Answers are grounded in this paper and cite their sources.
A key limitation of current artificial visual systems lies in their single-channel signal transduction, which fundamentally departs from the parallel processing architecture of the biological retinas enabled by rod and cone cells. To bridge this gap, we developed a retinomorphic vision device by ion-electron coupling transport based on carbon nanotube (CNT)/zinc oxide (ZnO) heterostructured nanotubes. Upon optical stimuli, this device generates two distinct ionic and electronic current signals, effectively mimicking the parallel phototransduction pathways in biological vision. This dual-channel property enables efficient machine vision perception, achieving luminance adaptation and color discrimination accuracies of 96.5% and 98.2%, respectively. Furthermore, traffic-light-guided closed-loop navigation of a mobile vehicle in a real scene has verified its application potential in embodied intelligence. This "single device enabled, distinguishable dual-channel readouts" principle opens a new avenue toward artificial vision systems.
No takes yet. Share an insight, caveat, or question.
Feng et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: