November 2022 – While Tesla has long sold and promised the Full Self-Driving package which allows turns and navigation on city streets, it’s still considered to be in an early-stage beta test. Drivers must remain attentive while using the software.
FSD Beta – Wider Releases
Starting November 23th, 2022, Tesla has opened up the FSD Beta to all those in North America, regardless of safety score as long as they request it from their screen in-car.
December 1, 2022 (estimate) FSD Beta V11 (2022.40.5) is the greatly anticipated “single stack” version of FSD that combines the city and highway code:
November 20th, 2022 FSD Beta 10.69.3.1 (2022.36.20) rolled out to Beta testers. This release includes the updated Energy App, Detailed Energy Charger Info, Enhanced Tesla App Integration, Blind Spot Camera Location, Alternate Routes, and additional improvements.
October 2nd, 2022 FSD Beta 10.69.2.3 (2022.20.18) rolled out and is a bug fix from 10.69.2.2 (2022.20.17).
September 11th, 2022 FSD Beta 10.69.2 (2022.20.15) rolled out to testers and promises to be a big improvement for intersections, roads without map data, roundabouts, and even unprotected left turns. Release notes below.
The price of Full Self-Driving also increased from $12,000 to $15,000 on September 5th.
May 28nd, 2022 FSD Beta 10.12.2 (2022.12.3.20) rolled out broadly to 100K cars and more Tesla Safety Score users. Release notes below. This version has improvements for unprotected left turns, heavy traffic, and complex intersections. In addition, 10.12.2 increases the maximum speed to 85 MPH.
April 12th, 2022 – FSD Beta 10.11.2 (2022.4.5.21) began rolling out more widely in the US and includes some significant improvements – see release notes below. The 10.11 rollout has seen several fits and starts, beginning with 10.11, then 10.11.1, and now finally 10.11.2 being paused and then turned on again as of 4/12. Of note from the Tesla team:
- Elon: “Vector lanes is a particularly significant architectural improvement to Tesla AI”
- Karpathy: “TLDR a GPT-like Transformer is now predicting the lanes and their connectivity. This “direct to vector space” framework allows predictions to be jointly coherent (due to sequential conditioning) and v easily used by planner (due to sparsity). Excellent work from the team.”
February 18th, 2022 – FSD Beta 10.10.2 (2021.44.30.21) was released and removes the rolling stop setting after a NHTSA safety recall where they said it can ‘increase the risk of a crash’. See release notes below.
In addition, the Canada FSD “request access” button has been released as of 2022.4.5.4, which rolled out on 2/26/2022.
January 17th, 2022 – FSD Beta 10.9 (2021.44.30.10) has rolled out – release notes below. Elon Musk also let it be known that v11 should be released in February.
January 7th, 2022 – FSD Beta 10.8.1 – (2021.44.30.5) Small update to fix some issues but is also more lenient with driver strikes that result in FSD Beta to no longer be available (e.g. after not paying attention). With this release, the strikes are reset after each FSD release and increases the number of strikes to 5.
December 23rd, 2021 – FSD Beta 10.8 – (2021.44.25.6) Tesla skipped a broader FSD Beta 10.7 release and went straight to FSD Beta 10.8 (release notes below), which included waypoints and holiday goodies (see software updates for the holiday release) … and rolled out to 97 Safety Score users.
December 5th, 2021 – FSD Beta 10.6 (2021.36.8.9 and 2021.36.8.10) FSD Beta 10.6 started being released on 12/5 and included. improved object detection among other things. However, the release was paused to push out 10.6.1 to fix ‘annoying’ issues according to Elon Musk.
November 22nd, 2021 – FSD Beta 10.5 (2021.36.8.8) FSD Beta 10.5 began rolling out to those with a Tesla Safety Score of 98 on November 22nd. The release notes are below.
November 7th, 2021 – FSD Beta 10.4 (2021.36.8.5) FSD Beta 10.4 began rolling out on November 7th. It had some incremental improvements, but not quite ready to roll out to additional testers yet.
