But Cosy, a startup founded by University of Pennsylvania graduate Jonas Cleveland, might have the solution — a software framework that combines neural networking and indoor positioning to pinpoint the location of robots, smartphones, and more in three-dimensional space.
And unlike most positional tracking technologies, which rely on infrared sensors to orient objects in rooms and hallways, Cosy’s platform works on any device with a camera. “It doesn’t matter the platform — it works on fixed cameras” Cleveland said. “A lot of folks claim you’re not able to achieve [3D mapping] without using lidar, but we’ve done it. It’s all post processing.”
It’s a little like Snapchat’s Lenses, which use computer vision and augmented reality to lay digital objects on top of real-world environments. Cleveland wouldn’t spill the beans on Cosy’s tech, some of which is patent-pending. But if it’s anything like Lenses, it takes into account factors like rotation, occlusion, and more in real time.
Cosy is applying it to retail — specifically inventory management. Robots that are four-and-a-half feet tall robots (“about the size of a fourth grader,” Cleveland said) equipped with the startup’s software navigate the store aisles autonomously, taking note of the products on shelves. They relay that data to a central server, which updates the store’s inventory database.
Cosy sees it lightening human workloads. “We’re empowering humans to do other things,” Cleveland said. “People are not very good at tedious and mundane tasks. Inventory is incredibly labor intensive, and very inaccurate — humans hate doing inventory. We’re checking stock and seeing where items are located in the store through software, without human assistance.”
It might not be long before you see a Cosy-powered ‘bot at your local department store. The startup is in talks with three of the top twenty U.S. retailers to deploy the inventory-checking robots in the next year. And in the future, Cosy hopes to adapt the tech to smartphones — Cleveland envisions an app that directs customers to a particular section of a store, for example, or even an exact shelf.
“When you can map indoor space and all the objects, that solves a lot of problems,” said Cleveland.
