Application of memristors for efficient neuromorphic computing in tactile sensing
TL;DR
The physical mechanisms of mainstream memristors are discussed, recent progress in memristor-based neuromorphic computing for tactile sensing is highlighted, and key challenges and future directions for neuromorphic tactile perception systems are outlined.
Abstract
With the rapid development of the Internet of Things (IoT) and wearable electronics, tactile sensors play an indispensable role in intelligent sensing systems. However, traditional tactile sensing systems follow the von Neumann architecture, where sensors and processing units are physically separated. This leads to frequent data transfer of large raw data volumes, causing high latency and energy consumption. Such bottlenecks cannot meet the requirements of real-time closed-loop control and edge intelligence. Inspired by the highly integrated "perception-storage-computation" mechanism of biological sensory systems, memristor-based neuromorphic computing offers a groundbreaking solution beyond conventional approaches. Memristors combine non-volatile storage with tunable resistance. They enable in-situ emulation of synaptic plasticity, in-memory computing, and brain-inspired processing, thereby holding the potential to significantly improve the energy efficiency and response speed of tactile systems. This review systematically discusses the physical mechanisms of mainstream memristors, highlights recent progress in memristor-based neuromorphic computing for tactile sensing, and outlines key challenges and future directions for neuromorphic tactile perception systems.