Jason71
Well-Known Member
I have never set myself as an "advanced" user. I test software as part of my job I don't feel the need to do it in my off hours. But despite there being an army of willing
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Have you lost faith in Tesla?
- Thread starter DeejUK
- Start date
No, people are not infuriated by improvements to their car, whether software or physical. They are infuriated when Tesla remove features, such as USS with no alternative in sight. They are infuriated when Tesla can’t solve simple thing like auto wipers. Tesla have managed to improve AHB, though it took them several years, but my wipers are still complete crap. Luckily for Tesla, though, there are plenty of apologists who don’t think this is unreasonable at all.I don’t.
but then I work in tech building software for companies who are agile, maybe I just get it and enjoy the changes because this is normal to me, where the same things infuriate others.
Not to provoke / poke the bear, but the wipers...
In day light, are they *really* any worse than a regular sensor? I 100% agree they are rubbish in the dark, but I've really not had much of an issue in the day / early evening, apart from the occasional (and it is fairly rare) mad wiping session when setting off.
In day light, are they *really* any worse than a regular sensor? I 100% agree they are rubbish in the dark, but I've really not had much of an issue in the day / early evening, apart from the occasional (and it is fairly rare) mad wiping session when setting off.
Jason71
Well-Known Member
For me they are OK in daylight in moderate to heavy rain but that is about it.
In drizzle / spray from wet roads I personally find they do not wipe often enough leaving me with what I consider to be an unacceptable levels of visibility.
And yes I do think this is worse than other brands I have used e.g. my wife's 2016 mini.
They were rubbish in the drizzle/spray on the M4 yesterday, it's kind of inexplicable as it's clear to me that the screen is totally obscured but they won't wipe.
For the most part my experience with auto wipers has been ok.
I have had instances where I've got into the car and because the sun has been low (I guess?) they have inexplicably just come on full tilt. Other times, as expressed above, it is not wiping fast enough to clear a moderate amount of drizzle. I feel like if I have to think "they shouldve swept the screen by now" then they've failed. At this point in time this sort of technology should (and does - on other cars) disappear into the background so you don't even know it's there, it just does its thing.
The technology - as used on other cars, e.g. rain sensors - is so refined at this point, both in a technological and cost sense, that I remain unconvinced that an alternative can do the job any better, or needs to. Let's not forget that there is not a zero amount of effort involved in either the programming or execution of code that tries to solve rain with AI, or whatever it's doing.
People might argue that not paying Bosch $5 or whatever per car is, over time, a good thing for Tesla, when you multiply it by the number of cars they sell. That might be the case, but it comes at our cost. I'd gladly pay $5 extra for auto wipers that used a traditional rain sensor.
I have had instances where I've got into the car and because the sun has been low (I guess?) they have inexplicably just come on full tilt. Other times, as expressed above, it is not wiping fast enough to clear a moderate amount of drizzle. I feel like if I have to think "they shouldve swept the screen by now" then they've failed. At this point in time this sort of technology should (and does - on other cars) disappear into the background so you don't even know it's there, it just does its thing.
The technology - as used on other cars, e.g. rain sensors - is so refined at this point, both in a technological and cost sense, that I remain unconvinced that an alternative can do the job any better, or needs to. Let's not forget that there is not a zero amount of effort involved in either the programming or execution of code that tries to solve rain with AI, or whatever it's doing.
People might argue that not paying Bosch $5 or whatever per car is, over time, a good thing for Tesla, when you multiply it by the number of cars they sell. That might be the case, but it comes at our cost. I'd gladly pay $5 extra for auto wipers that used a traditional rain sensor.
First car I had with auto wipers was a 1995 Peugeot 306. Its sensor was better than the current Tesla version. As has every subsequent car I have had with auto wipers, from Fords to BMWs to multiple French brands. It is clearly programmed for the benefit of the camera not for the driver.
