IEEE Electrification Magazine - December 2019 - 28

hazards in the busy waterways outside Istanbul. Please
refer to Campbell et al. in the "For Further Reading" section for a thorough review of collision-avoidance techniques for autonomous ships.
Weather-related issues play a critical role in maritime
optimization and require a dynamic approach based on
real-time data. In many cases, the corresponding optimization problem will involve multiple objectives, for which
specialized evolutionary algorithms (e.g., NSGA-II) can be
applied. A recent review is available in Walther et al.

Security and Crime Fighting
AI techniques can be employed in marine applications
involving security and crime fighting. For instance, according to a report published by Allianz in 2012, 470 crew
members were held as hostages as a result of pirate
attacks in Somalia in 2011. Jakob et al. discussed AI techniques that were used for random route selection in highrisk areas (i.e., in the region near the Horn of Africa) based
on recent statistical data on the geographical distribution
of pirate attacks. However, the problem can also be
researched by using multiagent simulation methods or
applying game theory to develop a system for scheduling
Coast Guard route and deployment planning, with the aim
of minimizing security risks at U.S. ports.
Another risk factor mentioned in the Allianz report is
that human error and fatigue account for between 75
and 96% of all maritime accidents. Here, too, AI techniques could play an important role in the future of maritime operations.

Hull Optimization
Another application of machine learning concerns the
design of ship hulls, taking into account various optimization criteria, i.e., drag, displacement, and so on. This is a
typical application where stochastic-optimization algorithms, i.e., genetic algorithms, can be applied.
The goal of hull form optimization is to minimize the
vertical motion of a ship's bow, or e.g., to target resistance
minimization. In both cases, a genetic algorithm can be
applied to a simplified mathematical model. It is also possible to combine several optimization criteria, such as resistance minimization, seakeeping performance, and stability
performance by applying a multiobjective genetic algorithm to find pareto-optimal solutions for use in designing
the hull of a fishing vessel. A similar approach can be used
also for hull optimization applied to fast naval vessels.

Theories and Methods for Ship Simulation
Stochastic-optimization algorithms require many iterative
calculations, thus imposing limits on the time allowed for
a single simulation. In the examples discussed previously,
indirect measures were used for determining, e.g., resistance and seakeeping performance. It would, however,
also be possible to simulate an entire ship directly,
although traditional fluid dynamics simulation techniques

I E E E E l e c t r i f i cati o n M agaz ine / DECEMBER 2019

(e.g., grid-based approaches including the finite-volume,
finite-element, and finite-difference methods) are typically too computationally intensive to be useful in such situations. An alternative approach that has gained popularity
in recent years is to use grid-free, particle-based methods,
i.e., smoothed particle hydrodynamics (SPH). The SPH
approach was originally developed for astrophysical applications by Lucy et al. After undergoing many modifications and refinements, today, SPH is also used for general
fluid simulations. In SPH, a scalar quantity (e.g., density) is
interpolated by integrating overcontributions from various
particles using a smoothing kernel function to extend the
influence of any given particle. In practice, because the
number of particles is finite, the integral is reduced to a
weighted sum. A similar procedure can be used for finding
derivatives of any order by differentiating the kernel function. SPH can easily handle distortions of any magnitude
and also guarantees mass conservation by construction
because the particles bear the mass in SPH simulations.
SPH methods are particularly suitable for violent phenomena, such as large waves hitting a vessel. There are also
drawbacks with SPH, but, a clear advantage is the possibility of quickly evaluating a scenario (e.g., during optimization) using a rather small number of particles, and then
either increasing the number of particles or employing a
grid-based method to achieve more accurate simulations
once the optimization procedure has been completed.

Case Study: Styrsöbolaget
A prestudy was carried out in partnership with a shipping
company, AB Göteborg Styrsö Skärgårdstrafik, commonly
known as Styrsöbolaget. The company is a subsidiary
company of Transdev Sverige AB, and its business is
based on contracted traffic on behalf of a local transportation company (Västtrafik) and Göteborgs Stad (The city
of Gothenburg). Styrsöbolaget operates Västtrafik passenger traffic with archipelago boats and river traffic with
service from two ferries: Älvsnabben and Älvsnabbare. The
traffic is procured by Västtrafik, which also decides on
traffic volume and tariffs. On behalf of the city of Gothenburg (i.e., the Traffic office), Styrsöbolaget also carries
freight traffic with goods and commercial vehicles within
the Gothenburg archipelago.
The vessels are owned by Styrsöbolaget, except for the
two new commuter ferries used for river traffic. These ferries, operated by Styrsöbolaget on behalf of Västtrafik,
were put into service at the beginning of 2015. The terminals, bridges, and waiting halls are all owned by Västtrafik, with the exception of a few that operate in the
archipelago, where some bridges are privately owned by
local associations.
Traffic in the southern archipelago is operated by four
different lines from the port of Saltholmen: Line 281,
between Donsö and Vrångö; Line 282, from Styrsö to Brännö Husvik; Line 283, between Asperö and Brännö Rödsten;
and Line 284, from Knarrholmen to Stora Förö. The river

IEEE Electrification Magazine - December 2019

Table of Contents for the Digital Edition of IEEE Electrification Magazine - December 2019