Hindawi Publishing Corporation

Mathematical Problems in Engineering

Volume 2015, Article ID 854372, 12 pages
http://dx.doi.org/10.1155/2015/854372

Research Article
Multirunway Optimization Schedule of Airport Based on
Improved Genetic Algorithm by Dynamical Time Window

Hang Zhou and Xinxin Jiang

College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Jiangsu, Nanjing 210016, China

Correspondence should be addressed to Hang Zhou; [email protected]

Received 15 January 2015; Accepted 24 April 2015

Academic Editor: Wei (David) Fan

Copyright © 2015 H. Zhou and X. Jiang. 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.

Reasonable airport runway scheduling is an effective measure to alleviate air traffic congestion. This paper proposes a new model
and algorithm for flight scheduling. Considering the factors such as operating conditions and flight safety interval, the runway
throughput, flight delays cost, and controller workload composes a multiobjective optimization model. The genetic algorithm
combined with sliding time window algorithm is used to solve the model proposed in this paper. Simulation results show that the
algorithm presented in this paper gets the optimal results, the runway throughput is increased by 12.87%, the delay cost is reduced
by 61.46%, and the controller workload is also significantly reduced compared with FCFS (first come first served). Meanwhile,
compared with the general genetic algorithm, it also reduces the time complexity and improves real-time and work efficiency
significantly. The analysis results can provide guidance for air traffic controllers to make better air traffic control.

1. Introduction flight scheduling and sequencing, and reduce flight delays,

thus increasing air traffic flow [2–8]. In the year 2011, Bennell
With the rapid development of the aviation industry, the et al. designed algorithms for runway scheduling, thereby
passenger throughput of the busiest airport increases signif- increasing runway capacity and reducing flight delays [9]. In
icantly and flight delays are becoming a hot focus. Air traffic the year 2013, Samà et al. adopted a dynamical time window
congestion problem is attracting more and more attention as
algorithm to solve the problem of a lot of flight delays in case
an important aspect of air traffic control.
of heavy traffic [10].
Countries in the world take measures to reduce flight
delays. The most simple way to solve the problem is to Besides the above methods, many scholars in China also
improve hardware facilities, such as making further expan- do a lot of research on the flight scheduling. Aircraft terminal
sion of airports and runways and improving the performance controllers (ATC) usually apply FCFS (first come first served)
of communication navigation and surveillance equipment. strategy to make flight scheduling, but the fact shows that
It needs a long period and large investment to make it and FCFS is not the best strategy to maximize the use of existing
the efficiency is low. By learning a variety of optimization airport capacity and to reduce the average delay costs [11].
algorithms, we analyze airport capacity and make a model In the year 2004, Jiang and Yun presented an efficient opti-
of flight delays and allocate air traffic flow. How to dispatch mization algorithm for aircraft landing scheduling problem,
approach and departure flights in the terminal area reason- based on depth-first-search implicit enumeration algorithm,
ably, how to reduce flight delays, and how to make full use of which is subject to MPS (maximal position shift) limiting,
resources in the airspace have become the focus of the study position lock limiting, moving time window, and other ATC
[1]. Therefore, domestic and foreign scholars have done a lot restrictions. The results indicated that the algorithm could
of research: in the foreign countries, Vranas, D’Ariano, and apply to dynamic ATC automation system with flexible
many other scholars made a model of air traffic control in the goal cost function, efficiently minimized traffic delay, and
airspace of the airports and designed appropriate algorithms improved landing capacity [12]; Wang et al. analyzed the
to reduce the time that the flight waits in the air, optimize impact of the relationship between the window dimension,
2 Mathematical Problems in Engineering

the position constraints, and the step-length, put forward a the weather, air traffic control, route, and other factors. The
sliding window method, and gave a theoretical manifestation. reasons of using this innovative algorithm are shown as
The results showed that the quantity of calculation declined, follows.
which had been proved in an authentic simulation test [13].
In the year 2005, Jiang and Zhang proposed a new approach (i) Genetic algorithm has strong global search capability,
scheduling model based on earliest estimated arrival time. which will help to find the optimal results.
A calculation example showed that the new model and its (ii) Dynamical time window algorithm can greatly reduce
algorithm gave better approach scheduling with obvious less the amount of calculation and also improve the
computation [14]. In the year 2009, Yang introduced the quality of the solution.
sliding window algorithm by considering the total landing (iii) Hence, putting these algorithms together to solve
time and the amount of calculation. Aiming at the theoret- the problem of flight scheduling, we cannot only get
ical manifestation and simulation results got in this paper, better results, but also increase operational efficiency
they analyzed amount of calculation and evaluation versus and improve the robustness of the flight sequence.
window dimension and the step-length. The results validated
the flexibility and practicability of this algorithm [15]. In the The remainder of this paper is organized as follows.
year 2011, Huang et al. developed an object function with total The multiobjective optimization model and the genetic algo-
delays and proposed an improved genetic algorithm based rithm combined with dynamical time window algorithm are
on Grefenstette coding and receding horizon control strategy developed in Section 2. After a simulation experiment is
according to the characteristics of ASS problem. Then, the conducted, the results analysis and comparison with other
scheduling model with genetic algorithm was simulated. algorithms are presented in Section 3. Finally, this paper
The results of the simulation indicated that the total delays concludes with a summary and future research direction in
were greatly reduced compared with FCFS algorithm and Section 4.
the proposed algorithm had better optimization performance
than traditional genetic algorithm [16]. In the year 2012, J. 2. Model and Methods
Zhang and W. Zhang constructed the arriving and leaving
aircraft scheduling model based on dynamic optimization. 2.1. Problem Description. At present, controllers usually use
They viewed the scheduling process of arriving and leaving FCFS policy to make flight scheduling in the terminal area,
as an integer. So the controller’s burden was alleviated which is mainly based on estimated leaving time or landing
and the loss caused by subjective factors was minimized. time of each flight. It is mostly dependent on the work experi-
The model runs effectively and efficiently in the simulation ence of controllers to make sorting of flight. For the approach
experiment compared with FCFS [17]. In the year of 2014, we flight, controllers need to make a reasonable allocation of
considered factors such as operating conditions and safety their runways and taxiways to make them land smoothly and
interval of multirunways; the maximum throughput and reach the apron safely; for the departure flight, controllers
minimum flight delays losses as well as robustness were taken need to make a reasonable allocation of their taxiways and
as objective functions; the model of optimization scheduling runways to make them leave the airport safely and take
of approach and departure flights was established. The genetic off smoothly. Flight scheduling problem considered in this
algorithm was introduced to solve the model. Although the paper is mainly based on the existing research and combines
algorithm can improve the efficiency and reduce delay losses genetic algorithm with dynamical time window algorithm to
effectively, the efficiency of algorithm is not high enough [18]. solve the problem. Different types of aircraft must comply
Although the above scholars have done a lot of research with specific “minimum safety interval standards,” rearrange
on flight scheduling, there is still some deficiency. Some the order of approach and departure flight in the queue.
existing models were established in an ideal situation and did Dynamic sorting method is used in this paper to search
not take into account the actual factors such as the weather, all possible flight sequences and find out the best one with
air traffic control, and airport surrounding environment. maximal runway throughput, minimal delay cost, and lowest
Moreover, most models were only suited for single runway, controller workload.
only a little for multirunway. These models did not meet the For the mixed types of aircraft, the International Civil
development trend of the airport. In addition, the scheduling Aviation Organization (ICAO) specifies the minimum inter-
problem of approach and departure flights on multirunway val standards, defined as 𝛿𝑖𝑗 ; the distance interval and its
will be more complex with the increasing number of runways, corresponding time interval are listed in Table 1 (it represents
and the impact of human factors will also be deepened. It is the condition of windless).
needed to make further improvement to the algorithm and Figure 1 presents the approach and departure time of 3
design better optimized program to get closer to the actual kinds of flight sequences. Queue 1 represents the estimated
operational requirements. approach and departure time queue; queue 2 represents the
To address the above problems, runway throughput, FCFS algorithm queue; queue 3 represents the optimized
flight delay cost, and controller workload are created to be algorithm queue. The delay time of queue 2 and that of queue
optimization objective, a multiobjective optimization model 3 are 221 s and 96 s, respectively. Obviously, rearranging the
is established, and the genetic algorithm combined with a order of the flight can reduce the total delay time as well as
dynamical time window algorithm is proposed to solve the delay cost (here, numbers 1 and 2 are approach flights and
problem of flight scheduling on multirunway considering numbers 3 and 4 are departure flight).
Mathematical Problems in Engineering 3

