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

Tesla Motors and street-mapping services

This site may earn commission on affiliate links.

replicant

Active Member
Aug 24, 2014
1,380
6,767
France
Uber has recently acquired Microsoft's Bing mapping unit. Nokia will probably sell its HERE business to Audi, BMW and Daimler. Apple keeps expanding its Maps data and software. Google also invests massively in this area, notably via their acquisition of Waze. All these companies are said to work on their own self-driving cars.

What are Tesla plans to improve their street-mapping and route-planning services?
How could they not rely on their competitors?
Will they be technically and legally able to mass-collect high-quality data from Autopilot-enabled cars?
 
These are good questions. Specifically for Waze, and for other evolving crowd-sourced updates/status reports, we do need faster connections. Tesla is now delivering LTE in new cars, but older ones survive so far with 3G. I suspect we'll not see major mapping improvements until faster cellular data and faster graphics cards have been installed. AFAIK, there has been no official announcement. Still, I fully expect to see a substantial updated package sometime around Model X shipping dates, with retrofit pricing and terms for other cars, similar to the battery pack and Ludicrous launch. When?, how much cost?, what capabilities?

Pure speculation: Google and Tesla CEO's are close;
1) we should expect to see Waze integration, just as Google/Waze integration is on the near horizon. They might even happen at the same time;
2) The probable problem with (1) is how to accommodate crowd-sourcing, essential for updating. Google should have a good solution in mind using the voice recognition of tesla, but upgraded in effectiveness and function;
3) Comprehensive routing improvements require far better multi-stop and Boolean functions. Oddly, Uber purchase of deCarta helps them with the same issues plus presumably better routing predictive algorithms. And, Uber has conflicts brewing with Google, Bing maps being an issue too;
4) So, since these issues are far more aligned with strategic views and expertise of Google than they are for Tesla, I expect doubling down on Google for solving the routing optimization problems of Tesla. Tesla can then focus on the drivetrain including batteries, really the major play for Tesla.

This logic can be extended, of course, to driving automation too...
 
More pure speculation: MS cars come with connectivity and basically all MS on the road since release, have been dropping bread crumbs. Not too difficult to imagine taking this geo data and basically have your own map system. There is no confusion in the data, since the car is basically on the road, unlike a cell phone. Overlay this data with existing map services... Makes the billion miles driven of much greater importance...
 
More pure speculation: MS cars come with connectivity and basically all MS on the road since release, have been dropping bread crumbs. Not too difficult to imagine taking this geo data and basically have your own map system. There is no confusion in the data, since the car is basically on the road, unlike a cell phone. Overlay this data with existing map services... Makes the billion miles driven of much greater importance...

This video from 2013 shows how much work is required to get high-quality maps:

This more recent presentation shows how Google gather detailed 3D-maps with LIDAR systems.

I fear these bread crumbs aren't worth much...
 
Last edited by a moderator:
There are two fundamentally different paradigms in self-driving cars and their need for good maps:

* Google relying on a centralized data/computation approach: it builds an incredibly detailed, 3-D representation of the road network, supplemented by real-time feedback, and then pushes instructions to its vehicles based on massive computing capacity; versus
* Tesla relying on in-car systems, looking at the conditions on the road and responding.

Google's approach is simply not scaleable, and it relies on continuous data streams between the car and the Googleplex. Huge swaths of my home state, Maine, don't have cellular data coverage, and I serious doubt that Google is going to get around to 3-D mapping all the little rural routes across the country. So, I think Tesla's distributed model is intrinsically more robust.
 
  • Like
Reactions: totoro722
I don't think that's how they work. They both need a) a map (it'd better be able to cache it, for going through tunnels or backwoods Maine), so the car knows what route to take, b) eyes/radar on the immediate surroundings.

You can't do only (b) and expect to navigate anywhere. You can't do only (a) and expect the car to get through traffic, cyclists, pedestrians, obstacles, road works, etc.

If you want an autonomous car, the map is 100% essential. Is Tesla OK with using someone else's? Are there competitive suppliers should the one turn bad?

For such a critical piece, should Tesla in fact be in full control by owning it? (Like BMW/Audi/Daimler are doing - Audi, BMW and Daimler Near Deal to Buy Nokia Mapping Service - WSJ)
 
Can Tesla enable autopilot cameras to build maps data with user permission? It will take a few years, but that's an easy way to build and update maps.

This is my thought on the subject. Record data, upload via wifi, only done with user's permission.

They could tie it into using autopilot in general, i.e., if you use autopilot, then the car will record and send data back to Tesla for features much more advanced than auto lane keeping.
 
