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

Anything that current AP hardware is not capable of?

This site may earn commission on affiliate links.
Hi,

I have had my classic S85 for a year (one of those built right before AP rolled out), and I'm tempting to upgrade to 85D. Autosteer and even TACC alone would be such a lifesaver in stop-and-go traffic.

Now, as the other thread suggested, AP 2.0 may be coming, and this is one day after I put down $2500 for an inventory 85D. Speaking of bad timing...

I agree with the idea that if there's something specific the current cars can't do, I should wait for the feature, but otherwise I should proceed with what makes me happy today. The problem is - I don't know what the current AP hardware can't do! EM said it could do everything - i.e. drive from parking lot to parking lot from SF to Seattle, but if that's the case why would there be the new 8-camera system anyway?

Question for you guys - what do you think the current AP hardware is incapable of? Like, detecting pedestrians popping up in the blind spots?
How soon do you think will there be a software feature that makes an APv2 car do more than APv1 car?

Thanks very much.
 
Hi,

I have had my classic S85 for a year (one of those built right before AP rolled out), and I'm tempting to upgrade to 85D. Autosteer and even TACC alone would be such a lifesaver in stop-and-go traffic.

Now, as the other thread suggested, AP 2.0 may be coming, and this is one day after I put down $2500 for an inventory 85D. Speaking of bad timing...

I agree with the idea that if there's something specific the current cars can't do, I should wait for the feature, but otherwise I should proceed with what makes me happy today. The problem is - I don't know what the current AP hardware can't do! EM said it could do everything - i.e. drive from parking lot to parking lot from SF to Seattle, but if that's the case why would there be the new 8-camera system anyway?

Question for you guys - what do you think the current AP hardware is incapable of? Like, detecting pedestrians popping up in the blind spots?
How soon do you think will there be a software feature that makes an APv2 car do more than APv1 car?

Thanks very much.

They've been building cars with hardware for AP since 2014. I'm sure you'll have it.

Yes, there is LOTS that AP doesn't do. It doesn't do any recognition AT ALL other than the sides of the road, and that depends on both sides being painted. Even then it gets fooled by intersections, turn lanes, tar strips, sunlight, and old paint.

It is short sighted: It doesn't see past a slight curve, can't see over a slight rise and dip. It cannot visualize or see the road more than about ten car lengths, if that. We, on the other hand, watch the road far, far ahead of where we are.

TACC is not autosteer. It uses radar to detect the car ahead of you and holds a set distance that you choose. It is very helpful in stop and go traffic, but it does not see people or read signs. It ONLY sees the car ahead of you, not past it.

AP is in very primitive beginner stages. That's why it's beta. It will improve, and Tesla will be among the first to get it, but right now AP is not that much of a much. Get your car, let it do what it can, and you will be very happy.
 
They've been building cars with hardware for AP since 2014. I'm sure you'll have it.

Yes, there is LOTS that AP doesn't do. It doesn't do any recognition AT ALL other than the sides of the road, and that depends on both sides being painted. Even then it gets fooled by intersections, turn lanes, tar strips, sunlight, and old paint.

It is short sighted: It doesn't see past a slight curve, can't see over a slight rise and dip. It cannot visualize or see the road more than about ten car lengths, if that. We, on the other hand, watch the road far, far ahead of where we are.

TACC is not autosteer. It uses radar to detect the car ahead of you and holds a set distance that you choose. It is very helpful in stop and go traffic, but it does not see people or read signs. It ONLY sees the car ahead of you, not past it.

AP is in very primitive beginner stages. That's why it's beta. It will improve, and Tesla will be among the first to get it, but right now AP is not that much of a much. Get your car, let it do what it can, and you will be very happy.

One slight addition: the AP camera will recognize the car in front and follow it. It adds this to the lane model.

I would dispute "AP is not that much of a much". In my usage it is stellar and a genuinely great thing to have.
 
My main interest in AP is auto parking. Until proven wrong, my opinion is that the current AP hardware cannot auto park in stall parking unless there are obstructions on its sides (e.g. cars or other objects) due to its reliance on sonar sensors. To put it in another way, it cannot auto park using lane markings only.
 
My main interest in AP is auto parking. Until proven wrong, my opinion is that the current AP hardware cannot auto park in stall parking unless there are obstructions on its sides (e.g. cars or other objects) due to its reliance on sonar sensors. To put it in another way, it cannot auto park using lane markings only.

Just curious, but that seems like a very minor feature to me. Why is it at the top of your list vs. semi-autonomous driving? I'm not sure I'd use auto-stall parking ever, given the kinds of weird outcroppings and obstructions that could be below sensor level on even a car bristling with cameras.
 
Hi,

I have had my classic S85 for a year (one of those built right before AP rolled out), and I'm tempting to upgrade to 85D. Autosteer and even TACC alone would be such a lifesaver in stop-and-go traffic.