Table 1: The safety interval of various types of aircraft.

Tailing
Types of aircraft Minimum distance interval/km Minimum time interval/s
L M H L M H
L 3 3 3 98 74 74
Leading M 4 3 3 138 74 74
H 6 5 4 167 114 94
H, M, and L represent three kinds of flight type; they are heavy, medium, and light.

Estimated approach or departure time (s)

1 2 3 4 (s)
H

M
L

L
0 60 100 150
FCFS (s)

1 2 3 4
H

M (s)

L
L

0 167 233 371

The optimized sorting (s)

2 4 3 1 (s)
M

H
L

0 96 170 246

Figure 1: Approach and departure time of 3 kinds of flight sequences.

As can be seen from Figure 1, the optimized flight (5) The capacity of the airport meets the assumption, the
sequence saved 125 s compared with FCFS. number of flight within the capacity of the airport.
(6) The approach flight does not delay when it takes off
2.2. Model Introduction. Flight scheduling is a dynamic con- at the departure airport and arrives on time at the
tinuous process. It is needed to make adjustments according terminal area waiting for landing.
to the change of real-time information. Assume that there is 𝑢
approach flight and V departure flight waiting for scheduling. 2.3. Parameters and Symbols. (i) 𝐹𝐴𝐴: the collection of
They all need to meet the minimum safety interval. Arrange approach flight, 𝐹𝐴𝐴 = {𝑓𝐴1 , 𝑓𝐴2 , . . . , 𝑓𝐴𝑢 }.
the order of flight to satisfy the desired object. (ii) 𝐹𝐷𝐷: the collection of departure flight, 𝐹𝐷𝐷 =
In this paper, we studied the approach and departure {𝑓𝐷1 , 𝑓𝐷2 , . . . , 𝑓𝐷V }.
flight scheduling on multirunway; assume the following: (iii) 𝐹𝑖 : the collection of all the flights 𝑓𝑖 , 𝑓𝑖 ∈ 𝐹𝑖 , 𝐹𝐴𝐴 ∪
𝐹𝐷𝐷 = 𝐹𝑖 , 𝑖 = {1, 2, . . . , 𝑢 + V}.
(1) There is 𝑢 + V flight of 𝑀 airlines waiting for (iv) 𝐿: the collection of independent parallel runways, 𝐿 =
scheduling. Controllers make a dynamic sorting of {1, 2, . . . , 𝑙}.
flight and assign a reasonable time for approach and 1, there is a flight arriving on the runway 𝑙
(v) 𝐻𝐴𝑙 = { 0, there is not a flight arriving on the runway 𝑙.
departure flights.
𝑙 1, there is a flight leaving the runway 𝑙
(vi) 𝐻𝐷 = { 0, there is not a flight leaving the runway 𝑙.
(2) The estimated time and actual time of each flight are
different and can be determined at the time 0. (vii) 𝑇𝐴𝑌max : the maximum delay time of approach flight.
(viii) 𝑇𝐷𝑌max : the maximum delay time of departure
(3) The airport studied in this paper contains multiple flight.
parallel runways, and each of them must comply with (ix) 𝑆𝑇𝐴 𝑖 : the actual time of flight 𝑓𝑖 arriving at the
independent operation standards. airport.
(4) The information of approach and departure flights (x) 𝐸𝑇𝐴 𝑖 : the estimated time of flight 𝑓𝑖 arriving at the
(including flight number, flight type, the estimated airport.
arriving or leaving time of flight, etc.) is known. (xi) 𝑆𝑇𝐷𝑖 : the actual time of flight 𝑓𝑖 leaving the airport.
4 Mathematical Problems in Engineering

