Optimization heuristics for scheduling observations in remote sensing missions: single-instrument case

MASTER
FINAL THESIS
Op imiza ion
Heu is ics
o
Scheduling Obse a ions in
Remo e Sensing Missions:
Single-Ins umen case
Au ho :
Pie o Di Sa no
Di ec o / Co-di ec o :
Manel So ia Gue e o / Da id de la To e Sang à
Deg ee:
Mas e in Space and Ae onau ical Enginee ing
Examina ion session:
Au umn 2024
Documen :
Repo
i
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
ii
Abs ac
The planning and scheduling o scien i ic obse a ions in space missions p esen complex
challenges due o ope a ional cons ain s, including limi ed esou ces, s ingen ime
windows, and he need o maximize scien i ic e u n. This hesis ocuses on de eloping
op imiza ion heu is ics, speci ically le e aging Gene ic Algo i hms (GAs), o add ess hese
challenges. The s udy u ilizes he ESA’s JUICE mission as a e e ence, wi h pa icula
emphasis on he JANUS mul ispec al came a asked wi h obse ing 141 Regions o
In e es (ROIs) on Ganymede du ing mul iple lybys.
The esea ch in eg a es ad anced me hodologies such as he "Online F on ie Repai "
mosaicing algo i hm o imp o e he accu acy o obse a ion scheduling and quali y
assessmen s. This algo i hm is combined wi h Py hon-based op imiza ion amewo ks and
NASA's SPICE oolki o ensu e obus handling o mission da a and cons ain s. Mul i-agen
logic is u he employed o enhance compu a ional e iciency and solu ion quali y.
The me hodology was alida ed h ough a wo-s ep app oach: ini ial es s on single ROIs o
e i y algo i hm unc ionali y, ollowed by comp ehensi e scheduling ac oss all Ganymede
ROIs. Resul s demons a ed he e ec i eness o he p oposed ools in gene a ing op imized
obse a ion schedules ha espec geome ic and empo al cons ain s. Compa isons wi h
p e ious app oxima ions e ealed imp o emen s in obse a ion quali y and scheduling
accu acy.
This wo k highligh s he po en ial o Gene ic Algo i hms o s eamline and enhance space
mission planning, p o iding mission planne s wi h e icien ools o op imizing complex
scheduling p oblems. The indings unde sco e he impo ance o in eg a ing ad anced
compu a ional echniques o maximize scien i ic ou pu while adhe ing o mission
cons ain s.
iii
Resumen
La plani icación y p og amación de obse aciones cien í icas en misiones espaciales
p esen an desa íos complejos debido a las es icciones ope a i as, como ecu sos
limi ados, en anas de iempo es ic as y la necesidad de maximiza el e o no cien í ico.
Es a esis se cen a en desa olla heu ís icas de op imización, especí icamen e u ilizando
Algo i mos Gené icos (AG), pa a abo da es os desa íos. El es udio u iliza como e e encia
la misión JUICE de la ESA, con un en oque pa icula en la cáma a mul iespec al JANUS,
enca gada de obse a 141 Regiones de In e és (ROIs) en Ganímedes du an e múl iples
sob e uelos.
La in es igación in eg a me odologías a anzadas, como el algo i mo de mosaico "Online
F on ie Repai ", pa a mejo a la p ecisión en la p og amación de obse aciones y en la
e aluación de la calidad. Es e algo i mo se combina con ma cos de op imización basados
en Py hon y el ki de he amien as SPICE de la NASA pa a ga an iza un manejo obus o
de los da os y las es icciones de la misión. Además, se emplea lógica mul iagen e pa a
mejo a la e iciencia compu acional y la calidad de las soluciones.
La me odología ue alidada median e un en oque de dos pasos: p uebas iniciales en ROIs
indi iduales pa a e i ica la uncionalidad del algo i mo, seguidas de una p og amación
in eg al de odas las ROIs de Ganímedes. Los esul ados demos a on la e icacia de las
he amien as p opues as pa a gene a p og amas op imizados que espe an las
es icciones geomé icas y empo ales. Las compa aciones con ap oximaciones p e ias
e ela on mejo as en la calidad de las obse aciones y en la p ecisión de la p og amación.
Es e abajo des aca el po encial de los Algo i mos Gené icos pa a simpli ica y mejo a la
plani icación de misiones espaciales, p opo cionando a los plani icado es he amien as
e icien es pa a op imiza p oblemas complejos de p og amación. Los hallazgos sub ayan
la impo ancia de in eg a écnicas compu acionales a anzadas pa a maximiza los
esul ados cien í icos cumpliendo con las es icciones de la misión.

Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
i
Table o con en s
Abs ac ................................................................................................................. ii
Resumen .............................................................................................................. iii
Table o con en s ..................................................................................................
Lis o ables ....................................................................................................... ii
Lis o igu es ..................................................................................................... iii
Lis o abb e ia ions / Glossa y ......................................................................... ix
1. In oduc ion ................................................................................................. 1
1.1 Objec ................................................................................................................. 3
1.2 Scope ................................................................................................................. 3
1.3 Requi emen s ..................................................................................................... 4
1.4 Jus i ica ion ......................................................................................................... 4
2 Backg ound .................................................................................................... 6
2.1 S a e o he A o Mission Scheduling................................................................. 6
2.1.1 Oppo uni y Analysis and Spacec a Cons ain s Checking ......................... 6
2.1.2 Obse a ion Scheduling ............................................................................... 6
2.1.3 Schedule Op imiza ion ................................................................................. 7
2.1.4 Ad ances in Au oma ion and Au onomy ....................................................... 7
2.1.5 Challenges and Fu u e Di ec ions ................................................................ 7
2.2 The JUICE mission ............................................................................................. 8
2.2.1 Main scien i ic objec i es ............................................................................. 8
2.2.2 Mission p o ile .............................................................................................. 9
2.2.3 JANUS........................................................................................................ 11
3 Me hodology ................................................................................................ 13
3.1 Mosaic heu is ics ...............................................................................................13
3.1.1 A ea Co e age Planning P oblem (ACPP) ..................................................13
3.1.2 Online F on ie Repai ................................................................................15
3.2 Gene ic Algo i hms .............................................................................................17
3.2.1 Geno ype and Pheno ype ...........................................................................18
3.2.2 Gene ic ope a o s .......................................................................................19
4 The Scheduling P oblem ............................................................................. 21
4.1 The Science Oppo uni y Analysis .....................................................................21
4.1.1 P oblem inpu s ............................................................................................21
4.1.2 Logical low and classes .............................................................................22
4.1.3 The oPlanRoi class and he Online F on ie Repai implemen a ion ...........26
4.2 Gene ic Algo i hm ..............................................................................................35
4.2.1 oplan class .................................................................................................35
4.2.2 aga class ....................................................................................................39
4.2.3 Conside a ions on benchma king ................................................................40
4.2.4 The mul i-agen oolbox ..............................................................................40
5 Resul s and discussion ............................................................................... 41
5.1 Valida ion: GA wi h a single ROI ........................................................................41
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
i
5.2 Single-Agen and Mul i-Agen : esul s and discussion ........................................45
6 Budge summa y ......................................................................................... 52
7 Analysis and assessmen o en i onmen al and social implica ions ..... 53
7.1 En i onmen al implica ions ................................................................................53
7.2 Social Implica ions .............................................................................................53
7.3 E hical Conside a ions .......................................................................................54
8 Conclusions ................................................................................................. 55
8.1 Main Resul s ......................................................................................................55
8.2 Con ibu ions and Implica ions ...........................................................................55
8.3 Limi a ions and Fu u e Pe spec i es ..................................................................56
9 Re e ences ................................................................................................... 57
ii
Lis o ables
Table 2.1: JUICE's main mission phases .........................................................................10
Table 2.2: JANUS's key ea u es ......................................................................................12
Table 4.1: O icial s a and end imes o JUICE lybys a ound Ganymede aimed a JANUS
expe imen s [33] .........................................................................................................22
Table 4.2: Angle pa ame e s and ela i e cons ain s o ge he complian ime window ...23
Table 4.3: Values o he Ins umen class objec ini ialized o JANUS..............................26
Table 4.4: GA pa ame e s' alues based on benchma king ..............................................40
Table 5.1: JUICE schedule (pa 1) ..................................................................................47
Table 5.2: JUICE schedule (pa 2) ..................................................................................48
Table 5.3: JUICE schedule (pa 3) ..................................................................................49
Table 5.4: JUICE schedule (pa 4) ..................................................................................50
Table 6.1: Budge summa y .............................................................................................52
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
4
1.3 Requi emen s
The equi emen s o his p ojec de ine he cons ain s ha he p og am mus adhe e o.
Speci ically:
 The code should be de eloped in Py hon, chosen o i s use - iendliness, lexibili y,
and open-sou ce na u e. The code should be clea , well-s uc u ed, s aigh o wa d
o ollow and code mus include ou pu p in op ions o clea ly display he inal
schedule and he quali y o he obse a ion plan.
 The code mus be uploaded o Gi Hub, no only o ensu e accessibili y o all use s
bu also o acili a e collabo a ion among mul iple indi iduals, gi en he complexi y
o he p oblem.
 Whene e possible, he code should ollow an objec -o ien ed app oach. Using
Py hon classes makes he code easie o unde s and and modi y, acili a ing he
addi ion o new ea u es h ough me hod and class o e iding.
 The implemen a ion mus suppo unning mul iple ins ances simul aneously,
enabling he use o mul iple agen s o enhance he op imiza ion p ocess.
 I should be adap able and designed o handle a ious ROIs, plane s, a ge s, o
missions.
1.4 Jus i ica ion
Based on he ajec o y da a om e sion 3 o he mission’s Consolida ed Repo on Mission
Analysis (CReMA), scien is s had ini ially calcula ed he su ace a eas whe e images could
be cap u ed a he highes esolu ion wi h he aim o de e mining which egions a ound
Ganymede would be obse able du ing he upcoming lybys.
As he s udy p og essed, he mission p o ile was upda ed o CReMA 4.2b, and la e o
CReMA 5 ( he cu en e sion is CReMA 5.1) [5]. These new models in oduced signi ican
changes o he lyby condi ions, in pa icula , he illumina ion condi ions wo sened, and he
egions co e ed by he lybys we e al e ed, which made obse a ions a la i udes abo e 40º,
al eady challenging in CReMA 3, e en mo e di icul [5].
This p esen s a unique oppo uni y, as i allows he use o he de eloped ools o assess
which egions a e now obse able, he speci ic TW (Time Window) when hey a e isible
and which ROIs will be included in he op imal schedule wi h he upda ed ajec o y, based
on he eal mission pa ame e s. This will p o ide aluable insigh in o he pe o mance o
he algo i hm in add essing p ac ical, eal-wo ld p oblems.
Addi ionally, he la ge numbe o obse able ROIs o he JUICE mission a Ganymede
makes i an excellen es case o e alua ing he algo i hms. The di e si y and complexi y
o hese egions will o e c i ical insigh s in o he s eng hs and weaknesses o applying
gene ic algo i hms o his ask.

5
Figu e 1.2: JANUS co e age (nadi looking only) du ing he Ganymede lybys based on CReMA
3.0: a) incidence angle, b) phase angle and c)spa ial esolu ion o e laid on he global base map by
[6].
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
6
2 Backg ound
2.1 S a e o he A o Mission Scheduling
The app oach used in his hesis o sol e he mission scheduling p oblem has been
in es iga ed by nume ous esea che s. Fo example, among he opics s udied du ing he
PhD, Paula Be iu Ro e [7] analyzed he use o gene ic algo i hms and men ioned hei
possible combined use wi h simula ed annealing o sol e he mission scheduling p oblem.
In he same line o esea ch, a UPC, Jo ge Simon Azna [8] wo ked on s udying he case
wi h mul iple ins umen s, and Diengo Andia [9], whose wo k included unc ions o selec
he bes egions o obse e and an algo i hm o co e age analysis and image mosaicing
o hese egions.
Mission scheduling is a c ucial aspec o space explo a ion, ensu ing ha scien i ic
objec i es a e achie ed wi hin he ope a ional and esou ce cons ain s o spacec a . I
ypically in ol es ou main ac i i ies: oppo uni y analysis, spacec a cons ain s
checking, obse a ion scheduling and schedule op imiza ion. These ac i i ies
collec i ely ensu e ha missions maximize scien i ic ou pu while adhe ing o echnical
and logis ical limi a ions. O e he yea s, space agencies and esea ch o ganiza ions ha e
de eloped ad anced ools and me hodologies o add ess hese challenges, enhancing
bo h e iciency and au onomy in mission ope a ions.
2.1.1 Oppo uni y Analysis and Spacec a Cons ain s Checking
Oppo uni y analysis ocuses on iden i ying a o able ime windows o obse a ions. This
p ocess in ol es analyzing spacec a ajec o ies, o bi al pa ame e s, and isibili y
condi ions o de e mine when speci ic ins umen s can conduc obse a ions unde
op imal condi ions. Tools like CLASP (Comp essed La ge-scale Ac i i y Scheduling and
Planning) ha e been used in missions such as NISAR ( o me ly DESDynI), le e aging
geome ic cons ain s and empo al pa ame e s o de elop easible obse a ion
schedules. CLASP enables he apid gene a ion o high-le el ope a ional plans while
accoun ing o esou ce a ailabili y and scien i ic p io i ies [10].
Spacec a cons ain s checking ensu es ha plans adhe e o spacec a capabili ies, such
as powe consump ion, memo y s o age, and da a ansmission bandwid h. Fo ins ance,
CASPER (Con inuous Ac i i y Scheduling, Planning, Execu ion, and Replanning)
dynamically e i ies and adjus s schedules in esponse o eal- ime upda es on spacec a
s a us. This ool has p o en e ec i e in Ea h-obse ing missions and plane a y
explo a ion scena ios, enabling spacec a o au onomously adap o unexpec ed changes
in condi ions o mission pa ame e s [11][12].
2.1.2 Obse a ion Scheduling
The scheduling o obse a ions in ol es alloca ing speci ic asks o ime slo s while
ensu ing he op imal u iliza ion o spacec a ins umen s and esou ces. Due o he limi ed
ope a ional windows and compe ing objec i es, obse a ion scheduling is o en a
complex, mul i-objec i e p oblem. Tools like Eagle Eye ha e been ins umen al in
imp o ing scheduling e iciency by analyzing co e age and mosaicing oppo uni ies. Fo
ins ance, Eagle Eye's algo i hms enable au onomous selec ion o obse a ion a ge s
based on p ede ined scien i ic p io i ies, ensu ing comp ehensi e co e age o egions o
in e es du ing plane a y lybys [13].
7
In missions such as Galileo and JUICE, obse a ion scheduling has also bene i ed om
ools like he Obse a ion Planning Tool o Ins umen and Mission Analysis
(OPTIMA). OPTIMA employs heu is ic algo i hms o balance compe ing objec i es, such
as maximizing scien i ic ou pu and minimizing esou ce consump ion. This ool has been
alida ed in scena ios in ol ing Eu opa and Ganymede lybys, whe e i success ully
op imized obse a ion imelines o enhance esolu ion and co e age [7].
2.1.3 Schedule Op imiza ion
Schedule op imiza ion ocuses on e ining mission schedules o achie e maximum
scien i ic e u ns. Va ious me hodologies, including heu is ic algo i hms, gene ic
algo i hms, and cons ain p og amming, ha e been applied o his end. Fo example,
ools like SciBox s eamline he op imiza ion p ocess by au oma ing he anking and
selec ion o obse a ion oppo uni ies based on c i e ia such as esolu ion, illumina ion,
and da a ansmission capaci y. SciBox was success ully deployed in he MESSENGER
mission, whe e i educed planning ime signi ican ly while maximizing obse a ion
e iciency [14].
CLASP u he illus a es he po en ial o op imiza ion ools by employing i e a i e epai
algo i hms o add ess scheduling con lic s. In missions equi ing mul i-sa elli e
coo dina ion, hese algo i hms dynamically adap schedules o accommoda e changes in
ope a ional condi ions o scien i ic p io i ies, ensu ing obus and con lic - ee imelines [10].
2.1.4 Ad ances in Au oma ion and Au onomy
Au oma ion has played a pi o al ole in ad ancing mission scheduling, educing manual
in e en ions and enabling spacec a o espond dynamically o e ol ing mission
equi emen s. Tools like CASPER exempli y his p og ess, allowing spacec a o
au onomously e-plan ac i i ies in esponse o eal- ime da a, he eby enhancing mission
lexibili y and e iciency [11]. Simila ly, he in eg a ion o machine lea ning algo i hms in o
scheduling sys ems is an eme ging end, enabling p edic i e p io i iza ion o asks based
on his o ical da a and eal- ime cons ain s [15].
Ano he no ewo hy inno a ion is Ea h Obse ing Au onomy (EOA), which combines
onboa d da a analysis wi h eal- ime scheduling. By au onomously analyzing collec ed
da a and e-p io i izing obse a ions, EOA enables spacec a o adap hei ac i i ies
wi hin minu es, signi ican ly educing eliance on g ound-based planning and inc easing
he scien i ic e u n o missions [12].
2.1.5 Challenges and Fu u e Di ec ions
Despi e hese ad ancemen s, mission scheduling emains a challenging domain. The
dynamic na u e o ask a i als, limi ed spacec a esou ces, and unce ain ies in
en i onmen al condi ions demand inno a i e solu ions. Hyb id models ha in eg a e
op imiza ion algo i hms wi h onboa d p ocessing a e being explo ed o add ess hese
challenges. Addi ionally, he adop ion o dis ibu ed scheduling app oaches o mul i-
sa elli e cons ella ions is expec ed o imp o e coo dina ion and esou ce sha ing ac oss
missions [16].
In conclusion, mission scheduling has e ol ed om manual p ocesses o highly
au oma ed sys ems capable o add essing complex cons ain s and objec i es. Tools like
CLASP, CASPER, SciBox, and Eagle Eye demons a e he ans o ma i e impac o
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
8
ad anced scheduling echnologies on space explo a ion. As missions become
inc easingly complex, hese ools will be indispensable in ensu ing he success and
e iciency o u u e endea o s.
2.2 The JUICE mission
2.2.1 Main scien i ic objec i es
The JUICE mission, led by ESA, is dedica ed o explo ing he Jupi e sys em wi h a
pa icula ocus on i s icy moons: Ganymede, Eu opa, and Callis o. These moons a e
belie ed o ha bo subsu ace oceans, making hem p ime candida es o s udying po en ial
habi abili y beyond Ea h (ei he pas o p esen ) [17].
Figu e 2.1: JUICE spacec a [19]
Launched in Ap il 14 h 2023, JUICE’s scien i ic objec i es include:
1. Cha ac e izing Ganymede: his in ol es mapping i s magne ic ield, su ace
ea u es, and subsu ace ocean o unde s and i s po en ial as a habi able
en i onmen . Scien is s a e pa icula ly in e es ed in Ganymede, he only na u al
sa elli e in he Sola Sys em known o ha e an in ensi e magne ic ield. This
unique cha ac e is ic makes i a na u al labo a o y o s udying magne osphe ic
phenomena and plasma in e ac ions wi h Jupi e [18].
2. Eu opa’s Su ace Chemis y: in es iga ing Eu opa's su ace composi ion and
sea ching o e idence o biosigna u es.
3. Compa a i e Analysis: s udying he simila i ies and di e ences among he
Galilean moons, ocusing on hei geology and in e ac ions wi h Jupi e 's
magne osphe e
9
2.2.2 Mission p o ile
The Jupi e Icy Moons Explo e (JUICE) ollows a me iculously designed mission p o ile
aimed a maximizing i s scien i ic e u n while na iga ing he complexi ies o in e plane a y
a el. Launched on Ap il 14, 2023, om Kou ou, F ench Guiana, aboa d an A iane 5, JUICE
emba ks on an eigh -yea jou ney ha le e ages mul iple g a i y-assis maneu e s o each
Jupi e in July 2031 [20].
The key mission phases a e:
1. Launch and Ea ly Ope a ions Phase (LEOP)
Following i s success ul launch, JUICE en e ed i s ea ly ope a ions phase, du ing
which he spacec a 's heal h was e i ied, and c i ical sys ems we e
commissioned. Wi h i s 85 squa e me e s o sola panels ully deployed, JUICE
began gene a ing he powe needed o i s long jou ney, e en in he dim
en i onmen o he ou e Sola Sys em.
2. In e plane a y C uise (2023-2031)
Du ing his eigh -yea phase, JUICE pe o ms ou g a i y-assis maneu e s o
conse e uel and adjus i s ajec o y. These include:
o Luna -Ea h Flyby #1 (Augus 2024): P o ides he i s eloci y boos
h ough a combined Ea h-Moon g a i y assis .
o Venus Flyby (Augus 2025): The closes app oach a 9500 km p o ides a
majo ajec o y co ec ion
o Ea h Flyby #2 (Sep embe 2026): A mid-cou se adjus men o ine- une
JUICE's ajec o y.
o Ea h Flyby #3 (Janua y 2029): The inal lyby deli e s a c i ical boos ,
inc easing JUICE's eloci y o he escape eloci y equi ed o i s jou ney o
Jupi e .
3. Jupi e A i al and O bi Inse ion (July 2031)
Upon a i al a Jupi e , JUICE will execu e a p ecise Jupi e O bi Inse ion (JOI)
maneu e , educing i s eloci y by app oxima ely 900 m/s an es ablishing an ini ial
o bi wi h a pe iapsis a 200,000 km om Jupi e ’s cen e.
4. Jupi e Sys em Tou (2031-2034)
JUICE will unde ake a se ies o lybys o he Jo ian moons o collec c i ical da a:
o Eu opa Flybys (July 2032): Two close lybys, a al i udes o app oxima ely
400 km.
o Callis o Flybys (2032-2034): Mul iple lybys, which will also adjus JUICE’s
o bi al inclina ion o 29° o pola obse a ions o Jupi e .
A sequence o Ganymede and Callis o lybys will adjus JUICE’s o bi so ha i can
en e o bi a ound Ganymede in Decembe 2024.
5. Ganymede O bi al Phase (Decembe 2034 - End o Mission)
The mission’s highligh will be JUICE’s ansi ion o o bi Ganymede, he la ges
moon in he Sola Sys em, and he i s ime a spacec a will o bi a moon o
ano he plane .
6. End o Mission (2035)
A e comple ing i s scien i ic objec i es, JUICE will execu e a con olled descen
on o Ganymede’s su ace, ensu ing compliance wi h plane a y p o ec ion p o ocols.
This inal maneu e ma ks he end o o e a decade o g oundb eaking explo a ion
and da a collec ion.

Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
10
Table 2.1: JUICE's main mission phases
Phase De ail Da e
Launch and LEOP Success ul launch aboa d
A iane 5 om Kou ou,
Guiana. Ini ial checks and
sola panel deploymen .
Ap il 14, 2023
In e plane a y C uise Eigh -yea phase wi h
g a i y assis s o adjus
ajec o y:
2023-2031
Luna -Ea h Flyby #1: Fi s
eloci y boos . Augus 2024
Venus Flyby: Closes
app oach a 9500 km o
ajec o y co ec ion.
Augus 2025
Ea h Flyby #2: Mid-
cou se adjus men .
Sep embe 2026
Ea h Flyby #3: Final boos
o escape eloci y owa d
Jupi e .
Janua y 2029
Jupi e a i al Jupi e O bi Inse ion
(JOI), 900 m/s
decele a ion, pe iapsis a
200,000 km.
July 2031
Jupi e Sys em Tou Flybys o Eu opa (400 km
al i ude in July 2032) and
Callis o (mul iple lybys o
29° inclina ion).
2031-2034
Ganymede O bi al Phase O bi al inse ion a ound
Ganymede.
Decembe 2034-2035
End o Mission Con olled descen on o
Ganymede’s su ace.
2035
11
Figu e 2.2: JUICE's ajec o y [21]
2.2.3 JANUS
As explained in he Sec ion 1, his hesis will ocus solely on he scheduling o a single
ins umen . Rega ding he eal-wo ld case, he JANUS ins umen has been selec ed.
JANUS is an op ical came a sys em onboa d JUICE designed o pe o m high- esolu ion
imaging o Jupi e ’s moons in he isible and nea -in a ed spec um. I is a elescope
coupled wi h a CMOS aming de ec o . I s p ima y scien i ic goals include s udying
geological s uc u es, ec onic p ocesses, c yo olcanism, and su ace composi ion, as well
as de e mining he ela i e ages o he Galilean moons [22][23].
Figu e 2.3: JANUS [24]
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
12
The ins umen has been speci ically designed o ope a e in he ha sh adia ion en i onmen
o he Jo ian sys em. I suppo s bo h nadi -poin ing and il ed obse a ion modes, wi h
a spa ial esolu ion anging om 400 m/pixel o less han 10 m/pixel, depending on he
dis ance o he obse ed a ge [22].
JANUS’s key ea u es a e:
Table 2.2: JANUS's key ea u es
De ec o Fo ma 2000 pixel × 1504 pixel
Field O View (FOV) 1.72° × 1.79°
Spec al Co e age Visible o nea -in a ed
wa eleng hs (13 il e s)
JANUS is equipped wi h 13 spec al il e s, allowing i o s udy he chemical composi ion
o Jupi e ’s moons in de ail. The sys em enables s e eo imaging o c ea e 3D models o he
su aces [25].
13
3 Me hodology
This chap e aims o p o ide a mo e de ailed explana ion o he concep s and ools equi ed
o add ess he mission scheduling p oblem. The code desc ibed in Chap e 4E o !
Re e ence sou ce no ound., ha was de eloped du ing he hesis wo k, is di ided in o
wo pa s: in he i s pa , he Online F on ie Repai algo i hm calcula es he da a ha will
be inpu o he gene ic algo i hm, which will gi e he op imal o nea -op imal schedule as
ou pu .
Fi s , wha he Online F on ie Repai algo i hm does will be desc ibed, hen he ocus will
be on he gene ic algo i hm, unde s anding why i is sui able o he mission scheduling
p oblem and jus i ying why he algo i hm’s pa ame e s we e aken om To es’s [2]
benchma king.
3.1 Mosaic heu is ics
3.1.1 A ea Co e age Planning P oblem (ACPP)
P io o implemen ing a comp ehensi e obse a ion plan o all ROIs, i is necessa y o
ini ially add ess he obse a ion o a single ROI u ilizing he payload ins umen s. In
pa icula , he ocus will be on aming ins umen s, i.e. op ical came as, as JANUS.
The appa en size o he ROI o en exceeds he ield o iew (FOV) o he came a onboa d
he spacec a . This disc epancy necessi a es he implemen a ion o sophis ica ed a ea
co e age planning algo i ms o ensu e comple e and e icien obse a ion o he ROI.
Among hese, mosaic heu is ics play a pi o al ole, pa icula ly in adap ing o he dynamic
cons ain s o spacec a mo ion and he e ol ing oo p in o onboa d ins umen s.
Mosaic heu is ics b eak he ROI in o smalle , ixed-size iles, c ea ing a uni o m g id. The
spacec a ’s came a hen a ge s he cen oid o each ile sequen ially, ge ing a mosaic as
inal esul , i.e., a composi e pic u e c ea ed by me ging mul iple indi idual images. This
sys ema ic app oach allows o e icien co e age while managing he cons ain s o o e lap
and esolu ion. The o e lap is c ucial o main ain con inui y and alignmen bew een adjacen
images, ensu ing high scien i ic alue o he esul ing mosaic. Con e sely, excessi e
o e lap can esul in was ed ope a ional esou ces and inc eased mission cos s.
Gi en a spacec a ’s ajec o y, ins umen obse a ion geome y and equi emen s, and a
designa ed ROI, he A ea Co e age Planning P oblem (ACPP) consis s o de e mining he
obse a ion pa h ha maximizes he ROI co e age while ensu ing a easonable mission
du a ion; speci ically [7], being P he ROI on a a ge body’s su ace and
[ 0, ]
a ime in e al
(in his p oblem, i is one in e al wi hin he ROI is isible om he spacec a ) he objec i e
is o iden i y a se o oo p in s whose union will collec i ely co e P:
𝑃
⊆
󰇌
𝑓



(3.1)
whe e 𝑓 is he ins an aneous ins umen oop in and N is he o al numbe o oo p in s in
he plan.
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
20
which expands he sea ch o unexplo ed a eas, while selec ion and c osso e end o
p omo e exploi a ion, which e ines exis ing p omising solu ion.
 Explo a ion is d i en by mu a ion, as i in oduces new gene ic ma e ial and p e en s
p ema u e con e gence
 Exploi a ion is acili a ed by selec ion and c osso e as hey enhance and combine
he ai s o well-adap ed indi iduals o ocus on he bes -pe o ming egions o he
sea ch space
The selec ion ope a o is he mechanism by which indi iduals in he popula ion a e chosen
o con ibu e o he nex gene a ion. Inspi ed by na u al selec ion, his ope a o a o s
indi iduals wi h highe i ness sco es, ensu ing ha bene icial ai s a e p ese ed and
ampli ied, while ensu ing di e si y is main ained o a oid p ema u e con e gence.
Be e -quali y selec ion s a egies encou age exploi a ion o high- i ness indi iduals in he
popula ion, wi hou losing di e si y in he popula ion oo quickly. A wide a ie y o selec ion
s a egies ha e been designed o he GA, among which he i ness-based selec ion is one
o he mos common app oaches: o en called oule e wheel selec ion o p opo ional
selec ion, i assigns selec ion p obabili ies based on an indi idual’s i ness sco e ela i e o
he popula ion. Indi idual wi h highe i ness ha e a g ea e chance o being selec ed. This
me hod can be biased owa ds dominan indi iduals, po en ially educing gene ic di e si y
ea ly in he sea ch, which can be mi iga ed by i ness scaling o using al e na i e selec ion
me hods.
O he common selec ion s a egies a ea ou namen selec ion [30], ank-based selec ion and
andom selec ion [29].
Mu a ion in oduces andom changes o indi iduals’ gene ic ep esen a ions, imi a ing he
andom e o s o al e a ions ha occu in DNA eplica ion. This ope a o is c ucial o
explo a ion, as i p e en s he popula ion om s agna ing a local op ima and ensu es
con inuous disco e y o new a eas in he solu ion space. I basically al e s he genes o
o sp ing by in oducing andom changes; a ious mu a ion ope a o s, such as bi - lipping,
andom, di ec ed o adap i e mu a ion a e ailo ed o add ess speci ic challenges and
imp o e pe o mance by balancing explo a ion and exploi a ion .
C osso e combines he gene ic ma e ial o wo o mo e pa en solu ions o c ea e o sp ing,
emula ing gene ic ecombina ion obse ed in biological ep oduc ion. Techniques like
single-poin c osso e (spli ing pa en genomes a one poin ) and uni o m c osso e
(swapping genes andomly) de e mine how gene ic ma e ial is exchanged. The
e ec i eness o c osso e depends hea ily on he solu ion ep esen a ion and he
compa ibili y o pa en al geno ypes. When pai ed wi h a well-chosen selec ion s a egy,
c osso e ac s as a powe ul mechanism o e ining solu ions and d i ing con e gence.