October 25th, 2021 – FSD Beta 10.3.1 (2021.36.5.3) was released after a fix to 10.3 which rolled out with a host of significant bugs (including false Automatic Emergency Braking, AEB at freeway speeds). Those FSD owners who requested access and have a greater than 99 Tesla Safety Score should be getting access. Release details are below.
October 11th, 2021 – FSD Beta 10.2 (2021.32.25) released to approximately 1,000 FSD owners who requested access AND had a safety score of 100/100.
FSD Beta Release Notes
FSD Beta 10.69.3 Includes the Following Updates:
- Upgraded the Object Detection network to photon count video streams and retrained all parameters with the latest auto-labeled datasets (with a special emphasis on low visibility scenarios).
- Improved the architecture for better accuracy and latency, higher recall of far away vehicles, lower velocity error of crossing vehicles by 20%, and improved VRU precision by 20%.
- Converted the VRU Velocity network to a two-stage network, which reduced latency and improved crossing pedestrian velocity error by 6%.
- Converted the non-VRU Attributes network to a two-stage network, which reduced latency, reduced incorrect lane assignment of crossing vehicles by 45%, and reduced incorrect parked predictions by 15%.
- Reformulated the autoregressive Vector Lanes grammar to improve the precision of lanes by 9.2%, recall of lanes by 18.7%, and recall of forks by 51.1%. Includes a full network update where all components were retrained with 3.8x the amount of data.
- Added a new “road markings” module to the Vector Lanes neural network which improves lane topology error at intersections by 38.9%.
- Upgraded the Occupancy Network to align with road surface instead of ego for improved detection stability and improved recall at hill crest.
- Reduced runtime of candidate trajectory generation by approximately 80% and improved smoothness by distilling an expensive trajectory optimization procedure into a lightweight planner neural network.
- Improved decision-making for short-deadline lane changes around gores by richer modeling of the trade-off between going off-route versus trajectory required to drive through the gore region.
- Reduced false slowdowns for pedestrians near crosswalks by using a better model for the kinematics of the pedestrian.
- Added control for more precise object geometry as detected by the general occupancy network.
- Improved control for vehicles cutting out of our desired path by better modeling of their turning/lateral maneuvers thus avoiding unnatural slowdowns.
- Improved longitudinal control while offsetting around static obstacles by searching over feasible vehicle motion profiles.
- Improved longitudinal control smoothness for in-lane vehicles during high relative velocity scenarios by also considering relative acceleration in the trajectory optimization.
- Reduced best-case object photon-to-control system latency by 26% through adaptive planner scheduling, restructuring of trajectory selection, and parallelizing perception compute. This allows us to make quicker decisions and improves reaction time.
FSD Beta 10.69 Includes the Following Updates:
- Added a new “deep lane guidance” module to the Vector Lanes neural network which fuses features extracted from the video streams with coarse map data, i.e. lane counts and lane connectivities. This architecture achieves a 44% lower error rate on lane topology compared to the previous model, enabling smoother control before lanes and their connectivities become visually apparent. This provides a way to make every Autopilot drive as good as someone driving their own commute, yet in a sufficiently general way that adapts for road changes.
- Improved overall driving smoothness, without sacrificing latency, through better modeling of system and actuation latency in trajectory planning. The trajectory planner now independently accounts for latency from steering commands to actual steering actuation, as well as acceleration and brake commands to actuation. This results in a trajectory that is a more accurate model of how the vehicle would drive. This allows better downstream controller tracking and smoothness while also allowing a more accurate response during harsh maneuvers.
- Improved unprotected left turns with more appropriate speed profile when approaching and exiting median crossover regions, in the presence of high-speed cross traffic (“Chuck Cook style” unprotected left turns). This was done by allowing optimisable initial jerk, to mimic the harsh pedal press by a human, when required to go in front of high speed objects. Also improved lateral profile approaching such safety regions to allow for better pose that aligns well for exiting the region. Finally, improved interaction with objects that are entering or waiting inside the median crossover region with better modeling of their future intent.
