Why AP 2.0 Won't Be Here Soon, and It Won't Be What You Think It Is

[Camera-Cruiser]

I think we are all talking off the top of our heads. All these problems are being worked on in parallel.

That's why I come here. Mostly. This the National Enquirer thread for Tesla Heads.

 

[Saghost]

OK, another scenario I thought about... think about secondary roads with a 45/50 or 55 mph speed limit and stop lights every few miles. In some states I know of (like New Jersey), these intersections have massively long yellow lights. What is the car to do if you're traveling along at 50mph, and the light turns yellow? Most people who travel these roads know that the long yellow lights allows you a lot of time to breeze right through long before a red light. But the car doesn't know that. If the car is close enough to detect a light changing from green to yellow, will it be too close to safely stop, when the driver behind you fully expects you to run the yellow light? At some point the car is going to have to decide that it's actually OK to run the yellow light to prevent being rear-ended or stopping so quickly it's unsafe for the passengers.

And the idea of Car-to-light infrastructure to communicate light status to the car is a pipe dream. The municipalities that have to pay for, install and maintain this equipment aren't going to spring for more complicated hardware so your car can know what your eyes already see. There's nothing in it for them to support that.

In most states, the length of that yellow is subject to specific regulations based on the speed limit of the road. Assuming that applies here, the car should be able to know enough to make the go/no go decision, no matter how long the yellow is.

Unless the car can't make out the light before it turns as you suggested in the second half. In that case, it'd have to plan to stop. That's the same thing a human would have to do, right?

At least, I hope none of the drivers around me see a yellow light for the first time and think they'll blow through, no matter how long it's been yellow...

 

[KublaConn]

I'm imagining the distant future, to when we have complete saturation of fully autonomous vehicles and these questions about stop signs and stop lights and such will be moot. By complete saturation, I mean both vehicles and requisite infrastructure, because the technology required to achieve this future, by necessity, will not be solely contained within the vehicles. In a world of fully autonomous cars, when it comes to the roadway and most traffic cues, there is no need for visual markers such as lights and signs that were designed for the human eye, because sensors and processors can "see" without such infrastructure and there will be a vital necessity for more precise "seeing" in order to achieve the stated goal of orders of magnitude in improved safety. If we want these vehicles to be safer than human-operated vehicles, then we will need to take full advantage their superior capabilities and not limit them to just what humans can achieve with their inferior senses. A fully autonomous vehicle will be in constant comunication with all the other vehicles around it on the roadways and it will also be in communication with the roadway itself. I imagine that along the roadway and at intersections, etc., all that will be needed is transmitters and receptors that will communicate with the vehicles and define the operating space for the vehicles. All the issues being brought up in this thread are focusing solely on the vehicles and that's understandable because what we're talking about here is about the different stages and progressions leading up to fully autonomous vehicles, and hey, it's a car forum, but many of these issues will be overcome by a paradigm shift in the basic infrastructure of our roadways. I imagine there will be fully autonomous "zones" where full saturation of the infrastructure will be in place and various stages of saturation in other areas and even some "wild west" frontier areas where the car is on its own as they roll out all of the changes. (Cue "Red Barchetta") Oh, and in this future world, there'll be spandex jackets, one for everyone...

Okay, back to talking about the changes to the Tesla and the expected rate of progression, sorry for the Arthure C. Clarke/Isaac Asimov/William Gibson/Rush/Donald Fagen distraction, but my muse got me to musing...again.

 

[msnow]

I'm imagining the distant future, to when we have complete saturation of fully autonomous vehicles and these questions about stop signs and stop lights and such will be moot. By complete saturation, I mean both vehicles and requisite infrastructure, because the technology required to achieve this future, by necessity, will not be solely contained within the vehicles. In a world of fully autonomous cars, when it comes to the roadway and most traffic cues, there is no need for visual markers such as lights and signs that were designed for the human eye, because sensors and processors can "see" without such infrastructure and there will be a vital necessity for more precise "seeing" in order to achieve the stated goal of orders of magnitude in improved safety. If we want these vehicles to be safer than human-operated vehicles, then we will need to take full advantage their superior capabilities and not limit them to just what humans can achieve with their inferior senses. A fully autonomous vehicle will be in constant comunication with all the other vehicles around it on the roadways and it will also be in communication with the roadway itself. I imagine that along the roadway and at intersections, etc., all that will be needed is transmitters and receptors that will communicate with the vehicles and define the operating space for the vehicles. All the issues being brought up in this thread are focusing solely on the vehicles and that's understandable because what we're talking about here is about the different stages and progressions leading up to fully autonomous vehicles, and hey, it's a car forum, but many of these issues will be overcome by a paradigm shift in the basic infrastructure of our roadways. I imagine there will be fully autonomous "zones" where full saturation of the infrastructure will be in place and various stages of saturation in other areas and even some "wild west" frontier areas where the car is on its own as they roll out all of the changes. (Cue "Red Barchetta") Oh, and in this future world, there'll be spandex jackets, one for everyone...

