Search papers, labs, and topics across Lattice.
This paper introduces BoxTwin, an interactive digital twin framework designed to learn the dynamics of elastoplastic articulated objects (EAOs) from video data. By reconstructing scenes and identifying physics-aware constitutive models, BoxTwin effectively tracks joint trajectories and simulates post-contact plastic behavior in manipulation tasks. The results demonstrate that BoxTwin significantly enhances the predictive and adaptive control capabilities of robots interacting with complex deformable objects in unstructured environments.
BoxTwin enables robots to accurately predict and adapt to the complex dynamics of elastoplastic articulated objects, revolutionizing manipulation in unstructured settings.
Digital twins enable robots to anticipate and adapt to physical interactions, but existing models struggle with elastoplastic articulated objects (EAOs) that exhibit nonlinear elasticity, plastic yielding, and damage accumulation. We present BoxTwin, an interactive digital twin framework that learns the full dynamics of EAOs from videos. Our pipeline reconstructs the scene, identifies a physics aware constitutive model for each EAO. Experiments on manual folding and dual arm manipulation of EAOs show that BoxTwin accurately tracks joint trajectories and reproduces post contact plastic behavior over long horizons. By integrating video driven reconstruction with elastoplastic damage modeling, BoxTwin advances digital twins toward predictive, adaptive control of deformable articulated objects in unstructured environments.
