Why does the UI miss so many oncoming vehicles?

[MikeyC]

OK, I've been scratching my head for a year now and kept my mouth shut, but I just have to ask...

When driving down a two lane road (one lane in each direction), I notice that the UI will detect several vehicles in the opposing lane passing by and then it'll miss the next two or three, then see one more, then miss a couple, and so on. This is when there is nothing but a double yellow between us and the opposing traffic is the same (small) distance from me.

Why is the UI so inconsistent: why does it miss so many vehicles and it seems to only detect "random" cars going by? The detection rate is probably less than 50%. Why would it detect some and not others? BTW, it has nothing to do with size or gap between the vehicles. It'll detect a Prius for example and then a full 10 seconds later, it'll miss a box truck for example.

Thanks,
Mike

 

[sightman]

Ive always wanted to ask the same but never got to it...

I'm also curious if the UI is missing it, or the autopilot computer is missing it. The former is a non issue but the latter can have consequences.

 

[MikeyC]

Ive always wanted to ask the same but never got to it...

I'm also curious if the UI is missing it, or the autopilot computer is missing it. The former is a non issue but the latter can have consequences.

Exactly my thought. I never depend on autopilot and I'm always in control but optimally I'd like AP to be able to move to the right a bit if it sees an oncoming vehicle start to come over the line toward my lane. That happened only once to me: truck started to veer across the double yellow. I saw it (I was in AP) but 100 feet or so before I passed by, AP hadn't taken any action so I just turned the steering wheel to move myself further to the right manually. The truck was only maybe 6-8 inches over the line but it looked like he was still veering further into my lane so if I hadn't intervened, I doubt we would have collided but still: I wanted the extra room. I didn't notice whether or not the truck appeared on the UI as my focus was on making sure there was no collision.

Mike

 

[Knightshade]

Autopilot, the public version, is not intended for use on 2-way, 2-lane roads like that- as the manual points out. So if it's doing anything "wrong" it's because it's not meant for that situation in the first place.


FSDBeta, currently being tested by roughly 60k owners, is meant for that- and even does that MikeyC is asking about regarding moving over to the right for big coming trucks.

 

[EVRider-FL]

Schedule a service visit to determine if there’s a problem with your cameras.

 

[M3DS]

FSD Beta sees and shows everything, but I think you may only be seeing cars in adjacent lanes with oncoming traffic that it is worried about? generally when you are lane keeping it is only concerned with your lane and immediate danger IMHO.

 

[Theone8181]

I've noticed this in mine, here in the UK generally it's only motorways that have a central reservation. The vast majority of our roads don't. I wonder if it's only picking up cars/trucks it thinks are too far over?

 

[pnwadventures]

Off topic and contentious comment, but: I wish that entire visualizer area would be replaced with status information and vehicle controls. I don’t see the point of devoting that much control surface real estate to a video game that only passengers should be looking at, especially when not even using AP/FSD. I’d rather have one-touch fog light & wiper controls, and at-a-glance trip computer info over there.

 

[medphys3]

Autopilot, the public version, is not intended for use on 2-way, 2-lane roads like that- as the manual points out. So if it's doing anything "wrong" it's because it's not meant for that situation in the first place.


FSDBeta, currently being tested by roughly 60k owners, is meant for that- and even does that MikeyC is asking about regarding moving over to the right for big coming trucks.

This.

“Warning

Autosteer is intended for use on controlled-access highways with a fully attentive driver.”

I look at it this way- it’s not really designed to keep you away from oncoming traffic. It’s designed to keep you in your lane and to avoid traffic running with you (much less speed difference as opposed to oncoming traffic).

 

[tm1v2]

Off topic and contentious comment, but: I wish that entire visualizer area would be replaced with status information and vehicle controls. I don’t see the point of devoting that much control surface real estate to a video game that only passengers should be looking at, especially when not even using AP/FSD. I’d rather have one-touch fog light & wiper controls, and at-a-glance trip computer info over there.

I think almost every Model 3 and Y driver agrees with this. The huge, driver-side, always-showing AP/FSD visualization is a terrible waste of space on the single central display. It's by far the most annoying aspect of the car in daily driving.

I can only guess that Tesla views it as an advertisement for selling AP/EAP/FSD upgrades.

 

[Shelburne]

I have had the same problem with my M3 since the day I bought it in September of 2020. Not long afterwards I asked about this issue and was told that the software in the M3 is intentionally designed to pay more attention to the cars coming up behind you, or in adjacent lanes. The rational, I was told, is that the human driving the car can easily see what's in the oncoming lane or in the lanes ahead of you, so they invest more processing power in detecting cars alongside or behind you. But, frankly, I'm not sure I buy that explanation.

 

[GreenHokie]

OK. Here is correct answer to the OP’s question.

Our vehicles see and recognize pretty much all incoming vehicles/objects, but simply chooses not to display all the information/detail on the screen. Before I was given access to FSD Beta, I observed the exact same thing as the OP (the display screen was not showing every vehicle as they passed by). Since getting the FSD Beta download, my vehicle now renders/shows pretty much every vehicle/object/pedestrian that I see. It’s surprisingly accurate and a joy to see that much detail rendered/displayed in real time. Nothing has changed on my end from a hardware standpoint. The only difference from before is I now have FSD Beta.
Join the FSD Beta club and then you’ll understand what I’m saying…

 

[smatthew]

OK. Here is correct answer to the OP’s question.

