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

MASTER THREAD: 2019.40.2 - FSD AutoSteer Stop Sign Warning and Adjacent Lane Speeds

This site may earn commission on affiliate links.
I have verified the Stop Light Warning but have not been able to get the Stop Sign Warning?
From my experience with HW2.5 + EAP, I've gotten better stop sign warning success at OSM-mapped stop signs, but I've also seen the warning for other places too. Looks like there either aren't a lot of stop signs around here or they're not well mapped on OSM:

Screen Shot 2019-12-13 at 1.35.12 PM.png

I would suggest going to an area that has a high concentration of these blue circles and test at 4-way stops (or at least a street where you can keep going straight as otherwise Autopilot will panic seeing no road ahead).
 
From my experience with HW2.5 + EAP, I've gotten better stop sign warning success at OSM-mapped stop signs, but I've also seen the warning for other places too. Looks like there either aren't a lot of stop signs around here or they're not well mapped on OSM:

View attachment 487932

I would suggest going to an area that has a high concentration of these blue circles and test at 4-way stops (or at least a street where you can keep going straight as otherwise Autopilot will panic seeing no road ahead).

This makes sense. I wanted to try it on a side road before trying it on a major thoroughfare :)
 
Yes, I am basing it on Elon’s definition of FSD. My bet is that once Stop Light/Sign recognition is released, Elon will declare FSD feature complete.so they can declare income.

This is similar to when Smart Summon was released and he declared EAP was feature complete. We all know how many issues remain with Smart Summon.
If FSD only stops at stop signs and stop lights, it is not feature complete. FSD needs to complete follow the Nav system which means it makes turns and recognizes that the stop light turned green and starts moving again.
 
Unless Tesla has been working on a completely secret parallel development program with a whole separate fork of software, I have no idea how they plan on tackling the challenges of city driving any time in the next several years.

Did you miss Autonomy Day? They've definitely got a development version of the firmware with city-driving features far beyond what the released firmware can do. And this was back in April:

Check out the little Autopilot UI during the demonstration, it's got a much better understanding of drivable space than the current firmware, has visualization of intersections, stop signs, stop lights, turn lanes.

Granted, the demonstration was in a well tested area, but it does show us that Tesla isn't pushing the latest and greatest to our cars all at once. They have a development branch in the works.
 
Full self driving cars won’t happen until we develop and implement vehicle to vehicle communications. Sadly, our “President” discontinued research into a car to car communication standard. This has set back true self driving several years.
I said this some time ago. And I do believe in that. So, guys, do not be so excited, it will not happen soon...
 
Has anyone who received the update earlier on and has had a chance to use it over a more extended period of time noticed any gradual improvement with lane ping pong? I've had updates in the past where it has been initially sketchy, but 'learned' or resolved over time.

Hello. The car doesn't "learn" locally, and requires that the software be updated to implement any newly trained or coded behavior.

The issue with ping-pong effect within a lane really raised its head for me in the 36.2 series, and I noticed it on a 500+ mile drive. I had 40.2 for around 350 miles out of that same 500+ mile drive I repeated almost exactly a month later, and the ping-pong effect got much worse. To the point where I didn't know if I could stand to make the whole drive with AP engaged.

I applied the 40.2.1 update during the last third of this trip, and noticed that wandering inside a lane had reduced but not been eliminated. It also seems like the reaction to the presence of a semi is slightly smoothed, but still way too dramatic for my taste. There are new issues with lane line detection, and what happens when lines disappear because the road surface is destroyed, and NoAP still makes haphazard lane choice suggestions. The vehicle also has too much lateral force when entering and exiting a curve, and it saws at the wheel while following a long bend in a road or on/off ramp.

These things need to be addressed. Some of them are regressions in behavior that had previously been working fine.
 
  • Informative
Reactions: Fernand and Char
Full self driving cars won’t happen until we develop and implement vehicle to vehicle communications. Sadly, our “President” discontinued research into a car to car communication standard. This has set back true self driving several years.

This will literally never be a requirement, and you should never want anything like this. If I can send something to your car that will make it react, then I can abuse that system and cause negative behavior. Imagine me at a 4-way intersection with a laptop, telling every car that I'm entering the intersection and they better not. I could grind an entire town to a halt by doing this at a major roadway.

V2V communication is a terrible idea.
 
