickness, and I can't ride in any roller coaster or play FPS games. So if the AP or TACC is jerky I think I would have noticed. I just AP when I am not in the carpool lane.. and use TACC when I am in the carpool lane (to not be too close to the center divider and let motorcycle passes).

I am in LA too same as OP and I work at El Segundo, so I get traffic everyday too. Maybe OP has a P3D? Also I set my following distance to 3.
 
Last edited: