Slot allocation for a Multiple Airport System: Equity

and Efficiency
Mengying Wang, Yanjun Wang, Chaohao Liao, Daniel Delahaye, Xu Hang

To cite this version:

Mengying Wang, Yanjun Wang, Chaohao Liao, Daniel Delahaye, Xu Hang. Slot allocation for a
Multiple Airport System: Equity and Efficiency. UAS-Europe ATM Seminar, FAA-Eurocontrol, Jun
2023, Savannah, United States. �hal-04119202�

HAL Id: hal-04119202

https://enac.hal.science/hal-04119202v1
Submitted on 6 Jun 2023

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 Fifteenth USA/Europe Air Traffic Management Research and Development Seminar (ATM2023)

Slot allocation for a Multiple Airport System:

Equity and Efficiency
Mengyin Wang, Yanjun Wang Chaohao Liao, Xu Hang Daniel Delahaye
College of Civil Aviation Airspace management office ENAC Research Lab
Nanjing University of Aeronautics and Astronautics Central-South ATM Bureau Ecole Nationale de l’Aviation Civile
Nanjing 210016, China Guangzhou 510405, China Toulouse 31400, France
{mengyinw, ywang}@nuaa.edu.cn hxatc@163.com daniel@recherche.enac.fr

Abstract—Airport slot allocation aims to distribute airport slots Airport slot allocation is a challenging resource alloca-
to the airlines under given procedures and rules. The objectives tion problem, which aims to allocate airport capacity to
of slot allocation are to minimize the total displacements of slot the airlines that are operating or plan to operate in the
requests, and/or to maximize airline’s preferences. Research ef-
forts have been devoted to slot allocation in a single airport over airport. Previous studies have mainly focused on single airport
decades. Slot allocation for a Multiple-Airport System (MAS) or airport network slot allocation problems [3], [4]. The
has been less addressed. In a single airport slot allocation, airport Worldwide Airport Slot Guidelines (WASG), endorsed by
capacity is the only resource that airlines compete for; while in the International Air Transport Association (IATA), play an
an MAS, there are several resources that should be considered: important role in the allocation process. In 2012, Zografos,
airport capacity, terminal airspace and fixes capacity. This
paper proposes an MAS slot allocation model that incorporates Salouras and Madas developed a slot allocation model with a
airline fairness. The objective of the model is to minimize single objective under rules in the Worldwide Slot Guidelines
the total slot displacements of the MAS, subject to airport (WSG), which is the old version of WASG. The model
capacity constraints, fixes capacity constraints, turnaround time allocates slot request series over a scheduling season at a
constraints, and fairness constraints. An MAS comprehensive slot-coordinated airport [6]. Zografos, Androutsopoulos and
fairness indicator is developed. The trade-off between efficiency
and fairness in an MAS slot allocation problem is explored. The Madas (2018) proposed two bio-objective slot allocation
model is tested using the data from the MAS of Guangdong- models according to the WASG [5], [7]. The results show
Hong Kong-Macao Greater Bay Area. Computational results that a flight schedule with better acceptance can be achieved
show that our model can effectively allocate the airport and by scarifying certain displacements. Further, Ribeiro et al.
airspace capacity of the MAS while considering airline fairness. incorporated more details of the IATA guidelines into their
model. The model fully complied with the slot priority classes
Keywords—demand and capacity management; airport slot specified in the IATA guidelines [8]. Pyrgiotis and Odoni
allocation; multi-airport system; slot-scheduling fairness; mixed proposed a demand smoothing optimization model to solve a
integer programming; hub airport slot scheduling problem [9]. Jacquillat and Odoni
then presented a model that achieves synergistic optimization
I. I NTRODUCTION of slot scheduling at the tactical and strategic levels by
Air transportation has always been a significant contribu- controlling runway configuration and flight service rates. The
tion to a nation’s modern transportation system and economy. results show that the method can significantly alleviate airport
In 2019, the global air traffic passenger reached 4.54 billion congestion [10].
[1]. Despite being disrupted by the COVID-19 pandemic, the There are other works that study slot allocation prob-
industry is recovering quickly and is expected to fully recover lems in airport networks. Castelli et al. developed a bi-
by 2024 [2]. However, the growth of demand has placed objective model that considers the capacity constraints of
pressure on the limited infrastructure capacity of airports, all airports within the airport network. However, due to a
leading to congestion and flight delays worldwide. Increasing large amount of data and computational complexities, it failed
capacity through the development of physical infrastructures, to achieve an exact solution [11]. Then, a meta-heuristic
such as runways and terminals, takes too much time and cost. algorithm has been used by Pellegrini et al., which solved the
Feasible options in the short term to mitigate the congestion problem successfully [12]. Later, Pellegrini et al. developed
have been considered: demand and capacity management an integer planning model for the simultaneous optimization
at the strategic level or air traffic flow management at the of European network-wide slot allocation [13]. Corolli et
tactical level. This paper focuses on airport slot allocation, al. extended the single-airport slot allocation model to the
which is an administrative-based mechanism for demand and airport networks level by developing a two-stage stochastic
capacity management. It proposes a slot allocation model for programming model [14].
a Multiple Airport System (MAS) that considers the balance However, in an MAS, there are multiple airports and dozens
of operation efficiency and airlines’ fairness. of operating airlines, which makes the interactions between
neighboring airports and their shared airspace complex. The the MAS. Previous studies have focused on the identification
slot allocation problem for an MAS differs from that for a of traffic patterns of the MAS, improving airspace operational
single airport in resource types and users. An MAS involves efficiency, and optimizing arrival and departure traffic at the
multiple types of resources, such as airport capacity, terminal tactical level [20]–[23]. As the MAS resource users, airlines
airspace, and fixes capacity, while airlines in the MAS may are most concerned with the fair and reasonable allocation
not necessarily operate at all member airports. The hetero- of resources. However, the capacity of an MAS relies on
geneity in airlines’ demands further increases the complexity many resources, including runways, airport terminals, avail-
