16 h In e na ional Symposium on P ac ical Design o Ships and O he Floa ing S uc u es PRADS 2025
Ann A bo , MI, USA, Oc obe 19 h-23 d 2025
Explo ing easibili y o UAV-Based Si ua ional Awa eness Suppo o
Ma i ime Au onomous Su ace Ships h ough Scena io-Based Sea T ials
Dong-eon Kim1, Hyun-Jae Jo1, Tae-Jun Youn1and Geun-Tae Yim1,*
1Ko ea Resea ch Ins i u e o Ships & Ocean Enginee ing, Ulsan, Republic o Ko ea
Abs ac . This s udy in es iga es he easibili y o employing unmanned ae ial ehicle (UAV) o en-
hance he si ua ional awa eness capabili ies o Ma i ime Au onomous Su ace Ships (MASS), he eby
imp o ing he eliabili y o hei ope a ions. Howe e , ixed si ua ional awa eness senso s on MASS,
including came as, LiDAR, and ada , a e subjec o inhe en physical limi a ions due o hei s a ic
ins alla ion, educing hei e ec i eness in dynamic and complex na iga ional en i onmen s. To ad-
d ess hese limi a ions, da a collec ed om scena io-based eal-sea ials we e applied in simula ions
o e alua e he e ec i eness o UAV-assis ed si ua ional awa eness. Fo his pu pose, we designed and
conduc ed ials using he sea- ial es bed essel HaeyangNu i. The ep esen a i e ope a ional scena io
was se in a geog aphically cons ained channel. Fo he simula ion, he essel’s mo ion was modeled
using he Veloci y Obs acle me hod, and a weigh ed exponen ial model, inco po a ing he dis ance and
ime o he closes poin o app oach (DCPA and TCPA), was de eloped o quan i a i ely assess colli-
sion isk. The esul s demons a ed educed collision isk and imp o ed ope a ional s abili y when UAV
we e employed. These indings highligh he po en ial o UAV o suppo MASS na iga ion, and u-
u e wo k will ocus on de eloping a obus MASS–UAV coope a i e amewo k o b oad ope a ional
applica ions o u he enhance he ope a ional eliabili y o MASS.
Key wo ds: Ma i ime Au onomous Su ace Ship, Sea ial, Simula ion, Si ua-
ional awa eness, Unmanned Ae ial Vehicle
1. In oduc ion
Recen ad ances in a i icial in elligence (AI), da a communica ions echnology, and sensing ha dwa e
ha e accele a ed esea ch and de elopmen on he au oma ion o a ious mobile pla o ms. In he ma i ime
sec o , se e al majo ini ia i es, such as KASS(Ko ea Au onomous Su ace Ship)p ojec in Ko ea, One
Sea Associa ion in Finland, Megu i 2040 in Japan among o he s, a e pu suing he comme cializa ion o Au-
onomous Ma i ime Su ace Ship (MASS). Acco ding o he In e na ional Ma i ime O ganiza ion (IMO), a
MASS is a essel ha is capable o ope a ing independen ly wi hou human in e en ion, al hough emo e
ope a ion may be equi ed depending on he le el o au onomy. Gi en ha con en ional ship ope a ions e-
qui e a conside able wo k o ce o specialized pe sonnel, and ha app oxima ely 85% o ma i ime acciden s
a e epo ed o be caused by human e o , MASS echnology is expec ed o deli e signi ican gains in bo h
e iciency and sa e y[1].
The si ua ional awa eness sys em is conside ed a co e echnology o MASS, designed o suppo he
de ec ion and ecogni ion o su ounding objec s h ough senso s and o assis in a oiding haza dous si ua-
ions. To acili a e complemen a y si ua ional awa eness, i ypically in eg a es mul i-senso da a, including
RADAR, LiDAR, AIS, and came as.[2]By p ocessing and using eal- ime in o ma ion, he sys em can
con ibu e o sa e na iga ion, collision a oidance, and mo e in o med decision-making. The sys ems ha e
been de eloped and demons a ed in a ious MASS echnology de elopmen p ojec s, and he e a e also
cases whe e hey ha e been de eloped and comme cialized as auxilia y sys ems o con en ional ships.
