Robust Models of Object Geometry

Precise and accurate models of the world are critical to autonomous robot operation. Just as robot navigation typically requires an accurate map of the world, robot manipulation typically requires accurate models of the objects to be grasped. However, the statistical inference tools that enable robot mapping have not yet had the same impact in geometric object modelling. We describe an inference algorithm for learning statistical models of objects from image data. We describe a representation that allows us to compute a distribution over the complete geometry of different objects, and describe how a library of object geometry models can be learned. Finally, we describe how learned object models can both be used to recognize new instances of objects and to infer the geometry of occluded parts of objects.