7 ms·

I'm glad that the article highlighted how valuable the mapping information is that Google collects. While Uber probably could still run its current operations w

I'm glad that the article highlighted how valuable the mapping information is that Google collects. While Uber probably could still run its current operations without Google map data, there is a much smaller probability that they would be able to run a self-driving car service without a highly accurate mapping product. And to build a mapping product is really difficult and expensive. I was surprised when Google starting putting money into mapping the streets because it's a giant effort with major barriers to entry. Nokia bought a company called Navteq that was the clear leader (and basically a monopolist) in driving the streets and collecting the data (they originally supplied data to Google Maps). Navteq supplemented the data with satellite images and other sources, but the roads change often enough that you need people driving on the streets. So, Uber really needs to start thinking about the mapping aspect. Perhaps they can incent their current drivers to start reporting back all the street info as they pick up their passengers. Of course, this would bring up problems if the drivers eventually realize that this data could be used to replace them. It would be interesting to see uber cars with Google StreetView style cameras on their hoods.

pyrocat 12y ago
Drivers have basically zero power here, so I'm guessing Uber would just update the driver side of the app to always report location. No need to add incentives when you can just make it mandatory with minimum pushback (from Uber's PoV). Assuming they don't already.
- tlrobinson 12y ago
  I assumed the parent comment meant user-provided data, not raw GPS data, since presumably the driver app is always reporting their location (how else would they know which drivers to dispatch?)
  - pyrocat 12y ago
    I'm not sure what user-provided data would even be beneficial? Stuff that's likely already collected encompasses where the user is picked up, dropped off, route taken, time of departure and arrival .etc
    - vidarh 12y ago
      Reasons for why drivers always take route X between streets A and B, for example (e.g. street C has been changed to one way only so route Y isn't actually legal/possible).
jfoutz 12y ago
I think the self driving car is a tougher problem to solve than the maps. If uber can buy self driving cars, they can make maps on a city by city basis. If you have a pile of money, you can get maps. If you have a pile of money, it's not really clear you can get an autonomous car. I'm the first to shout they're coming - but they aren't here yet.
- ericcumbee 12y ago
  And you have to get enough of those self driving cars for it to be effective. If uber is buying the cars then they cease to be a driver/car to passenger matching service and become a full on cab company, complete with the liability and maintenance costs that go with owning cars.
  - olefoo 12y ago
    And Uber may still wind up on the losing end of a lawsuit about pushing drivers to take on vehicles based on false pretenses ( Uber telling the drivers that there would be enough work to justify them financing a vehicle for Uber usage). Uber's efforts to shed itself of all liability and risk in terms of employees and assets is legendary; but they could still find themselves going up against the IRS on the question of whether drivers are employees or freelancers. Most especially with drivers that they referred to auto-financing. The theory being that having a requirement for a minimum standard of car and providing finance referrals would be similar to requiring "independent contractors" to wear uniforms.
    - Alupis 12y ago
      > Uber telling the drivers that there would be enough work to justify them financing a vehicle for Uber usage It seems incredibly unwise to finance a $20,000 USD vehicle hedging on some third party (of which you have zero control over) to provide enough for you to pay it through.
      - olefoo 12y ago
        Generally, if you tell someone that taking on a debt obligation is a condition of doing business with you and make representations to the effect that they will be able to earn enough to meet that obligation based on their projected earnings from your contractual agreement with them that would be considered an implied contract. If you then manipulate things such that they are effectively unable to meet their obligations because you are jerking them around they are certainly going to be able to find a lawyer who will take their case. Particularly if you were referring them to the auto-dealership and pre-qualifying them for financing.
TenJack 12y ago
Uber is already pretty much limited to major metropolitan areas, so maybe Uber would really only need map data in these consolidated areas. At least to start.
- vidarh 12y ago
  That's a major benefit to rolling this out as a cab company rather than as a car manufacturer - you can enforce limits like that and people won't complain because they won't even know about the 98% of roads you haven't automated yet. Assuming you can get regulators to accept it: Map area by area - send out human driven cars if a customer requests a journey you can't cover with driverless cars yet.. Assuming they collect GPS traces, they know exactly what areas will cover the most of their business with the least amount of mapping.
- shirro 12y ago
  They know the start and end points of the trip and can make a decision whether it can be handled with an autonomous car or a human driver at the outset. They could serve major routes like airports to hotels via autonomous vehicle to begin with and handle everything else with people. That could still represent a massive cost saving.
maxerickson 12y ago
For the US, the current TIGER data is pretty reasonable. They could get away with using that (especially as a starting point), or OpenStreeetMap. Waze demonstrated how well a crowd sourcing effort can work. It's also apparently possible for a small team to start doing sign recognition: http://blog.mapillary.com/update/2015/01/27/traffic-signs.html http://blog.mapillary.com/update/2015/01/27/traffic-signs.ht...
- vidarh 12y ago
