Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Region-aware Grasp Framework with Normalized Grasp Space for Efficient 6-DoF Grasping

Siang Chen, Pengwei Xie, Wei Tang, Dingchang Hu, Yixiang Dai, Guijin Wang

TL;DR

Leveraging the NGS, it is found that CNNs are underestimated for 3D feature extraction and 6-DoF grasp detection in clutter scenes and build a highly efficient Region-aware Normalized Grasp Network (RNGNet).

Abstract

A series of region-based methods succeed in extracting regional features and enhancing grasp detection quality. However, faced with a cluttered scene with potential collision, the definition of the grasp-relevant region stays inconsistent, and the relationship between grasps and regional spaces remains incompletely investigated. In this paper, we propose Normalized Grasp Space (NGS) from a novel region-aware viewpoint, unifying the grasp representation within a normalized regional space and benefiting the generalizability of methods. Leveraging the NGS, we find that CNNs are underestimated for 3D feature extraction and 6-DoF grasp detection in clutter scenes and build a highly efficient Region-aware Normalized Grasp Network (RNGNet). Experiments on the public benchmark show that our method achieves significant >20% performance gains while attaining a real-time inference speed of approximately 50 FPS. Real-world cluttered scene clearance experiments underscore the effectiveness of our method. Further, human-to-robot handover and dynamic object grasping experiments demonstrate the potential of our proposed method for closed-loop grasping in dynamic scenarios.

Region-aware Grasp Framework with Normalized Grasp Space for Efficient 6-DoF Grasping

TL;DR

Leveraging the NGS, it is found that CNNs are underestimated for 3D feature extraction and 6-DoF grasp detection in clutter scenes and build a highly efficient Region-aware Normalized Grasp Network (RNGNet).

Abstract

A series of region-based methods succeed in extracting regional features and enhancing grasp detection quality. However, faced with a cluttered scene with potential collision, the definition of the grasp-relevant region stays inconsistent, and the relationship between grasps and regional spaces remains incompletely investigated. In this paper, we propose Normalized Grasp Space (NGS) from a novel region-aware viewpoint, unifying the grasp representation within a normalized regional space and benefiting the generalizability of methods. Leveraging the NGS, we find that CNNs are underestimated for 3D feature extraction and 6-DoF grasp detection in clutter scenes and build a highly efficient Region-aware Normalized Grasp Network (RNGNet). Experiments on the public benchmark show that our method achieves significant >20% performance gains while attaining a real-time inference speed of approximately 50 FPS. Real-world cluttered scene clearance experiments underscore the effectiveness of our method. Further, human-to-robot handover and dynamic object grasping experiments demonstrate the potential of our proposed method for closed-loop grasping in dynamic scenarios.
Paper Structure (12 sections, 7 equations, 12 figures, 12 tables, 1 algorithm)

This paper contains 12 sections, 7 equations, 12 figures, 12 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Normalized Grasp Space
  5. Regional Normalized Grasp Network
  6. Dataset Experiments
  7. Realworld Experiments
  8. Limitation and Conclusion
  9. Additional Results
  10. Implemetation Details
  11. Discussion
  12. Limitations

Figures (12)

  • Figure 1: Region-aware Grasp Framework. With an RGBD image input, Grasp Heatmap Model (GHM) chen2023efficient predicts grasp location heatmap. Subsequently, graspable patches are extracted through an adaptive grid generator, which samples mesh grids based on the location heatmap and the gripper scale. Local patches are then converted into Normalized Grasp Space and fed into the Regional Normalized Grasp Network to predict regional rotation heatmaps and grasps. Finally, we apply the inverse of the normalization process and obtain the scene-level grasps.
  • Figure 2: Depth-Adaptive Patch Extraction and grasp representation.
  • Figure 3: Visualization for the characteristics of Normalized Grasp Space.
  • Figure 3: Normalized Grasp Space ablation. Tested on Realsense Split.
  • Figure 4: Detailed structure of RNGNet.
  • ...and 7 more figures