LEO-SLAM: A Multi-Level Scan Matching Approach with Submap-based Loop Closure Detection

LiDAR-based Simultaneous Localization and Mapping (SLAM) has become an essential capability for mobile robots and autonomous vehicles. Its versatility and reliability across various environments and conditions make it the preferred choice for handling diverse scenarios. LiDAR mapping techniques enable robots to generate detailed environmental representations for navigation. In this work, we present an alternative LiDAR SLAM approach that employs a multi-level alignment strategy and submap-based scan context loop closure detection to produce accurate maps suitable for navigation. We evaluated our method using a benchmark dataset and compared the results with two state-of-the-art algorithms, demonstrating strong accuracy and performance. Although our algorithm achieved outstanding results, we emphasize map coherence for reliable navigation, which is an essential yet often underestimated factor in state-of-the-art odometry systems.