Assessment of the eects of isolation, connement and hypoxia on space-ight
piloting performance for future space missions - The SIMSKILL experiment in
Miquel Bosch Brugueraa,∗
, Andreas Finka, Valerie Schrödera, Emilie Dessyb, Floris P. van den Bergc, Greig
Lawsond, Carole Dangoissec, Carmen Possnigc, Nadja Albertsenc, Nathalie Pattynb, Reinhold Ewalda
aInstitute of Space Systems, University of Stuttgart, Pfaenwaldring 29, 70569 Stuttgart, Germany
bRoyal Military Academy, Rue Hobbema 8, 1000 Brussels, Belgium
cEuropean Space Agency - ESTEC, Keplerlaan 1, 2201 AZ Noordwijk, Netherlands
dBritish Antarctic Survey, High Cross, Madingley Road, Cambridge CB3 0ET, United Kingdom
Interplanetary human missions to Mars and beyond will suppose a very demanding physical and psychological en-
vironment for future astronauts. Isolation, connement, hypoxia or hypercapnia in a reduced pressure atmosphere,
darkness and other factors are expected to endanger a mission’s success, directly inuencing human performance.
In order to study the eects of such environmental conditions on human beings, the SIMSKILL Experiment aims
to investigate how spacecraft piloting performance decays over time by deploying a Soyuz ight simulator on the
Antarctic research stations Halley VI and Concordia, which feature similar living conditions as those of a space
mission, leading to muscular atrophy, loss of cognitive capacities, and reduction of psycho-motor skills. This paper
oers an up-to-date analysis on the recorded data from the scientic campaigns in Antarctica, compared to those
of the control group subjects in Stuttgart. An overall total of 69 subjects and more than one thousand approach and
docking ights to the ISS performed in the simulator have been analysed using mathematical models. The results
obtained from this analysis show how the inuence of isolation, connement and hypoxia in Antarctica is crucial to
understand how dierences in performance appear between subjects. A thorough assessment of collective trends is
presented, by showing how reaction times, error events, visual perception, among others, are essential parameters
to understand a pilot’s skill evolution and propose optimal training and maintenance of acquired skills should be
dened in future space missions.
Keywords: Soyuz, Antarctica, Human Performance, Flight Simulator, Hypoxia, SIMSKILL
?This document is part of the research project funded by the Fed-
eral Ministry for Economic Aairs and Energy (BMWi) with Funding
??The presented article has been selected from the proceedings of
the 70th International Astronautical Congress, Washington D.C., 21-
25 October 2019.
Abbreviations: BAS, British Antarctic Survey; DLR, Deutsches
Zentrum für Luft- und Raumfahrt; ESA, European Space Agency;
FA, Final Approach; FF, Frequent Flyer Group; IF, Infrequent Flyer
Group; IPEV, Institut Polaire Français, Paul-Émile Victor; ISS, Inter-
national Space Station; NASA, National Aeronautics and Space Ad-
ministration; PNRA, Programma Nazionale die Ricerche in Antar-
tide; RF, Regular Flight; RMS, Root Mean Square
Email address: email@example.com
(Miquel Bosch Bruguera)
The investigation and assessment of astronaut skills
has been a key focus-point since the very beginning of
human space-ight. The search for an optimal train-
ing methodology with the aim of maximizing human
performance in space has become an essential aspect
for the success of crewed space missions. The method-
ologies to analyse and improve astronaut performance
have been constantly upgraded while new technologies
appeared. Consequently, the radical improvement of
computing performance has allowed, not only to dras-
tically improve the reliability and performance of rock-
ets and spacecraft, but also to enhance in the develop-
Preprint submitted to Acta Astronautica August 11, 2020
ment of training and evaluation tools for the crew, and
accordingly, improve the astronauts’ performance.
The analysis of piloting performance has undergone
a radical change with the availability of more real-
istic ight simulators and more sophisticated statisti-
cal models that require higher computational perfor-
mance. Statistical analyses by usage of linear models
have been proposed by Salnitski, Dudukin, Johannes et
al.  by assessing the relation between ight relia-
bility and the so-called "psychophysiological cost", in
order to understand which are the human factors that
dene a pilot’s performance. The ndings by Salnit-
ski and Johannes state that an intermission of about 30
days without training deteriorate ight reliability sig-
nicantly, although this can be recovered quickly with
latter training sessions. The assessment of piloting per-
formance is expected to play even a bigger role in the
future, where longer space missions might require long
training breaks among the crew.
The goal of space agencies and private companies
to perform interplanetary human missions in the up-
coming decades, principally within the Artemis pro-
gramme and ESA’s Lunar roadmap , requires a mul-
titude of preliminary research and development pro-
cesses, in order to enable the proper understanding
of the challenges presented by such missions and en-
sure their success. In the case of long duration human
space-ight, the existence of critical psychological and
physiological factors could put in risk the normal de-
velopment of a deep space mission. The environmen-
tal conditions in which astronauts will be exposed will
be partly unknown and require new approaches when
dening the training procedures, as explained by Ngo-
Anh and Rossiter . As a consequence, space agencies
and other scientic institutions have been performing
research on both real space missions in the ISS and pre-
vious orbital laboratories/stations (see Johannes et al.