(xii) 𝐸𝑇𝐷𝑖 : the estimated time of flight 𝑓𝑖 leaving the The objective function of the maximum runway through-
airport. put is
(xiii) 𝑓𝑖𝑝 : the initial position of flight 𝑓𝑖 in the flight queue. 𝑢+V 1
(xiv) 𝑓𝑖𝑞 : the location of flight 𝑓𝑖 in the optimized flight Capacity = min 𝑇all = min (𝑇𝑓𝑙 − 𝑇𝑓𝑙 ). (1)
queue.
(xv) 𝑑: the maximum displacement constraint MPS 2.4.2. The Objective Function of Flight Delays Cost. In most
(maximum position shift). cases, as for approach and departure flights, their actual
1, the flight 𝑓𝑗 kept landing after flight 𝑓𝑖
(xvi) 𝛼𝑖𝑗 = { 0, otherwise.
arriving or leaving time is almost impossible to be the
1, the flight 𝑓𝑗 kept leaving after flight 𝑓𝑖 same as their estimated time, so we make the following
(xvii) 𝛽𝑖𝑗 = { 0, otherwise. provision: supposing a certain range, the actual time later
1, the flights 𝑓𝑖 and 𝑓𝑗 are landing on the same runway
(xviii) 𝛾𝑖𝑗 = { 0, the flights 𝑓𝑖 and 𝑓𝑗 are not landing
than the estimated time within Δ𝑡 is not regarded as delay.
on the same runway. Furthermore, if the actual time is earlier than the estimated
(xix) 𝛿𝑖𝑗 : the minimum safety interval between two approach or departure time, the delay time is a negative value;
successive approaching flights (they are landing on the same it is not realistic, so we predetermine that if there is a flight
runway and 𝑖 ≠ 𝑗). Different flights have different safety arriving earlier than its estimated time, there is no delay.
interval; meanwhile, on the same runway, if the same type of The total delay time of the whole flight sequence could be
flight has different types of approach and departure, its safety expressed as follows.
interval is different too. (1) If flight 𝑓𝑖 is an approach flight, the delay time is
𝑖
(xx) 𝐶𝐴𝐻 : the unit time delay cost of approach flight 𝐹𝐴𝐴. 𝐴 delay :
𝑖
(xxi) 𝐶𝐷𝐻 : the unit time delay cost of departure flight 𝐹𝐷𝐷.
Different types of flight have different unit time delay 𝐴 delay
costs. In general, the larger flight has higher unit time delay
cost. {𝑆𝑇𝐴 𝑖 − 𝐸𝑇𝐴 𝑖 − Δ𝑡 𝑓𝑖 ∈ 𝐹𝐴𝐴, 𝑆𝑇𝐴 𝑖 − 𝐸𝑇𝐴 𝑖 > Δ𝑡 (2)
={
0 𝑓𝑖 ∈ 𝐹𝐴𝐴, 𝑆𝑇𝐴 𝑖 < 𝐸𝑇𝐴 𝑖 .
{
2.4. Modeling
2.4.1. The Objective Function of Runway Throughput. Suppose (2) If flight 𝑓𝑖 is a departure flight, the delay time is 𝐷delay :
that, on runway 𝑙, the actual scheduling time of the first flight
1
is 𝑇𝑓𝑙 𝑢+V
and the actual scheduling time of the last flight is 𝑇𝑓𝑙 , 𝐷delay
𝑢+V
so all the time for scheduling the flight queue is 𝑇all = (𝑇𝑓𝑙 −
1
{𝑆𝑇𝐷𝑖 − 𝐸𝑇𝐷𝑖 − Δ𝑡 𝑓𝑖 ∈ 𝐹𝐷𝐷, 𝑆𝑇𝐷𝑖 − 𝐸𝑇𝐷𝑖 > Δ𝑡 (3)
𝑇𝑓𝑙 ). In order to maximize the throughput of the runway, it is ={
0 𝑓𝑖 ∈ 𝐹𝐷𝐷, 𝑆𝑇𝐷𝑖 < 𝐸𝑇𝐷𝑖 .
required that the time of finishing the scheduling of the flight {
queue is minimum. So the smaller 𝑇all is, the larger runway
throughput is. So, the delay time of flight 𝑓𝑖 is

𝑙 𝑙
{𝐻𝐴 (𝑆𝑇𝐴 𝑖 − 𝐸𝑇𝐴 𝑖 − Δ𝑡) + 𝐻𝐷 (𝑆𝑇𝐷𝑖 − 𝐸𝑇𝐷𝑖 − Δ𝑡) 𝑆𝑇𝐴 𝑖 − 𝐸𝑇𝐴 𝑖 > Δ𝑡 or 𝑆𝑇𝐷𝑖 − 𝐸𝑇𝐷𝑖 > Δ𝑡
𝑇 (𝑓𝑖 ) = { (4)
0 𝑆𝑇𝐴 𝑖 < 𝐸𝑇𝐴 𝑖 or 𝑆𝑇𝐷𝑖 < 𝐸𝑇𝐷𝑖 .
{

The delay cost of the flight queue. The total delay cost of the entire flight queue is
Suppose that there is 𝑎 flight whose actual time is later
than estimated time; total delay cost is 𝑊𝐿 . Meanwhile, there 𝑊 = 𝑊𝐿 + 𝑊𝐸 . (7)
is 𝑏 flight whose actual time is earlier than estimated time,
Assume that there are 𝑔 kinds of sorting programs and the
because delay time is 0, so total delay cost is 𝑊𝐸 = 0.
delay cost of each program is 𝑊1 , 𝑊2 , . . . , 𝑊𝑔 , so the objective
So, the total number of 𝑎 + 𝑏 flights is all the approach and
function of flight delays cost is
departure flight 𝑢 + V:
𝑎 + 𝑏 = 𝑢 + V. (5) Cost = min (𝑊1 , 𝑊2 , . . . , 𝑊𝑔 ) . (8)

The delay cost of all the late flights is

2.4.3. The Objective Function of Controller Workload. The
𝑙 𝑚 object of controller workload can be achieved by the number
𝑖
𝑊𝐿 = ∑∑ [𝐶𝐴𝐻 𝐻𝐴𝑙 (𝑆𝑇𝐴 𝑖 − 𝐸𝑇𝐴 𝑖 − Δ𝜏) of the flight being adjusted; the larger it is, the heavier
𝑙=1 𝑖=1 (6) workload the controller has. Consequently, in order to reduce
𝑖 𝑙 the controller workload, we should try to keep the actual
+ 𝐶𝐷𝐻 𝐻𝐷 (𝑆𝑇𝐷𝑖 − 𝐸𝑇𝐷𝑖 − Δ𝜏)] . scheduling time of the flight consistent with the estimated
Mathematical Problems in Engineering 5

𝑢
time, so that we can reduce the number of flights being
∑ 𝛼𝑖𝑗 = 1
adjusted and also improve the robustness of the entire flight 𝑗=1
queue. The standard deviation of position changing value for (21)
each flight in the flight queue is introduced as a measurement V
for controller workload. ∑𝛽𝑖𝑗 = 1, ∀𝑖 ∈ 𝐹𝑖 , 𝑖 ≠ 𝑗
Here, 𝑥𝑖 is defined as the position changing value of the 𝑗=1

