Marine Traffic Simulation Using Ship Agent Clusters : Northbound Traffic in Tokyo Bay

Abstract

This paper presents a simulation study of the characteristics of the northbound traffic in Tokyo Bay using ship agents. A ship agent has four functions, i.e., sensor, decision-making, memory and effecter functions. The decision-making function performs circumstantial judgments and makes course and speed adjustments utilizing a ship cluster behavior model. We simulated the navigation characteristics of ship agent clusters in Tokyo Bay by manipulating the initial interval size among ships that formed a ship cluster. We found that when ship clusters enter the Uraga Suido Traffic Route from different directions, the initial interval size has little impact on the successful entry into the passage. Regardless of the interval size, the width of the new cluster was always smaller than the passage width. However, the simulation revealed that when the initial interval is small, ship agents sometimes deviate from the traffic route after entering. We also found that the initial interval size changed the point of heaviest congestion. When the interval is small, the congestion point emerged further south of the traffic route, but the point moved up north as the interval size was increased. No collision, congestion, or passage deviation was observed when the interval size was sufficiently large. However, the traffic (vessels/hour) under this "safe" condition in the simulation was smaller than the actual traffic volume. We conclude that the cluster condition before entering the passage has a large impact on the later navigation characteristics in the passage.