Hindawi

Journal of Advanced Transportation

Volume 2020, Article ID 8835201, 13 pages
https://doi.org/10.1155/2020/8835201

Research Article
A Tabu Search-Based Algorithm for Airport Gate Assignment:
A Case Study in Kunming, China

Jun Bi,1,2 Zhen Wu ,1 Lei Wang,3 Dongfan Xie,1,2 and Xiaomei Zhao1,2
1
School of Traffic and Transportation, Beijing Jiaotong University, Beijing 100044, China
2
Key Laboratory of Transport Industry of Big Data Application Technologies for Comprehensive Transport,
Beijing Jiaotong University, Beijing 100044, China
3
Information Science & Technology Department, Beijing Capital International Airport Co., Ltd., Beijing, China

Correspondence should be addressed to Zhen Wu; 18120767@bjtu.edu.cn

Received 16 July 2020; Revised 1 November 2020; Accepted 4 November 2020; Published 27 November 2020

Academic Editor: Luigi Dell’Olio

Copyright © 2020 Jun Bi et al. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

An airport gate is the core resource of an airport operation, which is an important place for passengers to get on and off the aircraft
and for maintaining aircraft. It is the prerequisite for other related dispatch. Effective and reasonable allocation of gates can reduce
airport operating costs and increase passenger satisfaction. Therefore, an airport gate assignment problem (AGAP) needs to be
urgently solved in the actual operation of the airport. In this paper, considering the actual operation of the airport, we formulate an
integer programming model for AGAP by considering multiple constraints. The model aims to maximize the number of
passengers on flights parked at the gate. A tabu search-based algorithm is designed to solve the problem. In the process of
algorithm design, an effective initial solution is obtained. A unique neighborhood structure and search strategy for tabu search are
designed. The algorithm can adapt to the dynamic scheduling of airports. Finally, tests are performed using actual airport data
selected from Kunming Changshui International Airport in China. The experimental results indicate that the proposed method
can enhance the local search ability and global search ability and get satisfactory results in a limited time. These results provide an
effective support for the actual gate assignment in airport operations.

1. Introduction time, manpower, land, etc., this method is not practical.

Secondly, we can optimize the gate assignment scheme,
The gate is an important resource for the airport and an which can improve the utilization efficiency of airport re-
important place for aircraft to receive ground services. The sources and reduce airport operating costs. Most airport gate
aircraft can park at the gate and the apron. For aircraft assignments are made manually and by simple computer-
parked at the gate, passengers can board and disembark aided methods by dispatchers based on personal experience.
directly through the covered bridge. For airplanes parked at This makes the gate assignment work under the condition of
the apron, passengers need to get on and off the plane by limited gate resources costly and inefficient.
shuttle bus; passengers need to use the ferry to get on and off The gate assignment plan is related to the operational
the plane. This way is not safe and convenient for passengers. safety of the airport. As a place to provide passenger and
However, the number of gates is limited, so not all flights can cargo transportation services for aircraft, there are usually
be parked at the gate. At present, there are two main ways to multiple aircraft and ground working vehicles on the airport
solve the shortage of airport gate resources. Firstly, we can parking lot for ground operations at the same time, which is
directly increase the resources of hardware facilities such as prone to safety accidents. Therefore, parking as many flights
expanding airports and aprons. Because the airport’s various as possible at the gate can reduce the use of vehicles and
hardware facilities are also impossible to expand indefinitely, reduce the probability of accidents. According to the reg-
and the expansion of the airport requires a lot of capital, ulations of the International Air Transport Association
2 Journal of Advanced Transportation