of the slot allocation problem. able airspace, and more. Therefore, fairness in the allocation
A key challenge of the slot allocation process is how of resources within an MAS should take into account multiple
to define the objectives that trade off the demands and factors.
preferences of different stakeholders. Several studies have This paper integrates fairness into the allocation of airport
attempted to balance efficiency, equity and other objectives. capacity and fix capacity among the airlines operating within
In a slot allocation problem, ‘efficiency’ often refers to the the MAS. The overall objective is to achieve the optimal
total schedule displacements; while fairness is usually defined allocation of airport slots in the MAS. The problem can be
as the distribution of displacements among airlines [15]. Nash briefly described as follows: Given airport capacities, critical
proposed and explored the efficiency-fairness equilibrium in terminal airspace capacities (i.e. fix capacities) of an MAS, as
the game problem for the first time [16]. In terms of fairness in well as operational constraints such as flying time between the
slot allocation, fairness exists in several dimensions: fairness airport and fix, and minimum and maximum flight turnaround
between flights, fairness between airlines, fairness between time, the multiple objectives of slot allocation for an MAS
airports, etc. The minimization of maximum displacement are (i) to allocate airport slots to the airlines that minimize
for single slot requests is used widely in the existing slot total displacements of airlines’ slot requests; (ii) to optimize
allocation model, which can be considered as the pursuit of airline’s fairness in the MAS throughout the whole allocating
inter-flight fairness [10]. Inter-airline equity has been taken process.
into account by Zografos and Jiang (2016, 2019) in their
single-airport slot allocation model. They construct fairness B. Assumptions
indicators depending on a flight proportionality principle [17], The following assumptions are made to simplify the prob-
[18]. Fairbrother et al. propose a ”peak request proportionality lem:
principle” by defining peak-period requests. Subsequently, 1) We only consider regular flights, excluding special
Zografos and Jiang diversified the construction of fairness circumstances such as extra flights, charter flights, and
indicators by proposing three different fairness metrics [18]. cancellations.
A common conclusion from the above studies is that a fairer 2) The term “slot” refers to a specific time interval with
flight schedule can be achieved by increasing a certain amount a minimum length of 5 minutes, rather than a precise
of schedule displacements. It can be found that most of the time point.
existing studies about fairness have been given to airlines 3) The definition of declared capacity is the number of
within a single airport. While in an MAS, consisting of slots that can be allocated to users per unit time in a
multiple types of resources, the airlines have heterogeneous coordinated airport [24].
demands at airports and fixes [19]. This further adds to the 4) The constant maximum and minimum turnaround time
complexity of the slot allocation. is set in this paper.
This paper contributes to the literature through the intro- 5) Only the flights operating within the MAS are consid-
duction of an MAS slot allocation model that (a) considers ered.
multiple types of resource including the airport capacity 6) We assume that the flight time for each flight from
and fixes capacity; (b) explore the comprehensive fairness the airport to the same fix remains constant within the
between airlines in the MAS; and (c) optimize the multiple MAS.
capacity allocation to users in a systematic way.
The remainder is organized as follows. In section II, we C. Fairness Definition
give the definition of slot allocation problem for an MAS and In this section, we introduce the concept of fairness in an
the definition of fairness. Section III formulates the MAS slot MAS. It is widely recognized that fairness is an important
allocation baseline model and fairness constraints. In Section factor that affects the results of resource allocation prob-
IV, we test our model with data from the MAS of Guangdong lems. The WASG emphasizes that “To ensure that slots are
- Hong Kong -Macao Greater Bay area. Finally, the paper allocated at congested airports in an open, fair, transparent
concludes in Section V. and non-discriminatory manner by a slot coordinator acting
independently”. Also, the slot regulation issued by the Civil
II. P ROBLEM D ISCRIPTION
Aviation Administration of China (CAAC) clearly points out
A. Brief discription that “the purpose of regulating slot management of civil
There are three main factors affecting the operation of flights, promoting the fairness, efficiency, competition, and
airports in an MAS: capacity constraints on shared arrival and integrity of the allocation of slot resources”. Interestingly,
departure fixes, limited airspace resources within the terminal a clear definition and measurement of fairness are absent
airspace, and constraints imposed by neighbor airports within from the guidelines provided by IATA, CAAC, and other
authorities. We refer to the existing research on slot allocation TABLE II
for a single airport while considering the airline’s fairness M ODEL INPUTS
[25]–[27]. Generally, fairness is defined as “to balance the
Notation Description
displacement of flights from their requested times fairly
S = {1, 2, ..., |S| + 1}: S represents the set of airports in
among the airlines”. the MAS. |S| is the total number
of airports in the MAS, and the
TABLE I airports outside are referred to as
E XAMPLES OF FAIR SLOT ALLOCATION IN AN MAS |S| + 1;
P = {1, 2, ..., |P |}: P represents the all fixes within the
MAS. |P | is the total number of
Allocation scheme fixes;
Arr/Dep Airline Airport Fix Request
I II III T = {1, 2, ..., |T |}: T is the set of time intervals t, the
length of which is 5 minutes;
Arr AL1 a P1 0800 0815 0815 0815 A = {1, 2, ..., |A|}: A represents the set of total airlines
Dep AL2 a P2 0800 0800 0800 0800 operating in the MAS, and |A| is
Dep AL1 a P1 0805 0805 0805 0805 the total number of them;
Arr AL2 a P2 0805 0820 0820 0820 M: M refers to the set of total slot
Dep AL1 a P2 0810 0825 0825 0830 requests in the MAS;
Arr AL2 a P2 0810 0810 0810 0810
- - - - - - - -
Dep AL1 b P1 0915 0915 0930 0930
Dep AL2 b P2 0915 0930 0915 0915 The objective of the model is to minimize total displace-
Arr AL1 b P1 0920 0905 0905 0900
Arr AL1 b P1 0920 0920 0920 0920 ments of all slot requests in the MAS.
Arr AL1 b P2 0925 0925 0925 0925 X
Dep AL2 b P1 0925 0910 0910 0910 dm = |t − tm | · xtm (2)
t
To illustrate the differences in fairness between slot allo- X X X
cation for a single airport and for an MAS, Table I shows an Da = dm = |t − tm | · xtm (3)
m∈Ma m∈Ma t
example of three allocation schemes. In the first two schemes,
we only consider the airport capacity. Fix capacity is taken D=
X
Da (4)
into account in scheme 3. Two airlines AL1 and AL2 submit a∈A
a total of 12 slot requests at two airports a and b during
two periods 08:00-08:30 and 09:00-09:30, with 8 requests min D (5)
from AL1 and 4 requests from AL2 . The 15min declared
where tm is the slot that flight m requests, while t is the
capacity at each airport is 3. Therefore, 3 slot requests have
slot that m has been allocated. Thus dm is the number of
to be displaced for each period. Each airline has to modify 3
displacements that are made to request m. The displacements
requests in Scheme 1. It seems to be a fair allocation from the
that airline a obtained are given by Da , and the total dis-
perspective of total displacements to each airline. However,
placements for the MAS are D.
AL1 requests for 8 slots, while AL2 only requests 4 slots. It is
Constraints:
unfair to AL2 if one considers the proportion of displacements
(1) Each flight can only be allocated one slot:
to the requests. Scheme 2 gives a more fair allocating result. X
4 slot requests of AL1 and 2 slot requests from AL2 are xtm = 1 (6)
adjusted, which aligns with the propotional principle. In an t∈T
MAS slot allocation problem, shared fix resources are one
(2) The maximum displacements of a single flight request:
of the critical airspace resources. So in scheme 3, the fix
The airline may reject the allocated slot if the displacements
capacity is taken into account. The flights pass two fixes,
to the requested slot were too large. To ensure the acceptance
both of which provide a 15min capacity of 3. The number of
of allocating result, the maximum displacements of a single
adjusted requests in scheme 3 is the same as that in Scheme 2,
flight are set to be tmax :
but the displacement minutes are increased in order to satisfy X
the fix capacity. It can be seen that the complexity would dm = |t − tm | · xtm ≤ tmax (7)
be much increased if the shared fix resources are considered t
during allocating airport slots. The problem is extended to (3) Turnaround time constraints: When an arrival flight
the fairness among various stakeholders in the allocating of landed at the destination airport, there will be time re-
multiple resources. quirements for turnaround processes. These processes in-
clude passengers and cargo disembarking, refueling, cleaning,
III. M ODEL
and passengers and cargo boarding, etc. A minimum time
A. Baseline model constraint is required to ensure the completion of all the
processes, while a maximum time constraint is to improve
Decision variables:
( the utilization of the gate or apron. Therefore, we have
1 if flight m is scheduled to slot t X X
t
xm = (1) f (m ,m ) ≤ txtm1 − txtm2 ≤ f¯(m1 ,m2 ) (8)
0 otherwise; 1 2
t∈T t∈T
 Arr Dep
t−lm,s,p +11 t+lm,s,p +11
where (m1 , m2 ) is a pair of slot requests that are operated X X X X
by the same aircraft. m1 is the preceding flight, while m2 is xtm + xtm ≤ CHpT otal ,
m∈MpArr Arr
t−lm,s,p m∈MpDep Dep
t+lm,s,p
the succeeding flight. f (m ,m ) is the minimum turnaround
1 2
time, while f¯(m1 ,m2 ) is the maximum turnaround time. t = 12 (th − 1) + 1, th ∈ TH
(4) Airport capacity constraints: Airport capacity include (16)
departure capacity, arrival capacity, and total airport capacity.
P = {1, 2, ..., |P |} is the set of all fixes that are considered
in the slot allocation model. MpArr is the set of arrival flights
t+2
X X that fly through p, while MpDep is the set of departure flights
xtm ≤ CQkS , Arr
that fly through fix p. lm,s,p is the time for arrival flight m
m∈M k t (9)
Dep
flying from fix p to airport s, while lm,s,p is the time for
t = 3 (tq − 1) + 1 , ∀k ∈ K, tq ∈ TQ departure flight m flying from airport s to fix p.
X t+11
X CQkp and CHpk refer to the capacity of fix p within 15
xtm ≤ CHSk , minutes and 60 minutes respectively, where k ∈ K, K =
m∈M k t (10) {Arr, Dep, T otal}.
t = 12 (th − 1) + 1 , ∀k ∈ K, th ∈ TH
B. Airline’s fairness constraints
Equations 9 and 10 are the capacity constraints for 15min
and 60min respectively. K = {Arr, Dep, T otal} stands for The baseline model developed in the previous section
arrival capacity, departure capacity and total. M k denotes the outputs an optimized flight schedule for the MAS. However, it
set of flights whose operation type is k. CQkS and CHSk are does not take into account of airline’s fairness. To investigate
the capacity of airport S during operation scenario k. how to allocate both airport capacity and fix capacity to
(5) Fix capacity constraint: Fix can be used only for arrival the airlines in a more equal and efficient way, we have to
flights or departure flights, or for both arrival and departure develop fairness indicators for airlines in the processes of
flights. Similar to airport capacity, we have 15min capacity airport slot allocation and fix capacity allocation, respectively.
and 60min capacity: Then by combining the two fairness indicators, we build a
Fix only serves for arrival flights: comprehensive fairness indicator for airlines in the MAS.
According to the “proportion principle” in section II-C, the
Arr
fairness indicators of an airline at the airport and at the fix
t−lm,p +2
X X are defined as follows:
xtm ≤ CQArr
p , (1) Airline fairness index at the airport ρa,s :
(11)
m∈MpArr Arr
t−lm,p D
t = 3 (tq − 1) + 1, tq ∈ TQ  Da,ss
ra,s ̸= 0
ρa,s = ra,s (17)
Arr 1 ra,s = 0
t−lm,p +11
X X
xtm ≤ CHpArr ,
(12) Da,s is the total displacements of airline a at airport s. Ds
m∈MpArr Arr
t−lm,p
is the total displacements to all the flights in airport s. ra,s is
t = 12 (th − 1) + 1, th ∈ TH the proportion of slot requests of airline a at airport s. When
Fix only serves for departure flights: ra,s = 0, then airline a has no slot request at airport s. In
this case, Da,s = 0, and ρa,s = 1. This is absolutely fair to
Dep
the airline a.
t+lm,p +2
X X (2) Airline’s fairness index at fix ρa,p :
xtm ≤ CQDep
p ,
(13)  Da,p
m∈MpDep Dep
t+lm,p  Dp
t = 3 (tq − 1) + 1, tq ∈ TQ ρa,p = ra,p ra,p ̸= 0 (18)
1 ra,p = 0
Dep
t+lm,p +11
X X Da,p is the total displacements of airline a at fix p. Dp is
xtm ≤ CHpDep ,
(14) the total displacements to all the flights through fix p. ra,p is
m∈MpDep Dep
t+lm,p
the proportion of slot requests of airline a flying through fix
t = 12 (th − 1) + 1, th ∈ TH p. Again, when ra,p = 0, airline a has no slot request using
Fix that can be used for both arrival and departure flights: fix p. In this case, Da,p = 0, and ρa,p = 1. This is absolutely
fair to the airline a.
Arr Dep
t−lm,s,p +2 t+lm,s,p +2 There are three possible scenarios for ρa,s and ρa,p :
X X X X
xtm + xtm ≤ CQTp otal , 
m∈MpArr Arr
t−lm,s,p m∈MpDep Dep
t+lm,s,p [0, 1)
 the airline is favoured
t = 3 (tq − 1) + 1, tq ∈ TQ ρ= 1 the airline is fairly treated (19)