), in ground-based, simulated mission environments
(the so-called analogue missions [5, 6, 7]) as well as in
certain scientic facilities such as Antarctic research
stations (Ngo-Anh et al. ) , natural caves, lava tubes
and underwater facilities such as the ESA CAVES and
NASA NEEMO programmes (Bessone et al. , von
Eherenfried  , Todd and Reagan ). Such envi-
ronments oer the possibility to develop and test new
frameworks and procedures based on the future Lunar,
Martian and other foreseen space ights. One of the
most singular platforms where to put in practice a real-
istic space simulation is Antarctica, which has captured
the interest of the space community for a long time, as
the living conditions are of full isolation, forcing the the
crew-members to be fully autonomous in every task.
Figure 1: Concordia Station. Credits: M. Bosch Bruguera
The European Space Agency is coordinating a series
of experiments in the Franco-Italian Concordia station
and the British Halley VI station in Antarctica (see Fig-
ure 1) in order to investigate the eects of isolation,
connement and hypoxia on astronauts during long-
term space missions. The implementation of space-
related experiments in such environments permits to
research of critical aspects of crewed missions and en-
hances the development of counteractive solutions for
current and future applications.
The dierent environmental conditions to which the
subjects are exposed, as well as their own personal con-
dition, are expected to play a big role on how their per-
formance evolves throughout the one year overwinter-
ing experiment, as explained by Sgobba and Sandal 
and Kanas . Hence, the eects of isolation and con-
nement in Concordia and Halley VI, plus hypoxia in
Concordia, should be stressors for a faster performance
decay. Table 1 shows the details of each research sta-
The SIMSKILL experiment (SIMulation SKILLs), is
a research project carried out by the University of
Stuttgart and the Royal Military Academy of Belgium,
and collaborating with the University of Rome "La
Sapienza", in the framework of the Antarctic research
opportunities oered by the ESA. SIMSKILL aims to
investigate how piloting performance evolves during
the overwintering campaigns in Antarctica, by mak-
ing use of an adapted and transportable Soyuz-TMA
space-ight docking simulator (see Figure 2). The main
research questions that this experiment targets are as
•Performance degradation: Understand and as-
sess in which measure does piloting performance
Research facility Environmental factors Crew origin
Halley VI Station Isolation, connement UK
Concordia Station Isolation, connement, hypoxia FRA, ITA + Research Medical Doctor (EU)
IRS Stuttgart Normal lifestyle DE
Table 1: Research platform details.
evolve and potentially decrease due to the eects
of isolation and connement, as well as hypoxia,
and extend the ndings to future long-term space
•Training Enhancement: Identify how previous
training aects a pilot’s performance and propose
eective solutions for the enhancement of astro-
Furthermore, this experiment expects to obtain a
better understanding on how subjects interact with the
cockpit and nd out which features should be present
in order to enhance mission safety and reliability.
Figure 2: Participant during a Soyuz simulator test at the Concordia
Station. Credits: M. Bosch Bruguera / I. Bruni
The subjects of the Halley VI and Concordia stations
were personnel already selected by the respective polar
institutes (BAS, IPEV and PNRA) meant to perform sci-
ence, maintenance and station keeping. The diversity
in tasks and qualication of the crew in all four Antarc-
tic campaigns was high. Each crew was provided by
at least two medical doctors, a cook, technical per-
sonnel responsible for station maintenance, and scien-
tic crew-members such as glaciologists, astronomers
and meteorologists. Consequently, age, qualication
and computer knowledge/gaming backgrounds were
implicitly very diverse. The selection of the control
group participants was tted to the crew distribution
in Antarctica, in order to oer comparable results. In
Figure 3 the age distribution of all experiment subjects
can be observed.
20 30 40 50 60 70
Figure 3: Age distribution of all 69 participants
2.2. Experiment Protocol
The experiment protocol proposed by SIMSKILL is
designed in a way that dierences between groups can
be recognized, depending on the living conditions the
participants experience during the campaign, thus un-
derstanding whether isolation and hypoxia, as well
training frequency, play a signicant role in the de-
velopment and reliability of piloting skills. Additional
parameters such as age, gender, background, etc. will
also be considered. As seen in Table 2, a total of 69 par-
ticipants completed the experiment successfully. Two
Campaign Subjects Female Male
Halley VI 2015 10 2 8
Halley VI 2016 11 3 8
Concordia 2018 12 1 11
Concordia 2019 12 2 10
IRS Stuttgart 2016 25 5 20
Total 69 (+1 drop-out) 13 (18,8%) 56 (81,2%)
Table 2: Subjects overview for each Antarctic campaign and control study.
overwintering campaigns in both Halley VI and Con-
cordia were performed. In addition, a control group in
Stuttgart was tested with the same protocol. A time-
line of the campaigns is shown in Figure A.1.
