类脑导航中基于差分Hebbian学习的网格细胞构建模型

doi:10.3969/j.issn.1001-506X.2020.03.023

系统工程与电子技术 ›› 2020, Vol. 42 ›› Issue (3): 674-679.doi: 10.3969/j.issn.1001-506X.2020.03.023

类脑导航中基于差分Hebbian学习的网格细胞构建模型

韩昆(), 吴德伟(), 来磊()

空军工程大学信息与导航学院, 陕西西安 710077

收稿日期:2019-07-26 出版日期:2020-03-01 发布日期:2020-02-28
作者简介:韩昆 (1990-),男,博士研究生,主要研究方向为智能自主导航理论与技术。E-mail: hk199009@126.com|吴德伟 (1963-),男,教授,博士,主要研究方向为智能导航与协同控制。E-mail: wudewei74609@126.com|来磊 (1983-),男,讲师,博士,主要研究方向为智能导航与协同控制。E-mail: lailei@mail.nwpu.edu.cn
基金资助:
国家自然科学基金(61603409);博士后科学基金(2017M623352);博士后科学基金(2018T111148)

Model of generating grid cell based on difference Hebbian learning in brain-inspired navigation

Kun HAN(), Dewei WU(), Lei LAI()

Information and Navigation College, Air Force Engineering University, Xi'an 710077, China

Received:2019-07-26 Online:2020-03-01 Published:2020-02-28
Supported by:
国家自然科学基金(61603409);博士后科学基金(2017M623352);博士后科学基金(2018T111148)

摘要/Abstract

摘要：

网格细胞是动物大脑中与空间认知和导航有关的重要神经元,具有拓展至整个空间的六边形放电野。位置细胞是网格细胞重要的信息源,可以通过Hebbian学习生成网格细胞,但是现有模型对Hebbian学习的脉冲自适应函数或学习窗函数做了预先的墨西哥帽模型假设。针对该问题提出一种基于差分Hebbian学习的位置细胞至网格细胞模型,利用细胞放电率的变化自发产生墨西哥帽模型的输入关联,然后通过对位置细胞至网格细胞的突触权重进行竞争性非线性限制,生成具有六边形放电野分布的网格细胞。仿真结果表明,该模型可以为无人运行体类脑导航系统的构建提供借鉴。

关键词: 类脑导航, 网格细胞, 位置细胞, 差分Hebbian学习, 竞争性非线性限制

Abstract:

The grid cell is a kind of important neuron cell related to spatial cognition and navigation in animal brain. It has hexagonal firing field extending to the whole two-dimensional space. The place cell is one of the main information sources of grid cells. The place cells can generate the grid cell through Hebbian learning, but the existing learning methods have assumed the Mexican hat model in advance for the adaptive function or learning window function of Hebbian learning. A place-to-grid cell model based on difference Hebbian learning is proposed. The input correlation with the Mexican hat model is generated spontaneously by using the difference of cells firing rates, and then the grid cell with the hexagonal firing field is generated through competitive nonlinear constraint of synaptic weights from place cells to the grid cell. The simulation results show that the difference Hebbian learning model can provide reference for the construction of the brain-inspired navigation system of unmanned platform.

Key words: brain-inspired navigation, grid cell, place cell, difference Hebbian learning, competitive nonlinear constraint

中图分类号:

TN96
Q811

韩昆, 吴德伟, 来磊. 类脑导航中基于差分Hebbian学习的网格细胞构建模型[J]. 系统工程与电子技术, 2020, 42(3): 674-679.

Kun HAN, Dewei WU, Lei LAI. Model of generating grid cell based on difference Hebbian learning in brain-inspired navigation[J]. Systems Engineering and Electronics, 2020, 42(3): 674-679.

