AET-LIO: An Adaptive Event-Triggered LiDAR–Inertial Odometry System for Localization and 3-D Reconstruction in Unmanned Surface Vehicle

Abstract

Accurate localization and 3-D reconstruction are critical for unmanned surface vehicles (USVs) operating in dynamic marine environments. However, water surface reflections, global positioning system (GPS) interruptions, and motion disturbances caused by wind and waves significantly degrade LiDAR–inertial odometry (LIO) performance. This article proposes an adaptive event-triggered (AET) LIO (AET-LIO) framework that dynamically adjusts the pose estimation model based on a motion-intensity metric, enabling robust state estimation under varying sea conditions. Furthermore, a real-time water-reflection filtering method is introduced to enhance point cloud integrity without sacrificing computational efficiency. Experiments conducted in both calm and wind-and-waves scenarios demonstrate that the proposed system reduces localization root-mean-square error (RMSE) compared with state-of-the-art LIO methods, while achieving high-fidelity structures above water reconstruction. The results validate the effectiveness of the AET in mitigating environmental uncertainties and ensuring reliable mapping in GPS-challenged maritime settings.