Dynamical modeling of hippocampal-basal ganglia interactions for spatial navigation

Coordinated interactions between the hippocampus and basal ganglia are known to support navigational decision-making, yet their precise collaborative mechanisms remain elusive. Based on biological theories, this study establishes a hippocampal-basal ganglia circuit model for navigation. Unlike existing neural reinforcement learning models, the proposed model aims to investigate how the interaction between the hippocampus and basal ganglia influences navigation. The model incorporates spike-timing-dependent plasticity (STDP) and dopamine-mediated reinforcement learning, enabling the hippocampal module to learn environments and retain goal memories in an allocentric (world-centered) coordinate system. Additionally, it integrates a cortico-basal ganglia network to address choice conflicts. This network receives egocentric (self-centered) landmark inputs and establishes stimulus-action associations through synaptic plasticity. By combining the hippocampus’s spatial representation and the basal ganglia’s action selection strategy, the model simulates the decision-making process from spatial learning to motor execution. Furthermore, the model successfully reproduces rodent navigation behaviors in Morris water maze and plus maze paradigms, demonstrating lesion-induced deficits matching biological observations. Finally, validation through mobile robot navigation task confirms physical realizability. The model demonstrates biological plausibility, mechanistically explaining how action sequences are generated during biological navigation. It provides a novel computational perspective for understanding the neural basis of navigational behavior.