The flexible use of prior experience to guide future behavior is a core feature of biological intelligence. However, the neural mechanisms that support such adaptive learning, especially in multi-goal environments, remain insufficiently understood. A hierarchical optimization navigation model driven by hippocampal reverse replay was proposed. The model was built on the spatial representation of place cells. Task-specific memory was formed through synaptic weight modulation induced by reverse replay. Cross-task optimization was achieved by loss regulation based on replay duration, thereby enabling the construction of a transferable navigation strategy. Unlike traditional retraining-based approaches, the proposed method indirectly optimizes the initial parameters of the model by adjusting the duration of replay, which allows rapid transfer of task knowledge. The model was systematically validated through simulations of mouse-like exploration in environments with dynamically changing goals. The results demonstrate that navigation time to known targets is significantly reduced, while high generalization performance to previously unseen goal locations is maintained. These findings confirm the effectiveness of the reverse-replay-based hierarchical optimization strategy for rapid multi-task adaptation. From the interdisciplinary perspective of neuroscience and computational modeling, this study provides new insights for the development of adaptive navigation systems in artificial intelligence and offers a theoretical basis for understanding and simulating hippocampal function.
ScovilleW B, MilnerB. Loss of recent memory after bilateral hippocampal lesions[J]. Journal of Neurology, Neurosurgery & Psychiatry, 1957, 20(1): 11-21.
[2]
MullerR U, SteadM, PachJ. The hippocampus as a cognitive graph[J]. The Journal of General Physiology, 1996, 107(6): 663-694.
LiuYanping, ZhengRongyan, SongFuhong, et al. A multi-strategy improved hippocampal optimization algorithm and its applications[J]. Science Technology and Engineering, 2025, 25(10): 4216-4228.
LiaoYishen, YuNaigong. A spatial navigation method based on a mouse-brain-inspired olfactory-hippocampal-prefrontal information transmission circuit[J]. Journal of Biomedical Engineering, 2024, 41(1): 80-89.
[7]
MorrisR G M, GarrudP, RawlinsJ N P, et al. Place navigation impaired in rats with hippocampal lesions[J]. Nature, 1982, 297(5868): 681-683.
[8]
OltonD S, WalkerJ A, GageF H. Hippocampal connections and spatial discrimination[J]. Brain Research, 1978, 139(2): 295-308.
WangLixiao, LiuChen, SunXiaohui. Spatial cognition analysis of pedestrians in subway stations based on cognitive maps[J]. Science Technology and Engineering, 2022, 22(4): 1588-1594.
RuanXiaogang, ChaiJie, WuYue, et al. Cognitive map construction and navigation based on hippocampal place cells[J]. Acta Automatica Sinica, 2021, 47(3): 666-677.
YuNaigong, YuanYunhe, LiTi, et al. A cognitive map construction method for bionic robots based on hippocampal cognitive mechanisms[J]. Acta Automatica Sinica, 2018, 44(1): 52-73.
[17]
PfeifferB E, FosterD J. Hippocampal place-cell sequences depict future paths to remembered goals[J]. Nature, 2013, 497(7447): 74-79.
[18]
ÓlafsdóttirH F, BushD, BarryC. The role of hippocampal replay in memory and planning[J]. Current Biology, 2018, 28(1): 37-50.
XuYajun, HaoBoyang, GaoTianlu, et al. Frequency assignment of aeronautical navigation aids based on a multi-objective genetic local search algorithm[J]. Science Technology and Engineering, 2025, 25(15): 6530-6537.
[21]
DragoiG, TonegawaS. Selection of preconfigured cell assemblies for representation of novel spatial experiences[J]. Philosophical Transactions of the Royal Society B: Biological Sciences, 2014, 369(1635): 20120522.
[22]
MckenzieS, RomanH, EnglishD F, et al. Preexisting hippocampal network dynamics constrain optogenetically induced place fields[J]. Neuron, 2021, 109(6): 1040-1054.
FengHaoran, ShangZhigang, YangLifang, et al. An evaluation method of animal learning ability based on Q-learning parameter identification[J]. Science Technology and Engineering, 2022, 22(27): 11842-11849.
[25]
TseD, LangstonR F, KakeyamaM, et al. Schemas and memory consolidation[J]. Science, 2007, 316(5821): 76-82.
[26]
DragoiG, TonegawaS. Preplay of future place cell sequences by hippocampal cellular assemblies[J]. Nature, 2010, 469(7330): 397-401.
[27]
GrosmarkA D, BuzsákiG. Diversity in neural firing dynamics supports both rigid and learned hippocampal sequences[J]. Science, 2016, 351(6280): 1440-1443.
[28]
Berners-LeeA, FengT, SilvaD, et al. Hippocampal replays appear after a single experience and incorporate greater detail with more experience[J]. Neuron, 2022, 110(11): 1829-1842.
[29]
WidloskiJ, Foster, DavidJ. Flexible rerouting of hippocampal replay sequences around changing barriers in the absence of global place field remapping[J]. Neuron, 2022, 110(9): 1547-1558.
[30]
McNaughtonB L, BarnesC A, GerrardJ L, et al. Deciphering the hippocampal polyglot: the hippocampus as a path integration system[J]. Journal of Experimental Biology, 1996, 199(1): 173-185.
CavalieriD, AngelovaA, IslahA, et al. CA1 pyramidal cell diversity is rooted in the time of neurogenesis[J]. eLife, 2021, 10: 69270.
[35]
PoulopoulosA, DavisP. Faculty opinions recommendation of preconfigured dynamics in the hippocampus are guided by embryonic birthdate and rate of neurogenesis[J]. Nature Neuroscience, 2022, 25(9): 1201-1212.
[36]
HarveyR E, RobinsonH L, LiuC, et al. Hippocampo-cortical circuits for selective memory encoding, routing, and replay[J]. Neuron, 2023, 111(13): 2076-2090.
[37]
AmbrogioniL, ÓlafsdóttirH F. Rethinking the hippocampal cognitive map as a meta-learning computational module[J]. Trends in Cognitive Sciences, 2023, 27(8): 702-712.
[38]
HagaT, FukaiT. Recurrent network model for learning goal-directed sequences through reverse replay[J]. eLife, 2018, 7: 217422.