Tokyo scientists develop a self-powered solar device mimicking synaptic behaviour to be employed for edge AI processing.
In a latest development, researchers at Tokyo University of Science (TUS) have developed a self-powered dye-sensitised solar cell-based device that mimics human synaptic behaviour. This novel invention, inspired by the eye’s afterimage phenomenon, promises to enhance edge AI processing by tackling time-series data with unparalleled efficiency.
Physical reservoir computing (PRC) is a cutting-edge framework designed to process time-series data, crucial for real-time tasks like monitoring infrastructure and health systems. While existing PRC systems have shown potential, they struggle to handle data across multiple timescales. The TUS team, led by associate professor Takashi Ikuno, addressed this limitation by introducing a device with light intensity-controllable time constants, enabling high-performance data classification.
This innovation is expected to cater to a wide audience, including industries relying on surveillance systems, wearable technology developers, and automotive manufacturers. Its potential applications in car cameras, health monitors, and standalone devices align with the increasing demand for energy-efficient, high-accuracy AI solutions.
Dr. Ikuno explained, “Inspired by the afterimage phenomenon of the eye, we developed an optoelectronic human synaptic device that serves as a power-saving computational framework for edge AI sensors.” The device integrates solar-powered AI computation and analog output at the material level, achieving remarkable energy efficiency.
The device’s squarylium derivative-based dyes exhibit synaptic plasticity, enabling features like paired-pulse facilitation and depression. This adaptability enhances its effectiveness in tasks such as human motion classification, achieving over 90% accuracy while consuming just 1% of the energy required by conventional systems.
By merging sustainability with cutting-edge technology, this innovation addresses the needs of industries striving for low-power solutions in AI. As Dr. Ikuno highlighted, “This device can operate at low power while delivering high accuracy, making it ideal for smartwatches, car-mounted cameras, and medical devices.”
This development represents a progressive growth in AI, setting standards for sustainable, and versatile edge computing solutions.
