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

10.9 FSD

This site may earn commission on affiliate links.
Last night I drove from Monterey Peninsula to San Diego (~7 hours). This requires crossing some mountains to get from 101 to 5 and in this 20-30 mile stretch was heavy fog. Must be common as there is a persistent sign that lights up warning about this condition.

This fog persisted onto route 5 for some 10 miles maybe. In that space, no less than 1/2 dozen times (maybe more) auto steer disabled due to visibility issues. Eventually so did cruise control as well.

How is FSD going to be able to drive me in bad weather when the humans were able to safely navigate, at no greater than 50mph I'd say, but the cameras alone could not (was wishing for radar to be used I guess)

This event followed some earlier days in the week where, at one time or another, both side mounted cameras were blinded by sun.

Just made me wonder
 
Last night I drove from Monterey Peninsula to San Diego (~7 hours). This requires crossing some mountains to get from 101 to 5 and in this 20-30 mile stretch was heavy fog. Must be common as there is a persistent sign that lights up warning about this condition.

This fog persisted onto route 5 for some 10 miles maybe. In that space, no less than 1/2 dozen times (maybe more) auto steer disabled due to visibility issues. Eventually so did cruise control as well.

How is FSD going to be able to drive me in bad weather when the humans were able to safely navigate, at no greater than 50mph I'd say, but the cameras alone could not (was wishing for radar to be used I guess)

This event followed some earlier days in the week where, at one time or another, both side mounted cameras were blinded by sun.

Just made me wonder


Probably the first baby steps.
IMG_0215.jpeg
 
Is it me or is there still no improvement on “random” “phantom” braking in that changelog…. Lol. There’s no way their internal test team doesn’t experience this crap.
It’s past time where this problem should be captured into some number of automated test cases. Sure the test cases won’t achieve 100% coverage but doing so would be moving the problem to new & eventually more complex conditions to occur.

The problem being solved is far to complex to solve without a good testing methodology.
 
  • Like
Reactions: pilotSteve
From reddit the full release notes ..

  1. Improved intersection extents and right of way assignment by updating modeling of intersection areas from dense rasters ("bag of points") to sparse instances. Increased intersection region IOU by 4.2%. The sparse intersection network is the first model deployed with an auto-regressive architecture that runs natively with low latency on the TRIP Al accelerator chip, through innovations in the Al compiler stack.
  2. Upgraded generalized static object network to use 10-bit photon count streams rather than 8-bit ISP tonemapped images by adding 10-bit inference support in the Al compiler stack. Improved overall recall by 3.9% and precision by 1.7%.
  3. Made unprotected left turns across oncoming lanes more natural by proceeding straight into intersection while yielding, before initiating the turn.
  4. Improved lane preference and topology estimation by 1.2% with a network update and a new format for navigation clues.
  5. Improved short deadline lane changes with better modeling of necessary deceleration for maneuvers beyond the lane change.
  6. Improved future paths for objects not confined to lane geometry by better modelling of their kinematics.
  7. Made launches from stop more calm when there is an imminent slowdown nearby.
  8. Improved gap selection when yielding to a stream of oncoming cars on narrow roads.
Lots of planning related updates. But the first 2 are beta 11 items pulled into 10.9.

Have to chuckle at (7) - "launches from stop more calm when there is an imminent slowdown nearby". Though I've to say there has been a steady improvement on this. FSD no longers tries to speed upto max when there is a car infront a few hundred feet away. They have to do that when there are turns upcoming too ...
 
Is it me or is there still no improvement on “random” “phantom” braking in that changelog…. Lol. There’s no way their internal test team doesn’t experience this crap.

They devoted an entire release to it (10.8 ?) ... I won't be surprised if its the #1 issue they have on their list now.

But, phantom braking is one of the toughest to fix. They may need lot of other improvements they are talking about for Beta 11 to make a big dent in phantom braking.
 
From reddit the full release notes ..

Have to chuckle at (7) - "launches from stop more calm when there is an imminent slowdown nearby". Though I've to say there has been a steady improvement on this. FSD no longers tries to speed upto max when there is a car infront a few hundred feet away. They have to do that when there are turns upcoming too ...
Interesting to see your feedback because 10.8 was a regression for me. To leave or return to my neighborhood I have several turns about 200 feet apart and it now accelerates way too much just to stop and then make a turn. That problem had been fixed but returned. Once I get 10.9 I'll know right away if the behavior is improved for me. I suspect my neighbors are wondering why I've all of a sudden started driving faster in the neighborhood:)