Autopilot 3.0 Hardware: When and what?

[calisnow]

Who here thinks that roughly 24 months from now there will be a third major leap in both sensor quality/quantity and computing horsepower?

And - because speculation is our raison d'être here at TMC - what will it be?

My money is on more radar and perhaps even more cameras for redundancy, perhaps coupled with fully redundant computers and various fail-safe devices for the steering and braking systems.

At that point I also speculate we may see a way to fold the steering wheel/yoke away from the driver along with a fold-out tray to use as a mobile desk, and even a driver's seat which can recline into a bed with a footrest.

If Tesla pulled this off 24 months from now it will be one way to throw more egg in the Germans' faces and stay a step ahead of them - we are now two years into lane-keeping that has still been equaled by nobody. 24 months from now Benz will just be rolling out their first long range EV and there are no indications it will be fully self driving, let alone have anything like a driver's desk or sleeping chair.

As long as I'm guessing, I'll speculate we are also 24-36 months away from a longer wheelbase flagship Tesla sedan with roomier rear seats which also recline. Yes, Tesla is on a mission to go mass market but that doesn't mean they aren't simultaneously developing the next generation of their upmarket offerings also.

What do y'all think?

 

[3Victoria]

Waaa ... we haven't stopped speculating on Model 3 yet .... relax, "let the dog see the rabbit", grasshopper.

 

[ecarfan]

@calisnow, my interpretation of the description of Full Self Driving Capability on Tesla's website right now is that it is the equivalent of Level 5 autonomous driving. Since Telsa has been very clear that the currently shipping sensor suite will do FSDC once the software is released then it appears that no further hardware is needed. You can't go over Level 5.

But feel free to speculate...

 

[calisnow]

@ecarfan - I agree with you, however Tesla is a company. A three year old MacBook Pro far exceeds most humans' needs but that doesn't stop Apple from building more powerful machines.

The obvious sales/marketing argument for more sensors and computers would be claiming even further reduction in accident risk.

Perhaps also some kind of mechanical systems for clearing dirt from camera lenses.

 

[ecarfan]

Sure, I agree. Of course there will be sensor changes/improvements in the future. My point was that Tesla appears to be claiming that the current sensor suite will achieve Level 5 autonomy. Which amazes me. We shall see...

 

[3Victoria]

I do think that the suite will be augmented in the future, for redundancy and added features.

 

[Saghost]

My guess is that shortly after they get stable, reliable full self driving, or possibly right after they get legislative approval for it, you'll see a third hardware generation.

That's the first time they'll really know what they need to do the job right, and Tesla tends to change things until they have it really right.

And, of course, the hardware available will continue to improve, so a replacement of obsolete pieces will make sense at some point anyway.

I'm hoping to see rear looking or 360 radar, and possibly thermal imaging forward, along with a deletion of the side mirrors and of course faster processors and higher resolution cameras.

 

[bswn1]

I think there's a non zero chance an accident occurs or regulatory approval doesn't happen because the current sensor suite is inadequate in some unforeseen way that becomes obvious after the fact. There will need to be another upgrade to the suite to make it fully level 5.

 

[Gen3]

There will definitely be updates to hardware.

1. Because Elon said a while ago that we should have 10x improvement on humans to accept full autonomy, and he recently said this suite should give 2x improvement.

2. Redundancy to allow minor faults to be accommodated, as well as reduce accident potential with extra "eyes" looking out.

3. Bad weather. Precipitation getting in the way of current cameras, and just 1 forward radar. Until you have full autonomy in in a blizzard, hardware will evolve, and even then it will still improve.

4. Smart traffic. I hope this comes soon, but could be delayed for regulation and standards that all manufacturers can agree on. The car should talk talk to cars around itself and be notifed of braking or maneuvering events. It should also talk to grid to be aware of and avoid potential conflicts at intersections, as well as receive timing instructions to slow down/speed up to flow traffic better than the current dumb traffic lights allow.