Our vehicles see and recognize pretty much all incoming vehicles/objects, but simply chooses not to display all the information/detail on the screen. Before I was given access to FSD Beta, I observed the exact same thing as the OP (the display screen was not showing every vehicle as they passed by). Since getting the FSD Beta download, my vehicle now renders/shows pretty much every vehicle/object/pedestrian that I see. It’s surprisingly accurate and a joy to see that much detail rendered/displayed in real time. Nothing has changed on my end from a hardware standpoint. The only difference from before is I now have FSD Beta.
Join the FSD Beta club and then you’ll understand what I’m saying…

Plus if you look at videos where they overlay what the computer sees... the car sees absolutely everything. It just doesn't show it all to you.

 

[FLMan]

I think almost every Model 3 and Y driver agrees with this. The huge, driver-side, always-showing AP/FSD visualization is a terrible waste of space on the single central display. It's by far the most annoying aspect of the car in daily driving.

Not me. 99% of my driving is on the highway and the "wasted space" you refer to is my blind spot monitor.

 

[medphys3]

Not me. 99% of my driving is on the highway and the "wasted space" you refer to is my blind spot monitor.

Same. I use that screen as much as I use the mirrors. In fact, I’d ditch the stupid map that’s constantly displayed. That to me is a waste of space. At least make it user configurable.

 

[tm1v2]

I stand corrected!

Maybe we all agree on making more of the layout user configurable though! Like in the old Model S UI with split top & bottom sections...

 

[3sr+buyer]

Maybe we all agree on making more of the layout user configurable though!

Yes, it would be nice to be able to put any of several things in various parts of the display.

It would also be nice to have more options of things that can be put on the bottom bar. V11 added that user configurable bottom bar, but some useful things are not offered as options to put there. For example headlight control, wiper control, heated seat, HVAC fan speed would be nice to be able to have quickly usable controls there.

 

[OxBrew]

What are we allowed (like, not prevented from...) using AP on city streets or 2 lane roads, if we're not supposed to?

I understand the warnings, and only sometimes try it out as a curiosity in those places, if no one's around and I'm ready to pay extra attention to monitoring it. But if it's not really "released" for that, then prevent me from using it, right? This baffles me.

 

[FLMan]

Maybe we all agree on making more of the layout user configurable though!

^^ Yes.

I can't understand how a company can create such an amazing car with all of the weird and unique engineering concepts it has along with the stylish looks and then completely punt on the software stack.

We had customizable (themed) UI's in the 1990's so it's a little baffling that they can't figure out how to let you theme the entire interface the way you like it and then draw inside the borders the user has defined for it. I'm not even asking to theme the icons, just let me place the icon controls for what I want where I want. A true object-oriented design would bring the functionality along with the widget no matter where you put it.

It also shouldn't take 6 months to a year to combine software stacks unless it's spaghetti code. It's likely too late since the base was probably not built with the future in mind - so it's a complete re-write and re-test at this point.

I feel like they are ahead of the competition at this point in time, but they're really only a good hire by the competition ahead. Whoever builds a clean, updateable control and UI stack will win out in the end. Farming it out to Apple or Google is not the answer, that should always be an add-on (i.e. Car Play) and it SHOULD be an add-on. Who cares if people use it or not, I'd use that instead of Steam / Netflix / etc. GM, Ford, VW, etc. all know by now that the OS is what will win the EV game in the end and I'm sure they have some Tesla's and know they are not out of it.

Since features take so long to arrive with Tesla and are only partially functional (things like auto-high beam and auto-wipers come to mind) it hints that there are some large chinks in the armor. It really shouldn't take more than a Sprint (4-6 weeks) for simple features like these to be delivered, fully tested and fully functional. We're all living in beta-land anyway, so I'd give them 2 Sprints to get it right. But months / years? That's way too long.

 

[smatthew]

^^ Yes.

I can't understand how a company can create such an amazing car with all of the weird and unique engineering concepts it has along with the stylish looks and then completely punt on the software stack.

We had customizable (themed) UI's in the 1990's so it's a little baffling that they can't figure out how to let you theme the entire interface the way you like it and then draw inside the borders the user has defined for it. I'm not even asking to theme the icons, just let me place the icon controls for what I want where I want. A true object-oriented design would bring the functionality along with the widget no matter where you put it.

It also shouldn't take 6 months to a year to combine software stacks unless it's spaghetti code. It's likely too late since the base was probably not built with the future in mind - so it's a complete re-write and re-test at this point.

I feel like they are ahead of the competition at this point in time, but they're really only a good hire by the competition ahead. Whoever builds a clean, updateable control and UI stack will win out in the end. Farming it out to Apple or Google is not the answer, that should always be an add-on (i.e. Car Play) and it SHOULD be an add-on. Who cares if people use it or not, I'd use that instead of Steam / Netflix / etc. GM, Ford, VW, etc. all know by now that the OS is what will win the EV game in the end and I'm sure they have some Tesla's and know they are not out of it.

Since features take so long to arrive with Tesla and are only partially functional (things like auto-high beam and auto-wipers come to mind) it hints that there are some large chinks in the armor. It really shouldn't take more than a Sprint (4-6 weeks) for simple features like these to be delivered, fully tested and fully functional. We're all living in beta-land anyway, so I'd give them 2 Sprints to get it right. But months / years? That's way too long.

You think that whoever builds the prettiest UI will win? You could put lipstick on Supercruise, but it will still be a pig.