SUMO Use Con e ence 2025
Con e ence pape
h ps://doi.o g/10.52825/scp. 6i.2610
© Au ho s. This wo k is licensed unde a C ea i e Commons A ibu ion 3.0 DE License
Published: 15 Jul. 2025
SUMO-UAV-Py: A SUMO Plugin o UAV-Based Road
T a ic Sensing
Cha alambos Tsiou is1,* , Ch is os Mak idis1,2 , and
S elios Timo heou1,2
1KIOS Cen e o Excellence, Uni e si y o Cyp us, Cyp us
2Depa men o Elec ical and Compu e Enginee ing, Uni e si y o Cyp us, Cyp us
*Co espondence: Cha alambos Tsiou is
Abs ac . In ecen yea s, Unmanned Ae ial Vehicles (UAVs) ha e eme ged as e ec-
i e ools o a ic moni o ing and con ol by o e ing high- esolu ion, ae ial obse a-
ions o ehicula mo emen . Al hough UAV simula ion is well es ablished, ools o cap-
u e mic oscopic a ic measu emen s om UAV-based obse a ions emain limi ed.
This pape in oduces SUMO-UAV-Py, an open-sou ce SUMO plugin ha in eg a es
UAV-based sensing in o mic oscopic a ic simula ions in Py hon. SUMO-UAV-Py cap-
u es de ailed ehicle obse a ions by dynamically employing mul iple UAVs o obse e
a ic measu emen s based on hei posi ion and ield-o - iew (FoV). Pe o mance e al-
ua ions on a mid-sized ne wo k demons a e ha SUMO-UAV-Py main ains simula ion
pe o mance compa able o s anda d pos -p ocessing me hods, con i ming i s sui abil-
i y o la ge-scale a ic moni o ing esea ch.
Keywo ds: UAV-Based Sensing, Open-Sou ce Tools, T a ic Moni o ing
1. In oduc ion
T a ic simula ion is an essen ial ool o analyzing and imp o ing mode n anspo a ion
sys ems. As u ban en i onmen s g ow mo e complex, he need o accu a e and scal-
able a ic moni o ing solu ions has become inc easingly e iden . Among eme ging
sensing echnologies, Unmanned Ae ial Vehicles (UAVs) ha e gained signi ican a en-
ion o hei abili y o moni o oad ne wo ks om an ae ial pe spec i e using onboa d
came as. UAV-based sensing has been widely explo ed o a a ie y o anspo a ion-
ela ed applica ions, including a ic moni o ing, con ol, and da a collec ion [1], [2], [3].
Howe e , despi e hei po en ial, conduc ing eal-wo ld UAV expe imen s can be chal-
lenging due o ope a ional cons ains and egula o y es ic ions, making simula ion-
based app oaches an a ac i e al e na i e o he de elopmen and alida ion o UAV-
based a ic sensing.
UAV-based mic oscopic simula ion can p o ide ine-g ained a ic da a ha many
exis ing s udies cu en ly lack. Fo ins ance, some wo ks on UAV-based a ic s a e
es ima ion ely solely on mac oscopic measu emen s (e.g., a e age densi y and speed)
[4], [5], hus missing he oppo uni y o cap u e indi idual ehicle in e ac ions. Such
65
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
e o s could bene i om a UAV-based mic oscopic amewo k, which enables highe -
esolu ion da a collec ion o mo e accu a e alida ion in a ic moni o ing app oaches.
Despi e he g owing body o wo k on UAV-based a ic es ima ion, ew a ic sim-
ula o s amewo ks explici ly ocus on in eg a ing UAV sensing a a mic oscopic le el
wi hin la ge-scale anspo a ion simula ions. SUMO (Simula ion o U ban MObili y)[6]
is a well-es ablished, open-sou ce a ic simula o designed o mic oscopic modeling
o complex oad ne wo ks. I s open-sou ce na u e is po en ially sui able o UAV e-
sea ch, as i enables cus omizable da a ex ac ion and manipula ion h ough ools like
he T aCI in e ace.