(IATA) [1], the proportion of passengers who board and assigned gates and the smallest total walking distance (or
disembark the plane over the boarding bridge should be time) of passengers, and uses Perato's simulated annealing
90%–95%. Whether passengers board or disembark through algorithm to solve. Marinelli et al. [11] aim to minimize
the boarding bridge is an important factor affecting pas- the total walking distance of passengers and minimize the
senger satisfaction. Since the number of flights that can be number of flights at parking apron, and uses a meta-
parked at the gate is much larger than the number of gates, heuristic algorithm to solve it. Mokhtarimousavi et al. [12]
how to assign the gate to maximize the number of people used the second version of Nondominated Sorting Ge-
boarding and disembarking through the bridge is an im- netic Algorithm (NSGA-II) to solve the problem with the
portant point in AGAP. The gate assignment is mainly a goal of minimizing the total walking distance of pas-
process of allocating inbound and outbound flights to sengers. Liu and Chen [13] proposed an optimization
various gates by the dispatch center under the conditions of model that considers operational safety constraints and
flight departure and arrival times, aircraft types, and the size solved them using genetic algorithms. Zhao and Cheng
of parking spaces. Ensure the normal operation of flights and [14] proposed a mixed integer model and designed a
maintain normal order at the airport. colony algorithm to solve this problem. Wang et al. [15]
The paper is organized as follows. A comprehensive proposed an optimization model based on the charac-
analysis of the existing literature is given in Section 2. In teristics of flights and airport gates and solved it using
Section 3, a model of aircraft allocation considering security immune genetic algorithms. Benlic et al. [16] established
constraints is constructed. In Section 4, a tabu search al- multiple objective functions and proposed a heuristic
gorithm based on the new neighborhood search structure is algorithm based on the breakthrough local search (BLS)
designed as well as a dynamic scheduling algorithm. In framework to solve this problem. Yan et al. [17] solved
Section 5, the actual airport data is used to verify and analyze AGAP by using a random boarding gate assignment
the algorithm. Finally, conclusions are proposed in Section 6 model and real-time assignment rules, as well as two
and future research directions are discussed. punishment adjustment methods. Lim and Rodrigues [18]
proposed a model with a time window and then solved it
using Tabu Search and Memetic Algorithms. Ding and
2. Literature Review Lim [19] designed a tabu search algorithm to solve the
Due to the continuous development of the aviation field, problem of the shortest distance traveled by passengers
research on boarding gates has continued to increase in with assigned gate. Noyan and Seker [20] added random
recent years. There are also many research methods mainly interference to establish a large-scale mixed integer
using heuristic algorithms and precise algorithms. programming problem and used a tabu search algorithm
For the exact solution methods, Yan and Huo [2] used to solve the problem.
the travel distance of the passenger as the objective In addition to applications in airports, optimization
function to establish a 0-1 integer programming model algorithms are also used in other fields. Zuo et al. [21] took
and then used the branch and bound method to solve the patients' flow cost and departments' closeness as the
model. Bolat [3–5] used the gate free time equilibrium as objective function and designed a tabu search algorithm,
the objective function to establish the gate allocation which contains a penalty function to deal with infeasible
model and used the branch and bound method to solve solutions. Wang et al. [22] proposed a competitive
this problem. Mangoubi and Mathaisel [6] take the memetic algorithm which includes the search operator in
walking distance of the transit passengers as the objective the variable neighborhood search framework and the
function and use the linear relaxation method of the simulated annealing strategy to solve the capacitated
mixed integer programming and the heuristic algorithm green vehicle routing problem. Xu et al. [23] used the
to solve the problems. Diepen et al. [7] proposed a new Delaunay-triangulation-based Variable Neighborhood
integer programming model and solved this linear re- Search algorithm to solve the large-scale A colored
laxation problem through column generation. traveling salesman problem. Wang et al. [24] took the
As the scale of flights is getting larger and larger and number of unserved requests, total traveling cost, and the
cannot be solved with accurate solutions, heuristic al- workload deviation as objective functions and proposed a
gorithms are increasingly used to solve this problem. For multiobjective iterated local search algorithm with
the heuristic algorithms, Cheng and Ho [8] use multiple adaptive neighborhood selection to solve this problem.
metaheuristic algorithms to solve the problem of shortest For the gate allocation problem, establishing an ef-
distance traveled by passengers. Deng and Zhao [9] fective gate allocation model is crucial to the actual op-
established a multiobjective aircraft allocation model and eration of the airport. The gate assignment model
solved it using an improved PSO algorithm. Drexl and proposed by some scholars and experts mainly has the
Nikulin [10] established a multitarget gate assignment following three problems. First, some constraints of the
model that includes the smallest number of flights without actual operation of the airport have not been fully
Journal of Advanced Transportation 3

considered; for example, flights parked at two adjacent (1) Each flight is only assigned to one gate:
gates cannot enter and leave at the same time. This ng
constraint is essential for the safe operation of the airport, 􏽘 xij � 1, ∀i ∈ F, j ∈ G, (2)
but most of the literature did not take this constraint into j�1
account. Secondly, the number of test cases in most lit-
eratures is relatively small or simulation data is used, where F is the set of flights and G is the set of gates.
which is not enough for the actual situation. Third, most
papers only considered the state where all the gates are (2) There must be a safety interval between two adjacent
initialized to empty and did not consider the real-time flights at the same gate:
changes in the status of the gates. In this study, a new nfc
neighborhood search algorithm is proposed with the 􏽘 xkj ≤ 1, k ∈ FC, (3)
objective of maximizing the number of passengers on k�1
flights parked at the gate under the consideration of se-
curity constraints. The gate assignment model was tested where FC represents a collection of flights that
using a large amount of data from Kunming Changshui conflict with each other in time and nfc represents
International Airport in China. the set size.
(3) Flights between two adjacent gates cannot enter and
3. Problem Formulation leave the gate at the same time, and a safety interval is
required between them:
The airport gate assignment has significant impacts on the nng nfc
quality of airport operations. Reasonable boarding gate
assignments can improve the safety and efficiency. Be- 􏽘 􏽘 􏼐xij + xkm 􏼑 ≤ 1, ∀i ∈ F, k ∈ FC, m ∈ NG,
m�1 k�1
cause of the limited number of gates, it is difficult to assign
all the flights in them. Furthermore, the size of the gates (4)
also limits the types of flights that can be parked. Two
aircraft on the same gate need to have a safe time interval. where NG is the set of gates adjacent to gate j and
Due to security considerations, gates in some areas cannot nng is the size of the gates.
enter and leave flights at the same time. As remote stands (4) The attributes of the flight (domestic/international)
are far away from the terminal, passengers on flights need to match the gate:
parked at remote stands need to take a shuttle bus to get
on and off the plane, which will seriously affect passenger xij � 0, ∀i ∈ F, ∀j ∈ CGi , (5)
satisfaction. Therefore, the goal is to maximize the
number of people boarding and disembarking using the where CGi represents the set of gates that cannot
boarding bridge, and to make full use of the gate to match flight i in the attributes of the flight (domestic/
improve the utilization of airport resources. international).
(5) The type of the gate and the type of the aircraft need
to match:
3.1. Objective Function. Maximize the total number of
passengers on flights parked at the gate: xij � 0, ∀i ∈ F, ∀j ∈ CGi , (6)

