GNM: GridCell navigational model. (15th June 2020)

• Record Type: Journal Article
• Authors: Naghizadeh, Alireza; Berenjian, Samaneh; Margolis, David J.; Metaxas, Dimitris N.
• Is Part Of: Expert systems with applications. Volume 148(2020)
• Publication Date: 2020-06-15
• Subjects: Gridcell -- Navigation -- Exploration -- Robotic -- Computational model -- Path-planning; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2020.113217
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: We get inspiration from patterns of GridCells to propose a new navigational model. Our goal is to provide a better navigation algorithm, not being biological realistic. We are concerned with exploring unknown environments with unknown end locations. It is desired to reach the goal with a small number of steps and memory usage. Our method is memory efficient and is faster than current solutions. Abstract: It has been shown that grid cell firing patterns in the medial entorhinal cortex, can be used as a mapping reference for spatial navigation in mice and other mammalian species. In this paper, we propose a novel computational model for patterns of grid cells and combine it with a mechanism to tune the weights of cells, which we use to create a decision-making process for robot navigation. The method is used as an unsupervised method for uninformed online search with unknown goal positions and unknown environments, such as finding the exit out of a maze or help a robot to find its way in a jungle where there is no clue about the exit. The results of this approach in simulated and real environments show superior algorithmic steps over current search methods. In addition, the typical size of the memory can be reduced without compromising completeness of the method. Our results show that the number of steps is stable in terms of variations in memory allocations.