Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

FSD Beta 10.69

This site may earn commission on affiliate links.
Thats so interesting. In California you can use center turn lanes to start and to complete left turns. I had no idea that other states don’t allow you to use it to complete left turns.
You can't use it to pull into to stop and for a merging into traffic. It is for left turning traffic (on the road) only. Here is Georgia code.

40-6-126 - Central lane for turning​

Whenever a highway or roadway has a central lane in which traffic may enter from either direction for the purposes of making a left turn, no vehicle shall be driven into such central lane except for the purpose of making a left turn, and no vehicle shall enter into such central lane at a location which is more than 300 feet from the location where the vehicle will turn left across one or more lanes of oncoming traffic. No vehicle which has been driven into such a central lane shall be operated in such central lane for more than 300 feet.
 
And what is “perfection”?

My takeaway from watching the first couple of videos is that it works great. The comments on the video are all like “ a little slow, a little jerky, a little close, a little…” I bet if my wife made a video of how I drive she could have made the same responses as could I if she would have driven.
If you expect a carbon copy of how you would have driven a route than you will be in for a disappointment. The main objective is that it drives safely from A to B and I think Tesla is on the verge of accomplishing that.
For me, ’perfection’ would be no interventions and no interjections from the passengers in my car saying “what was that?!”

But my comment was not about perfection, rather it was about people who are complain about the lack of perfection In the face of significant improvements.
 
But I don't know why Chuck keeps calling the shared left turn lane a median. That lane is only for making left turns off the highway, and it's illegal to use it as a merging lane like he's expecting FSD to do. Yes, people use it for that purpose, but that doesn't mean it's a legal maneuver.

Here's an example where someone asked a police officer about the legality of using a shared left turn lane to merge into traffic: Answer Man: Can I drive in the center turn lane?
This is somewhat analogous to the rolling stop debate. For those who don’t recall, Tesla briefly allowed the car to perform a rolling stop in a restricted set of circumstances (good/clear sight lines, no other traffic or pedestrians.) This mimicked what most humans do but was technically illegal. It was quickly (and understandably, IMO) quashed by authorities.

The center turn lane is often used as an acceleration/merging lane by people when there is heavy traffic, even though it may be technically illegal in many states. It’s also another example of how tedious it will be to program FSD not only to drive safely but to follow 50 different sets of laws. ANd that’s just in the U.S.
 
To all the geniuses and lawyers and experts on believing what the media has to say; here is the official DOT document from FLORIDA explaining what MEDIANS are and what they do. In the image specifically it says TO PROTECT PEOPLE MAKING LEFT TURNS aka that means you're expected to stop there but hey, yall know better than the DOT right? If you're gonna Google random bs at least up your googling skills or you won't be able to face the tesla keyboard warriors.
I just spent 30+ minutes looking at individual State and Provincial motor vehicle policy online documents concerning TWTLs (not medians in general) and whether it's legal to enter a TWTL from a side street in order to merge. I found that it's legal in some jurisdictions and illegal in others. There's no pattern in either geography or underlying safety logic. What is illegal everywhere is prolonged travel in a TWTL.
 
