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

2017.50.3.f3425a1 is out!

This site may earn commission on affiliate links.
Another Surprise!
Smooth Stops too!!!
This just keeps getting better!
Very impressed with Tesla Autopilot AP2 2017.50.3 f3425a1 ability to stop smoothly now. Had to disingage in 50.2 for fear of slamming into distant cars stopped at intersections. No sign of slowing at all. Felt like I was playing chicken. It’s nearly perfect now!

I just put over 700 miles on 50.3 today. Wow is all I can say. 99% of the drive was on AP, and it was solid as a rock. In my opinion, they've got it pretty well nailed down now.
 
Have you tested it in the light or dark yet?

Not yet, just got it last thing last night. I have presented my analysis and estimate. We shall see over time if I was right of course. :)

BTW first version may well be bad in dark. But my belief is they will get it working eventually - for the reasons outlined above... (But I've also explained why I believe it is/has not been exactly easy.)
 
Last edited:
Here’s something to help see clearly tomorrow, just in case :).

And I knew what that was gonna be even before I clicked on it. It's like the TMC version of being rick-rolled. :)

Looks like there's no love for AP1 cars with this version, but oh well, no update for us means no opportunity to break anything either.

Bruce.
 
The use of an existing sensor as the reference data source is interesting, but I think does not achieve the real requirements of auto-wipe.

Your argument seems to be that, because AP needs to detect a malfunctioning camera, it must therefore be automagically capable of detecting rain as well.

Those are two different problems. The rain sensor not only needs to recognize rain, but it needs to estimate how often to run the wipers. It's job is to keep the windshield clear.

AP's job is to detect a malfunctioning camera. In response, it may try to run the wipers, but rain estimation is not the goal.

Division of labor would so well here. Use the right tool for the right job. A separation of concerns also enhances safety: if the front camera did go offline, forcing AP to disengage, or if AP crashed due to a hardware or software issue, you don't want the wipers to stop working in a storm. The driver already has their hands full regaining control of the car from AP; they shouldn't have to think about why they can longer see through the windshield.
 
I really can't figure out why this took so long. Most other manufacturers have figured this out long ago. My Lexus from 10 years ago had this feature and it worked flawlessly. I wonder if Telsa is using a different set of sensors than most traditional auto manufacturers which led to a more ground up design?

They are indeed using different sensors from the rest of the industry. They abandoned the existing rain sensor and decided to use the camera(s) and image processing to detect rain.

This change provoked a number of complaints from folks with this revised hardware as they received expensive cars without something many consider a basic function.

Looks like it might all be OK now that the software development has caught up with the hardware changes. And just in time for me to take delivery in the next few weeks of possibly one of the last batch of 2017 VIN model S RHD cars to come off the production line.
 
Before we assume his rain sensing is not working at night, let's be sure he has it set up properly. It seems some are anxious to show it doesn't work as it should at night. There's plenty of time for that.

Setting switched to ON - check
Stalk set to either of two options - check
No sun in sky - check
Pissing down outside - check
Windshield wiped - negative

Hopefully it will rain today. Last time I said that was during a cricket match that England were about to [inevitably] lose.
 
  • Funny
Reactions: PrGrPa
Can someone explain why the need for a neural net for rain-sensing wipers? I don't understand. Seems like an overengineering feat to me?

That was my original thought: why abandon a ‘simple’, proven technology for software. I suppose in the end it is simplification of a sort. If you can emulate lots of individual systems using the Tesla’s ‘big brain’ computer, not only are you reducing your hardware, wiring, and interface burden, you are also making the control software ‘easily’ adjustable. I suppose it also potentially reduces the number of parts suppliers involved in the vehicle.
 
Your argument seems to be that, because AP needs to detect a malfunctioning camera, it must therefore be automagically capable of detecting rain as well.

Those are two different problems. The rain sensor not only needs to recognize rain, but it needs to estimate how often to run the wipers. It's job is to keep the windshield clear.

AP's job is to detect a malfunctioning camera. In response, it may try to run the wipers, but rain estimation is not the goal.

Division of labor would so well here. Use the right tool for the right job. A separation of concerns also enhances safety: if the front camera did go offline, forcing AP to disengage, or if AP crashed due to a hardware or software issue, you don't want the wipers to stop working in a storm. The driver already has their hands full regaining control of the car from AP; they shouldn't have to think about why they can longer see through the windshield.

Not malfunctioning camera, obscured camera.
AP needs to be aware if there is rain, snow, ice, salt, dew, etc altering the image. A rain sensor does one thing well, detecting rain in front of the sensor. It cannot detect rain or anything else in front of the camera (unless the moisture distribution is uniform and of the type the sensor is triggered by). It cannot report a moth impact for instance.
So a rain sensor would allow auto-wipe of the type cars currently have, but would not further the development of the type AP needs.

Side idea: If Tesla added an IR LED (activated once every X frames) to the camera module offset from the windshield, that would allow checking of optical clarity via back scatering/ reflection at night.
 
  • Informative
  • Love
Reactions: lunitiks and Rouget
I still hate how it takes so long to start up again after traffic starts moving. Maybe I’m just impatient but that drives me nuts.

I think that the TACC separation between me and the car ahead has to established for 'awhile' before it thinks of moving forward 'chill' mode may help.
In my case I have the separation set to 5 car lengths for TACC the 'awhile ' time is not adjustable as far as I know.
 
Not yet, just got it last thing last night. I have presented my analysis and estimate. We shall see over time if I was right of course. :)

BTW first version may well be bad in dark. But my belief is they will get it working eventually - for the reasons outlined above... (But I've also explained why I believe it is/has not been exactly easy.)

The first part of my commute this morning was along a dark country lane with no streetlights. The autowipers did not really activate on either setting however light from oncoming traffic triggered them straight away. I'd say this initial release does not work well in the dark, sure it will get better though and glad to finally have it.