Скачать или смотреть RMDO 2024: UniGarment: A Unified Simulation and Benchmark for Garment Manipulation

Spotlight talk at 4th Workshop on Representing and Manipulating Deformable Objects @ ICRA 2024
Workshop website: https://deformable-workshop.github.io...

Abstract: Manipulating garments and fabrics has long been a critical endeavor in the advancement of home-assistant robots, serving as a focal point for researchers in the fields of vision and robotics. However, due to complex dynamics and topological structures, garment manipulations pose significant challenges. Recent successes in reinforcement learning or vision-based methods offer promising avenues for learning garment manipulation. Nevertheless, these approaches are severely constrained by current benchmarks, which exhibit unrealistic simulation behavior and offer only a limited number of tasks. So we present UniGarment, a benchmark designed for deformable object and garment manipulation within realistic 3D indoor scenes. Our benchmark encompasses a diverse range of garment types, robotic systems and manipulators including dexterous hands. The multitude of tasks included in the benchmark enables further exploration of the interactions between garments, deformable objects, rigid bodies, fluids, and avatars. Furthermore, by incorporating multiple simulation methods, including FEM and PBD, along with our proposed sim-to-real techniques, we aim to significantly narrow the sim-to-real gap. We evaluate state-of-the-art vision methods, reinforcement learning (RL), and imitation learning (IL) techniques on these tasks, highlighting the challenges faced by current algorithms, notably their limited generalization capabilities. Our comprehensive analysis and provision of open-source environments lay the foundation for future research in garment manipulation, unlocking the full potential of these methodologies.

Extended abstract: https://deformable-workshop.github.io...