nf ng
where CGi represents the set of gates that cannot
max 􏽘 􏽘 xij ∗ gj ∗ pi , (1) match flight i in the attributes type.
i�1 j�1
(6) 0-1 variable constraints:
where xij indicates whether flight i is assigned to gate j;
xij ∈ {0, 1}. (7)
when the flight is only assigned to a gate, the value of xij is
equal to 1. Otherwise, the value of xij equals 0. gj indicates
whether there is a boarding bridge at gate j; when gate has a
covered bridge, the value of gj is equal to 1. Otherwise, the 4. A Tabu Search Algorithm
value of gj equals 0. nf represents the total number of
AGAP is an NP-hard problem [25]. As the number of flights
flights. ng represents the total number of seats. pi represents
increases, it is difficult to obtain accurate solutions using
the number of passengers on flight i.
some exact solution methods. Unlike conventional solution
methods, the tabu search algorithm is an intelligent search
3.2. Constraints. Constraints are mainly generated accord- algorithm that simulates the thinking of people. That is,
ing to some restrictions in the actual operation of the airport people will not search for the places that they have searched
and then described using mathematical expressions. The for, but search for other places. If they are not found, they
purpose is to make the calculation result meet the practical can search for the places they have already been. The tabu
limit and improve the operation efficiency. The constraints search algorithm starts from an initial feasible solution. It
in the model are listed as follows. selects a series of specific search directions based on trial and
4 Journal of Advanced Transportation

selects the movement which could improve the objective Gate

value furthest. In order to avoid falling into the local optimal
solution to the tabu search, a flexible “memory” technique is Num Flight 3
used, which records the optimization process that has been Gate2 spij nspij
performed and guides the next search. These are achieved by
establishing a tabu table. The tabu list stores the movements Gate1 Flight1 Flight2
achieved in the last few iterations. In the current iteration,
the movement in the tabu list is prohibited, so that the
algorithm can be prevented from revisiting the solution that
has been accessed in the previous iteration and avoid falling Start End Time
time time
into the local optimal solution. In order to avoid missing the
optimal solution, the tabu search also has a strategy of Figure 1: Structure of the solution.
“aspiration criterion.” Compare the candidate solution of
one iteration with “best so far.” If the target value is better
than “best so far,” you can unban the candidate solution, 65000
replace “best so far” with the current optimal solution, and
then add the tabu list and update tabu list. 60000

Number of people
The most important factors affecting the performance of 55000
tabu search are the quality of the initial solution, the design of
the neighborhood structure, and the processing of the tabu 50000
list. Therefore, getting a good initial solution, designing a
suitable neighborhood structure, and designing an effective 45000
tabu list are crucial. The structure of the solution to the airport
40000
gate assignment problem is shown in Figure 1; that is, each
boarding gate will be occupied by multiple non-conflicting 35000
flights under the condition that the constraints are met, where 120 140 160 180 200 220 240
spij represents the time interval between adjacent flights on Number of flights
the same gate and nspij represents the time interval between
flights on adjacent gates. Figure 2: Correlation.

Gate
4.1. Initial Solution. Figure 2 shows that the number of num
passengers on flights parked at the gate and the number of Flight3 Assign results by
Gate1
flights stopping at the gates are positively correlated. Assign arrival time
gates to flights in accordance with certain rules, and make as
Assign results by
few flights as possible to the apron, so that you can get a Gate1 Flight1 Flight2 departure time
better initial solution. The following two rules are used to
assign gates, and the literature [3] shows that, in the case of
partial constraints, more flights can be parked on the gate Start End Time
through such rules. time time

(1) Sort the flights according to the corresponding de- Figure 3: Gate assignment in two ways.
parture time.
(2) For each flight, you need to retrieve each gate and made as small as possible, so that the idle time of the gate
then find a gate that can be put into the flight. Then, can be reduced, thereby improving the utilization rate of
the corresponding time interval spij is compared, the gate.
where spij is the time interval of adjacent flights of
the same gate, the gate corresponding to the smallest
4.2. New Neighborhood Search Methods. The objective
spij is selected, and the flight is placed on the gate.
function of the model is to maximize the total number of
For the first rule, manual assignment is usually based on passengers on flights parked at the gate. Due to the initial
the arrival time of the flight. If the assignment is based on solution obtained, only one-by-one allocation is performed
the departure time of the flight, it can avoid the situation in the order of flight departure time, and there are con-
that only a few flights can be allocated when the gates are straints between adjacent gates, so two problems arise. The
allocated according to the arrival time in Figure 3. first problem is as shown in Figure 2. Assume that the
Therefore, the number of flights docked at the apron can be number of flight2 is greater than the number of flight1. This
reduced, and there is also a certain chance to increase the gate will be assigned flight1 directly on gate1 according to
total number of passengers on flights parked at the gate. For the time of arrival. This will cause flight2 not to be assigned
the second rule, while ensuring a safe interval, the time to a gate, so it will as a result; the objective function is
interval between adjacent flights at the same gate can be reduced. To solve this problem, this paper defines a “flight
Journal of Advanced Transportation 5