flight 𝑓𝑖 , so 𝑥𝑖 = |𝑓𝑖𝑞 − 𝑓𝑖𝑝 |. 𝑥 is the average variable quantity 󵄨󵄨 󵄨

󵄨󵄨𝑓𝑖𝑞 − 𝑓𝑖𝑝 󵄨󵄨󵄨 ≤ 𝑑. (22)
of displacement: 󵄨 󵄨

Equation (13) indicates that adjacent flight should meet

1 𝑢+V 1 𝑢+V 󵄨󵄨 󵄨
𝑥= ∑ 𝑥𝑖 = ∑ 󵄨󵄨𝑓 − 𝑓𝑖𝑝 󵄨󵄨󵄨󵄨 . (9) certain wake separation and vortex separation. 𝑢 + V is the
𝑢 + V 𝑖=1 𝑢 + V 𝑖=1 󵄨 𝑖𝑞 number of flight, 𝑡𝑓𝑖 is the scheduling time allocated to flight
𝑓𝑖 , 𝑚𝑖,𝑖−1 is the vortex separation, and 𝑤𝑖,𝑖−1 is the wake
𝜎2 is the variance of flight variable quantity of displacement: separation.
Equation (14) is to restrain the delay time of the approach
flight; it should not be larger than the maximum approach
1 𝑢+V 2 delay time.
𝜎2 = ∑ (𝑥𝑖 − 𝑥) . (10)
𝑢 + V 𝑖=1 Equation (15) is to restrain the delay time of the departure
flight; it should not be larger than the maximum departure
delay time.
𝜎 is the standard deviation of flight variable quantity of
Equations (16) and (17) are the constraint of runways;
displacement:
each flight only has one runway for arriving or leaving.
Equation (18) indicates the safety interval the flights need
1 𝑢+V 2
to meet if there are two successive flights arriving or leaving
𝜎=√ ∑ (𝑥 − 𝑥) . (11) the same runway (suppose that flight 𝑓𝑖 is the former and the
𝑢 + V 𝑖=1 𝑖
flight 𝑓𝑗 is the latter).
Equation (19) indicates that, for two approach flights 𝑓𝑖
The larger the 𝜎 value is, the greater workload the controller and 𝑓𝑗 , either 𝑓𝑖 is the leading and 𝑓𝑗 is the tailing or 𝑓𝑖 is the
has. So, the objective function of controller workload is tailing and 𝑓𝑗 is the leading.
Equations (20) and (21) indicate that any approach flight
only has one leading/tailing.
1 𝑢+V 2 Equation (22) indicates that the variable quantity of
Workload = min 𝜎 = √ ∑ (𝑥 − 𝑥) . (12)
𝑢 + V 𝑖=1 𝑖 displacement should not exceed maximum displacement
constraint.

2.5. Constraints. Consider

2.6. Design of Sliding Time Window Algorithm. Flight sorting
is a continuous dynamic process. Make sorting of 𝑢 + V
𝑡𝑓𝑖 − 𝑡𝑓𝑖−1 ≥ max (𝑚𝑖,𝑖−1 , 𝑤𝑖,𝑖−1 ) , approach and departure flights; there will be (𝑢 + V)! results.
(13) Among these results, there must be a flight sequence meeting
𝑖 = 1, 2, . . . , 𝑢 + V
all objectives, but the calculation is great and computational
𝑆𝑇𝐴 𝑖 − 𝐸𝑇𝐴 𝑖 ≤ 𝑇𝐴𝑌max , (14) complexity is high. However, the dynamical time window
algorithm can largely reduce the amount of computation;
𝑆𝑇𝐷𝑖 − 𝐸𝑇𝐷𝑖 ≤ 𝑇𝐷𝑌max , (15) its idea is as follows: according to the current real-time
information, the former 𝑝 flight waiting to be sorted forms
𝐻𝐴𝑙 ≤ 1, (16) a time window, makes full permutation of these 𝑝 flights, and
selects the 𝑞 flight with the best result as the former 𝑞 flight
𝑙
𝐻𝐷 ≤ 1, (17) for the final sorting result and keeps its order. Then, the time
window moves 𝑞 positions backward, reconstitutes a time
𝑆𝑇𝐷𝑗 ≥ 𝑆𝑇𝐷𝑖 + 𝛿𝑖𝑗 , 𝑖 ≠ 𝑗, (18) window which contains 𝑝 flight, seeks for the optimal 𝑝 flight
in the time window, and picks out the best 𝑞 flight adding to
𝛼𝑖𝑗 + 𝛼𝑗𝑖 ≤ 1, the final flight sorting result. A similar operation is repeated
(19) until all flight sorting has been finished, thus achieving the
𝛽𝑖𝑗 + 𝛽𝑗𝑖 ≤ 1, ∀ (𝑖, 𝑗) ∈ 𝐹𝑖 , 𝑖 ≠ 𝑗,
final total flight sorting results. Here, 𝑝 is the size of time
𝑢 window, 𝑞 is the sliding step, and 0 < 𝑝 < 𝑛.
∑𝛼𝑖𝑗 = 1 For example, 𝑢 + V = 9, 𝑝 = 5, and 𝑞 = 2; the process is
𝑖=1 shown in Figure 2.
(20)
V
∑𝛽𝑖𝑗 = 1, ∀𝑗 ∈ 𝐹𝑖 , 𝑖 ≠ 𝑗 Step 1. In the initial flight queue, put flights 1–5 into the time
𝑖=1 window; make full permutation of these five flights to find
6 Mathematical Problems in Engineering

Step 1 1 2 3 4 5 6 7 8 9 ··· 3 1 5 2 7 6 4 8

Step 2 4 1 2 5 3 6 7 8 9 ··· 0 1 1 0 0 1 0 1

Step 3 4 1 3 7 2 5 6 8 9 ··· Figure 3

Step 4 4 1 3 7 2 8 5 6 9 ···
3 1 3 1 3 2 1 1
Step 5 4 1 3 7 2 8 5 9 6 ···
0 1 1 0 0 1 0 1
Figure 2: The diagrammatic sketch of dynamical time window
algorithm. Figure 4

