Tesla FSD Beta 11.3 goes official, check the complete release notes

According to the latest report, Tesla is now released the Tesla FSD Beta 11.3. This update was originally planned in November 2022 but some reasons, the update delayed 2 month. This update will be available to all Tesla owners in the near future. You can check the complete changelog which is mentioned below.

Tesla FSD Beta 11.3 Updatelog:

  • oversampling 180K challenging videos including rain reflections, road debris and high curvature.
  • Increased the recall of close stopper cases by 20% by adding 40,000 automatically labeled fleet clips of this scene to the dataset. Handling of jammed cases has also been improved by improving their motion model, leveraging the same model for smoother lateral and vertical control of jammed objects.
  • Added “Lane Guidance Module and Perceptual Loss” for road edges and lane networks, improving absolute recall for lanes by 6% and absolute recall for road edges by 7%.
  • Improved the overall geometry and stability of lane predictions by updating the Lane Guidance module representation with information about predicted intersections and oncoming lanes.
  • Improved handling at high speeds and high curvature situations by offsetting inward lane lines.
  • Improved lane changes, including: earlier detection and handling of simultaneous lane changes, better gap selection when approaching blocked lanes, better integration of speed-based and navigation-based lane change decisions, and FSD driving profiles with speed More differences between lane changes.
  • Improved the smoothness of the longitudinal control response when following a vehicle in front by better simulating the possible impact of the brake lights of the vehicle in front on its future speed profile.
  • Improves rare object detection by 18% and reduces depth error for large trucks by 9%, mainly due to migration to more supervised auto-labeled datasets.
  • Improves semantic detection of school buses by 12% and that of stationary to moving vehicles by 15%. This was achieved by improving dataset labeling accuracy and increasing dataset size by 5%.
  • Improved decision-making at pedestrian crossings by replacing approximate kinematic models with neural network-based self-trajectory estimation.
  • Improved the reliability and smoothness of merge control by deprecating the traditional merge region task in favor of a merge topology derived from vector lanes.
  • Unlocks longer fleet telemetry clips (up to 26%) by balancing compressed IPC buffers and optimized write scheduling across dual SOCs.

Leave a Comment