exchange” operation. Another problem is that, after the Gate

exchange of flights, there may be more idle time on the
corresponding gate, and more flights may be put on this gate. Num Flight2
This operation is defined as the “flight insert” operation.
Gate1 Flight1 Flight3
4.2.1. Flight Exchange. The flights scheduled to be parked on
the apron are screened for the gates that can be put in
through various constraints, the flights that conflict with the
Start End Time
flight at the corresponding gates are found, and the flights time time
that conflict with them are exchanged. Define this exchange
as (fa ) ⟷ (gi |(fb , fc , . . .)), where fa is the flight assigned Figure 4: Status of flight exchange.
to the apron, gi is the gate number, and (fb , fc , . . .) rep-
resents the set of conflicting flights at the corresponding
Gate Flight2
gate. As shown in Figure 4, exchanging flight1 and flight2 num
can be expressed as (flight1)↔(gate1|(flight2)). Through Flight3
this exchange, flights with a larger number of people can be spk
allocated to the gate thereby obtaining a better solution.
Gate1 Flight1

4.2.2. Flight Insert. The result after the flight exchange may
produce the result as shown in Figure 5. After flight1 and
flight3 are exchanged, the free time spk will be generated. If Start End Time
there is a flight similar to flight2 on the apron, corresponding time time
gates can be inserted under various constraints; then, the Figure 5: Status of flight insert.
flight with the largest number of people from these flights
needs to be inserted into this gate. Define this exchange as
(fa ) ⟶ (gi ). Through this exchange, you can increase the greater than the historical optimal value; if it is satisfied,
number of flights and the number of people on the gate and then replace the historical optimal solution with the
obtain a better solution. current solution, and then go to Step 2; or, go to Step 4.
Step 4. Update tabu list. Determine whether the exchange
object corresponding to the candidate solution exists in
4.3. Tabu Short-Term Memory. In order to prevent the loop
the tabu list. If it does not exist, use the candidate solution
and the local optimization from appearing in the search
as the new current solution, put the corresponding ex-
process, a tabu list is designed. The tabu objects in this
change object in the tabu list, and go to Step 5. Con-
article are mainly for the process of “flight exchange.” If a
versely, perform the m � m − 1 operation on the tabu
neighborhood is generated by the transformation
length of the corresponding tabu object in the tabu list,
(fa ) ⟷ (gi |(fb , fc , . . .)), then this exchange will be
and then reselect the suboptimal solution from the
added to the tabu table and a tabu length l will be added to
candidate set, and then perform this step again.
the tabu object. This process has been forbidden until l � 0
and the tabu object is deleted from the tabu list, or when Step 5. Determine the end condition. Determine
the “aspiration criterion” is met; the tabu object will not whether the algorithm has reached the maximum
work. number of iterations, or the optimal solution of each
iteration remains unchanged for several consecutive
times; if so, the algorithm ends and the optimal solution
4.4. Tabu Search Algorithm Steps is output; otherwise, go to Step 2.
Step 1. Initialize the tabu search algorithm. Initialize the
tabu list as empty, use the method in this paper to
4.5. Dynamic Allocation Algorithm. During the actual op-
generate the initial feasible solution, and give the tabu
eration of the airport, the state of the gates must not be
search algorithm parameters.
empty and will change in real time. The method mentioned
Step 2. Get neighborhood solution. A neighborhood in the existing literature does not apply to the actual state of
solution is generated for the current solution according all airports. Therefore, on the basis of the above tabu search
to the above-mentioned method, and a part of the algorithm, it is improved to apply to the actual operating
candidate solutions is selected from them, and the status of the airport.
candidate solutions are sorted according to the ob- There are two statuses for the assigned flight in the actual
jective function from large to small. operation of the airport. One state is that the gate assigned to
Step 3. Choose a better solution. Select the candidate the flight cannot be changed, and the other is that the gate
solution corresponding to the largest objective function has been assigned to the flight but the gate can be changed.
value, and first determine whether the amnesty criterion In response to this situation, the improvements to the above
is met, that is, whether the objective function value is tabu search are as follows:
6 Journal of Advanced Transportation