To e ec i ely enable UAV ope a ions, in SUMO, we p esen a new open-sou ce
plugin, SUMO-UAV-Py, ha ex ends SUMO’s capabili ies o inco po a e UAV-based
sensing. The plugin’s goal is o simula e he ae ial an age poin and da a collec ion
p ocesses o UAV a ic moni o ing in an e icien and lexible manne . While SUMO-
UAV-Py does no explici ly ocus on high- ideli y UAV ligh physics, i s modula a chi-
ec u e allows he in eg a ion o ex e nal modules o ex ended lib a ies o accommo-
da e ad anced ae odynamic and en i onmen al models. The plugin enables mul iple
UAVs o be eely posi ioned abo e he oad ne wo k, collec ing de ailed ehicle-le el
in o ma ion and hus allowing esea che s o e alua e he pe o mance o UAV-based
algo i hms be o e deploying hem in eal-wo ld scena ios.
Ou con ibu ions in his pape include he de elopmen o SUMO-UAV-Py, an
open-sou ce and ligh weigh plugin o SUMO ha enables UAV-based mic oscopic
a ic obse a ions. SUMO-UAV-Py o e s a lexible, modula amewo k ha suppo s
dynamic UAV posi ioning while cap u ing de ailed ehicle-le el in o ma ion du ing sim-
ula ions, p o iding a obus pla o m o ae ial a ic moni o ing.
The emainde o his pape is o ganized as ollows. Sec ion 2 e iews exis ing
so wa e and simula ion amewo ks ha combine UAVs and a ic ne wo ks, highligh -
ing he gap add essed by ou p oposed plugin. Sec ion 3 de ails he a chi ec u e and
design p inciples o SUMO-UAV-Py, while Sec ion 4 p esen s he plugin’s use op ions
and ea u es. Sec ion 5 p esen s an expe imen al use case and pe o mance esul s.
Finally, Sec ion 6 concludes he pape and ou lines u u e wo k di ec ions.
2. Rela ed Wo k
A a ie y o simula ion pla o ms and co-simula ion s a egies ha e been p oposed o
s udy UAVs in ei he au onomous na iga ion o b oade a ic managemen con ex s.
High- ideli y en i onmen s like Ai Sim [7] le e age he Un eal Engine o simula e eal-
is ic isual and physical condi ions o au onomous ehicles, while o he en i onmen s
couple SUMO and Ai Sim o explo e lane-based Unmanned T a ic Managemen (UTM)
concep s [8]. Simila ly, UTSim [9] a ge s UAV ai a ic con ol and communica ion as-
pec s, and comme cial a ic simula o s such as PTV Vissim ha e been adap ed o
3D UAV isualiza ion [10].
In he con ex o open-sou ce UAV simula ion, Ua Sim [11] p o ides a ligh weigh
pla o m o compa ing mul i-UAV pa h planning algo i hms, and [12] in oduces a mod-
ula UAV simula ion amewo k ha p o ides kinema ic and ene gy models o mul iple
UAVs. Ano he s udy [13], explo es UAV swa m na iga ion in u ban en i onmen s, us-
ing sensing o eloci y and posi ion es ima ion wi hin he swa m while [14] in oduces
a co-simula ion amewo k o UAV physics and wi eless communica ion modeling. In
con as , he wo k in [15] aims o simula e a swa m o d ones o a ic moni o ing in
he con ex o a Sma Ci y de eloped in Uni y game engine.
66
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
While hese ools and amewo ks p o ide obus solu ions o UAV ligh dynamics,
swa m coo dina ion, o communica ion modeling, hey gene ally do no p io i ize o
ully implemen UAV-based g ound a ic sensing a a ehicle-le el scale. In con as ,
SUMO-UAV-Py aims o ill his gap by in eg a ing a ligh weigh , open-sou ce ex ension
o a mic oscopic UAV-based a ic obse a ion.
