Tesla Rolls Back Latest FSD Beta Version

Tesla Rolls Back Latest FSD Beta Version

Before the latest release of FSD Beta over the weekend, Tesla's intention to test its Full Self Driving sophisticated driver aid software with untrained vehicle owners on public roads drew scrutiny and controversy.

FSD Beta version 10.3 began going out late Saturday and early Sunday. It include a number of notable features. However, on Sunday afternoon Tesla's Technoking, Elon Musk tweeted that Tesla is “Seeing some issues with 10.3, so rolling back to 10.2 temporarily.”


Apparent Issues

While numerous drivers have already published videos and thoughts of the release on social media, some testers claim that the rollback update completely eliminates the FSD beta features from their cars. Some other users claimed that the 10.3 upgrade introduced phantom forwards collision warnings (FCW), additional complaints included a missing Autosteer option, TACC troubles, and occasional AutoPilot panic.

If the problem is widespread within the test group, a phantom FCW would almost certainly warrant a rollback. A Mazda 3 recall was issued in 2019 to resolve issues with the Smart Braking System incorrectly recognizing obstacles in the car's path. Vehicles that suddenly slam on the brakes for no apparent reason, as some social media posts claim, could easily cause an accident if another automobile follows closely behind. Another issue for beta testers is that some claimed the bogus FCW occurrences dropped their Tesla-graded "safety score" to the point where they might not be able to continue in the programme.

Tesla issued the latest FSD Beta, v10.3, early Sunday morning that included several features and existing beta testers were able to get the update and users who had a high Safety Score of at least 100.


FSD Beta 10.3 Features

In 10.3, Tesla added Full Self-Driving Profiles, or driving styles. There are three options to pick from, Chill, Average and Assertive.

Choosing a driving style will alter how Autopilot navigates and interacts with its surroundings. Changing your FSD profile, for example, affects your car's following distance, acceleration, lane changes, and more.

    Your automobile will leave a longer following space and be more 'mellow' in Chill mode. If the automobile in front of you slows down instead of going around them, it will stay in its lane more often. Complete pauses will be possible, as well as more gradual acceleration.

    The car will have a medium following distance in Average mode, and Tesla claims that it will be capable of rolling stops. Tesla is referring to choice stops rather than stop signs, such as driving out of a driveway or parking lot. If the road is clear, instead of coming to a complete stop, your car may simply slow down. When compared to Chill mode, you can also expect speedier acceleration.

    The last driving profile is Assertive. The following distance will be shorter in this profile, and your automobile will try to keep as much speed as possible by moving into open lanes more frequently. According to Tesla, the vehicle will not exit passing lanes. As a result, if you're travelling on the highway, your automobile will stay in the left lane.

    Other features included:

    • Driving along oncoming lanes to avoid path blockage has been included to the planning capability.
    • Increased creeping speed by tying it to visibility network estimation and the distance to the crossing lane's incursion point.
    • By adding more data to the surround video vehicle network, we were able to improve crossing object velocity estimate by 20% and yaw estimation by 25%. Additionally, the system frame rate was improved by 1.7 frames per second.
    • Added +25k video clips to the training data set, which improved vehicle semantic detections (e.g. brake lights, turn indicators, and hazards).
    • Improved static obstacle handling by adding 6k more video clips to the generalised static object network (+5.6% precision, +2.5% recall).
    • Increased acceleration while merging from on-ramps to major roadways and changing lanes from slow to rapid.
    • Improved the model of interaction between pedestrians and the static world, which reduced erroneous slowdowns and improved offsetting for pedestrians.
    • Improved unprotected turning profile by allowing ego to cross lane lines more easily when it's safe to do so.
    • A better speed profile for boosting onto high-speed roadways by imposing stronger longitudinal and lateral acceleration limitations to beat the crossing object.

    Unfortunately, users will have to wait for Tesla to iron out some kinks before making these features available again to the FSD Beta customers.


    *Images source credit to @chazman/Twitter.