A Belief representation for Active Uncertain Pose Estimation via Simulation-based Inference

Aditya Kamireddypalli, Matias Mattamala, Joao Moura, Russell Buchanan, Sethu Vijayakumar, Subramanian Ramamoorthy

Keywords: Sampling based Inference, Simulation based Inference, Perception for Dexterous Manipulation

Abstract:

Contact-rich manipulation tasks require more accu-
rate scene understanding than is normally possible solely using
visual sensing. For peg-insertion, the pose of the hole must be
estimated; however, camera noise can easily exceed insertion
part tolerance. By leveraging force and torque measurements
during part interaction, contact sensing can provide valuable
information that refines estimates based on noisy visual sens-
ing and enable contact-rich peg insertion. Here we propose a
Simulation-based Inference framework to fuse visual information
and contact information via probabilistic program proposals
utilising rendering and physics as generative functions

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 15