With the advance of robotic technology, unmanned surface vessels (USVs) have attracted much attention since these can perform hazardous or time-consuming missions, not only for military purposes but also for civilian purposes such as monitoring of the environment or pollution and inspection of marine infrastructure after natural disasters or large-scale accidents. For these operations, localization and mapping of unknown environments is an essential capability.
The integrated Global Positioning System (GPS) and Inertial Navigation System (INS) is a great tool. However, GPS is often restricted near large structures such as bridges, waterside buildings, cranes, and various other man-made structures. A localization method using an INS without GPS cannot accurately estimate the positions of vehicles, because drift errors grow in time. Hence, additional correction information is needed for precise localization in such GPS-restricted areas.
In addition, obstacle detection is critically required for safe navigation in cluttered environments. In fact, obstacles can be utilized to better estimate vehicle position, as the presence of obstacles enables the vehicle to perform relative navigation or localization with respect to the obstacles detected. In this study, the aim is to use bridge components for relative localization. Bridge columns are common obstacles that obstruct USVs autonomously passing under bridges, but these also can be used as landmarks for relative localization at the same time.
This study proposes a novel filter system that allows simultaneously estimating the positions of a vehicle and the surrounding obstacles, including the geometric parameters of the obstructive structures. Additionally, 3D mapping of the surrounding environment is carried out using cameras and lidar. The feasibility of the algorithm is demonstrated through indoor and outdoor experiments.