paper/arxiv/2608.28570

Robotics Dataset

ChainSplat is a physics-inspired framework that jointly learns the 3D geometry, appearance, kinematics, and dynamics of deformable linear objects (DLOs) from multi-view RGB videos, using a screw-theoretic model with Gaussian splatting for real-time state and force estimation and model-based trajectory optimization.

Catalog metadata

Repository owner

Seungyeon Kim, Noémie Jaquier

Source

Research paper

Task

ChainSplat: A Physics-Inspired Screw-Theoretic Model for Learning Deformable Linear Object Dynamics from Multi-View RGB Videos

Cameras

0

Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.