Simulation research on grid cell discharge characteristics based on feedforward neural network

In this study, an adaptive network model of grid cells was used to simulate the formation of grid cell networks during animal development. Each grid cell receives feedforward input from place cells and head direction cells. Feedforward connections use Hebbian learning to adjust connection weights to correlate the firing state of grid cells with location inputs. The simulation draws the conclusion that the grid cells are modulated by the position cells and the head facing cells, resulting in a stable periodic discharge pattern, and the partial discharge range covers the entire space, which is consistent with the biological experiment. When studying the firing pattern of grid cells, the common method only uses the input of place cells to grid cells, and in this study we added the effect of head direction cells on grid cell firing frequency. The coding method developed in this project can be applied to reinforcement learning to explore new path planning methods and provide reference for bionic navigation systems.