21
4 The Scheduling P oblem
The code explained in his chap e , which cons i u es he cen al pa o his hesis, is
a ailable on Gi Hub unde an open-sou ce license [32]. To summa ize, gi en a speci ic se
o Regions o In e es (ROIs) on Ganymede and he ime in e als o JUICE lybys a ound
Ganymede, he aim is o ind he op imal obse a ion schedule.
The code is di ided in o wo sequen ial pa s:
 Science Oppo uni y Analysis (p ecompu a ion): gi en he inpu da a ela ed o
he mission, such as lyby ime in e als, he ROIs, he spacec a and he celes ial
body o in e es , he ime in e als wi hin hese lybys du ing which he ROIs can be
obse ed (acco ding o some geome ical cons ain s), as well as he quali y o
hese obse a ions (speci ically, esolu ion and co e age), a e calcula ed. These
da a will be hen used by he Gene ic Algo i hm.
 Gene ic Algo i hm: s a ing om he da a compu ed in he SOA (Science
Oppo uni y Analysis), a gene ic algo i hm is execu ed, which gene a es
popula ions o indi iduals o e successi e gene a ions. Each indi idual ep esen s
a schedule and is cha ac e ized by a unique i ness alue aking in o accoun
esolu ion and co e age. The algo i hm aims o ind he minimum i ness alue and
e mina es a e a speci ied numbe o i e a ions o when he i ness alls below a
ce ain h eshold.
I is wo h no ing ha unning he gene ic algo i hm does no necessi a e execu ing he
Science Oppo uni y Analysis (SOA) e e y ime. The da a a e sa ed in iles wi hin he
eposi o y and he necessa y in o ma ion is subsequen ly loaded by he objec s in he
second pa o he code whene e he algo i hm is un.
4.1 The Science Oppo uni y Analysis
The sc ip JUICE_Obse abili y.py [32] ollows he main logic explained in his pa ag aph and
is he one ha mus be un o pe o m he ope a ions desc ibed he ein. Gi en he
compu a ionally in ensi e na u e o he calcula ions pe o med in his sc ip , i was execu ed
on he JUNO supe compu ing clus e a ESEIAAT, Uni e si a Poli ècnica de Ca alunya [33]:
i is a high-pe o mance compu ing clus e , enabling ad anced simula ions and da a
p ocessing o esea ch in mechanical and ae ospace enginee ing.
4.1.1 P oblem inpu s
The celes ial body on whose su ace he ROIs o be obse ed a e loca ed is speci ied as
GANYMEDE. Fu he mo e, i is assumed ha he spacec a is JUICE.
The oi_in o olde (in da a) con ains a ex ile, ganymede_ oi_in o. x , speci ying he de ails
o he ROIs o be obse ed. Speci ically, each line in he ile will con ain:
 ROI_key: an iden i ica ion o each ROI, de ined by ESA. Fo ins ance, conside ing
JUICE_ROI_GAN_1_0_14, GAN is he moon whe e he ROI is loca ed, while he
nume ical code X_X_XX deno es he ype o egion;
 ROI_name;
 ROI_la i udes: he la i udinal coo dina es o he ou e ices o each ROI a e
p o ided. The ROIs ha e a squa e o ec angula shape.
 ROI_longi udes: he longi udinal coo dina es o he ou e ices
The o icial da es o JUICE’s lybys a ound Ganymede ha e al eady been inse ed, as
ob ained om [33]. These ime in e als, which cons i u e he ini ial sea ch space ime ame,
do no accoun o any spacec a a i ude cons ain s o geome ic limi a ions. By imposing
hese, i will be possible o disca d any ime in e als du ing he lybys whe e, despi e
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
22
po en ial science oppo uni ies due o good esolu ion, he spacec a canno ake
ad an age o hem due o ac o s such as excessi e o insu icien illumina ion, he need o
ex eme il ing o he spacec a , and so on. The use selec s which cons ain s apply o he
p oblem, such as he accep able ange o alues o bo h, he eby ob aining mul iple
complian ime windows – one pe ROI – s o ed in a Py hon lis . The ime windows a e
ep esen ed as objec s o class SPICEDOUBLE_CELL om he SPICE’s lib a y.
Table 4.1: O icial s a and end imes o JUICE lybys a ound Ganymede aimed a JANUS
expe imen s [33]
Flyby # S a da e End da e
0 2033 NOV 26 18:22:11 2033 NOV 27 18:22:11
1 2034 JAN 14 06:38:51 2034 JAN 15 06:38:51
2 2034 JUN 05 18:53:51 2034 JUN 06 18:53:51
3 2034 JUL 11 19:50:31 2034 JUL 12 19:50:31
4 2034 SEP 07 06:03:51 2034 SEP 08 06:03:51
5 2034 SEP 28 18:48:51 2034 SEP 29 18:48:51
6 2034 NOV 18 09:58:51 2034 NOV 19 09:58:51
4.1.2 Logical low and classes
A ke nelFe ch class objec downloads he mission- ela ed in o ma ion in he o m o SPICE
ke nels. A SPICE ke nel is a da a ile whe e in o ma ion abou he posi ion and o ien a ion
o spacec a , as well as he posi ion, o ien a ion, shape, and size o sola sys em bodies,
is s o ed. These da a a e essen ial o he execu ion o he code, o example, when using
he unc ions o he SPICE lib a y o calcula e imes and dis ances.
Nex , a ROIDa aBase class objec is ins an ia ed. I akes as inpu he pa h o he ex ile
con aining he ROIs’ da a and he name o he a ge body. This objec cleans and s o es
he ROIs in o ma ion as a lis , named _ROIs, o Py hon dic iona ies (a Py hon dic iona y
s o es da a in key- alue pai s). Each dic iona y, ela i e o one ROI, has 6 keys and is
s uc u ed as ollows:
 # oi_key: he s ing wi h he ROI’s iden i ica ion
 body: he s ing speci ying he celes ial body
 la and lon: wo ec o s con aining la i udinal and longi udinal coo dina es o he
ROI’s e ices
 oi_name: he s ing o he ROI’s name
 e ices: e ices’ coo dina es a anged as an nda ay
This class also keeps a sepa a e lis o dic iona ies called indices. These dic iona ies s o e
he posi ion o each ROI wi hin he _ROIs lis . This app oach allows o quicke e ie al o
ROI de ails om ou side he class, a oiding he need o scan h ough he en i e lis . A block
diag am ep esen ing he class is shown (Figu e 4.1).
23
Figu e 4.1: Block diag am o ROIDa aBase class
Fo each ROI, he checkOneROI.py sc ip is called, which iden i ies he oppo uni y windows
o each ROI. Speci ically, he ollowing cons ain s mus be me o ge he complian ime
windows wi hin he ini ial sea ch space.
Table 4.2: Angle pa ame e s and ela i e cons ain s o ge he complian ime window
Pa ame e min max
Emission angle 0° 75°
Illumina ion-zeni h angle 0° 180°
Phase angle 0° 180°
The unc ions ha calcula e he alues o hese angles we e w i en by Paula Be iu and
ansla ed om MATLAB o Py hon by Jo ge Simon, whose hesis wo k ocuses on he same
opic; hey can be ound in he Gi Hub eposi o y Py hon Science Oppo uni y Analysis Tool
(PSOA) [35]. An explana ion and pseudo-code o each o hese unc ions a e p esen ed in
he ollowing pages, while he angles a e shown in Figu e 4.2.
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
24
Figu e 4.2: Obse a ion geome y [7]
The emissionang unc ion e u ns he phase angle be ween he ROI’s su ace no mal in i s
cen oid and he ec o joining he obse e and he ROI’s cen oid.
Algo i hm 2: emissionang
Inpu : su ace poin coo dina es, ET, a ge , obse e
Ou pu : o -nadi angle in deg ees
1: a ge ame ← Re u ns he a ge ame ID as known by SPICE;
2: obs ec ← Ge s he obse e posi ion ela i e o he s udied su ace poin wi hin he
a ge s ame using gobs ec
3: i s poin is as ec o wi h dimension 2 hen
4: s poin ← T ans o m om deg ees o adians
5: s poin ← Con e om plane ocen ic coo dina es o ec angula coo dina es
using SPICE’s s ec
6: end i
7: i con ains mo e han one imes ep hen
8: n m ec ← Ini ialize a no mal ec o o each imes ep: size( ) × 3 ma ix
9: o i ← 0 o leng h( ) do
10: n m ec[i] ←
Compu e he su ace no mal wi hin he a ge ame using SPICE’s
S n m
11: end o
12: angle ← Ini ialize he angle o each imes ep
13: o i ← 0 o leng h( ) do
14: angle[i] ← Compu e he emission angle using SPICE’s sep and ans o m
o deg ees
15: end o
16: else
17: n m ec ← Compu e he su ace no mal wi hin he a ge ame using SPICE’s
s n m, jus one ime
18: angle ← Compu e he emission angle using SPICE’s sep and ans o m o
deg ees, jus one ime
19: end i
20:
e u n
angle
25
The illzeni hang calcula es he angle o med by he ec o om he illumina ion sou ce o
he cen oid o he ROI and he su ace no mal a ha poin . This can be done by using he
same logic behind he emission angle sub ou ine.
Algo i hm 3: illzeni hang
Inpu : su ace poin coo dina es, ET, a ge
Ou pu : illumina ion-su ace no mal angle
1: obse e ← ‘SUN’
2: angle ← emissionang (s poin , , a ge , obse e )
3:
e u n
angle
The phaseang unc ion compu es he angle be ween he di ec ion o he Sun and he
di ec ion o he obse e as seen om he ROI’s cen oid.
Algo i hm 4: phaseang
Inpu : su ace poin coo dina es, ET, a ge , obse e
Ou pu : phase angle
1: i s poin is as ec o wi h dimension 2 hen
2: s poin ← T ans o m om deg ees o adians
3: s poin ← Con e om plane ocen ic coo dina es o ec angula coo dina es
using SPICE’s s ec
4: end i
5: obs ec ← Compu e he obse e ’s posi ion ec o e e ed o he su ace poin using
SPICE’s gobs ec sc ip
6: i con ains mo e hen one imes ep hen
7: angle ← Ini ialize an angle o each imes ep
8: o i ← 0 o leng h(ET) do
9: angle[i] ← Compu e he phase angle using SPICE’s sep ou ine
10: end o
11: else
12: angle ← Compu e he phase angle using SPICE’s sep ou ine, jus once.
13: end i
1
4
:
e u n
angle
The in o ma ion abou he cons ain s is p o ided as inpu o he my w inde lis unc ion,
con ained in he pySPICElib lib a y de eloped by Manel So ia [36]. This unc ion applies
Bolzano’s heo em o ind he complian ime window in he ini ial sea ch space. Each ime
window o a gi en ROI can ha e none, one, o mul iple in e als; u he mo e, he ini ial
in e al(s) o he uncons ained ime window a e di ided in o smalle segmen s. As a esul ,
he gene al sea ch space is no longe u ilized. The new ime windows consis o a subse
o n in e als de i ed om he o iginal sea ch space.
An objec o he Ins umen class is de ined, wi h he ollowing a ibu es se h ough he
cons uc o me hod __ini __:
 i o
- De ini ion: deno es he Ins an aneous FOV
- Da a ype: loa

Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
26
 npix
- De ini ion: indica es he numbe o pixels in he came a’s senso a ay. I has been
assumed ha he image senso has a squa e shape de i ed om he squa e oo o
he eal a ea o JANUS
- Da a ype: in
 imageRa e
- De ini ion: indica es he numbe o images he came a cap u es pe second
- Da a ype: loa
 s
- De ini ion: p o ides a ma gin o sage y o accomoda e po en ial unce ain ies
- Da a ype: loa
In his objec , he cha ac e is ics o JANUS ha e been included, as i is he ins umen used
in he cu en p oblem:
Table 4.3: Values o he Ins umen class objec ini ialized o JANUS
i o 15
×
10-6 ad
npix 1735 pixels
imageRa e 0.1 s-1
s 20
The ROIDa aBase class objec con ains a me hod, called ge ROIs, which ou pu s one o
mo e objec s o he oplanROI class, depending on he inpu . In each checkOneROI.py sc ip ,
which is called o each ROI, he me hod is in oked by p o iding as inpu he s ing
con aining he name o he speci ic ROI, he eby gene a ing an objec o he OplanROI class.
The oPlanRoi class is no only used o consolida e all he da a ega ding he ROI collec ed
so a bu also includes a ibu es and me hods ha allow he calcula ion o he i ness o an
obse a ion, so, h ough his, co e age and esolu ion a e compu ed.
The oPlanRoi class objec con ains he me hod ini ializeObse a ionDa aBase, which sa es
in o ma ion abou he complian ime windows o he ROI, e alua es and sa es co e age,
esolu ion and du a ion o obse a ions wi hin hese complian ime windows using he
Online F on ie Repai algo i hm. The oPlanRoi class is explained in g ea e de ail in he
Sec ion 4.1.3.
These da a a e also sa ed in wo se s o iles (loca ed in da a/ oi_ iles) wi hin he eposi o y:
o a gi en ROI, he e will be as many . x iles as he e a e lybys du ing which i is
obse able, and one .c g ile o all he lybys in which i is obse able. The in o ma ion om
he .c g iles is hen used o ini ialize he oPlanRoi class objec s in he gene ic algo i hm.
4.1.3 The oPlanRoi class and he Online F on ie Repai implemen a ion
The oPlanRoi class calcula es and sa es da a h ough i s me hods in o i s a ibu es, which
a e all de ined in he cons uc o me hod __ini __. I is s uc u ed as ollows:
 A ibu e/ROI_TW
- De ini ion: Va iable whe e he complian ime window o he selec ed ROI is sa ed.
- Da a ype: Spice Time Window (i.e., SPICEDOUBLE_CELL).
27
 A ibu e/ROI_ObsET
- De ini ion: F om he complian ime window con ained in he ROI_TW a ibu e,
he in e als a e ex ac ed and s o ed in his a iable. Each in e al is disc e ized
in o a housand poin s.
- Da a ype: A ay lis .
 A ibu e(s)/ROI_ObsLen, ROI_ObsImg, ROI_ObsRes, ROI_ObsCo
- De ini ion: Va iables ha sa e, espec i ely, he du a ion, he numbe o images,
he a e age esolu ion and he co e age o he obse a ions ha s a a he ime
ins an s con ained in he a ibu e ObsET_.
- Da a ype: A ay lis
 A ibu e/ROI.mosaic
- De ini ion: Va iable ha de e mines whe he he me hod o calcula ing he alues
o obse a ion du a ion, a e age esolu ion and co e age should be he Online
F on ie Repai o an app oxima ed me hod de eloped by Pablo To es [2]. In his
ex , he use o Online F on ie Repai has been s udied.
- Da a ype: Boolean
 Me hod/ini ializeObse a ionDa aBase
- De ini ion: This unc ion assigns da a o he class a ibu es and makes calls o
o he me hods as necessa y. I hese da a a e p o ided as inpu , he unc ion sa es
hem di ec ly; o he wise, i is necessa y o i s calcula e hem be o e hey a e
inse ed. Addi ionally, in he inpu s, i is possible o speci y he boolean alue o
ROI.mosaic.
 Me hod/compu eObse a ionET
- De ini ion: This me hod disc e izes he in e als wi hin he ime windows o he
ROI’s TW, c ea ing a linea space o 1000 poin s be ween he s a and end imes o
he in e al.
 Me hod/compu eObse a ionDa a
- De ini ion: This me hod calcula es he leng h, he numbe o images, he a e age
esolu ion and he co e age o he obse a ions, aking each ime in ROI_ObsET as
he s a ing ime o he obse a ion.
 Me hod/in e pola eObse a ionDa a
De ini ion: Gi en a ime wi hin a alid ime window, his me hod compu es he
leng h, numbe o images, a e age esolu ion and co e age o an obse a ion a
ime using ROI_ObsET, ROI_ObsLen, ROI_ObsImg, ROI_ObsRes and
ROI_ObsCo . I he gi en inpu ime alue does no coincide wi h one o he
housand poin s o disc e iza ion, he desi ed alue will be ound h ough
in e pola ion. Plo ing he alues ob ained om he p ecompu a ion e eals no
anomalies o i egula ends; he e o e, in e pola ion can be used as an app op ia e
me hod o model he unc ion.
In he compu eObse a ionDa a me hod, called by ini ializeObse a ionDa aBase, he
alues o obse a ion du a ion, numbe o images, a e age esolu ion and co e age
s a ing a ime ins an a e p e-compu ed. In he case unde conside a ion, wi hin a o loop
i e a ing o e all he complian in e als o he ROI, he mosaicOnlineF on ie unc ion is
called. This unc ion ecei es as inpu :
o complian In e al: a ay o 1000 poin s in which he complian in e al has been
disc e ized.
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
28
o ins : s ing speci ying he ins umen
o sc: s ing speci ycing he spacec a ( he obse e )
o ROIob: cu en oPlanRoi class objec
o ins umen : Ins umen class objec
o in : in ege numbe speci ycing he numbe o he in e al in he ROI’s complian TW
The mosaicOnlineF on ie unc ion pe o ms he mosaicing algo i hm using he
on ie Repai unc ion o each o he housand poin s ha cons i u e he ime in e al o he
ROI.
Focusing on he on ie Repai unc ion ( ixed obse a ion’s s a ing ins an and ROI) i is
assumed ha he obse a ion o a g id poin las s 8.5 seconds and ha he slew a e ( he
a e a which he spacec a ’s o he ins umen pla o m’s poin ing di ec ion slews be ween
g id poin s, in [º/s]) is 3 × 10⁻³ s.
A he end o he p ocess, a lis called pLis is gene a ed. This is a Py hon lis o dic iona ies,
whe e each dic iona y con ains he in o ma ion ela ed o a single oo p in . The union o all
he oo p in s will be used o c ea e he comple e image o he ROI (Figu e 4.3).
Figu e 4.3: Example o Online F on ie Repai algo i hm wi h Galileo Regio
(JUICE_ROI_GAN_5_0_09), obse a ion s a ed on 2034 JAN 14 18:22:31 UTC
In pa icula , he pLis will no be emp y since he ROI is obse able, espec ing he
p e iously imposed cons ain s.
mosaicOnlineF on ie ou pu s he lis s makespan, nimg, esROI and co . They a e speci ic
o each in e al, o his eason, each will con ain 1000 alues; i mus be emphasized ha
he in o ma ion used below (such as he s a ime o he obse a ion o he i s poin in he
g id) o calcula e he alues o he ollowing lis s is aken o m he pLis co esponding o
he speci ic obse a ion, bea ing in mind ha each pLis co esponds o one o he 1000
obse a ions.
 makespan: Con ains all he obse a ion du a ions. Each du a ion is ob ained by
adding he obse a ion ime o he s a ing ins an o he las g id poin ’s obse a ion
and sub ac ing he s a ing obse a ion ins an o he i s g id poin .
29
 nimg: The numbe s o images cap u ed du ing he obse a ion and is simply equal
o he numbe o dic iona ies in pLis (i.e., he numbe o oo p in s).
 esROI: A e age esolu ions o each obse a ion. Speci ically, he a e age
esolu ion is calcula ed by calling he compu eResMosaic unc ion, which akes
pLis and ins umen .i o as inpu s, hen calls he poin es unc ion [35] o compu e
he esolu ion using he ollowing o mula:
𝑅
=
𝑖𝑓𝑜𝑣
×
𝑑