  The TIGER data is a starting point. But it's outdated the day it is complete, and rapidly diverges. For many purposes that's perfectly fine, but as stated above, you need someone driving the street. Even something as major as postcodes changes multiple times in between TIGER updates, and tons of streets are changed or added or removed. The TIGER dataset is fantastic, and certainly helps levelling the playing field by giving people somewhere to start from, but it's the cheap 80%, except in this case the last 20% doesn't take the other 80% of the time, but more like 800%.
  - maxerickson 12y ago
    Have you looked at the last couple years of updates? It's way more accurate and complete than it used to be, and for getting names on that last 1%, they have drivers and the ability for the app to say "I don't know that street name, is it new?".
    - vidarh 12y ago
      I do some work for a company that does mapping, so yes, I've looked at it regularly. It may be way more accurate and complete, but it is still updated too rarely to ever be up to date. Cover any reasonable area, and road situations change daily, especially when you're doing things like driving which also needs to account for temporary road closures and the like. As I say, it's a good starting point, but it's not in any way sufficient, and keeping it up to date is a huge amount of work, which has been reflected in the value of companies that have set themselves up to be able to keep up to date data sets of this. That said, as I've commented elsewhere I do agree that Uber is in a good position, but mostly because they don't need full coverage. They need thorough and up to date information, but they need it in very tiny areas to begin with, and as you say they can rely on drivers to cover areas they don't currently know how to deal with. They can apply GPS traces to analyse exactly what areas are most profitable to improve map information for to make self driving viable first, and roll it out area by area.
threeseed 12y ago
Uber can simply license their maps from various third parties the same way every car company does today for their GPS. Google has added some decent additions to Google Maps but really 95% of what they had they purchased. For example Whereis/Yellow Pages data here in Australia.
- kumarm 12y ago
  There are too many sources and too many processes in world wide Maps data (Apple learnt hard way). I doubt Uber has the capability to have decent maps solution in short term (1-2 years).
- jedberg 12y ago
  Licensed maps are great for GPS units telling a human where to go, but a self driving car needs much more detailed information that most likely only google has.
  - threeseed 12y ago
    If Google is the only company with this information how come all the car manufacturers have active projects that don't rely on their data. It's because Google has nothing that other companies don't have/need. It's not like Google magically acquired all their data. They bought most of it.
- runeks 12y ago
  Map data from third parties requires manual verification, as we learned from the Apple Maps debacle. Google does this [1]: > Google employs a small army of human operators (they won’t say exactly how many) to manually check and correct the maps using an in-house program called Atlas. A Maps app isn't just another app. Nobody gets killed if the Stocks app shows an incorrect price, but someone could die if the Maps app uses incorrect mapping information. [1] http://www.wired.com/2014/12/google-maps-ground-truth/ http://www.wired.com/2014/12/google-maps-ground-truth/
toomuchtodo 12y ago
> So, Uber really needs to start thinking about the mapping aspect. Perhaps they can incent their current drivers to start reporting back all the street info as they pick up their passengers. Of course, this would bring up problems if the drivers eventually realize that this data could be used to replace them. It would be interesting to see uber cars with Google StreetView style cameras on their hoods. Most likely they would either start having drivers drive Uber cars with steetview-esq data collection gear on it, or pay a premium to drivers who mounted a tray on the roof of their vehicle that contained high resolution cameras, a laser scanner, and a controller to process and forward that data back. I would not be surprised if Uber was looking at acquiring https://www.mapillary.com https://www.mapillary.com for this functionality.
- trhway 12y ago
  Self-driving cars will solve the issue of collecting street data. Though of course one need some data for the cars to start with.
jlarocco 12y ago
One possibility, is that if they can get good enough inital mapping data, and good enough self-driving cars, the cars could update the mapping data while they're working. That's a lot of trust on the self-driving cars, though.
- JTon 12y ago
  I think a reporting system could solve that problem. Classify potential changes, and force human eyes before approving all but the most minor updates
downandout 12y ago
Self driving cars will be available to purchase from multiple manufacturers, and those manufacturers will have to deal with the mapping issues. Accordingly, while Google will be a formidable and perhaps dominant competitor in this space, I don't see mapping as the deciding factor here. The fact that Android, with its massive install base, will likely come with Google's service built in seems much more likely to help them destroy competitors.
Vik1ng 12y ago
Give it a few years and OpenStreetMap will be superior. Here in Germany it will probably beat google maps most of the time. There is of course still the routing part.
avodonosov 12y ago
Do you think OpenStreetMaps can be a reasonable substitute? IMHO their maps are quite accurate.
- seren 12y ago
  The issue with OSM is that the information level is uneven. Dense urban center, you have every bench and bin mapped. Go into the countryside, there is not even the buildings. It might not be an issue for a taxi company that is mainly operating in urban center though.
  - andygates 12y ago
    It's in the taxi company's interests to improve the maps for everyone: popular routes will be refined to a T, and gaps will quickly become obvious. These days the most common mapping gap is new developments, and they're faster to appear on OSM than the commercials.

