On November 22nd, on November 21st, a Tesla employee who received the update confirmed that Tesla has begun to release the beta version of the FSD Beta 10.5 advanced driver assistance system, but it is currently only open to internal employees. As can be seen from the update instructions, FSD Beta 10.5 has greatly improved the functions of the previous system.
Earlier this month, Musk hinted that FSD Beta 10.5 will be released around November 20th. On November 19, Musk said on his personal social media Twitter that he hoped that FSD Beta 10.5 would be ready soon. A tweet earlier in the day said We have confirmed to a Tesla employee who received this update tonight that this is FSD Beta 10.5. Tesla is pushing the FSD Beta 10.5 system update to employees’ vehicles. They have stated that if the final inspection goes well, they are expected to announce a new product on Sunday night.
It didn’t take long for the Tesla community to report that the latest generation of advanced driver assistance systems will be released soon. The media tracking Tesla’s update service subsequently said on Twitter that they have confirmed that FSD Beta 10.5 has indeed been released, but it is currently only open to employees.
In these reports, FSD Beta 10.5 related update instructions are also announced, accurately describing what aspects of the system have been improved during the update. It can be seen from the updated information that Tesla has once again improved the test version of FSD’s ability to deal with vulnerable road users (VRU), as well as its ability to detect traffic cones and traffic signs.
More importantly, the FSD Beta 10.5 system can also automatically change lanes to stay away from other vehicles that are merging, without slowing down. Under the control of the FSD Beta 10.5 system, the “false deceleration” of the vehicle has also been reduced. In general, related improvements will make FSD Beta 10.5 behave more like human driving.
The following is a partial update description of FSD Beta 10.5:
- Improve the system’s automatic marking function to reduce the speed error of overtaking vulnerable road users (pedestrians, bicycles, motorcycles) by 20%.
- Adopt new still life automatic marking and add 165K automatic marking video to improve the recognition accuracy of static markings such as lane dividing lines, road edge lines and lane connecting lines by 13%.
- Improve the detection capabilities of traffic cones and traffic signs.
- Improve the detection capability of vehicles in the queue by 5.5% to help reduce false decelerations
- Enable “emergency evasion maneuver” function in shadow mode
- Allow to change lanes under safe conditions to avoid merging vehicles
- The use of multi-modal target prediction at intersections improves the error rate of parallel target detection
- Improve the deceleration threshold in a short time and improve the lane change function
- Improve the curve modeling function to make the vehicle maneuver more precise
- Avoid unnecessary detours or re-planning of routes.