Hybrid navigation based on GPS data and SIFT-based place recognition using Biologically-inspired SLAM

Abstract

RatSLAM is a type of SLAM system inspired by the hippocampus of mammals which is used to localize and create maps. Based on the Scale Invariant Feature Transform (SIFT) algorithm, an improved RatSLAM is proposed to improve place recognition in indoor environments. In addition, to navigate and route the moving agent, a method based on GPS data and pose cell network of the improved RatSLAM has been proposed, in which the robot’s head direction is coordinated with GPS data. Also, an indoor database was created using MatLab and Webots to evaluate the proposed method. In comparison to main RatSLAM, the improved RatSLAM by SIFT demonstrated higher recall rate (53% to 71%) and precision rate (93% to 100%) with fewer memory cells used, resulting in faster recognition. Furthermore, the robot selected an appropriate path to reach the target points in the proposed method for navigation.