Liatris is an open-source project that determines any object’s exact pose and identity using a touch screen and RFID reader. Liatris includes a global API powered by a database of objects. An object’s 3D CAD model can be added to the robot’s 3D perception. The API also provides recommendations about grasping objects relative to the CAD model and pose, resulting in an accurate mobile manipulation solution.
Liatris is released under the BSD license.
At this time, Liatris is a proof of concept. A stable library is in development.
Although computer vision algorithms are powerful, determining an object’s pose and identity exclusively using cameras, LiDAR or structured light is a major obstacle to robotics becoming part of our everyday lives. Also, the introduction of thousands of new products every day further compounds the vision-only approach, as many of these products are visually very similar in shape, color, etc., if not identical.
The ability to reliably read an object’s pose and identity without relying on vision will help to break down barriers for robotics and many other applications such as augmented reality.
In the example code, a robot is programmed to identify and grasp any object placed on a touch screen. Regardless of what object is placed on the table, Liatris identifies its position, orientation and identity using capacitive touch and RFID technologies. Using the CAD model downloaded from Liatris’ API, the robot can perceive the object. Finally, the robot can grasp the object using instructions from Liatris’ API that state the ideal way to interact with that specific object, while avoiding collisions with other objects. This results in an accurate 3D perception and mobile manipulation solution.