5. Convoys. I'm surprised there is little talk about "car trains" or convoys given everyone's understanding of reducing drag and therefore energy. Cars, or tesla semi trucks could be a few inches to a few feet behind each other and have huge energy savings. Not to mention the safety of having the lead car or "pilot" leading a massive train of cars/trucks in full autonomy with the passengers sleeping, or having just cargo and no driver.

I guarantee tesla is working on this convoy, and I hope they shock the world soon when they cut freight costs by 10x, and emissions by the same or better.

I cant call who will win the World Series in 23 years, but this will be normal in 23 years.

 

[kurdakov]

nvidia plans to launch update of their autonomous driving hardware by the end of 2017 Introducing NVIDIA Xavier | NVIDIA Blog .

As Tesla likely will follow cooperation with nvidia, that means either by the end of 2017 or beginning of 2018 there will be next iteration.

How it will be called (2.1 or 3.0 ) is another question

 

[calisnow]

This is a picture of a Volvo concept - my guess is that this is 24-36 months out - max - for the Model S/X. Couple a sleeper seat with 400-500 mile battery and automatic supercharger connections (also perhaps only 24 months away given that the 100D will be 320+) and I think we are looking at airline disruption for short haul flights like LA/OC/SD to SF/Sacramento/Phoenix. Hop in your Tesla at 10 pm, pop an ambien - wake up in your destination city.

Me personally - I would even take the Tesla over a plane for the 700 miles, 11 hour LA -> SLC drive to ski. Sleep 8 hours on the road, wake up, have breakfast in the car, change into ski clothes and be floating through fresh powder at Alta by 9 am. Woot!

This kind of efficient luxury transport would be the freedom and time-saving machine I would gladly pay $150K+ for.

 

[LargeHamCollider]

Who here thinks that roughly 24 months from now there will be a third major leap in both sensor quality/quantity and computing horsepower?

And - because speculation is our raison d'être here at TMC - what will it be?

My money is on more radar and perhaps even more cameras for redundancy, perhaps coupled with fully redundant computers and various fail-safe devices for the steering and braking systems.

At that point I also speculate we may see a way to fold the steering wheel/yoke away from the driver along with a fold-out tray to use as a mobile desk, and even a driver's seat which can recline into a bed with a footrest.

If Tesla pulled this off 24 months from now it will be one way to throw more egg in the Germans' faces and stay a step ahead of them - we are now two years into lane-keeping that has still been equaled by nobody. 24 months from now Benz will just be rolling out their first long range EV and there are no indications it will be fully self driving, let alone have anything like a driver's desk or sleeping chair.

As long as I'm guessing, I'll speculate we are also 24-36 months away from a longer wheelbase flagship Tesla sedan with roomier rear seats which also recline. Yes, Tesla is on a mission to go mass market but that doesn't mean they aren't simultaneously developing the next generation of their upmarket offerings also.

What do y'all think?

Sounds about right.

 

[brkaus]

I don't get the obsession with radar? I understand it can see through fog, rain, etc.

But it isn't accurate enough to place the vehicle on the road with enough confidence to stay in a lane. It can't see lane lines. GPS isn't accurate enough either.

 

[Saghost]

I don't get the obsession with radar? I understand it can see through fog, rain, etc.

But it isn't accurate enough to place the vehicle on the road with enough confidence to stay in a lane. It can't see lane lines. GPS isn't accurate enough either.

Where do these pieces of information come from?

Some modern GPS systems are accurate within a couple centimeters - certainly plenty to steer along a lane if the car has a high precision map of where that lane is.

I don't have enough information about the capability of the radar unit Tesla is using to comment, but I suspect it has ample precision as well - provided the car has a radar map of the correct path.

That radar map is what I expect the whitelist we're developing right now to evolve into - which added to high precision map downloads will give the car triple redundancy in reading where it is on known routes in the future, and some capability to drive when snow blocks the lane lines.

