Tesla launches big fully autonomous driving beta update release

Tesla has launched a major new Fully Automated Driving (FSD) Beta Software Update with many improvements.

According to CEO Elon Musk, if this update “works”, automakers may have more access to the beta.

Tesla’s fully autonomous driving beta

Since October 2020, Tesla has been slowly rolling out what is known as the “Fully Automated Driving Beta” (FSD Beta). This is an early version of self-driving software currently being tested by a fleet selected by Tesla owners. Through the company and its “safety test scores”.

The software allows the vehicle to drive autonomously to the destination entered in the vehicle’s navigation system, but the driver must always be vigilant and controllable.

Despite its name, it is considered a Level 2 driver assistance system because the responsibility lies with the driver, not the Tesla system. In some updates, it was a “two-step forward, one-step backward” type program. Regression seen in terms of driving ability.

Tesla frequently releases new software updates for the FSD beta program, adding owners.

The last important update was FSD Beta 10.10 in early February.

As of the fourth quarter of 2021, automakers said they had nearly 60,000 owners in the FSD beta program.

Tesla FSD Beta 10.11

Currently, Tesla is starting to push the new FSD Beta 10.11 update to the Early Access program, which is an important update based on the release notes.

The release notes are as follows:

  • We have upgraded the modeling of lane geometry from a high density raster (“bag of points”) to an autoregressive decoder that directly predicts and connects “vector space” lanes point by point using a trans-neural network. This allows you to predict the intersecting lanes and reduce computational costs. It reduces post-processing-prone errors and opens the way for joint end-to-end prediction of many other signals and their relationships. More accurately predict where vehicles will turn or meet, reducing unnecessary slowdowns for non-intersecting vehicles. path.
  • Improved understanding of right of way when the map is inaccurate or the car cannot track the navigation. In particular, intersection range modeling is entirely based on network predictions and no longer uses map-based heuristics.
  • The accuracy of VRU detection has been improved by 44.9% and false positive pedestrians and bicycles (especially around tar seams, skid marks and raindrops) have been dramatically reduced. This was achieved by increasing the data size of the next generation automatic labeler and training network parameters. It was previously frozen and changed the network loss function. You can see that this reduces the incidence of VRU-related false slowdowns.
  • Reduced predicted speed error of very close motorcycles, scooters, wheelchairs and pedestrians by 63.6%. To do this, we introduced a new dataset of simulated hostile high-speed VRU interactions. This update improves autopilot control for high speed movement and disconnection. -At VRU.
  • Increased jerk at start and end of creeping and improved creeping profile.
  • Improved control of nearby obstacles by predicting the continuous distance to static geometry using a common static obstacle network.
  • By increasing the dataset size by 14%, we have reduced the vehicle’s “parking” attribute error rate by 17%. Also, the accuracy of the brake lights has been improved.
  • The speed error in the clear scenario was improved by 5% and the speed error in the highway scenario was improved by 10%, achieved by adjusting the loss function to improve performance in difficult scenarios.
  • Improved detection and control of open car doors.
  • Smoothness by turn by using an optimization-based approach to determine which road line is control-independent, taking into account lateral and longitudinal acceleration and jerk limits, and vehicle kinematics. Has been improved.
  • By optimizing the Ethernet data transfer pipeline by 15%, the stability of the FSDUl visualization has been improved.

CEO Elon Musk says that if this update “works”, Tesla will “probably” reduce access to drivers who score a “95” safety score.

For Canada, Tesla owners who purchased the FSD had access to the driver’s safety score for over a week. Automakers are expected to soon begin releasing beta to top-scoring people.

FTC: We use affiliate links for earning cars. more.

Subscribe to Electrek on YouTube to get exclusive videos and subscribe to podcasts.

About the author


Leave a Comment