HADRO-SoC: Reinventing AI Acceleration with Neuromorphic, In-Memory, and Graph-Aware Chip Design
Year: 2024
Type: article
Abstract: This paper presents a novel SoC architecture tailored for implementing Transformer-GNN-based AI models across domains such as Earth-based smart grids, spacecraft, UAVs, and commercial aviation. The proposed chip integrates recent hardware design strategies including In-Memory Computing (IMC) [3], Neuromorphic Coprocessing [5], and NoC-based modularity [8] to address latency, power, and domain adaptation challenges. Our contribution fills hardware-software integration gaps identified in 20 IEEE chip design papers and introduces a patentable blueprint for unified edge-AI deployment [1]–[20].
Author: ZAIDI, ADNAN HAIDER
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6893427717
FWCI:
Cites:
Cited by:
Related to: