Tesla has released a new beta software version of its Full Self-Driving Beta with a set of technical improvements for vehicles enrolled in the trial program.
According to ETNA via ArenaEV, this update, build number 2022.12.3.10, is considered one of the more extensive updates to the system and was developed based on more than 250,000 training clips extracted from the movements of Tesla fleet vehicles. At the time of this release, about 100,000 vehicle owners were actively participating in testing the software.
One significant change is improved vehicle decision-making in complex situations, particularly when turning left without a dedicated signal and crossing intersections. Tesla has also enhanced obstacle detection, understanding of road line geometry, lane edge recognition, and prediction of other vehicles’ movements to make system performance more reliable in various conditions.
In this version, detection of motorcycles, pedestrians, and cyclists has been improved and the error in estimating the speed and direction of some of these road users has been reduced. Detection of parked cars, vehicles with open doors, and distant moving objects has also been upgraded.
By adding new training data and redesigning some neural networks, Tesla has tried to make the vehicle’s reactions to yellow lights, lane changes, braking vehicles, and restricted paths more precise. The removal of three old neural networks has also increased the system’s processing rate by 1.8 frames per second.
Despite the name Full Self-Driving, this software remains at Level 2 Driver Assistance. Therefore, the driver must always monitor the road and be ready to take control of the vehicle. This update represents Tesla’s progress in processing driving data and improving system performance, but it does not eliminate the need for human supervision.