Tesla finally started rolling out the latest iteration of its much anticipated self-driving beta software FSD Beta 10.69.3. This new version was due after AI Day 2022 last month but was delayed due to code updates and architectural upgrades.
FSD Beta 10.69.3 has a long list of improvements that Tesla has published in the release notes of this update (below).
Tesla started rolling out this new version of FSD Beta to the company’s internal testing team yesterday. Some external beta testers are now reporting that they are able to download FSD Beta 10.69.3 (firmware version: 2022.36.15) this morning.
FSD Beta 10.69.3 Release Notes
- Upgraded the Object Detection network to photon count video streams and retrained all parameters with the latest auto-labeled datasets (with a special emphasis on low visibility scenarios).
- Improved the architecture for better accuracy and latency, higher recall for faraway vehicles, lower velocity error of crossing vehicles by 20%, and improved VRU precision by 20%.
- Converted the Non-VRU Velocity network to a two-stage network, which reduced latency and improved crossing pedestrian velocity error by 60%.
- Converted the Non-VRU Attributes network to a two-stage network, which reduced latency, reduced incorrect lane assignment of crossing vehicles by 45%, and reduced incorrect parked predictions by 15%.
- Reformulated the autoregressive Vector Lanes grammar to improve the precision of lanes by 9.2%, recall of lanes by 18.7%, and recall of forks by 51.1%. Includes a full network update where all components were re-trained with 3.8x the amount of data.
- Added a new “road markings” module to the Vector Lanes neural network which improves lane topology error at intersections by 38.9%.
- Upgraded the Occupancy Network to align with road surface instead of ego for improved detection stability and improved recall at hill crest.
- Reduced runtime of candidate trajectory generation by approximately 80% and improved smoothness by distilling an expensive trajectory optimization procedure into a lightweight planner neural network.
- Improved decision-making for short-deadline lane changes around gores by richer modeling of the trade-off between going off-route vs trajectory required to drive through the gore region.
- Reduced false slowdowns for pedestrians near crosswalk by using a better model for the kinematics of the pedestrian.
- Added control for more precise object geometry as detected by general occupancy network.
- Improved control for vehicles cutting out of our desired path by better modeling of their turning / lateral maneuvers thus avoiding unnatural slowdowns.
- Improved longitudinal control while offsetting around static obstacles by searching over feasible vehicle motion profiles.
- Improved longitudinal control smoothness for in-lane vehicles during high relative velocity scenarios by also considering relative acceleration in the trajectory optimization.
- Reduced best-case object photo-to-control system latency by 26% through adaptive planner scheduling, restructuring of trajectory selection, and parallelizing perception compute. This allows us to make quicker decisions and improves reaction time.
Tesla CEO Elon Musk was right when he said 10.69.3 is not just another point update, it’s a major upgrade to the FSD Beta stack.
As we can see in the release notes above, architectural changes and improvements have been a big part of the FSD Beta version 10.69.3.
Tesla has even retrained the neural network with new datasets to improve self-driving behavior in lower visibility. We will see these low-visibility upgrades in the future as well because now Autopilot is solely relying on Tesla Vision after the removal of the radar and ultrasonic sensors.
An interesting thing to notice in the above screenshot taken from the FSD Beta 10.69.3 driving visualizations is how far the Tesla Vision is now able to see (check the tiny cars at a distance).
The other area that Tesla has again focused on in this update is the safety of vulnerable road users (VRUs). The safety of these road users either at a crosswalk, footpath, or any unexpected scenario is of utmost importance if Tesla wants to succeed in the realization of Full Self-Driving (FSD).
FSD Beta 10.69.3 Testing Videos
As this new version of FSD Beta has been available to a very limited number of Tesla owners as of this writing, the test videos and reviews are limited. But the ones coming in on Twitter and YouTube are mostly positive about the improvements in v10.9.3.
Another short but interesting new feature in FSD Beta 10.69.3 is the addition of the estimated distance and time to the destination. This feature is only visible when the FSD Beta visualization is turned to full-screen mode.
Omar/Whole Mars Catalog drove his cars on the streets of San Francisco, California. In 3 of his attempts last night, he did not need to take over the car from Autopilot.
Navigating busy city streets is the final challenging task that Tesla has to overcome to give this beta software the shape of a final product. Let’s watch how FSD Beta 10.69.3 did in its first attempts at public trials.