Content uploaded by Michael Schultz
Author content
All content in this area was uploaded by Michael Schultz on Sep 24, 2021
Content may be subject to copyright.
Fourteenth USA/Europe Air Traffic Management Research and Development Seminar (ATM2021)
COVID-19: Passenger Boarding and Disembarkation
Michael Schultz∗, Majid Soolaki†, Elnaz Bakhshian††, Mostafa Salari‡, J ¨
org Fuchte§
∗Dresden University of Technology
Insitute of Logistics and Aviation
Dresden, Germany
University College Dublin
†Lochlann Quinn School of Business
†† School of Civil Engineering
Dublin, Ireland
‡University of Calgary
Department of Civil Engineering
Calgary, Canada
§Diehl Aviation
Hamburg, Germany
Abstract—Boarding and disembarking an aircraft is a time-
critical airport ground handling process. Operations in the
confined aircraft cabin must also reduce the potential risk
of virus transmission to passengers under current COVID-
19 boundary conditions. Passenger boarding will generally
be regulated by establishing passenger sequences to reduce
the influence of negative interactions between passengers (e.g.,
congestion in the aisle). This regulation cannot be implemented
to the same extent when disembarking at the end of a flight. In
our approach, we generate an optimized seat allocation that
takes into account both the distance constraints of COVID-
19 regulations and groups of passengers traveling together
(e.g., families or couples). This seat allocation minimizes the
potential transmission risk, while at the same time we calculate
improved entry sequences for passengers groups (fast boarding).
We show in our simulation environment that boarding and
disembarkation times can be significantly reduced even if a
physical distance between passenger groups is required. To
implement our proposed sequences during real disembarkation,
we propose an active information system that incorporates the
aircraft cabin lighting system. Thus, the lights above each group
member could be turned on when that passenger group is
requested to disembark.
Keywords—Passenger disembarkation, virus transmission,
COVID-19, pandemic requirements, passenger groups in aircraft
cabin, two-objective mathematical modeling
I. INTRODUCTION
The COVID-19 situation will have a lasting impact on air
transportation in general and on airport operations (aircraft
handling) and passenger handling in particular. The current
pandemic situation requires two major changes to normal
aircraft handling procedures: (a) passengers must maintain
a certain distance when boarding and disembarking, and (b)
in addition to normal cleaning procedures, the aircraft cabin
must be disinfected to avoid potential virus transmission via
surface contact. Fig. 1illustrates that the required process
changes will have a significant impact on aircraft turnaround
time, as these processes are part of the critical operating path.
Studies that consider COVID-19 constraints for passenger
boarding [1] and aircraft cleaning [2] highlight the need
for appropriate process adjustments to mitigate the impact
of significantly increased process times. However, there has
not yet been an intensive scientific discussion of passenger
disembarkation sequences under COVID-19 conditions [3].
There are several approaches to infrastructural changes in
the aircraft cabin to reduce transmission risks, but most of
these ideas are far from being a flexible and standardized
Fueling
Catering
Cleaning
Passenger Handling
Critical Path
Disembarkation Boarding
COVID-19 Impact
Standard Turnaround Time
Figure 1: Impact of COVID-19 regulations on aircraft turnaround
operations, in particular during disembarkation, cleaning, boarding.
solution for the aviation industry. From an operational per-
spective, adapted boarding strategies are more likely to be
put into practice by airlines and airports than modified cabin
equipment. Disembarkation is more difficult to control by
regulation, and passengers show poor discipline and limited
willingness to behave in a compliant manner while disem-
barking from the aircraft. This is particularly noteworthy
because the risk of virus transmission is much higher during
uncontrolled disembarkation than during controlled boarding
of the aircraft [4].
In our approach, we assume that passengers travel in groups
and that boarding and disembarkation could be controlled by
the present technical infrastructure. The idea behind consider-
ing group constellations is that group members, in the form of
families or couples, were already in close contact with each
other before entering the aircraft and should not be subject
to spatial spacing rules. While we propose a standard call-in
procedure for boarding, we assume that groups of passengers
can be instructed to stand up and leave the aircraft through the
seat-based lighting system of the aircraft cabin. In addition,
passengers will initially be responsible for maintaining a
minimum distance from other groups themselves, with cabin
crew monitoring this process. In a later phase, this may
also be supported by new technologies for precise passenger
location and the associated automated monitoring and control
of passenger movements via personal devices (see Fig. 2).
Figure 2: Estimation of passenger position based on localization
framework in a digital connected cabin using stationary anchors (red
and green circles) and signals from mobile devices [5].
Aircraft cabins have difficult conditions for wireless signals
due to possible reflections, scattering and attenuation of trans-
mitted signals. In this context, ultra-wideband (UWB) tech-
nology will enable precise, real-time localization for indoor
applications and could provide reliable range measurements
for compliance to COVID-19 regulations [5]. UWB is already
being integrated into personal devices for the consumer mar-
ket. In the context of future passenger handling in the confined
aircraft environment, an efficient sensing environment will be
a key element for improved situational awareness of system
conditions (e.g., aisle occupancy or baggage compartment
status). This information can be used to further improve
operational efficiency and enable new product developments
and passenger-oriented services.
A. Review of state of the art
Comprehensive overviews are provided for passenger
boarding research [6–8] and aircraft ground operations [9].
Only a few aircraft boarding and disembarkation tests have
been conducted to provide data for the calibration of input
parameters and validation of simulation results: using a mock
Boeing 757 fuselage [10], time to store hand luggage items in
the overhead compartments [11], small-scale laboratory tests
[12], evaluation of passenger perceptions during boarding
and disembarkation [13], operational data and passenger data
from field trial measurements [14], field trials for real-time
seat allocation in connected aircraft cabin [15], and using a
B737-800 mock-up (1/3 size) to explore the factors affecting
the time of luggage storage [16]. Although these field data
are used for simulation experiments, they only cover regular
behavior in a pre-pandemic situation.
The particular movement behavior of pedestrians depends
significantly on group constellations (e.g. friends or families)
and impacts the self-organization capabilities of crowds [17–
19]. Also in the context of passenger dynamics in the airport,
it is an important fact that up to 70% of the tourists and 30%
of the business passengers are traveling in groups [20]. Thus,
group constellations are important to understand granular flow
patterns during boarding and disembarkation (e.g. couples or
families are not separated). Group behavior may shorten the
process times since conflicts during the seating process are
internally solved [21] and aircraft boarding by rows should
be a recommended practice [11]. An approach of dynamically
optimized boarding indicates that the boarding process bene-
fits from the consideration of groups [22]. Furthermore, less
complex boarding sequences (e.g. random or block boarding)
benefit more from the consideration of passenger groups
(approx. 5% faster boarding), while seat-based sequences
(separation of the window, middle, and aisle seats) lead to
longer boarding times [6].