sin
(
90°
−
𝜃
)
(4.1)
This o mula is speci ic o each g id poin . The i o is gi en as inpu , d ep esen s
he dis ance om he g id poin and 𝜃 is he emission angle calcula ed by he
unc ion explained in he Algo i hm 2. Addi ionally, he e is a condi ion on he
maximum esolu ion o a oid singula i ies (in cases whe e 𝜃 is equal o 90°).
Subsequen ly, in compu eResMosaic, he a i hme ic mean o all he esolu ions o
he obse a ion is compu ed, esul ing in he a e age esolu ion o he obse a ion.
 co : The co e ages (in pe cen age) o he ROI when obse ed om each ins an
wi hin he in e al, whe e co e age e e s o how much o he ROI’s a ea is
e ec i ely co e ed by all he oo p in s. The co e age is calcula ed using he
oico e age unc ion ( om he olde a ea_co e age_planning_py hon), which akes
as inpu he name o he a ge celes ial body, he e ices o he ROI and he ela i e
pLis .
Speci ically, a h eshold has been se : i he co e age is g ea e han 95%, i is
au oma ically se o 100%.
A e comple ing he p e-compu a ion, ha is, a e unning he JUICE_Obse abili y.py
sc ip , i is possible o ind he minimum esolu ion alues o each ROI in each lyby
(p ecisely, wi hin he lyby in e als whe e he ROI is isible, i isible); hese a e shown in
Figu e 4.4 and Figu e 4.5.
Mo eo e , i is possible o display 7 g aphs ep esen ing he spacec a ’s g ound ack and
he isible ROIs o he 7 lybys. Conside ing ha each ROI has minimum esolu ion alues
o each lyby in which i is isible, as shown in he p e iously desc ibed hea map ma ix, i
is possible o speci y in each o he 7 g aphs which ROIs, among hose isible du ing ha
lyby, can be obse ed wi h he minimum esolu ion among all he minimum esolu ion
alues o he lybys in which hey a e isible. In o he wo ds, i is possible o indica e
whe he ha lyby con ains he bes possible momen o obse e ha ROI: in his case, he
ROI polygon is g een, o he wise i is ligh b own; he g ound ack is ed.
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
36
- Da a ype: Lis o 2 in ege numbe s
The oplan class con ains he ollowing me hods:
 Me hod(s)/ge ObsLeng h, ge ObsNumImg:
- De ini ion: These unc ions ake as inpu a ROI (oPlanRoi objec ) and a speci ic
ime ins an , e . A e iden i ying he ime in e al o he complian TW whe e he
ins an e is loca ed (by calling he me hod indIn e alInTw), hese unc ions ou pu ,
espec i ely, he du a ion and he numbe o images o he obse a ion o his ROI
s a ing om e . In his speci ic case, since he mosaic a ibu e is se o T ue, hese
alues a e ob ained by in e pola ing wi h he housand du a ion and image numbe
alues compu ed o he ele an in e al using he Online F on ie Repai .
 Me hod/ anFun:
- De ini ion: Gi en he oiL lis , i andomly assigns a s a ime o he obse a ion
o each ROI in he lis and he du a ion o he obse a ion o ha ROI s a ing om
he assigned ime. The s a ime and du a ion alues o each ROI a e p o ided by
he uni o mRandomInTw me hod and a e s o ed in he a ibu es s ol and obsLeng h.
 Me hod/uni o mRandomInTw:
- De ini ion: Gi en a ROI, his me hod andomly selec s one o he in e als wi hin
he complian ime window, wi h he p obabili y o selec ing an in e al equal o:
𝑃

=
𝛿

∆

(4.2)
whe e, wi h 𝑡, and 𝑡, being he s a and end imes o he i- h in e al and N he
o al numbe o in e als in he TW:
𝛿

=
𝑡

,

−
𝑡

,

(4.3)
∆

=

𝛿



(4.4)
In his way, a highe p obabili y is assigned o la ge in e als.
Wi hin he selec ed in e al, he me hod andomly ex ac s a poin ollowing a
uni o m dis ibu ion, ensu ing ha i he obse a ion begins a his poin , i concludes
wi hin he selec ed in e al. I his condi ion is no sa is ied, ano he poin in he
in e al ha mee s he equi emen is sea ched.
 Me hod/mu Fun:
- De ini ion: This me hod pe o ms he mu a ion, implemen ing a sligh modi ica ion
o each obse a ion s a ing ins an , which in u n a ec s he co esponding
obse a ion du a ion. The new alues a e gene a ed calling he
andomSmallChangeInTw me hod and a e upda ed in he s ol and obsLeng h
a ibu es.

37
 Me hod/ andomSmallChangeInTw:
- De ini ion: Gi en a ROI and he ini ial obse a ion ime 𝑡 o he ROI, his
me hod iden i ies he in e al o he complian ime window con aining 𝑡. I hen
samples, om he in e al, a new obse a ion ime om a Gaussian dis ibu ion
cen e ed a 𝑡, wi h a s anda d de ia ion 𝜎 equal o he smalle o he wo alues:
𝜎

=
min
(
𝑡

−
𝑡

,
𝑡

−
𝑡

)
(4.5)
Whe e he i s one is he dis ance o he s a o he in e al om 𝑡 and he
second one is he dis ance om 𝑡 o he end o he in e al.
I he new obse a ion lies wi hin he in e al, 𝑡 is upda ed in he co espondan
posi ion o he ROI in s ol. O he wise, i he condi ion is no sa is ied a e 50
a emp s, 𝜎 is hal ed a each subsequen i e a ion, up o a maximum o 500
a emp s. I he condi ion emains unme , he p ocess e mina es, and he execu ion
is hal ed.
 Me hod/ epFun:
- De ini ion: This me hod pe o ms he c osso e and akes as inpu an indi idual,
p1. Le p2, o example, be he objec o class oplan om which he me hod epFun
is called. He e, p1 and p2 ep esen he pa en s. F om hese, an indi idual is
gene a ed such ha a, o each ROI, he ini ial obse a ion ime is se as he midpoin
be ween he ini ial obse a ion imes o he wo pa en s:
𝑡

,



=
𝑡

,


+
𝑡

,


2
(4.6)
I he esul ing ime alls wi hin an in e al o he complian ime window o he gi en
ROI, he obse a ion leng h is compu ed using he ge ObsLeng h me hod.
O he wise, a andom ini ial ins an o he ROI is assigned using he
uni o mRandomInTw me hod and he makespan is e alua ed.
A he end o he p ocess, wo lis s a e ob ained: one o he ini ial imes and one o
he obse a ion leng hs. These lis s a e s o ed in s ol and obsLeng h a ibu es o
he cu en indi idual. Consequen ly, p2, which ini ially ep esen ed a pa en , now
ep esen s he o sp ing.
 Me hod/e alResPlan:
- De ini ion: This me hod calcula es he esolu ions o obse a ions s a ing a he
ins an s speci ied in s ol. Fo each ROI, he unc ion e alResRoi is called o compu e
he esolu ion alue as ou pu . These esolu ions a e a e age alues, as explained
in Sec ion 4.1.3.
 Me hod/e alResRoi:
- De ini ion: I akes as inpu he index associa ed wi h he ROI in he oiL lis and
he ins an a which he obse a ion o ha ROI begins. I e u ns he esolu ion alue
o he obse a ion as ou pu . The esolu ion is de e mined ia in e pola ion using he
in e pola eObse a ionDa a me hod o he ROI. Speci ically, since he mosaic
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
38
a ibu e o he ROI is se o T ue in his p oblem, he in e pola ion is pe o med using
he da a p o ided by he Online F on ie Repai .
 Me hod/e alCo Plan:
- De ini ion: I compu es he co e ages o all obse a ions wi hin he plane, s o ing
he esul ing alues in he c oi a ibu e. To e alua e he co e age o each obse ed
ROI, he e alCo Roi me hod is in oked.
 Me hod/e alCo Roi:
- De ini ion: This me hod akes as inpu he ini ial obse a ion ins an e o a ROI
and he oiL index co esponding o ha ROI. I subsequen ly calls he
in e pola eObse a ionDa a me hod o he oPlanRoi class objec , assessing he
co e age a he gi en epoch by means o in e pola ion.
 Me hod/ge To alO e lapTime:
- De ini ion: I compu es and ou pu s he o al o e lap ime, de ined as he sum o
o e laps be ween consecu i e obse a ions. O e laps in ol ing mo e han wo
obse a ions a e no conside ed.
The me hod i s so s all he imes con ained in s ol. Then, o each ime in his
so ed ec o , s a ing om he second (included) o he las (included), i checks ha
he ime does no occu be o e he end o he p e ious obse a ion, which is
de e mined by adding he du a ion o he p e ious obse a ion o i s s a ime.
Finally, all he du a ions du ing which one obse a ion o e laps wi h he p e ious
one a e summed, esul ing in he o al o e lap ime.
 Me hod/ i Fun:
- De ini ion: The i ness o an indi idual is compu ed. Since he cu en p oblem
in ol es single-objec i e op imiza ion, meaning only one i ness alue is e alua ed
o each indi idual, a single-objec i e unc ion is used. Speci ically, as simul aneous
obse a ions a e no allowed, indi iduals wi h obse a ions o e lapping o a ce ain
du a ion a e excluded.
Fi s he o al o e lap ime is calcula ed by calling he ge To alO e lapTime unc ion:
i he o e lap ime is g ea e han 0, he indi idual is assigned an ex emely high
i ness alue, ensu ing i is no a o ed by he selec ion mechanism, which in his
p oblem a o s indi iduals wi h he lowes i ness alues. O he wise, wi h
𝑅,𝑅,𝑅 ep esen ing, espec i ely, he esolu ion alue o he obse a ion o
he i- h ROI in he schedule, he maximum and he minimum esolu ion alues
possible ac oss all obse a ions o all ROIs, 𝐶
he co e age alue o he obse a ion
o he i- h ROI in he schedule and
n
he numbe o ROIs obse ed in he plan, he
i ness o he indi idual
is gi en by:
𝑓
=
𝑤