A few entries from 10.13 were deleted / didn't make the cut for the already long 10.69 release notes:
  • Improved the “is parked” attribute for vehicles by 5% by adding 20% more examples to the training set.
  • Reduced false lane changes for cones or blockages by preferring gentle offsetting in-lane where appropriate.
  • Improved in-lane positioning on wide residential roads.
A couple made it unchanged:
  • Made speed profile more comfortable when creeping for visibility, to allow for smoother stops when protecting for potentially occluded objects.
  • Enabled creeping for visibility at any intersection where objects might cross ego’s path, regardless of presence of traffic controls.
A bunch had a lot of edits (and reordering within the list) and extra improvement for left turns and occupancy network:
  • Improvedstopping pose while yielding for crossing objects at “Chuck Cook style” 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 utilizingallowing optimizable initial jerk, to mimic the medianharsh pedal press by a human, when required to go in front of high speed objects. Also improved lateral profile approaching such safety regions. Improved decision makingregions to allow forunprotected left turns using better estimation of ego’spose that aligns well for exiting the region. Finally, improved interaction withother objects throughthat are entering or waiting inside the maneuver.median crossover region with better modeling of their future intent.
  • Improved lane positiongeometry error of crossing and mergingego-relevant lanes by 22% by adding long-range skip connections and a more powerful trunk to the network architecture. Improved lane position error by 5%34% and lane recallcrossing lanes by 12%21% with a full vector lanesVector Lanes neural network update. Information bottlenecks in the network architecture were abatedeliminated by increasing the size of the per-camera feature extractors, video modules,and internals of the GPTautoregressive decoder, and by adding a hard attention modules.mechanism which greatly improved the fine position of lanes.
  • Improvedpedestrian and bicyclist 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.
  • Improvedanimal detection recall of animals by 34%and decreased false positives by 8% by doubling the size of the auto-labeled training set.
  • Improveddetection recall of object detection, eliminating 26% of missing detections for far away crossing vehicles by4% by tuning the loss function used during training and improving label quality.
  • Upgraded Occupancy Network to use video instead of images from single time step. This temporal context allows theoccupancy network to detect dynamic objectsbe robust to temporary occlusions and enables prediction of occupancy flow. Also, improved performance by adding a video module, tuning the loss function,ground truth with semantics-driven outlier rejection, hard example mining, and adding 37k new clips toincreasing the training set.dataset size by 2.4x.
  • Reduced false slowdowns around crosswalks by better classificationnear crosswalks. This was done with improved understanding of pedestrianspedestrian and bicyclists as not intending to interact with ego.bicyclist intent based on their motion.
Most are new with 10.69:
  • 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 connectivites. This architecture achieves a 44% lower error rate on lane topology compared to the previous model, enabling smoother control before lanes and their connectivities becomes 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. 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 manevuers.
  • 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 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.
  • 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 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.
I'm really surprised that many of these features, such as latency, are just now being accounted for!
 
  • Like
Reactions: Bpseattle
You can't use it to pull into to stop and for a merging into traffic. It is for left turning traffic (on the road) only. Here is Georgia code.

40-6-126 - Central lane for turning​

Whenever a highway or roadway has a central lane in which traffic may enter from either direction for the purposes of making a left turn, no vehicle shall be driven into such central lane except for the purpose of making a left turn, and no vehicle shall enter into such central lane at a location which is more than 300 feet from the location where the vehicle will turn left across one or more lanes of oncoming traffic. No vehicle which has been driven into such a central lane shall be operated in such central lane for more than 300 feet.
Literally none of what you quoted makes using that lane while turning left onto that road illegal.

Whenever a highway or roadway has a central lane in which traffic may enter from either direction for the purposes of making a left turn,

This just defines the turning lane in question. The bolded part differentiates it from left turning ‘bays’ where only one direction uses them. In essence it says ‘can both directions on a road use this lane? Then it’s the one these laws apply to.’

no vehicle shall be driven into such central lane except for the purpose of making a left turn,

If the car isn’t making a left, it’s illegal. Turning into the road is making a left, so this clause does not make it illegal.

and no vehicle shall enter into such central lane at a location which is more than 300 feet from the location where the vehicle will turn left across one or more lanes of oncoming traffic.

Can’t enter the lane far away from the turn. For turning onto the road… is the car turning across oncoming traffic? If yes, it obviously isn’t more than 300ft. If no (T intersection), then the conditions for illegality are still not met. This clause doesn’t apply.


No vehicle which has been driven into such a central lane shall be operated in such central lane for more than 300 feet.

Can’t stay in the lane for a long time after turning into it. Obviously doesn’t apply to the turning into the road unless you stay there too long.



Now I’m not saying it’s legal in Georgia, but this quoted statute certainly doesn’t make it illegal!
 
Now I’m not saying it’s legal in Georgia, but this quoted statute certainly doesn’t make it illegal!

Georgia lawyers have interpreted that statute differently from you. If it intended to allow merging, it would say "all directions" not "either direction."


The law in Georgia says whenever there is a central lane which traffic may enter for purposes of making a left turn, no vehicle might be driven into that line, except for the purpose of making a left turn and no vehicle shall enter that lane more than 300 feet from where they plan to make their left turn. The biggest violation I see of this law is people who are wanting to get out into the street and use that central lane to pull into, and then merge into the lane that they’re ultimately trying to get into. Now the reason this is a safety problem is if a person is pulling onto the street, from say a driveway, and they pull into the central turn lane at the same time a person is rightfully trying to get into that center lane to make a left-hand turn, then they’re probably going to be a head-on collision. Which is the exact reason the legislature doesn’t permit you to use that lane for purposes of merging into the street.
 