(1) In the process of obtaining the initial solution, for the shown in Figure 6; the optimal value reaches the maximum
initial gates state, all assigned flights need to be added when l � 18 and no longer increases with the increase of l.
to the corresponding seats, and the corresponding
state is added for each flight. There are three main
5.3. Case Analysis. We selected 356 flights with arrival and
statuses, including the inability to change the camera
departure times within one day, with a total of 98,606 people,
position, the possibility to change the camera po-
and tested the algorithm. According to the actual allocation
sition, and the unassigned camera position.
result of the airport, the proportion of passengers on the
(2) During the flight exchange and flight insert process, flight stopping at the gate on that day reached 85.63%. The
it is necessary to determine the status of the flight. initial solution obtained first includes 342 flights assigned
When the flight status cannot be changed, the ex- to the gate, 14 flights assigned to the apron, and a total of
change is stopped directly. When the flight status is 94430
exchangeable, exchange is performed when the ex- √�� people
√��� assigned to the gate. Set tabu length
l � m � 342 ≈ 18. The results obtained by the tabu
change can improve the objective function. It can be search algorithm are shown in Figure 7 and Table 3. In the
exchanged when the flight status is unassigned. end, there were 96,197 people assigned to the gate at 346
flights, an increase of 1767 people compared to the initial
5. Experimental Results and Analysis solution. Figure 8 represents the iterative process of the
algorithm. At the beginning of the algorithm, the number of
5.1. Test Data. This paper uses actual operating data from people grew faster and finally reached the optimal value at
Kunming Changshui International Airport, including 65 the 128th time. The result made the proportion of passengers
gates and 356 flights per day. The detailed information of the on flights parked at the gate on this day reach 97.56%.
gate is shown in Table 1, where gateno indicates that the Through this case, it can be proved that the boarding gate
number of the gate is from 101 to 165; mdl indicates the assignment scheme obtained by the algorithm in this paper
types of aircraft that can be docked, including C, D, and E improves the total number of passengers on flights parked at
types. The three types are backward compatible. For ex- the gate and reduces the airport’s operating costs.
ample, E-type gates can park C and D aircraft. Nation in-
dicates whether the type of parked flight is international or
domestic. Similarly, the seat that can park international 5.4. Case Analysis of Dynamic Allocation. For a day’s flight,
flights can also park domestic flights. Flight details are the flight that arrived yesterday and the flight that departed
shown in Table 2, where planenumber is the unique iden- today and the flight that arrived and departed early need to
tifier of the aircraft; arrivetime and leavetime indicate the be allocated in advance. The allocation of this flight to the
time of flight in and out of the airport; arrivepeople and new day is an initial state and needs to be considered
leavepeople indicate the number of flights in and out of the separately. The allocation of these flights to the new day is an
airport; and nation indicates the domestic and international initial state and needs to be considered separately. Select 370
attributes of the flight. I indicates an international flight, and flights in one day, 41 of which are already allocated flights.
D indicates a domestic flight. For example, I/D indicates that The results obtained using the above dynamic allocation
entering the airport is an international flight and leaving the algorithm are shown in Figure 9. The green flight in the
airport is a domestic flight. As long as one of the flights is an figure represents the assigned result. The yellow flight in the
international flight, the aircraft should be placed in a place figure represents the result of algorithm allocation. In the
where international flights can be parked. mdl indicates the end, 313 flights were allocated to the gate, and the result
type of aircraft. made the proportion of passengers on flights parked at the
gate on this day reach 89.65%. The decrease in the bridge rate
is mainly due to the initial state of the aircraft limiting the
5.2. Parameter Settings. The parameters of tabu search are scope of understanding. Through this case, it can well
set as follows. In order to ensure that there is no impact represent the application of this algorithm in the actual
between flights, the time interval for the actual allocation of dispatch of the airport and has a good reference for the
gates at the airport is adopted. Set the security time interval actual dispatch of the airport.
spij between flights of the same gate to 20 minutes, and the
security time interval nspij between flights between adjacent
5.5. Comparative Analysis. We implement these algorithms
gates to 20 minutes. In order to ensure the efficiency of time
in Python and run them in Windows machines with i7-8700
and that the algorithm can reach a better value within 200
CPU and 16G memory. They are tested with different data,
times, the maximum number of iterations is Tmax � 200. The
and then the different test results are analyzed and researched.
most important parameter is the tabu length l. If the tabu
The algorithm is compared with the precise algorithm and
length is too short, it may fall into the local best advantage,
genetic algorithm. The calculation time and the quality of the
which will cause a cycle. The length of the tabu is too long,
solution are analyzed to verify the rationality and effectiveness
and all the candidate solutions are tabu, which will cause the
of the tabu search algorithm in the paper.
calculation time to increase. When 356 flights were selected
for testing, 342 flights were
√�� assigned
√��� to gate. According to Test1 (363 Flights and 65 Gates). In order to test the
the size of the data, l � m � 342 ≈ 18 was selected. The performance of the algorithm when solving large-scale
optimal value obtained near the tabu length of l � 18 is data, 65 gates of 363 flights were used for testing.
Journal of Advanced Transportation 7

Table 1: Detailed information of gates in hub airport.

Gateno mdl Nation
101 C I
102 C I
103 D I
104 D I
··· ··· ···
163 E D
164 E D
165 D D

Table 2: Detailed information of flights in hub airport.

Planenumber Arrivetime Leavetime Arrivepeople Leavepeople Nation mdl
B8395 2019-11-24 00 : 05 : 00 2019-11-24 01 : 10 : 00 122 143 I/D C
B6142 2019-11-24 00 : 05 : 00 2019-11-24 14 : 30 : 00 77 111 D/I C
B8318 2019-11-24 00 : 10 : 00 2019-11-24 09 : 10 : 00 214 208 D/D D
··· ··· ··· ··· ··· ··· ···
B6371 2019-11-24 22 : 45 : 00 2019-11-24 23 : 50 : 00 120 48 D/I C
B1461 2019-11-24 22 : 45 : 00 2019-11-24 23 : 45 : 00 114 112 D/D C
B1565 2019-11-24 22 : 55 : 00 2019-11-24 23 : 50 : 00 131 147 D/D C

96200

96150
Optimal_value

96100

96050

96000

95950

14 16 18 20 22 24 26 28
Tabu_length
Figure 6: Tabu length selection process.

