Theory: Autopilot hardware upgradeability likely now a priority?

[Vitold]

Unless AP 2.0 is fully autonomous there's no need to rush it - 1.0 is the best there is right now and it still being tweaked.

 

[Vitold]

I don't see lane holding autopilots stopping for lights or stop signs for this reason. There is always one hidden behind a branch or something.

Only way would be if it was implemented as a warning. Traffic light (sign), detected. Will stop in X feet. Countdown. Start to brake. If driver didn't see countdown then they are on their own.

Most likely in the future traffic lights/intersections will communicate with the car itself.

 

[RogerHScott]

Tesla will have an economies of scale advantage if this is the same hardware going into Model 3

Except that there are no economies of scale for labor (or, at least, much less so than for parts).

 

[sillydriver]

A couple people here have suggested that AP functionality will follow high-tech industry practice and eventually be end-of-lifed for sensor 1.0 cars, so that AP will eventually be turned off for these cars. I doubt this will happen for many years because it is so far out of the norm of what is acceptable in the car industry. It would be rather like selling cars with air conditioning and announcing a year later that the customers' air conditioning will cease to function at the end of next year. I believe what will happen instead is either 1) continuous hands-on will be required for AP 1.0 or 2) AP 1.0 will be geofenced to operate only on limited access highways, where vehicles are unlikely to cross the road.

I hope and believe (and predict) the outcome will be (2). I think Tesla will do (1) only if the regulators force them to.

 

[sandpiper]

Tesla's private high resolution lane maps may be the key to this. Tesla knows or will know exactly where all the stop signs are.

Until a new one is added or removed. Or until there is construction in the area. Right now the nav maps in my car don't include some new roads built 12 months. They will never be able to keep up with the new stop signs & lights that city council adds every time some busybody is unhappy that actual cars are driving past their house.

 

[commasign]

Until a new one is added or removed. Or until there is construction in the area. Right now the nav maps in my car don't include some new roads built 12 months. They will never be able to keep up with the new stop signs & lights that city council adds every time some busybody is unhappy that actual cars are driving past their house.

True, but that's going to be an issue for everyone whether they are driving a regular car, fully autonomous car, or a Tesla with autopilot 2.0. If the stop sign isn't visible, it's not there. What I meant to say was that Tesla will know every stop sign and traffic light that a prior Tesla has stopped at with autopilot off. Combine that with improved camera system that can recognize stop signs and traffic lights most of the time and you've got a fairly redundant and safe, continuously improving system.

 

[sandpiper]

True, but that's going to be an issue for everyone whether they are driving a regular car, fully autonomous car, or a Tesla with autopilot 2.0. If the stop sign isn't visible, it's not there. What I meant to say was that Tesla will know every stop sign and traffic light that a prior Tesla has stopped at with autopilot off. Combine that with improved camera system that can recognize stop signs and traffic lights most of the time and you've got a fairly redundant and safe, continuously improving system.

The law of large numbers suggests that at some point, a new stop sign won't be in the map, will be obscured and somebody will be hit. And even if performance is better than human, there will be hell to pay - as we see this week. We won't see full automation for 20 years+.

 

[thegruf]

imho the hardware no way will the hardware be upgraded.

Any upgrades would be complex, labour intensive, invite new criticism and possible reliability issues, rework for rattles and creaks and other resulting damage from the work. And who would pay for it?

Plus, Tesla simply do not have the resource to do so, their SCs can barely keep up with the service work.

Not only that but Tesla damage their forward order book too.

Hardware upgrades are a pipe dream, sorry.

 

[Gizmotoy]

Except that there are no economies of scale for labor (or, at least, much less so than for parts).

Right. Can you imagine the labor required to retrofit a bunch of new sensors to, what, 80k+ vehicles? The SCs can barely handle the current load, let alone requiring a huge portion of the fleet to come back in for new hardware.

 

[JoshG]

My thoughts: I think you are nuts.

Do you expect Mercedes, Ford, GM, etc. to retro-fit their older cars as a new model year gets new/better features than the prior model year?

You don't seem to follow what I said, or I'm not being clear. I do NOT expect Tesla to upgrade everyone with new hardware for all kinds of new features as new things come out. I didn't expect upgrades when my car didn't have parking sensors or folding mirrors. I didn't expect a free center console. I didn't expect a free front motor to get me AWD.

And your analogy is flawed... Mercedes, GM, Ford haven't released cars with auto-steer or autopilot and described it the way Tesla have. My point is that this feature from Tesla MAY be different than past hardware upgrades and different from other auto makers that also come out with nice convenience upgrades. It may prove to be less costly to upgrade as many people as possible to prevent confusion about the capabilities here (and that confusion is already showing itself in the market). And, Tesla may have realized this a while back and started making it easier to upgrade AP hardware to newer version with existing wiring harnesses or connectors that could be in the car already.

Here's maybe another example to make the point. If Tesla comes out next week with ventilated seats for Model S, and the new console with rear cupholders and USB, I'm going to be slightly bummed I didn't get those things, but I completely accept it and have been enjoying my 2nd Model S these weeks as is. It has everything they said it would have when I bought it. Believe it or not, I'd feel the same about new AP hardware. Mine does what I thought it would do when I bought it and I'm not going to demand I get the new stuff that comes in the future. But my point is that Tesla may see reasons (financial, legal, regulatory, or ethical) to upgrade AP1 to AP2 to avoid problems where people misunderstand what their car's capabilities are in steering and braking itself. There's little danger of serious issues, collisions, or injuries arising from confusion about whether one's car has rear cupholders... but there might be with something sold as "Autopilot."

I accept that the majority in this thread think that AP hardware upgrades are unlikely... in their opinion, the costs to Tesla outweigh the benefit of this upgrade. I see the calculus slightly differently given what Autopilot does or doesn't do, but that's fine.

 

[shokunin]

A couple people here have suggested that AP functionality will follow high-tech industry practice and eventually be end-of-lifed for sensor 1.0 cars, so that AP will eventually be turned off for these cars. I doubt this will happen for many years because it is so far out of the norm of what is acceptable in the car industry. It would be rather like selling cars with air conditioning and announcing a year later that the customers' air conditioning will cease to function at the end of next year. I believe what will happen instead is either 1) continuous hands-on will be required for AP 1.0 or 2) AP 1.0 will be geofenced to operate only on limited access highways, where vehicles are unlikely to cross the road.

I hope and believe (and predict) the outcome will be (2). I think Tesla will do (1) only if the regulators force them to.

I don't think Tesla is going to deprecate AP 1.0 capabilities and turn them off. Rather AP 1.0 will have a defined use-case and possibly with more restrictions based on geofenced data on where/when it can be used. I have to believe that even with AP2.0, the algorithms and mapping will also improve AP 1.0 capabilities on the highway.