Computational mechanism of spatial disambiguation via dual-channel cue integration in the hippocampal-entorhinal circuit

It remains not fully understood how the brain dynamically integrates heterogeneous sensory information to build robust cognitive maps in perceptually ambiguous environments. To investigate this mechanism, we constructed a computational model inspired by the entorhinal-hippocampal circuit. This model integrates local geometric features encoded by boundary vector cells and global self-motion cues integrated by grid cells. Our results reveal how dynamic interactions of multimodal inputs shape robust spatial representations in perceptually ambiguous environments. Specifically, in geometric aliasing scenarios, global metric inputs from grid cells can break the symmetry of place fields induced by local visual cues. Conversely, in metric aliasing scenarios, distinct boundary features can act as spatial anchors to stabilize the firing representations of place cells. Furthermore, based on these robust spatial representations, the model combined with the successor representation method achieves stable navigation in ambiguous environments. This study reveals the dynamic mechanisms by which the nervous system utilizes the complementarity of heterogeneous information to resolve perceptual aliasing. Consequently, it provides a new computational perspective for understanding information processing in the hippocampus.