Wi hin he KASS P ojec , he in elligen Si ua ional Awa eness Sys em (iSAS)has been de eloped[3].
iSAS uses senso s moun ed on he ship and p ocesses aw senso da a wi h deep-lea ning-based de ec ion
algo i hms o gene a e eal- ime objec ecogni ion and s a us es ima ion. I is ins alled on he 25-me e
*Co espondence o: [email p o ec ed]
1
es bed essel HaeyangNu i, which has been es ablished and ope a ed o he e i ica ion and alida ion
(V&V)o MASS echnologies. Using his pla o m, eal-sea ials ha e been conduc ed o e alua e he
sys em’s objec de ec ion pe o mance, collec ope a ional da a, and pe o m in eg a ed ials in conjunc ion
wi h he au onomous na iga ion sys em.[4][5]
In in e na ional na iga ion, essels may encoun e si ua ions whe e shipboa d senso s, cons ained by
hei ixed ins alla ion, exhibi inhe en limi a ions in de ec ion co e age. These limi a ions a e pa icula ly
e iden in a eas wi h geog aphical cons ain s, such as na ow channels, b eakwa e s, o nea by islands.
Gi en ha such condi ions a e known o pose ele a ed acciden isks o con en ional essels[6][7], he
In e na ional Ma i ime O ganiza ion (IMO)has manda ed he use o he Au oma ic Iden i ica ion Sys em
(AIS)—a sys em ha b oadcas s eal- ime essel iden i y, posi ion, and cou se using adio wa es—and
es ablished egula ions o he deploymen and ope a ion o Vessel T a ic Se ices (VTS)in high- isk ma -
i ime a eas. Howe e , despi e hese measu es, ma i ime acciden s con inue o be epo ed, highligh ing
he inhe en limi a ions o VTS—i s eliance on human ope a o s and passi e da a sou ces such as ada ,
AIS, and olun a y ship epo ing—and aising pe sis en conce ns abou he eliabili y and in eg i y o AIS
da a[8][9]. Al hough ex ensi e esea ch is unde way o add ess he limi a ions o bo h VTS and AIS[10]
[11], a MASS mus be capable o independen ly ensu ing obus si ua ional awa eness o ope a e sa ely in
di e se and dynamic en i onmen s wi hou elying solely on ex e nal aids. This independence is pa icula ly
c i ical o MASS, as i s da a-d i en decision-making a chi ec u e can be highly ulne able when dependen
on po en ially un eliable ex e nal inpu s.
To help add ess he limi a ions o ixed shipboa d senso s, he use o Unmanned Ae ial Vehicle (UAV)
as ac i e and mobile si ua ional awa eness pla o ms has been sugges ed as a po en ial app oach. UAV, wi h
hei mobili y, may enhance si ua ional awa eness beyond he capabili ies o ixed shipboa d senso s and
con ibu e o imp o ed ope a ional sa e y o MASS.
Mul i-domain ope a ional cases in eg a ing land and ai sys ems ha e been epo ed, such as mili a y
ope a ions, logis ics suppo , and au onomous ehicle assis ance. In he ma i ime domain, a ious indepen-
den s udies ha e been conduc ed on he use o UAV, such as inspec ion o ship hulls and anks, deli e y o
supplies o wa ship[12], UAV landing echnologies unde dynamic sea condi ions[13]. Resea ch on coope -
a i e USV–UAV sys ems o na iga ion suppo o man-o e -boa d sea ches by using image-based objec
ecogni ion and dis ance es ima ion has also been ac i ely pu sued[14][15][16]. Howe e , esea ch on he
use o UAV o imp o e he sa e ope a ion o MASS emains in i s ea ly s ages[17].
Based on his backg ound, his s udy pe o ms simula ion using da a om sea ial conduc ed wi h he
HaeyangNu i o examine he easibili y o u ilizing UAV o enhance he si ua ional awa eness capabili ies
o MASS. Sec ion 2 in oduces he pla o ms used in he sea ials. Sec ion 3 desc ibes he me hodology, in-
cluding he applied algo i hms and analy ical amewo k. Sec ion 4 discusses he esul s, d aws conclusions,
and ou lines di ec ions o u u e esea ch.
