TITLE:
Bio-Inspired Physical Reservoir Computing Based on Organic Electrochemical Transistors for Temporal Pattern Recognition
AUTHORS:
Jingqi Wang, Dongsheng Zhang
KEYWORDS:
Organic Electrochemical Transistors (OECT), Physical Reservoir Computing, Neuromorphic Computing, PEDOT:PSS, MNIST Classification, Bio-Electronic
JOURNAL NAME:
Journal of Computer and Communications,
Vol.14 No.6,
June
18,
2026
ABSTRACT: The increasing demand for real-time and energy-efficient artificial intelligence (AI) processing exposes fundamental limitations of traditional von Neumann architectures. Neuromorphic computing, inspired by biological neural systems, offers a promising alternative by integrating memory and computation within dynamic physical substrates. Here, we demonstrate organic electrochemical transistors (OECTs) based on PEDOT:PSS as dynamic nodes for physical reservoir computing (PRC). OECTs exhibit coupled ionic-electronic transport, volumetric doping, and intrinsic short-term memory arising from ion diffusion, enabling nonlinear temporal signal transformation without recurrent circuitry. Flexible high-resolution OECT devices were fabricated via multilayer photolithography and characterized electrically. Under pulsed gating, the devices exhibit synaptic behaviors including paired-pulse facilitation, spike-timing-dependent plasticity, and frequency-dependent plasticity. Temporally encoded 4 × 4 MNIST digit images were applied as gate voltage sequences. The resulting drain current waveforms were analyzed to extract dynamic features and classified using a softmax regression model. The system achieved an average recognition accuracy of 78.6% without internal weight training. These results establish OECT-based PRC as a scalable, low-power, and bio-compatible neuromorphic hardware platform for temporal information processing.