pyrocat 12y ago

Drivers have basically zero power here, so I'm guessing Uber would just update the driver side of the app to always report location. No need to add incentives when you can just make it mandatory with minimum pushback (from Uber's PoV). Assuming they don't already.

tlrobinson 12y ago
I assumed the parent comment meant user-provided data, not raw GPS data, since presumably the driver app is always reporting their location (how else would they know which drivers to dispatch?)
- pyrocat 12y ago
  I'm not sure what user-provided data would even be beneficial? Stuff that's likely already collected encompasses where the user is picked up, dropped off, route taken, time of departure and arrival .etc
  - vidarh 12y ago
    Reasons for why drivers always take route X between streets A and B, for example (e.g. street C has been changed to one way only so route Y isn't actually legal/possible).

jfoutz 12y ago

I think the self driving car is a tougher problem to solve than the maps. If uber can buy self driving cars, they can make maps on a city by city basis. If you have a pile of money, you can get maps. If you have a pile of money, it's not really clear you can get an autonomous car. I'm the first to shout they're coming - but they aren't here yet.

ericcumbee 12y ago
And you have to get enough of those self driving cars for it to be effective. If uber is buying the cars then they cease to be a driver/car to passenger matching service and become a full on cab company, complete with the liability and maintenance costs that go with owning cars.
- olefoo 12y ago
  And Uber may still wind up on the losing end of a lawsuit about pushing drivers to take on vehicles based on false pretenses ( Uber telling the drivers that there would be enough work to justify them financing a vehicle for Uber usage). Uber's efforts to shed itself of all liability and risk in terms of employees and assets is legendary; but they could still find themselves going up against the IRS on the question of whether drivers are employees or freelancers. Most especially with drivers that they referred to auto-financing. The theory being that having a requirement for a minimum standard of car and providing finance referrals would be similar to requiring "independent contractors" to wear uniforms.
  - Alupis 12y ago
    > Uber telling the drivers that there would be enough work to justify them financing a vehicle for Uber usage It seems incredibly unwise to finance a $20,000 USD vehicle hedging on some third party (of which you have zero control over) to provide enough for you to pay it through.
    - olefoo 12y ago
      Generally, if you tell someone that taking on a debt obligation is a condition of doing business with you and make representations to the effect that they will be able to earn enough to meet that obligation based on their projected earnings from your contractual agreement with them that would be considered an implied contract. If you then manipulate things such that they are effectively unable to meet their obligations because you are jerking them around they are certainly going to be able to find a lawyer who will take their case. Particularly if you were referring them to the auto-dealership and pre-qualifying them for financing.

TenJack 12y ago

Uber is already pretty much limited to major metropolitan areas, so maybe Uber would really only need map data in these consolidated areas. At least to start.

vidarh 12y ago
That's a major benefit to rolling this out as a cab company rather than as a car manufacturer - you can enforce limits like that and people won't complain because they won't even know about the 98% of roads you haven't automated yet. Assuming you can get regulators to accept it: Map area by area - send out human driven cars if a customer requests a journey you can't cover with driverless cars yet.. Assuming they collect GPS traces, they know exactly what areas will cover the most of their business with the least amount of mapping.
shirro 12y ago
They know the start and end points of the trip and can make a decision whether it can be handled with an autonomous car or a human driver at the outset. They could serve major routes like airports to hotels via autonomous vehicle to begin with and handle everything else with people. That could still represent a massive cost saving.

maxerickson 12y ago

For the US, the current TIGER data is pretty reasonable. They could get away with using that (especially as a starting point), or OpenStreeetMap. Waze demonstrated how well a crowd sourcing effort can work. It's also apparently possible for a small team to start doing sign recognition: http://blog.mapillary.com/update/2015/01/27/traffic-signs.html http://blog.mapillary.com/update/2015/01/27/traffic-signs.ht...

