-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
The same way a colour blind person knows the colour of a light — by its position. Red is always at the top (when vertical) or on the driver's side (when horizontal).
MarkS22
Member
Promises made by Musk disagree:
"This version of Autopilot does not take into account stop signs and red lights, but a future software update will,” Musk says.
Source: Tesla’s Cars Now Drive Themselves, Kinda
Before anyone dismisses this, we're only going by the same promises for Level 5 autonomy. They must be accountable for their feature promises.
The comment above isn't "may" it's "will" and it's specifically a software update mentioned in 2015. But I've seen it mentioned a couple places in 2014 and after.
"This version of Autopilot does not take into account stop signs and red lights, but a future software update will,” Musk says.
Source: Tesla’s Cars Now Drive Themselves, Kinda
Before anyone dismisses this, we're only going by the same promises for Level 5 autonomy. They must be accountable for their feature promises.
The comment above isn't "may" it's "will" and it's specifically a software update mentioned in 2015. But I've seen it mentioned a couple places in 2014 and after.
In terms of improvement there is very little they can realistically do to improve it. It simply doesn't have the capability to do much beyond what it does right now regardless of what was claimed.
The biggest improvement that is potentially possible is with AEB/AP with the radar+white listing, and to me this would be a dramatic improvement.
I do expect Tesla to try to balance the desire for safety (in encouraging people not to use AP on undivided highways) with the desire to sell CPO cars. Most of the Model S/X's that have AP1 are going to be resold through Tesla. So they do have incentive to keep AP1 operating at a decent level. They also have incentive because an existing customer is very likely (especially with Tesla) to be a repeat customer.
They did get a lot of negative flack for what they pulled with the latest update. When they added limitations to AP without mentioning them in the release notes. A lot of owners including me sent some rather stern emails to Tesla stating how this wasn't acceptable. For a lot of us our biggest concern isn't that AP1 won't get updated, but that it will continue to get cut/tweaked to where there isn't anything left.
For me personally my biggest concern is not having unity among the owners of AP1 to fight back against some of the recent rationality used to limit AP. For example Elon Musk said he felt that people were going too fast on twisty roads so that's why they limited AP to the speed limit on undivided roads. That to me is NOT an acceptable reason to monkey with the entire fleet. It's like we're going to be continually penalized for what some idiots do. There was no regulatory reason or anything that really raised to the level of what I see as a justifiable reason. As a consumer base I do think we need to do a better job pushing back.
The biggest improvement that is potentially possible is with AEB/AP with the radar+white listing, and to me this would be a dramatic improvement.
I do expect Tesla to try to balance the desire for safety (in encouraging people not to use AP on undivided highways) with the desire to sell CPO cars. Most of the Model S/X's that have AP1 are going to be resold through Tesla. So they do have incentive to keep AP1 operating at a decent level. They also have incentive because an existing customer is very likely (especially with Tesla) to be a repeat customer.
They did get a lot of negative flack for what they pulled with the latest update. When they added limitations to AP without mentioning them in the release notes. A lot of owners including me sent some rather stern emails to Tesla stating how this wasn't acceptable. For a lot of us our biggest concern isn't that AP1 won't get updated, but that it will continue to get cut/tweaked to where there isn't anything left.
For me personally my biggest concern is not having unity among the owners of AP1 to fight back against some of the recent rationality used to limit AP. For example Elon Musk said he felt that people were going too fast on twisty roads so that's why they limited AP to the speed limit on undivided roads. That to me is NOT an acceptable reason to monkey with the entire fleet. It's like we're going to be continually penalized for what some idiots do. There was no regulatory reason or anything that really raised to the level of what I see as a justifiable reason. As a consumer base I do think we need to do a better job pushing back.
I'm beginning to think this is right as I've had the same experience. Also, based on earlier Elon comments, I think there will be a greater dependence on radar. If this is the case, we may see AP1 improvements.
The stop sign detection is already built into the MobileEye stuff. All Tesla has to do is activate it, but I don't think they will. The problem is it comes with a lot of liability, and there is only one front sensor. Plus if they were going to do it we would have likely seen it by now.