While passenger boarding research exhibits a broad range
of improvements (e.g. group boarding, sequence optimiza-
tion, infrastructural changes), research in the specific field
of passenger disembarkation is quite limited and findings
often arise as a side product. Two general concepts are
addressed to analyze the efficacy of block-wise (aggregated
seat rows) or column-wise (e.g. all aisle seats) sequences.
Here, column-wise disembarkation was found to be more
effective for narrow-body aircraft [23]. These two structured
disembarkation sequences are analyzed in small-scale field
trials applying inside-out (column-wise) and back-to-front
(block-wise) sequences [24], but in contrast to the prior
simulation experiment [25], no significant improvements of
the disembarkation time could be demonstrated.
Currently, the research focus is set on efficient passenger
handling in the aircraft cabin during pandemic situations.
Standard boarding sequences are analyzed considering the
quantity and quality of passenger interactions and evaluated
with a virus transmission model to provide a more detailed as-
sessment. The implementation of physical distances indicates
that conventional boarding sequences take longer and trade-
offs between economic efficiency (seat load) and process
duration must be made to minimize the impact on various
health risks [26]. Adjusted seat allocation sequences are de-
veloped considering both distances to the aisle (ensure lower
transmission risks caused by aisle movements) and distance
between the occupied seats [27]. Furthermore, investigation
shows that physical distances between passengers decrease
the number of possible transmissions by approx. 75% for
random boarding sequences, and could further decreased by
more strict reduction of hand luggage items (less time for
storage, compartment space is always available) [4]. Further-
more, standard process times could be reached if the rear
aircraft door is used for boarding and disembarkation. This
investigation also points out that disembarkation consists of
the highest transmission potential and only minor benefits
from distance rules and hand luggage regulations. The op-
timized consideration of passenger groups in the context of a
pandemic boarding scenario will significantly contribute to a
faster process (reduction of time by about 60%) and a reduced
transmission risk (reduced by 85%), which reaches the level
of boarding times in pre-pandemic scenarios [1]. The results
of the passenger process evaluation considering the current
COVID-19 situation were taken as input to further inves-
tigate the impact of pandemic requirements on the aircraft
turnaround [2]. Here an integrated cleaning and disinfection
procedure was developed and optimized. Under COVID-19
constraints, aircraft turnarounds require between 10% (with
additional personnel) and 20% (without additional personnel)
more ground time. Finally, aircraft disembarkation has not
yet been properly addressed and is missing to complete the
picture of operational impacts of COVID-19 [3].
B. Focus and structure of document
We provide in our contribution an approach for aircraft
boarding and disembarkation considering a physical distance
between groups of passengers (e.g. families or couples). In
previously conducted research, it was already shown that the
consideration of passenger groups shorten boarding time. The
paper is structured as follows. After the introduction (Sec-
tion I), we briefly introduce a stochastic cellular automaton
approach, which is used for modeling the passenger move-
ments in the aircraft cabin (Section II). A transmission model
is implemented to evaluate the virus transmission risk during
passenger movements. In Section III, we motivate and intro-
duce a problem description to derive optimized sequences of
passenger groups during boarding. This approach is extended
in Section IV to allows weighting disembarkation time and
a transmission risk indicator in the objective function. This
sequence is then implemented in the cellular automaton model
to verify the results. Finally, we conclude our research with
a summary and an outlook (Section V).
II. MOD EL FOR PASSEN GER M OVE MEN T IN THE CA BIN
The individual movement behavior of passengers in the
aircraft cabin is modeled by a cellular automaton approach
[6], which covers short (e.g. avoid collisions, group behavior)
and long-range interactions (e.g. tactical wayfinding). This
cellular automaton model is based on an individual transition
matrix, which contains transition probabilities to move to
adjacent positions around the current passenger position [28].
A. Operational constraints and rules of movement
The implemented cellular automaton model considers oper-
ational conditions of aircraft and airlines (e.g. seat load factor,
conformance to the boarding procedure) as well as the non-
deterministic nature of the underlying passenger processes
(e.g. hand luggage storage) and was calibrated with data from
the field [14]. The cellular automaton for aircraft boarding
and disembarkation is based on a regular grid (Fig. 3), which
consists of equal cells with a size of 0.4 x 0.4 m, where a
cell can either be empty or contain exactly one passenger.
Passengers can only move one cell per timestep or must stop
if the cell in the direction of movement is occupied.
front door rear door
135729
27
2523
... ...
seat row
seat aisle
F
E
D
C
A
B
A,B,C,D,E,F seat columns
Figure 3: Grid-based aircraft model with 29 seat rows and 6 seats per
row (reference layout for single-aisle, narrow-body configurations).
Layout shows one door in use for disembarkation.
The boarding progress consists of a simple set of rules
for the passenger movement: (a) enter the aircraft at the
assigned door (based on the current boarding scenario), (b)
move forward from cell to cell along the aisle until reaching
the assigned seat row, and (c) store the luggage (aisle is
blocked for other passengers) and take the seat. The storage
time for the hand luggage depends on the individual number
of hand luggage items. The seating process depends on the
constellation of already used seats in the corresponding row.
The agents are sequenced concerning the active boarding
sequence. From this sequence, a given percentage of agents
are taken out of the sequence (non-conforming behavior)
and inserted into a position, which contradicts the current
sequence (e.g. inserted into a different boarding block).
For the disembarkation the movement rules are: (a) all
passengers are seated in the aircraft according to an initial
seat configuration, (b) passengers could enter the aisle if the
seats at their corresponding row are free and the aisle is
not blocked by other passengers, (c) if passengers enter the
aisle, they take their hand luggage items out of the overhead
compartment and block the corresponding aisle cell, (d) if
all hand luggage items are taken, passengers move in the
direction of the assigned aircraft door by entering empty aisle
cells in front of them.
The maximum, free walking speed in the aisle is 0.8 m/s
[15], so a simulation timestep is 0.5 s. In each simulation
step, the list of passengers to be updated is randomly shuffled
to emulate a parallel update behavior for the discrete time
dynamics (random-sequential update) [28,29]. Each boarding
and disembarkation scenario is simulated 125,000 times, to
achieve statistically relevant results defined by the average
boarding/disembarkation time. Further details regarding the
general model, parameter setups, and the simulation environ-
ment are provided in [6].