(15) (1, ∞) the airline is treated unfairly

 (3) The comprehensive airline fairness indicator in the TABLE IV
MAS ρa : A IRPORT THROUGHPUT IN THE MAS OF GBA
X X
ρa = ws ρa,s + ρa,p (20) Airport Pax(M) Cargo(M tons) Movements(M)
s p ZGGG 73.378 1.920 0.491
X X ZGSZ 52.932 1.283 0.356
ws + wp = 1 (21) ZGSD 12.283 0.051 0.085
VMMC 9.611 0.042 0.077
s p ZGHZ 2.554 0.009 0.020
where ws and wp are the weights of fairness at airport s M: million;
and fix p, which can be adjusted based on the preferences of
relevant authorities.
A. Problem setting
To measure the fairness of the total resource allocation of
the MAS, the Gini index is introduced. Based on the fairness 1) Model input: The main inputs into the model include
indices developed for the airlines, we can evaluate the overall the capacity of each member airport in the MAS, fixes
fairness of slot allocation of an MAS. capacity, the flying time between each airport and each fix,
P P and the maximum displacements to a single flight.
i∈A j∈A |ρi − ρj | Among the five airports in the MAS of GBA used in the ex-
Gini ρ = P (22)
2 × |A| × i∈A ρi ample, there are two coordinated/Level 3 airports (Guangzhou
Baiyun Airport and Shenzhen Bao’an Airport), one Level
where ρi , ρj are the comprehensive fairness indices of
2 airport (Zhuhai Jinwan Airport) and one non-coordinated
airline i and j. |A| is the number of airlines operating in the
airport (Huizhou Pingtan Airport). To simplify the expression,
MAS. Ginip is the Gini index of the MAS. The Gini index
we use declared capacity as the airport capacity for all five
is closer to 0, the allocation scheme is more fair. The detail
airports of the MAS. The capacity data provided by the air
implication of Gini index is given in Table III.
traffic control authorities are shown in the table V.
TABLE III
TABLE V
R ANGE OF G INI COEFFICIENT AND ITS MEANING
D ECLARED CAPACITY OF FIVE AIRPORTS ( NUM . OF MOVEMENTS )
Range Meaning
< 0.2 Perfect equity Airport Dep. (15min) Arr. (15min) Total (15min) Total (1hr)
0.2-0.3 Good equity ZGGG 20 19 32 67
0.3-0.4 Average equity ZGSZ 15 15 20 55
0.4-0.5 Poor equity ZGSD 5 5 8 20
> 0.5 Absolute inequity VMMC 5 5 8 22
ZGHZ 5 5 8 18
Dep.:Departure; Arr.: Arrival;
By the leverage of ϵ constraint, we add the fairness into
the baseline model, where ϵ ∈ [0, 1]. Until here, one can trade
off the fairness and efficiency of slot allocation by setting ϵ Fix GYA, LMN, YIN are the fixes that we focus on in this
in the appropriate range. paper. Their capacity is given by the air traffic authorities, as
shown in the table VI.
P P
i∈A j∈A|ρi − ρj | TABLE VI
ε1 ≤ Gini ρ ≤ ε2 ⇒ ε1 ≤ P ≤ ε2 F IX CAPACITY (N UM . OF MOVEMENTS )
2 × |A| × i∈A ρi
(23)
Fix Capacity (15min) Capacity (1hr)
GYA 8 27
IV. R ESULTS LMN 8 32
We use the data from the MAS of the Guangdong-Hong YIN 9 36
Kong-Macao Greater Bay Area (GBA) to validate and exam-
ine the performance of the proposed model. Five airports are Flying time between an airport and a fix is affected by
considered in the MAS: Guangzhou Baiyun Airport (ICAO multiple factors such as weather, air traffic control strategies,
code: ZGGG), Shenzhen Bao’an Airport (ICAO code: ZGSZ), or aircraft performance. Therefore, the flying time is stochas-
Zhuhai Jinwan Airport (ICAO code: ZGSD), Macao Airport tic actually. The model in this paper does not consider the
(ICAO code: VMMC), and Huizhou Pingtan Airport (ICAO uncertainty of flying times. The median of historical flying
code: ZGHZ). Hong Kong Airport (ICAO code: VHHH) time during 2018-2019 between an airport and a fix is selected
is excluded from this study because its arrival traffic and as the flying time between the airport and the fix in the model.
departure traffic is separated from the other airports. In other 2) Slot requests: Due to data confidentiality, we are not
words, there is no conflict between the traffic of VHHH and able to obtain airlines’ slot request data. To validate the model
of other airports within the MAS. The throughput of the five and examine its performance, we use the flight schedules of
airports in 2019 is presented in Table IV. We can see that the a typical day in 2019 as the slot request data. Follow-up
throughput of ZGGG and ZGSZ are significantly higher than research can use actual slot request data. The presented work
that of the other three airports, accounting for more than 80% uses a total of 3055 slot requests, including 1160 connecting
of the MAS. flights.
 TABLE VII
S LOT REQUESTS OF THE AIRLINES AT THE AIRPORT AND FIX