It's not that at all. The rain sensor used on other cars isn't automatic. There's generally* a sensitivity setting on a stalk used to calibrate it by the driver. That might need tweaking up or down on different days, certainly in different seasons etc. To Tesla they are building a fully automated system so can't expect a human to tweak that setting. If you can solve the calibration then you don't really need the sensor, so no sensor.
They need to work more on the system that detects rain, as people observe there are clear scenarios where it's not good enough, clearly the model has gaps in it's training as if I can tell the screen needs wiping it should be able to pick up the same details.
You can have the opinion that it's an unsolvable problem with software, but would also mean therefore that driverless cars is unsolvable, so not in Tesla's perspective.
* there are some cars, typically lower end that don't have the calibration, read their owners forums and it's hit of miss whether they get lucky with the right settings.
Zilla91
Active Member
My biggest gripe with the wipers is that they go mad whenever the sun is out
Nubo
Test Mule
How adept the auto-wipers are is of secondary importance. The larger issue is Tesla's insistence of keeping the human driver at arms length from various essential manual controls that should be operable quickly by muscle memory without distracting them from the road ahead. Competent auto-wipers are a nice-to-have but in no way are they an excuse to strong-arm the driver away from manual control when needed (or simply wanted).
And with the announcement that stalks are being done any with altogether, it will make it even harder to over-ride the useless automated settings.
They can be ok at times, but they can be very inconsistent. I recall I used to adjust the sensitivity on previous cars which helped work around any such issues but Tesla don't allow that on auto.
What sort of functionality are you talking about? If you're talking about simple binary functions like 'if I press this button X should happen but it didn't' then yeah, I'd agree. Personally I haven't encountered any of those but then I have only had the car a month.
If you're talking about something like wipers not handling changing rain conditions correctly then I'd say you're being really optimistic about how easy testing is. Testing against complex scenarios is hard. Where wipers are concerned (and I realise you haven't said you're talking about wipers - just using it as an example) you've got different levels of rain intensity, different types of rain (heavy drops vs misty stuff) different light levels, different droplet impact patterns due to vehicle speed and wind, different ambient lighting conditions, different levels of dirt on windscreens, different coatings on windscreens (use of RainX, etc seems to correlate with an increased number of issues), different quality wiper blades, and that's just off the top of my head.
Now factor in that most of the above are not binary states and all of them can occur alone or in combination with any of the others AND the success criteria are quite subjective - the point where I want the wipers to go faster may not be the same as someone else, and testing is massive even for that one function.
Tesla does seem to recognise this - if you watch their FSD presentations half of that discusses how massive the testing task for FSD is, and how much R&D they've put in to finding new ways to test the system. Eg - gathering telemetry from real cars, using that to build 3D models of real world road systems in Unreal Engine, then fuzzing that environment to add in complex scenarios like kamikaze cyclists, exceptional weather and differing ambient foliage. It may not be perfect, but that leaves us in to the standard Tesla owner position of being forced to live with the ups and downs as they develop the platform.
MrBadger
Badger out
Do a search for 16A charging bug for starters… and map recalls… and…
MashinBenzin
Member
Wipers. It wasn’t raining today. I got in the car in daylight and pressed the wiper button when parked, to clear the windscreen of condensation. Wipers went mental, full speed for about ten wipes before I switched them off.
Then used autopilot on the motorway heading into the sun, had a bunch of dry wipes, most of them about 2-5 seconds apart over a distance of a mile or two.
Nothing extreme or unusual in the weather.
Then used autopilot on the motorway heading into the sun, had a bunch of dry wipes, most of them about 2-5 seconds apart over a distance of a mile or two.
Nothing extreme or unusual in the weather.
I think a lot of the complaints with Tesla wipers/headlights etc comes from two psychological factors:
1) They're trying to build a complex, sophisticated system that 'does the right thing'. Every time it doesn't 'do the right thing' (and that's very subjective, remember) users perceive that it has failed. Contrast that with old school technology. It had a much more predictable behaviour and users get used to working around its limitations. Eg - turning wipers on and off. My previous car had auto wipers and had you asked me I would have said they worked really well. But I also used to use the little switch for turning up/down the sensitivity all the time, and I used to manually turn them on and off in edge-case scenarios.