Test2 (363 Flights and 50 Gates). In order to test the gates will directly affect the initial solution. Comparing
efficiency and results of the algorithm when the number Test1 and Test2 and comparing Test3 and Test 4 show that
of gates is reduced so that more flights are allocated on when the number of gates decreases, the number of flights
the apron, 363 flights and 50 gates were used for testing. allocated to the gate decreases, which leads to an increase
Test3 (111 Flights and 65 Gates). In order to test the in calculation time. By analyzing the design of the domain
quality and efficiency of the solution for small-scale data, structure in this paper, let n denote the total number of
flights at peak hours of the day were selected for testing, flights, and m is the number of flights assigned to the gate
which included 111 flights and 65 gates for testing. in the initial solution. It is necessary to compare the
number of passengers on each flight parked at aprons with
Test4 (111 Flights and 50 Gates). In order to test the
those on a near seat. So, the time complexity of the al-
efficiency and quality of the solution of the algorithm
gorithm is T(n) � m ∗ (n − m). Based on the results ob-
when more flights are allocated on the apron under
tained, this formula is also verified.
small-scale data, 111 flights and 50 gates were used for
In order to further verify the effectiveness of the algo-
testing.
rithm, the algorithm in this paper is compared with other
Compare the results of the four cases in Table 4. Since algorithms. Analyze the differences between the algorithms
the initial solution obtained is directly obtained through and their respective advantages. The above four types of test
certain rules, the number of flights and the number of data are solved by precise algorithms and genetic algorithms,
8 Journal of Advanced Transportation

165
164
163
162
161
150
159
158
157
156
155
154
153
152
151
150
149
148
147
146
145
144
143
142
141
140
139
138
137
136
The number of gate

135
134
133
132
131
130
129
128
127
126
125
124
123
122
121
120
119
118
117
116
115
114
113
112
111
110
109
108
107
106
105
104
103
102
101

0 200 400 600 800 1000 1200 1400

Time (min)
Figure 7: Gantt chart of airport gate assignment.

respectively. For the genetic algorithm, the maximum algorithm uses the cutting plane method and the complex is
number of iterations is set as 200, and the results of the used for the solution. The exact algorithm solution results
genetic algorithm solution are shown in Table 5. The exact are shown in Table 6. As the amount of data increases and
Journal of Advanced Transportation 9

Table 3: Detailed information of gate assignment result.

Gate number Flight number People
101 B8395, 9MLNT, B5271, B1211, B5797, B2375, B9973, B1723, B1565 2420
102 B7167, B6947, B8563, B5813, B5308, B6787, B2759 1776
103 VNA641, B1660, B7991, B1907 1050
104 B300L, B30C8, B1611, B20A9, B5796, B6373, B1106 1971
105 9MAFB, B1811, B207A, B6943 1188
106 B7867, B1792, B5842, B1361, B1480, B5292, B302Y 2071
107 RPC9368, B5936, B1029, B6732, B1001, B3355, B5811, B205Z 2139
108 B5292 228
109 HSBBB, B6147, B5255, RDPL34223, B1307, B300K, B6636, B8856, B6617, B9927, B5385 2873
110 MU5813, B209N, B1108 866
111 B5732, B6347, B308W, B8700, B5775 1421
112 B6736, B6146, B6249, B5823, B1004, B6143 1875
··· ··· ···
163 B302Y, B5282, B8035, B5842, B5308, B1470, B5243, B5493 2332
164 B1563, B1703, B1790 810
165 B8470, B1226, B6862, B5820, B1309, B5265, B1076 1852

96250

96000

95750
Optimal_value

95500

95250

95000

94750

94500

0 50 100 150 200 250 300 350 400

Iteration
Figure 8: Iteration process.

the number of variables increases, the number of feasible algorithms. The algorithm in this paper can get a solution
solutions will increase exponentially. When solving this with a high number of passengers on flights parked at the
problem, there will be a dimensional disaster, and no feasible gate. Although it is not necessarily the optimal solution, it
solution will be obtained. has certain value for practical applications. The genetic
According to the data in Tables 5–7 and Figures 10 and algorithm is used to solve the problem. Because the ge-
11, when the amount of data is small, using an accurate netic algorithm is random, the results obtained are also
algorithm to solve can get an accurate solution in a limited random. For example, in Test1, the results obtained
time. For example, in Test3, the number of flights parked through multiple experiments are 63 less than the results
at the gate is equal to the solution obtained by the method of the algorithm in the paper, and the calculation time is
in the article. The number of passengers on flights parked 850 s longer. For Test2 when the gate is reduced, the
at the gate has only increased by 21, but the calculation solution time of the genetic algorithm is 590 s longer than
time required has nearly doubled. In Test4, the objective the solution time of the algorithm in the paper. In short,
function value increased by 141, but the calculation time whether it is an exact algorithm or a genetic algorithm, the
increased by 11 s. When the amount of data increases, it is algorithm in this paper has advantages for the actual gate
difficult to find the optimal solution using accurate assignment.
10 Journal of Advanced Transportation

168
167
166
165
164
163
162
161
160
159
158
157
156
155
154
153
152
151
150
149
148
147
146
145
144
143
142
141
140
139
138
The number of gate

137
136
135
134
133
132
131
130
129
128
127
126
125
124
123
122
121
120
119
118
117
116
115
114
113
112
111
110
109
108
107
106
105
104
103
102
101

0 200 400 600 800 1000 1200 1400

Time (min)
Figure 9: Gantt chart of airport gate dynamic assignment.

