Images and videos captured by millions of Pokémon Proceed players over nearly a decade are now guiding autonomous robots on city streets, reducing reliance on GPS during urban navigation.
Millions of people taking to the streets with their smartphones in 2016 to hunt for Pokémon may not have realized they were simultaneously building one of the largest photographic archives ever assembled. Niantic, the creator of Pokémon GO, was leveraging each image and video captured by players to map streets, buildings, landmarks, and points of interest in cities around the world. Now, that collective effort is being put to work training delivery and meal-service robots.
Niantic Spatial, the company’s artificial intelligence (AI) division, announced a partnership with Coco Robotics to optimize the startup’s fleet of autonomous robots. The collaboration aims to utilize the augmented reality data collected by players to power a visual positioning system (VPS), capable of identifying a robot’s location based on its surroundings, rather than relying solely on GPS signals. This development underscores the growing trend of repurposing large datasets for unexpected applications in robotics, and AI.
According to Niantic Spatial, the underlying logic is surprisingly straightforward. “Making Pikachu run realistically and making Coco’s robots move safely and accurately around the world is, in fact, the same problem,” the company stated. In both cases, the task requires associating an element—virtual or physical—with real-world space as precisely as possible.
This solution is particularly valuable in urban environments, where GPS signals are often unreliable. Brian McClendon, Chief Technology Officer at Niantic Spatial, explained to MIT Technology Review: “The urban canyon is the worst place in the world for GPS.” Tall buildings reflect and interfere with radio signals, causing navigation errors precisely where delivery robots need the most accuracy.
Over the years, players have captured images and videos of PokéStops, gyms, and landmarks, contributing to the creation of detailed three-dimensional models of numerous cities. The system was further enhanced with missions that encouraged players to explore and digitize environments in exchange for in-game rewards, which accelerated data collection and improved the quality of urban mapping.
Niantic states it has been transparent about data collection, ensuring the terms were explicitly outlined in the application’s policies. However, few users likely connected photographing a statue to earn points in the game with contributing to the training of AI systems with commercial applications.
