LiDAR-Inertial SLAM With Voxel-Information Filter for Autonomous Surface Vehicles

Due to the high cost and limited onboard computational capabilities, real-time simultaneous localization and mapping (SLAM) for autonomous surface vehicles (ASVs) in complex water environments has long been a technology challenge. To address this issue, a tightly coupled voxel-informed gradient Anderson light detection and ranging (LiDAR)-inertial odometry is designed to facilitate high efficient ASV SLAM. Therein, a manifold-constrained iterative information filtering optimization scheme is devised with adaptive outlier rejection design. To enhance ASV SLAM efficiency, a voxel-based acceleration mechanism featuring voxel search pruning and Anderson extrapolation is proposed. Finally, extensive experiments are conducted on a self-established ASV platform to verify the effectiveness of our SLAM scheme.