Development and Evaluation of ICP-based Map Matching Localization for Small Autonomous Marine Mobility

Abstract

With the recent advancement of autonomous marine mobility, various studies have been conducted focusing on four modules of H/W and S/W architecture: localization, perception, decision/planning, and maneuvering/control. In marine mobility such as large ships and offshore structures, localization technique is often simply constructed solely using signals measured by Global Navigation and Satellite systems (GNSS). However, small autonomous marine mobility designed for surveillance and reconnaissance are developed with constraints on budget, size, and weight, as well as a high risk of loss during mission execution. In this paper, it was aimed to develop a localization technique applicable to small autonomous marine mobility considering a low-cost, low-precision GNSS with a 3D LiDAR point cloud. Furthermore, metrics were proposed for qualitative and indirect quantitative evaluation of the developed localization technique. First, a positioning technique was developed using an Extended Kalman filter (EKF) and Iterative Closest Point (ICP) map matching. Second, residuals were used for the qualitative evaluation of the developed localization technique. Third, the Intersection over Union (IoU) was used for indirect quantitative evaluation. Finally, it was confirmed that localization performance was significantly improved using the proposed indices.