∑
𝑅





𝑛
−
𝑅

𝑅

−
𝑅

+
𝑤

󰇧
1
−
1
100
∑
𝐶





𝑛
󰇨
(4.7)
This is a weigh ed a e age ha akes in o accoun esolu ion and co e age o de ine
he single- alue i ness o an indi idual. Rega ding esolu ion, he a e age esolu ion
o he en i e plan is i s calcula ed and i is hen no malized be ween 0 and
1.Simila ly, o he co e age, he a e age co e age is i s compu ed and also
exp essed be ween 0 and 1. 𝐶 is, in ac , exp essed as a pe cen age anging om
0 o 100.
39
Since esolu ion is be e when lowe and co e age is be e when highe , and gi en
ha in his p oblem he i ness is minimized, he complemen o co e age wi h
espec o 1 is used, he eby minimizing i and maximizing co e age.
Bo h esolu ion and co e age alues a e no malized be ween 0 and 1 due o hei
di e en o de s o magni ude. This no maliza ion ensu es ha hei ela i e and no
absolu e a ia ions a e accoun ed o .
In he cu en p oblem, he weigh s w1 and w2 a e se o 0.5, meaning ha esolu ion
and co e age ha e equal impo ance in he inal i ness alue.
4.2.2 aga class
The aga class (“A Gene ic Algo i hm”) pe o ms he ope a ions o he gene ic algo i hm. This
class was en i ely de eloped by Manel So ia [3].
Speci ically, in his p oblem, a e andomly gene a ing an ini ial popula ion o 1000
indi iduals (wi h he anFun me hod o he oplan class), he algo i hm i e a es h ough
successi e popula ions o ng (500) gene a ions. Each popula ion consis s o :
 ne bes indi iduals (wi h he lowes i nesses) om he p e ious popula ion. These
a e he eli es.
 nd indi iduals ob ained h ough ep oduc ion om pa en s in he p e ious popula ion.
These a e he descendan s.
 nm indi iduals ob ained by mu a ing indi iduals om he p e ious popula ion. These
a e he mu an s.
I he combined coun o descendan s, mu an s and eli es does no ma ch he o al
popula ion size, he algo i hm ills any emaining slo s wi h newcome s, which a e andomly
gene a ed indi iduals. Mo eo e , he ollowing pa ame e s a e signi ican :
 nCanMu a e: The numbe o indi iduals eligible o se e as he sou ce o mu an s.
These a e he Mu an candida es.
 nCanP oce a e: The numbe o indi iduals ha a e selec ed as pa en s o
ep oduc ion. These a e he Pa en candida es.
The pa ame e s men ioned abo e can be modi ied by he use . The algo i hm class aga is
p oblem-independen , consequen ly, he class de ining he indi idual – oplan in his case –
mus include a me hod o compu e he i ness. aga is also independen o he unc ion o
gene a ing andom indi iduals, ep oduc ion and mu a ion, which mus also be implemen ed
wi hin he indi idual class.
The selec ion mechanism, on he o he hand, is included wi hin he aga class. The aga class
has an a ibu e, pop, whe e he indi iduals a e s o ed and so ed based on hei i ness
alues, om he lowes o he highes , so he indi iduals wi h he lowes i ness ( he bes
ones) a e a he on .The pop a ibu e is upda ed each ime wi h he popula ion
co esponding o he cu en gene a ion. Conside ing his, he selec ion mechanism is as
ollows:
 The i s ne indi iduals o pop a e p ese ed in he nex gene a ion
 Among he i s nCanP oc ea e indi iduals o pop, nd imes, wo pa en s a e
andomly selec ed o gene a e nd child indi iduals
 Among he i s nCanMu a e indi iduals o pop, nm imes, a andom indi idual is
selec ed o undee go mu a ion
 As jus men ioned, newcome s could be added.
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
40
A he end o he las gene a ion, he bes indi idual will ep esen he bes schedule and i
will be he ou pu o he code.
4.2.3 Conside a ions on benchma king
The goal o benchma king he algo i hm is o es how use ul i is o sol ing complex
p oblems and o de e mine he op imal se o pa ame e s o imp o emen s ha can be made
o achie e be e solu ions.
A benchma king s udy has al eady been conduc ed by Pablo To es [2]; despi e he
modi ica ions be ween he wo codes, he benchma king wo k ha could be pe o med he e
would be iden ical and would p oduce he same esul s. In he p esen wo k, he
combina ion o pa ame e s ha , as shown by his benchma king, yields he bes inal i ness
is chosen.
As w i en in he p e ious sec ion, 1000 indi iduals and 500 gene a ions we e chosen. A
high numbe o indi iduals helps main ain gene ic di e si y, educes he isk o p ema u e
con e gence and imp o es he quali y o he inal solu ion, especially in complex p oblem
domain, while i was obse ed ha a e he chosen numbe o gene a ions he i ness does
no dec ease u he . The ollowing able con ains he pa ame e s chosen based on
benchma king:
Table 4.4: GA pa ame e s' alues based on benchma king
Pa ame e Value Pe cen age o popula ion
ne 50 5%
nd 750 75%
nm 150 15%
nCanMu a e 200 20%
nCanP oc ea e 300 30%
4.2.4 The mul i-agen oolbox
S a ing om he Py hon modules ha come om PMOT [3], he sc ip JUICE_ unMany.py
has been de eloped in he eposi o y, which implemen s mul i- h eading: i is possible o
un he same sc ip mul iple imes in pa allel, c ea ing di e en ins ances o agen s ha
execu e he gene ic algo i hm independen ly om each o he .
The ad an age is ha , om a ce ain gene a ion onwa ds, hese ins ances communica e
wi h each o he , allowing he good indi iduals o p opaga e among he a ious ins ances
and using hese indi iduals o ob ain be e esul s. In he same compu a ional ime, be e
esul s a e achie ed.
The ollowing classes a e used:
 spawnAgen s one se s up and uns he agen s.
 agen Da aSha ing allows he da a sha ing be ween he agen s.
41
5 Resul s and discussion
The cu en chap e p esen s and discusses he esul s ob ained by unning he algo i hm
explained in he p e ious sec ion. Speci ically, he esul s o a simple case, wi h known
solu ion, will i s be shown o e i y he alidi y o he algo i hm. Subsequen ly, he esul s
ob ained o he p oblem o in e es will be p esen ed in bo h single-agen and mul i-agen
scena ios. Finally, hese esul s will be compa ed wi h hose ob ained wi hou he Online
F on ie Repai [2], highligh ing any di e ences and simila i ies.
5.1 Valida ion: GA wi h a single ROI
A i s example o e i y he e ec i eness o he algo i hm jus de eloped should consis o
he ollowing:
 Conside ing he esolu ion alue as he i ness o each indi idual; hus, he objec i e
unc ion is he esolu ion o mula (4.1).
 Selec ing only one o he lybys ha JUICE pe o ms a ound Ganymede. In his case,
he chosen lyby begins on 2034 JUN 05 18:53:51 UTC and ends on 2034 JUN 06
18:53:51 UTC.
 Fo each ROI on Ganymede, he emission angle and dis ance om he ROI’s
cen oid du ing he selec ed lyby can be plo ed.
 A e analyzing he plo s, a ROI isible du ing he lyby is selec ed, p e e ably one
exhibi ing highly a iable esolu ion alues. Fo example, he minimum emission
angle (<90° o sa is y he isibili y condi ion) should be nea o he minimum dis ance
and inc ease as he dis ance inc eases. The bes esolu ion in he lyby, acco ding
o he p oposed o mula (4.1), will occu nea he wo minima, while esolu ion will
wo sen, along wi h he indi idual’s i ness, as bo h emission angle and dis ance
inc ease. Addi ionally, a check is pe o med using he p ecompu ed iles o ensu e
ha he wo minima a e wi hin a complian in e al o he lyby (also cons ain s on
illumina ion zeni h and phase angles mus be espec ed).
 De e mining he op imal schedule o obse ing only he selec ed ROI wi hin he
complian in e al o he lyby ha has he minimum esolu ion alue.
The chosen ROI is JUICE_ROI_GAN_5_0_09, which exhibi s he emission angle and
dis ance alues du ing he lyby as shown:
Figu e 5.1: Plo o emission angle ( alues on he le ) and dis ance ( alues on he igh ) om he
JUICE_ROI_GAN_5_0_09's cen oid o he i s alida ion case

Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
42
This ROI indeed adhe es o he emission angle and dis ance ends desc ibed abo e. The
esolu ion end o e ime is such ha i eaches a minimum alue and hen con inuously
inc eases o e ime:
Figu e 5.2: Resolu ion o e JUICE_ROI_GAN_5_0_09 in he lyby om 2034 JUN 05 18:53:51 o
2034 JUN 06 18:53:51
In his case, each oplan class objec will con ain a single elemen in he lis s s ol, q oi and
obsLeng h. Assuming ha i will de ini ely be possible o obse e he ROI s a ing om he
ins an o minimum esolu ion wi hou he end o he obse a ion alling ou side he in e al
– and knowing he e a e no o he ROIs o obse e – he bes indi idual will be he one whe e
he elemen in s ol coincides wi h he ins an o minimum esolu ion.
Howe e , he op imal solu ion in his case is al eady eached in he second gene a ion,
making he scena io oo simple o be o in e es o alida ion pu poses.
To make he p oblem mo e challenging o he algo i hm, all he lybys in which
JUICE_ROI_GAN_5_0_09 is isible a e conside ed. Fu he mo e, o ind he complian
in e als, i is assumed ha he only cons ain o be espec ed in he p ecompu a ion is ha
he emission angle < 90°, in o de o ha e mo e in e als in which o sea ch o he
minimum.The p ecompu a ion was he e o e pe o med again o his case, bu only o he
ROI o in e es and wi h he emission angle cons ain only.
Fo he selec ed ROI he e a e 8 complian ime in e als du ing he lybys (Table 4.1). As
can be seen om he analysis o he g aphs in Figu e 5.3 and Figu e 5.4, he ins an o
minimum esolu ion lies wi hin he complian in e al o he i s lyby and, again, in absence
o o he ROIs and obse a ion ca ied ou en i ely wi hin he in e al, his ins an is he
op imal one o s a he obse a ion.
The s a ing obse a ion ins an p o ided by he GA coincides wi h he ac ually op imal one,
hus alida ing he algo i hm jus de eloped. I is wo h no ing ha , as shown in Figu e 5.5
he bes esul is achie ed a ound gene a ion numbe 5 ( he e o e, 500 gene a ions a e no
equi ed o achie e he op imal esul and only 20 will be inse ed), which makes i a case
ha is no o e ly simple o he algo i hm.
43
Figu e 5.3: Resolu ion end in he i s 4 complian in e als o he ROI JUICE_ROI_GAN_5_0_09.
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
44
Figu e 5.4: Resolu ion end in he las 4 complian in e als o he ROI JUICE_ROI_GAN_5_0_09.
45
Figu e 5.5: E olu ion o bes i ness ( esolu ion) in he case o he es whe e only
JUICE_ROI_GAN_5_0_09 is obse ed in all he complian in e als ob ained by conside ing
emission angle as he only cons ain
5.2 Single-Agen and Mul i-Agen : esul s and discussion
A e pe o ming he p ecompu a ion o ob ain he complian in e als ha espec he
cons ain s on emission angle, illumina ion zeni h angle and phase angle (Table 4.2), he
GA was execu ed i s o a single agen and hen wi h mul iple agen s in pa allel, wi h he
pa ame e s speci ied in Sec ion 4.2.3.
The i ness (which, as a eminde , is a weigh ed a e age o co e age and esolu ion) o he
bes indi idual o each gene a ion o he single agen a e shown:
Figu e 5.6: Fi ness wi h gene a ion, single agen
The inal bes indi idual’s i ness is 0.986.
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
52
6 Budge summa y
In Table 6.1 he es ima ed budge o he Mas e ’s Thesis, comp ising p o essional ees
and elec ici y cos , is summa ized.
The S a cos was de e mined using he annual sala y o a junio ae ospace enginee wi h
no p io expe ience, calcula ed as 2299 € [37]. The hou ly a e was de i ed based on a ypi-
cal wo k schedule, assuming 12 mon hs in a yea , 4 weeks pe mon h, and 40 wo king
hou s pe week. Elec ici y cos s we e es ima ed using da a om [38]. The calcula ions con-
side a lap op consuming 130 W o e 300 hou s and he JUNO se e ope a ing a 500 W
o 336 hou s. These assump ions o m he basis o he alues p o ided in Table 6.1.
Table 6.1: Budge summa y
Ca ego y Value Uni cos Cos
S a cos 300 hou s 12.0 € / hou 3594.0 €
Elec ici y - Lap op 39 kWh 0.119 € / kWh 4.64 €
Elec ici y - JUNO
se e
168 kWh 0.119 € / kWh 19.99 €
To al es ima ed budge 3.618,63 €

