-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Wiki FSD Beta 10.4
- Thread starter MilesAR
- Start date
alstoralset
Man from another place
I've been wondering the same thing. With "traditional" product development, typically you iterate on releasing to an audience just large enough to uncover bugs, then fix those and and expand the audience as appropriate.
But, NN training is (at least to me) a dark art, so who knows how much data for the same issues may be required. Or, maybe it's a marketing or FSD revenue realization thing.
EVNow
Well-Known Member
I don't think FSD revenue realization is all that important anymore. Tesla demand is so high, even with price increases (some 10 times this year !) - they don't need any revenue padding to hit their earnings goals. They wouldn't care about "marketing" for the same reason ...
I think it is more about
- Data for NN
- Some competetive pressure from Waymo, MobilEye etc
- Elon wants to move ahead, its something he gets criticized the most, so its a ego thing for him now to prove his haters wrong
The weather this weekend was amazing in the Bay Area and many cyclists were out.
I was very, very happy to see this--10.4 edges over and routes around nearby cyclists instead of just braking and creeping past.
It also does this with trucks.
I was very, very happy to see this--10.4 edges over and routes around nearby cyclists instead of just braking and creeping past.
It also does this with trucks.
It's been my experience in the last 3 1/2 years that whenever AP capabilities are expanded it gets nervous and jerky for a few months until it smooths out. I've seen half a dozen of these cycles and 10.4 looks to be an extreme case. This is all digital software but the rules of servo mechanisms still apply. It seems the gain/bandwidth are too high. These are all adjustable but one wonders if they're close to a memory or processing limit. Don't let the jerking continue as it can't be good for the steering drive belt.
Nah. How much stress and wear is that belt really taking? It's also encased and isolated from the environment.
EVNow
Well-Known Member
The weather this weekend was amazing in the Bay Area and many cyclists were out.
I was very, very happy to see this--10.4 edges over and routes around nearby cyclists instead of just braking and creeping past.
View attachment 733261
It also does this with trucks.
I had an interesting edge case. We have road sweeper trucks here and FSD was trying to go around the slow moving sweeper. I disengaged because I wasn't sure whether the car would go over the brushes on the road
I had an interesting edge case. We have road sweeper trucks here and FSD was trying to go around the slow moving sweeper. I disengaged because I wasn't sure whether the car would go over the brushes on the road
View attachment 733285
When I tried taking this type of picture my Y started giving me the "emergency takeover" warning and then waved a stick at me when I parked indicating it would remove me from the Beta if I don't pay better attention. It left me a little shocked but mostly impresssed.The weather this weekend was amazing in the Bay Area and many cyclists were out.
I was very, very happy to see this--10.4 edges over and routes around nearby cyclists instead of just braking and creeping past.
View attachment 733261
It also does this with trucks.
EVNow
Well-Known Member
I take videos of all me fsd drives using a GoPro mounted to the roof of the car. I've never, ever tried to take a photo while driving (FSD or otherwise) in 2+ decades.
I don't want to get into an accident and spend months in a hospital or worse bring that misfortune on someone else because I wanted to take a photo.
It's the rapid reversals that are hardly ever encountered in manual driving. It is a flexible toothed belt sort of like a miniature timing belt. Timing belts last a long time but they don't have these rapid reversals.
AlanSubie4Life
Efficiency Obsessed Member
You are setting yourself up for disappointment. Think a timeframe of 12-24 months for improvements to “much better” here. And that is a long road with many perils (see below!) along the way. 10.5 will be largely indistinguishable from 10.4 - there may be new minor features added, but expect the same basic experience for quite a while now.
Maybe in 11 we’ll see some marked changes in behavior (and maybe single stack - which may not be an improvement) but I still expect it will be a pretty rough ride.
I generally have the idea that in a year or two we may have something that provides overall a fairly pleasant drive and behaves naturally. Not sure if that is too optimistic - it’s really hard to predict. We are at 3+ years with NOA and we’re not at that point yet, so this timeline may well be optimistic.