In addi ion o he UAV simula ion amewo ks discussed abo e, se e al es ablished
mic oscopic a ic simula o s a e widely used o modeling u ban a ic. Comme -
cial p oduc s such as Aimsun p o ide de ailed mic oscopic simula ion and analysis,
including suppo o pedes ians, bicycles, and ehicles exhibi ing non-lane-based be-
ha io [16]. Also, VISSIM o e s a mul imodal a ic simula ion en i onmen wi h a
use - iendly in e ace and high-quali y 3D isualiza ion [17]. Among he open-sou ce
op ions, MITSIMLab is a mic oscopic a ic simula ion model ha assesses he e ec s
o di e en a ic managemen sys em designs on ope a ional pe o mance [18], and
MATSim suppo s la ge-scale, agen -based simula ions [19]. SUMO is an open-sou ce
mic oscopic a ic simula o unde ac i e de elopmen ha is highly con igu able and
p o ides di ec da a access ia he T aCI in e ace, making i pa icula ly well-sui ed o
in eg a ing ou UAV-based sensing plugin and achie ing de ailed, eal- ime mic oscopic
a ic obse a ions.
Th ough di ec access o SUMO’s in e nal s a es ia T aCI, SUMO-UAV-Py aims
o suppo eal- ime ehicle-le el da a collec ion, while allowing use s o modi y UAV
ajec o ies dynamically. Mo eo e , i s modula design acili a es coupling wi h ex e nal
simula o s o ad anced UAV models, ensu ing lexibili y o esea che s who equi e
e ined ae odynamics, pa h planning, o en i onmen al condi ions.
3. Plugin A chi ec u e
SUMO-UAV-Py is a Py hon-based plugin designed o in eg a e SUMO’s a ic simu-
la ion wi h UAV-based sensing. I accep s use -de ined d one pa ame e s, launches
SUMO wi h he speci ied ne wo k con igu a ion, and gene a es de ailed ae ial obse -
a ions o ehicula a ic a each simula ion s ep. In his sec ion, we examine SUMO-
UAV-Py’s implemen a ion, including i s communica ion amewo k, he implemen ed
UAV mo ion model, and ou isualiza ion echnique.
3.1 Implemen a ion O e iew
Figu e 1. High-le el o e iew o SUMO-UAV-Py: ini ial pa ame e s om con igu a ion iles a e used o
ini ialize SUMO and SUMO-UAV-Py, which communica es wi h SUMO-UAV-Py, o allow he
communica ion be ween he wo en i ies. The ou pu o SUMO-UAV-Py is s o ed as a ile con aining
Mic oscopic UAV Obse a ions.
67
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
Figu e 1 p esen s a high-le el diag am o SUMO-UAV-Py’s a chi ec u e. Two con-
igu a ion iles se e as he ini ial inpu s: he SUMO con igu a ion ile, which p o ides
he ne wo k opology and simula ion pa ame e s, and he SUMO-UAV-Py con igu a ion
ile, which speci ies all UAV- ela ed se ings, including he o al numbe o UAVs, hei
speed, o a ion speed, and ield-o - iew (FoV) angles. SUMO supplies ne wo k and
a ic da a o SUMO-UAV-Py, while SUMO-UAV-Py sends upda ed UAV posi ions and
o ien a ions back o SUMO. The plugin hen eco ds all d one-based obse a ions in
an ou pu ile.
The UMO-UAV-Py con igu a ion ile includes a lis o 5D waypoin s o each UAV,
S
gi en by , x, y, z, ϕ . He e, is he simula ion ime in which he UAV s a s i s mo e-
men o a posi ion (x, y z), wi h a yaw o ien a ion ϕ.
The ou pu ile con ains a ibu es ha mimic eal-wo ld UAV senso da a o each
simula ion s ep, as de ailed in Table 1.
Table 1. Ou pu File A ibu es.
A ibu e Name Type Desc ip ion
S ep Numbe Cu en simula ion s ep numbe .
Seconds Numbe Simula ion ime in seconds.
UAV ID S ing Unique iden i ie o each UAV.
UAV Pos (x, y, z, ϕ)4D coo dina es o each UAV.
VehicleID S ing IDs o he de ec ed ehicle.
Veh Pos (x, y)2D coo dina es o each de ec ed ehicle.
Veh Speed Numbe Speed o each de ec ed ehicle.
