Progress of Artificial Olfactory Systems in Computing Architectures: From Von Neumann to Neuromorphic Computing

Abstract

ABSTRACT Artificial olfactory systems (AOS) are emerging as a pivotal technology for intelligent odor detection, drawing inspiration from the remarkable sensitivity and adaptability of the biological olfactory system. This review provides a comprehensive summary of recent advances in AOS, from traditional designs based on von Neumann architectures to cutting‐edge neuromorphic platforms that closely integrate sensing, memory, and computation. We introduce a systematic classification of AOS according to their computational architecture: (1) von Neumann‐based systems that combine sensor arrays and advanced pattern recognition algorithms for odor classification and multi‐modal analysis; (2) integrated perception‐memory devices that emulate synaptic plasticity, enabling adaptive response and memory for complex olfactory environments; and (3) neuromorphic olfactory systems featuring near‐sensor and in‐sensor computing, leveraging memristor arrays and monolithic integration for real‐time, energy‐efficient, and robust odor processing. This review focuses on the recent progress, working mechanisms, and advanced features of AOS from traditional von Neumann architectures to neuromorphic computing. Finally, we offer a critical outlook on overcoming systemic bottlenecks, and propose a strategic roadmap for realizing robust, autonomous olfactory intelligence.