“Visual positioning shouldn’t be a really recent technology,” says Konrad Wenzel at ESRI, an organization that develops digital mapping and geospatial evaluation software. “Nevertheless it’s obvious that the more cameras we have now on the market, the higher it becomes.”
Niantic Spatial has trained its model on 30 billion images captured in urban environments. Particularly, the pictures are clustered around hot spots—places that served as essential locations in Niantic’s games that players were encouraged to go to, corresponding to Pokémon battle arenas. “We had a million-plus locations world wide where we are able to locate you exactly,” says McClendon. “We all know where you’re standing inside several centimeters of accuracy and, most significantly, where you’re looking.”
The upshot is that for every of those million locations, Niantic Spatial has many hundreds of images taken in roughly the identical place but from different angles, at different times of day, and in numerous weather conditions. Each of those images comes with detailed metadata that pinpoints where in space the phone was on the time it captured the image, including which way the phone was facing, which way up it was, whether or not it was moving, how briskly and by which direction, and more.
The firm has used this data set to coach a model to predict exactly where it’s by considering what it’s taking a look at—even for locations apart from those million hot spots, where good sources of image and placement data are scarcer.
Along with GPS, Coco’s robots, that are fitted with 4 cameras, will now use this model to attempt to determine where they’re and where they’re headed. The robots’ cameras are hip-height and point in all directions directly, so their viewpoint is a little bit different from a Pokémon Go player’s, but adapting the information was straightforward, says Rash.
Rival firms use visual positioning systems too. For instance, Starship Technologies, a robot delivery firm founded in Estonia in 2014, says its robots use their sensors to construct a 3D map of their surroundings, plotting the sides of buildings and the position of streetlights.
But Rash is betting that Niantic Spatial’s tech will give Coco an edge. He claims it would allow his robots to position themselves in the right pickup spots outside restaurants, ensuring they don’t get in anybody’s way, and stop just outside the shopper’s door as a substitute of a number of steps away, which may need happened previously.
A Cambrian explosion in robotics
When Niantic Spatial began work on its visual positioning system, the thought was to use it to augmented reality, says Hanke. “When you are wearing AR glasses and you would like the world to lock in to where you are looking, then you definitely need some method for doing that,” he says. “But now we’re seeing a Cambrian explosion in robotics.”