 

[LargeHamCollider]

I don't get the obsession with radar? I understand it can see through fog, rain, etc.

But it isn't accurate enough to place the vehicle on the road with enough confidence to stay in a lane. It can't see lane lines. GPS isn't accurate enough either.

GPS with SBAS or LBAS and inertial smoothing is. I fly an aircraft fairly regularly that uses SBAS and inertial smoothing, I've never seen the system more than 18 inches off (except right after it starts up and all the satellites have not yet been acquired), LBAS is even more accurate. Tesla will definitely use SBAS with inertial smoothing, I wouldn't be surprised if they have some sort of LBAS built in as well.

SBAS: Satellite Based Augmentation System
LBAS: Local (no, not "Locally") Based Augmentation System.

 

[brkaus]

Best info I can find in gps. If you have better, please share.

---
The actual accuracy users attain depends on factors outside the government's control, including atmospheric effects, sky blockage, and receiver quality. Real-world data from the FAA show that their high-quality GPS SPS receivers provide better than 3.5 meter horizontal accuracy.

Higher accuracy is attainable by using GPS in combination with augmentation systems. These enable real-time positioning to within a few centimeters, and post-mission measurements at the millimeter level.

---

Augmentation systems to the best of my knowledge are location specific and requires extra hardware.

 

[brkaus]

GPS with SBAS or LBAS and inertial smoothing is. I fly an aircraft fairly regularly that uses SBAS and inertial smoothing, I've never seen the system more than 18 inches off (except right after it starts up and all the satellites have not yet been acquired), LBAS is even more accurate. Tesla will definitely use SBAS with inertial smoothing, I wouldn't be surprised if they have some sort of LBAS built in as well.

SBAS: Satellite Based Augmentation System
LBAS: Local (no, not "Locally") Based Augmentation System.

Thank you. 18 inches is pretty good! Much better than I had read. I still don't feel like it's good enough to fit in a 12 ft lane.


What type of gps does Tesla currently have in the cars?

 

[Saghost]

Best info I can find in gps. If you have better, please share.

---
The actual accuracy users attain depends on factors outside the government's control, including atmospheric effects, sky blockage, and receiver quality. Real-world data from the FAA show that their high-quality GPS SPS receivers provide better than 3.5 meter horizontal accuracy.

Higher accuracy is attainable by using GPS in combination with augmentation systems. These enable real-time positioning to within a few centimeters, and post-mission measurements at the millimeter level.

---

Augmentation systems to the best of my knowledge are location specific and requires extra hardware.

Differential GPS - Wikipedia

There's apparently already a nationwide augmentation system run by the government in the U.S.

This isn't the one centimeter I remember reading about, but ten centimeters is probably good enough if they can achieve it regularly.

 

[kurdakov]

I don't get the obsession with radar? I understand it can see through fog, rain, etc.

But it isn't accurate enough to place the vehicle on the road with enough confidence to stay in a lane. It can't see lane lines. GPS isn't accurate enough either.

accuracy of radar can be ~4 centimeters MIT Lincoln Laboratory: News: Lincoln Laboratory demonstrates highly accurate vehicle localization under adverse weather conditions then radar can be used more efficiently, than now link: Toyota Bets Big On Autonomous Tech, Swallows Millimeter Wave Radar Maker

for gps there are also some possible improvements
Green Car Congress: UC Riverside team developing nav system that uses signals of opportunity; support for autonomous vehicles ( also for dead reckoning see : Green Car Congress: New Telit autonomous navigation IoT module relies on internal sensors to deliver class-leading dead reckoning accuracy )

there are other approaches though like
Green Car Congress: Oryx Vision raises $17M to create novel depth-sensing solution for autonomous vehicles; LiDAR replacement

so not only lidar is a way to improve navigation

 

[number12]

Wonder if Tesla will ever release a 3 month lease option?

Every quarter they get a new sale, and I won't feel constantly outdated. Win-win in my book.