vidarh 12y ago
The TIGER data is a starting point. But it's outdated the day it is complete, and rapidly diverges. For many purposes that's perfectly fine, but as stated above, you need someone driving the street. Even something as major as postcodes changes multiple times in between TIGER updates, and tons of streets are changed or added or removed. The TIGER dataset is fantastic, and certainly helps levelling the playing field by giving people somewhere to start from, but it's the cheap 80%, except in this case the last 20% doesn't take the other 80% of the time, but more like 800%.
- maxerickson 12y ago
  Have you looked at the last couple years of updates? It's way more accurate and complete than it used to be, and for getting names on that last 1%, they have drivers and the ability for the app to say "I don't know that street name, is it new?".
  - vidarh 12y ago
    I do some work for a company that does mapping, so yes, I've looked at it regularly. It may be way more accurate and complete, but it is still updated too rarely to ever be up to date. Cover any reasonable area, and road situations change daily, especially when you're doing things like driving which also needs to account for temporary road closures and the like. As I say, it's a good starting point, but it's not in any way sufficient, and keeping it up to date is a huge amount of work, which has been reflected in the value of companies that have set themselves up to be able to keep up to date data sets of this. That said, as I've commented elsewhere I do agree that Uber is in a good position, but mostly because they don't need full coverage. They need thorough and up to date information, but they need it in very tiny areas to begin with, and as you say they can rely on drivers to cover areas they don't currently know how to deal with. They can apply GPS traces to analyse exactly what areas are most profitable to improve map information for to make self driving viable first, and roll it out area by area.

threeseed 12y ago

Uber can simply license their maps from various third parties the same way every car company does today for their GPS. Google has added some decent additions to Google Maps but really 95% of what they had they purchased. For example Whereis/Yellow Pages data here in Australia.

kumarm 12y ago
There are too many sources and too many processes in world wide Maps data (Apple learnt hard way). I doubt Uber has the capability to have decent maps solution in short term (1-2 years).
jedberg 12y ago
Licensed maps are great for GPS units telling a human where to go, but a self driving car needs much more detailed information that most likely only google has.
- threeseed 12y ago
  If Google is the only company with this information how come all the car manufacturers have active projects that don't rely on their data. It's because Google has nothing that other companies don't have/need. It's not like Google magically acquired all their data. They bought most of it.
runeks 12y ago
Map data from third parties requires manual verification, as we learned from the Apple Maps debacle. Google does this [1]: > Google employs a small army of human operators (they won’t say exactly how many) to manually check and correct the maps using an in-house program called Atlas. A Maps app isn't just another app. Nobody gets killed if the Stocks app shows an incorrect price, but someone could die if the Maps app uses incorrect mapping information. [1] http://www.wired.com/2014/12/google-maps-ground-truth/ http://www.wired.com/2014/12/google-maps-ground-truth/

toomuchtodo 12y ago

> So, Uber really needs to start thinking about the mapping aspect. Perhaps they can incent their current drivers to start reporting back all the street info as they pick up their passengers. Of course, this would bring up problems if the drivers eventually realize that this data could be used to replace them. It would be interesting to see uber cars with Google StreetView style cameras on their hoods. Most likely they would either start having drivers drive Uber cars with steetview-esq data collection gear on it, or pay a premium to drivers who mounted a tray on the roof of their vehicle that contained high resolution cameras, a laser scanner, and a controller to process and forward that data back. I would not be surprised if Uber was looking at acquiring https://www.mapillary.com https://www.mapillary.com for this functionality.

trhway 12y ago
Self-driving cars will solve the issue of collecting street data. Though of course one need some data for the cars to start with.

jlarocco 12y ago

One possibility, is that if they can get good enough inital mapping data, and good enough self-driving cars, the cars could update the mapping data while they're working. That's a lot of trust on the self-driving cars, though.

JTon 12y ago
I think a reporting system could solve that problem. Classify potential changes, and force human eyes before approving all but the most minor updates

downandout 12y ago

Self driving cars will be available to purchase from multiple manufacturers, and those manufacturers will have to deal with the mapping issues. Accordingly, while Google will be a formidable and perhaps dominant competitor in this space, I don't see mapping as the deciding factor here. The fact that Android, with its massive install base, will likely come with Google's service built in seems much more likely to help them destroy competitors.

Vik1ng 12y ago

Give it a few years and OpenStreetMap will be superior. Here in Germany it will probably beat google maps most of the time. There is of course still the routing part.

avodonosov 12y ago

Do you think OpenStreetMaps can be a reasonable substitute? IMHO their maps are quite accurate.

seren 12y ago
The issue with OSM is that the information level is uneven. Dense urban center, you have every bench and bin mapped. Go into the countryside, there is not even the buildings. It might not be an issue for a taxi company that is mainly operating in urban center though.
- andygates 12y ago
  It's in the taxi company's interests to improve the maps for everyone: popular routes will be refined to a T, and gaps will quickly become obvious. These days the most common mapping gap is new developments, and they're faster to appear on OSM than the commercials.