3.2 Communica ion
The co e o SUMO-UAV-Py’s implemen a ion elies on a TCP-based communica ion
loop be ween SUMO, T aCI, and he SUMO-UAV-Py Py hon sc ip , as illus a ed in
Figu e 2. A each simula ion s ep, SUMO-UAV-Py compu es he UAV’s new posi ion,
sends i o SUMO ia T aCI, and e ie es he lis o ehicles ha all wi hin he UAV’s
calcula ed FoV. This connec ion enables con inuous moni o ing o bo h UAV mo ion and
ne wo k-wide ehicle s a es. To op imize pe o mance, SUMO-UAV-Py uses T aCI’s
objec a iable subsc ip ion o speci y a se o a iables (e.g., posi ion, speed) ha
should be au oma ically e u ned in each s ep, a oiding he o e head o indi idually
que ying e e y ehicle.
3.3 Mo ion Model
To de ine each UAV’s ajec o y, SUMO-UAV-Py eads an ini ial 5D waypoin
0, x0, y0, z0, ϕ0and he nex 5D waypoin , x, y, z, ϕ. A simula ion ime , he
UAV o a es om ϕ0 o align wi h he di ec ion o (x, y, z), hen mo es om (x0, y0, z0)
owa d (x, y, z). Upon a i al, he UAV pe o ms a inal o a ion o ma ch he yaw angle
ϕ. This sequence o mo es ” o a e-mo e- o a e” p o ides a basic ye lexible app oach
o mos UAV pa h scena ios.
In SUMO-UAV-Py he ansi ions be ween waypoin s ollow a ligh weigh linea ap-
p oach model. Gi en wo 3D poin s x1, y1, z1and x2, y2, z2, he a el dis ance is
compu ed as he Euclidean dis ance
d=p(x2−x1)2+ (y2−y1)2+ (z2−z1)2.(1)
68
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
Figu e 2. Sequence diag am o TCP communica ion among SUMO and SUMO-UAV-Py, using T aCI.
The numbe o s eps equi ed o each he nex waypoin , wi h a simula ion s ep leng h
∆ and ho izon al speed , is calcula ed using
Nmo e =d
∆ .(2)
Simila ly, o calcula e he numbe o o a ion s eps needed o co e a yaw-angle di e -
ence ∆ϕwi h yaw a e ωwe use
N o a e =∆ϕ
ω∆ .(3)
Fo mo e complex ligh pa hs he plugin is in eg a ed wi h a disc e e mo emen op-
ion ha o e ides he buil -in dynamics, in which use s manually supply 5D waypoin s
( , x, y, z, ϕ), enabling seamless coupling wi h ex e nal ajec o y planne s.
Fu he mo e, SUMO-UAV-Py is able o de e mine he ehicles obse able o each
deployed UAV by calcula ing hei came a FoV. The came a is ixed a a 90-deg ee
angle ela i e o he g ound, yielding a ec angula p ojec ion whose wid h and heigh
scale linea ly wi h al i ude h. As illus a ed in Fig. 3, o came a angles α(ho izon al)
and β( e ical), he hal - iew angen geome y gi es
FoVx= 2 h an
α
2, FoVy= 2 h an
β
2,(4)
which a e used o compu e he FoV ec angle.
The compu ed dimensions F oVxand F oVyde ine a ec angle cen e ed a he UAV’s
cu en posi ion (xu, yu), o ien ed acco ding o he UAV’s yaw angle ϕ. To e alua e
which ehicles a e isible du ing each simula ion s ep, SUMO-UAV-Py compu es he
global coo dina es o he ec angle’s co ne s a e applying a plana o a ion a ound he
UAV cen e .
To compu e he o a ed posi ion o a FoV co ne , he poin (x, y), de ined in he
local axis-aligned FoV ame, is i s ansla ed so ha he UAV cen e (xu, yu)is mo ed
69
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
Figu e 3. Rec angula FoV ep esen a ion o a 90◦came a angle a heigh h.
o he o igin. A o a ion by he yaw angle ϕis hen applied using a 2D o a ion ma ix.
A e he o a ion, he poin is ansla ed back o i s posi ion ela i e o he UAV in he
SUMO coo dina e ame.