This will literally never be a requirement, and you should never want anything like this. If I can send something to your car that will make it react, then I can abuse that system and cause negative behavior. Imagine me at a 4-way intersection with a laptop, telling every car that I'm entering the intersection and they better not. I could grind an entire town to a halt by doing this at a major roadway.

V2V communication is a terrible idea.

So then, your idea is that full autonomy is a no-go.
 
So then, your idea is that full autonomy is a no-go.

No, just like V2V isn't necessary for you to drive a car, it isn't necessary for full autonomy. Even if a V2V standard was made, how many years would it be before every car on the road incorporated it. (Likely never unless they outlaw non-V2V compliant cars.)

You couldn't rely on V2V so you have to be able to handle everything via local sensors. Be that radar, LIDAR, and/or camera. And V2V doesn't help with pedestrians, bicyclist, animals, etc. So it becomes additional complexity for almost no value. (And a lot of risk.)
 
Last edited:
It’s more than that. You also have to have auto steer activated.
.

How do you activate auto steer? I think my understanding of these terminology maybe a bit off. Doesn't double clicking down the right stock activates autopilot; which automatically activates auto steer and TACC? One single click down down on the right stock activates the cruise control (TACC). So I'm confused on this terminology.
 
No, just like V2V isn't necessary for you to drive a car, it isn't necessary for full autonomy. Even if a V2V standard was made, how many years would it be before every car on the road incorporated it. (Likely never unless they outlaw non-V2V compliant cars.)

You couldn't rely on V2V so you have to be able to handle everything via local sensors. Be that radar, LIDAR, or camera. And V2V doesn't help with pedestrians, bicyclist, animals, etc. So it becomes complexity for almost not value. (And a lot of risk.)

How then would an "autonomous" car know when there is another car about to enter the intersection? Or that the other driver is about to go out of turn on a four way stop? Or about to run a red light?
 
.

How do you activate auto steer? I think my understanding of these terminology maybe a bit off. Doesn't double clicking down the right stock activates autopilot; which automatically activates auto steer and TACC? One single click down down on the right stock activates the cruise control (TACC). So I'm confused on this terminology.

You down click once for TACC. Then you down click again for auto steer. You can cancel auto steer by turning the wheel. You can cancel both by up clicking or hitting the brake.
 
Pulled into Las Vegas from Barstow. Arrived with 5% charge and had been driving for about 2 hours.

Hit the peak of 170kW for a few minutes. Dropped below 100kW between 40 and 50% SOC.

Edit: I have June 2019 SR+ with FSD.

Took about 40 minutes to get from 5 to 90%.

SR+ 40.2
14 % charge on arrival.

I hit 133 at a SC rated at 120 (Must really be be a 150).

Only lasted for a minute then started dropping.
Outside temp 50. Only 10 minutes of conditioning before arriving at SC.
Lots of thermal thumps from pack while charging.

20 minutes to charge from 14 % to 70%.

Some data points. Video screwed up. No video until about 5 min into charge.
99 kw at 32%
97 kw at 33%
95 at 34%
93 35
92 36
90 37
88 39
87 40
85 41
84 43
83 44
82 46
81 47
80 48
79 49
78 50
77 51
71 56
70 57
69 58
66 60
64 61
64 62
62 63
62 64
61 65
59 66
58 67
57 68
56 69
 

Attachments

  • 20191213_145937.jpg
    20191213_145937.jpg
    296.6 KB · Views: 45
Last edited:
Jerky accel, decel can be improved a bit by using chill mode and decreased brake regen. I do this whenever I find myself in stop and go driving.

As I said in my post chill mode made no difference in the jerkiness.

The abrupt accel/decal in slow and go traffic (5-15 MPH) is making me car sick and for the first time in nearly two years I manually drove my car through an hour of traffic because the aggressive (and unnecessary) accel/brake was making me nauseous.
 
  • Like
Reactions: Skione65
The abrupt accel/decal in slow and go traffic (5-15 MPH) is making me car sick and for the first time in nearly two years I manually drove my car through an hour of traffic because the aggressive (and unnecessary) accel/brake was making me nauseous.[/QUOTE]

Same today for me. Literally turned it off and drove myself for the first time in a long time out of frustration. Lane changes are SWEET tho!

Ski
 
Last edited:
  • Like
Reactions: mtndrew1