For the COVID-19 scenarios, an additional assumption is
that a cell is blocked if entering or moving in the aisle
would violate the separation distance between passengers or
groups of passengers (e.g., families or couples). In the context
of physical separation, the International Aviation Transport
Association (IATA) requires a minimum separation distance
of 1 m [30] and the Federal Aviation Administration (FAA)
requires a minimum separation distance of 6 feet (2 meters)
[31]. Considering the cellular automaton model with its
regular grid structure (cell spacing of 0.4 m) and to maintain
comparability with our previous results [1,2,4], the minimum
physical spacing was set at 1.6 m (4 cells). At this point, we
assume that passengers are informed that a distance of 1.6 m
corresponds to the distance between 2 rows of seats, which
provides sufficient visual orientation for the passengers.
B. Transmission model
The basic cellular automaton developed for stochastic pas-
senger movements is extended to include an approach for
assessing the risk of virus transmission during boarding and
disembarkation. Transmission risk can be defined by two main
factors: Proximity to the index case and duration of contact
time. A simplified approach is to count both the individual
interactions (passengers in adjacent cells) and the duration
of these contacts in the aisle and during the seating process.
However, counting individual contacts provides only an initial
indication of possible ways of infection. Our approach is
based on a transmission model [32], which defines the spread
of SARS-CoV2 coronavirus as a function of distance, using
different distance measures [33]. Here, the probability of a
person nbeing infected by a person mis described by (1).
Pn= 1 −exp −θX
mX
t
SRm,t inm,t tnm,t!(1)
defined by:
PnProbability of person nto receive an infectious dose.
Not “infection probability”, which depends highly on
the immune response of the affected person.
θCalibration factor for the specific disease.
SRm,t Shedding rate, the amount of virus the person m
spreads during the timestep t.
inm,t Intensity of the contact between nand mduring the
timestep t, which corresponds to their distance.
tnm,t Time person ninteracts with person mat timestep t.
Considering this idea, we define the shedding rate SR as
a normalized bell-shaped function (2) with z∈(x, y)for
both longitudinal and lateral dimensions, respectively. The
parameters are a(scaling factor), b(slope of leading and
falling edge), and c(offset) to determine curve shape.
SRxy =Y
z∈(x,y) 1 + |z−cz|
az
2bz!−1
(2)
SR was calibrated in a prior study [4] based on the
transmission events of an actual flight [34]. We have applied
the corresponding parameter setting with ax= 0.6,bx= 2.5,
cx= 0.25,ay= 0.65,by= 2.7, and cy= 0. This causes
the footprint in the y-direction (lateral to the direction of
motion) to be smaller than in the x-direction (in the direction
of motion). When passengers reach their seat row and start to
store the hand luggage or enter the seat row, the direction of
movement is changed by 90◦, heading to the aircraft window.
f(x; 0.6, 2.5,
0.25)
f(y; 0.65, 2.7,
0.0)
-2 2
-1 10
x, y (m)
probability
1.0
0.8
0.6
0.4
0.2
0.0
0.8
0.6
0.4
0.2
probability
-1 1
0
y (m)
-1
1
0
x (m)
0.40
0.20
0.80
0.04
Figure 4: Transmission probability for longitudinal xand lateral y
components (left), and (x,y) probability field (right) [1].
Finally, the individual probability for virus transmission Pn
corresponds to Θ, the specific intensity per timestep (3).
Pn= Θ SRxy α(3)
In accordance with [4], Θis set to 1
20 , which means a
passenger reaches a probability of Pn= 1 after standing 20 s
in closest distance in front of an infected passenger (SRxy =
1). The parameter α∈ {1,2}is 1 and changed to 2 when
the passenger stores the luggage or enters the seat row. This
doubled shedding rate reflects the higher physical activities
within a short distance to surrounding passengers.
III. PASSENGER BOARDING
A. Nominal case - single passengers
We introduce a baseline setup to depict the results for the
evaluation of transmission risks [6]. Table Ishows the com-
prehensive evaluation of transmissions around one infected
passenger, which is randomly seated in the aircraft cabin.
Two different scenarios are evaluated against the reference
implementation (R) of the boarding strategies: (A) applying a
minimum physical distance between two passengers of 1.6 m,
and (B) additionally to the physical distance, the number
of hand luggage items is reduced by 50% (implemented by
reducing the storing time by 50%). Scenarios A and B are
additionally extended by the use of two aircraft doors (one
in the front and at the rear) during boarding, scenarios A2
and B2. The transmission risk and the boarding time are
used as evaluation criteria [4]. The analysis points out that, in
particular, the back-to-front sequence (2 blocks: front block
with rows 1-15, rear block with rows 16-29) exhibits lower
values for the transmission probability than the optimized
block sequence (using 6 blocks of aggregated seat rows)
(see [6]). When passengers board (block-wise) from the back
to the front, the chance to pass an infected person is reduced to
a minimum, which is confirmed by the reduced transmission
probability exhibited in Table I. This effect is also a root
cause of the low transmission risks of the outside-in, reverse
pyramid, and individual boarding sequence.
TABLE I. Transmission risk assessment assuming a SARS-CoV2
passenger in the cabin, graded by four categories: random, block-
based, row- and individual-based, and disembarkation.
boarding strategies: R A B A2 B2 R A B A2 B2
boarding sequence Transmission risk (a.u.) Boarding time (%)
Random 5.9 1.6 1.1 1.4 1.0 100 198 154 133 103
Back-to-front
(2 blocks) 5.6 1.4 1.0 1.2 0.8 96 220 169 153 116
Optimized block
(6 blocks) 6.5 2.3 1.5 1.5 1.0 95 279 210 166 125
Outside-in 3.5 0.4 0.2 0.3 0.1 80 161 116 107 77
Reverse pyramid 3.0 0.2 0.1 0.2 0.1 75 185 128 119 82
Individual 2.0 0.2 0.1 0.2 0.1 66 114 104 103 74
Disembarkation 10.0 9.7 7.8 7.6 6.0 55 97 68 52 36
The use of two aircraft doors for boarding will provide
an appropriate solution for a reduced transmission risk inside
and outside the cabin, if near apron stands could be used
and passengers could walk from the terminal to the aircraft.
This kind of walk boarding also prevents passengers from
standing in the badly ventilated jetway during the boarding.
Disembarkation is difficult to control by specific procedures
given that passengers demonstrated little discipline and high
eagerness to leave the aircraft. More attention should be paid
to this process and consideration should also be given to
procedural or technical solutions to provide passengers better
guidance and control.
B. Passenger Groups
We consider passenger groups as an important factor to
derive an appropriate seat allocation and boarding sequence.
The main idea behind the group approach is that members
of one group are allowed to be close to each other, as they
are already in close contact with each other before boarding,
while different groups should be separated as far apart as
necessary.
