Bioinspired Ultralow‐Power Flexible Electro‐Optically Configurable Synaptic Transistors: An Integrated Perception‐Processing‐Tracking Visual Chain

Abstract

Biological visual tracking closely parallels the autonomous navigation that requires motion recognition, typically progressing through a cascade of optical perception, adaptive attention, and real‐time tracking. However, most efforts focus on low‐level perception and decoupled algorithms, while overlooking the crucial intermediate mechanism of in situ background suppression for salient target separation. Here, inspired by the Nile tilapia visual system, a perception‐processing‐tracking visual chain architecture based on flexible, electro‐optically configurable synaptic transistors is first proposed. Leveraging the sharp absorption edge of quantum dots (QDs), the device achieves prominent RGB‐NIR band recognition. Notably, the introduction of a phosphoric acid‐mediation strategy significantly enhances the stability of blue‐emitting CsPbCl 3 QDs while preserving their excellent optical performance. Combined with a poly(vinylidene fluoride‐ co ‐hexafluoropropylene) (PVDF‐HFP) dielectric layer, the system enables efficient cross‐modal optoelectronic signal processing. Electric field tuning of photocarrier tunneling assigns an attention score as high as 95.88% to target regions and merely 3.79% to backgrounds, with the score gap widening over time. Assisted by reservoir computing (RC), the device classifies six target motion trajectories with 98.89% accuracy, operating at an ultra‐low energy cost of 0.011 fJ per synaptic event with exceptional mechanical flexibility. These attributes open up promising prospects for future autonomous driving applications.