Deep Reinforcement Learning Based on Local GNN for Goal-conditioned Deformable Object Rearranging

TL;DR

This paper introduces a local GNN-based learning framework for deformable object rearranging that effectively transfers from simulation to real robots, improving generalization and manipulation capabilities.

Contribution

The novel use of local GNNs with attention mechanisms for encoding keypoints enables better simulation-to-real transfer in deformable object manipulation.

Findings

Effective in multiple simulation tasks involving ropes and cloths

Successfully transferred to real robot with fine-tuning

Outperforms previous CNN-based approaches in transferability

Abstract

Object rearranging is one of the most common deformable manipulation tasks, where the robot needs to rearrange a deformable object into a goal configuration. Previous studies focus on designing an expert system for each specific task by model-based or data-driven approaches and the application scenarios are therefore limited. Some research has been attempting to design a general framework to obtain more advanced manipulation capabilities for deformable rearranging tasks, with lots of progress achieved in simulation. However, transferring from simulation to reality is difficult due to the limitation of the end-to-end CNN architecture. To address these challenges, we design a local GNN (Graph Neural Network) based learning method, which utilizes two representation graphs to encode keypoints detected from images. Self-attention is applied for graph updating and cross-attention is applied for generating manipulation actions. Extensive experiments have been conducted to demonstrate that our framework is effective in multiple 1-D (rope, rope ring) and 2-D (cloth) rearranging tasks in simulation and can be easily transferred to a real robot by fine-tuning a keypoint detector.