Object Pose Distribution Estimation for Determining Revolution and Reflection Uncertainty in Point Clouds
This addresses pose uncertainty for robotics in industrial settings where color information is often unavailable, representing a novel approach but with a current focus on specific symmetries.
The paper tackles the problem of object pose uncertainty estimation in robotic perception by proposing a neural network-based method that uses only 3D colorless data, achieving validation in a real-world bin picking scenario with geometrically ambiguous objects.
Object pose estimation is crucial to robotic perception and typically provides a single-pose estimate. However, a single estimate cannot capture pose uncertainty deriving from visual ambiguity, which can lead to unreliable behavior. Existing pose distribution methods rely heavily on color information, often unavailable in industrial settings. We propose a novel neural network-based method for estimating object pose uncertainty using only 3D colorless data. To the best of our knowledge, this is the first approach that leverages deep learning for pose distribution estimation without relying on RGB input. We validate our method in a real-world bin picking scenario with objects of varying geometric ambiguity. Our current implementation focuses on symmetries in reflection and revolution, but the framework is extendable to full SE(3) pose distribution estimation. Source code available at opde3d.github.io