Research On Spatial Expression Method Based On Rat's Spatial Cognition Mechanism

Autonomous locating and map building capability are important foundations for research on autonomous mobile robots.Currently,robot locationg and map building capabilities generally rely on a large number of sensor information collections or rely on a wealth of prior knowledge to determine the position of their own space environment.Compared with robots,mammals and humans seem to have endogenic spatial cognitive ability.They can realize cognition of space environment through exploring the environment in the absence of prior knowledge and external clues,and can quickly and accurately locate their own positions.Anatomical and biological studies have confirmed that the rat's environmental cognitive processes occur mainly in the hippocampal formations of the rat brain.With the discovery of space selective cells such as place cells,head-direction cells,grid cells,and border cells,we have gained much understanding of the biospatial cognitive mechanism.More and more researchers are trying to explain the bio-spatial cognition mechanism by means of mathematics and engineering methods.This provides new possibilities for the intelligent algorithms of bionic navigation of mobile robots.Based on the spatial cognition mechanism of rat brain hippocampal formation,our research group conducted a relevant research on the transmission process of spatial information and its dynamics,and proposed a spatial expression method that can be used for bionic environment recognition algorithms.The main research work of this article includes:(1)A grid cell model combining CAN-OI is proposed to generate grid cells' map blocks at various scales.Firstly,a perceptual speed expression method is proposed based on speed cells and perceptual speed influence factors that accurately encode absolute rates.Then,a recurrent spike neural network(SNN)is constructed based on the Izhikevich model,and network parameters are configured to keep the network's attractor characteristics.The input of network is controlled by the perceptual speed of the spiking oscillation sequence.Finally,the network outputs several θ waves with different phases and amplitudes,and the hexagonal regular grid fields with different scales are generated through mutual oscillation interference.(2)A multi-channel positional cell location wild model is proposed to generate a place cell map consisting of discrete firing fields.First,according to the analysis of the dependence of the entorhinal cortex and hippocampal CA1 region,and the direct mapping of other hippocampal formations to the CA1 region,a multi-channel input method for the place cells' firing fields computation model is proposed to maintain the single mapping capacity of the place cells while the entorhinal cortex damaged(grid cells inactivation),with only the direct input of the population of head direction cells and the population of border cells.And based on the study of the place cell's re-mapping characteristic,the remapping mechanism is achieved through changing of border information.The place cells' map updating is realized according to the remapping characteristics of the place cells.(3)The concept of a complete hippocampus cognitive map is proposed and a complete hippocampus cognitive map is organized and constructed.Firstly,the characteristics of grid cell map discretization and modularity are analyzed from physiology and engineering.Using the concept of map tile technology,grid cell map blocks are organized in the form of tiles;then,feature in grid spacing,the grid map blocks are clustered through K-means clustering algorithm and classified into several types of grid cell map tiles with similar spacing.Each type of tiles are ordered in phase order to achieve a sparse scale in the global spatial information coding.Several layers of global sparse spatial information coding are constructed from the clustering results to form a grid cell map pyramid level model.Finally,based on the irreplaceability of grid cell maps and location cell maps,based on parallel levels The model unites grid cell maps and location cell maps to achieve complete hippocampal cognitive map information representation.This model vercomes the storage pressure of grid cell map disorder storage,and ensures the integrity of information and supports the flexible call of map information at different scales.