Texas adds yet another wrinkle, allowing the shared lane to be used to pass slower traffic:

If a roadway is divided into three lanes and provides for two-way movement of traffic, an operator on the roadway may not drive in the center lane except:
(1) if passing another vehicle and the center lane is clear of traffic within a safe distance;
(2) in preparing to make a left turn; or
(3) where the center lane is designated by an official traffic-control device for movement in the direction in which the operator is moving.
 
But my comment was not about perfection, rather it was about people who are complain about the lack of perfection In the face of significant improvements

Yep, it has always been clear this would take years. We’re only about a year in! Got maybe as many as 3-4 years to go! (As I said in 2019, I’d be really surprised if we didn’t have something decent by 2029.)

Patience!
 
  • Like
Reactions: sleepydoc
Yep, it has always been clear this would take years. We’re only about a year in! Got maybe as many as 3-4 years to go! (As I said in 2019, I’d be really surprised if we didn’t have something decent by 2029.)

Patience!
Yes, this is a process of gradual improvement with plenty of things that still need worked. ULTs will need tweaking for some time as issues are discovered, for example.

I expect that, with 100,000 testers, finding issues isn't a problem! Having enough developer bandwidth to work on them is.

In any event, it should be very interesting to see how this unfolds.
 
  • Like
Reactions: Bpseattle
Georgia lawyers have interpreted that statute differently from you. If it intended to allow merging, it would say "all directions" not "either direction."

Like I said, legal or not, that statute has no language that makes it illegal.

Interestingly I found this page about Michigan, which uses the exact same statute that was quoted above.


And the ‘traffic expert’ Church has this to say about why it is indeed illegal:

“The tricky part of all of this is that this is not specified in the Michigan Vehicle Code but is found in the Michigan Manual on Uniform Traffic Control Devices (MMUTCD) and applied to the Michigan Vehicle Codethrough (this section to this section).”

The Michigan Vehicle Code does not say whether driving in the turn lane is allowed. But, Church said, the vehicle code does say the MMUTCD determines what signs, lane markings, etc. mean. And, under the MMUTCD, driving in a turn lane is not allowed.

I mean just look at that statute. There’s absolutely no language making it illegal. But it is common for traffic laws to be scattered and spread out through a variety of MV regulations.
 
Literally none of what you quoted makes using that lane while turning left onto that road illegal.
Yes it does. "... from either direction for the purposes of making a left turn..." is NOT from ANY direction. Also you can't left turn into the center turn lane and come to a stop because then you are making a RIGHT merge turn (You are NO longer making a left turn since that part is done) which is clearly NOT allowed. "....no vehicle shall be driven into such central lane except for the purpose of making a left turn..." Once in the lane you MUST make a left turn from the lane, so clearly making a right merge is illegal.
 
Last edited:
Looks like @Daniel in SD owes me a beer after he gets back online. Haven’t done exact scoring but that is two failures (intervention, and butt sticking out incorrectly, I have not watched the video so this is a minimum but probably just two) in less than 20 attempts, in light traffic.

I’m glad he made the bet apply to any unprotected lefts, otherwise I would be in potential trouble (who wins at 90% is unclear though). I thought that would hurt me but I got lucky.

Will be interesting to see how heavy traffic goes.

Overall it looks like a substantial improvement in these videos. Lots of intriguing incremental improvements.

Whole Mars video, 6:50
Steering wheel: Jerks back and forth in completely unnatural manner.
Whole Mars: That was good.

Lol.

Still looks like we’re gonna be mashing that accelerator all the way to the stop line, though hard to tell exactly how much of an issue there is in the videos.

Two interventions in Whole Mars video (signaling).
Whew! Needless to say I'm extremely upset. Not only does it appear that @AlanSubie4Life will win the bet but the value of his car has increased by $3k. My only hope is that the overfit works better in heavy traffic (perhaps the NN is not good at detecting single cars).
I don't see how anyone can see this as anything other than a massive failure. After years of development they can't even achieve a single 9 of performance on a maneuver that requires near perfection. I'd really like to know why real world testing was necessary at all, I don't see any reason that dangerous failure wouldn't have shown up in simulation.
 
No full video yet…

Looks like it is ready for pushing across LA traffic lanes. @Daniel in SD said this was hard!

I rate this turn a pass assuming it actually knew no one was coming (visualization said yes it did, but I would like to see 1000+ more tests with situations that are more heavily occluded and traffic IS there).

 
Last edited: