What is it about?
The rapid expansion of AI in edge applications, such as AIoT, robotics, and electric vehicles, has enabled a new era of context-aware devices. Yet, the computationally demanding and power-hungry nature of AI poses major hurdles for practical implementations. To address this, this paper introduces a low-power, mixed-signal in-memory computing architecture driven by a RISC-V controller, tailored to accelerate AI tasks directly at the edge.
Featured Image
Photo by Immo Wegmann on Unsplash
Why is it important?
Our work revolutionizes edge AI by combining analog in-memory computing with an open-source RISC-V controller. It solves the severe energy disparity of traditional computing by performing signed matrix multiplications directly within the memory array. This mixed-signal approach shrinks power consumption while enabling highly accurate, single-cycle computations for smart, battery-powered devices.
Perspectives
Writing this article has been a great privilege, providing the opportunity to collaborate with a team of highly competent and innovative researchers. More importantly, it served as a fond reminder that back in 1988 at Imperial College, I was the last person to use a massive analog computer to evaluate the accuracy of various transfer function estimators.
Professor W.M. To
Macao Polytechnic University
Read the Original
This page is a summary of: A Low-Power Mixed-Signal Differential In-Memory Matrix–Vector Computing Circuit Architecture with RISC-V Control for Edge AI, Journal of Low Power Electronics and Applications, June 2026, MDPI AG,
DOI: 10.3390/jlpea16030022.
You can read the full text:
Contributors
The following have contributed to this page