The pedestrian detection is within AP. Not sure if they'll add it to TACC, or AEB. I also don't know how good this would be since my main concern with pedestrians is hitting them at an intersection while turning right in the dark/rain. I don't know if the single front sensor would even be able to see them in time. Not only does the camera have to see them, but it has to have enough visual information for the neural net to detect it as a pedestrian.
But, even when AP1 works in the rain it still doesn't operate as well as a good driver. The problem is AP lane-steering tries to center the car and that means placing the wheels in the worst possible location. Where there is a greater depth of water than the edges of the lane.
Suffice to say I don't use AP in heavy rain or after a heavy rain. I do love using TACC in the rain though.
A big reason Tesla will continue to improve AP 1.0 is to avoid refunding the $2500 every owner paid for it. Remember - summon, autopark and autosteer are all still considered "beta" software. If those features are never released then users will be able to demand their money back.
They would likely take everything away if one opted to get your proposed $2500 refund. I would not want to give up TACC, which would be included in what would be lost.
Dborn
Confirmed
There could easily be financial advantage to Tesla in allowing the central screen to be updated to the latest computer and also the sonar suite. The rear camera now is also a higher definition one and easily swapped out. There is thus significant potential for improvement to AP1. Mobileye sells a stand alone unit that detects pedestrians and has done for years now. There is no reason that could not be activated.
I agree with this. TACC works exceptionally well and I no longer want to drive a car without it.
In its current form, autosteer works well in nice weather and heavy stop and go traffic (even with the god-awful non-divided highway speed limitations). Even without a single improvement, AP1 is well worth the additional $2500 -- a bargain.
I hope Tesla enhances our AP1 experience but I'm not expecting it. I plan to simply enjoy my Tesla like pre-AP Tesla owners did (and are doing) until autonomous driving is fully operational. By that time, there will likely be other technology improvements that will make upgrading even more enticing.
ig_epower
Member
I am anxious to see when they will release 8.1 to see if it improves my situation. The speed limit on undivided highways is one thing but yesterday, I experienced (while on AP) an automatic reduction from going at 100 kph (on a 100 kph limit) to 85 kph while on an HOV lane! HOV lanes have added space from the regular lanes. For a brief section, the speed just dropped to 85 without warning and there were no cars in front of me. Luckily there weren't any cars behind me either. I tried to override the speed setting back to 100 without any luck. I will test that section of highway again in the next couple days to see what's going on. I should have used the online "Note:" function but I forgot about it.
I think that there are improvements that can be made within the AP 1.0 hardware suite:
- Improved reliability and stability of Autosteer. (Ping Pong, Truck love--you know.)
- More graceful stopping when approaching a stopped car from a distance
- Stop for traffic signs and signals
- Handle highway intersections more reliably when not following another car and not wander off the road. (Generally I think the car should keep going in a straight line and ask for help if it can't decide what to do, rather than go off on a tangent looking for lines.)
- Improved Summon. Currently Autopark can do highly accurate and sophisticated maneuvers to get into parallel and perpendicular parking places. There's no reason the car couldn't back itself into a garage without elaborate pre-positioning.
- Implement the documented feature to slow down when signaling to exit a freeway (documented in Owners Manual in 7.1 and 8.,0)
- Take an exit ramp when signaled (promised 8.0)
- Take an exit ramp under navigation (promised 8.1)
Yes there is a reason. TSLA stock price.
A good rule of thumb going forward is if it does not help Tesla stock price - it will not happen.
RogerHScott
Active Member
Driving westward at sunset there are times when I can't reliably read a red light. How are the HW1 monochrome cameras going
to do this reliably enough to be anything more than "backup" to the driver? What about when there's a tall truck between you and the
light and it chooses to run the light (or the very "ripe" yellow)?
RogerHScott
Active Member
Good luck getting any court in the land to rule that the word "beta" carries any meaning whatsoever.
ElectricTundra
P85D AP1
A few thoughts...
Without a camera there is no lane keeping? Can radar (or LIDAR) see painted lines? How do those trucks see the lines without a camera?
Stop sign recognition was in the original AP1 code (and some betas?) but was not rolled out. My guess is that there are far too many instances where it did not recognize a stop sign (like covered with snow, branches in front of it, etc.) or did not adequately account for slippery road conditions. I wouldn't count on ever seeing this capability for HW1 as the liability would be far too great even with gobs of verbiage about how the driver is still in control. There would be fatalities, bad PR and lawsuits when it didn't see a sign in time and the driver didn't take over soon enough. They can certainly augment w/ map data (or likely use map data as primary and augment with vision) but no idea how difficult that would be to maintain for both versions nor how reliable. Elon doesn't bow to lawyers often, but...
I'd guess that swapping many of the sensors could result in overload on the AP CPU that is apparently already at its limit. How much more processing is required for a higher resolution image? Or conversely, if it treats the higher resolution image as the original lower resolution then what advantage was there in getting the new sensors? On top of this is the problem of too many hardware variations requiring too much code and QA.
I'm not sure current hardware can stop much more gracefully as it apparently can't see far enough ahead to do so. We might possibly see some improvement (and I'd very much welcome it) but I'm not holding my breath for it responding as smoothly as I do manually. I think we could and hopefully will see some improvement in the smoothness of rubber-banding when following in stop & go traffic.
Without a camera there is no lane keeping? Can radar (or LIDAR) see painted lines? How do those trucks see the lines without a camera?
Stop sign recognition was in the original AP1 code (and some betas?) but was not rolled out. My guess is that there are far too many instances where it did not recognize a stop sign (like covered with snow, branches in front of it, etc.) or did not adequately account for slippery road conditions. I wouldn't count on ever seeing this capability for HW1 as the liability would be far too great even with gobs of verbiage about how the driver is still in control. There would be fatalities, bad PR and lawsuits when it didn't see a sign in time and the driver didn't take over soon enough. They can certainly augment w/ map data (or likely use map data as primary and augment with vision) but no idea how difficult that would be to maintain for both versions nor how reliable. Elon doesn't bow to lawyers often, but...
I'd guess that swapping many of the sensors could result in overload on the AP CPU that is apparently already at its limit. How much more processing is required for a higher resolution image? Or conversely, if it treats the higher resolution image as the original lower resolution then what advantage was there in getting the new sensors? On top of this is the problem of too many hardware variations requiring too much code and QA.
I'm not sure current hardware can stop much more gracefully as it apparently can't see far enough ahead to do so. We might possibly see some improvement (and I'd very much welcome it) but I'm not holding my breath for it responding as smoothly as I do manually. I think we could and hopefully will see some improvement in the smoothness of rubber-banding when following in stop & go traffic.
Part of that could be that not everyone sees that the same way you do. You see it as them taking away a feature, others see it as adding safety features(identifying roads on which Tesla cannot guarantee the safety of using autopilot over the speed limit and adjusting speed accordingly) that weren't there before.
MarkS22
Member
Then you're too close to the truck. I know people who have been ticketed for exactly this. Autopilot can already categorize trucks and could simply be more conservative in following distance.
We've grown so impatient and believe we all have the right to break the laws of the road "because everyone does it so it's dangerous not to." Many people are going to be disappointed when autonomous cars drive defensively and follow the laws.
ElectricTundra
P85D AP1
What does Tesla do? If they've made promises of features (stopping at stop signs, summon to front door, etc.) that they will not be able to implement? At what point are such public promises a contract?
People chose to purchase a $100k+ car, some based on getting the promised features. Will Tesla have any legal or ethical obligation? Is refunding $2500 sufficient if the impact is having purchased an entire car not just a single AP product? Would offering a substantial discount on upgrading to a HW2 be more appropriate or completely over the top?
People chose to purchase a $100k+ car, some based on getting the promised features. Will Tesla have any legal or ethical obligation? Is refunding $2500 sufficient if the impact is having purchased an entire car not just a single AP product? Would offering a substantial discount on upgrading to a HW2 be more appropriate or completely over the top?
Similar threads