- Added control for arbitrary low-speed moving volumes from Occupancy Network. This also enables finer control for more precise object shapes that cannot be easily represented by a cuboid primitive. This required predicting velocity at every 3D voxel. We may now control for slow-moving UFOs.
- Upgraded Occupancy Network to use video instead of images from single time step. This temporal context allows the network to be robust to temporary occlusions and enables prediction of occupancy flow. Also, improved ground truth with semantics-driven outlier rejection, hard example mining, and increasing the dataset size by 2.4x.
- Upgraded to a new two-stage architecture to produce object kinematics (e.g. velocity, acceleration, yaw rate) where network compute is allocated O(objects) instead of O(space). This improved velocity estimates for far away crossing vehicles by 20%, while using one tenth of the compute.
- Increased smoothness for protected right turns by improving the association of traffic lights with slip lanes vs yield signs with slip lanes. This reduces false slowdowns when there are no relevant objects present and also improves yielding position when they are present.
- Reduced false slowdowns near crosswalks. This was done with improved understanding of pedestrian and bicyclist intent based on their motion.
- Improved geometry error of ego-relevant lanes by 34% and crossing lanes by 21% with a full Vector Lanes neural network update. Information bottlenecks in the network architecture were eliminated by increasing the size of the per-camera feature extractors, video modules, internals of the autoregressive decoder, and by adding a hard attention mechanism which greatly improved the fine position of lanes.
- Made speed profile more comfortable when creeping for visibility, to allow for smoother stops when protecting for potentially occluded objects.
- Improved recall of animals by 34% by doubling the size of the auto-labeled training set.
- Enabled creeping for visibility at any intersection where objects might cross ego’s path, regardless of the presence of traffic controls.
- Improved accuracy of stopping position in critical scenarios with crossing objects, by allowing dynamic resolution in trajectory optimization to focus more on areas where finer control is essential.
- Increased recall of forking lanes by 36% by having topological tokens participate in the attention operations of the autoregressive decoder and by increasing the loss applied to fork tokens during training.
- Improved velocity error for pedestrians and bicyclists by 17%, especially when ego is making a turn, by improving the onboard trajectory estimation used as input to the neural network.
- Improved recall of object detection, eliminating 26% of missing detections for far away crossing vehicles by tuning the loss function used during training and improving label quality.
- Improved object future path prediction in scenarios with high yaw rate by incorporating yaw rate and lateral motion into the likelihood estimation. This helps with objects turning into or away from ego’s lane, especially in intersections or cut-in scenarios.
- Improved speed when entering highway by better handling of upcoming map speed changes, which increases the confidence of merging onto the highway.
- Reduced latency when starting from a stop by accounting for lead vehicle jerk.
- Enabled faster identification of red light runners by evaluating their current kinematic state against their expected braking profile.
FSD Beta 10.12.2 Includes the Following Updates:
- Upgraded decision making framework for unprotected left turns with better modeling of objects’ response to ego’s actions by adding more features that shape the go/no-go decision. This increases robustness to noisy measurements while being more sticky to decisions within a safety margin. The framework also leverages median safe regions when necessary to maneuver across large turns and accelerating harder through maneuvers when required to safely exit the intersection.
- Improved creeping for visibility using more accurate lane geometry and higher resolution occlusion detection.
- Reduced instances of attempting uncomfortable turns through better integration with object future predictions during lane selection.
- Upgraded planner to rely less on lanes to enable maneuvering smoothly out of restricted space.
- Increased safety of turns with crossing traffic by improving the architecture of the lanes neural network which greatly boosted recall and geometric accuracy of crossing lanes.
- Improved the recall and geometric accuracy of all lane predictions by adding 180k video clips to the training set.
- Reduced traffic control related false slowdowns through better integration with lane structure and improved behavior with respect to yellow lights.
- Improved the geometric accuracy of road edge and line predictions by adding a mixing/coupling layer with the generalized static obstacle network.