In order to understand which performance dier-
ences appear due to lack or reduced frequency of train-
ing during a mission, each crew was split in two groups:
frequent and infrequent yers (FF and IF, respectively).
The FF group would perform all the tests in the proto-
col every four weeks, whereas the IF group would still
perform basic experiments as frequently as the FF, but
wouldn’t be tested in the space-ight simulator only
in a twelve-weeks basis, as represented in Figure A.2.
The division of both groups was done after the train-
ing sessions, in order to balance the amount of pro-
cient and less-procient pilots in each group. The train-
ing phase took place during the Antarctic summer and
was performed by an instructor on site. It included 3
hours of instructed training and 6 hours of unsuper-
vised ight. After training completion, a check-ight
was performed by all the crew, which set the ocial
start of the experimental campaign.
Each session of the experiment was performed with
the same procedure, as shown in Figure 4. First, a set of
two ight scenarios was own (three at RF0, RF3, RF6
and RF9. See Figure A.2), which contained an instru-
mental ight, where the pilot could y with the support
of the docking instrumentation on the cockpit panels;
a visual ight, where the pilot could only rely on the
periscope view, and if doing a 3-ights session, an "ISS-
Pitch" scenario, where the target station was spinning
with a constant local angular rate of 2 deg/s. This last
scenario was meant to challenge the pilot under an un-
expected situation and required continuous correction
of the approach path. After the ight tests, two well-
known tests were performed, in order to assess both
the psychomotoric and the cognitive status of the sub-
ject. The Vienna Test System test battery "MLS - Motor
Performance Series" was used to test the motor skills
(see [14, 15]). Additionally, a mental cognition test with
the "Cognition" software battery was performed (see
Basner et al. ). Finally, two questionnaires were
handled to the participants: the Pittsburgh Sleep Qual-
ity Index (PSQI) and the State-Trait Anxiety Inventory
(STAI). The results obtained from the described tests
are not in the scope of this paper and should be later
clustered with the ight performance results, in order
to nd correlations between the mentioned measure-
Simulator Flig hts Vienna Test
System (VTS) Cognition Questionn aires
Instrumental Vis ual ISS Pitch
(every 3 sessi ons)
Figure 4: Session’s testing procedure.
2.2.1. The Soyuz-TMA simulator
At the Institute of Space Systems of the University of
Stuttgart, a Soyuz-TMA spacecraft simulator has been
used for training and research during the last decade
(see Noll et al. ). Dierent ight phases have been
under focus, most especially the close-range approach
and docking to the Russian segment of the ISS. The
spacecraft docking simulator, a self developed set-up
with a core based on the Orbiter Spaceight Simula-
tor (Schweiger ) is capable of recreating a realis-
tic docking manoeuvre and requires a high level of
skill performance from the pilot, who needs to con-
trol a 6 degrees-of-freedom spacecraft while monitor-
ing other ight parameters displayed in the featured
cockpit. The simulator’s hardware is a simplied ver-
sion of the real Soyuz cockpit, but it features the main
and necessary controls and systems that are needed to
perform a rendezvous and docking procedure (propul-
sion, radar, communications,...). The limited space and
weight to be shipped to Antarctica lead to the design
of an adapted one-seater simulator, which nevertheless
provided the same features for the pilot, as seen in Fig-
The ight procedure is represented in Figure B.1, as
it shows a typical ight prole for an approach tra-
jectory with the Soyuz-TMA spacecraft to one of the
four Russian docking ports of the ISS (Zvezda, Pirs,
Poisk, Rassvet). This procedure is a simplied trajec-
tory which doesn’t consider far-range staging points,
as the ight duration would be too long.
The ights to be performed during the experiment
were divided in a set of pre-dened scenarios, to en-
hance comparability. Table 3 describes how the scenar-
ios were distributed during the campaign. The ight
data was recorded and saved for posterior analysis. No
feedback on ight performance was given in real-time
but the parameters provided by the simulator’s instru-
mentation. As explained in Bosch Bruguera et al. ,
a ight analysis methodology has been developed in
order to process the raw parameters obtained from the
simulator. Such parameters were lineal and angular o-
sets to the relative target docking port, lineal and angu-
lar rotation rates of the spacecraft, fuel status and con-
trol inputs in each axis. More signicant parameters
regarding a certain ight phase, such as lateral oset
to the ideal trajectory, amount of inputs, among oth-
ers, were calculated out of the raw parameters recorded
every simulation time-step (0.2 s). See Table C.1 for a
description of the recorded and processed ight param-
The pilots had to follow a previously learned proce-
dure by performing the ight phases shown below:
•Final approach (FA)
A detailed description of the phases can be found in
Appendix B and in Bosch Bruguera et al. .