Okay, back to talking about the changes to the Tesla and the expected rate of progression, sorry for the Arthure C. Clarke/Isaac Asimov/William Gibson/Rush/Steely Dan distraction, but my muse got me to musing...again.

Which one was Steely Dan?

 

[KublaConn]

Which one was Steely Dan?

Oops! My bad. I must've been too high on the future, it addled my brain. Not Steely Dan, but rather, Donald Fagen's "I.G.Y." Thanx, fixed it.

 

[Todd Burch]

I'm mostly in agreement with the OP, even though I strongly disagree that some of the specific items on his 10-point list are in
any way technology blockers.

To be clear, I didn't mean to imply those items were technology blockers--just that a year into release and several years into development, most of those items haven't even been attempted yet. And in the big scheme of driving in the world, they're the "easy" scenarios.

 

[Todd Burch]

I think the OP missed a few of key points --

1. It is of great benefit to Tesla to ship hardware that will be level 4 capable in order that they can collect/analyze the resulting data, even if those cars never reach level 4.

I absolutely agree. I didn't miss that.

2. A key gating item to level 4 will be regulatory approval.

Also mentioned in the original post.

3. A key gating item to level 4 will be the inclusion of redundant systems.

Mentioned that as well, I think.

I 100% agree that the AP2.0 hardware is going to be enough for level 3 (concentrating on sensors so they can accurately see the world around the car and test scenarios). However, I don't think it will have fully redundant systems. To do so, especially recognizing that Level 4 is still many years off, would be a big waste of money for Tesla. Lots of hardware in the car that would never need to be used.

AP 2.0 cars will probably never see full autonomy.

Reminder: I think a little under a year ago (I might be a little off on the time frame), Elon mentioned that a Tesla would be able to go from NY to LA to pick you up. (He did not imply it would be legal).

Does anyone really believe that in about a year a Tesla (even a research version) will be technically capable of driving from NY to LA? When that car comes to a construction zone and encounters a flagger that is controlling traffic to alternate back and forth on one lane, will it handle that?

When it encounters an accident with emergency vehicles crossing all travel lanes and remaining cars drive around on the shoulder, will Tesla's most secret research software and vehicle will handle these situations in just a year? Highly doubt it.

 

[Todd Burch]

It certainly is, and yet Google achieved it years ago on the San Francisco peninsula with a tiny number of vehicles. I have seen Google cars in my area for years. By law they are required to have a licensed driver at the wheel, of course, but Google has made clear that the cars drive themselves and they have had no fatalities and no serious accidents, just a handful of low speed impacts almost all caused by other cars with human drivers.

In contrast, Tesla is now engaged in using tens of thousands of cars in the US and Western Europe (and growing numbers in other regions) that are all sending data back to Tesla. In a few years Tesla will have over a million cars on the road which will allow them to map large areas with high precision. The amount of data Tesla will gather will grow exponentially.

If you watch a TED conference by the chief Google autonomous driving engineer, he talks about a lot of scenarios that still can't be handled....or that it will sit at an intersection for an excessively long time...or about how the cars don't drive very fast...and how he hopes that his less than 10 year old children will never have to drive.

You're hearing from the top engineer of the group furthest along in autonomy talking about another 8+ years. And Google is not known for having slack coders...

I'd also add that while Tesla is WAY ahead of everyone else as far as a data-gathering infrastructure is concerned, data doesn't solve everything. There's still some seriously challenging software to develop. Certainly the most complex software ever written.

 

[Todd Burch]

The argument in the opening message is, because level 4 autonomy won't be here for years, people should stop waiting for autopilot 2.0 hardware and buy the car now. I see too many problems with this argument.

No, that is not my argument. Buy your car whenever you want, I don't care. I already have (and love) my Tesla.

My point was that level 4 isn't coming soon, and Level 3 is still years off.

 

[Todd Burch]

The argument that "I'm not smart enough to think of a solution to this outlier problem." does not mean that the problem is not easily solvable....This is not going to be all that difficult (technologically). ...