Now, as the other thread suggested, AP 2.0 may be coming, and this is one day after I put down $2500 for an inventory 85D. Speaking of bad timing...

I agree with the idea that if there's something specific the current cars can't do, I should wait for the feature, but otherwise I should proceed with what makes me happy today. The problem is - I don't know what the current AP hardware can't do! EM said it could do everything - i.e. drive from parking lot to parking lot from SF to Seattle, but if that's the case why would there be the new 8-camera system anyway?

Question for you guys - what do you think the current AP hardware is incapable of? Like, detecting pedestrians popping up in the blind spots?
How soon do you think will there be a software feature that makes an APv2 car do more than APv1 car?

Thanks very much.

I wouldn't worry about AP 2.0. It's a long way off even if they deploy the hardware in 6 months. Look how long AP1 took from HW deployment to software availability with most (not all even yet) of the promised features.
 
My main interest in AP is auto parking. Until proven wrong, my opinion is that the current AP hardware cannot auto park in stall parking unless there are obstructions on its sides (e.g. cars or other objects) due to its reliance on sonar sensors. To put it in another way, it cannot auto park using lane markings only.

If it sees the lines on the road for Autosteer, why isn't it theoretically possible that it could see parking lines in your "stall parking" scenario?
 
Just curious, but that seems like a very minor feature to me. Why is it at the top of your list vs. semi-autonomous driving? I'm not sure I'd use auto-stall parking ever, given the kinds of weird outcroppings and obstructions that could be below sensor level on even a car bristling with cameras.
I do use auto steer and TACC, and I follow to 99.9% the rules of the release notes. So in that aspect, I am quite satisfied. While I enjoy leisure driving, I don't enjoy parking in office parkade and shopping center tight parking stalls.
 
If it sees the lines on the road for Autosteer, why isn't it theoretically possible that it could see parking lines in your "stall parking" scenario?
Because I park back-in to the stall. I am not sure that the rear camera is integrated into AP. Any way to find out?

- - - Updated - - -

Fair enough. I have to admit, I still haven't gotten good at parking this thing after almost 3 years.
I must admit that parking MS is much easier in USA than in Vancouver, BC. MS is a large vehicle compared to most of the other Japanese, Korean, European sedans.
 
Hi,

I have had my classic S85 for a year (one of those built right before AP rolled out), and I'm tempting to upgrade to 85D. Autosteer and even TACC alone would be such a lifesaver in stop-and-go traffic.

Now, as the other thread suggested, AP 2.0 may be coming, and this is one day after I put down $2500 for an inventory 85D. Speaking of bad timing...

I agree with the idea that if there's something specific the current cars can't do, I should wait for the feature, but otherwise I should proceed with what makes me happy today. The problem is - I don't know what the current AP hardware can't do! EM said it could do everything - i.e. drive from parking lot to parking lot from SF to Seattle, but if that's the case why would there be the new 8-camera system anyway?

Question for you guys - what do you think the current AP hardware is incapable of? Like, detecting pedestrians popping up in the blind spots?
How soon do you think will there be a software feature that makes an APv2 car do more than APv1 car?

Thanks very much.

Have you tested an auto-pilot car? There is no reason you shouldn't be fully informed, unless there is no way for you to get a test drive where you are. Auto-pilot is both "great" - because its first and it works well if the lanes are well-marked, but also "not-so-great", in that you just want so much more when you use it!

I want it to know not to take off ramps. I want it to change freeways and follow navigation. I want it to never TOE (take-over-immediately). I want it to travel smoothly through curves. I want it to never jerk suddenly away from the divider it doesn't see until the last second, when any human would have seen it long ago. I want it to avoid cars that pull over on top of me, like I do (there's going to be a wreck with auto-pilot where a car pulls a dick-move on somebody and assumes the human in the Tesla will slam on the brakes, but the car's going to be on auto-pilot so it won't happen)

I guess I'm trying to tell you that as great as auto-pilot is, its got a LONG way to go, and it may not really get much better than it is with the sensors the current model S has. Sounds like you're already worried about this. I can assure you that 2 years from now you'll be able to buy a car with much better auto-pilot capabilities, if not fully autonomous.

This technology is going to improve fast. But maybe not on the current hardware in the Model S today.
 
I want it to avoid cars that pull over on top of me, like I do (there's going to be a wreck with auto-pilot where a car pulls a dick-move on somebody and assumes the human in the Tesla will slam on the brakes, but the car's going to be on auto-pilot so it won't happen)

There's already video of the car handling this quite well. Not sure where now, but it was from Russia I think. The car hit the brakes when a Taxi cut them off (and would have hit them).
 
Because I park back-in to the stall. I am not sure that the rear camera is integrated into AP. Any way to find out?

I don't know but initially it could be set up similar to auto parallel parking with defined rules. Sensors have to see both lines, centers between lines or cars and you have to park face in. Back in phase 2.
 