a meaningful flight sequence and reduce the complexity of

the best sorting result. At last, put flight 4 and flight 1 as the the algorithm.
top two flights in the final result. According to Grefenstette coding method, the double
chromosome given above could be described as shown in
Step 2. The time window moves 2 positions backward; flight Figure 4.
6 and flight 7 are included into the time window. Make full
permutation of flights 2, 3, 5, 6, and 7 and find the best sorting 2.7.2. Generate Initial Population. Apply Grefenstette coding
result. Put the former two flights 6 and 7 into the final result. method to generate double chromosomes randomly, and each
of them is meaningful.
Step 3. The time window continues moving 2 positions
Judge if the chromosomes are qualified or not: firstly,
backward; flight 8 and flight 9 are included into the time
decode the chromosomes above; then, make sorting of the
window. Make full permutation of flights 2, 5, 6, 8, and 9 and
flights on each runway according to their estimated approach
find the best sorting result. Put the former two flights 2 and 8
or departure time, and assign scheduling time for each flight
into the final result.
on the basis of constraints. If it is prior to the earliest time,
Step 4. The time window keeps on moving 2 positions make the scheduling time equal to the earliest time. If it
backward; the subsequent two flights are included into the is later than the latest time, then the individual is invalid
time window. Make full permutation of these five flights in and needs to generate new individual, until it meets the
the time window and find the best sorting result. Put flight 5 constraints.
and flight 9 into the final result.
2.7.3. Fitness Function. Fitness plays the role of judging if
Step 5. The time window goes on moving backward, repeats the individual is good or bad and drives the group moving
the above steps until the last flight finishes sorting, and then to the best evolution direction in the process of genetic
gets the best final sorting result, that is, 4-1-3-7-2-8-5-9-6. algorithm. Because the fitness function is always nonnegative,
under any circumstances, the bigger it is, the better it is. The
model established in this paper is a multiobject optimization
2.7. Design of Improved Genetic Algorithm. According to the model; make improvement of general genetic algorithm; the
characteristics of multirunway flight scheduling problem, it sorting method according to pros and cons of all objectives is
is needed to make improvement of general genetic algorithm. determined to calculate fitness. Make descending order of the
Conduct encoding, selection, crossover, and mutation opera- individual according to their quality of the performance. As
tions of the flights in the window, thereby generating optimal for each object 𝑓𝑖 (𝑥), all the individuals will generate a feasible
sequence. sequence 𝑋𝑖 on the basis of the value of the object function.
Sorting for each object, we can get overall performance of the
2.7.1. Coding Scheme. The flight scheduling problem studied individual for all object functions. The calculation of fitness
in this paper is a multirunway problem, so make improve- can be calculated by [19]
ment of coding scheme and design double string gene coding
2
based on the flight number and runway number. The first {(𝑁 − 𝑅𝑜 (𝑋𝑜 )) 𝑅𝑜 (𝑋𝑗 ) > 1
code string is flight chromosome, showing flight number; 𝐹𝑜 (𝑋𝑗 ) = { (23)
2
the second code string is runway chromosome, showing the {𝑘𝑁 𝑅𝑜 (𝑋𝑗 ) = 1
corresponding runway number of the flight (assuming there 𝑛
are two runways, resp., 0 and 1). Assign a runway for each 𝐹 (𝑋𝑗 ) = ∑𝐹𝑖 (𝑋𝑗 ) . (24)
flight randomly as shown in Figure 3. 𝑖=1
In the process of genetic operation for double chromo-
somes, the crossover, mutation operation makes it easy to In (23), 𝑜 = 1, 2, . . . , 𝑛; 𝑗 = 1, 2, . . . , 𝑁; 𝑛 is the number
produce meaningless flight sequence. In order to avoid it, of objective functions; 𝑁 is all individuals of the population;
Grefenstette coding method is used to produce efficient chro- 𝑋𝑗 is the individual 𝑗 of the population; 𝑅𝑜 (𝑋𝑗 ) is the serial
mosomes. So each individual chromosome could represent number of the individual 𝑗 after all individuals are being
Mathematical Problems in Engineering 7

sorted in view of the objective 𝑜; 𝐹𝑜 (𝑥𝑗 ) is the fitness value of A 3 1 3 1 3 2 1 1

the individual 𝑋𝑗 on account of the objective 𝑜; 𝐹(𝑋𝑗 ) is the 0 1 1 0 0 1 0 1
comprehensive fitness value of the individual 𝑋𝑗 on account
of all objectives; 𝑘 is a constant within (1, 2), for increasing the B 2 2 1 3 1 1 2 1
fitness value when the performance of individual is optimal.
1 1 0 0 1 0 1 1
As can be seen from the equations above, optimal overall
performance of individuals can get better fitness to gain more Figure 5
opportunities to participate in evolution.

2.7.4. Selection. Selection is the process to choose the individ- A󳰀 3 1 3 1 1 1 2 1

ual with strong vitality to generate new population. Selection 0 1 1 0 1 0 1 1
operator is used to select the survival of the fitness: the
individual with higher fitness has more probability inherited B󳰀 2 2 1 3 3 2 1 1
to next generation, while the individual with lower fitness has 1 1 0 0 0 1 0 1
little probability inherited to next generation. The roulette
selection operator is used in this paper; namely, the proba- Figure 6
bility of the fitness in proportion decides the possibility of its
descendants going or staying. If a certain individual is 𝑋𝑗 , its
fitness is 𝐹(𝑋𝑗 ); the probability to be selected can be expressed
operation again, until producing the offspring that is consis-
as
tent with the constraints.
𝐹 (𝑋𝑗 )
𝑃𝑋𝑗 = . (25)
∑𝑁
𝑗=1 𝐹 (𝑋𝑗 ) 3. Simulation and Analysis
3.1. Simulation. In order to verify the performance of the
2.7.5. Crossover. Crossover is to pick up two individuals algorithm proposed in this paper, the data of 24 flights in
from the population with higher fitness and change a bit or a continuous period based on Chengdu Shuangliu Interna-
some bits of them to produce new individuals. The offspring tional Airport is shown as an example. There are 12 approach
inherits the basic characteristics of their parents. Due to flights and 12 departure flights and 2 mutually independent
the use of Grefenstette coding method, there is no need to parallel runways. Moreover, Δ𝜏 = 120 s. The initial flight data
worry about the meaninglessness of the chromosome after is shown in Table 2.
crossover. Single-point crossover method is used in this paper Run the algorithm in this paper. Based on the selected
[19]. For example, write two flight sequences A and B as 24 approach and departure flights, set the size of time
parents and randomly select crossover point, shown as in window 𝑝 = 15 and sliding step 𝑞 = 3. According to the
Figure 5. estimated time, number 1–7 flights and number 12–20 flights
After performing a simple single-point crossover, we can are included in the window. Apply genetic algorithm in the
get what is shown in Figure 6. window for these 15 flights; select the former 3 flights of the
As for the new generated individuals after crossover, best sequence as the former 3 flights of the final sequence.
the constraints are applied to make a judgment; if the The time window moves 3 positions backward and repeats
offspring does not meet the constraints, then execute the the above steps until the last flight sorting is finished. Set
crossover operation again, until producing the offspring that displacement constraint 𝑑 = 8. In the genetic algorithm,
is consistent with the constraints. the crossover probability is 0.8, the mutation probability is
0.2, the generation gap is 0.9, the elimination rate is 0.2,
2.7.6. Mutation. In order to speed up the operation efficiency 𝑘 in the fitness function is 1.5, the size of population is
of the algorithm, design two kinds of mutation operator 100, and evolution algebra is 2000. After several rounds of
according to the coding scheme. selection, crossover, and mutation, the evolutionary trends of
the genetic algorithm are shown in Figure 7.
(1) Exchange two consecutive flight chromosomes on the
same runway. Select a certain position of a runway
sequence randomly and exchange two consecutive 3.2. Analysis. Make sorting of the flight according to the
flight chromosomes of the position. optimization algorithm and select 6 kinds of optimization
(2) Exchange two flight chromosomes at any position on schemes; their values of each objective function are shown
the different runway. Select a certain position of two in Table 3.
runway sequences randomly and exchange their flight Draw the comparison diagram of 6 optimization schemes
chromosomes. according to Table 3, and compare them with the results of
FCFS algorithm. The delay cost of FCFS is 101115.8/CNY;
As for the new individuals after mutation, the constraints the runway throughput is 1172/s. The comparison results are
are applied to make a judgment; if the offspring generated shown in Figure 8. As can be seen from Figure 8, compared
does not meet the constraints, then execute the mutation with FCFS, the results of the optimization algorithm have
8 Mathematical Problems in Engineering

Table 2: The initial data of approach and departure flights.

