Light Dark

Now Reading: Pokémon Go players helped build new maps of cities

01
Pokémon Go players helped build new maps of cities

Light Dark

Home
Top Posts
Technology
Pokémon Go players helped build new maps of cities

Pokémon Go players helped build new maps of cities

AhmadJunaidTechnologyMay 19, 2026358 Views

In your hometown, you can glance at a building or landmark to orient yourself. A new computer system uses artificial intelligence to do much the same thing. Called VPS 2.0, it’s a type of visual positioning system. It lets machines “look through a camera and know exactly where they are,” explains Brian McClendon. This system went global in April.

And if you’ve played Pokémon Go, you might have helped make it possible.

McClendon led the team that developed VPS 2.0 for Niantic Spatial, based in San Francisco, Calif.

In some cities, delivery robots will use VPS 2.0 to bring pizza or groceries to someone’s door. “We’re really excited [for the new system],” says George O’Brien. He leads product development at Coco Robotics, also based in San Francisco.

Robots delivering pizza sounds fun, says Kathleen Tuite, who does not work with either company. But she worries about our privacy. A system like this “could probably figure out where you are from certain kinds of photos,” she points out. Tuite is a software developer at ODK. This company, based in San Diego, Calif., helps make tools people can use to collect data.

Since VPS 2.0 only works live, there’s no way to put someone’s photos into the system, says a spokesperson for Niantic Spatial. As the system receives images from a camera in real time, it sends back a location. It also identifies the direction that camera had been facing — such as north, south, east or west.

The role of Pokémon Go players

To play Pokémon Go, someone walks through the real world looking at the screen on their device. It displays a map of their area. Virtual critters — called Pokémon — will appear there. You might capture a Pikachu, Charizard or (if you’re super lucky) a Galarian Zapdos. The popular game has been around for 10 years. Tens of millions of people around the world still play it each month.

Players can also scan public landmarks with their phones — such as a statue or a library. Scanning typically earns players rewards in the game.

But that’s not the end of the story.

Together, all those images and videos that players have captured end up showing what landmarks look like from all angles — and in all kinds of weather and lighting. These data were essential for building out VPS 2.0. If you did any of this scanning, McClendon says, you helped “to build technology that will now begin to guide real robots through city streets.” He stresses that scanning was always optional. (Players were also told their scans would be used to develop new technology.)

Last year, the company that makes Pokémon Go split into two. Niantic Labs now runs Pokémon Go and several other games. Niantic Spatial focuses instead on spatial intelligence — such as tools for mapping and finding locations.

In places where Niantic Spatial has detailed scans of buildings or landmarks, VPS 2.0 can determine the position of a camera on a robot or other device that is using the system to within centimeters (inches), McClendon says.

But the map also works near landmarks that people have not yet scanned. And that’s new.

To do this, the company developed an AI system that learns how to find a position from less detailed visual data. This system doesn’t need millions of scans of every new location. All it essentially needs is “a quick cheat sheet,” says McClendon.

Where did data for the cheat sheet come from? That’s a company secret, he says.

And this concerns Tuite. She wonders, what parts of our data are getting slurped up and used in ways that we don’t know about? We just have to hope, she says, that tech companies will handle our data with care.

Do you have a science question? We can help!

Submit your question here, and we might answer it an upcoming issue of Science News Explores

Robot deliveries

VPS should make it easier to navigate than GPS does, especially in dense cities.

That’s where Coco robots operate.

The robot looks like a pink “cooler on wheels,” notes O’Brien. Each can hold up to eight large pizzas or several bags of groceries — and has cup holders for drinks. To pick up or deliver an order, a Coco robot trundles along sidewalks until it reaches its destination.

Coco robots will soon make deliveries in some U.S. cities — Los Angeles, Miami, Chicago and Jersey City — as well as in Helsinki and Turku, Finland. “We’re always expanding to new cities,” says George O’Brien. Coco Robotics

Currently, to keep track of where they are, these robots rely on detailed maps and many different types of sensors. They also use GPS. It relays signals between several satellites in space and a single device on Earth. To identify where it is, the GPS system does some math, based on the time it takes those signals to arrive.

In open areas, GPS can figure out where a robot (or a phone or car) is to within about 4.9 meters (16 feet). But that system doesn’t work as well in cities. For one thing, city sidewalks are around 1.2 meters (4 feet) wide. So a GPS system usually can’t tell if a robot is on the sidewalk or in the street. Plus, if a GPS signal hits a big building, “it bounces off it a few times,” notes O’Brien. That can leave GPS-directed robots lost or confused. “Sometimes we might take a wrong turn and then need to go back,” he says.

The new VPS system, however, shines in this type of dense city environment. The reason: A city has recognizable landmarks everywhere. “Our VPS gives the robot a visual anchor,” says McClendon. “Instead of trusting a bouncing GPS signal, it looks at what its cameras see, matches that to our spatial model and knows within centimeters where it is.”

He and the Niantic Spatial team have been working with Coco to outfit its robots with VPS 2.0 tech.

Most mapping apps use a global positioning system (GPS), left, to figure out where you are. GPS bounces signals to satellites and back. A visual positioning system (VPS), right, instead looks at camera views of landmarks, then matches these to a 3-D model of the area. Niantic Spatial

Grassroots mapping

Five years before Pokémon Go launched, Tuite pioneered the idea that ordinary people could help build a virtual map. Her game, Photo City, came out in 2011. Back then, she was a graduate student at the University of Washington in Seattle. Few cell phones had cameras at the time. So players snapped photos of buildings with digital cameras and uploaded them.

At one point, her university faced off against Cornell (in upstate New York) in a battle to see who could map the most of their campus this way. “I think Cornell ultimately won,” she says.

Once Pokémon Go launched, Tuite got into that game. She recalls scanning landmarks, but never got feedback on them. Players never learned if they had supplied good data, she says, or what their scans might be used for.

In Photo City, she points out, players knew exactly why they snapped pictures of buildings. They got to see the new section of a building they’d photographed appear on the virtual map. Tuite thinks that such an open, community-based approach to collecting data leads to more trustworthy tech.

What do you think?

Upvote0PointsDownvote

0 Votes: 0 Upvotes, 0 Downvotes (0 Points)