The article introduces AdaFold, a model-based framework for optimizing cloth folding trajectories through feedback-loop manipulation. It leverages semantic descriptors from pre-trained visual-language models to enhance the particle representation of cloth. The experiments demonstrate AdaFold's ability to adapt folding trajectories to cloths with varying physical properties and generalize from simulation to real-world execution. The content is structured into sections covering Introduction, Related Work, Problem Formulation, Cloth Perception, Trajectory Optimization, Implementation Details, Experimental Results, Ablation Study, Real World Experiments, and Semantic Cloth Representation.
Na inny język
z treści źródłowej
arxiv.org
Głębsze pytania