Answers are grounded in this paper and cite their sources.
Edge smart gas-sensing systems process data closer to where they are generated. They face hardware and software challenges due to quasi-static sensory responses; physical separation of sensors, processors, and memory; and tedious training algorithms. We present gallium nitride (GaN)-based high-electron-mobility transistors (HEMTs) for in-sensor reservoir computing (RC) to learn and recognize complex gas patterns at the edge. The Pt-AlGaN/GaN HEMT features dynamic responses to encode gas patterns through dipole generation/annihilation and their impact on two-dimensional electron gas (2-DEG). These responses can be modulated by the sensor's gate bias (V G ) to adapt to different gas dynamics. The devices exhibited reliable electrical behavior and repeatable responses to gas across a wide range of temperatures, which were also theoretically modeled. We also developed an RC system for learning complex gas patterns, which outperforms and reduces training costs compared with existing systems.
No takes yet. Share an insight, caveat, or question.
Jiang et al. (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: