항적데이터 기반 경험적 확률 로드맵 기법을 이용한 연안해역 경로 탐색 연구

다국어 초록 (Multilingual Abstract)

Coastal navigation involves complex spatial constraints such as narrow channels, irregular shorelines, and shallow waters, which require safe and efficient route planning. Although existing path planning algorithms have demonstrated effectiveness in terms of optimality or computational efficiency, they often fail to reflect actual vessel operating patterns, resulting in unrealistic routes, particularly in constrained coastal environments. This limitation mainly arises from defining the search space as a global free space without considering empirical navigational characteristics. To address this issue, this study proposes an Empirical Probabilistic
Roadmap (EPRM)–based path planning framework that redefines the search space using real ship trajectories derived from Automatic Identification System (AIS) data. First, a navigational environment was constructed from electronic nautical charts by incorporating coastlines, islands, obstacles, and depth constraints, which was then converted into a binary occupancy grid. Historical AIS trajectories were analyzed to identify regions repeatedly used by vessels, and these regions were defined as anempirical navigable space. The EPRM algorithm generates roadmap nodes and edges exclusively within this trajectory-based space, thereby
structurally eliminating inefficient global sampling inherent in conventional probabilistic roadmap methods. Performance evaluations were conducted in a coastal area of the southwestern sea of Korea, and the proposed method was compared with A*, PRM, RRT, and RRT* algorithms in terms of success rate, path length, planning time, and trajectory deviation. The results demonstrate that EPRM achieves stable path planning with fewer samples, produces shorter paths with significantly reduced computation time, and generates routes with high spatial consistency with actual AIS trajectories. These findings indicate a substantial improvement in both navigational realism and computational efficiency over existing approaches. Beyond algorithmic performance enhancement, this study presents a new perspective on coastal path planning by emphasizing search space design based on empirical navigational behavior. The proposed EPRM framework can be effectively applied to autonomous ship route planning systems, S-100–based electronic navigational charts, and vessel traffic service
decision-support systems, providing a practical and reliable solution for safe and efficient coastal navigation.