The radar cruise control was the same - it had a very narrowly defined function; 'stay x seconds behind the vehicle in front', and therefore the situations where it was going to fail were obvious - car indicating to pull in to an already narrow gap ahead = preemptive manual intervention to stop the car panicking when it realises a car has just appeared right in front of it. As a user you get used to these and work around them without really realising it. Tesla's Autopilot behaviour is much less stable because it's aiming much higher and as a user you're never entirely sure what it is 'thinking' at any given moment. You perceive that it behaves differently in the 'same' situation because you aren't entirely sure what the car's looking at.
2) Tesla's software changes over time. You never get chance to get firmly in a rut of knowing what the car can and can't do because that envelope is always evolving. We've become used to this with eg. smart phones, but this is not the common experience for car ownership. Incidentally, if you wind back the clock 10 years you would find ALL the same conversations going on in IT about software - I don't want a cloud/subscription service, I don't want automatic updates because I want to know that what works today will continue to work tomorrow. That isn't really a thing that gets talked about in the consumer space anymore because we've all just gotten used to both its upsides and downsides.
Obviously I realise there will immediately be a deluge of responses pointing out I am empirically wrong and rain/headlight sensors are flawless in all other vehicles and have been since Henry Ford's first model rolled off the production line. All I can say is this is not my experience.
1) They're trying to build a complex, sophisticated system that 'does the right thing'. Every time it doesn't 'do the right thing' (and that's very subjective, remember) users perceive that it has failed. Contrast that with old school technology. It had a much more predictable behaviour and users get used to working around its limitations. Eg - turning wipers on and off. My previous car had auto wipers and had you asked me I would have said they worked really well. But I also used to use the little switch for turning up/down the sensitivity all the time, and I used to manually turn them on and off in edge-case scenarios.
The radar cruise control was the same - it had a very narrowly defined function; 'stay x seconds behind the vehicle in front', and therefore the situations where it was going to fail were obvious - car indicating to pull in to an already narrow gap ahead = preemptive manual intervention to stop the car panicking when it realises a car has just appeared right in front of it. As a user you get used to these and work around them without really realising it. Tesla's Autopilot behaviour is much less stable because it's aiming much higher and as a user you're never entirely sure what it is 'thinking' at any given moment. You perceive that it behaves differently in the 'same' situation because you aren't entirely sure what the car's looking at.
2) Tesla's software changes over time. You never get chance to get firmly in a rut of knowing what the car can and can't do because that envelope is always evolving. We've become used to this with eg. smart phones, but this is not the common experience for car ownership. Incidentally, if you wind back the clock 10 years you would find ALL the same conversations going on in IT about software - I don't want a cloud/subscription service, I don't want automatic updates because I want to know that what works today will continue to work tomorrow. That isn't really a thing that gets talked about in the consumer space anymore because we've all just gotten used to both its upsides and downsides.
Obviously I realise there will immediately be a deluge of responses pointing out I am empirically wrong and rain/headlight sensors are flawless in all other vehicles and have been since Henry Ford's first model rolled off the production line. All I can say is this is not my experience.
I also find the weirdest wiper behaviour is when there is no/very little rain rather than when it's actually raining. At a guess I'd say the windscreen in front of the sensor gets smeared in some way and the car's trying to wipe it off believing it to be rain.
It would be nice if that was fixed. Normally I find that using the washer to clean the screen sorts it though.
MashinBenzin
Member
What you’ve described seems like a significant waste of time and money when a perfectly good sensor exists.
Unless… hypothetically a sensor costs $2 per car and adds $2 cost in build. $4m per million cars. If I put minimal dev and test resources into what you describe, two FTS engineers, two FTE testers, $600k per annum. Starts to look a compelling argument to tolerate the lower quality wiping experience. (Noting that these FTEs are just a small part of the larger FSD team)
