March 22, 2023

Tesla releases new Full Self-Driving Beta update with high-level changes

Tesla is starting to roll out a new Full Self-Driving (FSD) Beta software update that includes many high-level changes that should positively impact performance.

FSD Beta allows Tesla vehicles to drive autonomously to a destination entered into the car’s navigation system, but the driver must remain alert and ready to take control at all times.

Since the responsibility rests with the driver and not Tesla’s system, it is still considered a level two driver assistance system despite its name. It’s been kind of a “two steps forward, one step back” kind of program, as some updates have seen a drop in ride quality.

Tesla has frequently released new software updates to the FSD Beta program and added new owners to it.

The company now has more than 100,000 people in the FSD Beta program and plans to expand it to everyone who buys access in North America by the end of the year with a few software updates to improve the system.

Considering we’re already in November and Tesla’s new FSD Beta update usually takes at least a month to deliver, we expect Tesla to be one or two updates away from the promised wider release.

Today, the automaker has started pushing the new FSD Beta update (v10.69.3) to employees for internal testing, which usually means it will soon expand to beta testers in the customer fleet.

According to the release notes below, the update doesn’t include any new features, but it does include a lot of high-level updates to Tesla’s neural networks to improve overall system performance.

Tesla Full Self-Driving Beta Release Notes v10.69.3 Release Notes via Not a Tesla App:

– Updated the Object Detection network for photon counting video streams and retrained all parameters with the latest auto-labeled datasets (especially in low visibility scenarios).

– Improved architecture improves accuracy and latency, better distant vehicle recall, reduced speed error at intersection by 20% and improved VRU accuracy by 20%.

– Changed the VRU Velocity network to two-phase, reducing latency and improving pedestrian velocity error by 6%.

– Converted the non-VRU Attributes network to two-phase, which reduced latency, reduced incorrect lane allocation of overtaking vehicles by 45% and incorrect parked predictions by 15%.

– Redesigned autoregressive Vector Lanes grammar to improve lane accuracy by 9.2%, lane recovery by 18.7% and fork recovery by 51.1%. Includes a full network update where all components were retrained with 3.8x the amount of data.

– Added a new “road marking” module to the Vector Lanes neural network that improves lane topology error at intersections by 38.9%.

– Updated the Occupancy Network to align with the road surface instead of the ego, for improved detection stability and improved recall on hill crest.

– Reduced candidate trajectory creation time by about 80% and improved fluidity by distilling an expensive trajectory optimization procedure into a lightweight designer’s neural network.

– Improved decision-making for short-term lane changes around explosions by modeling a richer trade-off between deviating from the route and the travel path required to drive through the driving zone.

– Reduced false pedestrian decelerations near intersections by using a better model for pedestrian kinematics.

– Added control to get a more accurate target geometry detected by the general use network.

– Improved maneuverability for vehicles that deviate from our intended route by better modeling their turning/lateral movements, avoiding unnatural decelerations.

– Improved longitudinal control when passing static obstacles by looking for possible vehicle motion profiles.

– Improved smoothness of longitudinal steering for in-lane vehicles in high relative speed scenarios by also considering relative acceleration in trajectory optimization.

– Reduced the best possible delay between the target’s photons and the control system by 26% due to adaptive planner timing, re-organizing trajectory selection and parallelizing observation computation. This way we can make faster decisions and improve reaction time.

FTC: We use automatic affiliate links that generate income. More.

Subscribe to Electrek on YouTube for exclusive videos and subscribe to the podcast.

#Tesla #releases #Full #SelfDriving #Beta #update #highlevel

Leave a Reply

Your email address will not be published. Required fields are marked *