Wow. Appreciate the insult to my intelligence first of all, but that must be the understatement of the century. I'm sure it's not going to be all that difficult. Clearly. That's why everyone's driving around in fully autonomous cars.

Do you believe Elon's tweet that a Tesla will be able to drive from NY to LA in about a year?

That would be great. Autopilot's awesome, but I'm still waiting for it to crest a hill without heading toward the car next to me. Once it can do that, then maybe I'll consider that it can cross the country on its own.

 

[Alketi]

I 100% agree that the AP2.0 hardware is going to be enough for level 3 (concentrating on sensors so they can accurately see the world around the car and test scenarios). However, I don't think it will have fully redundant systems. To do so, especially recognizing that Level 4 is still many years off, would be a big waste of money for Tesla. Lots of hardware in the car that would never need to be used.

I'm in total agreement, especially for the Model 3. Introducing redundant systems that would have no immediate use on a car that needs to be kept "simple" enough to manufacture on an extremely accelerated timeframe would seem to be fairly crazy.

Musk alluded to this at the Mossberg conference when he said that they were deferring bells & whistles until v2.0. Thus, I think it follows that the initial production run of Model 3s will not be capable of driverless operation, unless regulators allow such operation without full safety redundancy.

However, I wouldn't call Level 3 years off. It's important to distinguish that hardware will likely be shipping within the next 6 months and almost certainly before July 2017, and that software will progress steadily but rapidly toward 99% self-driving capability (with a driver present).

 

[RogerHScott]

I'm imagining the distant future, to when we have complete saturation of fully autonomous vehicles and these questions about stop signs and stop lights and such will be moot. By complete saturation, I mean both vehicles and requisite infrastructure, because the technology required to achieve this future, by necessity, will not be solely contained within the vehicles. In a world of fully autonomous cars, when it comes to the roadway and most traffic cues, there is no need for visual markers such as lights and signs that were designed for the human eye, because sensors and processors can "see" without such infrastructure and there will be a vital necessity for more precise "seeing" in order to achieve the stated goal of orders of magnitude in improved safety. If we want these vehicles to be safer than human-operated vehicles, then we will need to take full advantage their superior capabilities and not limit them to just what humans can achieve with their inferior senses. A fully autonomous vehicle will be in constant comunication with all the other vehicles around it on the roadways and it will also be in communication with the roadway itself. I imagine that along the roadway and at intersections, etc., all that will be needed is transmitters and receptors that will communicate with the vehicles and define the operating space for the vehicles. All the issues being brought up in this thread are focusing solely on the vehicles and that's understandable because what we're talking about here is about the different stages and progressions leading up to fully autonomous vehicles, and hey, it's a car forum, but many of these issues will be overcome by a paradigm shift in the basic infrastructure of our roadways. I imagine there will be fully autonomous "zones" where full saturation of the infrastructure will be in place and various stages of saturation in other areas and even some "wild west" frontier areas where the car is on its own as they roll out all of the changes. (Cue "Red Barchetta") Oh, and in this future world, there'll be spandex jackets, one for everyone...

Okay, back to talking about the changes to the Tesla and the expected rate of progression, sorry for the Arthure C. Clarke/Isaac Asimov/William Gibson/Rush/Donald Fagen distraction, but my muse got me to musing...again.

Can I loan you my copy of Colossus the Forbin Project? I assume you already have access to The Terminator.

 

[KublaConn]

Can I loan you my copy of Colossus the Forbin Project? I assume you already have access to The Terminator.

Ah ain't skeert uh dem dare consarnit mush-eens!

 

[987S4]

It feels like Tesla's learned quite a bit from AP1. They've learned enough that radar is now the primary vs the camera. That's a huge paradigm change. That switch, coupled with the lack of multi-cameras to fill in the Model X's housing makes me think that new hardware is still awhile away.

They'll need to evaluate OS 8 and see how well it works in wide release. Then maybe new hardware will be extra radar instead of extra cameras, or different placement of sensors.

Whatever the case, my gut feeling is that we won't see Level 3 autonomy until 2018 at the earliest.

 

[Todd Burch]

It feels like Tesla's learned quite a bit from AP1. They've learned enough that radar is now the primary vs the camera. That's a huge paradigm change. That switch, coupled with the lack of multi-cameras to fill in the Model X's housing makes me think that new hardware is still awhile away.

They'll need to evaluate OS 8 and see how well it works in wide release. Then maybe new hardware will be extra radar instead of extra cameras, or different placement of sensors.

Whatever the case, my gut feeling is that we won't see Level 3 autonomy until 2018 at the earliest.