There are two fundamentally different paradigms in self-driving cars and their need for good maps:

* Google relying on a centralized data/computation approach: it builds an incredibly detailed, 3-D representation of the road network, supplemented by real-time feedback, and then pushes instructions to its vehicles based on massive computing capacity; versus
* Tesla relying on in-car systems, looking at the conditions on the road and responding.

Google's approach is simply not scaleable, and it relies on continuous data streams between the car and the Googleplex. Huge swaths of my home state, Maine, don't have cellular data coverage, and I serious doubt that Google is going to get around to 3-D mapping all the little rural routes across the country. So, I think Tesla's distributed model is intrinsically more robust.

There is also a danger in the centralized push approach. Hackers could cause havoc by just jamming the transmission frequency. Could you imagine New York City or Los Angeles with thousands of Google cars that suddenly don't know what to do? It wouldn't even need to be a sophisticated hack either, all it needs to do is overwhelm the actual signals with noise.
 
There is also a danger in the centralized push approach. Hackers could cause havoc by just jamming the transmission frequency. Could you imagine New York City or Los Angeles with thousands of Google cars that suddenly don't know what to do? It wouldn't even need to be a sophisticated hack either, all it needs to do is overwhelm the actual signals with noise.

Google's self-driving car certainly have a better autopilot system than Tesla (which works without detailed street maps), but Google make their cars more reliable by using 3D maps. Their system is about autonomous driving augmented with high quality data.
 
  • Disagree
Reactions: bkp_duke
* Google relying on a centralized data/computation approach: it builds an incredibly detailed, 3-D representation of the road network, supplemented by real-time feedback, and then pushes instructions to its vehicles based on massive computing capacity;

...

Google's approach is simply not scaleable, and it relies on continuous data streams between the car and the Googleplex. Huge swaths of my home state, Maine, don't have cellular data coverage, and I serious doubt that Google is going to get around to 3-D mapping all the little rural routes across the country. So, I think Tesla's distributed model is intrinsically more robust.

This is blatant lie. Or at least misinformation based on not understanding technology.

Google cars are designed to be perfectly autonomous and able to drive when connection(3g/4g etc) is not present. Moreover, all computations are done onboard, nothing in the cloud. When Google car see roadside construction etc it detects it using onboard chips and electronic, NOTHING relies on the cloud.

If you know anything about Google, you are know that Google research arm is a leader in convolutional neural net technology. And convnets are currently hold state of the art on major computer vision benchmarks. And there is a lot of research done by Google in an area of model distillation, essentially technology that let one squeeze high performance mathematical models into something that could be run by weaker hardware.

I hope Tesla collaborate with Google already, Tesla could learn a lot. As for the Google Maps, they are good and they help, but mostly used with so called SLAM. SLAM algorithms essentially help car to know where exactly car located. If there are bugs in maps(and there are some for sure!), for example due to new construction project etc and map was not updated, car will still know roughly where it is and will immediately recover it precise position on the map the moment it will get into properly mapped territory. BUT IT IS IMPORTANT TO UNDERSTAND that precise position on the map do not contribute to real time decisions of the self driving car. Well, map info helps with decision as to which exit to take or what turn to do, but how exactly lets say left turn is executed is up to internal real time feed from sensors and electronics. So maps do help Google cars with meta decision, but all actual driving is done by in car electronics and sensor suite. And there is no way around it, self driving cars have to be resilient to sudden construction road blocks or road closures. And this philosophy is exactly what helped Google to become a leader in autonomous driving technology. Plus tons of research in the field of convolutional neural nets of cause.
 
Can Tesla enable autopilot cameras to build maps data with user permission? It will take a few years, but that's an easy way to build and update maps.

I sincerely wish they would notice that the street I live on actually connects to another street. Every day my car is horrified that I apparently off-road for a kilometer before it recognizes I am on a road. It also refused to route me home, or it does but it routes me through a closed road. Really, Ms. car, I drive this road every day. Please assume it is paved, kthnksbai.
 
Tesla should buy TOMTOM

Tomtom current market cap is 2.51B now. Assuming a 60% premium, it would cost 4B to acquire. Quite out of reach for TM I guess.

TomTom Weighs Options Amid Nokia Maps Deal - Bloomberg Business:
"TomTom is attracting companies and investors who were looking at Nokia Oyj’s maps business HERE, which it agreed to sell to German carmakers on Monday, the people said, asking not to be identified because the discussions are private. While the $2.5 billion Dutch firm is speaking to possible advisers, there is no formal process underway and TomTom may decide against pursuing a sale and instead focus on investments, they said."
 
  • Disagree
Reactions: bkp_duke