Table 4: Information for comparison of results.

Initial solution (number of flights parked at the gate ∗ number of flights Running time The number of people added by the
Case
parked at the apron) (s) optimal solution
Test1 340 ∗ 23 131 1178
Test2 288 ∗ 79 360 2976
Test3 103 ∗ 8 20 1079
Test4 86 ∗ 35 41 1794
Journal of Advanced Transportation 11

Table 5: Results of the genetic algorithm.

Case Number of flights Number of people Running time (s)
Test1 342 95193 986
Test2 280 81796 950
Test3 103 24087 173
Test4 87 21368 158

Table 6: Results of the cutting plane method.

Case Number of flights Number of people Running time (s)
Test1 — — —
Test2 — — —
Test3 103 24175 59
Test4 88 21576 52

Table 7: Results of individual cases.

Number of people in the initial Number of flights in the initial Number of people in the final Number of flights in the final
Case
solution solution solution solution
Test1 94078 340 95256 341
Test2 79465 288 82441 289
Test3 23075 101 24154 103
Test4 19641 86 21435 88

Time consumed Time consumed

1000 986 950

800
800

600 600
Time (s)

Time (s)

400 400 360

200 200
131

0 0
TS GA Exact model TS GA Exact model
Algorithm Algorithm
(a) (b)
Time consumed Time consumed
175 173 160 158

150 140

120
125
100
Time (s)

Time (s)

100
80
75
59 60 52
50 41
40
25 20
20

0 0
TS GA Exact model TS GA Exact model
Algorithm Algorithm
(c) (d)

Figure 10: The comparison of time consumption. (a) Test1. (b) Test2. (c) Test3. (d) Test4.
12 Journal of Advanced Transportation

The optimal value The optimal value

100000 82441 81796
95256 95193
80000

80000 70000

60000

Number of people
Number of people

60000 50000

40000
40000
30000

20000
20000
10000

0 0
TS GA Exact model TS GA Exact model
Algorithm Algorithm
(a) (b)
The optimal value The optimal value
25000 24154 24087 24175 21435 21368 21576
20000
20000
Number of people
Number of people

15000
15000

10000
10000

5000
5000

0 0
TS GA Exact model TS GA Exact model
Algorithm Algorithm
(c) (d)

Figure 11: The comparison results of objective values. (a) Test1. (b) Test2. (c) Test3. (d) Test4.

6. Conclusion algorithms and precise algorithms and more practical than

them, especially when the amount of data is relatively large.
In this paper, the total number of flights docked at the gate The constructed gate assignment model and algorithm can
is used as the objective function. According to some re- significantly increase the total number of passengers on
strictions in the actual operation of the airport, some flights parked at the gate, thereby improving airport service
constraints are constructed, and a mathematical model of levels and passenger satisfaction. The proposed algorithm
gate assignment is established. We have designed a better can provide an effective reference for airport gate
initial solution algorithm and a special domain search al- allocation.
gorithm which includes two operations: flight exchange Since the flight data of the airport changes in real time,
and flight insertion. We have improved the related algo- our allocation plan will also change. The algorithm proposed
rithm which has been applied to dynamic gate allocation. in this paper can only recalculate this problem, but in actual
The tabu search algorithm designed in this way can improve scheduling, it cannot change too many scheduling plans. It is
the accuracy and efficiency of the solution. The practicality necessary to further study the change of the flight and
of the algorithm is analyzed through actual flight data and propose a more effective method to solve the gate reas-
gate data of various airports, and then the genetic algorithm signing problem.
and the cutting plane method are used to solve the problem.
The obtained solutions are compared to analyze the ef- Data Availability
fectiveness of the algorithm. The experimental results show
that the optimization performance of the algorithms The data used to support the findings of this study are in-
proposed in this paper is generally better than genetic cluded within the article.
Journal of Advanced Transportation 13