2. Expe imen al Pla o ms o Real-Sea T ial
The Haeyang Nu i and a comme cially a ailable UAV we e u ilized in his s udy. The Haeyang Nu i,
buil unde he KASS p ojec , se es as a sea- ial es bed essel and is ac i ely used o he demons a ion
and pe o mance alida ion o MASS echnologies unde eal-sea condi ions.
A la , unobs uc ed su ace is essen ial o s able UAV ake-o and landing. Howe e , he uppe s uc-
Figu e 1. The HaeyangNu i, a sea- ial es bed essel
2
Figu e 2. UAV deck & senso s ins alled on HaeyangNu i
u e o he Haeyang Nu i is densely i ed wi h senso s and an ennas, c ea ing spa ial cons ain s o UAV
deploymen . To add ess his, a 4 m × 4 m UAV deck was ins alled on he essel’s op deck. In line wi h
CAP 437: S anda ds o O sho e Helicop e Landing A eas, he deck mee s he ecommended s anda d o
o o diame e s up o 3.2 m, based on he 1.25× sa e y ma gin[18].
The deck is cons uc ed wi h FRP–PVC sandwich panels o weigh educ ion and s uc u al s eng h,
inished wi h an i-slip, damping ma e ials o sa e y. A clea ance a he hull–deck junc ion imp o es ai low
and educes he in luence o ela i e wind du ing na iga ion.
2.1. Da a acqui ed by sea- ial
A scena io was de eloped o e lec na iga ional si ua ions commonly expe ienced by essels engaged
in in e na ional ade, pa icula ly in geog aphically cons ained en i onmen s such as channel app oaches
o he Po . In such wa e s, obs uc ions—such as headlands, islands, indus ial acili ies, and b eakwa e s—
can es ic he line o sigh o essels, which may esul in limi ed isibili y, delayed ecogni ion o po en ial
collision isks, and inc eased h ea s o sa e na iga ion.
To eplica e hese condi ions, a scaled-down sea ial was conduc ed a coas al si e in Ulsan, Republic
o Ko ea, unde WMO Sea S a e Code 2 (Smoo h)condi ions, wi h signi ican wa e heigh s below 0.5 m
and wind speeds no exceeding 8 kno s. The HaeyangNu i and he UAV we e manually con olled as pe
scena io by a c ew and a UAV ope a o , espec i ely. Came as moun ed on he HaeyangNu i and he UAV
used in he ial we e comme cial p oduc s. Mo ion da a o bo h he own ship and a ge essel, as well as
de ec ion esul s om ixed shipboa d senso s and UAV, we e collec ed o use in he simula ion. Figu e 3
p o ides an o e iew o he sea- ial p ocess.
The ial si e was si ua ed nea he b eakwa e o Bangojin Po (app ox. 35°28’38”N, 129°25’33”E),
an a ea cha ac e ized by e ain-induced line-o -sigh limi a ions. Figu e 4(a)p esen s he Elec onic Cha
Display and In o ma ion Sys em (ECDIS) iew o he Haeyang Nu i app oaching he b eakwa e be o e
Figu e 3. O e iew o he scena io
3
Figu e 4. S ill image o ial o Scena io; (a)ECDIS, (b)Moun ed came as,(c)UAV
en e ing he po , whe e he iangula ma ke indica es he a ge ship’s posi ion based on AIS signals. Fig-
u e 4(b)shows a s ill image om he essel-moun ed came as, while Figu e 4(c)p o ides a UAV-cap u ed
image.
2.2. Simula ion design
In he simula ion, he dynamic mo ion o he essels was modeled using he Veloci y Obs acle (VO)
algo i hm[19]. Fo he own ship, he mo ion cha ac e is ics o he Haeyang Nu i we e inco po a ed o e lec
i s maneu e ing beha io . The ini ial posi ions and speeds o bo h he own ship and he a ge ship we e
ep oduced om da a collec ed du ing he sea ial, and he simula ion en i onmen was designed o eplica e
he condi ions o he ial si e.
