IEEE Computational Intelligence Magazine - May 2022 - 87

network, traffic flow on the road can be
monitored in real time. Some studies further
used machine learning methods to
predict short-term traffic flow, in which
EC algorithms were used for parameter
optimization [28], [29]. With this information,
the signal pattern of the traffic
light can be adjusted intelligently. Traffic
light scheduling belongs to the scope of
municipal administration; thus, this application
is considered from the government
perspective.
Related studies focused on optimizing
the duration of each signal phase. A
signal phase represents the signal of each
traffic light in an intersection. Two
examples of a signal phase are shown in
Fig. 3. The decision variables for an
intersection are a sequence of signal
phases and the duration of each phase.
Typically, several intersections are simultaneously
considered to generate an efficient
traffic network. Sanchez-Medina
et al. [30] used a binary GA to solve this
problem with the optimization objective
of the average travel time for cars to
arrive at their destination. The duration
of the signal phase is restricted as an
integer for binary encoding. The cellular
automata technique is used to simulate
the traffic flow and obtain the fitness of
a solution (i.e., the average travel time).
Li et al. [31] also used GA and set the
same optimization objective (i.e., the
average travel
time) but with integer
encoding. To simulate traffic flow, a
modified Dijkstra algorithm that considers
congestion is proposed to assign a
route to each car. The integer encoded
PSO algorithm proposed in [32] maximizes
the number of cars that can reach
their destinations and minimizes the
traveling time of cars. Ferrer et al. [33]
focused on the adaptation of the traffic
light scheduling approach to highly
dynamic and uncertain traffic flow. Bi
et al. [34] proposed a multi-agent type-2
fuzzy logic control system to address
uncertainties in traffic flow, where DE
was used to optimize the parameter
configuration of the system. Bie et al.
[35] dealt with traffic light scheduling
for controlling the headway between
buses. Two buses on the same bus line
must maintain a certain distance and
time interval, but the uncertain number
of passengers makes the stop time of a
bus unpredictable. Thus, they used GA
to schedule the traffic light and the
speed of buses to maintain the regularity
of the bus line. Rather than only considering
the traffic flow of vehicles, some
studies also considered the traffic flow of
pedestrians during traffic light scheduling.
Zhang et al. [36] proposed a discrete
harmony search algorithm to optimize
the weighted sum of the delay times of
vehicles and pedestrians. In [37], the
delay times of vehicles and pedestrians
were set as two separate objectives, and
the harmony search and artificial bee
colony algorithms (both integrated with
a local search operator) were used to
solve this multiobjective model. Some of
the studies mentioned above conducted
experiments based on real-world road
networks; for example, the road network
in Saragossa, Spain is used in [30], and
the road network in Jurong, Singapore is
used in [37].
C. Intelligent Electric Vehicle
Infrastructure
Electric vehicles (EVs) are a new type of
vehicle powered by electric or traction
motors. EVs are more environmentally
friendly (e.g., can reduce greenhouse gas
emissions) than traditional petrol vehicles
[38] and are thus considered to be the
trend of future vehicle development. EV
charging requires specific facilities so new
infrastructure should be deployed to support
the usage of EVs. First, the government
and EV manufacturers should
deploy charging facilities in parking lots
and construct specific charging stations.
Intelligent charging strategies should also
be developed to reduce expenditure and
improve service quality. In addition, energy
companies should deploy a power grid
for the charging of EVs. The government
often engages in the construction and
management of EV infrastructure, related
companies want to earn profits, and citizens
benefit from the intelligent EV infrastructure
in terms of travel convenience.
Thus, the intelligent EV infrastructure is
considered from the government, business,
and citizen perspectives.
In the domain of infrastructure
deployment, Zhang et al. [39] used a
PSO-based algorithm to optimize the
deployment of charging facilities. Specifically,
the algorithm determines
the
number of fast-charging stations along a
roadside and the location and charging
capacity of each station. The optimization
objective consists of the following
four factors: investment cost, operation
and maintenance cost, electricity cost,
and time efficiency. Herein, time efficiency
refers to the time that drivers
spend in arriving at the charging facilities
and waiting for the charging process
to be complete. Zeng et al. [40] investigated
the location of parking lots, the
number of charging facilities, and the
incentive policy to optimize the profit.
Alegre et al. [41] optimized the location
(a)
(b)
FIGURE 3 Two examples of a traffic light signal phase. (a) Signal phase for the up-down direction.
(b) Signal phase for the left-right direction.
MAY 2022 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE 87

IEEE Computational Intelligence Magazine - May 2022

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