Tesla started rolling out the latest FSD Beta version 11.4.2 (2023.7.10) last week. Although Tesla made a number of improvements in lane handling in this release, it failed in U-turn tests massively.
Chuck Cook who has been testing FSD Beta for the last three years now, took his Tesla Model 3 running on v11.4.2 for some U-turn testing (video below). Chuck had to intervene each time either to initiate a U-turn by pressing the accelerator pedal or turn the car himself by taking control of the steering wheel (disengagement).
Let’s first go through the FSD Beta v11.4.2 release notes to know which specific improvements Tesla has made in this version. It’s always interesting to read FSD Beta release notes as Tesla provides accurate measures of the progress and improvements it has made in specific areas and scenarios.
FSD Beta v11.4.2 (2023.7.10) Release Notes
- Improved control through turns, and smoothness in general, by improving geometry, curvature, position, type, and topology of lanes, lines, road edges, and restricted space. Among other improvements, the perception of lanes in city streets improved by 36%, forks improved by 44%, merges improved by 27% and turns improved by 16%, due to a bigger and cleaner training set and updated lane-guidance module.
- Added lane-guidance inputs to the Occupancy Network to improve detections of long-range roadway features, resulting in a 16% reduction in false negative median detection.
- Improved ego’s assertiveness for crossing pedestrians in cases where ego can easily and safely cross before the pedestrian.
- Improved motorbike recall by 8% and increased vehicle detection precision to reduce false positive detections. These models also add more robustness to variance in vision frame-rate.
- Reduced interventions caused by other vehicles cutting into ego’s lane by 43%. This was accomplished by creating a framework to probabilistically anticipate objects that may cut into ego’s lane and proactively offset and/or adjust speed to position ego optimally for these futures.
- Improved cut-in control by reducing lane-centric velocity error by 40-50% for close-by vehicles.
- Improved recall for object partial lane encroachment by 20%, high yaw-rate cut-in by 40%, and cut-out by 26% by using additional features of the lane-change trajectory to improve supervision.
- Reduced highway false slowdowns related to underestimated velocities for faraway objects by adding 68K videos to the training set with improved auto-labeled ground truth.
- Smoothed in-lane offsetting for large vehicles by tuning the amount of lateral jerk allowed for the maneuver.
- Improved lateral control for upcoming high-curvature merges bias away from the merging lane.
U-Turns on FSD Beta v11.4.2
The last time Chuck Cook tested U-turns was on FSD Beta v11.3.4. Since then, there have been a lot of improvements in lane handling as we read in the above release notes. But the U-turn scenario seems to have been either a super-difficult task for Tesla Autopilot AI or it is currently not a high priority on the list of tasks.
It was interesting to see that after multiple failed attempts at U-turns. The Tesla map planner AI started to plan re-routing/alternate routes instead of taking U-turns. This is a good strategy in the short term for safety but costs time and battery range until U-turns become seamless.
Chuck also took a series of unique tests that include a B-pillar camera range and forward camera range test. Both of these tests were also performed on FSD Beta 11.4.2 (watch below).
According to Chuck Cook’s test, Tesla has extended the range
Related
- Watch FSD v13 first impressions by Tesla owners (unpark, park, reverse park, more)
- Tesla starts FSD v13 rollout to early access non-employees (Release Notes)
- Tesla Cybercab spotted testing and training FSD v13 at Giga Texas
- Cybertruck FSD 12.5.5.3 (2024.39.5 and 2024.38.4) Release Notes — tilted mirrors, Spotify Free, Sentry Mode improvements
- Tesla releases Cybertruck FSD for Canada (FSD 12.5.5.3)
- FSD v13 is coming soon, says Tesla CEO Elon Musk