Airline Number Flight number Type Unit time delay cost Estimated Runway
of departure departure time
CES 1 MU5990 M 2 0:00:00 0
CES 2 MU2342 L 1 0:00:00 1
CCA 3 CA2342 L 1.1 0:00:00 0
CSC 4 3U8731 M 2.1 0:00:00 1
CSZ 5 ZH1407 M 2.2 0:04:00 1
CSZ 6 ZH4307 M 2.2 0:04:00 0
CCA 7 CA426 L 1.2 0:04:00 1
CCA 8 CA4307 H 4 0:04:00 0
CSC 9 3U8668 L 1.1 0:08:00 0
CES 10 MU2414 L 1 0:08:00 1
CES 11 MU3005 M 2.1 0:08:00 0
CSC 12 3U8869 H 4.1 0:08:00 1

Airline Number Flight number Type Unit time delay cost Estimated Runway
of approach approach time
CSC 13 3U8676 M 42 0:00:00 0
CCA 14 CA4434 H 62 0:01:06 1
CSC 15 3U8648 L 23 0:01:47 0
CCA 16 CA1415 L 23.1 0:02:35 1
CSZ 17 ZH1415 M 42.2 0:03:19 0
CSC 18 CA1945 L 23 0:03:55 1
CES 19 MU5401 M 41.3 0:04:07 0
CSZ 20 ZH1915 H 62.1 0:05:14 1
CSC 21 3U8702 M 42 0:02:52 0
CSC 22 3U8628 L 23 0:06:23 1
CCA 23 CA4392 L 23.1 0:07:34 0
CSZ 24 ZH2306 M 42.2 0:08:48 1
Note: CCA: Air China; CSC: Sichuan Airlines; CSZ: Shenzhen Airlines; CES: China Eastern; CHH: Hainan Airlines.

×104 According to the value of each objective function in

4.5 Table 3, suppose that 6 kinds of schemes are 𝑥1 , 𝑥2 , . . . 𝑥6 ; the
4 three objective functions are 𝐴 1 , 𝐴 2 , 𝐴 3 , so that
Change of solutions

3.5
3 𝑦11 = 𝐴 1 (𝑥1 ) = 1018,
2.5
2 𝑦12 = 𝐴 1 (𝑥2 ) = 1100,
1.5
1
𝑦13 = 𝐴 1 (𝑥3 ) = 1072,
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Genetic algebra 𝑦14 = 𝐴 1 (𝑥4 ) = 1022,
Figure 7: The evolution trend graph of genetic algorithm. 𝑦15 = 𝐴 1 (𝑥5 ) = 994,

𝑦16 = 𝐴 1 (𝑥6 ) = 1066,

been improved greatly compared with the FCFS algorithm
with lower delay cost and higher runway throughput. 𝑦21 = 𝐴 2 (𝑥1 ) = 64988.8,
As for the selected 6 optimization schemes, based on mul-
tiobjective decision, the optimal cardinal number summation 𝑦22 = 𝐴 2 (𝑥2 ) = 56622.5,
method is adopted to select the best scheme in this paper [20].
Specific method is described as follows. 𝑦23 = 𝐴 2 (𝑥3 ) = 59488.7,
Mathematical Problems in Engineering 9

𝑦24 = 𝐴 2 (𝑥4 ) = 41369.5, ×103

120 1200
𝑦25 = 𝐴 2 (𝑥5 ) = 67218.4, 1173
100 1150
𝑦26 = 𝐴 2 (𝑥6 ) = 44634.6,

Runway throughput (s)

Delay costs/CNY
80 1100 1100
𝑦31 1
= 𝐴 3 (𝑥 ) = 2.7346, 1072 1066
60 1050
𝑦32 2
= 𝐴 3 (𝑥 ) = 2.6812, 1018
1022
40 1000
994
𝑦33 = 𝐴 3 (𝑥3 ) = 2.3886,
20 950
𝑦34 4
= 𝐴 3 (𝑥 ) = 1.9539,
0 900

Optimization 1

Optimization 2

Optimization 3

Optimization 4

Optimization 5

Optimization 6

FCFS
𝑦35 5
= 𝐴 3 (𝑥 ) = 2.8974,

𝑦36 = 𝐴 3 (𝑥6 ) = 2.0648.

(26)
The comparison of optimization schemes and FCFS
The quantitative matrix is Delay cost/CNY
Runway throughput (s)
𝑌
Figure 8: Comparison of 6 optimization schemes with the FCFS of
1018 1100 1072 1022 994 1066 object function value.
[ ] (27)
= [64988.8 56622.5 59488.7 41369.5 67218.4 44634.6] .
[ 2.7346 2.6812 2.3886 1.9539 2.8974 2.0648 ]
𝑢5 = 0.7846,
Find the minimum value of the objective function; the
calculation of optimal cardinal number is shown in the 𝑢6 = 0.8776.
following formula: (31)
𝑦𝑘𝑖 Because of 𝑢4 > 𝑢6 > 𝑢3 > 𝑢2 > 𝑢1 > 𝑢5 , the
𝑧𝑘𝑖 = 1 − . (28)
√∑𝑠𝑖=1 𝑦𝑘𝑖 2 number 4 scheme is the best scheme under the given weight
coefficient. In addition, the decision makers can change the
Here, 𝑖 is the number of optimization schemes; 𝑘 is the weight coefficient of each objective function according to
number of object functions; that is, 𝑖 = 1, 2, . . . , 6, 𝑘 = 1, 2, 3. their preferences, thereby selecting the best scheme to meet
So the optimal cardinal number matrix is their demands.
The results of the number 4 scheme are shown in Table 4.
𝑍 The genetic algorithm combined with the dynamical
time window algorithm is applied in this paper, making
0.8421 0.8157 0.8249 0.8409 0.8495 0.8269 optimization scheduling of the approach and departure
[0.7799 0.8329 0.8156 0.9108 0.7646 0.8962] (29)
=[ ]. flights, and obtaining ideal results. Compared with the FCFS
algorithm, the runway throughput is improved by 12.87%;
[0.7971 0.8049 0.8452 0.8964 0.7722 0.8843] delay cost is reduced by 61.46%. In addition, in order to
Now suppose that the weight coefficient of each object verify the effect of the algorithm, make a comparison of the
function is 0.2, 0.4, and 0.4. According to formula (30), make genetic algorithm combined with dynamical time window
weighted sum and sorting of these optimal cardinal numbers: optimization algorithm used in this paper and the general
genetic algorithm used in other literature. The comparison
𝑚 results are shown in Table 5.
𝑢𝑖 = ∑ 𝑤𝑘 ⋅ 𝑧𝑘𝑖 . (30) According to Table 5, the delay cost of three algorithms
𝑘=1 for each flight is compared; the 24 flight delays’ cost compar-
So, ison chart is shown in Figure 9.
Compared with the FCFS algorithm, the runway
𝑢1 = 0.7992, throughput of genetic algorithm was improved by 13.21%
compared to FCFS; the delay cost decreased by 62.95%;
𝑢2 = 0.8183, compared with the general genetic algorithm, the runway
𝑢3 = 0.8293, throughput of the genetic algorithm combined with
dynamical time window algorithm only reduced by 0.39%;
𝑢4 = 0.8911, delay cost only increased by 3.86%, while the controller
10 Mathematical Problems in Engineering