This esul s in he ollowing ans o ma ion:
x′
y′!=R(ϕ)"x
y−xu
yu#+xu
yu,(5)
whe e R(ϕ)is he 2D o a ion ma ix de ined as:
R(ϕ) = cos ϕ−sin ϕ
sin ϕcos ϕ.(6)
He e, (x, y)a e he local coo dina es o a FoV co ne , and (x′, y′)a e he co e-
sponding coo dina es in he global SUMO ame. This p ocedu e is applied o all ou
co ne s o he FoV ec angle o compu e i s o a ed p ojec ion on he plane. Vehi-
cles whose posi ions all wi hin his ans o med ec angle a e ma ked as isible and
eco ded, based on da a e ie ed ia T aCI subsc ip ions a each simula ion s ep.
3.4 Visualiza ion
When SUMO uns in g aphical mode, SUMO-UAV-Py isually ep esen s each UAV
and i s FoV by d awing polygons and poin s-o -in e es (POIs). The ec angula FoV
polygon mo es and o a es wi h he UAV’s posi ion (x, y, z)and yaw angle ϕ. Also, i
changes size ela i e o he UAV’s al i ude. The UAV posi ion is ma ked wi h a POI
icon. Figu e 4 shows an example sc eensho o hese polygons and POIs in SUMO’s
GUI.
70
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
Figu e 4. Example sc eensho o SUMO-UAV-Py’s polygon- and POI-based isualiza ion in SUMO’s
GUI.
4. Use Op ions and Fea u es
The plugin o e s a wide ange o use -o ien ed ea u es ha maximize lexibili y and
ease o con igu a ion. In addi ion o he abili y o bypass UAV g aphics o pe o mance-
o ien ed expe imen s, use s can ailo mul iple aspec s o he simula ion o i speci ic
esea ch scena ios. These op ions can be selec ed ia he con igu a ion ile.
4.1 UAV Model Pa ame e s
The UAV dynamics gene a ed by SUMO-UAV-Py a e desc ibed by a ligh weigh linea
mo ion model as s a ed in Sec ion 3.3. Howe e in espec o ealis ic modeling o
speed, FoV cha ac e is ics and ba e y li e app oxima ions SUMO-UAV-Py comes wi h
wo p ese op ions o well-known comme cial UAV models: Ma ic 2e [20] and Mini 3
p o [21]. Fo u he cus omiza ion, he plugin o e s a Manual op ion, which allows
use s o speci y cus om alues o maximum speed, yaw a e, and FoV dimensions.
The op ional ba e y mode can be enabled, o assign a ini e ligh du a ion o each
UAV. When ba e y le els d op o i e minu es emaining, a wa ning is issued, and i
deple ed, he UAV ceases ope a ion. This ea u e is essen ial o simula ing ealis ic
mission cons ain s.
4.2 UAV Obse a ion Mode
SUMO-UAV-Py suppo s h ee obse a ion modes:
•Ho e ing: The UAV o a es and mo es owa d he nex waypoin , mo ing wi h a
uni o m speed. When i eaches he desi ed waypoin , i ho e s un il i s ime o
mo e o he nex waypoin . Ho e ing mode collec s da a con inuously ega dless
o being s a iona y o in mo ion.
•Sampling: The UAV ollows he same mo ion pa e n as in Ho e ing mode, bu
eco ds da a only when s a iona y, hus cap u ing isola ed sampling poin s.
•Spinning: Upon eaching a waypoin , he UAV o a es in place o maximize i s
FoV by co e ing a ull o pa ial o a ion. This mode is pa icula ly use ul when a
wide-angle obse a ion is desi ed as illus a ed in igu e 5.