We develop a mathematical model to determine an optimal
strategy for assigning seats in the cabin under the objective
to minimize the virus transmission risk. The idea to create an
appropriate seat allocation for a pandemic situation includes
three assumptions. The first one is that an airline could assign
just a percentage of the available seats (e.g. 50%) to reduce
the virus transmissions in the cabin and this strategy will
be the primary solution to face the pandemic situation. The
next assumption is about minimizing passenger contacts or
maximizing the distances between passengers in the cabin
and guaranteeing at the same time that the confined space
inside the aircraft is used efficiently.
Although complex boarding sequences, such as outside-in,
reverse pyramid, and individual lead to better boarding times,
there will be an issue. The boarding process is driven by the
willingness of passengers to follow the proposed strategy. We
will assume a group of four members (e.g. a family) to be
seated. If one of these boarding strategies is applied, they will
have just two options. The first one is seating near each other,
therefore they have to split during the boarding. The next
option is remaining as one group in the boarding sequence and
as a result, they have to seat in different rows. Both options are
inconvenient for group members (families). Here we propose
to look at the group members as a community since they
were already in close contact before boarding. The strategy
that is used in Fig. 5depicts a general solution, but it could
be improved considering groups. Without loss of generality,
we could suppose that the transmission rate for the members
of each group is zero, which will result in better use of space
and create a new pattern.
front door
135729
27
2523
... ...
seat row
occupied seats aisle
empty seats
Figure 5: Fifty percent of the seats will be allocated to passengers
during the pandemic situation according to a pattern with maximum
physical separation.
The introduced concept of a shedding rate of infected pas-
sengers will be used here as well. If an infected passenger was
assigned to different columns, the several shedding rates must
be counted based on the location of the adjacent locations.
Taking Fig. 6as an example, when a passenger seated in row
i= 21 and column C (aisle), we compute the shedding rate
for the passenger from other groups that seat in the same row
(i= 21 at column A (window), B (middle), and D (aisle))
and previous row i−1 = 20 (column B (middle), C (aisle),
and D (aisle)).
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
F
B
E
C
D
1
1
1
1
1
1
3
3
3
3
3
3
3
3
A
1
2
2
2
2
2
2
1
5
5
4
4
4
6
6
4
Figure 6: Types of passenger interactions (orange) in the aircraft
cabin around the infected passengers (red) considering different seat
positions: besides (1 and 4), in front (2), diagonally in front (3), and
across the aisle (5 and 6).
Based on the assumptions of the problem description, we
set up an optimization model [1]. We tried to solve the
mathematical model for a medium-size problem (e.g. 10
groups, 10 seat rows) but the run time increased significantly
and the optimization software (GAMS with CPLEX solver)
could not find an optimal solution in a reasonable time (10
hours). Therefore, we designed a Genetic Algorithm (GA)
for the real-sized problem. The problems run on a computer
with AMD Ryzen 7, 3700U, 2.30GHz CPU, 16 GB RAM, and
Matlab 2013 software is used for running the GA. We choose
six scenarios for the optimization and used the optimal seat
allocation in the passenger boarding simulation to derive an
appropriate boarding sequence with a low transmission risk.
The corresponding solutions for the three scenarios with a
50% seat load (87 passengers) are illustrated in Fig. 7. The
values of the objective function (O.F.) of these three scenarios
indicate that our approach (scenario 3) for an improved seat
allocation should also result in a significant reduction of the
transmission risk: a reduction of 91% of O.F. value compared
to scenario 1 (seats are assigned randomly to passengers with
a maximum distance), and a reduction of 85% compared to
scenario 2 (seats are assigned randomly, group members in
the same zone). The implementation of the mixed-integer
Scenario 1:
Pass. No.= 87
O.F.= 147.320
Scenario 2:
Pass. No.= 87
O.F.= 90.533
Scenario 3:
Pass. No.= 87
O.F.= 13.335
1
17
23 4 567 8
9 10 11 12 13 14 15 16
18 19 20 21 22
23 24 25
26 27 28
29
31
30
Scenarios 1, 2, and 3
Figure 7: Seat allocation in the aircraft cabin considering 87
passengers (31 groups) assigned to different groups using regular
pattern (scenario 1 and 2) and improved group-based seat allocation
(scenario 3).
linear programming model and the genetic algorithm leads
to an improved allocation for the passengers to be seated
in the aircraft cabin. This seat allocation is used as input
for the passenger boarding model, which was extended by
a transmission module to evaluate transmission risk during
aircraft boarding, to derive an optimum sequence to board
the passengers.
Analyses in the context of appropriate boarding sequence
accompanied by the introduction of infrastructural changes
showed that an improved sequence comprises a mix of
boarding per seat (from window to aisle) and per seat row
(from the rear to the front) [35]. First and foremost, per-seat
boarding (window seats first) is the most important rule to
ensure seating without additional interaction in the seat rows.
Starting with an outer seat in the last row, the number of
group members and the necessary physical distance between
passengers (1.6 m) defines the subsequently following seat
row, which could be used in parallel (e.g. 6 passengers with
seat row 29 will block the aisle until seat row 27 (waiting),
the physical distance requires to block row 26 and 25, the next
group must have seats in front of row 25). This process of
seat and row selection is repeated until the front of the aircraft
is reached and is repeated until all passengers are seated. We
further assume that the passengers in each group will organize
themselves appropriately to minimize local interactions.
If the sequencing algorithm is applied to the optimized
seat allocation from scenario 3, the passenger groups are
boarded in five segments. Inside each group, the distance
between passengers is not restricted but between groups, it is
constrained by 1.6 m (last member of the first group and the
first member of the following group). The first segment starts
with group 31 and the last segment with group 14 (see Fig. 8).
As an example, the passengers inside group 31 (yellow) are
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
123 4 567817
910 11 12 13 14 15 16 18 19 20 21 22
23 24 25 26 27 28 29 31
30
Figure 8: Optimized boarding of 31 groups considering a physical
distance of 1.6 m between passengers of different groups.
organized by the following sequence of seats, which results
in a minimum of individual seat and row interactions: 29A,
29B, 28A, 28B, 27A, 27B, and 27C. Considering distances
between groups, the best candidate will be group 27 (red) with
seats 23F, 23E, 22F, 22E, and 22D. This sequence allows both
groups to start the seating process in parallel, without waiting
time due to a too small distance between the seat rows.
In the three scenarios, 87 passengers are boarded with
different sequences (see Fig. 7): individual passengers in
a regular pattern (scenario 1), groups in a regular pattern
(scenario 2), and groups in an optimized seat allocation
(scenario 3). Scenario 1 is used as a reference case to evaluate
the performance (boarding time) and the transmission risk of
scenarios 2 and 3 (Table II).
