LOAM ODOMETRIC ALGORITHM INNOVATED FROM C++ WITHOUT ROBOT OPERATIVE SYSTEM

Abstract

Odometry plays a crucial role in estimating the relative position of a mobile robot, which motivates this article to implement the method used in the LOAM algorithm, developed within the ROS (Robot Operating System) framework. The focus of this work is on the analysis of the LOAM algorithm and its subsequent implementation in C++ without dependency on ROS, using libraries such as PCL (Point Cloud Library), Eigen 3.0, and some functionalities adapted for use in C++, such as ROS. ::tf. To carry out the analysis of the LOAM algorithm, a reverse engineering process was applied and the existing literature was reviewed, based on the articles published by the authors of the aforementioned algorithm. Likewise, Visual SLAM (Simultaneous Localization and Mapping) methods were examined in order to compare them and propose possible improvements and observations in the algorithm studied and subsequently implemented. It was shown that the LOAM algorithm has greater efficiency in straight sections compared to curves, due to the absence of an inertial measurement sensor that is capable of compensating for the rotation of the laser in curves. Therefore, it is recommended to perform a segmented analysis in scenarios with various types of paths to improve the results and reduce the total error of the experiment.