Electrophysiologically informed spiking neural networks for fish-inspired navigation with boundary vector cells and hydrostatic pressure cues

Spatial memory is a fundamental cognitive capacity governing the mental encoding of spatial information, which can be used to navigate through intricate terrains. While most navigation models rely on place- and grid cells as key building blocks, in fish, the largest vertebrate class, the neural basis of navigation was suggested to primarily comprise boundary vector cells (BVCs) and hydrostatic pressure (HP) cues. In this work, we used experimental neural data recordings from the telencephalon of the goldfish to implement a neuromorphic (brain-inspired) spiking neural network-based navigation framework. BVCs were used for collision avoidance and HP for trajectory control toward the target. We show that the model supports reliable goal-directed navigation in a depth-constrained task without requiring explicit place-cell position encoding. Navigation performance emerges from the interaction between boundary-related signals, HP cues, and a fixed initial directional bias. These results provide a computational account of how biologically grounded cues can support efficient navigation under constrained sensory assumptions.