Sot requests at the airport Slot requests at the fix

Rank Airlines ICAO Code Total requests ZGGG ZGSZ ZGSD VMMC ZGHZ YIN GYA LMN
1 China Southern Airlines CSN 1061 757 241 57 4 2 161 128 166
2 Shenzhen Airlines CSZ 403 111 270 12 10 0 49 48 42
3 China Eastern CES 210 124 68 12 0 6 17 37 29
4 Air China CCA 202 119 61 16 4 2 60 13 37
5 Hainan Airlines CHH 180 86 88 4 0 2 38 14 30
6 Xiamen Air CXA 68 24 28 8 0 8 6 11 3
7 Spring Airlines CQH 63 24 31 2 0 6 7 7 7
8 Air Macau AMU 60 0 0 0 0 60 11 0 5
9 SF Airlines CSS 59 8 51 0 0 0 9 2 4
10 Shandong Airlines CDG 50 18 12 20 0 0 4 9 1
11 All other airlines \ 699 370 197 73 29 30 81 173 2
Total 3055 1641 1047 204 47 116 443 442 326

A total of 100 airlines operated flights at the airports in

the MAS. Among them, the top 10 airlines in terms of the
number of flights requested 2,356 slot requests, accounting
for 41.2% of the total request. Thus, we consider 100 airlines
when testing the model but give our focus on the top 10
airlines when analyzing the results. The information on these
10 airlines, as well as the number of slot requests of each
airline at the five airports, is shown in table VII. It can be
found that slot requests at ZGGG and ZGSZ are significantly
higher than that of the other three airports, while slot requests
at ZGHZ are the lowest. China Southern Airlines (CSN) and
Shenzhen Airlines (CSZ) have their main bases at ZGGG and
ZGSZ respectively, thus the slot requests of these two airlines
are obviously large; While at VMMC, Air Macau (AMU) is
the main slot user. The number of slots requested by airlines
at the three fixes is given in Table VII. It can be seen that
Figure 1. MAS total displacements and Gini-based fairness
almost all airlines have flights passing through three fixes,
except that AMU at LMN. According to the proportion of
slot requests, ten airlines can be simply divided into three
categories: a large amount (CSN), a medium amount (CSZ, Figure 1 plots the efficient frontier for MAS slot allocation.
CES, CCA, CHH), a small amount (CXA, CQH, AMU, CSS, We use solid and dashed lines to distinguish the feasible
CDG). and optimal solutions. When the Gini-based fairness is con-
The model in this paper is a mixed integer program- strained below 0.1199, only feasible solutions can be obtained
ming(MIP) model, which involves a huge amount of flight due to the time limits, with the Gurobi gap value of around
data at five airports and is complex to solve. The model is 0.08. Recall that the Gini coefficient is defined to measure the
solved using Gurobi 9.5, and the computer used is equipped fairness of resource allocation among the users. It commonly
with eight-core Intel(R) Core(TM) i7-10700 CPU, 2.90GHz. employs 0.4 as a threshold for quantifying fairness. When
the Gini coefficient is less than 0.4, the allocating result is
B. Results considered to be relatively fair to all the users. Otherwise, it
is considered to be unfair to some of the users. In addition,
1) Trade-off between slot displacements and Gini-based the smaller the Gini fairness, the more equitable the slot allo-
fairness: cation of the MAS. From figure 1, it can be observed that the
In this section, we analyze the trade-off between slot total displacements increase with the decrease of Gini-based
displacements and fairness from three aspects: the MAS, the fairness below 0.1668. The result suggests that achieving a
airlines and airports. highly fair flight schedule would require more adjustments
(1) The MAS total displacements and Gini-based fair- to airlines’ requests. When the Gini-based fairness of the
ness MAS is greater than 0.1668, the total displacements remain at
335 minutes. By comparing their specific optimized schedule, number. In general, the more slot requests, the more slot
we found that there are differences in their adjustment. For displacement. This finding is consistent with the ”proportional
example, one slot request from airline CSC is shifted when principle” used in the model.
Gini-based fairness is 0.1668 but has not been shifted when
Gini-based fairness is 0.2701. This observation suggests the
existence of multiple solutions, or flight schedules, with the
same total displacements. Different flight schedules lead to
different Gini fairness. However, when the Gini-based fairness
is set over 0.3200, the optimized schedule for the MAS
remains the same. In other words, the slot allocation results
are no longer influenced by the Gini-based fairness constraint.
Overall, it is crucial for the slot coordinator or managing
body to balance fairness and efficiency during the MAS slot
allocation.
(2) Slot displacements and Gini-based fairness of top
ten airlines
(a) Airport

(b) Fix
Figure 2. Airline’s total displacements and MAS Gini-based fairness Figure 3. Total displacements and MAS Gini fairness

Figure 2 presents the total displacements of the top 10 (3) Slot displacements at the airports & fixes, and MAS
airlines with the MAS Gini-based fairness. Overall, each Gini-based fairness
airline’s slot displacements fluctuate up and down when Figures 3(a) and 3(b) depict the changes in slot displace-
Gini fairness changes. When the MAS Gini Fairness reaches ments for the five airports and three fixes in the MAS, respec-
0.3200, the airlines’ displacements remain stable since the tively. In Figure 3(a), the ranking of the slot displacements
optimized flight schedule are the same. The 10 airlines can be of the airport aligns with the ranking of airport slot demand.
classified into three groups based on their total displacements: The slot displacement of ZGGG and ZGSZ is significantly
(1) Group one: CSN. The total displacements of CSN are higher than that of the other three airports. Besides, the slot
always the highest among the 10 airlines for each Gini- displacements of ZGSD, ZGHZ and VMMC are always small,
based fairness, fluctuating within [90 min, 180 min]. This never exceeding 50 minutes. Finally, ZGHZ has the fewest
is reasonable because the slot requests from CSN are far slot displacements, only changing between 0 and 10 minutes.
more than other airlines. The total displacements of CSN Compared with other airports, ZGHZ has the fewest slot
decrease as the MAS Gini fairness narrows, indicating that the requests and few flights passing through congested fixes, so
burden of adjusted slots for CSN is shared by other airlines, the slot requests of it rarely need to be shifted. In terms of
which improve the fairness between airlines. (2) Group two: fixes, figure 3(b) shows that the slot displacements of the
CSZ, CCA, CES and CHH. The slot displacements of these three fixes also fluctuate in their ranges. The YIN fix has
four airlines stay at a medium level, with a maximum of 90 the largest slot displacements, ranging between [120 min,
minutes and a minimum of 5 minutes. The minimum bound 160 min]. The GYA fix follows with a range of [60 min,
indicates that slot adjustments always occur in these airlines. 100 min], and the LMN is the smallest, almost staying at
(3) Group three airlines are CXA, CQH, AMU, CSS and 0. It is true that when the Gini-based fairness of the MAS
CDG, with the slot displacement remaining at a low level. is less than 0.1668 and continues to decrease, the total slot
These five airlines’ slot displacement changed in the range of displacements of the MAS as a whole increase significantly.
[0,15 min]. It can be found that there is a clear relationship The obvious raise also happen to airports or fixes, with some
between the airlines’ slot displacement and their slot request fluctuation to some extent. This may be due to the definition
of the MAS Gini-based fairness, which is aimed to measure
whether the resources are allocated fairly among the airlines.
The 10 airlines in the case study have slot applications in 5
airports, and most of them pass through YIN, GYA and LMN
fixes. So when the Gini-based fairness narrows, the increased
slot displacement will be distributed to each airport and each
fix.
In general, a fairer flight schedule of an MAS can be
achieved with a slight increase in the number of total dis-
placements. The pursuit of “extremely fair” requires a greater
sacrifice of total displacements. The results are consistent with
current research on single airport slot allocation. The amount
of displacement obtained by the airline is related to the slot
requests of the airline itself operating in the MAS. A similar
conclusion can be made from the perspective of airports
and fixes. Generally, the more slot requests, the greater the
possibility of slots being shifted. Figure 4. Total displacements of the MAS under different λ
2) Sensitivity analysis: When building the comprehensive
fairness indicator of airlines, we used two parameters ws
and wp . The parameters respectively represent the weight of three fixes before and after optimization to verify the
of the airport s in the MAS and the weight of the fix p. effectiveness of the model in optimizing the allocation of
According to the theory of fair distribution of multiple types MAS capacity resources.
of resources, the weight reflects the preference of resource
allocation subject for the fair allocation of this resource. When
handling slot allocation, the weight reflects the preference of
the slot coordinator for the fair allocation of airport capacity
and fix capacity. For airlines, different airports and fixes imply
different importance to the company’s efficiency and benefits.
In an MAS, airports and fixes have their specific functioning
position. Therefore, how to measure the preference is worth
studying but it is not the focus of this paper. Considering the
influence of weights on the slot allocation results, we define
λ as the ratio of the two weight parameters, namely
ws
λ= (24)
wp
Under the same ϵ constraint, we change the value of
λ(λ ∈ {0.1, 0.5, 1, 2, 10}) to initially explore the influence
Figure 5. Airline’s displacements and fairness indicators in different Gini-
of different weight ratios on the MAS Gini fairness and the based fairness
total slot displacements.
Figure 4 displays the change curve of the total slot dis- Figure 5 shows the slot displacements and comprehen-
placements of the MAS as the value of λ varies. The graph sive fairness indicator for ten airlines, under two different
illustrates that the value of λ has an impact on the problem MAS Gini-based fairness values. The figure contains two
of MAS slot allocation. Specifically, when the value of λ scenarios: (1) Gini-based fairness of 0.1668, which is the
decreases, a final schedule with larger total displacements best Gini fairness the model can achieve with the smallest
will be obtained. This means that when the weight of fixes total slot displacements (335 minutes); (2) Gini-based fairness
in the fairness indicator is greater than that of airports, the of 0.1199, which is the smallest Gini-based fairness value
slots required to be adjusted are fewer. We surmise this for which the model can obtain an optimal solution within
may be because the capacity and demand conflict of fixes the time limit. It corresponds to total slot displacements
is more prominent than that of airports within the MAS. This of 365 minutes. Firstly, we compare the MAS total slot
conclusion also highlights the need to consider the allocation displacements under the two Gini-based fairness values. It
of fixes capacity within the MAS. For different MAS with can be concluded that the model can achieve better Gini-based
specific congestion characteristics, different weight values fairness by sacrificing a certain amount of slot displacements.
should be set. In future research, it is necessary to develop When the Gini fairness is 0.1668, none of the two airlines
rules for setting appropriate weight values. (CXA, CQH) have slots to be displaced. The airline with the
3) Capacity allocation optmization: A major difference most slot requests, CSN, receives a total of over 100 minutes
in MAS slot allocation is the consideration of critical fix of slot displacements, while the other airlines only have no
capacity. This section analyses the changes in the traffic flow more than 75 minutes. In contrast, when the Gini fairness is
0.1199, the airlines have a smaller gap in the number of slot busyness of the busy periods and increasing the workload
displacements and slot displacements occur to all airlines. In in the relatively ”free” periods. Finally, it realizes demand
terms of the comprehensive fairness indicator, when the Gini- management at a strategic level.
based fairness is 0.1668, the comprehensive fairness indicator Overall, the model is effective in achieving the optimal
varies more across airlines but is still less than 1.0 (airlines allocation of multiple resources in an MAS. When the MAS
are favored). At a Gini fairness of 0.1199, the differences Gini-based fairness is reduced, the additional displacements
between airlines’ fairness indicators reduce obviously. Thus, a from a fairer flight schedule can be more evenly distributed
lower Gini fairness allows for a more even distribution of slot across the airlines. The adjusted flight schedules result in a
displacements between airlines. Finally, the airlines’ fairness more even and smooth distribution of traffic. By coordinating
indicators reach a better balance as well. flights from busy to idle periods, the model alleviates the
busyness of busy periods to some extent.

