A Segmented Robot Grasping Perception Neural Network for Edge AI

TL;DR

This paper presents a hardware-optimized neural network for robotic grasp detection that runs entirely on low-power edge devices, enabling real-time autonomous manipulation.

Contribution

It introduces a heatmap-guided grasp detection neural network optimized for edge AI deployment on RISC-V chips, demonstrating fully on-chip inference.

Findings

Successful on-chip inference on GAP9 RISC-V SoC

Achieved real-time grasp detection with low power consumption

Validated on GraspNet-1Billion benchmark

Abstract

Robotic grasping, the ability of robots to reliably secure and manipulate objects of varying shapes, sizes and orientations, is a complex task that requires precise perception and control. Deep neural networks have shown remarkable success in grasp synthesis by learning rich and abstract representations of objects. When deployed at the edge, these models can enable low-latency, low-power inference, making real-time grasping feasible in resource-constrained environments. This work implements Heatmap-Guided Grasp Detection, an end-to-end framework for the detection of 6-Dof grasp poses, on the GAP9 RISC-V System-on-Chip. The model is optimised using hardware-aware techniques, including input dimensionality reduction, model partitioning, and quantisation. Experimental evaluation on the GraspNet-1Billion benchmark validates the feasibility of fully on-chip inference, highlighting the potential of low-power MCUs for real-time, autonomous manipulation.