In expansive apple orchards, autonomous mobile robots encounter substantial obstacles in achieving high-precision environmental mapping and real-time pose estimation; primarily due to uneven terrain, elevated environmental noise; and repetitive scene features. This paper aims to introduce GoLE-SLAM, an innovative LiDAR-based Simultaneous Localization and Mapping (SLAM) system tailored for such challenging outdoor environments.
To mitigate the impact of rugged orchard terrain, GoLE-SLAM uses a seed-growth algorithm leveraging LiDAR scan-line structures; coupled with a one-dimensional composite filtering strategy; This approach effectively segments the ground while filtering out prominent weeds, and surface irregularities. Furthermore, capitalizing on the availability of reliable GPS signals in outdoor settings, a GPS-Informed Selective Correspondence weighted matching strategy is implemented to optimize computational efficiency during loop-closure detection.
Extensive field tests in apple orchards demonstrate that GoLE-SLAM achieves efficient processing with minimal computational overhead, robust terrain segmentation for nonplanar surfaces and effective drift compensation in localization trajectories through its loop-closure module.
A tightly coupled SLAM framework, termed GoLE-SLAM, is proposed, integrating LIDAR, GPS and inertial measurement unit (IMU). This framework combines ground-assisted LiDAR odometry, IMU preintegration and loop closure enhancement to achieve superior state estimation and mapping performance.
