A Hybrid FAST-LIO2 and SC-A-LOAM SLAM Algorithm for Autonomous Vehicles

Incorporating simultaneous localization and mapping (SLAM) techniques empowers robots to construct environmental maps while concurrently pinpointing their own locations within these maps. SLAM, utilized across robotics, virtual reality, and geographic information systems, is an evolving field, aimed at refining robustness, real-time performance, and scalability. Ongoing research addresses challenges encompassing sensor expenses, algorithm intricacies, and the absence of standardized datasets. This study explores the scalability of integrating diverse SLAM methodologies to alleviate their specific limitations. Further, it compares efficiency among LiDAR-based SLAM techniques and evaluates LiDAR data's applicability through point cloud processing. The methodology encompasses data acquisition, preprocessing, SLAM algorithm selection and implementation, system integration and optimization, and comprehensive testing and evaluation. Additionally, this work enhances odometry algorithms loop closure detection capability. Integration of SCA-LOAM and FAST-LIO2 yields advantages surpassing their individual applications. FAST-LIO2 gives a superior APE of 0.91 m and RPE of 0.35%. SC-A-LOAM's loop detection outperforms FAST-LIO2's default algorithm, while GTSAM-based pose graph optimization offers marked improvements. This integrated system is particularly well-suited for long-term localization tasks, such as those required for autonomous vehicles.