Thousands of Tesla customers have been anxiously awaiting the arrival of Tesla’s FSD beta software update and today, we have the first concrete evidence the software has arrived.
Andy Dowling first noticed it this morning, posting a map from the 3rd party service TeslaFi that tracks software updates across the world. This was quickly followed up by Teslascope which offers a competing service.
The software build in question is version 2023.12.10 which we know contains FSD Beta.
- TeslaScope‘s data shows the build was installed on a Tesla Model 3 Standard Range at 2023-05-24 9:21PM.
- TeslaFi‘s data shows the build was installed on a Tesla Model Y Standard Range Plus (LFP) at 05/24/2023 04:21 PM. TeslaFi provides some additional information, including that the MY has 945km on the odometer and that the car is located in the ACT.
It is incredibly likely that these vehicles are both Tesla company cars, used for testing the release.
Tesla’s Full Self Driving package has been sold in Australia for a number of years and while the software offers a number of benefits like Navigation on Autopilot, Smart Summon, Traffic Light and Stop Sign Control, the most significant change to the driving experience, Autosteer on city streets is still listed in the Upcoming section on the website.
For those who pay attention to the development of the software, they look to the US and Canada where FSD Beta, the latest, most capable builds are deployed and watch with envy, hoping it’ll be released here.
While today’s confirmation that FSD Beta is on at least 2 cars is exciting, it doesn’t confirm a timeline for rollout to customer cars, but does give hope it’ll be this calendar year.
FSD Beta allows the car to turn corners, take roundabouts, and drive on unmarked roads and has dramatically increased performance when it comes to lane selection to follow a route. The biggest difference between FSD (preview) that we have now, versus the FSD Beta builds, is that the code base has shifted considerably, from a largely heuristics-based combination of if this then that logic, to run almost entirely on Neural Networks.
This enables the car to have very different abilities, like crossing lane lines when safe to do so, to move around an obstacle like a double-parked car. FSD Beta is much closer to human capabilities and in terms of monitoring the changing environment around the car, far exceeds human ability.
FSD Beta V11 was a significant shift in capability and it is important to note that the build seen today on Australian cars is v11.3.6, not even the latest build which is FSD Beta v11.4.3 (or 2023.6.15). In theory, now the first Beta cars have arrived, they should get updated with new builds and we should see that reflected through services like TeslaFi and TeslaScope.
Teslascope is also reporting the release has been seen in other locations outside the US and Canada, including Germany and Belgium.
2023.12.10 / Full Self-Driving (Beta) v11.3.6 Release Notes
- Enabled FSD Beta on highway. This unifies the vision and planning stack on and off-highway and replaces the legacy highway stack, which is over four years old. The legacy highway stack still relies on several single-camera and single-frame networks, and was setup to handle simple lane-specific maneuvers. FSD Beta’s multi-camera video networks and next-gen planner, that allows for more complex agent interactions with less reliance on lanes, make way for adding more intelligent behaviors, smoother control and better decision making.
- Improved recall for close-by cut-in cases by 15%, particularly for large trucks and high-yaw rate scenarios, through an additional 30k auto-labeled clips mined from the fleet. Additionally, expanded and tuned dedicated speed control for cut-in objects.
- Improved the position of ego in wide lanes, by biasing in the direction of the upcoming turn to allow other cars to maneuver around ego.
- Improved handling during scenarios with high curvature or large trucks by offsetting in lane to maintain safe distances to other vehicles on the road and increase comfort.
- Improved behavior for path blockage lane changes in dense traffic. Ego will now maintain more headway in blocked lanes to hedge for possible gaps in dense traffic.
- Improved lane changes in dense traffic scenarios by allowing higher acceleration during the alignment phase. This results in more natural gap selection to overtake adjacent lane vehicles very close to ego.
- Made turns smoother by improving the detection consistency between lanes, lines and road edge predictions. This was accomplished by integrating the latest version of the lane-guidance module into the road edge and lines network.
- Improved accuracy for detecting other vehicles’ moving semantics. Improved precision by 23% for cases where other vehicles transition to driving and reduced error by 12% for cases where Autopilot incorrectly detects its lead vehicle as parked. These were achieved by increasing video context in the network, adding more data of these scenarios, and increasing the loss penalty for control- relevant vehicles.
- Extended maximum trajectory optimization horizon, resulting in smoother control for high curvature roads and far away vehicles when driving at highway speeds.
- Improved driving behavior next to row of parked cars in narrow lanes, preferring to offset and staying within lane instead of unnecessarily lane changing away or slowing down.
- Improved back-to-back lane change maneuvers through better fusion between vision-based localization and coarse map lane counts.
- Added text blurbs in the user interface to communicate upcoming maneuvers that FSD Beta plans to make. Also improved the visualization of upcoming slowdowns along the vehicle’s path. Chevrons render at varying opacity and speed to indicate the slowdown intensity, and a solid line appears at locations where the car will come to a stop.
- Improved the recall and precision of object detection, notably reducing the position error of semi-trucks by 10%, increasing the recall and precision of crossing vehicles over 100m away by 3% and 7%, respectively, and increasing the recall of motorbikes by 5%. This was accomplished by implementing additional quality checks in our two million video clip autolabeled dataset.
- Reduced false offsetting around objects in wide lanes and near intersections by improving object kinematics modeling in low speed scenarios.
- Adjusted position of Automatic Blind Spot Camera when FSD Beta is active to prioritize the Autopilot visualization. Drag the camera to save custom positions.