Marine Traffic Model – tool for effective decision-making

Research Project, 2015 – 2017

Marine transportation has a large impact on the global economy and environment. Several legal entities have concluded that the current environmental damage caused by marine transportation is unacceptable, and therefore they have started to intervene in it [1]. Unfortunately, marine transportation is a complex system, and well-intended interventions can easily have negative repercussions.  Trial and error is out of the question. The development and implementation time of most legislative and technological interventions is in the order or decades, and the costs of errors, in terms of lives, money, and environmental degradation, are unacceptable. Any type of intervention must be evaluated according to a model of the marine transport system.

            Of all the elements of the marine transport system that must be modeled, marine traffic is a particularly challenging one. It consists of the movement of a large number of vessels, differing with respect to their sailing capacities and economical incentives in a dynamic environment. Currents, waves, wind, rough weather, ice, visibility, traffic density, fuel prices, contractual obligations, and legislation influence the vessel’s route, sailing speed, and behavior in risk situations. Marine traffic accounts for most of the environmental damage and energy consumption of the marine transport system, and therefore it is an attractive place to intervene in order to improve its performance.

The objective of this project is to research the techniques necessary to develop a marine traffic model that is appropriate for devising, developing and evaluating interventions that improve marine traffic performance–its energy efficiency, transportation times and costs, environmental impact, resilience, and safety.

Such model must account for the differences between the sailing vessels; otherwise, it would render unrealistic traffic simulations. Also, the model must be able to predict the traffic flows in new situations where there is no relevant historical traffic data. To fulfill both requirements we plan to build profiles of individual vessels through machine-learning techniques and to simulate the vessel movements in new scenarios through multi-agent based simulations.