[공지사항]을 빙자한 안부와 근황 
Show more

Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

A segmented robot grasping perception neural network for edge AI

Created by
  • Haebom

Author

Casper Brocheler, Thomas Vroom, Derrick Timmermans, Alan van den Akker, Guangzhi Tang, Charalampos S. Kouzinopoulos, Rico Mockel

Outline

This paper presents an end-to-end framework for Heatmap-Guided Grasp Detection (6-DoF grip pose detection) that enables low-power and low-latency inference in edge environments to solve the robotic gripping problem of stably grasping and manipulating objects of various shapes, sizes, and orientations, on GAP9 RISC-V SoC. The model is optimized using hardware-aware techniques such as input dimensionality reduction, model partitioning, and quantization, and the feasibility of full on-chip inference is verified through the GraspNet-1Billion benchmark, demonstrating the potential of real-time autonomous manipulation using low-power MCUs.

Takeaways, Limitations

•
Takeaways:
◦
We present the possibility of real-time robot gripping using low-power MCUs.
◦
We present an effective method for efficient 6-DoF grip pose detection on edge devices.
◦
Contributes to the implementation of energy-efficient robotic systems through hardware-aware optimization techniques.
•
Limitations:
◦
These experimental results are limited to a specific hardware platform called GAP9 SoC. Further research is needed to determine generalizability to other platforms.
◦
Only performance evaluation on the GraspNet-1Billion dataset is presented, and generalization performance on other datasets or real-world environments is not verified.
◦
There is a lack of detailed analysis of the optimization level and limitations of the hardware recognition techniques used.
👍
Made with Slashpage