图/表 5

表1

图1

图2

图3

图4

参考文献 28

1	MYROSHNYCHENKO M , SEAMANS J K , PHILLIPS A G , et al. Temporal dynamics of hippocampal and medial prefrontal cortex interactions during the delay period of a working memory-guided foraging task[J]. Cerebral Cortex, 2017, 27 (11): 1- 12.
2	EPSTEIN R A , PATAI E Z , JULIAN J B , et al. The cognitive map in humans: spatial navigation and beyond[J]. Nature Neuroscience, 2017, 20 (11): 1504- 1513. doi: 10.1038/nn.4656
3	王琳, 王亮. 认知地图的神经环路基础[J]. 生物化学与生物物理进展, 2017, 44 (3): 187- 197.
	WANG L , WANG L . Neural circuit basis of cognitive map[J]. Progress in Biochemistry and Biophysics, 2017, 44 (3): 187- 197.
4	于乃功, 苑云鹤, 李倜, 等. 一种基于海马认知机理的仿生机器人认知地图构建方法[J]. 自动化学报, 2018, 44 (1): 52- 73.
	YU N G , YUAN Y H , LI T , et al. A cognitive map building algorithm based on cognitive mechanism of hippocampus[J]. Acta Automatica Sinica, 2018, 44 (1): 52- 73.
5	邹强, 丛明, 刘冬, 等. 仿鼠脑海马的机器人地图构建与路径规划方法[J]. 华中科技大学学报(自然科学版), 2018, 46 (12): 88- 93.
	ZHOU Q , CONG M , LIU D , et al. Robotic map building and path planning methods based on the hippocampus of rat's brain[J]. Journal of Huazhong University of Science and Technology (Nature Science Edition), 2018, 46 (12): 88- 93.
6	TANG H J , HUANG W W , NARAYANAMOORTHY A , et al. Cognitive memory and mapping in a brain-like system for robotic navigation[J]. Neural Networks, 2017, 87 (C): 27- 37.
7	MILFORD M, JACOBSON A, CHEN Z, et al.RatSLAM: using models of rodent hippocampus for robot navigation and beyond[C]//Proc.of the 16th International Symposium ISRR, 2016: 467-485.
8	BANINO A , BARRY C , URIA B , et al. Vector-based navigation using grid-like representations in artificial agents[J]. Nature, 2018, 557 (7705): 429- 433. doi: 10.1038/s41586-018-0102-6
9	OMER D B , MAIMON S R , LAS L , et al. Social place-cells in the bat hippocampus[J]. Science, 2018, 359 (6372): 218- 224. doi: 10.1126/science.aao3474
10	ZUTSHI I , LEUTGEB J K , LEUTGEB S . Theta sequences of grid cell populations can provide a movement-direction signal[J]. Current Opinion in Behavioral Sciences, 2017, 17, 147- 154. doi: 10.1016/j.cobeha.2017.08.012
11	PAGE H J I , WILSON J J , JEFFERY K J . A dual-axis rotation rule for updating the head direction cell reference frame during movement in three dimensions[J]. Journal of Neurophysiology, 2017, 119 (1): 192- 208.
12	JULIAN J , RYAN J , HAMILTON R , et al. The occipital place area is causally involved in representing environmental boundaries during navigation[J]. Current Biology, 2016, 26 (8): 1104- 1109. doi: 10.1016/j.cub.2016.02.066
13	YAMAMOTO J , TONEGAWA S . Direct medial entorhinal cortex input to hippocampal CA1 is crucial for extended quiet awake replay[J]. Neuron, 2017, 96 (1): 217- 227. doi: 10.1016/j.neuron.2017.09.017
14	ZHOU Y , WU D W . Grid-to-place cells model based on radial basis function network[J]. Electronics Letters, 2017, 53 (3): 200- 201. doi: 10.1049/el.2016.1750
15	KANTER B R , LYKKEN C M , AVESAR D , et al. A novel mechanism for the grid-to-place cell transformation revealed by transgenic depolarization of medial entorhinal cortex layer Ⅱ[J]. Neuron, 2017, 93 (6): 1480- 1492. doi: 10.1016/j.neuron.2017.03.001