V. C ONCLUSIONS

A multiple airport system consists of several airports that

provide air transport service to the metropolitan area. To
improve the utilization of resources of an MAS, this paper
aims to allocate slots for all the airlines in the MAS con-
sidering trading off equity and efficiency. A baseline model
is first proposed with the objective of minimizing the total
slot displacements of the MAS. The baseline model considers
airport capacity constraints, fixes capacity constraints, flight
turnaround time constraints, and the maximum adjustments
for a single flight. Then, comprehensive fairness indicators
for airlines are developed to measure airlines’ fairness in
resource allocation at airports and fixes. A new slot allocation
model for an MAS considering the airline’s fairness is then
Figure 6. The 15min traffic flow at YIN when Gini-based fairness is 0.1199,
0.1668 and 0.3200
formulated by adding fairness constraints into the baseline
model. The models are validated and tested using flight
schedule data of the MAS in the Guangdong-Hong Kong-
After the data analysis, we found that the optimization
Macao Greater Bay Area. The results demonstrate that our
pattern of capacity allocation was similar for the three fixes.
model can improve the balance of efficiency and fairness to
Thus, we selected the fix YIN as the representative to analyze
some extent during the MAS slot allocation. In general, the
the optimization effect of the model, whose capacity conflict
more the airline’s demand, the greater the possibility of the
is more intense. Figure 6 shows the 15-minute traffic flow
slot being displaced. The model is shown to be an effective
of fix YIN for three different MAS Gini fairness. The green
tool for optimally allocating the airport and airspace capacity
line in the figure is the fix capacity line with a value of 9,
of the MAS.
indicating a 15-minute capacity constraint of 9 flights at the
fix. As can be seen from the graph, the model can achieve The findings of this study have a number of important
a reasonable allocation of capacity resources at the fix. The implications for future practice. First, the type of airlines
result satisfies the capacity limit of the fix and optimizes the and operations can be taken into account to enhance the
flight distribution. Taking the fix YIN as an example, slots validity of the fairness indicator. Second, this paper focuses
are adjusted around 09:00, 12:00-14:00, 16:00 and 17:00. on fairness among airlines, the airport is yet another important
Clearly, the scheduled slots during these periods exceed the stakeholder in the MAS. Continued efforts are needed to
fix capacity. They are considered the peak periods for YIN investigate the impact of the scheduling results on fairness
that highly require to mitigate demand-capacity conflicts. among the member airports. In addition, the priority rules on
Figure 6 also shows a comparison between three different slot requests that are listed in WASG, for example, historic
Gini fairness. The optimization effect under them is almost flights enjoy the first priority, should be included in the model.
the same with tiny differences as highlighted in red boxes. The Finally, both the MAS and airport networks involve multiple
smaller Gini fairness optimization solution achieves a gentler airports, thus the slot allocation model proposed in this paper
flight distribution curve, indicating a more even distribution can be extended for the slot allocation problem of airport
of flights. Interestingly, the difference between 0.1668 and networks.
0.3200 only happened during the period of 6:00-8:00 in the
morning. However, the MAS fairness experienced an obvious ACKNOWLEDGMENT
improvement. This may indicate that the adjustment of a
specific period of slots could improve the whole fairness. This research was supported by the National Natural Sci-
Overall, our approach has the advantage of smoothing out ence Foundation of China (Grant Nos. U2033203, 52272333,
the flight distribution during the periods, ”reducing” the U1833126, 61773203).
 R EFERENCES [23] W Liao, X Cao, Y Liu, Y Huang, “Investigating differential effects
of airport service quality on behavioral intention in the multi-airport
[1] International Air Transportation Association, “An- regions”, Research in Transportation Business & Management, Vol. 45,
nual Review 2020”. Accessed January 1, 2022, Part C, 2022.
https://www.iata.org/contentassets/c81222d96c9a4e0bb4ff6ced0126f0bb/ [24] C Barnhart, Fearing D, A Odoni, et al. Demand and capacity man-
iata-annual-review-2020.pdf. agement in air transportation [J]. EURO Journal on Transportation and
