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

Wiki FSD Beta 10.4

This site may earn commission on affiliate links.
FSD 10.4

holding till next version. Was on single lane with double yellow line divider. The car tried to cross the lines into oncoming traffic....could have been bad. also getting disengagements when making right turns.

plus, I just dont use fsd if I see cars close behind me in case of phantom braking...I have to constantly check the rear view mirror. Just too much going on with this version but 10.3.1 was better.

waiting for 10.5 to restart testing....which is what this is...testing.
 
I'll try to get used to this. The throttle doesn't disable FSD though; I think that was my instinctive hesitation, that I'd rather disable FSD than fight with it. But I'll try the throttle the next time it happens
I have been using AP for at least three years. I have always kept my foot near the gas pedal, incase of the sudden braking. As soon as it starts I hit the gas. the amount of times it happened decrease. Now with Beta it is doing it a lot. I feel like AP has trained us how to react to beta. when we are charging red lights I'll hit the stalk down to minimize the excessive braking.

The expanded screen shows all the intersections and or large driveways, I notice the braking when it gets close. It makes me wonder if I'm the first one driving beta through that area and I swear the system is hunting for stop signs. After I have traveled that road two or three times, I notice less braking.

On the flip side, when at 4 way stops, the car is noticing a parked car crowding the intersection. My car will not move until I hit the gas. when the car balks at crossings I also hit the gas to encourage it through.
 
The expanded screen shows all the intersections and or large driveways, I notice the braking when it gets close. It makes me wonder if I'm the first one driving beta through that area and I swear the system is hunting for stop signs. After I have traveled that road two or three times, I notice less braking.
This just may be a coincidence because you have too few data points to determine a trend. If the FSD computer is truly learning from your repeat locations, then once it understands your correction within the same software version then it should never suffer a flaw at that location again. If you do 10 trials and the failed ones show up at random, then you have just an uncertainty in the FSD decision. If it fails every time then that is an instruction set in the build.

I wanted to test the FSD-beta to recognize speed bumps. So, in the first beta test versions I did it failed to see them and slow down after several dozen trials. Conclusion- there was nothing in the FSD beta version to instruct to recognize speed bumps. Later, in version 10.4 I ran my test drive with 20 speed bumps of all shapes and markings. Surprise! v10.4 saw every one with no failures. I repeated the test route the next two days and speed bumps were properly executed with a slowing to 10-19mph from 30-35mph roads. Conclusion: Tesla wrote it into the latest version. FSD-beta wasn't using some human level intelligence to figure out speed bumps on it's own. The Tesla Team added it to the FSD computer software.

Do more than 2-3 trials to see if you have a real conclusion.

In the case of random ( or phantom) braking, I believe but can't verify that phantom braking is caused by a large variety of triggers. Hopefully the guys at Tesla can see some trends in the data that is being reported because they have many more trials than I could possibly do so they can find out what causes it and fix it.
 
Last edited:
I'll try to get used to this. The throttle doesn't disable FSD though; I think that was my instinctive hesitation, that I'd rather disable FSD than fight with it. But I'll try the throttle the next time it happens.

I have been using AP for at least three years. I have always kept my foot near the gas pedal, incase of the sudden braking. As soon as it starts I hit the gas. the amount of times it happened decrease. Now with Beta it is doing it a lot. I feel like AP has trained us how to react to beta. when we are charging red lights I'll hit the stalk down to minimize the excessive braking.

I've been trained by the vehicle not to keep my foot close to the brake when monitoring AP but instead close to the throttle/accelerator. Disengaging would be smoother as repeated brake-then-acceleration "waves" are not appreciated by vehicle occupants.
 
This just may be a coincidence because you have too few data points to determine a trend. If the FSD computer is truly learning from your repeat locations, then once it understands your correction within the same software version then it should never suffer a flaw at that location again. If you do 10 trials and the failed ones show up at random, then you have just an uncertainty in the FSD decision. If it fails every time then that is an instruction set in the build.

I wanted to test the FSD-beta to recognize speed bumps. So, in the first beta test versions I did it failed to see them and slow down after several dozen trials. Conclusion- there was nothing in the FSD beta version to instruct to recognize speed bumps. Later, in version 10.4 I ran my test drive with 20 speed bumps of all shapes and markings. Surprise! v10.4 saw every one with no failures. I repeated the test route the next two days and speed bumps were properly executed with a slowing to 10-19mph from 30-35mph roads. Conclusion: Tesla wrote it into the latest version.

Do more than 2-3 trials to see if you have a real conclusion.

In the case of random ( or phantom) braking, I believe but can't verify that phantom braking is caused by a large variety of triggers. Hopefully the guys at Tesla can see some trends in the data that is being reported because they have many more trials than I could possibly do so they can find out what causes it and fix it.
It's crazy the differing behavior of speed bumps. Out of probably 60-70 speed bumps passes in my neighborhood (I go over many each time in and out) on 10.4, FSD has slowed for exactly 1 speed bump. They are all rendered; just no slowing.
 
