Spatial Representation Model Based on Grid Cell to Place Cell

Abstract

A biologically heuristic map construction and autonomous localization algorithm is proposed inspired by the cognitive mechanism of hippocampal structures to address the problems of limited localization accuracy and environmental noise interference in simultaneous localization and mapping of mobile robots in unknown environments. The self-motion information of the mobile robot and the external visual landmark information were used as inputs to build the location-aware model from the grid cell to the place cell and visual landmark model, respectively. The visual information extracted by the visual landmark model is combined with the location information obtained by the location-aware model through Hebbian network to construct the spatial topology map. A grid cell reset mechanism is added to the system to limit the number of error nodes generated, and reduce the adverse effects of environmental noise on robot positioning. The experimental results show that the proposed algorithm in this paper enables the construction of spatial topology maps with a low number of error nodes generated when there is a grid cell reset mechanism.