4.3 Real- ime in e ac ions
In pa allel wi h he T aCI loop, SUMO-UAV-Py implemen s a sepa a e h eaded mecha-
nism o handle eal- ime in e ac ions o modi y he exis ing UAV ligh pa hs o add new
71
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
Figu e 5. FoV co e age isualiza ion in he 2D plane in he Spinning UAV Obse a ion Mode o
di e en simula ion s eps.
ones. This h ead allows use s o submi new waypoin s in eal- ime (ei he h ough a
local in e ace o an op ional emo e connec ion). Such in e ac ions immedia ely upda e
he d one’s ajec o y wi hou pausing he simula ion, allowing he use s o SUMO-UAV-
Py o implemen hei own open-loop o closed-loop logic, o dynamically adjus UAV
posi ions as he simula ion p oceeds. The local in e ace o eal- ime modi ica ions i s
shown in igu e 6.
Figu e 6. SUMO-UAV-Py in e ace o eal- ime UAV upda es.
4.4 GUI Dialog
SUMO-UAV-Py p o ides a GUI dialog ha includes all use -de ined pa ame e s in a
single in e ace. This in e ace allows use s o con igu e he o al numbe o UAVs, se-
lec ligh modes, se ba e y op ions, and speci y he ne wo k and SUMO con igu a ion
ile among o he s. Figu e 7 shows he dialog window, illus a ing how a ious simula ion
and UAV se ings a e easily accessible in one place.
72
Tsiou is e al. |SUMO Con P oc 6 (2025) ”SUMO Use Con e ence 2025”
Figu e 7. GUI dialog o SUMO-UAV-Py con igu a ion and use op ions selec ion.
5. Pe o mance E alua ion
To e alua e SUMO-UAV-Py’s compu a ional o e head, we conduc ed a benchma k us-
ing a eal-wo ld a ic scena io om he ci y o Bologna [22] (A.Cos a scena io om
iTe is). The scena io includes 8,600 ehicles, 179 edges, 112 nodes, 267 lanes and
a o al lane leng h o 33.52 km. All simula ions co e ed h ee hou s o i ual ime and
we e execu ed on a machine wi h a 12 h Gen In el®Co eTM i9-12900K p ocesso and
32 GB o RAM.
We compa ed wo app oaches o cap u ing ehicle-le el measu emen s o a ic:
•SUMO-UAV-Py Logging (No GUI): SUMO-UAV-Py eco ds all ehicles wi hin
each UAV’s FoV in eal ime. The numbe o UAVs was a ied om 0 o 32 ac oss
di e en uns.
•Pos -Analysis Pa sing: SUMO was un in an iden ical se up, gene a ing loa ing
ca da a ( cd-ou pu ) ile ins ead o he SUMO-UAV-Py implemen a ion. A Py hon
sc ip using he lib a y xml.e ee.Elemen T ee hen pa sed he ile o iden i y he
same ehicles ha would ha e been obse ed by he UAV(s).
Each scena io was execu ed i e imes o each UAV coun (0 o 32) unde iden ical
condi ions. We eco ded he o al simula ion ime, including pos -p ocessing in he
second app oach, and hen compu ed a e age un imes and s anda d e o s. Figu e 8
shows he esul s, whe e he “Plugin” line co esponds o ae ial measu emen s logged
by SUMO-UAV-Py and he “Pos -Analysis” line ep esen s o line pa sing me hod.
Impo an ly, he UAV=0 case e lec s he base simula ion pe o mance o SUMO
wi hou any UAV logging. In his case, SUMO-UAV-Py simply ini ializes bu pe o ms
no addi ional compu a ions. This allows us o measu e he plugin’s o e head ela i e o
SUMO’s base pe o mance. In he igu e, a ho izon al line indica es his base SUMO
iming as a e e ence.
Fo he case o ze o UAVs, SUMO-UAV-Py uns as e han he pos -analysis ap-
p oach, because he la e equi es addi ional p ocessing o w i e and expo he loa -
ing ca da a ile. As he numbe o UAVs inc eases, he un imes o bo h me hods
emain compa able, and he e a e ins ances in which SUMO-UAV-Py ou pe o ms he
pos -analysis sc ip . This esul is pa ly due o SUMO-UAV-Py logging only hose
ehicles wi hin each UAV’s FoV, while he pos -analysis me hod p ocesses he en i e
ne wo k ile. O e all, he ex a compu a ional cos o SUMO-UAV-Py’s eal- ime da a
73