2.3. Statistical Analyses
A total of 1298 ights have been performed through-
out the experiment. In order to facilitate the assessment
of piltoing performance, the obtained ight parame-
ters and post-processed data, as shown in Appendix C,
have been analysed, rst, by visualizing trends, ight
trajectories and, posteriorly, by means of linear mixed-
eects models (LME). Such mathematical models oer
the possibility to correlate a predictor variable to mul-
tiple xed and random eects, as explained by Bates et
al. and the UCLA . In this case, LME models al-
low to correlate which experiment factors have played
a bigger role in the evolution of the ight performance
of the pilots. The presented models focus on the ef-
fect of the main grouping variables Campaign, Group,
Flight Type, Session ID and Pilot, which should show
how the application of the same experimental proto-
col in dierent conditions leads to signicant ight per-
formance dierences. Figure 6 depicts how the experi-
ment grouping variables are linked to each other. Note
that each pilot belongs to a specic group (FF or IF), and
this group is directly linked to a Campaign. Such rela-
tion is considered as "nested", as will be shown when
implementing the model.
The LME models run to analyse the ight data have
been assessed by performing ANOVA comparisons,
which showed in which extent a model provided better
predictions. The analysis of the models has been car-
ried out with the use of the lme4 package for R by Bates
et al.  and the data has been processed with help
of the R package sjPlot by Lüdecke . Additionally,
the odds ratios for ight success haven been estimated
by means of Generalized LME (GLME) using a bino-
mial function, which allows to understand which are
the probabilities of mission failure. The odd ratio rep-
resents the ratio between successful and failed ights.
Hence, a higher value is desired for the best reliabil-
ity. Further detail on the statistical analysis is available
3.1. Flight Trajectories
A useful tool to understand how pilots performed is
to plot the ight trajectories of each ight according
to the scenario own. Figure 7, Figure 8 and Figure 9
show the ight trajectories of all the own scenarios
in the SIMSKILL experiment. Note how the the tra-
jectory distribution diers between scenarios. Instru-
mental scenarios are own with much higher reliabil-
ity, as most of the pilots aligned to the center of the
approach corridor with better accuracy. Moreover, the
approach speed was monitored and adapted with ex-
actitude, causing a rainbow-coloured alignment path
since the majority of pilots kept the required speed.
It is however important to remark that a considerable
amount of pilots accelerated towards the ISS before en-
tering the safety zone dened by the virtual corridor, as
seen by the more diagonal trajectories.
The visual ights shown in Figure 8 present a rather
dispersed distribution, both in terms of alignment and
Left MFD Right MFDPeriscope
Figure 5: Overview of the mobile Soyuz-TMA simulator. MFD: Multi-Functional Display. RHC and LHC: Right and Left Hand Controllers.
Scenarios Frequency Initial conditions Description
P os[m]; ˙
P os[m/s]; ~
Instrumental Every session (68.201,5.211,-12.670); (-0.033,-
Use of docking instrumentation al-
Visual Every session (69.741, 17.865,5.055); (0.030,-
Only use of periscope view.
ISS-Pitch Every 3 sess. (79.797,-24.279,3.517); (-0.063,-
Use of instrumentation limited to ap-
proach velocity monitoring. Addi-
tional diculty due to ISS pitch rota-
Table 3: Scenario description
Of f : Relative position to target’s port (Long., Vert., Horiz.).
Ang: Relative angle to target’s port main axis (Roll, Yaw, Pitch).
approach velocity. Note how some of the trajectories
are still outside the approach corridor within the last
meters to contact the station.
Finally, the ISS-Pitch scenario is depicted in Figure
9. The use of the instrumentation enhanced the capa-
bility of the pilots to align correctly inside the corridor,
although with less precision to the rst scenario. The
fact that the target station rotated continuously proves
how some trajectories permanently oscillate, even tres-
passing the corridor’s limits.
3.2. Performance Assessment
The grouping variables Campaign, Group, Session
ID, Flight Type and Pilots presented in section 2.3
have been implemented as factors for the analysis
of three performance-relevant parameters: Alignment
time, amount of Steering errors and, nally, the amount
of unsuccessful ights. The data trends have been
visualized by using a combination of box-plots (see
MATLAB "boxplot"), histograms and, later, anal-
ysed with LME and GLME models.
