Rethinking GPS: Engineering Next-Gen Location at Uber

Rethinking GPS: Engineering Next-Gen Location at Uber

• Location and navigation using GPS
  • deeply embedded in our daily lives,
• crucial to Uber’s services.

 

• To orchestrate quick, efficient pickups,
• need to know
  • locations of matched riders and drivers,
  • provide navigation guidance
    • from driver’s current location
    • to
    • where the rider needs to be picked up,
  • then to rider’s destination.

rider指的是打车的人啊？这么用的吗？

 

• For this process to work seamlessly,
  • location estimates for riders and drivers
  • need to be precise

第二段

• Since (literal!) launch of GPS in 1973
• we have advanced our understanding of the world,
  • experienced exponential growth in the computational power
  • developed powerful algorithms to
    • model uncertainty from fields like robotics.

啥意思啊？啥叫model uncertainty from fields like robotics

 

• fundamentals of how GPS works have not changed that much,
  • leads to significant performance limitations.

 

• rethink some of the starting assumptions
  • that were true in 1973
    • regarding where and how we use GPS,
  • as well as the computational power
  • additional information we can bring to bear to improve it.

第三段

• GPS works well under clear skies,
• wildly inaccurate (with a margin of error of 50 meters or more) when we need it the most:
  • densely populated
  • highly built-up urban areas,
• where many of our users are located.

 

• a software upgrade
  • to GPS for Android
• improves location accuracy in urban environments
  • via a client-server architecture
  • utilizes 3D maps
  • performs sophisticated probabilistic computations
    • on GPS data
    • available through Android’s GNSS APIs.

第四段

• why GPS perform poorly in urban environments
• how we fix it using
  • advanced signal processing algorithms
  • deployed at scale on our server infrastructure.

• standard GPS (red)
• improved location estimate (blue) for a pickup
  • from Uber HQ in San Francisco.
• Our estimated location
  • follows the true path taken by the rider,
  • GPS shows very large excursions.

A bit of background on GPS/GNSS

• a quick recap of how GPS works
  • in order to understand
  • why it can be inaccurate in high-rise urban environments.

第二段

• GPS
  • network of more than 30 satellites
  • an altitude of about 20,000 kilometers.
• (Most cell phones can pick up similar Russian “GLONASS” satellites too.)

 

• satellites send out
  • radio frequency signals
  • that GPS receivers, those found in cell phones, can lock onto.

 

• these satellites advertise the time at which
  • they launch their signals.

第三段

• For each satellite
• the difference between reception time and launch time (time-of-flight),
  • multiplied by the speed of light,
  • the pseudorange.

 

• If the satellite and receiver’s clocks are synchronized,
• signal travels along
  • the straight line-of-sight path,
• this equal the actual distance to the satellite.

 

• the clocks are not synchronized,
  • so the receiver needs to solve for four unknowns,
  • its own 3D coordinates on the globe,
  • and its clock bias.

 

• a minimum of four satellites (four equations)
  • to solve for these four unknowns.

第4段

• ignore bias
• the location estimate
  • intersecting
  • spheres centered at the satellites with the radius of each sphere given by the pseudorange.

 

• GPS receiver processes signals
  • from a number of satellites (20 GPS and GLONASS satellites are visible in an open field),
• more than the minimum number of equations provides extra robustness to noise, blockages, etc.

 

• some new/future receivers
  • can/will process signals from other satellite systems.

 

• Other navigation satellite systems coming online
  • are Galileo, operated by the European Union,
  • IRNSS in India,
  • and BeiDou, operated by China.

 

• GNSS (global navigation satellite systems)
  • encompasses these systems.
  • (use this term in the remainder)

• In this simplified interpretation of GPS receiver computation,
• spheres intersect at the center of known satellite locations.

球相交于已知卫星位置的中心？？这什么鬼？

Why GNSS location is inaccurate in urban environments

• behind GNSS-based positioning
  • receiver has a direct line-of-sight to each satellite whose pseudorange it is computing.

 

• really breaks down in urban environments, as shown in Figure 3

• Line-of-sight blockage and strong reflections can cause large GPS error

第二段