- Improved geometric accuracy and understanding of visibility by retraining the generalized static obstacle network with improved data from the autolabeler and by adding 30k more videos clips.
- Improved recall of motorcycles, reduced velocity error of close-by pedestrians and bicyclists, and reduced heading error of pedestrians by adding new sim and autolabeled data to the training set.
- Improved precision of the “is parked” attribute on vehicles by adding 41k clips to the training set. Solved 48% of failure cases captured by our telemetry of 10.11.
- Improved detection recall of far-away crossing objects by regenerating the dataset with improved versions of the neural networks used in the autolabeler which increased data quality.
- Improved offsetting behavior when maneuvering around cars with open doors.
FSD Beta 10.11.2 Includes the Following Updates:
- Upgraded modeling of lane geometry from dense rasters (“bag of points”) to an autoregressive decoder that directly predicts and connects “vector space” lanes point by point using a transformer neural network. This enables us to predict crossing lanes, allows computationally cheaper and less error-prone post-processing, and paves the way for predicting many other signals and their relationships jointly and end-to-end.
- Use more accurate predictions of where vehicles are turning or merging to reduce unnecessary slowdowns for vehicles that will not cross our path.
- Improved right-of-way understanding if the map is inaccurate or the car cannot follow the navigation. In particular, modeling intersection extents is now entirely based on network predictions and no longer uses map-based heuristics.
- Improved the precision of VRU detections by 44.9%, dramatically reducing spurious false positive pedestrians and bicycles (especially around tar seams, skid marks, and rain drops). This was accomplished by increasing the data size of the next-gen auto-labeler, training network parameters that were previously frozen, and modifying the network loss functions. We find that this decreases the incidence of VRU-related false slowdowns.
- Reduced the predicted velocity error of very close-by motorcycles, scooters, wheelchairs, and pedestrians by 63.6%. To do this, we introduced a new dataset of simulated adversarial high-speed VRU interactions. This update improves autopilot control around fast-moving and cutting-in VRUs.
- Improved creeping profile with higher jerk when creeping starts.
- Improved control for nearby obstacles by predicting continuous distance to static geometry with the general static obstacle network.
- Reduced vehicle “parked” attribute error rate by 17%, achieved by increasing the dataset size by 14%.
- Improved clear-to-go scenario velocity error by 5% and highway scenario velocity error by 10%, achieved by tuning loss function targeted at improving performance in difficult scenarios.
- Improved detection and control for open car doors.
- Improved smoothness through turns by using an optimization-based approach to decide which road lines are irrelevant for control given lateral and longitudinal acceleration and jerk limits as well as vehicle kinematics.
- Improved stability of the FSD Ul visualizations by optimizing the ethernet data transfer pipeline by 15%.
FSD Beta 10.10 Includes the Following Updates:
- Smoother fork maneuvers and turn-lane selection using high fidelity trajectory primitives.
- Disabled rolling-stop functionality in all FSD Profiles. This behavior used to allow the vehicle to roll through all-way-stop intersections, but only when several conditions were met, including: vehicle speed less than 5.6 mph, no relevant objects / pedestrians /bicyclists detected, sufficient visibility and all entering roads at the intersection have speed limits below 30mph.
- Improved generalized static object network by 4% using improved ground truth trajectories.
- Improved smoothness when stopping for crossing objects at intersections by modeling soft and hard constraints to better represent urgency of the slowdown.
- Enabled lane changing into an oncoming lane to maneuver around static obstacles, when safe to do so.
- Improved smoothness for merge handling by enforcing more consistency with previous cycle’s speed control decisions.
- Improved handling of flashing red light traffic controls by adding more caution for events where crossing vehicles may not stop.
- Improved right of way understanding at intersections with better modeling of intersection extents.
FSD Beta 10.9 Includes the Following Updates:
- Improved intersection extents and right of way assignment by updating modeling of intersection areas from dense rasters (“bag of points”) to sparse instances. Increased intersection region IOU by 4.2%. The sparse intersection network is the first model deployed with an auto-regressive architecture that runs natively with low latency on the TRIP Al accelerator chip, through innovations in the Al compiler stack.