3.2.1. Training Frequency - Pilot Groups
A dierence on ight performance between frequent
and infrequent yers is expected in the later months
after the rst checkight (RF0). Figure 10 shows how
the percentage of failed ights in all the scientic cam-
paigns decreases clearly for FF pilots: initially, approx-
imately a 12% of the performed ights couldn’t be
docked to the target docking port. A much lower per-
Stugart 2016 Halley VI 2015 Halley VI 2016 Concordia 2018 Concordia 2019
FF IF FF FF IFFF IF
Only on RF0,3,6,9
RF0 RF1 RF2 RF3 RF4 RF5 RF6 RF7 RF8 RF9
Figure 6: Representation of the relation between the experiment grouping variables.
centage of unsuccessful ights happens for the IF group
in the session RF0. The initial dierence between FF
and IF is caused by a singular pilot in one of Halley
VI campaigns. The use of a Generalized Linear Mixed-
Eects binomial model allows to integrate the eects
of singular pilots as random variables and obtain pre-
ciser estimations on how the risk of mission failure
should evolve, according to the data recorded. The ob-
tained model results prove that the odds-ratio of mis-
sion success for frequent yer pilots is higher on all
subsequent ight sessions, with an estimate value rang-
ing from 35,02 (p<.001, condence interval 95%) up to
59,50 (p<.001) on the last experiment’s session RF9. On
the other hand, the infrequent yer group starts very
reliably, but decreases its performance already at the
next session RF3, with a value of 14,82 (p<.001). On RF6
and RF9 the reliability slightly improves, ending with a
value of 29,16 (p<.001).
The eects of interrupted training frequency can
also be assessed by observing the duration of the Sys-
tems Activation phase or "Alignment Start" in this doc-
ument. This can be observed more especially in the in-
strumental scenarios of each session, since they are run
rst. As seen in Figure 11, the time that pilots required
for activating the spacecraft prior to alignment and ap-
proach is clearly aected by the training frequency.
The rst check-ight (instrumental, RF0) shows very
similar times with a median of approximately 36 sec-
onds (p<.001), whereas the later sessions tend to show
an increase of time required by IF pilots up to 70 sec-
onds (p<.001). Instead, FF pilots take more time until
session RF3, but improve their systems activation dura-
tion and keep it constant to a value of about 40 seconds
during the whole experiment.
The amount of steering input errors that happen in
each ight has been evaluated along the experiment
time-line, as shown in Figure 12. Note how frequent
yer pilots do more errors during the RF3, RF6 and
RF9 sessions, which include the more dicult scenario
"ISS-Pitch". The presented data has been analysed by
means of an LME that takes into consideration Cam-
paign, Group and Pilots as nested random eects, plus
an additional random eect on the intersect for the
Flight Type. The dierences between FF and IF groups
have proven to be statistically signicant at the level of
Figure 7: Flight Trajectories on the Instrumental Scenario. Flight samples: 406.
Figure 8: Flight Trajectories on the Visual Scenario. Flight samples: 377.
3.3. Environmental conditions: Isolation, Connement
and Hypoxia - Campaigns
In order to understand how piloting performance
and reliability are aected by factors such as isola-
tion, connement or hypoxia, a comparison between
the campaigns in Concordia, Halley VI and Stuttgart
needs to be done. Figure 13 displays the distribu-
tion of amount of steering input errors for each cam-
paign. The black boxes representing the groups of fre-
quent yers show how each of the campaigns kept an
amount of approximately 25 to 30 errors. However,
the blue boxes representing the infrequent yers pi-
lots group are placed at higher values, diering de-
pending on which campaign. Note that the Concordia
campaigns show a higher dierence between FF and
IF. In Halley VI, the dierence between both groups
still exists, whereas in Stuttgart both groups deliver
the same amount of errors. The only campaign that
presents statistically signicant dierences is Concor-
dia 2018 (p<.05), with an estimate of 56 input errors per
Figure 9: Flight Trajectories on the ISS Pitch Scenario. Flight sam-
The systems activation phase has been also com-
puted in order to compare which role could the ex-
perimental environment play, that is, the campaign.
As seen in Figure 14 . A statistical signicance has
been found out in the IF group of Concordia 2018 (90
s, p<.05) and the results of Stuttgart 2016 are also real-
itvely signicant with p=0.152. See how the frequent
yer groups deliver a similar value for all campaigns,
while the Antarctic campaigns have shown to lead to a
longer Alignment Start duration for infrequent yers.
An extended analysis of the ight data described in
Appendix C has not been considered in the scope of
this paper and are expected to be analysed in further
4.1. Training frequency
The presented results show how training periodicity
is a key factor to ensure piloting reliability, as an activ-
ity that requires high concentration and psycho-motor
skills. As proven by Salnitski et al. in previous ISS and
Mir missions, a three-month period of training causes
a critical decay of skills and endangers mission success.
SIMSKILL conrms such a statement by providing a de-
tailed comparison by FF and IF pilots, where dierences
between both groups are signicant in later phases of
isolation. It is important to note that studied param-
eters have shown substantial dierences between fre-
quent yers and infrequent yer groups. In the rst
case, a contstant performance has been delivered by
most of the subjects, although an initial degradation
happened from phases RF0 to RF3. One of the reasons
for that behaviour can be explained by the sudden in-
terruption of training, since subjects had been trained
Percentage of failed ights
Figure 10: Percentage of non-docked ights in each experimental
phase. Flight samples: 1298.
in a one to two day basis before performing the initial
check-ight (RF0). A test after one month was at rst
challenging, but then proved to stabilize throughout
the experiment. On the other hand, infrequent yers
were applied a three-month training frequency, show-
ing a clear decay of performance. Thus, the application
of training sessions with at maximum a month of dif-
ference ensures a constant piloting performance, even
an improvement in certain pilots. However, a longer
interruption of training leads to erratic piloting and sig-
nicantly reduces the mission success rate.