Table 3: The objective function value of the 6 optimization schemes.

Optimization schemes 1 2 3 4 5 6
Runway throughput/s 1018 1100 1072 1022 994 1066
Delay cost/CNY 64988.8 56622.5 59488.7 41369.5 67218.4 44634.6
Controller displacement 2.7346 2.6812 2.3886 1.9539 2.8974 2.0648

Table 4: The optimized flight data.

Runway 0 Runway 1
Airline Number Flight number Type Time/s Delay cost/CNY Airline Number Flight number Type Time/s Delay cost/CNY
CSC 13 3U8676 M 0 0 CSC 15 3U8648 L 0 0
CSZ 17 ZH1415 M 74 0 CCA 3 CA2342 L 96 0
CES 19 MU5401 M 148 0 CCA 14 CA4434 H 162 0
CES 1 MU5990 M 224 208 CES 2 MU2342 L 302 182
CSC 21 3U8702 M 300 336 CCA 16 CA1415 L 398 2841.3
CSC 4 3U8731 M 376 537.6 CCA 8 CA4307 H 464 416
CSZ 24 ZH2306 M 452 0 CSZ 20 ZH1915 H 542 6706.8
CSZ 5 ZH1407 M 528 369.6 CCA 7 CA426 L 682 386.4
CES 11 MU3005 M 602 4.2 CSC 12 3U8869 H 756 639.6
CSC 18 CA1945 L 722 8441 CCA 23 CA4392 L 896 7438.2
CSZ 6 ZH4307 M 788 941.6 UEA 9 EU2705 L 992 431.2
CES 10 MU2414 L 926 326
CSC 22 3U8628 L 1022 11164
Runway throughput: 1022 s
Delay cost: 41369.5CNY

25000 4. Conclusion
20000 For the air traffic congestion problem, this paper develops
a new algorithm based on multiobjective optimization for
Delay cost/CNY

15000 flight scheduling model. The runway throughput, flight delays

cost, and controller workload are regarded as objective,
10000 respectively. A new improved genetic algorithm combined
with a dynamical time window algorithm is created in the
5000 paper. Compared with FCFS and general genetic algorithm,
the main advantages of the model and improved algorithm
0
proposed in this paper are as follows:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

Flight number (1) The runways studied in this paper are multiple parallel
FCFS Optimization algorithm runways, which accords with the development trend
Genetic algorithm of airport nowadays. According to the characteristic
of multirunway, double string gene coding based on
Figure 9: Comparison of the delay costs of 24 flights for three the flight number and runway is created to make
algorithms.
improvement of genetic algorithm, and the optimal
results getting in this method are more ideal.
workload has decreased by 26.35%; meanwhile, as for the (2) The multiobjective optimization could find an opti-
complexity of the algorithm, the optimization algorithm mum solution of the flight scheduling. According
increased nearly by 62.5% compared with general genetic to the characteristic of multiobjective optimization
algorithm, which greatly saves the calculation time. Although model, define the fitness value of individual by the
the optimization algorithm used in this paper did not achieve comprehensive values of all optimization objects;
the best global optimization results as the general genetic then, get the better offspring. Besides, the decision
algorithm, it reduced the computational complexity, makers could change the weight coefficient of the
improved efficiency, and reduced controller workload to a objective functions according to their need, so they
certain degree. Thus, the optimization algorithm used in can get the satisfying flight sequence.
this paper could also get a good optimization result and (3) The displacement constraint of the flight moving
significantly saved the calculation time. reflected the fairness among the flights taking off
Mathematical Problems in Engineering 11

Table 5: The delay cost of 3 kinds of algorithms.

FCFS Genetic algorithm Optimized algorithm

Number Estimated Actual Departure Runway Optimized Runway
Runway Delay cost Delay cost Delay cost
departure time departure time time departure time
1 0:00:00 0:00:00 0 0 0:10:50 1 1060 0:03:44 0 208
2 0:00:00 0:00:00 1 0 0:00:00 1 0 0:05:02 1 182
3 0:00:00 0:02:34 0 37.4 0:01:38 1 0 0:01:36 1 0
4 0:00:00 0:02:02 1 4.2 0:09:36 1 957.6 0:06:16 0 537.6
5 0:04:00 0:08:54 1 382.8 0:11:44 0 756.8 0:08:48 0 369.6
6 0:04:00 0:09:43 0 490.6 0:10:30 0 594 0:12:48 0 941.6
7 0:04:00 0:11:12 1 374.4 0:03:14 0 0 0:11:22 1 386.4
8 0:04:00 0:10:57 0 1188 0:07:34 0 376 0:07:44 1 416
9 0:08:00 0:16:29 0 427.9 0:15:20 0 352 0:16:32 1 431.2
10 0:08:00 0:16:41 1 401 0:16:58 0 418 0:15:26 0 326
11 0:08:00 0:17:43 0 972.3 0:08:22 1 0 0:10:02 0 4.2
12 0:08:00 0:17:55 1 1947.5 0:13:56 1 967.6 0:12:36 1 639.6
FCFS Genetic algorithm Optimized algorithm
Number Estimated Actual Approach Runway Optimized Runway
Runway Delay cost Delay cost Delay cost
approach time approach time time approach time
13 0:00:00 0:03:50 0 4620 0:01:14 0 0 0:00:00 0 0
14 0:01:06 0:03:23 1 1054 0:02:44 1 0 0:02:42 1 0
15 0:01:47 0:05:59 0 3036 0:04:50 0 1449 0:00:00 1 0
16 0:02:35 0:06:10 1 2194.5 0:06:28 0 3072.3 0:06:38 1 2841.3
17 0:03:19 0:08:27 0 7933.6 0:00:00 0 0 0:01:14 0 0
18 0:03:55 0:07:48 1 2599 0:12:50 1 9545 0:12:02 0 8441
19 0:04:07 0:12:35 0 16024.4 0:05:52 1 0 0:02:28 0 0
20 0:05:14 0:12:18 1 18878.4 0:08:52 0 4693.6 0:09:02 1 6706.8
21 0:02:52 0:07:15 0 6006 0:04:38 1 0 0:05:00 0 336
22 0:06:23 0:15:05 1 9246 0:16:16 1 9756.2 0:17:02 0 11164
23 0:07:34 0:14:53 0 7368.9 0:13:44 0 5775 0:14:56 1 7438.2
24 0:08:48 0:19:33 1 22155 0:07:06 1 0 0:07:32 0 0
Total 101115.8/CNY 39773.1/CNY 41369.5/CNY
Controller workload 2.6531 1.9539

