Ride-Hailing Demands Good GPS & AVs Will Need It Even More

If you feel like GPS lets you down sometimes — say, when you’re hailing a car downtown and the driver can’t find you — how do you think it’ll be when there’s no driver? That’s a problem ride-hailing companies are already working on.

A phone’s Geographic Positioning System (GPS) radio needs clear lines of sight to satellites to determine its location. In built-up areas, those views can be hard to get because of tall buildings, and satellite signals bouncing off walls can make things even more confusing. That’s why GPS readings in cities sometimes place users a block away from where they really are.

Ride-hailing companies have been trying to tackle this problem for years, because if a driver goes to the wrong spot to pick up a passenger, the two may never connect and no one will get paid. GPS accuracy will get even more important when autonomous vehicles are picking up the passengers – or just trying to drive on the correct side of the street.

Uber even acquired a company, ShadowMaps, that uses software and 3D maps to calculate where someone really is. ShadowMaps compares the GPS satellites that a phone can see with the ones that would be blocked by certain buildings in the area. Using that information, it can narrow down the phone’s location to where those particular satellites would be hidden from view, The Verge reported last week.

Uber’s rival, Lyft, faces the same issue and has its own solutions.

“GPS is not terribly accurate. It tends to have all kinds of errors,” Vinay Shet, a Lyft product manager focused on mapping, told The Connected Car. “When tall buildings crowd you out, your GPS can go all over the place.”

It can be a problem both for riders trying to show where they are and drivers trying to get there. (Shet said it’s rare for a Lyft to miss a pickup, but he wasn’t able to share numbers.) Lyft has taken steps for both, Shet said.

One solution asks riders to adjust, more than the technology. When a rider requests a car in an area where a lot of other Lyft customers have hailed rides, the app can tap into Lyft’s historical data and suggest spots nearby where pickups often happen. That place may be easier for drivers to find if they’ve picked up passengers there before, and it may be an easier spot for cars to stop.

Then there’s the driver’s location. On the way to a pickup point or a ride destination, GPS interference may make a Lyft car appear to jump from block to block in an illogical sequence. To get around this, Lyft can use map data and route history to “snap” the car’s location to the street the car’s been traveling on, Shet said.

The app can also use history and machine learning to bias its estimates of a driver’s location toward routes that other Lyft drivers have used in the area. Any given driver might be taking a path that’s never been taken, but the goal is to filter out GPS readings that are more likely to be incorrect.


Want to hear more about the leading operator use cases for AI technologies? Join us in Austin from May 14-16 at the fifth-annual Big Communications Event. There’s still time to register and communications service providers get in free!


For autonomous cars, accurate GPS is even more essential. Lyft is developing and testing AV technologies itself and offers a software platform to partners, which currently include General Motors, Ford and Jaguar.

Human-driven cars can get by with GPS readings that are accurate down to a few meters, but AVs will need to know their location down to a few centimeters, Shet said. Being in the correct lane depends on it. Other tools can supplement GPS to make this possible.

One is an inertia sensor, which allows the car’s navigation system to perform dead reckoning: determining location by sensing which direction it’s been moving, how long, and how fast. Another is lidar, the radar-like sensors that can build a real-time map of a vehicle’s surroundings by bouncing lasers off every object.

Because Lidar builds a 3D map of a given street every time an AV drives there, and those maps can be combined and updated, the next AV going down the street has a good idea of where it is in relation to all the objects around it. This helps achieve five- to ten-centimeter precision, Shet said.

As a result, AVs will actually have less trouble finding waiting customers, he said. What will be hard for AVs are the nuances of where to stop: red zones are illegal, white zones are legal but have to be available, it has to be an easy place for the rider to get into the car, and so on. Human drivers get these nuances, Shet said, but AVs will have to learn them.


Leave a comment

Your email address will not be published. Required fields are marked *