- Upgraded generalized static object network to use 10-bit photon count streams rather than 8-bit ISP tonemapped images by adding 10-bit inference support in the Al compiler stack. Improved overall recall by 3.9% and precision by 1.7%.
- Made unprotected left turns across oncoming lanes more natural by proceeding straight into intersection while yielding, before initiating the turn.
- Improved lane preference and topology estimation by 1.2% with a network update and a new format for navigation clues.
- Improved short deadline lane changes with better modelling of necessary deceleration for maneuvers beyond the lane change,
- Improved future paths for objects not confined to lane geometry by better modelling of their kinematics.
- Made launches from stop more calm when there is an imminent slowdown nearby.
- Improved gap selection when yielding to a stream of oncoming cars on narrow roads.
FSD Beta 10.8 and 10.8.1 Include the Following Updates:
- Improved object attributes network to reduce false cut-in slowdowns by 50% and lane assignment error by 19%.
- Improved photon-to-control vehicle response latency by 20% on average.
- Expanded use of regenerative braking in Autopilot down to O mph for smoother stops and improved energy efficiency.
- Improved VRU (pedestrians, bicyclists, motorcycles, animals) lateral velocity error by 4.9% by adding more auto-labeled and simulated training examples to the dataset.
- Reduced false slowdowns for crossing objects by improved velocity estimates for objects at the end of visibility.
- Reduced false slowdowns by adding geometric checks to cross-validate lane assignment of objects.
- Improved speed profile for unprotected left turns when visibility is low.
- Added more natural behavior to bias over bike lanes during right turns.
- Improved comfort when yielding to jaywalkers by better modelling of stopping region with soft and hard deadlines.
- Improved smoothness for merge control with better modelling of merge point and ghost objects positioned at the edge of visibility.
- Improved overall comfort by enforcing stricter lateral jerk bounds in trajectory optimizer.
- Improved short deadline lane changes through richer trajectory modeling.
- Improved integration between lead vehicle overtake and lane change gap selection.
- Updated trajectory line visualization.
FSD Beta 10.6 Includes the Following Updates:
- Improved object detection network architecture for non-VRUs (eg.cars, trucks, buses). 7% higher recall, 16% lower depth error, and 21% lower velocity error for crossing vehicles.
- New visibility network with 18.5% less mean relative error
- New general static object network with 17% precision improvements in high curvature and nighttime cases.
- Improved stopping position at unprotected left turns while yielding to oncoming objects, using object predictions beyond the crossing point.
- Allow more room for longitudinal alignment during merges by incorporating modelling of merge region end.
- Improved comfort when offsetting for objects that are cutting out of your lane.
FSD Beta 10.5 Includes the Following Updates:
- Improved VRU (pedestrians, bicyclists, motorcycles) crossing velocity error by 20% from improved quality in our auto-labeling,
- Improved static world predictions (road lines, edges, and lane connectivity) by up to 13% using a new static world auto-labeler and adding 165K auto-labeled videos
- Improved cone and sign detections by upreving the generalized static object network with 15K more video clips and adjusting oversampling and overweighting strategies (+4.5% precision, +10.4% recall).
- Improved cut-in detection network by 5.5% to help reduce false slowdowns.
- Enabled “emergency collision avoidance maneuvering” in shadow mode.
- Enabled behavior to lane change away from merges when safe to do so.
- Improved merge object detection recall by using multi-modal object prediction at intersections.
- Improved control for merges by increasing smoothness of arrival time constraints and considering possible merging objects beyond visibility.
- Improved lane changes by allowing larger deceleration limit in short-deadline situations.
- Improved lateral control for creeping forward to get more visibility.
- Improved modeling of road boundaries on high curvature roads for finer maneuvers.
- Improved logic to stay on-route and avoid unnecessary detours/rerouting.
