予測型エージェントを用いた船舶の群行動に関する研究

概要

論文の詳細を見る
This paper presents an analysis of ship-group navigation simulations based on an autonomous ship behavior cluster model. In this model, maneuvering decisions of a ship agent are computed by combining four independent forces : goal attainment force, contracting force, expanding force, and cluster collision avoidance force. The magnitude of each force is determined by a sigmoid function which takes distance to a relevant point/vessel as input. Although fairly simple, the sigmoid function led to flexible decision-making. The characteristic of group behavior was compared using two kinds of agents, the estimation type agent which can estimate risk of collision, and the non-estimation type agent which lacks this function. The estimation type agents exhibited better performance in terms of tracks, the distance of CPA and the total amount of course altered.
社団法人日本航海学会の論文
2004-09-25