IEEE Computational Intelligence Magazine - May 2022 - 94

maintenance work, respectively. Ip et al.
[134] focused on the scheduling of
maintenance tasks in an airport. The
maintenance tasks of airplanes should be
assigned to several maintenance teams,
and the maintenance sequence should
also be determined. They proposed a
GA-based algorithm to minimize the
total delay of all flights. Eltoukhy et al.
[135] considered disruptions when planning
the maintenance schedule, and
proposed an ACO-based algorithm to
minimize the delay of flights. The proposed
algorithm is validated based on
real-world data from a major Middle
Eastern airline.
4) Crew Scheduling
Crew scheduling consists of two parts:
pairing and rostering [136]. With a set of
flights that an airline plans to operate, a
pairing is a flight sequence in which the
first flight begins from the base of the
crew, the other flights start from the destination
of their previous flight, and the
last flight should return to the base of
the crew. The airline must construct
pairings that can cover all planned
flights; this situation describes the crew
pairing problem. Each pairing is operated
by a crew, and there are two
important constraints in the construction
of pairings. One is that the duty
time of a crew over a given period
cannot exceed the maximum working
hours in law. The other is that the crew
members, particularly the pilots, must
have sufficient rest time between two
flights. The resulting pairings are then
assigned to airline employees to determine
the crew members for each pairing,
which describes the crew rostering
problem. The primary objective of
crew pairing is to minimize the
employment cost of crews, while crew
rostering typically considers the fairness
of workload assignment and the
satisfaction of employees.
To solve the crew pairing problem,
Deng and Lin [137] proposed an ACObased
algorithm. Rather than the direct
construction of pairings, Deveci and
Demirel [138] first enumerated all possible
pairings via Depth-First-Search;
then, a GA-based algorithm is proposed
to pick up certain pairings that can
cover all planned flights and with minimal
cost. To solve the crew rostering
problem, Zhou et al. [139] constructed a
multiobjective model that simultaneously
considered the fairness and satisfaction
of the workload assignment. The
proposed multiobjective ACO algorithm
is based on the multiple populations for
multiple objectives framework [140],
[141], which uses two ant colonies to
optimize the two objectives. Chen and
Chou [142] considered disruptions in
crew rostering, such as mechanical malfunctions
of airplanes and crew member
sickness. If a disruption occurs, the rostering
plan must be rearranged. The
overall objective is to minimize the
changes between the new and original
plans. Six optimization objectives are
considered, and the proposed algorithm
is based on NSGA-II.
The studies mentioned above only
deal with one of the two problems in
crew scheduling. However, crew pairing
and crew rostering are not two independent
problems. The crew pairing is the
premise of crew rostering, i.e., solving
the crew rostering problem requires a set
of pairings as input. Due to the large
scale of data and the complex features of
these two problems, it is very challenging
to solve these two problems simultaneously.
Researchers have made some
attempts. Souai and Teghem [143] proposed
a GA-based algorithm that contained
three heuristic strategies to make
the solution feasible. Chen et al. [144]
encoded the solution by two matrixes
that represented the pairing and rostering
schemes, respectively. Five optimization
objectives are considered: maximizing
the day-offs of employees and
minimizing the number of deadheads,
the overnight cost, the waiting time
between two consecutive flights, and the
workload deviation among employees. A
deadhead represents that a crew member
should take a flight as a passenger to get
to a specific destination for work. Overnight
cost refers to the cost needed for
crew members to stay overnight in a
place other than their base. An improved
NSGA-II algorithm is proposed to solve
this multiobjective optimization problem.
94 IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE | MAY 2022
5) Shuttle Arranging
Airports are usually located in suburbs.
Airport shuttle service is a common
approach that makes it convenient for
people to travel from downtowns to airports,
which typically refers to the
arrangement of shuttle buses. Öner et al.
[145] proposed a simulated annealingbased
algorithm to arrange the schedule
of a fixed bus line between the airport
and city center, aiming to maximize the
profit. The travel time through shuttle
buses may be influenced by traffic conditions,
but passengers must arrive at the
airport on time to catch the flight. Thus,
Bao et al. [146] focused on optimizing
the reliability of the airport shuttle service.
The reliability refers to the probability
that shuttle buses arrive on time. A
hybrid algorithm of GA and a hillclimbing
algorithm is proposed to solve
this problem. Usually, the airport shuttle
bus lines are between airports and city
centers because there are many potential
passengers. For people who do not live
in the city center, certain airlines provide
demand-responsive services, in which
people request pick-up service via telephones
or the Internet and the airlines
arrange vehicles and routes to deliver
these people to the airport. Wei et al.
[147] used NSGA-II to optimize the
operating cost, fuel consumption, and
carbon emissions of the arrangement.
C. Intelligent Route Design
Route design includes the design of
the flight route in various scenes, and
the optimization objectives of route
design are often related to the safety of
flight, fuel consumption, and impact
on residents. Thus, route design is considered
from both the business and citizen
perspectives. There are two typical
issues to consider: noise handling and
danger handling.
The flight of airplanes will generate
much noise, which has a marked impact
on nearby residents, particularly when
airplanes fly at low altitudes. A direct
solution is designing flight routes far
away from residential areas. However,
such routes may induce more fuel consumption
that reduces the profit of the
airline. To address this problem,
IEEE Computational Intelligence Magazine - May 2022

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