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

2024-10-30

arXiv.org

AI Summary

To address the insufficient real-time performance of 6-DoF grasp detection on edge devices, this paper proposes E3GNet, a lightweight end-to-end framework. Its core contributions are: (1) a novel hierarchical heatmap representation that jointly encodes grasp position, orientation, and size; (2) a cross-scale feature propagation mechanism integrating top-down semantic enhancement with bottom-up local geometric constraints; and (3) a lightweight RGB-D CNN backbone coupled with a multi-level heatmap regression head. Evaluated on a Jetson AGX Orin, E3GNet achieves >15 FPS real-time inference while attaining a 94% grasp success rate in cluttered scenes—surpassing state-of-the-art methods. Crucially, it reduces computational overhead by 42% without compromising grasp diversity or 6-DoF pose accuracy.

Technology Category

• Computer Vision: Biometrics, Face, Gesture & Pose
• Intelligent Robots: Manipulation
• Machine Learning: Learning on the Edge & Model Compression

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments

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.

Problem

• Develop efficient 6-DoF grasp detection for edge devices.
• Overcome computational challenges in real robot deployment.
• Achieve real-time grasp detection in cluttered environments.

Innovation

• Hierarchical heatmap representations for grasp detection
• End-to-end network design for real-time efficiency
• Optimized for edge devices with high success rate