Why signal out of courtesy? What benefit does that serve? Honestly it’s discourteous as it is extraneous noise that could interfere with legitimate signals.
Tesla obviously needs an automotive manners consultant
The difference between simulation and reality is greater in reality than in simulation.
My version has always been, "In theory, there is no difference between theory and practice. But in practice, there is.
Asking only because I don‘t recall ever seeing any. Prior to .69 was Chuck ever able to actually complete this turn to provide any useable data? I recall it giving up and turning right repeatedly. Unless I missed some vids. Unless he had many or any data fail vids my guess is they had to force collect their own.
Extrapolating from your statement, FSDb has been out for almost a year so at this point Tesla should have all the data they need from the current beta testers and should therefore be able to make perfect simulations eliminating the need for further beta testing.
I have seen FSD beta make this turn successfully many times on Chuck's videos. I've also seen him take over many times. There was one version, 10.12.1, maybe, that repeatedly rejected doing a left turn and bailed to the right.
Chuck's complete video library is on YouTube. I might have missed one, but as I recall, there are many videos of the car completing the ULT. The early videos did include many attempts where the car bailed and turned right. There are also many attempts where Chuck disengaged and took control. But there are also successful attempts. You'd have to ask Chuck for the exact percentage.
Dude constantly says it was just a bump in the video comments, his rim is fine.
He was. He also made the turn many times manually. The usable data is the camera data which can be used to create a model of the area. The weather, time of day, vehicle types and trajectories are all generated by the simulator.
Yeah, that's ridiculous. I thought I made it clear that I'm not talking about edge cases or getting near human performance. I'm only talking about getting the turn to ~99.9% success.
update. Cheers!Appreciate the clarification. Yeah I only watched his last video that I suspect triggered this interest by Tesla. Let’s hope they continue taking aggressive bites out of FSD improvements like they committed on this.
What I would like to know is why it often signals when going around a curve (sometimes even the wrong signal direction) BUT rarely signals when moving into a turn lane.
I've noticed that, too. In MN you're required to signal a turn 100 feet prior and it seems like FSD takes that very literally. If there's a 200' turn lane it won't signal until it's 100' from the intersection. Personally, I signal prior to entering the turn lane, as soon as I start to slow down to let other drivers know my intent. I don't know if that's legally required, but it's what most people do around here and seems like the courteous thing to do.What I would like to know is why it often signals when going around a curve (sometimes even the wrong signal direction) BUT rarely signals when moving into a turn lane.
Yeah, that applies to what I said, too. If they could get to 99.9% success by simulation they would have done it by now.
Agree. I've always thought FSD turned on blinkers too late. I will sometimes, depending on the turn, initiate the blinker earlier if there is a car close behind me,
Maybe, I'm just curious why simulation doesn't work in this case. It's hard for me to imagine them getting to the performance required for driverless operation if they haven't figured out how to use data from the fleet to solve Chuck's ULT.