Conflicts of Interest [16] U. Benlic, E. K. Burke, and J. R. Woodward, “Breakout local

search for the multi-objective gate allocation problem,”
The authors declare that they have no conflicts of interest. Computers & Operations Research, vol. 78, pp. 80–93, 2017.
[17] S. Y. Yan and C.-H. Tang, “A heuristic approach for airport
Acknowledgments gate assignments for stochastic flight delays,” European
Journal of Operational Research, vol. 180, no. 2, pp. 547–567,
This research was supported by an Open Fund of Key 2007.
Laboratory of Airports Cluster Intelligent Operation under [18] A. Lim, B. Rodrigues, and Y. Zhu, “Airport gate scheduling
with time windows,” Artificial Intelligence Review, vol. 24,
Grant no. KLAGIO20180801.
no. 1, pp. 5–31, 2005.
[19] H. Ding, A. Lim, B. Rodrigues, and Y. Zhu, “Aircraft and gate
References scheduling optimization at airports,” in Proceedings of the
37th Hawaii International Conference on System Sciences, Big
[1] E. Piazza, “Increasing airport efficiency: injecting new tech- Island, HI, USA, January 2004.
nology,” IEEE Intelligent Systems, vol. 17, no. 3, pp. 10–13, [20] N. Noyan and M. Seker, “Stochastic optimization models for
2002. the airport gate assignment problem,” Transportation Re-
[2] S. Yan and C.-M. Huo, “Optimization of multiple objective search, Part E. Logistics and Transportation Review, vol. 48,
gate assignments,” Transportation Research Part A: Policy and no. 2, pp. 438–459, 2012.
Practice, vol. 35, no. 5, pp. 413–432, 2001. [21] X. Zuo, B. Li, X. Huang et al., “Optimizing hospital emergency
[3] A. Bolat, “Assigning arriving flights at an airport to the department layout via multiobjective tabu search,” IEEE
available gates,” Journal of the Operational Research Society, Transactions on Automation Science and Engineering, vol. 16,
vol. 50, no. 1, pp. 23–34, 1999. no. 3, pp. 1137–1147, 2019.
[4] A. Bolat, “Models and a genetic algorithm for static aircraft- [22] L. Wang and J. Lu, “A memetic algorithm with competition
gate assignment problem,” Journal of the Operational Re- for the capacitated green vehicle routing problem,” IEEE/CAA
search Society, vol. 52, no. 10, pp. 1107–1120, 2001. Journal of Automatica Sinica, vol. 6, no. 2, pp. 516–526, 2019.
[5] A. Bolat, “Procedures for providing robust gate assignments [23] X. Xu, J. Li, and M. Zhou, “Delaunay-triangulation-based
for arriving aircrafts,” European Journal of Operational Re- variable neighborhood search to solve large-scale general
search, vol. 120, no. 1, pp. 63–80, 2000. colored traveling salesman problems,” IEEE Transactions on
[6] R. Mangoubi and D. F. Mathaisel, “Optimizing gate assign- Intelligent Transportation Systems, vol. 99, pp. 1–11, 2020.
ments at airport terminals,” Transportation Science, vol. 19, [24] J. Wang, Y. Sun, Z. Zhang, and S. Gao, “Solving multitrip
no. 2, pp. 173–188, 1985. pickup and delivery problem with time windows and man-
[7] G. Diepen, J. W. Smeltink, J. A. Hoogeveen, and power planning using multiobjective algorithms,” IEEE/CAA
J. M. v. d. Akker, “Finding a robust assignment of flights to Journal of Automatica Sinica, vol. 7, no. 4, pp. 1134–1153,
gates at amsterdam airport schiphol,” Journal of Scheduling, 2020.
vol. 15, no. 6, pp. 703–715, 2012. [25] A. Bouras, M. A. Ghaleb, U. S. Suryahatmaja, and
[8] C.-H. Cheng, S. C. Ho, and C.-L. Kwan, “The use of meta- A. M. Salem, “The airport gate assignment problem: a survey,”
heuristics for airport gate assignment,” Expert Systems with The Scientific World Journal, vol. 2014, Article ID 923859,
Application, vol. 39, no. 16, pp. 12430–12437, 2012. 27 pages, 2014.
[9] W. Deng, H. Zhao, X. Yang, J. Xiong, M. Sun, and B. Li,
“Study on an improved adaptive PSO algorithm for solving
multi-objective gate assignment,” Applied Soft Computing,
vol. 59, pp. 288–302, 2017.
[10] A. Drexl and Y. Nikulin, “Multicriteria airport gate assign-
ment and pareto simulated annealing,” IIE Transactions,
vol. 40, no. 4, pp. 385–397, 2008.
[11] M. Marinelli, M. Dell’Orco, and D. Sassanelli, “A meta-
heuristic approach to solve the flight gate assignment prob-
lem,” Transportation Research Procedia, vol. 5, pp. 211–220,
2015.
[12] S. Mokhtarimousavi, D. Talebi, and H. Asgari, “A non-
dominated sorting genetic algorithm approach for optimi-
zation of multi-objective airport gate assignment problem,”
Transportation Research Record: Journal of the Transportation
Research Board, vol. 2672, no. 23, pp. 59–70, 2018.
[13] S. Liu, W.-H. Chen, and J. Liu, “Robust assignment of airport
gates with operational safety constraints,” International
Journal of Automation and Computing, vol. 13, no. 1,
pp. 31–41, 2016.
[14] H. Zhao and L. Cheng, “Ant colony algorithm and simulation
for robust airport gate assignment,” Mathematical Problems
in Engineering, vol. 2014, Article ID 804310, 7 pages, 2014.
[15] L. Wang, Q.-L. Zhu, and X.-F. Xu, “Optimised assignment of
airport gate configurations using an immune genetic algo-
rithm,” Mathematical Structures in Computer Science, vol. 24,
no. 5, Article ID e240517, 2014.