16	NEHER T , AZIZI A H , CHENG S . From grid cells to place cells with realistic field sizes[J]. PLoS ONE, 2017, 12 (7): e0181618. doi: 10.1371/journal.pone.0181618
17	HU J , TANG H , TAN K C , et al. How the brain formulates memory: a spatio-temporal model research frontier[J]. IEEE Computational Intelligence Magazine, 2016, 11 (2): 56- 68.
18	MUESSING L , HAUSER J , WILLS T J , et al. Place cell networks in pre-weanling rats show associative memory properties from the onset of exploratory behavior[J]. Cerebral Cortex, 2016, 26 (8): 3627- 3636. doi: 10.1093/cercor/bhw174
19	BONNEVIE T , DUNN B , FYHN M , et al. Grid cells require excitatory drive from the hippocampus[J]. Nature Neuroscience, 2013, 16 (3): 309- 317. doi: 10.1038/nn.3311
20	TRIMPER J B , TRETTEL S G , ERNIE H , et al. Methodological caveats in the detection of coordinated replay between place cells and grid cells[J]. Frontiers in Systems Neuroscience, 2017.
21	RENNO-COSTA C , ABL T . Place and grid cells in a loop: implications for memory function and spatial coding[J]. Journal of Neuroscience, 2017, 37 (34): 8062- 8076. doi: 10.1523/JNEUROSCI.3490-16.2017
22	GIL M , ANCAU M , SCHLESIGER M , et al. Impaired path integration in mice with disrupted grid cell firing[J]. Nature Neuroscience, 2018, 21 (1): 81- 91. doi: 10.1038/s41593-017-0039-3
23	DORDEK Y , SOUDRY D , MEIR R , et al. Extracting grid cell characteristics from place cell inputs using non-negative principal component analysis[J]. ELife, 2016, 5, e10094. doi: 10.7554/eLife.10094
24	STEPANYUK A . Self-organization of grid fields under supervision of place cells in a neuron model with associative plasticity[J]. Biologically Inspired Cognitive Architectures, 2015, 13, 48- 62. doi: 10.1016/j.bica.2015.06.006
25	D'ALBIS T , KEMPTER R . A single-cell spiking model for the origin of grid-cell patterns[J]. PLoS Computational Biology, 2017, 13 (10): e1005782. doi: 10.1371/journal.pcbi.1005782
26	MONSALVE-MERCADO M M , LEIBOLD C . Hippocampal spike-timing correlations lead to hexagonal grid fields[J]. Physical Review Letters, 2017, 119 (3): 038101. doi: 10.1103/PhysRevLett.119.038101
27	WEBER S N , SPREKELER , H . Learning place cells, grid cells and invariances with excitatory and inhibitory plasticity[J]. ELife, 2018, 7, e34560. doi: 10.7554/eLife.34560
28	STOVOLD J , SIMON O . Advances in unconventional computing[M]. Cham, Switzerland: Springer, 2017: 141- 166.

参数	变量	数值
正方形区域	L×L	80 m×80 m
位置细胞个数	N×N	80×80
位置细胞放电率系数	a	10
b	1/8
突触权重更新率	η	5×10^-6
位置细胞放电率缩放系数	p	3
q	3
突触权重边界	Th_down	0
突触权重边界	Th_up	0.1

[1]	赵辰豪, 吴德伟, 韩昆, 代传金. 无环境信息下多尺度网格细胞群空间表征模型[J]. 系统工程与电子技术, 2021, 43(3): 814-822.
[2]	赵辰豪, 吴德伟, 韩昆, 朱浩男, 代传金. 基于多尺度网格细胞的路径整合模型[J]. 系统工程与电子技术, 2021, 43(10): 2961-2967.

类脑导航中基于差分Hebbian学习的网格细胞构建模型

Model of generating grid cell based on difference Hebbian learning in brain-inspired navigation

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 5

参考文献 28

相关文章 2

编辑推荐

Metrics

本文评价