TABLE II. Evaluation of average boarding times and transmission
risk during boarding assuming a randomly selected contagious
passenger at 50% seat occupancy (87 passengers).
Scenario Sequence Time Transmission
(%) risk (a.u.)
1 random 100.0 0.58
best sequence 45.2 0.00
2 groups, random sequence 68.0 0.62
groups, best sequence 51.9 0.20
3 groups, optimal allocation
random sequence 69.0 0.57
groups, optimal allocation
best sequence 41.1 0.09
Therefore, a passenger sequence is established for both
random and individual boarding sequences (optimized). The
boarding time for the random sequence is set to 100%, as
reference. As shown in Table II, the implementation of the
individual sequence will reduce the boarding time to 45.2%
at a minimum of transmission risk. The consideration of
groups (scenarios 2 and 3) using the random sequence already
reduces the boarding time by about a third at a comparable
level of transmission risk. If the optimized seat allocation is
used together with the individual groups the boarding time
could be further reduced to 41.1% at a low transmission risk
of 0.09 new infected passengers on average (85% reduction).
IV. PASSENGER DISEMBARKATION
While boarding can be controlled to some degree, pas-
senger disembarkation takes place in a less controllable
environment. In a first and simplified approach, passenger
groups were disembarked in batches of passenger groups. In
this case, groups within a batch are allowed to enter the aisle
in parallel, while ensuring that the minimum distance between
groups is maintained at all times. After a group has been
allowed to disembark, all members of the group enter the
aisle, take their hand luggage, and wait until the group in
front of them begins disembarking. Subsequent groups may
begin when the last group in the current batch has passed
their row. In future operational scenarios, passenger groups
could be notified directly via personal devices or active lights
on their seats to begin disembarking.
To show how the time for disembarkation changes, a sensi-
tivity analysis is performed. To do this, a group of passengers
from the example case above (Fig. 8) is randomly selected
and added to the current disembarkation batch with a certain
probability. This probability increases in 10% increments
from 0% (one group per batch) to 100% (all groups in one
batch). In addition, three levels for hand luggage amount are
considered: 100% (standard), 50% (reduced), 0% (no items).
The scenario with only one group per batch and no hand
luggage is chosen as the reference case. The result of this
analysis is shown in Fig. 9using the average disembarkation
times and the associated standard deviations as measurements.
Figure 9: Random group sequence for disembarkation: (a) given
probability that the following group is part of the same disembarka-
tion batch, and (b) three level of hand luggage quantity (100% -
standard, 50% - reduced, 0% - no items).
The analysis shows that disembarkation will benefit from
the superior organization of groups into batches up to a
certain point. As a result, average disembarkation times
decrease until a minimum is reached at approximately 80%.
In general, the standard deviations exhibit also a minimum
in that region (70% - 90%) but show both a significant
increasing and decreasing behavior. The reason is that if all
passenger groups are in one batch, the disembarkation is a
front-to-back process, where groups in the front are leaving
the aircraft first. This a somehow a much more stable process
than a batch-organized prioritization but also results in higher
disembarkation times.
The obtained minimum times indicate a significant poten-
tial to reduce the disembarkation time by about 40%. To
provide a more improved disembarkation process, a manual
assignment of groups will be derived from the introduced
seat allocation. Each batch of passenger groups is generated
starting with the last occupied seat row. The group of this row
will be placed into the aisle, assuming the individual personal
space (one cell per passenger). Considering the distance of
1.6 m per group, the nearest group to that location (down-
stream) is added to the current batch. This process is finished
when the first seat row is reached. Fig. 10 exhibits this batch
assignment process, where 6 batches are created for the given
example. This algorithm-based batch sequence results for
the three hand luggage scenarios in reduced disembarkation
times by 7%, 27%, and 35% for the zero luggage, 50%,
and 100% hand luggage items accordingly (referring to the
corresponding minimum times presented at Fig. 9).
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
Batch 1
Batch 2
Batch 3
Batch 4
Batch 5
Batch 6
123 4 567817
910 11 12 13 14 15 16 18 19 20 21 22
23 24 25 26 27 28 29 31
30
11
11
6
13
13
16
16
17
17
24
24
2424
27
27
2727
27
31
31 31
3131
31
31
Figure 10: General, batch-oriented disembarkation considering a
minimum distance of 1.6 m between passenger groups. The current
aisle status corresponds to the first batch of passenger groups ready
for disembarkation. The following batches are depicted below.
These results lead to the question of how this process can be
further improved. In the following section, we develop a two-
objective mathematical model, which includes an optimized
disembarkation strategy incorporating passenger groups and
physical distance requirements from the current COVID-19
situation. In the development, we are not considering hand
luggage items assuming that the average pick-up process will
not influence the sequence generation.
The concept of the shedding rates is now used for the
disembarkation process. Here, the corresponding shedding
rate depends on the position of the infected passenger in the
aisle and the positions of the passengers of other groups that
will leave their seats afterward. We calculate the transmission
risk function for all passengers based on the locations of other
passengers. Three types of passenger interference are imple-
mented in our approach. The first type is defined based on the
concept of physical distancing. We suppose a close distance
between members of each group in the aisle. According to the
COVID-19 regulations a physical distance of 1.6 m between
groups is implemented (see Fig. 11, part 1).
The second type of interference considers the concept
of shedding rates. If an infected passenger leaves the seat,
corresponding shedding rates must be calculated based on
the location of the other group members who will leave the
aircraft after that passenger. Fig. 11 (part 2) demonstrates the
interference generated in the two middle rows (i= 15 and
i= 16) when the first member of group 30 (coded orange) is
walking in the aisle at period h, where the shedding rates for
the passengers from other groups that seat in the related row
(i= 15 at column D (aisle), E (middle), and F (window)) are
calculated. Similarly, the third member of this group could
generate different types of interactions for passengers who
seat in the next row at the same time (i= 16 at column C
(aisle), B (middle), and A (window)).
As a result, if the passengers of the first group leave the
aircraft earlier, the transmission risk which is the sum of
shedding rates of all passengers could be minimized. On the
other hand, this strategy leads to longer disembarkation time.
Therefore, we have proposed a two-objective mathematical
model to handle these two conflicting objectives. The third
type of interference is the scheduling process. Taking Fig. 11
(part 3) as an example, if the first member of group 31 arrives
in the nineteenth row (i= 19) in four seconds then, the
passenger who seated in this row and in column F (window)
could not leave the seat at that moment. The following
assumptions are based on the different types of interference
and considering general disembarkation procedures.
•Group members leave their seats at the same time.