Logistics, 2012, 1(1-2):135-155.
[2] International Air Transportation Association, “An-
[25] R de Neufville, A R Odoni. Airport systems: Planning, design, and
nual Review 2022”. Accessed January 1, 2023,
management. 2003.
https://www.iata.org/contentassets/c81222d96c9a4e0bb4ff6ced0126f0bb/
[26] Y Jiang, K G. Zografos, “A decision making framework for incor-
annual-review-2022.pdf.
porating fairness in allocating slots at capacity-constrained airports”,
[3] F. Katsigiannis, K. Zografos, “Multi-objective airport slot scheduling Transportation Research Part C: Emerging Technologies, Vol. 126, 2021.
incorporating operational delays and multi-stakeholder”. [27] K N. Androutsopoulos, M A Madas, “Being fair or efficient? A
[4] K G Zografos, M A Madas, K N Androutsopoulos. Increasing airport fairness-driven modeling extension to the strategic airport slot scheduling
capacity utilisation through optimum slot scheduling: review of current problem”, Transportation Research Part E: Logistics and Transportation
developments and identification of future needs [J]. Journal of Scheduling, Review, Vol. 130, pp. 37-60, 2019.
2017, 20(1):3-24.
[5] International Air Transport Association (IATA), Worldwide Air-
port Slot Guidelines (WASG), 2020. Accessed 3 January 2023.
https://www.iata.org/contentassets/4ede2aabfcc14a55919e468054d714fe
/wasg-edition-1-english-version.pdf.
[6] K. G. Zografos, Y. Salouras, M. A. Madas, “Dealing with the efficient
allocation of scarce resources at congested airports”, Transportation
Research Part C: Emerging Technol.
[7] K. G. Zografos, K. N. Androutsopoulos, M. A. Madas, “Minding
the gap: Optimizing airport schedule displacement and acceptability”,
Transportation Research Part A: Policy and Practice, vol. 114, pp. 203–
221, 2018.
[8] N. A. Ribeiro, A. Jacquillat, A. P. Antunes, A. R. Odoni, J. P. Pita, “An
optimization approach for airport slot allocation under IATA guidelines”,
Transportation Research Part B: Methodological, vol. 112, pp. 132-156,
2018.
[9] N. Pyrgiotis, A. Odoni, “On the impact of scheduling limits: A case study
at newark liberty international airport”, Transportation Science, vol. 50,
pp. 150-165, 2016.
[10] A. Jacquillat, A. R. Odoni, “An integrated scheduling and operations
approach to airport congestion mitigation”, Operations Research, vol. 63,
pp. 1390–1410, 2015.
[11] L. Castelli, R. Pesenti, A. Ranieri, “The design of a market mechanism
to allocate air traffic flow management slots”, Transportation Research
Part C: Emerging Technologies, vol. 19, pp. 931-943, 2011.
[12] P. Pellegrini, L. Castelli, R. Pesenti, “Secondary trading of airport slots
as a combinatorial exchange”, Transportation Research Part E: Logistics
and Transportation Review, vol. 48, pp. 1009-1022, 2012.
[13] P. Pellegrini, T. Bolić, L. Castelli, R. Pesenti, “SOSTA: An effective
model for the simultaneous optimisation of airport slot allocation”,
Transportation Research Part E: Logistics and Transportation Review, vol.
99, pp. 34–53, 2017.
[14] L. Corolli, G. Lulli, L. Ntaimo, “The time slot allocation problem
under uncertain capacity”, Transportation Research Part C: Emerging
Technologies, vol. 46, pp. 16–29, 2014.
[15] J. Fairbrother, K. G. Zografos, K. D. Glazebrook, “A slot-scheduling
mechanism at congested airports that incorporates efficiency, fairness, and
airline preferences”, Transportation Science, vol. 54, pp. 115-138, 2020.
[16] J. F. Nash, The bargaining problem, Econometrica, vol. 18, pp.
155–162, 1950.
[17] K. G. Zografos, M. A. Madas, K. N. Androutsopoulos, “Increasing
airport capacity utilisation through optimum slot scheduling: review of
current developments and identification of future needs”, Journal of
Scheduling, vol. 20, pp. 3-24, 2017.
[18] K. G. Zografos, Y. Jiang, “A bi-objective efficiency-fairness model for
scheduling slots at congested airports”, Transportation Research Part C:
Emerging Technologies, vol. 102, pp. 336-350, 2019.
[19] R de Neufville, “Management of multi-airport systems: A development
strategy”, Journal of Air Transport Management, Vol. 2, Issue 2, pp. 99-
110, 1995.
[20] S Sidiropoulos, A Majumdar, K Han, “A framework for the op-
timization of terminal airspace operations in Multi-Airport Systems”,
Transportation Research Part B: Methodological, Vol. 110, pp. 160-187,
2018.
[21] Y Wang, Y Zhang, “Prediction of runway configurations and airport
acceptance rates for multi-airport system using gridded weather forecast”,
Transportation Research Part C: Emerging Technologies, Vol. 125, 2021.
[22] L Ruan, A Gardi, R Sabatini, “Operational efficiency analysis of Beijing
multi-airport terminal airspace”, Journal of Air Transport Management,
Vol. 92, 2021.