Robot localization performance using different SLAM approaches in a homogeneous indoor environment

Abstract

Robot localization has always been a significant concern in the robotic field. This paper presents robot localization performance for a ROS-based differential platform using four distinct measuring tools. The majority of mobile robots deploy the odometry data for localization, which in turn is highly influenced by the accuracy of encoders and IMUs as well as the quality of tracking surface. The main objective of this research is to eliminate the odometry and to estimate the robot's position and orientation (Pose) merely through an image or a laser scan. Several SLAM methods have been proposed to extract the robot's Pose out of visual data. Despite the emergence of visual sensors in recent years, no pertinent study could be found on the influence of camera types on the Pose accuracy. In this work, Visual SLAM is implemented using Xbox 360 Kinect, Xbox One Kinect, Realsense D435, and Laser SLAM is implemented utilizing Hokuyo UTM30-LX. At last, the benefits and drawbacks of using each are discussed by referring to their performance.