It’s fine - there’s no rush! Just have to try to make sure no one totals my car over the next few years. Everyone out there is trying as far as I can tell.
For this accident, seems like a lot of assumptions to make. While it’s possible for EPAS to outmuscle the driver, apparently, I have never experienced that even with emergency road departure assist (which I have seen on a few occasions). So I don’t really see anything wrong with their current level of lane departure correction (and does it even apply during a turn?).
It’s really hard to say what happened in this case. My guess would be that FSD jerked the wheel suddenly during a turn into the adjacent turn lane, putting the car on a collision trajectory, which is 100% normal and fully expected. Then the driver disengaged and did not recover it properly and completed the collision. But there’s really no way to know without video or data from the vehicle. I would be shocked if it happened as described, in any case. And all of the above is pure speculation.
Last edited:
TravelFree
Active Member
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.
AlanSubie4Life
Efficiency Obsessed Member
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.
TravelFree
Active Member
I'm really happy with FSD NOA on the highway. Last summer I drove 370 miles on NOA on I-95 and never had one intervention. Mostly stayed the speed limit to avoid the many radar traps but it was so pleasant just keeping a little pressure on the wheel. I don't ever recall driving almost 6 hours non stop and hardly feeling it. When I got out to Supercharge, my bladder was telling me forget charging run in to the mens room.
Highway AP is so simple compared to this city and even rural driving. We'll get there but not on Elon's optimistic timeline.
TravelFree
Active Member
My few data points showed it to be very accurate at the traffic light but the reaction of that slamming on the brakes when the car and truck lights came on indicates it's the interpretation and on what to do that needs work. But that's what we're here for. To let them know where to focus.
AlanSubie4Life
Efficiency Obsessed Member
Quiet interstate freeway driving is a best-case scenario for NOA and I agree it works in that scenario with very few issues. I’ve done nearly 1100 miles in a day and NOA definitely helped. I’m referring to more complicated scenarios on interstate freeways where there is significant other traffic and high-speed interchanges - that is where there’s still a lot of work needed to provide a normal driving experience.I'm really happy with FSD NOA on the highway. Last summer I drove 370 miles on NOA on I-95 and never had one intervention. Mostly stayed the speed limit to avoid the many radar traps but it was so pleasant just keeping a little pressure on the wheel. I don't ever recall driving almost 6 hours non stop and hardly feeling it. When I got out to Supercharge, my bladder was telling me forget charging run in to the mens room.
EVNow
Well-Known Member
I've not found a significant difference between pure TACC behavior pre-FSD and FSD Beta (i.e. radar + vision and vision only). So, my guess is VO speed detection is fairly good.
AlanSubie4Life
Efficiency Obsessed Member
Just when it is safe to do so (or preferably have a passenger assist you or use a GoPro) pay attention to how often the brake lights are detected, when there are no brake lights. It happens a lot! It does depend on lighting conditions, etc., so I think there are scenarios where it works pretty well. But there are also scenarios where it does not work at all. It’ll definitely show brake lights in these low specificity scenarios.
Sure but the question is how can you know that your car did not detect a slowdown in the vehicle ahead due to brake application. You’d have to have a test of a vehicle ahead applying brake lights without slowing down (could perhaps do this at dusk by having the lead car turn on and off headlights which would potentially emulate brake light application - or rewire the brake lights on a test vehicle temporarily).
Last edited:
Knightshade
Well-Known Member
Well, the lower speed limit is noticeable... and the longer min follow distance even moreso.
From the data green has shown comparing the VO vs radar values VO is pretty close, but not quite at parity (which explains why it's still more restricted I guess)
EVNow
Well-Known Member
My observation is mostly on city streets, so yes, lower speeds.
Though, as time goes by with FSD - Radar will start becoming a dim memory ...
Similar threads