53
7 Analysis and assessmen o en i onmen al and social
implica ions
Space missions ep esen a pi o al s ep in humani y's ques o expand he bounda ies o
scien i ic knowledge. Howe e , hese endea o s ca y signi ican en i onmen al and social
implica ions ha mus be c i ically analyzed o ensu e sus ainable de elopmen and
esponsible explo a ion. This chap e discusses he key conside a ions associa ed wi h he
scheduling o emo e sensing obse a ions, ocusing on he en i onmen al impac and
socie al con ibu ions o space explo a ion.
7.1 En i onmen al implica ions
The en i onmen al oo p in o space missions p ima ily s ems om he manu ac u ing,
launch, and ope a ion phases. The launch p ocess, in ol ing he combus ion o ocke
p opellan s, eleases subs an ial quan i ies o g eenhouse gases and pa icula e ma e
in o he a mosphe e, po en ially a ec ing he ozone laye and con ibu ing o clima e
change. Fo ins ance, he Falcon 9 launch by SpaceX in 2018 bu ned app oxima ely
112,184 kilog ams o ke osene, eleasing abou 336,552 kilog ams o ca bon dioxide in o
he Ea h's a mosphe e [40].
Mo eo e , he inc easing accumula ion o space deb is in Ea h's o bi poses long- e m
isks o bo h ac i e sa elli es and u u e missions. Space junk, comp ising emnan s o
ocke s and spacec a , h ea ens ope a ional sa elli es and can plumme owa ds Ea h,
po en ially ha ming he en i onmen and human popula ions. Despi e guidelines om he
Uni ed Na ions O ice o Ou e Space A ai s [41], only abou 40% o space missions ad-
he e o olun a y space deb is mi iga ion p ac ices.
In he con ex o plane a y explo a ion missions, adhe ence o plane a y p o ec ion
p o ocols is c i ical. P e en ing con amina ion o celes ial bodies wi h Ea h-based
biological ma e ial is essen ial o p ese e he in eg i y o u u e scien i ic in es iga ions.
The op imiza ion o obse a ion schedules, as explo ed in his hesis, indi ec ly suppo s
hese goals by maximizing he scien i ic e u n pe mission and educing he need o
edundan ope a ions, he eby minimizing esou ce usage.
Da a ansmission in Space missions has no able en i onmen al implica ions. G ound s a-
ions, such as he Deep Space Ne wo k (DSN), equi e signi ican ene gy o main ain con-
ac wi h spacec a , while onboa d sys ems mus balance powe consump ion wi h limi ed
ene gy esou ces like sola panels o nuclea gene a o s. Ine icien da a managemen
can deple e hese esou ces, isking mission objec i es and inc easing he en i onmen al
oo p in .
Op imizing ansmission schedules mi iga es hese impac s by educing he equency and
olume o da a ans e s. P io i izing high- alue da a and using comp ession echniques
minimizes ene gy usage bo h onboa d and on he g ound, aligning wi h sus ainabili y
goals
7.2 Social Implica ions
The socie al impac o space missions ex ends beyond scien i ic ad ancemen s, in luencing
educa ion, echnology, and global coope a ion. By enhancing ou unde s anding o
plane a y sys ems, hese missions inspi e u u e gene a ions o pu sue ca ee s in Science,
Technology, Enginee ing, and Ma hema ics (STEM). Addi ionally, echnologies de eloped
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
54
o Space explo a ion o en ind applica ions in a ious indus ies, om heal hca e o
en i onmen al moni o ing, d i ing inno a ion and economic g ow h.
Howe e , he apid inc ease in space ac i i ies, pa icula ly by p i a e companies, has
aised conce ns abou equi y and en i onmen al jus ice. The localized en i onmen al
damage caused by equen launches disp opo iona ely a ec s nea by communi ies and
ecosys ems. Fo ins ance, SpaceX's ope a ions in Boca Chica, Texas, ha e epo edly
dis up ed he habi a s o endange ed species, such as bi ds and sea u les, while also
a ec ing local esiden s [42]. These cases highligh he need o a balanced app oach o
space explo a ion ha conside s bo h echnological ambi ions and he well-being o a ec ed
popula ions.
The op imiza ion echniques discussed in his hesis ha e he po en ial o democ a ize
access o space by imp o ing he e iciency and cos -e ec i eness o mission planning. By
educing he compu a ional esou ces equi ed o scheduling, smalle ins i u ions and
na ions could pa icipa e in plane a y explo a ion, os e ing inclusi i y and collabo a ion.
7.3 E hical Conside a ions
The e hical dimensions o Space explo a ion, pa icula ly ega ding esou ce u iliza ion
and po en ial coloniza ion, wa an ca e ul delibe a ion. As humani y explo es new on-
ie s, i is c ucial o balance he pu sui o scien i ic disco e y wi h a espec o he in insic
alue and p ese a ion o celes ial bodies. Space is no me ely a esou ce o exploi bu a
sha ed he i age ha holds cul u al, scien i ic, and philosophical signi icance.
One key conside a ion is he po en ial con amina ion o ex a e es ial en i onmen s. In-
oducing Ea h-based mic oo ganisms o ma e ials o o he plane s could dis up na i e
ecosys ems—i hey exis —and comp omise he in eg i y o u u e scien i ic in es iga ions.
Adhe ing o s ic plane a y p o ec ion p o ocols is essen ial o minimize hese isks while
main aining he scien i ic eliabili y o missions.
Resou ce ex ac ion also aises signi ican e hical ques ions. The u iliza ion o ex a e es-
ial ma e ials, such as mining as e oids o he Moon, may p o ide subs an ial bene i s o
space missions and indus ies on Ea h. Howe e , such ac i i ies mus be go e ned by in-
e na ional ag eemen s o ensu e equi able access and a oid monopoliza ion by a ew
powe ul en i ies.
The ools and me hodologies de eloped in his hesis align wi h hese e hical impe a i es
by p omo ing he e icien use o esou ces and minimizing unnecessa y in e en ions.
This app oach no only enhances mission sus ainabili y bu also e lec s a b oade com-
mi men o esponsible and inclusi e explo a ion.
55
8 Conclusions
This hesis has explo ed he use o gene ic algo i hms (GAs) o op imize he planning o
scien i ic obse a ions in Space missions, ocusing speci ically on he applica ion o a sin-
gle ins umen scena io. The main objec i e was o demons a e he e ec i eness o hese
compu a ional echniques in add essing he challenges posed by ope a ional cons ain s
while maximizing scien i ic e u n. The wo k emphasizes he po en ial o GAs as a obus
and adap able solu ion o single-ins umen scheduling p oblems.
The s udy began wi h he iden i ica ion o a key issue in space mission planning: he need
o e icien scheduling o obse a ions o ensu e compliance wi h s ic mission con-
s ain s. Unlike adi ional manual planning app oaches, which a e ime-in ensi e and
p one o subop imal solu ions, he p oposed me hod le e ages ad anced compu a ional
ools o au oma e and op imize he p ocess. To ackle his challenge, a model was de el-
oped ha in eg a es NASA's SPICE oolki o mission da a managemen and Py hon-
based lib a ies o op imiza ion. The gene ic algo i hm was ca e ully ailo ed o iden i y ea-
sible and op imal obse a ion schedules o a single ins umen , conside ing ac o s such
as obse a ion windows and scien i ic p io i ies.
8.1 Main Resul s
The case s udy cen ed on he Ganymede lybys o he JUICE mission, aiming o op imize
he obse a ion schedule o he JANUS ins umen . These lybys p esen ed unique
challenges due o he limi ed ime a ailable o high- esolu ion obse a ions and he need
o align wi h scien i ic objec i es. The esul s demons a ed he algo i hm’s abili y o:
1. Gene a e easible schedules: The GA success ully iden i ied obse a ion windows
ha complied wi h he mission’s geome ic and empo al cons ain s. This ensu ed
ha all obse a ions adhe ed o he p ede ined pa ame e s wi hou con lic .
2. Op imize obse a ion quali y: The algo i hm e ec i ely p io i ized obse a ion
oppo uni ies, selec ing hose ha maximized image esolu ion and scien i ic alue.
This esul ed in a signi ican imp o emen in he quali y o he obse a ion plan
compa ed o baseline me hods.
3. E icien ly alloca e esou ces: The model e ec i ely balanced he ade-o s
be ween ope a ional cons ain s and scien i ic objec i es, ensu ing op imal esou ce
u iliza ion. This was pa icula ly e iden in he a oidance o o e lapping asks.
8.2 Con ibu ions and Implica ions
This hesis con ibu es o he ongoing de elopmen o compu a ional ools o space mission
planning, showcasing he po en ial o gene ic algo i hms o op imize single-ins umen
obse a ion schedules. The esea ch unde sco es he impo ance o le e aging ad anced
compu a ional echniques o add ess complex ope a ional challenges. Key implica ions
include:
 P ac ical ele ance: The de eloped model o e s a scalable and adap able
app oach ha can be applied o o he single-ins umen scena ios in a ious space
missions. I s modula design acili a es in eg a ion wi h exis ing mission planning
amewo ks, p o iding a p ac ical ool o mission designe s.
 Enhanced mission e iciency: By au oma ing he scheduling p ocess, he
me hodology educes he eliance on manual planning, he eby imp o ing
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
56
ope a ional e iciency and decision-making. This app oach also minimizes he isk
o human e o , which can ha e signi ican implica ions o mission success.
8.3 Limi a ions and Fu u e Pe spec i es
While he esul s a e p omising, he scope o his s udy is limi ed o a single-ins umen
applica ion. This p oblem has been he s a ing poin o adap he algo i hm o mul i-
ins umen missions, add essing mo e complex scheduling scena ios, including con lic ing
objec i es among ins umen s.
Fu u e esea ch could expand by:
1. Inco po a ing o he op imiza ion echniques: O he algo i hms, such as
Simula ed Annealing, could be used o add ess he mission scheduling p oblem.
One me hod could in ol e assigning di e en heu is ics o each agen in a mul i-
agen sys em.
2. Valida ing on addi ional missions: Tes ing he model on di e en mission p o iles
and ins umen s o e alua e i s gene alizabili y and obus ness. Compa a i e s udies
wi h exis ing planning ools could also p o ide aluable insigh s in o i s ela i e
pe o mance.
3. Explo ing addi ional heu is ics: In oducing al e na i e heu is ic algo i hms such
as Sidewinde o Replanning Sidewinde o compa e hei pe o mance and
e ec i eness in op imizing obse a ion schedules. These me hods could
complemen he GA app oach by p o iding al e na i e solu ions o enhancing
speci ic aspec s o he planning p ocess.
4. Enhancing eal- ime adap abili y: De eloping ex ensions o enable eal- ime
eplanning capabili ies, allowing he model o espond o dynamic mission
en i onmen s and un o eseen cons ain s.
The app oach de eloped he e ep esen s a aluable s ep o wa d in ad ancing space
mission planning, p o iding a ounda ion o u u e esea ch and applica ions in he ield o
compu a ional op imiza ion. By add essing he limi a ions and explo ing new di ec ions, his
wo k opens he doo o mo e sophis ica ed and e icien planning ools, ensu ing ha u u e
Space missions achie e hei scien i ic goals wi h maximum e iciency and p ecision.
57
9 Re e ences
[1] PATERNA, S.; SANTONI, M.; BRUZZONE, L.: An App oach Based on
Mul iobjec i e Gene ic Algo i hms o Schedule Obse a ions in Plane a y Remo e
Sensing Missions. IEEE Jou nal o Selec ed Topics in Applied Ea h Obse a ions
and Remo e Sensing [online]. 2020, ol. 13 [ isi ed on 2025-01-04].
[2] TORRES RUBIO, P.: Gene ic algo i hms o schedule obse a ions in plane a y
emo e sensing missions [online]. Bachelo ’s Thesis, UPC, Escola Supe io
d'Enginye ies Indus ial, Ae oespacial i Audio isual de Te assa, Depa amen de
Física, 2024. A ailable a : h ps://upcommons.upc.edu/handle/2117/416191
[3] SORIA GUERRERO, M.: Py hon Mul iheu is ic Op imiza ion Tools [online]. A ailable
a : h ps://gi hub.com/ManelSo ia/PMOT. [N.d.]. [ isi ed on 2025-01-07].
[4] ACTON, C.: An o e iew o SPICE. Je P opulsion Labo a o y: Oak G o e, KY, USA,
1998.
[5] STEPHAN, K. e al. Regions o In e es on Ganymede’s and Callis o’s su aces as
po en ial a ge s o ESA’s JUICE mission. Plane a y and Space Science [online].
2021, ol. 208, p.105324 [ isi ed on 2025-01-05]. ISSNN 0032-0633. A ailable a
DOI: 10.1016/j.pss.2021.105324
[6] KERSTEN, E.; ZUBAREV, A.E.; ROATSCH, Th.; MATZ, K.-D.: Con olled Global
Ganymede mosaic om Voyage and Galileo images. Plane a y and Space Science
[online].2021, ol. 206, p.105310 [ isi ed on 2025-01-05]. ISSN 0032-0633.
A ailable a DOI: 10.1016/j.pss.2021.105310
[7] BETRIU, P.: Op imiza ion Techniques in Science Planning o Plane a y Explo a ion
missions [online]. PhD Thesis, UPC, Escola Supe io d'Enginye ies Indus ial,
Ae oespacial i Audio isual de Te assa, Depa amen d'Enginye ia Mecànica, 2024.
A ailable a : h ps://upcommons.upc.edu/handle/2117/420573
[8] SIMÓN AZNAR, J.: Gene ic algo i hms o schedule obse a ions in plane a y
emo e sensing missions [online]. Mas e ’s Thesis, UPC, Escola Supe io
d'Enginye ies Indus ial, Ae oespacial i Audio isual de Te assa, Depa amen de
Física, 2024. A ailable a : h ps://upcommons.upc.edu/handle/2117/418588.
[9] ANDÍA VON LIGNAU, Diego. A new se o obse a ions o Eu opa o he Galileo
spacec a [online]. 2022. [ isi ed on 2025-01-05]. A ailable a :
h ps://upcommons.upc.edu/handle/2117/374159. Bachelo ’s Thesis. Uni e si a
Poli ècnica de Ca alunya. Accep ed: 2022-10-07T10:40:11Z.
[10] DOUBLEDAY, Joshua; KNIGHT, Russell: Science Mission Planning o DESDynI
wi h CLASP. 2014. ISBN 978-1-62410-221-9. A ailable a DOI: 10.2514/6.2014-
1757.
[11] KNIGHT, S.; RABIDEAU, G.; CHIEN, S.; ENGELHARDT, B.; SHERWOOD, R.
Caspe : space explo a ion h ough con inuous planning. IEEE In elligen Sys ems.
2001, ol. 16, no. 5, pp. 70–75. A ailable a DOI: 10.1109/MIS.2001.956084.
[12] KNIGHT, R.; HU, S.: Comp essed La ge-scale Ac i i y Scheduling and Planning
(CLASP) Applied o DESDynI. Je P opulsion Labo a o y, Cali o nia Ins i u e o
Technology. A ailable a : smci .ecs.baylo .edu.

Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
58
[13] SHAO, E.; BYON, A.; DAVIES, C.; DAVIS, E.; KNIGHT, R.; LEWELLEN, G.;
TROWBRIDGE, M.; CHIEN, S.: A ea Co e age Planning wi h 3-axis S ee able, 2D
F aming Senso s. [N.d.].
[14] CHOO, T.; STEELE, R.; LUCKS, M.; NGUYEN, L.; ANDERSON, B.; BEDINI, P.
MESSENGER SciBox, An Au oma ed Closed-loop Science Planning and
Commanding Sys em. In: AIAA SPACE 2011 Con e ence & Exposi ion [online]. Long
Beach, Cali o nia: Ame ican Ins i u e o Ae onau ics and As onau ics, 2011 [ isi ed
on 2025-01-05]. ISBN 978-1-60086-953-2. A ailable a DOI: 10.2514/6.2011-7339
[15] GALUZIN, V.; KUTOMANOV, A. Yu.; MATYUSHIN, M. M.; SKOBELEV, P. : A
e iew o mode n me hods o planning and scheduling o he ope a ions in
ad anced space sys ems. In: Mekha onika A oma iza siya Up a lenie [online].
A ailable a DOI: 10.17587/mau.21.639-650
[16] KONAK, A.; COIT, D.W.; SMITH, A. E.: Mul i-objec i e op imiza ion using
gene ic algo i hms: a u o ial. Reliabili y Enginee ing & Sys em Sa e y. Volume 91,
Issue 9, Pages 992-1007. 2006. ISSN 0951-8320. A ailable a DOI:
10.1016/j. ess.2005.11.018.
[17] JUICE Science S udy Team: JUICE Explo ing he eme gence o habi able wo lds
a ound gas gian s. ESA(SRE)(2011)18, Issue 1, Decembe 2011. Eu opean Space
Agency [online]
[18] Eu opean Space Agency (ESA): The science: JUICE's key objec i es a Jupi e .
In: ESA Science & Explo a ion [online]. A ailable a :
h ps://www.esa.in /Science_Explo a ion/Space_Science/Juice/The_science_Juice
_s_key_objec i es_a _Jupi e
[19] Eu opean Space Agency (ESA). Segmen Visualisa ion and Co e age Tool
[online]. [N.d.]. [ isi ed on 2024-06-12]. A ailable om:
h ps://juicesoc.esac.esa.in /e en _ ool/segmen /segmen _ca alogue_c ema_5_1_
150lb_ 23_1
[20] Eu opean Space Agency (ESA): Juice: Jupi e Icy Moons Explo e [online].
[ isi ed on 2025-01-06]. A ailable a : h ps://www.esa.in /Science_Explo a ion
/Space_Science/Juice
[21] SARRI,G.; WITASSE, O.; CAVEL, C. : The JUICE Mission o Jupi e and i s Icy
Moons. 9 h EUROPEAN CONFERENCE FOR AERONAUTICS AND SPACE
SCIENCES (EUCASS). A ailable a DOI: 10.13009/EUCASS2022-7148
[22] SCHMITZ, N. e al.:JANUS on JUICE: A Came a o In es iga e Ganymede, Eu opa,
Callis o and he Jo ian Sys em. Deu sches Zen um ü Lu - und Raum ah e.V.
(DLR). Sep embe 2013.
[23] JUICE Science S udy Team: JUICE Explo ing he eme gence o habi able wo lds
a ound gas gian s. ESA(SRE)(2014)1, Issue 1, Sep embe 2014. Eu opean Space
Agency [online]
[24] HARUYAMA, J. : ONBOARD JUICE: JANUS. Japan Ae ospace Explo a ion
Agency (JAXA) / Ins i u e o Space and As onau ical Science (ISAS).Ap il 2023
A ailable a : h ps://cosmos.isas.jaxa.jp/onboa d-juice-janus/
59
[25] LARA, L. : JANUS, he op ical came a on ESA's JUICE p obe, cap u es s unning
images du ing i s i s luna and e es ial lyby. Ins i u o de As o ísica de Andalucía
(IAA-CSIC). Augus 2024. A ailable a : h ps://www.iaa.csic.es/en/news/janus-
op ical-came a-esas-juice-p obe-cap u es-s unning-images-du ing-i s- i s -luna -
and
[26] SHAO, E.; BYON, A.; DAVIS, E.; KNIGHT, R.; LEWELLEN, G.; TROWBRIDGE,
M.; CHIEN, S. : A ea co e age planning wi h 3-axis s ee able, 2D aming senso s.
In e na ional Con e ence on Au oma ed Planning and Scheduling (ICAPS), 2018.
[27] BETRIU, P.: A ea Co e age Planning, Gi Hub eposi o y. [ isi ed on 2025-01-06].
A ailable a : Gi Hub - PauBe /a ea-co e age-planning
[28] HOLLAND, J.H.: Gene ic Algo i hms, Scien i ic Ame ican, Vol. 267, N. 1. July 1992,
pp. 67-73.
[29] BRABAZON, A.; O’NEILL, M.; MCGARRAGHY, S.: Na u al Compu ing Algo i hms.
Na u al Compu ing Se ies, Sp inge , 2015. ISSN 1619-7127. A ailable a DOI:
10.1007/978-3-662-43631-8
[30] RAZALI, N. M.; GERAGHTY, J. : Gene ic Algo i hm Pe o mance wi h Di e en
Selec ion S a egies in Sol ing TSP. In e na ional Con e ence o Compu a ional
In elligence and In elligen Sys ems (ICCIIS'11). 2011. A ailable a :
h ps://www.iaeng.o g/publica ion/WCE2011/WCE2011_pp1134-1139.pd
[31] KARIMI-MAMAGHAN, M.; MOHAMMADI, M.; MEYER, P.; KARIMI-MAMAGHAN,
A. M.; TALBI, G. : Machine lea ning a he se ice o me a-heu is ics o sol ing
combina o ial op imiza ion p oblems: A s a e-o - he-a . Eu opean Jou nal o
Ope a ional Resea ch. Volume 296, Issue 2, 2022, Pages 393-422. ISSN 0377-
2217. A ailable a DOI: 10.1016/j.ejo .2021.04.032.
[32] DI SARNO, P.; FIENGO, I.F.: TFM_Naples_SingleIns . Gi Hub eposi o y[online],
2024. [ isi ed on 2025-01-08]. A ailable a h ps://gi hub.com/
Pie ooDiSa no/TFM_Naples_SingleIns
[33] In es igación y ecnología de la UPC en Te assa, TUAREG. A ailable a :
h ps:// ece ca e assa.upc.edu/es/equipamien os-cien i ico ecnicos/ ua eg
[34] Eu opean Space Agency (ESA), JUICE SOC E en Ca alogue. A ailable a :
h ps://juicesoc.esac.esa.in /e en _ ool/e en /e en _ca alogue_5_1_150lb_23_1_b
2. 2024. A ailable also a : h ps://juicesoc.esac.esa.in /e en _ ool
/e en /e en _ca alogue_5_1_150lb_23_1_b2.
[35] BETRIU, P. : Py hon Science Oppo uni y Analysis Tool. Gi Hub eposi o y [online],
2024. [ isi ed on 2025-01-08]. A ailable a h ps://gi hub.com/PauBe /PSOA
[36] SORIA, Manel. pySPICElib, Gi Hub eposi o y [online]. [N.d.]. A ailable a :
h ps://gi hub.com/ManelSo ia/pySPICElib. [N.d.]. Accessed: 2025-01-08.
[37] S ipendi pe Junio Enginee . 2024. [online]. A ailable a :
h ps://www.glassdoo .i /S ipendi/ba celona-junio -enginee -s ipendio-
SRCH_IL.0,9_IC2547194_KO10,25.h m. [ isi ed on 2025-01-12]
Op imiza ion Heu is ics o Scheduling Obse a ions in
Remo e Sensing Missions: Single-Ins umen case
60
[38] P ecio de la luz en Ba celona HOY. 12-01-2025. [online]. A ailable a :
h ps://www.menosdeluz.com/p ecio-luz-ba celona. [ isi ed on 2025-01-12]
[39] TORRES RUBIO, P.: oplan. Gi Hub eposi o y [online], 2025. [ isi ed on 2025-01-
12]. A ailable a : h ps://gi hub.com/Pablo-TR/oplan
[40] SPACE X: FALCON 9, FIRST ORBITAL CLASS ROCKET CAPABLE OF
REFLIGHT [online]. A ailable a : h ps://www.spacex.com/ ehicles/ alcon-9/ [ isi ed
on 2025-01-12].
[41] UNITED NATIONS, O ice o Ou e Space A ai s [online]. A ailable a :
h ps://www.unoosa.o g/ [ isi ed on 2025-01-12]
[42] DEFENDERS OF WILDLIFE, The Te ible I ony o Des oying Ea h in Sea ch o
Plan(e ) B: SpaceX’s Impac s o Boca Chica, Texas. A ailable a :
h ps://de ende s.o g/blog/2024/10/ e ible-i ony-o -des oying-ea h-sea ch-o -
plane -b-spacexs-impac s-boca-chica-0 [ isi ed on 2025-01-12]