  • Like
Reactions: EVNow and Itsuo-DC
hmmmm ...... I have the model x and use the smaller visualizations on the front screen,...is it possible the enhanced visualization view makes the fsd drive better in terms of smoother, less disengagements etc?
My guess is it is a graphics display part of the software and not how the vehicle control is limited. FYI- I also use the center visualization too. It also keeps the posted speed and vehicle speeds back at the top. I like having the wh/hr meter up and being an efficiency freak, I use it constantly. I also use the expanded map display too. If Tesla didn't give us that limited 1/3 display I probably would not have joined the beta program.
 
It's crazy the differing behavior of speed bumps. Out of probably 60-70 speed bumps passes in my neighborhood (I go over many each time in and out) on 10.4, FSD has slowed for exactly 1 speed bump. They are all rendered; just no slowing.
Interesting! I don't have FSD beta on my M Y but that may be a difference. Also how do you have your speed settings. Mine are + - 0 fixed. Can't think of any other differences.
 
Interesting! I don't have FSD beta on my M Y but that may be a difference. Also how do you have your speed settings. Mine are + - 0 fixed. Can't think of any other differences.
I suspect it's a mapping issue. I checked OSM and none of the speed bumps were mapped. I have since mapped them but it will be some time before Tesla integrates them into a map update. That's the only difference I can think of.
 
I suspect it's a mapping issue. I checked OSM and none of the speed bumps were mapped. I have since mapped them but it will be some time before Tesla integrates them into a map update. That's the only difference I can think of.
Maybe. Off the top of my head look on your map to see if Dean Road in Jacksonville, FL has speed bumps mapped. I've never used or heard of "OSM" Dean Road is just one road on my test route.

Dean Road.png
 
10.3.1 slowed for speed bumps. 10.4 might have included speed bump visualization to the UI.

This just may be a coincidence because you have too few data points to determine a trend. If the FSD computer is truly learning from your repeat locations, then once it understands your correction within the same software version then it should never suffer a flaw at that location again. If you do 10 trials and the failed ones show up at random, then you have just an uncertainty in the FSD decision. If it fails every time then that is an instruction set in the build.

I wanted to test the FSD-beta to recognize speed bumps. So, in the first beta test versions I did it failed to see them and slow down after several dozen trials. Conclusion- there was nothing in the FSD beta version to instruct to recognize speed bumps. Later, in version 10.4 I ran my test drive with 20 speed bumps of all shapes and markings. Surprise! v10.4 saw every one with no failures. I repeated the test route the next two days and speed bumps were properly executed with a slowing to 10-19mph from 30-35mph roads. Conclusion: Tesla wrote it into the latest version. FSD-beta wasn't using some human level intelligence to figure out speed bumps on it's own. The Tesla Team added it to the FSD computer software.

Do more than 2-3 trials to see if you have a real conclusion.

In the case of random ( or phantom) braking, I believe but can't verify that phantom braking is caused by a large variety of triggers. Hopefully the guys at Tesla can see some trends in the data that is being reported because they have many more trials than I could possibly do so they can find out what causes it and fix it.
 
I wasn't referring to a slowdown ahead, rather a brake light came on and the car or truck was not even in my lane. That's when I observed my car braking for no reason, ie. nothing ahead of ME.


Sure, but maybe the car was braking because the car ahead not in your lane was slowing down. Potentially nothing to do with brake lights. Hard to know without testing it without any slowing.

That being said, you’ll notice all cars are hitting the brakes all the time in the visualization, so I doubt they are using this. Or maybe they require a certain confidence level to use it, which is higher than the confidence level used in the visualizations. There is really no way to know without carefully testing it.
 
Had to go to the Dentist this morning and I did look for the brake lights action. At traffic red lights 4 lanes with two left turn and one right turn lane showed all cars and trucks with lit up brake lights. I saw the car in front of me in the left turn lane inch forward and when his lights went out I began to creep too as he drifted forward to tighten his gap. Then the center lane got the green and began to move and when their tail lights went out I didn't move. Seems to work as it should at the traffic light.

Traveling down the road 3 lanes with me in the center doing the 45mph speed limit a car in the right lane just ahead slowed down with brake lights. FSD-beta slammed on the brakes when that happened for no reason as the car to the right was not in my lane or trying to get in my lane. Later I saw a 6 wheel truck speeding faster in the left lane and then he spotted a police car way up ahead. he hit his brakes and again FSD beta slammed on the brakes for no reason. Good there were no cars tail gating me.

Still it is pretty cool seeing the other cars brake lights across all 4 lanes at a traffic red light.

