Research On Environmental Cognition Mechanism Of Rat's Brain Hippocampal Formation And Its Application In Navigation

Autonomous localization and target-oriented navigation capability are essential for animals and autonomous mobile robots.Compared with robots,animals and humans have the advantage of being able to quickly locate their current position by using incomplete space information from their senses without any prior knowledge.Along with the associated neurons related to environmental cognition such as stripe cells,grid cells,place cells in rats were found in succession,it gave some insight into the problem of how animals and humans' self-localization might happen in spatial environment.This paper studies the cognitive mechanism of hippocampal formation in rat's brain and applies it to navigation.The main research work of this paper includes:(1)Based on grid cell biology discovery mechanism,a novel grid cell's grid field model is constructed,which is grid cell's grid field model based on path integration.Based on biological studies,it is asssumed that there are many different path integrator modules in the entorhinal cortex.Each path integrator module has three vector directions and the vectors in three directions are the same size.Different path integrator modules have different vector directions and vector sizes.The rat's self-motion information is composed of the head direction and velocity of the rat.Firstly,the integrator integrates the rat's self-motion information to calculate the current position of the rat.Then calculate the amount of displacement along the direction of each vector and determine whether the grid cells are activated according to the grid cell spacing and grid size which are set in the experiment.Finally,it integrates the output of path integrator module's three vector directions to obtain the grid cell's grid field model which is consistent with the biological research.The simulation results show that grid cell's grid field in the proposed model is consistent with the grid field discharge characteristics of grid cells recorded in different positions of the dorsal ventral cortex in the entorhinal cortex.(2)According to the information transfer mechanism between the entorhinal cortex and hippocampus,a novel place cell model is constructed,which is a Gauss distribution activation function model of grid cells to place cell.Biological research found that a simple linear superposition mathematical model cannot express the mapping from multiple activated grid cell's grid fields to a single place cell's place field in the cerebral cortex of rodents.For this difficult problem,the Gauss distribution activation function is introduced to filter out the low activation rate's place field.The results of simulation experiments obtain a single place cell's place field which is consistent with the biological findings.(3)Based on the constructed place cell model in(2),a method for constructing hippocampus cognitive map is proposed.That is,with the rat's continuous exploration of the space environment,it gradually forms the single localized firing pattern of place cell at each position in the environment and finally forms the expression of the spatial environment in the rat's brain which is called hippocampus map.The simulation results show that the intrinsic expression of the outer space environment which is called hippocampus map in the brain of rats is generated with the free movement of the rat.(4)Biological studies have found that hippocampus map alone cannot predict the future movement direction of the rat.The ventral tegmental area mainly exists a dopaminergic neuron associated with the reward prediction error signals and dopaminergic neurons project information into the nucleus accumbens.The input of nucleus accumbens mainly comes from the hippocampus.There is bi-directional nerve fiber projection between the prefrontal cortex and nucleus accumbens.That is,on the one hand,neurons in the nucleus accumbens receive the rat's spatial information from the hippocampus.On the other hand,neurons receive the relevant prediction error information from the ventral tegmental area and then interact with the prefrontal cortex to correctly predict the future movement direction of the rat.The neurons in the prefrontal cortex of the rat are mainly related to motion-motor cells,while the neurons in the hippocampus are mainly place cells.The neurophysiological basis of target-oriented navigation task in rats is the synaptic modulation related to reward prediction error signals between hippocampal place cells and motor cells.The target-oriented navigation task is achieved by reinforcement learning algorithm between continuous state and action spaces.The continuous state spaces refer to the discharge activity of place cells.Based on this,a feed-forward neural network model including input layer,place cell layer,motor cell layer and output layer is constructed and Q-learning algorithm is used to achieve the target-oriented navigation task.Experimental results show that when the rat learns a period of time,it can quickly find the shortest path from the start position to the target position from the 9th round experiment and quickly realize the navigation to the target position,which fully demonstrates that the model constructed in this paper can quickly realize the target-oriented navigation task.