4.2. Isolation and Connement
Despite the existing individual dierences, a rst
assessment of the dierent crews and campaigns has
shown that isolation and connement, as well as hy-
poxia in Concordia, can be factors that increase the de-
cay of piloting performance as well of that of other
activities that require a high level of cognitive and
psycho-motor skills. The dierences in performance
between each campaign show that the crew-members
in Stuttgart delivered a more stable performance, both
FF and IF pilots, whereas in Antarctica, especially in
Concordia, the reliability of the pilots decreased sub-
stantially, as shown in Figure 13.
The so-called third quarter phenomenon (see Wil-
son ) has been assessed but no remarkable ndings
could be pointed out. Further analysis on that matter
will proceed after clustering the data from other tests.
Alignment Start [s]
Figure 11: Start of the alignment phase in instrumental ights. F light
The inuence of hypoxic conditions in Concordia
could be one of the performance degradation cataly-
sers. The higher values obtained in both Concordia
campaigns, compared to those of the Halley VI station
lead to consider the eects of hypoxia to skills degra-
dation, although the presented data is not conclusive,
as more factors need to be analysed in order to assure
the direct link of hypoxia to performance decay.
4.4. Additional factors
A multitude of extra variables play a big role on
the obtained results, mostly the personal dierences
between subjects. Age, qualication, gender, cultural
background, experience with gaming devices are pa-
rameters that dene how pilots performed. Some re-
sults have been found to be diering between crews,
which could imply a direct inuence of the training
procedure prior to session RF0. The eect of the train-
ing performed by a selected ight instructor can not be
underestimated, even if a very strict experimental ight
protocol had been designed. Finally, it is also important
to mention that besides the purely physical factors such
as hypoxia and isolation, the inter-personal relations
between each crew in Antarctica can be an additional
factor to be considered when analysing performance of
the crew. As an example, the already challenging con-
ditions that the Concordia Station poses to the crew
Total Input Errors
Figure 12: Total steering input errors per ight. Flight samples: 1298.
could be magnied by intra-cultural dierences in a
4.5. Performance degradation countermeasures
In order to ensure mission success, it has been recog-
nized that training frequency is the main factor, since
frequent yers in all campaigns performed similarly.
As already seen in astronaut training, the denition
of standard training procedures as well as contingency
planning are crucial for the astronaut to evaluate and
solve any issues on board. A predened training rou-
tine is consequently essential.
The inuence of the hardware set-up can’t be ig-
nored, as piloting performance can vary strongly de-
pending on the human-computer-interfaces such as the
joysticks and displays. Therefore, a ight simulator
with a realistic astronaut cockpit position has been de-
The use of an adapted, but realistic Soyuz-TMA
spaceight simulator in Antarctica and in the Institute
of Space Systems of the University of Stuttgart has pro-
vided a big dataset that allows to understand how pi-
loting performance is aected by isolation and con-
nement, hypoxia and interrupted training periodicity.
An experimental protocol was dened accordingly, and
has allowed to gather a valuable dataset, not only from
Total Input Errors
Figure 13: Total steering errors on all campaigns. Flight samples:
the simulator ights, but also from additional testing
devices.Parameters such as the cockpit activation dura-
tion, the amount of steering errors showcase how hu-
man memory and precision decrease after long periods
without training, and should be further correlated to
the Cognition and Vienna Test System tests performed
The application of linear mixed-eects models for
the data analysis has proven to be a reliable tool when
it comes to assess parameters aected by multiple vari-
ables. The possibility to integrate random variables
into the model allows not to only t the model more
precisely, but to nd out which additional factors play
a bigger role in the variability of the data. In this
experiment’s case the skill dierences between pilots
have shown the biggest variance, which has inevitably
caused the appearance of non statistically signicant
The presented paper is an overview of the obtained
data from the recently completed SIMSKILL experi-
ment. Further results and an enhanced performance
analysis methodology are expected to give deeper de-
tail on which are the most critical factors for the deni-
tion of a reliable human long duration space mission.
Alignment Start [s]
Figure 14: Start of the Alignment phase for the last experimental
session, RF9. Flight samples: 67.
