引用本文:刘晓德,郭宇飞,黄旭辉,马 喆.基于脉冲神经网络的智能控制研究进展[J].控制理论与应用,2024,41(12):2189~2206.[点击复制]
LIU Xiao-de,GUO Yu-fei,HUANG Xu-hui,MA Zhe.Research advance in intelligent control based on spiking neural networks[J].Control Theory and Technology,2024,41(12):2189~2206.[点击复制]
基于脉冲神经网络的智能控制研究进展
Research advance in intelligent control based on spiking neural networks
摘要点击 4707  全文点击 162  投稿时间：2023-05-17  修订日期：2024-09-13
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2024.30330
  2024,41(12):2189-2206
中文关键词  脉冲神经网络, 深度学习, 神经网络与智能控制, 神经形态计算
英文关键词  spiking neural networks (SNN)  deep learning  neural network and intelligent control  neuromorphic computing
基金项目  国家自然科学基金项目(12202413, 12202412)资助.
作者单位E-mail
刘晓德 航天科工集团智能科技研究院有限公司 lxde@pku.edu.cn 
郭宇飞 航天科工集团智能科技研究院有限公司  
黄旭辉 航天科工集团智能科技研究院有限公司  
马 喆* 航天科工集团智能科技研究院有限公司 jackson1952@126.com 
中文摘要
      近些年, 具备低功耗、高鲁棒、融合时空信息等优势的脉冲神经网络(SNN)在类脑研究与智能控制的交叉领域方兴未艾. 基于脉冲神经网络架构的智能控制方法是实现与环境自主交互并且高能效完成复杂控制任务的有效途径之一. 为此, 本文首先介绍了SNN的基本要素与研究动机; 然后, 详细介绍了近年来基于脉冲神经网络智能控制的研究进展以及在机器人、无人车、无人机等领域的应用情况; 接着, 总结了一些现有的硬件平台, 用以实现SNN算法的高效能实现; 最后, 总结展望了SNN控制发展的机遇与挑战. 本文旨在梳理出SNN控制发展的技术脉络, 为其快速发展提供借鉴与思路.
英文摘要
      In recent years, spiking neural networks (SNN) have garnered significant attention in the fields of braininspired learning and intelligent control due to their advantages in energy efficiency, robustness, and the ability to incorporate spatial-temporal information. In the field of brain-inspired learning and intelligent control, SNN architectures have shown promise in achieving complex control tasks with autonomous interaction with variations in the environment. This paper presents a comprehensive review of the development of intelligent control based on SNN and systematically summarizes relevant SNN control applications. Firstly, the basic concept of SNN, as well as the motivations and advantages of intelligent control based on SNN, is briefly introduced. Subsequently, the research progress of intelligent control based on SNN in recent years and its applications in fields such as robotics, unmanned vehicles, and unmanned aerial vehicles are systematically reviewed. Additionally, we summarize some hardware platforms that enable efficient implementation of SNN algorithms. Finally, the opportunities and challenges associated of SNN control are discussed. The purpose of this paper is to provide a technical framework for intelligent control based on SNN approach, and facilitate its rapid development and application.