Let's goooo
100% success in heavy traffic
So close!
6/7 per the rules (perfect ~6-7 second gap missed on the last turn). This is setting aside Chuck’s comments about cars hitting the brakes for him (which would also be failures, if true, and due to excessive creep speed rather than overly cautious behavior of other drivers). I don’t know how he could tell they were braking for him so I discount this, unless there is clear evidence somehow.
Anyway it looked pretty good, though the butt sort of hung out on one of the turns (not counting it because the car relatively quickly proceeded).
I think the two areas where the ULT is still challenged is that body language when creeping (both for the driver and other drivers), and the median pose localization consistency.
That body language at the start is probably the biggest issue. Also the whole “using the median” is a very limited use case (many medians are uncomfortable to use this way). The pose isn’t as big a deal, since it can be fixed with the accelerator (assuming a wide median!).
Now that I have done a few of these, I know that body language makes it very tricky. Doesn’t look like they fixed it here based on Chuck’s response but hopefully on a future version.
I’m also not counting the inconsiderate driving against it. Yes, that is not cool (have experienced it). But hard to tell if there were alternatives (like not creeping as far) without being there.
It is interesting how far it is creeping. To me it looks ok, it is just the approach to that point which is apparently too sudden. Hard to see in videos but I believe Chuck.
My prediction is that in 10.69.3 we’ll see “new” issues in heavy traffic that actually already exist here (just not a large enough sample size with enough scenarios). Won’t be a regression, just the same.
It’s still not crossing assertively enough when it goes to the median, or in general. Maybe in 10.69.3!