and landing, reduced the workload of the controllers, Acknowledgments

and improved the robustness of the flight sequence.
It improved the efficiency greatly with less time The authors thank the reviewers for helping them to improve
complexity, which provided good guidance for air this paper. This work is supported by the Fundamental
traffic controllers to make flight scheduling. Research Funds for the Central Universities, no. NS2014064.

However, because of the complexity of flight scheduling References

of multiobjective on multirunway, it is needed to make
adjustment according to real-time information. So, in the [1] D. A. Lee, C. Nelson, and G. Shapiro, The Aviation System
future study, make further improvement of the algorithm Analysis Capability Airport Capacity and Delay Models, National
Aeronautics and Space Administration, Langley Research Cen-
so as to fit the actual complex scheduling environment and
ter, 1998.
achieve a mutually satisfactory outcome.
[2] P. B. Vranas, D. Bertsimas, and A. R. Odoni, “Dynamic ground-
holding policies for a network of airports,” Transportation
Conflict of Interests Science, vol. 28, no. 4, pp. 275–291, 1994.
[3] J. C. Beck and P. Refalo, “A hybrid approach to scheduling with
The authors declare that there is no conflict of interests earliness and tardiness costs,” Annals of Operations Research,
regarding the publication of this paper. vol. 118, no. 1–4, pp. 49–71, 2003.
12 Mathematical Problems in Engineering

[4] A. Mukherjee, “Dynamic stochastic optimization models for

air traffic flow management,” Working Papers, Institute of
Transportation Studies, Berkeley, Calif, USA, 2004.
[5] A. D’Ariano, P. D’Urgolo, D. Pacciarelli, and M. Pranzo, “Opti-
mal sequencing of aircrafts take-off and landing at a busy
airport,” in Proceedings of the 13th International IEEE Conference
on Intelligent Transportation Systems (ITSC ’10), pp. 1569–1574,
Funchal, Portugal, September 2010.
[6] Y. Eun, I. Hwang, and H. Bang, “Optimal arrival flight
sequencing and scheduling using discrete airborne delays,”
IEEE Transactions on Intelligent Transportation Systems, vol. 11,
no. 2, pp. 359–373, 2010.
[7] Z.-H. Zhan, J. Zhang, Y. Li et al., “An efficient ant colony
system based on receding horizon control for the aircraft arrival
sequencing and scheduling problem,” IEEE Transactions on
Intelligent Transportation Systems, vol. 11, no. 2, pp. 399–412,
2010.
[8] M. Pistelli, A. D’Ariano, and D. Pacciarelli, “Optimization
models and algorithms for air traffic control in the airspace of
busy airports,” Tech. Rep. RT-DIA-185-11, Roma Tre University,
2011.
[9] J. A. Bennell, M. Mesgarpour, and C. N. Potts, “Airport runway
scheduling,” 4OR, vol. 9, no. 2, pp. 115–138, 2011.
[10] M. Samà, A. D’Ariano, and D. Pacciarelli, “Rolling horizon
approach for aircraft scheduling in the terminal control area of
busy airports,” Procedia—Social and Behavioral Sciences, vol. 80,
pp. 531–552, 2013.
[11] M. Ignaccolo and G. Inturri, “Modelli Analitici e di Simulazione
perla Valutazione delle Prestazioni in ambito Aeroportuale,” in
Metodi e Tecnologie perl'Ingegneria dei Trasporti, Franco Angeli,
Milan, Italy, 2001.
[12] Y. Jiang and P. Yun, “Mathematical model and analysis of run-
way landing capacity,” Journal of Southwest Jiaotong University,
vol. 39, no. 1, pp. 42–46, 2004.
[13] L.-L. Wang, Z.-K. Shi, and Z.-N. Zhang, “Optimizing slip win-
dow algorithm for sequencing of scheduled arriving aircrafts,”
Journal of Civil Aviation, vol. 22, no. 6, pp. 18–21, 2004.
[14] B. Jiang and F.-Q. Zhang, “Approach scheduling model based
on earliest estimated arrival time and its algorithm,” Journal of
Southwest Jiaotong University, vol. 40, no. 4, pp. 509–512, 2005.
[15] K. Yang, “Arriving aircrafts scheduling algorithm based on
adaptive slip window,” Journal of Sichuan University (Natural
Science Edition), vol. 46, no. 4, pp. 963–967, 2009.
[16] Z. Huang, C. Bai, and W. Zhang, “Genetic algorithm for
arrival sequencing and scheduling optimization,” Advances in
Aeronautical Science and Engineering, vol. 2, no. 2, pp. 236–240,
2011.
[17] J. Zhang and W. Zhang, “Arriving and leaving aircrafts schedul-
ing model based on dynamic optimize,” Journal of Sichuan
University (Natural Science Edition), vol. 49, no. 1, pp. 90–96,
2012.
[18] H. Zhou and X. Jiang, “Research on arrival/departure schedul-
ing of flights on multirunways based on genetic algorithm,”
Mathematical Problems in Engineering, vol. 2014, Article ID
851202, 13 pages, 2014.
[19] R. Han, Genetic Algorithms and Applications, Ordnance Indus-
try Press, Beijing, China, 2010.
[20] Y. Hu, Multi-Objective Decision: Practical Models and Optimal
Selection Method, Shanghai Science and Technology Press,
Shanghai, China, 2010.
Copyright of Mathematical Problems in Engineering is the property of Hindawi Publishing
Corporation and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.