Pokémon Go Data Powers Real-World Robotics

The Pokémon Go phenomenon, which captivated hundreds of millions of players a decade ago, has yielded a surprising legacy: a highly detailed, continuously updated map of the physical world. This dataset, built from player-submitted photos and videos, is now being utilized by robotics companies like Coco Robotics to improve the navigation and precision of delivery bots.

From Gaming Sensation to Mapping Revolution

Over the past decade, Pokémon Go players voluntarily contributed over 30 billion images of public landmarks, street corners, and urban environments. Niantic Spatial, the enterprise AI and mapping division spun out of Niantic Inc., has transformed this trove of data into a photorealistic, street-level model specifically designed for robotic applications. According to Niantic Spatial’s chief technology officer, Brian McClendon, one of the original creators of Google Earth, the player data serves as “particularly high-quality ground training data for other lower-quality datasets.”

Navigating the Urban Landscape

Traditional GPS systems struggle in dense urban environments due to interference from tall buildings. Niantic Spatial’s Visual Positioning System (VPS) overcomes this limitation by comparing live camera feeds from robots against its extensive image database to determine position in real-time. This is particularly crucial for autonomous delivery services, where precise positioning is essential for successful deliveries. A Niantic Spatial spokesperson stated that the VPS is “particularly resilient in urban canyons where GPS performs badly.”

How it Works: A Large Geospatial Model

The core of Niantic Spatial’s technology is a large geospatial model (LGM) trained on billions of images and hundreds of millions of real-world scans. This model provides three key capabilities: reconstructing spaces as navigable 3D models, localizing machines within those spaces, and understanding environments semantically. As Niantic CEO John Hanke wrote in a recent blog post, the LGM acts as a “living, breathing map of the world, one that is native to robots, and AI.”

Coco Robotics and the Last-Mile Delivery Challenge

Coco Robotics, with a fleet of approximately 1,000 delivery bots operating globally, is leveraging Niantic Spatial’s VPS to enhance its service. Zach Rash, CEO of Coco, highlighted the importance of precise delivery, noting that a poor customer experience results when a robot delivers to the wrong location. He emphasized the need for robots to possess a level of “intuition” similar to humans, particularly in challenging urban environments. Rash stated, “If I can more precisely figure out where to drop off food, my customers will be happy.”

The Future of Real-World Mapping

Niantic Spatial’s approach to mapping is unique, relying on crowdsourced data and continuous updates. Whereas initially built on scans from Pokémon Go players, the model now incorporates data from enterprise customers as well. This collaborative approach positions Niantic Spatial as a significant player in the emerging field of real-world mapping, offering a solution to the challenges faced by the robotics and autonomous vehicle industries.