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

Tesla crowd-sourced maps - how to overcome "small fleet" problem vs GM/VW?

This site may earn commission on affiliate links.
Anybody thought of this problem? Surely Tesla has thought of it and has some kind of plan. When crowdsourced HD maps are built GM/VW (the Mobileye mapping project partners I believe) will have a huge advantage if they choose to start building cameras into all their cars. Out in rural America every podunk dirt road gets driven frequently by somebody in a GM product. This is obviously not the case for Tesla. I am the only AP2 Tesla I know of for dozens of miles where I am right now in Wisconsin. I drive a pretty predictable route - and so there are hundreds if not thousands of miles of suburban and rural roads around me which may not be driven by an AP2 Tesla for years to come. What's the solution? Maybe the "problem" is not nearly as important as I imagine it to be. Perhaps even a single Tesla can construct a visual map - so that an owner in Nowheresville will gradually build out their own maps and after a few weeks of commuting a route Tesla will be able to "turn on" HD maps for that owner on those roads. Everyone will blaze their own mapping trail, so to speak. Possible?
 
  • Like
Reactions: kurdakov
Same things people do: slow down and map/learn as you go. Slowing down improves the confidence of image and feature recognition. At a low enough speed, an L5 system should be able to operate on its own, without any mapping data. That could even make sense as a legal requirement, perhaps with a minimum speed of 25 mph.

Sure, a vehicle doing this might inconvenience other vehicles. People who are new in town do that too. But this vehicle will upload the data it gathers. Next time it can drive a bit faster, and so can any other vehicle with access to the mapping data.
 
Seems like one of those things that shouldn't be patentable. Add a bunch of data to a database from a bunch of sources. That's a unique idea? Wha? Here we go again. No one else can now make their own database of mapping info from more than one (their own?) source?
 
  • Love
Reactions: croman
Well whatever Mobileye claims in that 215 (!) page document, the US Patent Office has granted it, so that means it's untouchable.

The diagrams on p. 36-39 (FIG 18-29), and p. 45-46 (FIG 27-28) sums up the general idea, I think.

Big patent
 
Last edited:
Anybody thought of this problem? Surely Tesla has thought of it and has some kind of plan. When crowdsourced HD maps are built GM/VW (the Mobileye mapping project partners I believe) will have a huge advantage if they choose to start building cameras into all their cars. Out in rural America every podunk dirt road gets driven frequently by somebody in a GM product. This is obviously not the case for Tesla. I am the only AP2 Tesla I know of for dozens of miles where I am right now in Wisconsin. I drive a pretty predictable route - and so there are hundreds if not thousands of miles of suburban and rural roads around me which may not be driven by an AP2 Tesla for years to come. What's the solution? Maybe the "problem" is not nearly as important as I imagine it to be. Perhaps even a single Tesla can construct a visual map - so that an owner in Nowheresville will gradually build out their own maps and after a few weeks of commuting a route Tesla will be able to "turn on" HD maps for that owner on those roads. Everyone will blaze their own mapping trail, so to speak. Possible?

Do not under estimate model 3s impact on rural areas, the number of AP2 cars that will be hitting the road in the next 18mo will be a force multiplier and will gather more data then even Intel, mobileye or Tesla can make use of. Also, don't forget the semi, which tend to drive upwards of 120,000 miles a year. How many Walmarts where you live? If someone like Walmart goes with Tesla, that will certainly help. And yes, your car should be able to gather enough raw data to create maps for you and your car can tell you when to take over well in advance of running out of mapping area. Missing maps could also be licensed from companies like lvl5 that crowd sources using Uber and Lyft drivers with cell phone mounted dash cams.

As far as parents go, I havn't read any patent so I'm not sure what processes mobileye are patenting, but Tesla and lvl5 are only crowd sourcing the gathering of raw GPS, radar and video data. The maps are not made by the cars processing power or software. That will happen at Tesla when all the data is combined to create both 3d maps and teach the neural networks.

Lastly, I believe 3d maps are required for level 4+. It's as much about redundancy as it is about the AI being able to drive like us, using only vision. The 3d maps will offer advantages in bad weather where road markings are obscured by snow and as a back up for stop lights, signs and turn Lanes and such. It's one thing to have the car see a red light and know what it is, but if it doesn't for some reason, like vision is obscured by a semi, then the 3d maps has that info as well. Level 4+ requires redundancy so 3d maps are required. Another potential requirement might be connected cars and connection to the environment it self. Imagine an intersection with cameras that broadcasts what's going on in the intersection and in all directions approaching the intersection. Then cars relaying that info down the road to you as you approach. A lot of situations get easier if vehicles can and by law must communicate, even it's is just telemetry data and intent based on gps nav.
 
Well whatever Mobileye claims in that 215 (!) page document, the US Patent Office has granted it, so that means it's untouchable.

The diagrams on p. 36-39 (FIG 18-29), and p. 45-46 (FIG 27-28) sums up the general idea, I think.

