SLAM Performance Evaluation on Uneven Terrain Under Varying Illuminance Conditions and Trajectory Lengths
SLAM Performance Evaluation on Uneven Terrain Under Varying Illuminance Conditions and Trajectory Lengths
Blog Article
Simultaneous Localization and Mapping (SLAM) here algorithms based on visual and light detection and ranging (LiDAR) technologies have shown remarkable capabilities in robot navigation, particularly for mapping and navigating unfamiliar environments.However, the performance of visual and LiDAR based SLAM in challenging settings, such as uneven terrain with complex environmental conditions, remains insufficiently explored.To address this challenge, the feasibility of two visual-based SLAM algorithms (i.e.real-time appearance-based mapping (RTAB-Map) and neuro-inspired 4DoF (degree of freedom) SLAM (NeuroSLAM)), and two LiDAR-based SLAM algorithms (i.
e.LiDAR inertial odometry via smoothing and mapping (LIO-SAM) and fast LiDAR-inertial odometry (FAST-LIO)) were investigated.The algorithms were tested on uneven terrain within a palm oil plantation under varying illuminance conditions (morning and noon) and across trajectories of different lengths, corresponding to map sizes of 400 m (small), 600 m (medium), 1000 m (large), and 1200 m (extra-large).Performance metrics, including distance error and loop closure detection, were used to assess the impact of trajectory length and illuminance.Additionally, the consistency and stability of the map estimation were analyzed.
Ground-truth data were obtained using the GeoMax Zoom 10 Total Station to ensure high-precision reference measurement.Results revealed that LIO-SAM consistently achieved the lowest distance errors across all map sizes during morning condition of 0.17 m, 1.08 m, 0.52 m, and 0.
71 m for small, medium, large, and extra-large maps, respectively.During noon, LIO-SAM also demonstrated superior performance for small, medium, and large map sizes, with distance error values of 0.11 m, 0.25 m, and 0.38 m, respectively.
Furthermore, LIO-SAM recorded the highest successful loop closure rates by achieving 75 % in the morning and 100 % at noon.Overall, the findings highlights the superior performance of LiDAR-based SLAM algorithms click here over visual-based SLAM approaches in complex, uneven terrains under varying environmental conditions.