FSD Beta 10.4 Includes the Following Updates:
- Improved handling when driving off navigation route by allowing better recovery, when safe to do so.
- Improved handling and detection of high speed objects when crossing high speed roads. Enabled faster acceleration across high speed roads.
- Improved speed through narrow spaces surrounded by high obstacles.
- Improved static obstacle control by upreving the generalized static object network with hyperparameter tuning and improvements for oversampling strategies (+1.5% precision, +7.0% recall)
- Improved VRU detection (e.g. pedestrians, bicyclists, motorcycles) by adding data from next generation autolabeler (precision +35%, recall +20%).
- Improved emergency vehicle detection network by adding new data and improved training regime (pass rate +5.8%).
- Improved VRU control relevance attribute by adding navigation route as input to object detection network (accuracy + 1.1%)
FSD Beta 10.3 Included the Following Updates:
- Added FSD Profiles that allow drivers to control behaviors like rolling stops, exiting passing lanes, speed-based lane changes, following distance and yellow light headway.
- Added planning capability to drive along oncoming lanes to maneuver around path blockage.
- Improved creeping speed by linking speed to visibility network estimation and distance to encroachment point of crossing lanes.
- Improved crossing object velocity estimation by 20% and yaw estimation by 25% by upreving surround video vehicle network with more data. Also increased system frame rate by +1.7 frames per second.
- Improved vehicle semantic detections (e.g. brake lights, turn indicators, hazards) by adding +25k video clips to the training data set.
- Improved static obstacle control by upreving the generalized static object network with 6k more video clips (+5.6% precision, +2.5% recall).
- Allowed more acceleration when merging from on-ramps onto major roads and when lane changing from slow to fast lanes.
- Reduced false slowdowns and improved offsetting for pedestrians by improving the model of interaction between pedestrians and the static world.
- Improved turning profile for unprotected turns by allowing ego to lane lines more naturally, when safe to do so.
- Improved speed profile for boosting onto high-speed roads by enforcing stricter longitudinal and lateral acceleration limits required to beat the crossing objects.
FSD Beta Request Button
On September 24th Tesla released the 2021.32.22 software update that included a new button for those who purchased the Full Self-Driving package and have compatible vehicles with Hardware 3. Those customers had a new “Request Full Self-Driving Beta” button under Controls > Autopilot.
For those interested in joining the FSD Beta program, you’ll need to agree to allow Tesla to monitor your driving and calculate a safety score.
Tesla will monitor your driving behavior for at least seven days.
Once you sign up to the program, your Tesla app will show you how you’re doing so you can improve (iPhone only for now):
If you participate, be sure to read the Tesla Safety Score Beta information posted on their website to understand exactly what they’re tracking and how it works.
- Keep a good amount of following distance (especially above 50 MPH)
- Brake gently, ideally just using regenerative braking.
- Keep an eye out for anything ahead on the road to avoid Forward Collision Warnings
- Don’t make aggressive turns.
Note that driving on Autopilot will not be included in safety score calculations but the miles driven will be included in the total, according to Tesla. This means that you may want to use Autopilot on the freeway, especially if you have a hard time maintaining following distances to other cars.
After at least seven days of good driving, you’ll be eligible to join the FSD Beta program. It’s still unclear whether that will happen immediately after the seven days or whether they will gradually add people to the program over time (guess the latter).
Where Tesla Drivers Have the Most Issues
According to the Tesla Safety Score data in the app, most people have problems in the following areas:
- Unsafe Following – 15%
- Aggressive Turning – 3%
- Hard Braking – 2%
- Forward Collision Warnings – 10 per 1,000 miles
So be sure to leave lots of room between you and the car in front of you when traveling greater than 50 MPH. Per Tesla:
Removal from FSD Beta
Once you’re in FSD Beta, you still have to pay careful attention while driving, otherwise, you may be kicked out. Some in the FSD Beta program have received emails from Tesla for inattention while driving:
Specifically, while using the FSD Beta feature, you or another driver of your vehicle received:
- Two or more “strikeouts,” which resulted in the loss of Autopilot availability for that drive; or
- At least one “strike” per 5 km (about 3 miles) driven on Autopilot, which is a visual and audible warning that requires attention.
