@footbag - Any discussion about autonomous driving in snowy conditions? IMHO it doesn't seem like Tesla has considered this very much as the front radar under the nose cone takes about 10 seconds to get covered with snow or slush in Canadian winters. Low contrast lines on I-405 are one thing, but a layer of ice and snow on Highway 2 is much harder to deal with.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Info on Autopilot + v7
- Thread starter Footbag
- Start date
Thanks for the pictures, Blurry_Eyed. Based on the indentation on the rear bumper and the laser scanner equipment that you spotted within, I think we now know what the indentation on the rear bumper is for on the Model X.
steph280
Member
No. GPS antenna (along with cell antennas) are inside the passenger side mirror housing.
Carbon fiber side view mirror caps and wha't under them
Carbon fiber side view mirror caps and wha't under them
Last edited:
rxlawdude
Active Member
What part of "Limitations: Autopilot components may not be fully functional or may provide inaccurate information in these situations: Visibility poor (heavy rain, SNOW, fog, etc.), bright light interfering with the camera, ... Model S is being driven in extremely hot or cold conditions" has eluded you?
Show me one from any manufacturer that doesn't have such disclaimers AND works in the conditions you describe. Good luck! :smile:
There were six sensors around the test car that I saw at Mt. Shasta (One by each wheelwell for a total of 4 on the sides of the car and one in the front nose of the car, where the current camera is for autopilot equipped cars, and one in the middle of the back bumper) and the Ibeo website says the LUX Fusion system can sync up to six sensors, so that makes total sense. Maybe Tesla is trying to do a version of fusing in software where they take the ultrasonic data they get from the ultrasonics around the car and then fuse it with the rear laser sensor and the tri-camera that is mounted on the inside of the rearview mirror of the car and come up with a lower cost solution to make the rear detection better than a mono camera system might allow. That would be pretty intriguing for the Model X autopilot system.
I'm very sure they won't obligate themselves to autopilot working in anything but optimal conditions. Everything will have a disclaimer attached.
Todd Burch
14-Year Member
And they shouldn't obligate themselves beyond that anyway. After all, it's a visual system. It's a camera detecting the lanes. If the camera can't see lane markings, I don't see how people would expect lane keeping to work.
Snow on the ground? Heavy rain? Fog? Faint lines? Lots of extraneous pavement markings? Don't get your hopes up about lane keeping working in these situations. They don't in other cars, and won't in any car for a long time.
Andyw2100
Well-Known Member
I'm not suggesting this thread has had no effect, but I think a little announcement about Ludicrous mode, and the distraction that has caused might just also be a factor.
No discussion about that, sorry. As others point out, not much Tesla can do about those types of issues in the near future. I was keeping an eye on our main freeways through town last winter... they typically were clear (what I would expect would be needed for autopilot) within a couple days of most snow falls, and we don't get snow every day. Granted, it may be different in Toronto. I think I would be satisfied getting to use Autopilot 50%-75% in the winter, and would understand that the remainder of it, perhaps it is better that a person is in control - even just for ice and stupid drivers in the conditions.
brucet999
Active Member
+1 Until the autonomous system can look ahead two or three cars looking for brake lights or squirrely driving, humans will do better in bad weather conditions.
lennier
Member
The MobilEye machine vision system Tesla uses is certainly working on methods that don't rely on seeing lane markings, but which interpret other features in the image to infer where the vehicle should be tracking, in the same way humans do in such conditions. Assuming they achieve a good enough result then I'd expect Tesla will start exploiting that information in future releases. There will always be limits of course, but they may well push further than people think based on assumptions about tracking lane markings.
kuttakamina
Member
Speak for yourself buddy. Perhaps in some parallel universe or another plane of time.
In this reality, Autopilot is late. Way late than it should have been. This thread makes no difference to that.
And 11 months after the announcement, and the website bragging about incredible features that just aren't there, is simply unjustifiable.
They should have rolled out Autopilot 6 months ago. If not, they need to roll that out NOW!
scottf200
Well-Known Member
So even at the risk of it not working very very good and perhaps some major accident happening with potential lives at stake.
But you're angry and just wanted it delivered "NOW" because it is late regardless of the consequences.
Pretty sad perspective you have.
kuttakamina
Member
So please tell me where did I say "please roll it out now even if it is unsafe". I don't think I ever said that.
The very fact that they don't have the capability to make a safe autopilot after 10 months of beating their chests about it .. is just flat out wrong.
I don't see anything wrong about being frustrated about not being delivered, what was promised, and what has been paid for.
You may give them slack for being the earth saver great company they are. Realize that some of us are just customers who want what we paid for.
This is not a kickstarter, neither is it charity, nor am I a Tesla investor. I am a Tesla customer, demanding for what I have paid for.
I don't see anything wrong with asking for what I was promised, and have paid for.
And just to be clear, I didn't pay for an unsafe autopilot. That would be ridiculous no?
The way I see it, I'd be happy for them to roll it out as beta and we can make it better by sending bug reports and providing statistics.
When it was first released, TACC had some pretty big issues (follow distance wasn't right, reaction was not correctly judged etc) but it got better over time and I think that'd be what we need with auto steering. Even just a rudimentary version that doesn't always work (but notifies you when it's disengaged) would still make it better than the Mercedes Benz system
When it was first released, TACC had some pretty big issues (follow distance wasn't right, reaction was not correctly judged etc) but it got better over time and I think that'd be what we need with auto steering. Even just a rudimentary version that doesn't always work (but notifies you when it's disengaged) would still make it better than the Mercedes Benz system
Okay, correction: I'm 95% sure that's a *high-precision* GPS antenna on the roof. I would think that either the standard GPS antenna doesn't give the desired precision, or they simply didn't want to tamper with the standard GPS antenna.
It's 9 months, not 11 months. Autopilot was announced in October.
- - - Updated - - -
You paid for a promise of something that would happen in the near future. We are still well within that time frame.
perkiset
... this one goes to 11
I think you obviously have never had to do a project around the house, where you had to take 3+ trips to Home Depot to actually get something done that was more difficult or involved than you had thought. Much less any software development. Further, developed anything where people's lives LITERALLY depend on it.
Your anger does not move the ball forward. We all paid for it and we will get it. And when we finally do, it will be probably 85% of what everyone wants and will STILL go through bug fixes. But it will be great.
I think this thread has been a great exposition of the fact that they are working their butts off, trying to get it right as quickly as possible. Thanks 'Foot for making such a diligent report.