To e alua e de ec ion pe o mance, isual da a acqui ed om bo h he essel-moun ed came as and he
UAV we e p ocessed using YOLO11[20], an open-sou ce AI-based objec de ec ion model. A ime delay
was obse ed be ween he wo pla o ms: while he UAV success ully de ec ed he a ge ship i s , he ixed
came a equi ed app oxima ely 110 seconds longe o achie e de ec ion unde he same condi ions.
To quan i y he collision isk (R)be ween he own ship and he a ge ship, a weigh ed exponen ial
unc ion model conside ing bo h he Dis ance o Closes Poin o App oach (DCPA)and he Time o Closes
Poin o App oach (TCPA)was employed. The isk index anges om 0 o 1, whe e alues app oaching 1
indica e highe isk. The index inc eases only when bo h dis ance and ime simul aneously indica e ele a ed
collision isk.
R=e−DCPA
d0×e−TCPA
0(1)
whe e d0is he e e ence DCPA and 0is he e e ence TCPA.
Figu e 5. Exponen ial decay o he R: (a)TCPA a ia ion o ixed DCPA alues, and (b)DCPA a ia ion o ixed
TCPA alues
4
Figu e 6. Compa ison o he own ship’s ajec o y and collision isk index: (a) ixed-came a case, (b)UAV-assis ed
case. The dashed line is he in ended pa h, and he ed line is he ac ual ack.
2.3. Resul
The simula ion esul s demons a e a signi ican pe o mance imp o emen in collision-a oidance be-
ha io when u ilizing UAV-assis ed de ec ion compa ed o he ixed came a se up. As shown in Table 1, he
UAV achie ed an ea lie de ec ion ime o 110 s, compa ed o he ixed came a. This ea lie de ec ion led o
a 27.8% educ ion in he maximum isk index ( om 0.4827 o 0.3485)and a 14.4% dec ease in he a iance
o he isk index. The s anda d de ia ion also dec eased by 7.2%, indica ing mo e s able collision-a oidance
esponses du ing he simula ion.
In e ms o na iga ional accu acy, he ack e o —de ined as he accumula ed dis ance de ia ing om
he in ended pa h—was educed om 383 m in he ixed came a case o 258 m in he UAV-assis ed case,
ep esen ing a 32.6% imp o emen . The educ ion in he maximum isk index sugges s ha he likelihood
o a po en ial collision was e ec i ely lowe ed when UAV da a we e inco po a ed. Simila ly, he dec eases
in a iance and s anda d de ia ion e lec mo e consis en and p edic able collision-a oidance beha io ,
minimizing ab up changes in isk le els. The ack e o imp o emen u he indica es ha he essel was
able o main ain a ajec o y close o i s in ended pa h, educing unnecessa y de ia ions du ing maneu e ing.
Table 1. Compa ison o collision-a oidance pe o mance be ween ixed came a and UAV-assis ed de ec ion
Me ic Fixed came a UAV Imp o emen
Max. isk index 0.4827 0.3485 −27.8%
Va iance 0.0097 0.0083 −14.4%
S anda d de ia ion 0.0984 0.0913 −7.2%
T ack e o [m]383 258 −32.6%
No e: The ixed came a exhibi ed a de ec ion delay o app oxima ely 110 seconds compa ed o he UAV.
5
3. Discussion & u he wo k
The esul s o his s udy demons a e ha UAV-assis ed de ec ion signi ican ly enhances he imeliness o
si ua ional awa eness, pa icula ly in en i onmen s wi h line-o -sigh cons ain s. The educ ion in de ec ion
ime di ec ly con ibu ed o a lowe maximum isk index and educed a iabili y, indica ing mo e p edic able
and s able collision-a oidance beha io . Fu he mo e, he dec ease in ack e o sugges s imp o ed ajec-
o y adhe ence, which is c i ical o sa e na iga ion in cons ained wa e ways whe e maneu e ing ma gins
a e limi ed.
