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.
I wonder why the five star humans can't drive at a constant speed without a pace car. Maybe it's five stars out of ten.
This is guard rails put on an early release to make it more conservative in some situations and that was likely a situation where it was not sure if it should be conservative or not.
How does "autolabel" know if laws were broken, specifically?
1. Release it wide
2. Find errors
3. Fix errors one by one

For example if they find that it doesn't stop at stop signs in CA and NHTSA says to them fix this, they ask autolabel for situations in CA where there was a stop sign and the driver didn't stop and remove them and then do the hard work of gathering data of drivers actually obeying the stupid law and adding them. Problem solved, customer are complaining but at least NHTSA are happy.

For the turn on red they first make a map of where the law applies, then ask for situations where driver does this and add them to the dataset. Then it looks for situation where the system fails and asks for even more data of those specific situations where the system behaves correctly.

If they cannot find enough data they ask their hired drivers to go there and manually gather the data, like they did with Chucks unprotected left turn. Fun story, I was once part of team gathering data. We realized we didn't have enough data of the lane for emergency vehicles only. Our professional test driver did 10 laps of driving in it, knowing he was breaking the traffic laws.
 
It seems the myriad folks insisting "everyone" crosses double yellows and rolls stop signs in all the places they drive would directly contradict that
You pretty much have to cross double yellows if people are already doing jt to wait for the turn so I agree with that.
For stopping at stop signs there is a cluster of people who do stop at stops signs and the people running them are doing it all different speeds so maybe you can filter that out.
It does seem like if you can select training data from the largest group of drivers that behave the same way in the same situation that is likely to be the right way to behave.
 
I'd love to hear what a cop would say is the correct and legal move here. Do you think a police officer would tell you to cross the double yellow? I'm not being glib, I'm seriously asking what the correct, legal move is.
As far as I can tell cops no longer enforce traffic laws so I’m not sure. I’ve never heard of someone getting a ticket for it before.
If I were trying to drive 100% legally, for example if I was transporting large quantities of illegal drugs or trying avoid upsetting the NHTSA I would reroute to avoid the left turn.
 
Is your experience that AUTO tends to go too fast? Or perhaps on sections of 101 where the speed limit is 55, it stays at 65? Other reports seemed to be more of it being too slow, but that might be more of behavior on residential streets?
Here's an example with Tessie (see sig) who has v11 FSD. Today there was a construction zone preceded with a Speed Limit 55 sign. There was also an electronic sign that said "SPEED LIMIT WILL BE ENFORCED."

In the zone, Tessie was going 63 mph, and the "63" was in blue. Other cars were going 65. I dropped it to 55.

On curvy roads the car will often drop below the speed limit (e.g. 48 instead of 55, 61 instead of 65) although it could safely do the limit. That's unrelated to traffic, of which there is none.
 
If you want FSD to drive like a human, it's going to get you a ticket now and then.
I actually don't want the car to drive like a human. Driving like a human is overrated.

I want it to drive like a perfect robot as long as I can take over when it's acting like a goody two shoes. Same with my tax software and robot surgeons.
 
As far as I can tell cops no longer enforce traffic laws so I’m not sure. I’ve never heard of someone getting a ticket for it before.
If I were trying to drive 100% legally, for example if I was transporting large quantities of illegal drugs or trying avoid upsetting the NHTSA I would reroute to avoid the left turn.
This is great news. Red lights and stop signs are annoying, and inefficient for my commute. 😁
 
This is great news. Red lights and stop signs are annoying, and inefficient for my commute. 😁
I couldn't find data for San Diego where I live but I see this report that traffic citations have dropped 97% in San Francisco. That's why Whole Mars tests there.
 
  • Informative
Reactions: Dewg
The question I have - is that a perception issue, a processing lag, or by design? The more "human" it drives, the more rubber banding it may have. In the past people complained the car overreacted on slow downs. Perhaps this is a cushion. Not sure.
Good question. I don't think I've ever seen v11.x respond faster than 1.5 secs so there seems to be something additional happening in the above case.
 
  • Like
Reactions: jebinc and Dewg
The question I have - is that a perception issue, a processing lag, or by design? The more "human" it drives, the more rubber banding it may have. In the past people complained the car overreacted on slow downs. Perhaps this is a cushion. Not sure.
I think the complicating factor there was that the car was in an intersection. I'm sure there's emphasis on not stopping in intersections. That may be so strongly emphasized that FSD was willing to maintain speed until hard braking was needed to avoid an accident. Interestingly, the driver didn't react to the event as it happened and didn't comment on it afterwards.
 
I've had 11.x do many quick lane changes to get past a slow vehicle
I think what surprised me with the 12.2.1 behavior is that the slow vehicles couldn't even be seen until the lead vehicle moved out of the way, and 12.x quickly followed with its own lane change before the lead vehicle even completed their change. Here's every other frame for half a second leading up to turn signal showing the repeater camera view:

12.2.1 change left.jpg


At least from this camera view, you can only really see the slow vehicle in the top-right frame, so about a quarter second for it to turn on the signal. Perhaps end-to-end has learned that a sudden lead vehicle changing lanes might reveal slow traffic, so prepare to change lanes too?
 
I'd love to hear what a cop would say is the correct and legal move here. Do you think a police officer would tell you to cross the double yellow? I'm not being glib, I'm seriously asking what the correct, legal move is.
What I remember, but could not find a reference, is that if you cannot fit into the marked turn lane, you must proceed straight and try again at the next intersection.
 
Maybe current training is forcing this?
Is the neural network doing lane planning for a route or is that planning still coming from some external component that feeds FSD? For example, "take the next left". If that directive comes late enough, then there's not much any system can do.

I find it so odd that lane selection is such a difficult problem to solve. The car is juggling pedestrians, cars, traffic controls, markings on the road, etc, in real time, but knowing when to get into the left lane for a left turn is somehow beyond it. Or is this that perennial favorite - flawed map data?