Agent-based Simulation in the Stoca Project

In Agent-Based Modeling and Simulations (ABMS) individual agents make their own decisions and  complexity emerges from the interaction between different agents (Bonabeau 2002). The agents might only posses a couple of different rules but it is not possible to know what the final interaction inside the whole system will be. Thus, only a couple of different rules can induce extremely complex behavior (Reynolds (1987) has shown how it is possible to simulate a flock of birds using only three different rules: separation, alignment, and cohesion).

During the Stoca -Project different simulation works were conducted. ABMS was among the chosen methods. Oil accidents are among the most disastrous accidents on seas. BP's oil accident in the Gulf of Mexico has destroyed the local economy. We were insterested in analyzing the impact of an oil spillage in the Gulf of Finland and created an Agent-based model. The model contains individual cargo ships which operate in the Baltic Sea Region. The model can be found in:

Cargo flows simulation

As one simulation run gives only partial information, a Monte Carlo -simulation model was constructed as well. The simulation model runs 100 times and calculates the mean waiting time in different runs. The model can be found in:

Monte Carlo simulation 

Literature review about seaport and maritime traffic simulations

Aim of this literature review was to find out different agent-based simulations mainly concerning maritime traffic. During the review it was found out that agent-based simulation models created for maritime or seaport traffic are very rare. It is a consequence of agent-based simulation being still a relatively new field of modeling. Henesey et al. (2009) have used simulation tool, called SimPort, to use agent-based simulation to estimate container traffic and transshipments in a container terminal. Output of the model is a schedule on how to use berths and how to arrange arriving ships. According to the study of Henesey et al. (2009) it is possible to decrease turn-around times of arriving ships by using the simulation tool and the output of it. Model proposes recommendations on how to manage the container terminal. In his doctoral dissertation, Henesey (2006), researches suitability of multi-agent based systems in supporting the management of container terminal to improve its efficiency. Vidal et al. (2010) have used agent-based simulation model to simulate a seaport container terminal, to be exact the model concerns container terminal’s yard crane operations. Finding of the study is that the efficiency of the cranes can be increased if trucks are served distance-based instead of time-based. In distance-based system the nearest truck will get its containers first, while the time-based system is based on first come first serve. Overall there seems to be very few studies regarding simulations on seaports using agent-based simulation models. However, as agent-based simulation modeling is an extremely flexible modeling method there are no reasons why it could not be used to simulate individual seaports or maritime traffic.

There are though more simulation models used to simulate container terminals and maritime traffic e.g. queues and risks, but they are not agent-based. Models created to simulate maritime traffic are usually based on small geographical locations e.g. different straits that are assumed to have large accident risks or capacity limitations. Cortés et al. (2007) have created a simulation model to research freight traffic in the Seville inland port. Aim of their study was to find out if port’s capacity is enough for its current maritime freight traffic. Finding of the research is that it has enough capacity. Port could even increase its freight traffic. A similar local maritime simulation model is created by Mavrakis and Kontinakis (2008) to research queuing of maritime traffic in Bosporus Straits. The model is discrete-event. The simulation model presented in their study supports previous research concerning Bosporus Straits and its maritime traffic. Benefit of the model is that it offers opportunity for more accurate conclusions. Result of the study is that Bosborus Straits is not congested at current yearly maritime traffic. The Straits could even serve increased amount of traffic without significant increase in the waiting time of freight ships. In addition, same strait is researched with a simulation model by Özbaş and Or (2007). Aim of the study is to research different risks in the Straits of Istanbul. Similar geographical simulation model is used in research by Ulusçu et al. (2009). They used the model for a risk analysis in the Strait of Istanbul, which is the other name for Bosporus Straits. Li et al. (2010) have used simulation model to study the impacts of investment projects in Pearl River Delta area. There is already a lot of competition at the Pearl River Delta area. There are also plans to create more container terminals at the highly competed area. According to the results of the simulation there is enough cargo flow estimated to be in the future for new container terminal to entry the market, especially if its land transport network is functional. van Dorp and Merrick (2009) have researched maritime risk of oil spill using simulation. Merrick et al. (2003) have created a traffic density analysis of proposed ferry service expansion in San Francisco Bay using a maritime simulation model. With the simulation model Merrick et al. (2003) were able to define geographical locations that have greatest risk spikes i.e. locations where possibility for a accident is larger than in the other areas. Another finding is that with increased ferries the number of accidents will increase without large investments in accident prevention.

As it was mentioned in the previous chapter, container terminal are researched with different simulation models. Legato et al. (2010) used a simulation-based optimization in their study. Aim of the simulation was to research operations at a container terminal, especially discharge and load of container ship by multiple quay cranes. With the simulation model Legato et al. (2010) were able to create a decision support tool that can optimize the schedule of quay crane operations so that the waiting times of quay cranes can be minimized. Lau and Lee (2007) and Bielli et al. (2006) have also studied container terminal with simulation.

In addition, there are some more rare study subjects that concern seaports or maritime traffic. Simulation models can be used to optimize fleet sizing when moving oil drilling rigs from one place to another. To move a rig it needs several number of anchor handling tug supply vessels, which are very expensive to rent. Daily hire price can be up to 100,000 euros. It is important to have right amount of supply vessels at the right time. Shyshou et al. (2010) have used simulation model to research the fleet sizing problem. More specifically they used simulation-based decision support tool. With the decision support tool it is possible to decrease the costs caused by fleet renting. Tool also gives a estimate on how many long term vessels the firm should have to decrease the costs by maximum. Ng and Kee (2008) have used simulation model in a similar research. In the study Ng and Kee (2008) used model to estimate the optimal sizes of the container liner feeder ships. Fagerholt et al. (2010) used decision support methodology based on maritime transportation simulation. Aim of the study is to reduce short-term routing and schedule planning problems. Authors also tested simulation-based decision support system with a real case company. With the decision support methodology companies can use valuable knowledge on important strategic planning problems. Most of the researches that use simulation models are created in the 21^st century. Only a small number of maritime simulations are created in the previous century. Table summarizes main studies reviewed. In the text below there are, in addition, more similar references that are not listed in Table below.

All the used references are listed in Microsoft Word document below:

 Agent-based modeling references.doc