Agree, and that paradigm change underscores why the Tesla/Mobileye partnership broke up.

Many other car companies will be going the route with Mobileye, but it appears Tesla now has the data it needs to determine that the camera as a primary sensor is the wrong approach.

 

[Canuck]

Whatever the case, my gut feeling is that we won't see Level 3 autonomy until 2018 at the earliest.

But to me this isn't even the question. The question to me, as someone who has held off on upgrading waiting for more hardware, is when will AP2.0 hardware be on a Tesla? I can't imagine that the Model 3 will come out with the same cameras, sensors and processing power of the current Model S/X. If it does not and has more hardware, then Elon Musk has broken his promise that the currently available Model S/X will always lead in technology over the Model 3. That means to me that either the Model 3 will have no changes over the current hardware, which I find extremely unlikely, or the Model S/X will have AP2.0 before the end of 2017 when the Model 3 is released.

No one knows when the AP2.0 release date will be, but according to my analysis above, it's my view that we are the very most months, and not years, away from new hardware.

but it appears Tesla now has the data it needs to determine that the camera as a primary sensor is the wrong approach.

Yes, that appears to be what they are telling us, along with:

After careful consideration, we now believe it [radar] can be used as a primary control sensor without requiring the camera to confirm visual image recognition.

Perhaps they have determined the camera is wrong approach, but what is concerning is that they have not determined that radar is the correct approach. They "believe" it is, but it's still a beta test to attempt to verify a belief, or at least in my view that seems to be what they are expressly telling us.

 

[987S4]

when will AP2.0 hardware be on a Tesla? I can't imagine that the Model 3 will come out with the same cameras, sensors and processing power of the current Model S/X. If it does not and has more hardware, then Elon Musk has broken his promise that the currently available Model S/X will always lead in technology over the Model 3.
No one knows when the AP2.0 release date will be, but according to my analysis above, it's my view that we are the very most months, and not years, away from new hardware.

Obviously it's all speculation, but if you believe:

1) Tesla was planning to use (more) cameras as the primary sensor in AP2 (windshield housing in Model X)
2) Tesla decided to investigate if it's possible to switch from cameras to radar only 3-4 months ago (per Elon Musk)
3) Tesla has only recently rolled out AP based on the radar vs cameras (speculation on how long it might be to code based on #2)
4) Tesla will need to time to evaluate success of radar vs camera to decide how to best implement hardware for AP2.

I don't know how it's possible to put new hardware in before 2018. Yes, I understand that this conflicts with Elon's statements re: Model 3 and Model S/X, but things have changed since those comments were made.

We have a fatality that EM believes would have been prevented if radar was the primary sensor AND we now have the NHTSA heavily involved, which will undoubtedly slow things up.

 

[Canuck]

I don't know how it's possible to put new hardware in before 2018. Yes, I understand that this conflicts with Elon's statements re: Model 3 and Model S/X, but things have changed since those comments were made.

New hardware is simply more radar, cameras and faster processing. It need not be activated when initially placed on vehicles (and very likely will not be) but being available for future activation is huge given the relatively low cost of the hardware and tons of Model 3's that will be pumped out from day 1. The change to radar does not at all change the fact that Tesla is very slim on hardware and I don't see Tesla putting the same hardware and processing on the Model 3 as in use today. But I've been wrong before. These are just guesses are my part, just like everyone else.

 

[987S4]

I thought that AP2 hardware was coming very very soon too, but after reading the transcripts of AP in v8 of the software update I've changed my mind completely. I can't conceive how they can figure out what hardware they'll need before Model 3 release, let alone source and install.

Maybe they'll need 2 radars in the front and 1 in the back. Maybe it'll be back to 3 cameras and radar will take the backseat. How much data will they need before they can make that decision?

I'm sure they'll put it in before it's activated, but they need to know what and where they're going to put in first.

 

[Saghost]

I thought that AP2 hardware was coming very very soon too, but after reading the transcripts of AP in v8 of the software update I've changed my mind completely. I can't conceive how they can figure out what hardware they'll need before Model 3 release, let alone source and install.

Maybe they'll need 2 radars in the front and 1 in the back. Maybe it'll be back to 3 cameras and radar will take the backseat. How much data will they need before they can make that decision?

I'm sure they'll put it in before it's activated, but they need to know what and where they're going to put in first.

Yeah, I think they would have rolled out the triple camera based system right about now - if they hadn't decided to go radar heavy. That switch means they may not even know what hardware they want for the next generation yet - and almost certainly means any plans of the mid-cycle triple cam but otherwise the same system are out the window.