For years, Pokémon Go trainers have explored streets and parks to capture creatures. Unknowingly, their millions of geolocated images have served to train Niantic's visual positioning system (VPS). Now, that technology takes a practical turn: a collaboration with Coco Robotics will use that VPS to guide delivery robots with centimetric precision. 🤖
From AR to logistics: the leap of VPS trained by crowdsourcing 🗺️
Niantic's VPS does not rely solely on GPS, which suffers from blockages and low precision in cities. It analyzes visual features of the environment, such as facades or signs, to determine an exact location. This model was trained with the vast photographic database generated by players. When applied to robots, they can navigate sidewalks and recognize portals autonomously, optimizing last mile routes.
Your Pidgey hunting run now funds someone's pizza 🍕
The next time a robot dodges you on the sidewalk on its way to a delivery, think that maybe you were the one who trained it. Your photos of a Snorlax in front of the portal, uploaded in 2016, are now training data for a machine to deliver food. The circle closes: you hunted virtual monsters for fun and, without asking permission, became an unpaid surveyor of the logistics revolution.