•Members of two different groups sitting in a close zone
have a certain time interval (e.g. three timesteps) to leave
their seats. This distance increases with the number of
group members.
•The length of the cabin aisle is 23.2 m.
•The transmission rates are calculated for all passengers.
•Passengers can only leave seats and enter the aisle if this
does not result in interference with other passengers.
A. Mathematical model
The following sets, parameters, and decision variables are
used in the optimization model.
Notation Definition
Sets and Indexes
iIndex set of row i∈ {1,2,...,I}
kIndex set of passenger group k∈ {1,2,...,K}
jIndex set of seat column j∈ {1,2,...,J }
hIndex set of time period h∈ {1,2,...,H}
rIndex set of interaction type r∈ {1,2,...,R}
Parameters
TkGroup number k
SR0
jRelated shedding rate for interaction jor r,
considering 6 different types based on the number
of columns
λij Disembarkation time (in period unit) required
for a passenger, seats in row iand column j
Yij Binary parameter, equals kif a passenger from
group k, seated in in row iand column j;
equals zero otherwise
Mkk0Binary parameter, equals one if the members of
group number kand k0could not leave their seats at
the same time (because they will block each other such
as groups 30 and 31), equals zero otherwise
w1Coefficient weight of the first objective function
w1∈[0,1]
w2Coefficient weight of the second objective function
which is equals to 1−w1
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
31
31
31
31 31
31
31
31
30
30
30
30
30
30
30
23
23
23
23
23
19
19
19
19
13
13
13
11
11
3
3
2
2
A
C
F
E
D
B
Part 1 Part 2 Part 3
1.6 m distance
Figure 11: Different types of interference between passengers in the aircraft: physical distance between groups (part 1), close interactions
covered by particular shedding rates (part 2, cf. Fig. 6), far-reaching interactions (passenger in row 19 have to wait for group 31 (part 3)).
Notation Definition
Decision Variables
pkh Number of members of group kthat leave their
seats at time period h
xijkh Binary variable, equals one if a passenger from
group kwho is seated in a seat in row iand
and column jleaves its seat at the time period h
(activation time); equals zero otherwise
qijkh Period of time from the moment when a passenger
from group kin row i, column jleaves the cabin
niji0Period of time at the moment when a passenger
who has seated in row i, column jreaches row i0
ukk0hh0Binary variable, equals one if group kand k0leave
their seats at time period hand h0, respectively
z1First objective function: disembarkation time
z2Second objective function: transmission risk
TZ Total objective function which is calculated based
on two conflicted objectives z
The proposed new multi-objective minimization model for
the problem is introduced as follows.
TZ =w1z1−z1∗
z1∗+w2z2−z2∗
z2∗(4)
Disembarkation Time:
z1=Max
i∈I,j∈J ,k∈K,h∈H qij kh (5)
Transmission Risk Indicator:
z2=
I
X
i=1
J
X
j=1
K
X
k=1
k=Yij
H
X
h=1
I
X
i0=i
J
X
j0=1
K
X
k0=1
k0=Yi0j0
H
X
h0=1
h0<h−ni0j0i
SR
0
jukk0hh0
(6)
The L−1metric method is used to solve the multi-objective
decision problem. Therefore, we run the model three times.
The first time, we minimize the average disembarkation time,
with (5) as an objective function, considering constraints (7)-
(16). Thus, the first solution z∗
1could be obtained. Similarly,
we minimize the problem concerning minimum transmission
risk (6), obtaining the second solution z∗
2. In the third step, we
put these two solutions into the general problem, minimize
equation (4) under constraints (5)-(16). Here we consider two
implementation weights (w1,w2) for competing objectives
evaluated and set by decision-makers.
I
X
i=1
J
X
j=1 xijkh
Tk
+
I
X
i=1
J
X
j=1
h+Tk+3
X
h0=h
xijk0h0
Tk0≤1,
∀h, k, k0, Mkk0= 1 (7)
0.4
I
X
i=1
J
X
j=1
K
X
k=1
h
X
h0=1
h0>h−λij
xijkh0+
1.6
I
X
i0=1
J
X
j0=1
K
X
k0=1
h
X
h0=1
h0>h−λi0j0
xi0j0k0h0
Tk0≤23.2,∀h > 1(8)
I
X
i=1
J
X
j=1
xijkh =pkh ,∀k, h (9)
Constraint (7) guarantees that in each timestep two groups
sitting in a nearby zone cannot leave their seats at the same
time. The physical distances between the passengers in the
aisle are determined by constraint (8). A distance of 0.4 m
between members of a group and 1.6 m between groups in the
aisle is considered. Constraint (9) determines the number of
members of each group that leave their place in each period.
H
X
h=1
xijkh = 1 ,∀i, j, k =Yij (10)
Tkxijkh ≤
I
X
i0=1
J
X
j0=1
xi0j0kh ,∀i, j, k, h (11)
pkh +pk0h0−1≤ukk0hh0,∀k, k0, h, h0(12)
qijkh ≥h+λij + 500(xijk h −1) ,∀i, j, k, h (13)
qijkh ≤500xijk h ,∀i, j, k, h (14)
In addition, constraints (10)-(11) guarantee that all mem-
bers of each group leave their place in the same period.
To calculate the decision variables for the shedding rates of
two different groups, we define the constraint (12), which is
implemented in the transmission risk function. The constraints
(13)-(14) represent the disembarkation time of each passenger
in each group sitting in row iand column j, and the
corresponding decision variable takes the value of zero if the
seat was not occupied.
niji0=λij −i0+h , ∀i, j, k =Yij , h, xij kh = 1 (15)
xi0j0k0h0≤|h0−niji0|,
∀i, i0, j, j0, h0, k =Yij , k0=Yi0j0, k 6=k0>0(16)
xijkh , ukk0hh0∈ {0,1}, qijkh, nij i0, pkh, z1, z2≥0,
∀i, i0, j, k, k0, h, h0(17)
The third type of interference is formulated in constraints
(15)-(16). For each passenger, the arrival time (i.e., niji0)
in the lower row i0is determined by (15). Consequently,
passengers in this row are not allowed to leave their seats
at this time (16). Constraint (17) represents the requirements
on the decision variables.
The proposed mathematical model as a scheduling problem
is a type of NP-hard. For our real-size problem, we consider
29 rows, 6 columns, 3 interaction types, 31 groups, and
500 periods of time (or seconds). To solve this problem,
we implement a meta-heuristic algorithm and used a genetic
algorithm to solve the disembarkation problem (cf. [36,37]).
B. Optimization results
We already use optimized seat layouts [1] and have
developed a new mathematical approach with two opposing
objective functions. The first objective function gives the
overall disembarkation time and the second objective function
targets the transmission risk indicator. In the following, we
present the solutions for four example scenarios.