Big patent

IANAL, especially not a patent lawyer. And I don't want to downplay this patent: it looks important. But patents aren't exactly untouchable. They can be overturned or restricted, for example due to prior art. I wouldn't be surprised to see Waymo contest this one. That could get interesting, with Intel's legal department backing Mobileye and Alphabet behind Waymo. They might decide to cross-license, which would be bad for smaller players like Tesla.

In case anyone wants to dig into the details, here's a google link alternative to the pdf, along with the text abstract

US20170010106A1 - Crowd sourcing data for autonomous vehicle navigation - Google Patents

Abstract
A method of processing vehicle navigation information for use in autonomous vehicle navigation is provided. The method includes receiving, by a server, navigation information from a plurality of vehicles. The navigation information from the plurality of vehicles is associated with a common road segment. The method also includes storing, by the server, the navigation information associated with the common road segment. The method also includes generating, by the server, at least a portion of an autonomous vehicle road navigation model for the common road segment based on the navigation information from the plurality of vehicles. The method further includes distributing, by the server, the autonomous vehicle road navigation model to one or more autonomous vehicles for use in autonomously navigating the one or more autonomous vehicles along the common road segment.
I notice claims 1-28 are "canceled". I wonder what they were?
 
  • Informative
Reactions: lunitiks
Well whatever Mobileye claims in that 215 (!) page document, the US Patent Office has granted it, so that means it's untouchable.

The diagrams on p. 36-39 (FIG 18-29), and p. 45-46 (FIG 27-28) sums up the general idea, I think.

Big patent
I got beat to it, but you can't patent a general idea, only a specific method of doing something. Patents can also be invalidated by prior art. Unless I am reading it wrong, the patent was filed September 22, 2016, so not that far back (but it says it's a continuation of some other patents, which I'm not sure what that means). @croman can probably speak more on this.
 
Here's a link to the original claims: Register Plus PDF viewer

Apparently the original claims were 1-28, and they were replaced in full by the final claims 29-48. I'd say they're generally similar. I'm not sure why these claims had to be rewritten, but that's what keeps patent lawyers in business.

If any patent lawyers are reading this thread.... Could I avoid infringing on this patent if my crowd-sourced mapping technology avoids the use of 3D splines? Or if it uses just a bit more than 10-kB/km?
 
  • Helpful
Reactions: lunitiks
Yeah, Mobileye's IP/patent portfolio is a major part of its valuation.

You won't be able to build an autonomous vehicle system without paying licensing fees all over the place. Tesla must know this -- it's almost inevitable that they'll end up paying Mobileye licensing fees to use something, somehow - Just as Apple and Google pay Microsoft for every phone they sell.

Maps are a vital part of self driving tech, so you can't really avoid someone patenting everything involved around the creation and use of them for driving systems.

It doesn't mean Tesla won't pay them if they need to - it depends how hardball they are I guess!
 
It’s easy. You create a point system/game/bounty for roads that need to be driven.

Never driven - 10 points per mile
Rarely driven - 5 points per mile
Frequently driven - 0 points per mile

Put it on the center screen with a colorful map that either shows what has been driven or only show what has been drive and let people figure it out.

Add as much granularity to point system as desired.

Exchange points for Tesla store items or supercharger credits or unlock special car powers.
 
Hmm... These inventors must've really hit the "specific" nail on the head then :confused::confused::confused:
Sure, there are one or two that will slip through the system (or the person approved thought it would be funny and just let it through), in all seriousness those can be easily thrown out in court if challenged (plenty of prior art to cite).

It was already pointed out, but the claims have a lot of very specific points that seems like they can be worked around.
 
It’s easy. You create a point system/game/bounty for roads that need to be driven.

Never driven - 10 points per mile
Rarely driven - 5 points per mile
Frequently driven - 0 points per mile

Put it on the center screen with a colorful map that either shows what has been driven or only show what has been drive and let people figure it out.

Add as much granularity to point system as desired.

Exchange points for Tesla store items or supercharger credits or unlock special car powers.

I think Waze holds a patent on that ;)
 
I got beat to it, but you can't patent a general idea, only a specific method of doing something. Patents can also be invalidated by prior art. Unless I am reading it wrong, the patent was filed September 22, 2016, so not that far back (but it says it's a continuation of some other patents, which I'm not sure what that means). @croman can probably speak more on this.

Continuation means that there is a parent application which was going to be either issued or abandoned and the applicant wasn't able to get the claims they wanted in it so they broke off those contested claims and continued arguments with the examiner in a child application(s). I haven't look at the file wrapper history of this patent but that's generally what is going on with a continuation.

Sometimes this is done to increase the number of patents or to get a parent registered and still maintain the child for optics purposes (even if those claims are likely doomed if faced with a final rejection).
 
I notice claims 1-28 are "canceled". I wonder what they were?

They all relate to autonomous navigation of a vehicle.

upload_2017-10-3_10-25-9.png



upload_2017-10-3_10-25-46.png


upload_2017-10-3_10-26-13.png


upload_2017-10-3_10-26-54.png



upload_2017-10-3_10-27-14.png