A decade after captivating the world, the augmented reality phenomenon Pokémon Go is having an unexpected impact on the future of robotics. Niantic Spatial, the company evolved from Pokémon Go creator Niantic, is now leveraging the game’s vast database of player-submitted data to train delivery robots, demonstrating the surprising synergy between mobile gaming and artificial intelligence.
The core of this innovation lies in the 30 billion images and videos voluntarily contributed by Pokémon Go players over the past ten years. These crowdsourced visuals, capturing street corners, landmarks and urban intersections, have been transformed into a highly detailed, continuously updated model of the physical world. This model is now being used to navigate the fleet of roughly 1,000 delivery bots operated by Coco Robotics, a US-based robotics firm, across cities including Los Angeles, Chicago, Miami, Jersey City, and Helsinki. The robots have already logged millions of miles of deliveries.
Niantic Spatial, rebranded to focus on enterprise AI and mapping solutions, isn’t simply repurposing gaming data. The company explains that the player-generated content provides “very high-quality ground training data for other lower-quality datasets,” according to Niantic Spatial’s chief technology officer, Brian McClendon. This approach allows for the creation of robust localization, reconstruction, and semantic understanding capabilities, even when relying on less precise data sources.
The challenge Niantic Spatial addresses is particularly relevant in urban environments. “The urban canyon is the worst place in the world for GPS,” explained Brian McClendon to MIT Technology Review, referencing how tall buildings interfere with GPS signals. The Visual Positioning System (VPS) developed by Niantic Spatial overcomes this limitation by pinpointing locations with centimeter-level accuracy, relying on visual data rather than satellite signals. This is especially useful in indoor spaces or areas with obstructed GPS access.
The partnership with Coco Robotics marks Niantic Spatial’s first foray into the robotics industry. “The promise of last-mile robotics is immense, but the reality of navigating chaotic city streets is one of the hardest engineering challenges,” said Niantic Spatial CEO John Hanke. The utilize of Pokémon Go data highlights a novel approach to solving complex navigation problems, showcasing how seemingly disparate technologies can converge to drive innovation.
The data collection was facilitated by in-game features like ‘Field Research,’ which incentivized players to scan real-world locations for rewards. These scans are then converted into 3D models, further enhancing the accuracy of Niantic Spatial’s VPS technology. This demonstrates a unique model of collaborative data gathering, where players unknowingly contribute to advancements in fields beyond gaming.
As noted in IGN on March 16, 2026, the underlying principle is surprisingly simple: getting a virtual Pikachu to realistically navigate a real-world environment shares fundamental challenges with guiding a delivery robot through a busy city street.