The current work is fruit of a wide range of partners
and participants. Thanks to the DLR and BMWi for
sponsoring the project. Acknowledgements to IPEV,
PNRA and BAS for their incalculable support on im-
plementing this experiment in both Concordia and Hal-
ley VI stations. Special thanks to the SciSpaceE Team
of the ESA for opening the possibility to perform sci-
ence in such a singular environment and coordinating
the many dierent participants. Finally and most im-
portantly, a big acknowledgement to all volunteer par-
ticipants of the SIMSKILL experiment as well as the
ESA Research Medical Doctors in Antarctica, who con-
tributed to this project during a whole campaign for the
sake of science and space research.
 V. Salnitski, A. Dudukin, and B. Johannes, “Evaluation of opera-
tor’s reliability in long-term isolation (the “pilot”-test),” Simula-
tion of extended isolation: advances and problems. Slovo, Moscow,
pp. 30–50, 2001.
 NASA, “Lunar gateway.” https://www.nasa.gov/
topics/moon-to- mars/lunar-gateway, Sept.
2019. Accessed: 2019-09-06.
 T. J. Ngo-Anh and A. Rossiter, “Platforms for stress and im-
mune research in preparation for long-duration space explo-
ration missions,” in Stress Challenges and Immunity in Space,
Springer, Jan. 2020.
 B. Johannes, V. Salnitski, A. Dudukin, L. Shevchenko, and
S. Bronnikov, “Performance assessment in the pilot experiment
on board space stations mir and iss,” Aerospace medicine and
human performance, vol. 87, no. 6, pp. 534–544, 2016.
 I. B. Ushakov, M. B. Vladimirovich, Y. A. Bubeev, V. I. Gushin,
G. Y. Vasil’eva, A. G. Vinokhodova, and D. M. Shved, “Main
ndings of psychophysiological studies in the mars 500 ex-
periment,” Herald of the Russian Academy of Sciences, vol. 84,
pp. 106–114, Mar 2014.
 IMBP, “Sirius experiment.” http://sirius.imbp.
info/, Sept. 2019. Accessed: 2019-09-06.
 I. L. Schlacht, B. Foing, O. Bannova, F. Blok, A. Mangeot,
K. Nebergall, A. Ono, D. Schubert, and A. M. Kolodziejczyk,
“Space analog survey: Review of existing and new proposal
of space habitats with earth applications,” in 46th International
Conference on Environmental Systems, 2016.
 T. J. Ngo-Anh, A. Rossiter, A. Suvorov, G. Vassilieva, and
V. Gushin, “Mars500: The rst preparation of long-duration
space exploration missions—results and implications for a
holistic stress and immune research approach,” in Stress Chal-
lenges and Immunity in Space, Springer, Jan. 2020.
 L. Bessone, K. Beblo-Vranesevic, Q. A. Cossu, J. De Waele,
S. Leuko, P. Marcia, P. Rettberg, L. Sanna, F. Sauro, and S. Taiti,
“Esa caves: Training astronauts for space exploration,” in Pro-
ceedings of the 16th International Congress of Speleology. Brno,
pp. 321–327, 2013.
 M. D. von Ehrenfried, “Terrestrial cave and lava tube research,”
in From Cave Man to Cave Martian, Springer, Jan. 2019.
 B. Todd and M. Reagan, “The neemo project: A report on how
nasa utilizes the“aquarius”undersea habitat as an analog for
long-duration space ight,” in Engineering, Construction, and
Operations in Challenging Environments: Earth and Space 2004,
pp. 751–758, 2004.
 M. B. G. Kanki, J.-F. Clervoy, and G. Sandal, Space Safety and
Human Performance. Elsevier Science & Technology, 2017.
 N. Kanas and D. Manzey, Space psychology and psychiatry,
vol. 22. Springer Science & Business Media, 2008.
 Schuhfried GmbH, “Vts motor performance series.” https:
//www.schuhfried.com/test/MLS, Sept. Accessed:
 N. C. H. Ong, “The use of the vienna test system in sport psy-
chology research: A review,” International Review of Sport and
Exercise Psychology, vol. 8, pp. 204–223, jan 2015.
 M. Basner, A. Savitt, T. M. Moore, A. M. Port, S. McGuire, A. J.
Ecker, J. Nasrini, D. J. Mollicone, C. M. Mott, T. McCann, D. F.
Dinges, and R. C. Gur, “Development and validation of the "cog-
nition" test battery for spaceight,” Aerospace Medicine and Hu-
man Performance, vol. 86, pp. 942–952, nov 2015.
 J. Noll, S. Belz, M. Schmitz, and E. Messerschmid, “Educa-
tional training on spacecraft operations in a soyuz mock-up
simulator,” in Proceedings of the 61st International Astronautics
Congress, Prague, CZ, International Astronautical Federation,
 M. Schweiger, “Spacecraft simulation and visualisation with
orbiter 2006,” in 3rd International Workshop on Astrodynamics
Tools and Techniques, 2006. Accessed: 2020-03-18.
 M. Bosch Bruguera, V. Schöneich, A. Fink, V. Schröder,
E. Dessy, N. Pattyn, and R. Ewald, “Spacecraft piloting perfor-
mance assessment - a computational evaluation methodology
for the simskill experiment,” in Proceedings of the 68th Interna-
tional Astronautical Congress, Adelaide, AUS, 2017.
