IEEE Computational Intelligence Magazine - May 2022 - 96

the solution was encoded by a set of
points and the corresponding path was
the polyline connecting these points.
Economic benefit issues refer to the
profit of shipping companies. The ports
of call during the sailing itinerary and
the total transportation task (i.e., the
number of goods loaded and unloaded
at each port) are known in advance. De
et al. [172] proposed a PSO-based algorithm
to determine ship routes, and the
optimization objective is set as the sum
of handling cost, docking cost, berthing
cost, fuel cost, total operating cost, bunkering
cost, and charges for berthing
outside the given time window. In addition,
De et al. [173] set profit and carbon
emissions as two optimization objectives.
Then, they used NSGA-II and a nondominated
sorting-based PSO to solve
the problem. Jeong et al. [174] considered
the imbalanced distribution of containers
because certain countries or
regions import more goods than they
export, while others import fewer
goods than they export. Thus, in addition
to minimizing transportation cost,
the transportation of empty containers
during navigation is also considered to
avoid the shortage or surplus of containers.
A PSO-based algorithm integrated
with a heuristic algorithm is proposed
to solve the problem. Kang et al.
[175] focused on a specific transportation
scene where a car manufacturer
planned to transport cars to other locations
by sea. The optimization objective
is to minimize transportation cost, handling
cost, and carry-forward penalties.
Carry-forward penalties occur when
certain cars produced in the current
time period cannot be transported
immediately and are postponed to the
next time period.
C. Intelligent Container Stowage
The cargo transported by ships is typically
stored in containers. The containers
are stacked on ships' decks, as
shown in Fig. 6(a). Fig. 6 (b) shows an
example that illustrates two issues in
container stowage, where each rectangle
represents a container, and the
number represents its weight. The first
issue is container movement. A ship
needs to load and unload containers in
a number of ports during a given sailing
itinerary. However, to move containers
in the lower layer (e.g., the blue
containers), the corresponding containers
in the upper layer (e.g., the yellow
containers) must be moved first. Since
the loaded and unloaded containers in
each port are known in advance, certain
studies have focused on using EC
algorithms to generate a stowage
scheme that minimized container
movements and thus reduced handling
time at berth [176], [177]. Jovanovic
et al. [178] used ACO to optimize the
handling time of quay cranes, while
Hottung and Tierney [179] proposed a
biased random-key GA to minimize
container movements.
The second issue is the stability of
ships. The containers are typically of the
same size, but the weight of cargos they
carry is not the same. Thus, the stowage
scheme should maintain the stability of
ships. For example, in Fig. 6(b), the
weight of containers on the left side is
far greater than that on the right side,
which will make the ship imbalanced and
increase the risk during navigation. Some
studies considered these two issues by
setting the weighted sum of container
movements and stability factor as the
optimization objective [180], [181]. Zhang
and Lee [182] directly considered these
two issues as two optimization objectives
and used NSGA-II as the solver.
VI. Future Research Directions
and Open Issues
In this survey, the literature on EC for
intelligent transportation in smart cities
is reviewed and discussed in detail. In
the future, EC algorithms will be promising
for solving optimization problems
of intelligent transportation in smart cities.
Several future research directions and
open issues are discussed, which include
the " problem " aspect (e.g., how to
model intelligent transportation problems
more accurately), the " algorithm "
aspect (e.g., how to enhance EC algorithms
to make them more efficient in
solving intelligent transportation problems),
and the " application " aspect (e.g.,
how to apply EC algorithms to broader
real-world scenes related to intelligent
transportation in smart cities). This section
hopes to inspire researchers in their
future work.
A. Multi-Source Data FusionEnhanced
Problem Model
In smart cities, many IoT devices (e.g.,
roadside cameras and wireless sensors)
collect multi-source transportation data
from the environment. Traditionally, raw
data related to the optimization problem
will be directly used to formulate the
problem model. However,
such an
8t
15t
(a)
10t
11t
(b)
FIGURE 6 Container stowage on ships. (a) A typical container ship. (b) An example of container
stowage on a ship.
6t
7t
3t
5t
approach has two shortcomings. First,
the scale of raw data is typically large,
which makes problem-solving challenging.
Second, raw data may be incorrect
or imprecise: unreliable data sources and
device malfunctions will cause incorrect
and imprecise data collection, respectively.
Data fusion is a promising
approach to address these shortcomings.
In data fusion, multi-source raw data are
analyzed first. Then, a subset of them is
extracted, and the extracted data is combined
and presented in a structured
manner for formulating the problem
model [183].
96 IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE | MAY 2022

IEEE Computational Intelligence Magazine - May 2022

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - May 2022