Despi e hese p omising esul s, se e al limi a ions emain. The sea- ial da a used o he simula ion
we e de i ed om a scaled-down es en i onmen , and u he e alua ions a e equi ed o e i y he op-
e a ional s abili y o UAVs unde di e se ma i ime condi ions, including high winds, p ecipi a ion, and
sal wa e exposu e. Mo eo e , he cu en e alua ion ocused on a single na iga ional scena io, which may
no ully cap u e he ange o condi ions encoun e ed in global ma i ime ope a ions.
Building on hese indings, a MASS–UAV coope a ion sys em is expec ed o u he enhance he ope a-
ional eliabili y o MASS by p o iding hi d-pe son pe spec i es o emo e ope a o s du ing na iga ion in
high- a ic a eas, suppo ing local ou e op imiza ion, secu i y moni o ing, and sa e y managemen . Build-
ing on he p o o ype de eloped and demons a ed wi h he Haeyang Nu i, u u e esea ch will b oaden he
ange o ope a ional scena ios and e ine his amewo k in o a obus , ully in eg a ed MASS–UAV coop-
e a ion sys em o p ac ical applica ion in di e se and dynamic ma i ime en i onmen s.
Acknowledgmen s
This esea ch was suppo ed by Ko ea Resea ch Ins i u e o Ships and Ocean enginee ing a g an om
Endowmen P ojec o “A s udy on sea demons a ion o ae ial d one ope a ion as auxilia y sys em o si u-
a ional awa eness o au onomous ships” unded by Minis y o Oceans and Fishe ies(2520000694)
Re e ences
[1]Jin Kim and Hwasup Jang. T ends and p epa a ions o ma i ime au onomous su ace ship echnology(자율운항
선박 기술동향 및 준비). Bulle in o he socie y o na al a chi ec s o ko ea,56(4):4–7, 2019.
[2]Les Elkins, D ew Selle s, and W. Reynolds Monach. The au onomous ma i ime na iga ion (amn)p ojec : Field
es s, au onomous and coope a i e beha io s, da a usion, senso s, and ehicles. Jou nal o Field Robo ics,
27(6):790–818, 2010.
[3]Hyun Taek Choi, Jeonghong Pa k, Jinwoo Choi, Minju Kang, Yeongjun Lee, Jongdae Jung, Jonghwi Kim,
Hyeokjun Kweon, Jinwhan Kim, Kuk Jin Yoon, e al. Design and p elimina y esul s o no el si ua ional awa eness
sys em o au onomous ship based on a i icial in elligence echniques. Jou nal o Ins i u e o Con ol, Robo ics
and Sys ems, 27(8):556–564, 2021.
[4]Minju Kang, Kibeom Choo, Jinwoo Choi, Hyun-Taek Choi, Hye jin Kim, and Jeonghong Pa k. A mul imodal
da ase o ma i ime objec de ec ion. In P oceedings o he Ko ean Socie y o Mechanical Enginee s Sp ing and
Au umn Con e ence, Jeju, Sou h Ko ea, 04 2025.
[5]Kibeom Choo, Jeonghong Pa k, Jinwoo Choi, Minju Kang, and Hyun-Taek Choi. Cu en s a us o si ua ional
awa eness sys em demons a ions o collision and acciden p e en ion in mass(자율운항선박 충돌 및 사고방지
상황인식 시스템 실증 현황). P oceedings o he Ko ean Ins i u e o Na iga ion and Po Resea ch Con e ence,
2024(11):43–44, 2024.
[6]N. End ina, J. C. Rase o, and J. Mon ewka. Risk analysis o ma i ime a ic in he s ai o gib al a and imp o e-
men p oposal. In C. Guedes Soa es and Angelo P. Teixei a, edi o s, Ma i ime T anspo a ion and Ha es ing o
Sea Resou ces, olume 1, pages 223–231, Uni ed S a es, Janua y 2016. CRC P ess. In e na ional Cong ess o
he In e na ional Ma i ime Associa ion o he Medi e anean, IMAM ; Con e ence da e: 09-10-2017 Th ough
11-10-2017.
