Deformable objects present a formidable challenge for robotic manipulation due to the lack of canonical low-dimensional representations and the difficulty of capturing, predicting, and controlling such objects . We construct compacttopological representations to capture the state of highly deformable objectsthat are topologically nontrivial. We develop an approach that tracks the evolution of this topological state through time. We prove that the topology of the scene and its evolution can be recovered from point clouds representing the scene. Our further contribution is a method to learn predictive models that take a sequence of past point cloudobservations as input and predict sequence of topological states, conditioned on target/future control actions. These models offerfast multistep predictions suitable for real-time applications.

Author(s) : Rika Antonova, Anastasiia Varava, Peiyang Shi, J. Frederico Carvalho, Danica Kragic

Links : PDF - Abstract

Code :

Keywords : representations - objects - deformable - topological - models -

Leave a Reply

Your email address will not be published. Required fields are marked *