I'd actually say the current level of AutoPilot is really limited in what it can do. Sure it can do lane-centering, and it can do adaptive cruise control. It can alleviate some of the burden of driving especially when in heavy stop and go traffic or out in the middle of nowhere. In some ways this is a good thing because in terms of regulations Tesla isn't really doing that much to cause a stir. Sure there is some mumbling, but there doesn't seem to be any OMG reactionary stuff. Maybe a little tightening down, but nothing major at this point.

With Autopilot 1 the driver really does have to be attentive and ready to take over. If you want to do a lane change you absolutely have to check. The side spot monitoring might or might tell you that there is a car next to you. The car also has absolutely no idea if someone is coming up quickly.

The Sensors with AutoPilot 1.0 are very limited. The front camera/radar/ultrasonic is the only section of the car that combines multiple sensors to get a really good feel for what's coming up. The front sensing works fairly well as a result.

The side monitoring - Not so much
The rear monitoring - haha

While there hasn't been any confirmation about what AutoPilot 2.0 will have I'm pretty sure it will eliminate those limitations by using multiple forms of sensing. It will also further enhance the front vision which will hopefully correct issues with shadows, etc and hopefully will give a larger FOV.

With that being said it's going to be a long time from AutoPilot 1.0 HW/SW (where we're at now) to AutoPilot 2 HW/SW.

With my car (a 2015) it's easy because I just keep it until Autopilot 2.0 HW/SW.

With pre-autopilot it's a bit trickier since it's going to be years (at least 2) until AutoPilot 2.0 HW/SW, and also with AutoPilot 2.0 regulatory stuff is going to take a much larger role. At first it will just do what AutoPilot 1.0 can do, but only better.

AutoPilot 2.0 is really about going from Level II autonomous driving (semi-autonomous with a 100% driver responsibility) to Level III (Limited Self-Driving Automation where the driver can cede control of all functions). Now I'm not saying it will achieve this, but it will be way closer than the AutoPilot 1.0 hardware allows. Level III is when the manufacture might start taking some level of liability.
 
Have you tested an auto-pilot car? There is no reason you shouldn't be fully informed, unless there is no way for you to get a test drive where you are. Auto-pilot is both "great" - because its first and it works well if the lanes are well-marked, but also "not-so-great", in that you just want so much more when you use it!

I want it to know not to take off ramps. I want it to change freeways and follow navigation. I want it to never TOE (take-over-immediately). I want it to travel smoothly through curves. I want it to never jerk suddenly away from the divider it doesn't see until the last second, when any human would have seen it long ago. I want it to avoid cars that pull over on top of me, like I do (there's going to be a wreck with auto-pilot where a car pulls a dick-move on somebody and assumes the human in the Tesla will slam on the brakes, but the car's going to be on auto-pilot so it won't happen)

I guess I'm trying to tell you that as great as auto-pilot is, its got a LONG way to go, and it may not really get much better than it is with the sensors the current model S has. Sounds like you're already worried about this. I can assure you that 2 years from now you'll be able to buy a car with much better auto-pilot capabilities, if not fully autonomous.

This technology is going to improve fast. But maybe not on the current hardware in the Model S today.

Thanks everyone for the insightful comments. Yes, I have test driven one of these, and I've seen its limitations (it was a stormy day). But I was nonetheless totally blown away by it. My commute is not terribly long - it's just 4 miles, but every day when I'm driving I would think "gosh, this trip would be so much nicer if I can just turn on Autopilot, sit back and watch the machine do its thing, until the next red light." That pushed me over to take the depreciation hit and order a new 85D.

Now, speaking of limitations - in my mind there are 3 categories of limitations:

- Type I: things that can be fixed by software. (e.g. stall parking, reading red lights and stop signs, not steering towards exits)
- Type II: things that can be fixed with a reasonably-priced hardware retrofit. (e.g. ??? Did they actually build any upgradability into this car at all?)
- Type III: things that can only be fixed by ordering a new car. (e.g. Full Autonomous Driving™, of course, but anything more specific? like handling roundabouts?)

So what I'm struggling with is 1) what goes into Type III and 2) when will there be a car on the road that has a Type III feature.
In other words, if I bought this 85D today, when will I be thinking "gosh, this trip would be so much nicer if my car could do X, so I don't have to take over every time at this place" again.
If that's more than 3 years away, then well... who cares :p but if that's in 2016, I'll need to seriously think twice about my order.
 
I still cannot believe that after google having had auto driving cars for what a decade now that they didn't take a page out of their book and use google maps, / traffic updates that are updated with all the users using the map / location software. granted those are designed with full computers all through the cockpit of the car but software wise. not sure why they aren't using a superiority map software IE google maps. which is prob the most documented map now days(which most show relatively updated speed data as in the speed limit etc and can tell what the road is. ie curves beyond where you are. i do not think that the AP is integrated into the GPS in either case here. but i think should be. obv place more bias on the sensors but for good prediction etc use the maps.