Deformable Cluster Manipulation Via Whole-Arm Policy Learning
Jayadeep Jacob, Wenzheng Zhang, Houston Warren, Paulo Vinicius Koerich Borges, Fabio Ramos, Tirthankar Bandyopadhyay
AI summary
Problem
Manipulating clusters of deformable objects is difficult due to high-dimensional state spaces, noisy perception, and the need for contact-rich whole-arm interactions rather than just end-effector control.
Approach
A model-free RL framework integrating 3D point clouds—encoded via efficient distributional kernel mean embeddings—and proprioceptive touch indicators, guided by a novel context-agnostic occlusion reward heuristic.
Key results
- Efficient distributional state representation enabling real-time inference and faster training
- Context-agnostic occlusion heuristic generalisable to various exposure applications
- Generation of creative strategies leveraging multiple arm links for de-occlusion
- Zero-shot sim-to-real transfer for clearing real branches with unknown topology and dynamics
Why it matters
Enables autonomous robotic clearance of hazardous vegetation around power lines and improves agricultural fruit exposure, reducing human risk.
Abstract
Manipulating clusters of deformable objects presents a substantial challenge with widespread applicability, but requires contact-rich whole-arm interactions. A potential solution must address the limited capacity for realistic model synthesis, high uncertainty in perception, and the lack of efficient spatial abstractions, among others. We propose a novel framework for learning model-free policies integrating two modalities: 3D point clouds and proprioceptive touch indicators, emphasising manipulation with full body contact awareness, going beyond traditional end-effector modes. Our reinforcement learning framework leverages a distributional state representation, aided by kernel mean embeddings, to achieve improved training efficiency and real-time inference. Furthermore, we propose a novel context-agnostic occlusion heuristic to clear deformables from a target region for exposure tasks. We deploy the framework in a power line clearance scenario and observe that the agent generates creative strategies leveraging multiple arm links for de-occlusion. Finally, we perform zero-shot sim-to-real policy transfer, allowing the arm to clear real branches with unknown occlusion patterns, unseen topology, and uncertain dynamics. Website: https://sites.google.com/view/dcmwap/