A Preliminary Study on DWA-based Collision-free Path Planning for Remote Navigation Support System of Autonomous Ships

Abstract

This paper proposes an autonomous navigation framework designed to ensure the safe navigation of remotely controlled autonomous ships within port areas characterized by dense maritime traffic. To effectively support autonomous ships remotely operated from a remote operating center (ROC), it is crucial to comprehend not only the localized surroundings of the autonomous ships but also the broader maritime traffic conditions within the maneuvering area. Typically, maritime traffic information is obtained at the ROC using radar and Automatic Identification Systems (AIS). However, as this information is collected from long distances, it inherently includes a degree of uncertainty, time latency, and deviations. These inherent issues must be considered to remote operators to facilitate safe and effective decision-making. The proposed collision-free path planning framework integrates the motion information from various objects (e.g., maneuvering ships, anchored ships, and buoys) acquired through radar and AIS. It then incorporates the associated uncertainties to construct a Gaussian distribution model. By leveraging this probabilistic distribution model, the framework quantifies uncertainties and incorporates the resulting metrics into the objective function of a dynamic window approach (DWA) to generate collision-free paths. To validate the practical feasibility of the proposed framework, we conducted a series of simulations and preliminary field tests, and we discussed the results.