1) Scenario A1: Minimum disembarkation time with 50%
occupancy, 87 passengers on the aircraft, divided into 31
groups. Therefore the coefficient weight of disembarkation
time equals one (w1= 1) and the coefficient weight of
the transmission risk is supposed to be zero (w2= 0).
The optimized disembarkation progress is shown in Fig. 12.
The disembarkation time for the first scenario is 139 s and
the transmission risk indicator (second objective function) is
calculated in the value of 480.
Scenario A1: 50% occupancy
Disembarkation time = 139 seconds
110 20 30 40 50 60 6555453525155
steps (0.5 s per step)
Figure 12: Disembarkation progress for scenario A1, 87 passengers.
2) Scenario A2: Minimum transmission risk with 50%
occupancy, 87 passengers (31 groups), w1= 0, and w2= 1.
Compared to scenario A1, the transmission risk value de-
creases from 480 to 30 but the disembarkation time increases
by 50% (209 s). Scenario A2 was created only for a general
comparison of the fitness functions. In the following, only the
disembarkation time shall be minimized.
Scenario A2: 50% occupancy
Disembarkation time = 209 seconds
Figure 13: Disembarkation progress for scenario A2, 87 passengers.
3) Scenario B: Minimum disembarkation time with 66%
occupancy, 116 passengers on the aircraft, divided into 38
groups. Therefore, the coefficient of the first objective func-
tion is supposed to be one. Fig. 14 shows the disembarkation
progress with a disembarkation time of 188 s.
d
Scenario B: 66% occupancy
Disembarkation time = 188 seconds
d
110 20 30 40 50 60 6555453525155
steps (0.5 s per step)
Figure 14: Disembarkation progress for scenario B, 116 passengers.
4) Scenario C: Minimum disembarkation time with 66%
occupancy, 174 passengers on the aircraft, divided into 62
groups. Fig. 15 shows the corresponding seat allocation and
indicates a disembarkation time of 299 s.
Scenario C: 100% occupancy
Disembarkation time = 299 seconds
110 20 30 40 50 60 6555453525155
steps (0.5 s per step)
Figure 15: Disembarkation progress for scenario C, 174 passengers.
C. Evaluation of disembarkation process
We must emphasize at this point that we assume each pas-
senger leaves the cabin immediately after entering the aisle.
As shown in the exemplary solutions presented (Figs. 12-
15), we assume a forward ordered-sequential update of the
passenger positions [38]. Thus, the update of positions starts
with the passenger closest to the exit (front door), and
the positions of subsequent passengers along the aisle are
updated next. This approach results in the passenger in the
position ahead of the current passenger (in the direction
of movement) always moving first, and the exit time is
too optimistic. Against this background, we implemented
the optimized group sequences in the calibrated stochastic
simulation environment to check the disembarkation times.
Each group is activated to disembark when the group im-
mediately ahead of it in the disembarkation sequence passes
its seat row. This minimizes the potential time/space buffers
between group calls as would be required in later operational
implementation. Group members enter the aisle only when
the physical spacing between groups is assured. Table III
summarizes the simulation results.
Disembarkation time (s)
Update behavior
Scenario Seat load Passengers Forward- Random Reference
(%) ordered
A1 50 87 139 163 286
B 66 116 188 219 377
C 100 174 299 331 571
TABLE III. Disembarkation of passenger groups considering three
scenarios with different seat load factors.
As expected, the result of implementing the forward
ordered-sequential update underestimates the disembarkation
times by about 30 s. To show the potential for improving
the disembarkation time, we use reference cases for each
scenario. These reference cases consider the same number
of passengers, no groups, no hand luggage items, mandatory
physical distance, and realistic random sequential updating
behavior. Concerning the reference cases, our optimization
strategy (focusing on a fast disembarkation process) acceler-
ated the process by about 40%.
V. CONCLUSION AND OUTLOOK
In the aircraft cabin, passengers must share a confined en-
vironment with other passengers during boarding, flight, and
disembarkation, which poses a risk for virus transmission and
requires risk-appropriate mitigation strategies. Spacing be-
tween passenger groups during boarding and disembarkation
reduces the risk of transmission, and optimized sequencing of
passenger groups helps to significantly reduce boarding and
disembarkation time. We considered passenger groups to be
an important factor in overall operational efficiency. The basic
idea of our concept is that the members of a group should
not be separated, since they were already traveling as a group
before entering the aircraft. However, to comply with COVID-
19 regulations, different passenger groups should be separated
spatially. For the particular challenge of disembarkation, we
assume that passenger groups will be informed directly when
they are allowed to leave for disembarkation. Today, cabin
lighting could be used for this information process, but
in a future digitally connected cabin, passengers could be
informed directly via their personal devices. These devices
could also be used to check the required distances between
passengers [5]. The implementation of optimized group se-
quencing has the potential to significantly reduce boarding
and disembarkation times, taking into account COVID-19
constraints. In future studies, we will also consider the dis-
tribution of carry-on baggage in the cabin. Finally, we intend
to further investigate the potential of technical systems that
can monitor, evaluate, and, if necessary, regulate passenger
boarding and disembarkation.
REFERENCES
[1] M. Schultz and M. Soolaki, “Analytical approach to solve the prob-
lem of aircraft passenger boarding during the coronavirus pandemic,”
Transp. Res. Part C: Emerg. Technol., vol. 124, p. 102931, 2021.
[2] M. Schultz, J. Evler, E. Asadi, H. Preis, H. Fricke, and C.-L. Wu,
“Future aircraft turnaround operations considering post-pandemic re-
quirements,” J. Air Transp. Manag., vol. 89, p. 101886, 2020.
[3] C.-Z. Xie, T.-Q. Tang, P.-C. Hu, and H.-J. Huang, “A civil aircraft
passenger deplaning model considering patients with severe acute
airborne disease,” J. Transp. Saf. Secur., pp. 1–22, 2021.
[4] M. Schultz and J. Fuchte, “Evaluation of Aircraft Boarding Scenarios
Considering Reduced Transmissions Risks,” Sustainability, vol. 12,
no. 13, p. 5329, 2020.
[5] P. Schwarzbach et al., “Evaluation of Technology-Supported Distance
Measuring to Ensure Safe Aircraft Boarding during COVID-19 Pan-
demic,” Sustainability, vol. 12, no. 20, p. 8724, 2020.
[6] M. Schultz, “Implementation and application of a stochastic aircraft
boarding model,” Transp. Res. Part C Emerg. Technol., vol. 90, pp.
334–349, 2018.
