Tesla has now pushed 2022.45.11 onboard software update in North America. This version mainly fixes the recall issue and the FSD Beta was updated to v11.3.2. Compared to the previous version, this push has a larger range of users.
Enabling FSD Beta on the freeway This unifies the visioning and planning stack on and off the freeway and replaces the traditional freeway stack, which is more than four years old. While Tesla’s traditional freeway stack relied on a single camera and single frame network and could only handle simple lane specific maneuvering logic, FSD Beta can leverage a multi-camera network and next generation planner to enable more complex agent interactions and less reliance on lanes, which is making way for more intelligent behavior, smoother control and better decision making, according to the report.
By capturing more data from the self-driving test fleet, recall in close-cut situations is improved by 15%, especially for large trucks and high yaw rate scenarios. In addition, FSD extended and tuned dedicated speed control for cut-in objects.
Improved body position in wide lanes by deflecting in the direction of an upcoming turn to allow other cars to go around.
Improved handling in high curvature or large truck scenarios by offsetting in the lane to maintain a safe distance from other vehicles on the road and to improve comfort.
Improved path blocking behavior for lane changes in dense traffic. It will now maintain a greater distance between vehicles in blocked lanes to reduce the probability of being blocked all around in congested traffic.
Improves lane changes in dense traffic scenarios by allowing higher acceleration during the alignment phase. This results in it being possible to more naturally find gaps to pass cars very close to itself in neighboring lanes.
Thanks to the lane guidance module integrated into the road edge and line network in the latest version, it enables smoother turns by improving the consistency of detection between lane, line and road edge predictions.
Improved accuracy in detecting the intention of other vehicles to move. Accuracy in the case of other vehicles turning is improved by 23%, and the probability of Autopilot falsely detecting a vehicle stopping in front of it is reduced by 12%.
Extended maximum trajectory optimization range for smoother control of high curvature roads and distant vehicles at high speeds.
Improved driving behavior when approaching a row of parked cars in a narrow lane; it will be more inclined to veer off but still stay in the lane rather than change lanes or slow down for no apparent reason.
Improved back-to-back lane change operations through better integration of vision-based positioning and coarse map lane data.
Text profiles have been added to the user interface to inform operations in the FSD scheme, and visualization before deceleration in the vehicle path has been improved to indicate the magnitude of acceleration during deceleration with different opacities and velocity-rendered v-words, and a solid line is also drawn at the location where the car is about to stop.
The recall and accuracy of target detection have been improved, with a significant 10% reduction in position error for half cards, 3% and 7% improvement in recall and accuracy for passing vehicles beyond 100 meters, and 5% improvement in motorcycle recall.
False offsets around objects in wide lanes and near intersections are reduced by improving object kinematic modeling in low-speed scenarios.
The position of the automatic blind spot camera is adjusted when FSD Beta is activated to prioritize Autopilot visualization, or you can drag the camera to save a custom position.
As a result of the recent recalls (#23V085 in the U.S. and #2023-063 in Canada), Tesla has improved the following specific behaviors in the FSD Beta.
- Improved the decision logic for passing through an intersection or stopping before a yellow light, taking into account the deceleration time required to stop, the time to enter and cross the intersection, and the distance to cross the intersection before the light transitions, resulting in more natural and user-friendly yellow light handling.
- Improved the longitudinal deceleration control mechanism at stop sign intersections to make the shifting process feel more intuitive and natural.
- Improved Tesla’s speed adjustment logic when entering certain speed zones by allowing early control of detected speed limit signs. The strength of the response when slowing down to a detected speed limit sign depends on the current speed and the difference between it and the speed indicated by the detected speed limit sign. Tesla will add a visual effect behind the speed limit icon in the user interface that alerts the driver when the vehicle is set to exceed the detected speed limit by more than 50%. Finally, Tesla removed the option for absolute speed limit offsets in FSD Beta; only percentage-based offsets are available.
- improved the behavior of certain situations where a Tesla may continue straight ahead in the turn lane.
Tesla has been working on a v11 release, which is critical because it’s the first time Tesla’s driver assistance system has used a single software stack to enable highway and city driving.
Last week, Tesla CEO Elon Musk said that FSD Beta v11 had significant improvements, but “one more small release needs to be polished before a big rollout,” and it looks like that’s the big release he was talking about.
Of course, this version has actually been delayed for several months, and Tesla originally planned to release it around Thanksgiving last year. About 400,000 owners in North America have beta access to FSD Beta, and it’s not clear when FSD will be available to more users or when it will expand to more countries and regions outside of North America.