SZU-Campus-GLVI Dataset

To evaluate the engineering practicality of SLAM algorithm in real complex environments and its robustness under extreme degradation conditions, we present the self-collected SZU-Campus-GLVI multi-sensor dataset. This dataset accompanies the paper "GLOBE-LIVO: A Global Robust Localization Method for Unmanned Systems Based on GNSS RTK-LiDAR-Inertial-Visual Fusion". The data acquisition platform integrates a Livox Mid-360 LiDAR, a global shutter camera, an ICM-40609 IMU, and a GNSS RTK receiver. SZU-Campus-GLVI dataset specifically covers avariety of severely degraded scenarios that are highly challenging for LIVO(LiDAR-Inertial-Visual Odometry) or GNSS RTK systems: Dynamic Crowded Areas: Data were collected at building entrances and exits, cafeterias, and plazas with dense populations, where a large number of moving pedestrians cause unstable visual features and severe dynamic interference in LiDAR point clouds; Long Indoor Corridors: Traversing windowless indoor corridors lacking texture and geometric structure, resulting in simultaneous degradation of both vision and LiDAR; Urban Canyons and Canopy Occlusion: Traveling along campus main roads and tree-lined paths, where tall buildings and dense tree canopies cause frequent GNSS RTK signal interruptions or multipath effects; Long-range Loop Closure Trajectories: Some sequences have a total travel distance exceeding 3 km, with explicitly designed start-end coincident motion trajectories to examine the algorithm’s capability in suppressing accumulated drift and maintaining global consistency over long-duration operation. For documentation and updates, see: https://github.com/SZU-Rob-IPNP-Lab/SZU-Campus-GLVI-Dataset