Pokémon Head’s Legacy: Mapping the World, One Pokémon at a Time
The 2016 augmented reality (AR) phenomenon, Pokémon Go, captivated hundreds of millions of players worldwide. In its first sixty days, the app amassed around 500 million users [1]. Beyond the thrill of catching virtual creatures, players unknowingly contributed to a massive, crowdsourced mapping project that is now powering advancements in robotics and artificial intelligence.
Players, while focused on the game, inadvertently mapped their surroundings using their smartphone cameras. Niantic, the game’s developer, capitalized on this activity, collecting roughly thirty billion photos and videos of public spaces from over twenty million locations globally – encompassing landmarks, streets, parks, and urban corners [1]. Each image is geotagged with precise location data, time of day, and other metadata.
This vast collection of visual data has created a highly accurate geospatial model, capable of pinpointing locations with centimeter-level precision, surpassing the accuracy of traditional satellite-based maps [1]. Unlike static maps, this system dynamically captures spaces under varying conditions, angles, and times of day.
Voluntary Scans, Unexpected Applications
Players voluntarily scanned their surroundings, often incentivized by in-game rewards, to enhance the AR experience. However, most were unaware that these images would eventually be used to train artificial intelligence and guide autonomous delivery machines [1]. This isn’t an isolated case; companies like Tesla, Amazon, Meta, and Google also employ similar data collection strategies.
Niantic Spatial, a spin-off from Niantic, is now commercializing this data. The company’s Large Geospatial Model (LGM) and Visual Positioning System (VPS) are at the core of its offerings [1].
Navigating the Real World with Pokémon Go Data
Traditional satellite navigation systems struggle in dense urban environments due to signal obstructions from tall buildings. Niantic Spatial’s Visual Positioning System (VPS) addresses this limitation by enabling orientation based solely on camera images, comparing them to the extensive image database [2].
Coco Robotics, a startup operating approximately 1,000 delivery robots across several cities including Los Angeles, Chicago, and Helsinki, is the first major partner utilizing this technology [2]. These robots, which navigate sidewalks at pedestrian speed, rely on the Pokémon Go-derived data for precise navigation and delivery services.
Data Ownership and the Future of Mapping
The utilize of crowdsourced data from Pokémon Go raises questions about data ownership and user consent. While players agreed to data-sharing terms as part of the game’s conditions, many were likely unaware of the extent to which their contributions would be utilized [4].
Niantic Spatial intends to continuously refine its model, with the robot fleets providing real-time feedback and updates, creating a dynamic and evolving map of the physical world [2]. As this technology matures, it’s crucial to consider the implications of our digital entertainment and who ultimately benefits from the data we generate.