Tesla’s Full Self-Driving Beta program gets a new version this weekend with update version 10.6, and the release notes show extra emphasis on vision, object detection and the precision of the vehicle’s movements.
The Tesla Full Self-Driving Beta program is a unique fleet of customer-owned vehicles that have exclusive access to the automaker’s latest and most robust releases of its semi-autonomous driving program. It has drivers with Tesla Safety Scores of 98, 99 and 100, and is likely to expand to lower scores in the future. However, the best way to be included in the program is to achieve scores as close to 100 as possible. Tesla̵[ads1]7;s FSD Beta updates do not increase the capabilities of the FSD Suite to make the vehicle fully autonomous. Failure to pay attention to the steering wheel when using the program may result in expulsion from the Beta fleet, and Tesla has made it clear that they will take relatively no risk in protecting the exclusivity of the Beta program.
Tesla enforces strict FSD Beta rules for inattentive drivers
Tesla FSD Beta updates come every other week, give or take a few days based on performance internally. Yesterday, CEO Elon Musk confirmed that 10.6 will roll out this weekend, with the exception of large or significant bottlenecks that will affect the performance of the semi-autonomous driving suite. It rolled out to employees last night and earlier today.
It is released internally today, and then rolls out to external beta in a few days if no significant issues are found
– Elon Musk (@elonmusk) December 3, 2021
It seems that the main concerns of Tesla’s FSD engineers for this release were object detection, visibility and improvement of the overall performance of the suite, especially through the accuracy and predictability of the movements. FSD Beta 10.6 Release Notes were shared on Reddit by u / 110110.
Improved network architecture for object detection for non-VRUs (eg cars, trucks, buses). 7% higher recall. 16% lower depth errors, and 21% lower speed errors for intersecting vehicles.
New visibility network with 18.5% less average relative error.
New general static object network with 17% precision improvements in cases of high curvature and night time.
Improved stopping position for unprotected left turns while giving way to oncoming objects, using object predictions outside the intersection.
Allow more space for longitudinal adjustment during mergers by incorporating modeling of the merge region end.
Improved comfort when shifting for objects that cut out of the lane.
The new FSD Beta version should roll out to Beta program members in the coming days.
I would love to hear from you! If you have any comments, concerns or questions, please email me at [email protected]. You can also reach me on Twitter @KlenderJoey, or if you have news tips, you can email us at [email protected].