[7] C. Delcea, L.-A. Cotfas, and R. Paun, “Agent-Based Evaluation of the
Airplane Boarding Strategies’ Efficiency and Sustainability,” Sustain-
ability, vol. 10, no. 6, p. 1879, 2018.
[8] F. Jaehn and S. Neumann, “Airplane boarding,” Eur. J. Oper. Res., vol.
244, no. 2, pp. 339–359, 2015.
[9] M. Schmidt, “A review of aircraft turnaround operations and simula-
tions,” Prog. Aerosp. Sci., vol. 92, pp. 25–38, 2017.
[10] J. H. Steffen and J. Hotchkiss, “Experimental test of airplane boarding
methods,” J. Air Trans. Manag., vol. 18, no. 1, pp. 64–67, 2012.
[11] A. Kierzkowski and T. Kisiel, “The Human Factor in the Passenger
Boarding Process at the Airport,” Procedia Eng., vol. 187, pp. 348–
355, 2017.
[12] S. M. V. Gwynne et al., “Small-scale trials on passenger microbe-
haviours during aircraft boarding and deplaning procedures,” J. Air
Transp. Manag., vol. 67, pp. 115–133, 2018.
[13] A. Miura and K. Nishinari, “A passenger distribution analysis model
for the perceived time of airplane boarding/deboarding, utilizing an ex-
Gaussian distribution,” J. Air Trans. Manag., vol. 59, pp. 44–49, 2017.
[14] M. Schultz, “Field trial measurements to validate a stochastic aircraft
boarding model,” Aerospace, vol. 5, no. 27, 2018.
[15] ——, “Fast aircraft turnaround enabled by reliable passenger boarding,”
Aerospace, vol. 5, no. 1, p. 8, 2018.
[16] X. Ren, X. Zhou, and X. Xu, “A new model of luggage storage time
while boarding an airplane: An experimental test,” J. Air Trans. Manag.,
vol. 84, p. 101761, 2020.
[17] M. Moussa¨
ıd, N. Perozo, S. Garnier, D. Helbing, and G. Theraulaz,
“The walking behaviour of pedestrian social groups and its impact on
crowd dynamics,” PLOS ONE, vol. 5, no. 4, pp. 1–7, 2010.
[18] M. Schultz, L. R¨
oßger, H. Fricke, and B. Schlag, “Group dynamic
behavior and psychometric profiles as substantial driver for pedestrian
dynamics,” in Pedestrian and Evacuation Dynamics 2012, 2013, pp.
1097–1111.
[19] F. Zanlungo, Z. Y¨
ucel, and T. Kanda, “Intrinsic group behaviour II: On
the dependence of triad spatial dynamics on social and personal fea-
tures; and on the effect of social interaction on small group dynamics,”
PLOS ONE, vol. 14, no. 12, pp. 1–28, 2019.
[20] M. Schultz and H. Fricke, “Managing passenger handling at airport
terminal,” in 9th ATM Seminar, Berlin, 2011.
[21] T.-Q. Tang, S.-P. Yang, H. Ou, L. Chen, and H.-J. Huang, “An aircraft
boarding model accounting for group behavior,” J. Air Transp. Manag.,
vol. 69, pp. 182–189, 2018.
[22] H. Zeineddine, “A dynamically optimized aircraft boarding strategy,”
J. Air Trans. Manag., vol. 58, pp. 144–151, 2017.
[23] A. Wald, M. Harmon, and D. Klabjan, “Structured deplaning via
simulation and optimization,” J. Air Transp. Manag., vol. 36, pp. 101–
109, 2014.
[24] S. Qiang, B. Jia, and Q. Huang, “Evaluation of Airplane Board-
ing/Deboarding Strategies: A Surrogate Experimental Test,” Symmetry,
vol. 9, no. 10, p. 222, 2017.
[25] S.-J. Qiang et al., “Symmetrical design of strategy-pairs for enplaning
and deplaning an airplane,” J. Air Transp. Manag., vol. 54, pp. 52–60,
2016.
[26] L.-A. Cotfas, C. Delcea, R. J. Milne, and M. Salari, “Evaluating
Classical Airplane Boarding Methods Considering COVID-19 Flying
Restrictions,” Symmetry, vol. 12, no. 7, p. 1087, 2020.
[27] M. Salari, R. J. Milne, C. Delcea, L. Kattan, and L.-A. Cotfas, “Social
distancing in airplane seat assignments,” J. Air Transp. Manag., vol. 89,
p. 101915, 2020.
[28] M. Schultz, “Stochastic transition model for pedestrian dynamics,” in
Pedestrian and Evacuation Dynamics 2012, 2013, pp. 971–985.
[29] ——, “Entwicklung eines individuenbasierten Modells zur Abbildung
des Bewegungsverhaltens von Passagieren im Flughafenterminal,” PhD
thesis, TU Dresden, Dresden, 2010.
[30] International Aviation Transport Association (IATA), “Guidance for
cabin operations during and post pandemic, edt. 4,” 2020.
[31] Federal Aviation Administration (FAA), “COVID-19: Updated interim
occupational health and safety guidance for air carriers and crews,”
2020.
[32] T. Smieszek, “A mechanistic model of infection: why duration and
intensity of contacts should be included in models of disease spread.”
Theor. Biol. Med. Model., vol. 6, no. 25, 2009.
[33] S. A. M¨
uller, M. Balmer, A. Neumann, and K. Nagel, “Mobility traces
and spreading of covid-19,” medRxiv, 2020.
[34] S. J. Olsen et al., “Transmission of the severe acute respiratory
syndrome on aircraft,” New England Journal of Medicine, vol. 349,
no. 25, pp. 2416–2422, 2003.
[35] M. Schultz, “Dynamic change of aircraft seat condition for fast board-
ing,” Transp. Res. Part C Emerg. Technol., vol. 85, pp. 131–147, 2017.
[36] M. H. L. van den Briel, J. R. Villalobos, G. L. Hogg, T. Lindemann,
and A. V. Mul´
e, “America West Airlines Develops Efficient Boarding
Strategies,” INFORMS J. Appl. Anal., vol. 35, pp. 191–201, 2005.
[37] M. Soolaki et al., “A new linear programming approach and genetic
algorithm for solving airline boarding problem,” Appl. Math. Model.,
vol. 36, pp. 4060–4072, 2012.
[38] A. Schadschneider, D. Chowdhury, and K. Nishinari, “Chapter two
- methods for the description of stochastic models,” in Stochastic
Transport in Complex Systems, A. Schadschneider, D. Chowdhury, and
K. Nishinari, Eds. Amsterdam: Elsevier, 2011, pp. 27–70.