| Work Detail |
A team at Tokyo University of Science has developed a solar cell-based computing device that mimics human synaptic behavior for use in AI processing. The researchers say the device can accelerate the development of energy-efficient AI sensors for applications such as health monitoring, surveillance, and automotive.
Scientists at Tokyo University of Science have fabricated a self-powered device based on dye-sensitized solar cells that mimics human synaptic behavior for artificial intelligence processing.
The research paper, “ Self-Powered Dye-Sensitized Solar-Cell-Based Synaptic Devices for Multi-Scale Time-Series Data Processing in Physical Reservoir Computing,” published in the journal ACS Applied Materials & Interfaces , explains that physical reservoir computing (RPC) using synaptic devices is a promising artificial intelligence device for predicting emergencies such as heart attacks, natural disasters, and pipeline failures.
The study details how the scientists fabricated a synaptic device based on dye-sensitized solar cells with time constants controllable by changing the light intensity. “To process time series of input optical data with different time scales, it is essential to fabricate devices consistent with the desired time scale,” explains Takashi Ikuno, one of the lead researchers. “Inspired by the eye afterimage phenomenon, we devised a novel optoelectronic human synaptic device that can serve as a computational framework for energy-efficient edge AI optical sensors.”
The device uses dyes based on squalium derivatives and incorporates optical input, AI computation, analog output, and material-level power delivery functions. It showed synaptic characteristics, such as paired-pulse facilitation and paired-pulse depression, in response to light intensity.
When the device was used as a fallback layer for the PRC, it classified human motions such as crouching, jumping, running and walking with over 90% accuracy. The scientists found that performance on a temporal processing task was improved by varying the intensity of the light, even when the input pulse width was varied.
Thanks to the solar cell, its energy consumption was 1% of that required by conventional systems. “We have demonstrated for the first time in the world that the developed device can operate with very low energy consumption and yet identify human movement with a high rate of accuracy,” Ikuno said, noting that the proposed innovation could revolutionize low-power AI applications in healthcare monitoring, surveillance and automotive technologies.
“This invention can be used as a mass-use edge AI optical sensor that can be attached to any object or person, and can impact the cost of energy consumption, such as car-mounted cameras and car-mounted computers,” Ikuno explained. “This device can function as a sensor capable of identifying human motion with low power consumption, so it has the potential to contribute to improving vehicle energy consumption.”
The system could also be used as a low-power optical sensor in smartwatches and autonomous medical devices, significantly reducing their costs to make them comparable or even lower than those of current medical devices. |