 D. Bates, M. Mächler, B. Bolker, and S. Walker, “Fitting linear
mixed-eects models using lme4,” Journal of Statistical Soft-
ware, vol. 67, no. 1, pp. 1–48, 2015.
 I. f. D. R. . E. UCLA, “Introduction to linear mixed models, linear
 D. M. Bates, lme4: Mixed-eects modeling with R, Linear Mixed-
 D. Lüdecke, sjPlot: Data Visualization for Statistics in Social Sci-
ence, Linear Mixed-Eects, 2020. R package version 2.8.3.
 MathWorks, “Box plot.” https://www.mathworks.
com/help/stats/boxplot.html, Sept. 2006. Ac-
 D. Wilson, “The third-quarter phenomenon in antarctic person-
Appendix A. Experiment Protocol
Brunt Ice Shelf
0 m AMSL
3200 m AMSL
500 m AMSL
Figure A.1: Chronological representation of the ve dierent SIMSKILL Campaigns performed in Concordia, Halley VI and Stuttgart (control
group). AMSL - Above Mean Sea Level
Initial Checkflights (RF0)
Initial Checkflights (RF0)
Final Checkflights (RF9)
Final Checkflights (RF9)
Figure A.2: Example of an experiment campaign. In this case, Concordia 2019. RF: Regular Flight
Appendix B. Flight Prole
20 m 0 m
Start of the
Figure B.1: Typical ight prole for the approach and docking procedure of the Soyuz spacecraft.
Phase Description and rules
Systems activation Start of the propulsion and docking systems: Activate Propulsion System’s valves and RCS;
Initiate docking instrumentation; Lock on to target docking port; Observe periscope and
plan a ight strategy.
Alignment Point spacecraft to target docking port. Keep zero approach velocity. Move laterally to
align with approach corridor. Avoid losing visibility in periscope.
Approach Once in approach corridor, approach is allowed. Fire THC forwards to the required ap-
proach speed. Correct attitude and alignment as needed. Reduce speed as distance to target
decreases. Monitor docking instrumentation if allowed.
Final Approach Once within the last 20 meters to target port, keep approach speed to 0,10 m/s. Do minimal
corrections and focus to the docking marker, if visibility is enough.
Table B.1: Flight Phases description
Appendix C. Flight Parameters
Each of the ight parameters shown in Table C.1 has been analysed for all three active ight phases (Alignment,
Approach and Final Approach (FA). Some parameters are provided with the average (AVG) and root mean square
(RMS) throughout a whole ight phase. Root Mean Square should provide a better insight to the variability of the
parameter during the time period, although not being able to show negative signs.
Parameter Units Description
Phase Start s Start time of the ight phase.
Phase Duration s Time elapsed between start and end of a phase.
Docking Time s Total time until docking contact.
Time out of corridor s Total time outside the approach corridor
Fuel spent kg Amount of propellant spent during a phase
Lateral Oset m Lateral oset from target’s docking port main axis, as a function of
Of fyand Offz, respective lateral osets on the target port’s local axes.
LatOf f =qOff 2
y+Of f 2
Time without visibility s Time of ight without direct periscope view to the target docking port.
Steering Inputs - Amount of steering inputs per phase for each axis of the RHC and LHC.
Steering Average Duration s Average duration of the steering inputs on each axis.
Steering Errors - Amount of steering errors on each phase.
Induced Errors - Amount of errors provoked on a joystick axis due to the input on another
axis of the same joystick.
Combined Joysticks - Number of times where both joysticks were actuated simultaneously.
Combined LHC Axes - Number of times where any of the LHC axes are activated simultane-
Combined LHC (YZ) Axes - Number of times combining axes Y and Z of the LHC.
Combined RHC Axes - Number of times where any of the RHC axes are activated simultane-
Combined RHC (YZ) Axes - Number of times combining axes Y and Z of the RHC.
Combined Time s Average time spent on each of the above combinations.
Lateral Oset AVG/RMS mLateral Oset calculated throughout a ight phase.
Approach Velocity AVG/RMS m/s Approach Speed calculated throughout a ight phase.
Lateral Velocity AVG/RMS m/s Lateral Velocity calculated throughout a ight phase.
Angle AVG/RMS ºPitch, Yaw and Roll calculated throughout a ight phase.
Angular Rate AVG/RMS º/s Pitch, Yaw and Roll rates calculated throughout a ight phase.
Fuel Spent on Error Inputs kg Amount of propellant spend for wrong inputs.
% Steering Errors - Percentage of steering error, in relation to the total amount.
% Fuel on Errors - Percentage of fuel spent on erratic steering inputs, in relation to the total
AVG fuel consumption kg/s Mass of fuel spent on average for the whole ight duration.
Table C.1: Flight parameters overview.