Covalent Organic Framework‐Based Photoelectric Memristive Neurons With Unambiguous Encoding Capacity for Advanced Driver Assistance Systems

ABSTRACT Advanced Driver Assistance Systems (ADAS), as an essential component of modern intelligent driving, play a crucial role in enhancing situational awareness and optimizing decision‐making. Nevertheless, due to their inherent limitations, such as modular optimization and the von Neumann architecture, these systems encounter challenges regarding large energy consumption, high data latency, and compromised information security. Based on the biological sensory neural network, developing an intelligent recognition system featuring multi‐stimuli sensing, unambiguous encoding, and neuromorphic computing functions is an effective approach to tackle these challenges. Herein, using their tunable structures and properties, we fabricated wafer‐scale two‐dimensional (2D) covalent organic framework (COF) films with photoelectric response for high‐performance threshold switching memristors (TSMs) and artificial visual neurons, applied to ADAS traffic sign recognition. The fabricated TSM shows high yield (97%), low operation voltage (0.4 V), fast response (100 ns), and low energy consumption (8.62 fJ). By integrating a green light sensor, the TSM can be applied for constructing the visual neuron, enabling unambiguous encoding of blue and green light. These fused spikes can be processed by a 3‐layer spiking neural network to identify traffic signs with a high accuracy of 96.99%, which demonstrates the great potential of COF‐based memristors in future ADAS applications.