Efficient End-to-End 6-Dof Grasp Detection Framework for Edge Devices with Hierarchical Heatmaps and Feature Propagation

TL;DR

This work tackles the challenge of real-time $6$-DoF grasp detection on edge devices by proposing E3GNet, an end-to-end framework that combines a Geometry-aware, lightweight encoder with a Global Location Heatmap FPN, Region Feature Propagation, and a Regional Rotation-Heatmap for precise pose estimation. The method achieves state-of-the-art performance on the GraspNet-1Billion dataset (average $52.18$ mAP) and demonstrates real-time inference on edge platforms, outperforming prior approaches in both speed and robustness. Real-world experiments with a UR-5e robot and a parallel-jaw gripper yield a $94\%$ grasp success rate across cluttered scenes, validating practical applicability. Overall, E3GNet enables efficient, scalable 6-DoF grasp detection suitable for embedded robotics and mobile platforms.

Abstract

6-DoF grasp detection is critically important for the advancement of intelligent embodied systems, as it provides feasible robot poses for object grasping. Various methods have been proposed to detect 6-DoF grasps through the extraction of 3D geometric features from RGBD or point cloud data. However, most of these approaches encounter challenges during real robot deployment due to their significant computational demands, which can be particularly problematic for mobile robot platforms, especially those reliant on edge computing devices. This paper presents an Efficient End-to-End Grasp Detection Network (E3GNet) for 6-DoF grasp detection utilizing hierarchical heatmap representations. E3GNet effectively identifies high-quality and diverse grasps in cluttered real-world environments.Benefiting from our end-to-end methodology and efficient network design, our approach surpasses previous methods in model inference efficiency and achieves real-time 6-Dof grasp detection on edge devices. Furthermore, real-world experiments validate the effectiveness of our method, achieving a satisfactory 94% object grasping success rate.

Figures (2)

• Figure 1: Overview of the proposed Efficient End-to-End 6-Dof Grasp Detection Framework.
• Figure 2: Real-world robot experiment settings.