航空学报 > 2023, Vol. 44 Issue (19): 28569-028569   doi: 10.7527/S1000-6893.2023.28569

类脑导航的机理、算法、实现与展望

朱祥维1, 沈丹2(), 肖凯1,3, 马岳鑫1, 廖祥4, 古富强5, 余芳文6, 高柯夫7, 刘经南7   

  1. 1.中山大学 电子与通信工程学院,深圳 518107
    2.中山大学 系统科学与工程学院,广州 510006
    3.中国人民解放军战略支援部队信息工程大学 地理空间信息学院,郑州 450001
    4.重庆大学 医学院,重庆 400030
    5.重庆大学 计算机学院,重庆 400044
    6.清华大学 精密仪器系,北京 100084
    7.武汉大学 卫星导航定位技术研究中心,武汉 430079
  • 收稿日期:2023-02-17 修回日期:2023-03-20 接受日期:2023-03-31 出版日期:2023-10-15 发布日期:2023-04-11
  • 通讯作者: 沈丹 E-mail:shend7@mail2.sysu.edu.cn
  • 基金资助:
    国家自然科学基金(61973328);教育部-中国移动科研基金(MCM2020-J-1);深圳市科技计划(GXWD20201231165807008)

Mechanisms, algorithms, implementation and perspectives of brain⁃inspired navigation

Xiangwei ZHU1, Dan SHEN2(), Kai XIAO1,3, Yuexin MA1, Xiang LIAO4, Fuqiang GU5, Fangwen YU6, Kefu GAO7, Jingnan LIU7   

  1. 1.School of Electronics and Communication Engineering,Sun Yat?sen University,Shenzhen 518107,China
    2.School of System Science and Engineering,Sun Yat?sen University,Guangzhou 510006,China
    3.Institute of Geospatial Information,PLA Strategic Support Force Information Engineering University,Zhengzhou 450001,China
    4.College of Medicine,Chongqing University,Chongqing 400030,China
    5.School of Computer Science,Chongqing University,Chongqing 400044,China
    6.Department of Precision Instruments,Tsinghua University,Beijing 100084,China
    7.Research Center of Satellite Navigation and Positioning Technology,Wuhan University,Wuhan 430079,China
  • Received:2023-02-17 Revised:2023-03-20 Accepted:2023-03-31 Online:2023-10-15 Published:2023-04-11
  • Contact: Dan SHEN E-mail:shend7@mail2.sysu.edu.cn
  • Supported by:
    National Natural Science Foundation of China(61973328);Ministry of Education-China Mobile Scientific Research Fund(MCM2020-J-1);Shenzhen Science and Technology Program(GXWD20201231165807008)

摘要:

脑与神经科学近几十年来的快速发展,初步揭示了动物导航的神经机理。借鉴大脑神经结构及信息处理机制,研究类脑仿生智能导航系统,为复杂环境下低功耗高鲁棒的自主智能导航提供了新的灵感。在详细综述了动物空间导航神经机理基础上,概述探讨了当前机器人仿生类脑导航智能算法,从利用脑启发的神经网络方法处理导航信息实现智能导航的角度,归纳为吸引子神经网络、深度强化学习以及脉冲神经网络等3类算法。接着梳理了类脑导航的实现途径,包括仿生智能传感器和智能类脑处理器平台。最后归纳并展望了类脑导航发展趋势,包括进一步探索生物界导航的脑神经机理及其信息处理过程、细分类脑导航概念内涵、评价指标完善的途径和统一实现框架。

关键词: 类脑导航, 导航神经机理, 导航细胞, 连续吸引子网络, 深度强化学习, 脉冲神经网络, 神经形态相机, 类脑芯片

Abstract:

The rapid development of brain and neuroscience in recent decades has initially revealed the neural mechanism of animal navigation. Drawing on the brain neural structures and information processing mechanisms, the study of brain-inspired intelligent navigation systems provides new inspiration for low-power, highly robust autonomous intelligent navigation in complex environments. Based on a detailed review of the neural mechanisms of animal spatial navigation, this paper then outlines and discusses current intelligent algorithms for robotic bionic brain-inspired navigation, which can be categorized into three types according to the three types of neural networks used to process navigation information for intelligent navigation: attractor neural networks, deep reinforcement learning, and spiking neural networks. Then, the ways for implementing brain-inspired navigation, including bionic intelligent sensors and neuromorphic processor platforms, are sorted out. Finally, the development trend of brain-inspired navigation is discussed, including further exploration of the brain neural mechanism of navigation in the biological world and its information processing process with low energy consumption and high robustness mechanism, subcategorization of the conceptual connotation of brain-inspired navigation, and the ways to improve the evaluation index and the unified implementation framework.

Key words: brain-inspired navigation, neural mechanism of navigation, navigation cells, continuous attractor network, deep reinforcement learning, spiking neural network, neuromorphic camera, brain-inspired chips

中图分类号: