中国空间科学技术

中国空间科学技术 ›› 2025, Vol. 45 ›› Issue (1): 113-123.doi: 10.16708/j.cnki.1000-758X.2025.0011

• 论文 • 上一篇    

基于位置细胞与网格细胞信息融合的类脑导航方法

刘晨1，熊智1,2,*，华冰3，张玲1，杨闯1，邹伟全1   

  1. 1.南京航空航天大学 自动化学院，南京211106
    2.先进飞行器导航、控制与健康管理工业和信息化部重点实验室，南京211106
    3.南京航空航天大学 航天学院，南京211106
  • 收稿日期:2023-04-18 修回日期:2023-11-07 录用日期:2023-11-09 发布日期:2025-01-24 出版日期:2025-02-01

Brain-inspired navigation method based on place cell and grid cell information fusion

LIU Chen1，XIONG Zhi1,2,*，HUA Bing3，ZHANG Ling1，YANG Chuang1，ZOU Weiquan1   

  1. 1.College of Automation Engineering,Nanjing University of Aeronautics and Astronautics，Nanjing 211106,China
    2.Key Laboratory of Navigation,Control and Health-Management Technologies of Advanced Aerocraft,Ministry of 
    Industry and Information Technology,Nanjing 211106,China
    3.College of Astronautics,Nanjing University of Aeronautics and Astronautics，Nanjing 211106,China
  • Received:2023-04-18 Revision received:2023-11-07 Accepted:2023-11-09 Online:2025-01-24 Published:2025-02-01

PDF (PC)

7

摘要： 哺乳动物具备根据大脑中的导航细胞感知外源性信息以及自运动信息进行自身定位与导航的能力，这为发展智能性高、自适应性强的无人机导航方法提供了较好的生物模型。对关键导航细胞神经机理以及信息融合建模方法进行研究，针对传统基于视觉的类脑认知地图构建方法回环检测实时性不高，检测点稀疏的问题，提出了基于位置细胞与网格细胞信息融合的类脑导航方法。首先，利用连续吸引子神经网络以及各向同性高斯网络分别对网格细胞以及位置细胞进行建模，实现了路径积分以及位置测量，在此基础上提出网格细胞置零算法提高了网格细胞大范围计算效率；其次，通过赫布学习规则得到两种细胞网络的连接权值矩阵，实现了位置细胞对网格细胞路径积分的实时校正过程；最后，基于神经元群体矢量加权平均以及网格细胞顶点位置处理解码方法得到无人机的三维位置。实验结果表明：所提方法能够在大范围空间内对无人机三维位置进行精确解算，相较于传统导航定位算法精度有所提升，进一步拓展了无人机三维类脑导航方法的应用范围。

关键词: 类脑导航, 位置细胞, 网格细胞, 赫布学习, 吸引子神经网络

Abstract: Mammals have the ability to sense exogenous information and self-motion information according to the navigation cells in the brain for self-positioning and navigation,which provides a good biological model for the development of intelligent and adaptive UAV(unmanned aerial vehicle) navigation methods.In this paper,we study the neural mechanism of key navigation cells and the information fusion modelling method,and propose a brain-inspired navigation method based on place cell and grid cell information fusion to address the problems of low real-time loop detection and sparse detection points in the traditional vision-based brain-inspired cognitive map construction method.Firstly,we use continuous attractor neural network and isotropic Gaussian network to model grid cell and position cell respectively,and achieve path integration and position measurement,based on which we propose grid cell zeroing algorithm to improve the efficiency of grid cell wide range calculation.Secondly,the connection weight matrix of the two cell networks is obtained by Hebb-learning rule,which realises the real-time correction process of the position cell to the grid cell path integral.Finally,the 3D position of the UAV is obtained based on the neuron population vector weighted average as well as the grid cell vertex position processing decoding method.The experimental results show that the method proposed in this paper can accurately decode the three-dimensional position of the UAV in a wide range of space,which improves the accuracy compared with the traditional navigation and positioning algorithms,and further expands the application scope of the three-dimensional brain-inspired navigation method for UAVs.

Key words: brain-inspired navigation, place cell, grid cell, hebb learning, attractor neural network