Researchers at the Institut des Matériaux, de Microélectronique et des Nanosciences de Provence, in collaboration with scientists from the Centre de Nanosciences et de Nanotechnologies, the Commissariat à l'énergie atomique et aux énergies alternatives and the Institut Photovoltaïque d'Ile-de-France, have demonstrated that memristors can be used to manufacture a solar-powered AI system that operates reliably even when the energy supply is low.
Reading time: 4 minutes
|
Things to remember:
|
High AI power consumption poses a problem
Artificial intelligence (AI) is widely used in various embedded applications such as patient monitoring. To guarantee safety and minimize energy consumption due to communication, it would be preferable to process data directly in these embedded systems. However, the high energy consumption of AIs is an obstacle to their deployment in environments requiring autonomy.
A promising solution to this problem is the design of systems based on memristors, electronic components that can be electrically programmed to store information by modulating their resistance. The use of these memristors can considerably reduce AI energy consumption, making it even conceivable to create self-powered AI systems by directly collecting energy from their environment, thus enabling the design of autonomous AIs, not requiring batteries.
The challenge: maintaining supply voltage stability
Most memristor-based AI circuits are based on an analog memory computing concept, exploiting the classical laws of electricity (Ohm's and Kirchhoff's laws) to perform the fundamental operation of neural networks, namely multiplication and accumulation (MAC). This concept is difficult to put into practice due to the high variability of memristors, the imperfections of CMOS analog circuits and the effects of supply voltage variation.
To overcome these difficulties, memristor-based integrated AI systems use complex peripheral circuits, which are tuned for a particular supply voltage. This requirement for supply voltage stability is in direct contradiction with the properties of energy collectors such as miniature solar cells, which provide fluctuating voltage and energy, a major obstacle to the realization of memristor-based self-powered AI.
An alternative approach resistant to power fluctuations
Researchers from the Institut des Matériaux, de Microélectronique et des Nanosciences de Provence (IM2NP, CNRS / Aix-Marseille Université), in collaboration with scientists from the Centre de Nanosciences et de Nanotechnologies (C2N, CNRS / Université Paris-Saclay), the Commissariat à l'énergie atomique et aux énergies alternatives (CEA-Leti) and the Institut Photovoltaïque d'Ile-de-France (IPVF) have designed a binarized neural network, fabricated in a CMOS/memristor hybrid process, with an alternative approach that is particularly resistant to power supply fluctuations. This robustness was illustrated by powering the circuit with a miniature broadband solar cell, optimized for indoor applications.
Remarkably, the circuit remains functional even in low-light conditions equivalent to 0.08 times the average solar flux, suffering only a modest drop in neural network accuracy. When energy availability is limited, the circuit seamlessly switches from precise to approximate calculation.
These results pave the way for the deployment of artificial intelligence in energy-independent embedded systems.
Contact à ajouter
Informations complémentaires
Reference: Jebali, F., Majumdar, A., Turck, C. et al. Powering AI at the edge: A robust, memristor-based binarized neural network with near-memory computing and miniaturized solar cell. Nat Commun 15, 741 (2024).
This publication is available in the arXiv open archive.
Article published on May 31, 2024.
Discover also
Share on social media or
