A decade after Pokémon Go's peak, Niantic Spatial, an AI company spun out of the game's developer, is leveraging billions of real-world images captured by players to develop navigation technologies for delivery robots and potential military drone applications. The company's models are trained on geolocated images from the Scaniverse app and Pokémon Go scans, which include metadata about location and phone orientation. This data has enabled the development of a visual positioning system that can help devices navigate in GPS-denied environments. The system is being used by Coco Robotics for delivery robots and by Vantor for military and defense applications. Source: arstechnica
Niantic Spatial trained its model on 30 billion images, mostly from urban areas, captured by players incentivized to scan public locations. The images were taken from various angles and under different lighting and weather conditions, providing rich data for training AI models. The company emphasized that the scans were optional and that users were informed since 2019 about how their data would be used to improve technology platforms. This transparency allowed Niantic Spatial to develop its visual positioning system, which can provide accurate position and orientation data for devices. Source: arstechnica
Niantic Spatial's partnership with Vantor, a company with US military contracts, has led to the development of a comprehensive positioning system that integrates visual data with 3D terrain information. During a conference in February 2026, the system was reported to reduce positioning error by 70% with accuracy within 1.5 meters in many scenarios. However, some players expressed concern about their data being used for military applications, highlighting the ethical implications of data collected for entertainment being repurposed for defense. Source: arstechnica
