Operational model for vessel traffic using optimal control and calibration

Due to the ever-increasing economic globalization, the scale of transportation through ports and waterways has increased sharply. As the capacity of maritime infrastructure in ports and inland waterways is limited, it is important to simulate vessel behavior to balance safety and capacity in restricted waterways. Currently many existing vessel simulation models focus mainly on vessel dynamics and maritime traffic in the open ocean. These models are, however, inapplicable to simulating vessel behavior in ports and inland waterways, because behavior in such areas can be influenced by many factors, such as waterway geometry, external conditions and human factors. To better simulate vessel behavior in ports and waterways, we developed a new maritime traffic model by adapting the theory of pedestrian models. This new model comprises two parts: the Route Choice Model and the Operational Model. The Route Choice Model has been demonstrated and calibrated in our recent study, in which the desired speed is generated. This paper presents the second part of the model, the Operational Model, which describes vessel behavior based on optimal control by using the output of the Route Choice Model. The calibration of the Operational Model is carried out as well. In the Operational Model, the main behavioral assumption is that all actions of the bridge team, such as accelerating and turning, are executed to force the vessel to sail with the desired speed and course. In the proposed theory, deviating from the desired speed and course, accelerating, decelerating and turning will provide disutility (cost) to the vessel. By predicting and minimizing this disutility, longitudinal acceleration and angular acceleration can be optimized. This way, the Operational Model can be used to predict the vessel speed and course. Automatic Identification System (AIS) data of unhindered vessel behavior in the Port of Rotterdam, the Netherlands, were used to calibrate the Operational Model. The calibration results produced plausible parameter values that minimized the objective function. The paths generated with these optimal parameters corresponded reasonably well to the actual paths.