[7]S Yildiz, F Tonoğlu, Ö Uğu lu, S Loughney, and J Wang. Spa ial and s a is ical analysis o ope a ional condi ions
con ibu ing o ma ine acciden s in he singapo e s ai . Jou nal o Ma ine Science and Enginee ing, 10(12),
Decembe 2022.
6
[8]And ej And ojna and Ma ko Pe ko ič. Ais da a alsi ica ion - how long will i be be o e we can no longe us ais?
In 2024 IEEE In e na ional Wo kshop on Me ology o he Sea; Lea ning o Measu e Sea Heal h Pa ame e s
(Me oSea), pages 301–306, 2024.
[9]Windwa d L d. Gps jamming alsely placed lcc on eagle in i an p io o collision, June 2025. Accessed:
2025-08-31.
[10]Xing Wu, A i a Rahman, and Vic o A. Zaloom. S udy o a el beha io o essels in na ow wa e ways using
ais da a – a case s udy in sabine-neches wa e ways. Ocean Enginee ing, 147:399–413, 2018.
[11]Lei Zhang, Jiahao Ge, Flo is Goe land , Lei Du, Tuowei Chen, Ting ing Gu, Langxiong Gan, and Xiaobin Li. A
me hod o enhancing he a ic si ua ion awa eness o essel a ic se ice ope a o s by iden i ying high isk
ships in complex na iga ion condi ions. Jou nal o Ma ine Science and Enginee ing, 13:379, 02 2025.
[12]Kai G ee . D ones o deli e ca go du ing oyal na y ca ie s ike g oup 25 deploymen . h ps://
hea ia ionis .com/2025/04/07/csg25-d one-ca go-deli e y/, Ap il 2025. Accessed: 2025-05-13.
[13]Shadi Abujoub, Johanna McPhee, and Rishad A. I ani. Me hodologies o landing au onomous ae ial ehicles on
ma i ime essels. Ae ospace Science and Technology, 106:106169, 2020.
[14]Chen Cheng, Dong Liu, Jin-Hui Du, and Yong-Zheng Li. Resea ch on isual pe cep ion o coo dina ed ai –sea
h ough a coope a i e us -ua sys em. Jou nal o Ma ine Science and Enginee ing, 11(10), 2023.
[15]Yuanda Wang, Wenzhang Liu, Jian Liu, and Changyin Sun. Coope a i e us –ua ma ine sea ch and escue wi h
isual na iga ion and ein o cemen lea ning-based con ol. ISA T ansac ions, 137:222–235, 2023.
[16]Xuesu Xiao, Jan Du ek, Tim Woodbu y, and Robin Mu phy. Ua assis ed us isual na iga ion o ma ine mass
casual y inciden esponse. In 2017 IEEE/RSJ In e na ional Con e ence on In elligen Robo s and Sys ems (IROS),
pages 6105–6110, 2017.
[17]D.-K. Kim, J.-H. Kim, H.-H. Son, S.-W. Choi, D.-H. Kim, C. Y. Yeo, and J.-Y. Pa k. A s udy on a eal- ime ae ial
image-based ua -us coope a i e guidance and con ol algo i hm. Jou nal o he Socie y o Na al A chi ec s o
Ko ea, 61(5):324–335, 2024. Published Oc obe 20, 2024.
[18]UK Ci il A ia ion Au ho i y. CAP 437: S anda ds o O sho e Helicop e Landing A eas. Ci il A ia ion Au-
ho i y, A ia ion House, Ga wick Ai po Sou h, Wes Sussex, RH6 0YR, UK, 9 edi ion, Feb ua y 2023. A ailable
a : h ps://www.caa.co.uk/CAP437.
[19]P. Fio ini and Z. Shille . Mo ion planning in dynamic en i onmen s using he ela i e eloci y pa adigm. In [1993]
P oceedings IEEE In e na ional Con e ence on Robo ics and Au oma ion, pages 560–565 ol.1, 1993.
[20]Ul aly ics. Yolo11: Real- ime objec de ec ion, 2024. Accessed: 2025-08-31.
7