This is your only warning to please keep your hands on the wheel and remain attentive at all times when using Autopilot. The car is not autonomous, and if you aren’t paying attention, a crash could happen, and you or others could get hurt, or worse, so failure to abide by this warning will result in removal of the FSD Beta feature from your vehicle.
The Tesla Team
So be sure to pay attention and keep your hands on the wheel (or yoke!).
Beta Tester Videos
In the meantime, check out the myriad of videos being uploaded to YouTube on a daily basis or the ones highlighted below.
Here’s a good overview of the current state in the form of a ‘training video’:
FSD Beta Overview
Tesla’s Full Self-Driving option has long held the promise of being able to automatically navigate with Autopilot on any road, including city streets; even so far as becoming a Level 5 (see What do Levels Mean?), fully autonomous system. Customers who have spent thousands of dollars for the expensive Full Self-Driving package expecting city street Autopilot are also eagerly awaiting this functionality.
That said, it’s been a long road to achieving autonomous driving as navigating on any open road under any condition, is one of the most complex tasks for any AI system to handle. Originally Elon Musk targeted a fully autonomous trip from Los Angeles to New York in 2017, only to drop that goal altogether later on. Then, Full Self-Driving was targeted to be “feature-complete” by the end of 2019, only to move the timeline further out.
Eventually, we learned that Tesla was doing a major overhaul of the AI-based Autopilot system, a “rewrite”, so-to-speak. In late 2020, this rewrite, known as “Full Self-Driving Beta” (or FSD Beta), started rolling out to a small select group of testers with a likely broader release in 2021 or early 2022. While the Full Self-Driving package will have the ability to automatically turn on city streets, despite its name, it will not be fully autonomous (Level 5), and will still require an attentive driver for the foreseeable future.
Full Self-Driving Rewrite Creates 4D Environment
At the heart of the new Full Self-Driving rewrite is the capability for the neural network computer to utilize all eight cameras around the car, allowing it to create a virtual 3D (4D with time) environment for better situational awareness.
Full Self-Driving on City Streets
Creating this 3D model (4D with time) is especially critical for navigating city streets since that environment is far more complex than freeway driving. It will allow the vehicle to make complex maneuvers as mentioned by Tesla in the Beta release notes:
“When Full Self-Driving is enabled your vehicle will make lane changes off-highway, select forks to follow your navigation route, navigate around other vehicles and objects, and make left and right turns.“
That said, because this is a software rewrite and the entire system needs to be tested, not only the new features, early beta testers are being warned to be extra vigilant:
“Full Self-Driving is in early limited access beta and must be used with additional caution. It may do the wrong thing at the worst time, so you must always keep your hands on the wheel and pay extra attention to the road. Do not become complacent.“
Early Look at Full Self-Driving Beta
Beta testers are uploading tons of videos on YouTube with each release so you can virtually ride along to see how things are improving:
As you can see, this is still very early release testing software, likely still a long way off from public release, but has impressive capabilities.
Rapid But Cautious Updates
The team at Tesla appears to be making rapid progress, releasing updates to Beta testers every 5 to 10 days, according to Elon Musk, so improvements should come quickly.
However, after the problematic FSD Beta 10.4 release, which causes phantom Automatic Emergency Braking on the freeway, dangerous enough that Tesla had to roll it back, Tesla is being more cautious with future releases and testing.
Full Self-Driving Rewrite Release
While the Full Self-Driving update still appears to be a little ways off, it’s great to see real progress being made. That said, it’s important to remember that “Full Self-Driving” still very much requires an attentive driver and is not a “Level 5” fully autonomous driving system, despite the name. However, it’s great that Tesla continues to push the envelope in this area and provide exciting new Autopilot updates and finally bring significant new assisted driving functionality to those who purchased the Full Self-Driving package. We’re excited to see what the future holds!