A Self-Powered Synaptic Device Based on InGaO Nanowires for Humanoid Robot Learning

Abstract

Inspired by the human brain, self-powered synaptic devices hold substantial promise in the fields of storage, learning, and computation, hence qualifying as indispensable constituents for building neuromorphic computing systems. In this work, a self-powered synaptic device based on InGaO nanowires is proposed and demonstrated successfully. By excitation of deep ultraviolet (DUV) light, this synaptic device can simulate the double-spike promotion, spike timing plasticity, and memory learning ability of biological synapses. Among them, the incident light, electrodes, and photogenerated carriers correspond to the action potentials, pre/postsynaptic membranes, and neurotransmitters of biological synapses, respectively. With an ultrahigh paired-pulse facilitation index of 185%, the memristor synapse shows an excellent learning performance under self-powered conditions. Moreover, the application potential of the self-powered artificial synaptic device is demonstrated by the successful manipulation of a humanoid intelligent robot. The control commands coming from the self-powered memristor synapse can drive the humanoid robot to perform the corresponding actions, which shows a unique “learning–forgetting–relearning” ability. In an artificial neural network, the synaptic device displays the ability in effective image denoising and a high image recognition accuracy surpassing 93%, indicating its robust learning and cognitive potential. Therefore, this study not only demonstrates the great potential of nanowire-based synaptic devices in the field of intelligent robotics but also opens a fresh avenue for the development of neuromorphic computing technologies and artificial intelligence systems requiring ultralow energy consumption.