On this trip FSD-beta attempted to run a red light and I had to hit the brakes. The light was yellow for about 10 seconds away and if the Yellow would have been on for 15 seconds it wouldn't have been a problem but the light turned red when I was about 50 ft away. My hard braking had me stopped in good time.

Finally, my NOA had me making a right turn at the next light with a right turn lane and the Nav announcement came on but FSD-beta ignored the opportunity to get in the right turn lane until traffic was too late. Missed the turn and had to go to the next block to get back on my route.

I’ll have to look at this too. I’d be a little surprised if they were actually using this since the specificity of brake light detection is pretty low (false positive rate seems to be near 100% over a large enough interval of time). However, can’t argue with the results - they’re about what you would expect if they were using it.

It just seems very difficult to distinguish between the car detecting & responding to a speed change in the lead vehicle (which it must do quite accurately even with no brake lights showing), and it detecting and responding to brake lights.

OK, now that you guys are primed with scenario 1 (seeing if your car immediately slows down when it "sees" brake lights on the preceding car), let's move on to scenario 2. :)

Quite a few guys have reported that their cars will inappropriately attempt to go around traffic in front of them that is stopped at a traffic light.

So I've been specifically watching for this behavior quite a bit, and have another observation for you guys to try out if you're up for it.

Normally, when you're stopped behind a vehicle stopped at a traffic light, the preceding car will have its brake lights on. In my experience so far, as long as that condition is met (preceding car showing brake lights) my car won't attempt to go around them. At least, most of the time it won't.

But if the preceding car's brake lights go out, within just a few seconds, my car will reclassify them as a "Parked Vehicle" and attempt to go around them.

Unfortunately, this isn't an always repeatable thing... sometimes my car will attempt to go around cars that do have brake lights on, but it is quite repeatable in that 90% of the time my car will attempt to go around them shortly after their brake lights go out (assuming that they don't then start moving forward).

I did have quite a few different things saved with TeslaCam, but unfortunately that USB drive gave up the ghost on me, so I'm back to square one. Although my new iPhone mount just arrived, so hopefully I'll be able to record a few of these scenarios for you guys to look at.

It would be so much easier to say, "hey, let this ten second clip explain the scenario" than expect you guys to read paragraphs of highly entertaining "so I'm stopped at a traffic light, three lanes one direction, a turning lane, a VRU to the right...." *snore*.
 
I suspect some kind of "this is a parked vehicle" bit associated with that vehicle object in FSD memory got stuck on.
Yeah, and not only does it get stuck on, but it gets flipped to on inappropriately.

They really need to work on their vehicle classification code as it applies to labeling a vehicle as "active traffic" (or whatever the Tesla label is) or "parked vehicle."
 
I noticed something in the morning drive I had not noticed before. The car drives too fast until close to the turning and brakes hard before turning. All this in "Chill" mode. I tried looking at my old videos - but those traffic light turnings are always busy and have some vehicles in front. So, probably for the first time the was clear and FSD sped up - I thought it would miss the turn, but slowed down from 35 to 20 quickly to make the turn.

ps : Another interesting thing today - when stopped behind a row of cars (behind a not-visible school bus) - the car didn't attempt to go around by going over the yellow center line.
 
Last edited:
It would be so much easier to say, "hey, let this ten second clip explain the scenario" than expect you guys to read paragraphs of highly entertaining "so I'm stopped at a traffic light, three lanes one direction, a turning lane, a VRU to the right...." *snore*.
I hear ya! Plus I am equipped to record with 3 GoPros. One on my drivers display, one on the main map screen and one in the windshield to see forward. I have the equipment and a YT channel but that is reserved for my travel videos and I don't want to clutter it with Tesla testing. Besides editing is time consuming and I don't collect a pay check from YT which gives me more freedom from copyright infringements. There are plenty of younger guys doing excellent video on YT and more published every week. They need the YT money. :)

The only reason I do record now is because it helps me do a quick look at a situation later since I don't have a log keeper riding along. I use the cameras for other studies I have been doing like my project to plot V3 Supercharging ramp curves for different SOC's.
 
  • Funny
  • Like
Reactions: EVNow and Phlier
Plus I am equipped to record with 3 GoPros.
I'm very jealous! My Chief Financial Acquisitions Officer refused my requisition request for just a single GoPro. I just got a mount that will hopefully allow me to suspend my iPhone from the 3's roof glass. Going to attempt to get a bit of video today.

my project to plot V3 Supercharging ramp curves for different SOC's.
Now that is a thread I look forward to reading!
 
I've seen that on YT videos but in some 50-75 test red lights I haven't had that happen to me.
It seems like it is far more likely to happen in scenarios where the camera can't see the traffic light well enough to determine its current state, but there are other situations that seem to trigger it, too.

Definitely something to record if you get a chance; we need to supply Tesla with as many videos of this as possible.

There are a couple of traffic lights close to me where this happens quite often. I'm going to attempt to get a bit of video of it today.