Enhancing Graph-SLAM 2D Maps Accuracy and Data Efficiency via Loop Closure through PGO and Map-Based Navigation

Abstract

This paper investigates the accuracy of 2D Graph-SLAM maps for the TurtleBot3 (TB3) robot in the Robot Operating System (ROS). We conduct two experiments to assess the impact of Pose Graph Optimization (PGO) and loop closure detection on map quality. In the first experiment, the robot navigates a single-loop trajectory in two indoor environments, generating maps with both standard Graph-SLAM and PGO-based Graph-SLAM methods. The second experiment involves multiple-loop navigation. Teleoperated paths result in non-optimal trajectories and large ROSbag files due to extensive data collection. To address this, we use a map-based navigation method to optimize the robot's trajectory, reducing simulation time and file sizes. Results show that PGO-based Graph-SLAM significantly improves map accuracy and trajectory quality, especially when combined with loop closure detection.