Ac as de las XV Jo nadas
de Ingenie ía Telemá ica
(JITEL 2021),
A Co uña (España),
27-29 de oc ub e de 2021.
This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0)
Fede a ed lea ning o sma cha ging o
connec ed elec ic ehicles
Yaqoob Al-Zuhai i, P ashan h Kannan, M´
onica Aguila Iga ua
Depa men o Ne wo k Enginee ing
Uni e si a Poli ´
ecnica de Ca alunya (UPC), Ba celona, Spain
yaqoob[email p o ec ed], [email p o ec ed], [email p o ec ed]
Due o ising conce ns o e clima e change, ai
pollu ion and clean ene gy awa eness, he demand
o elec ic ehicles (EVs) and enewable ene gy
gene a ion has inc eased in ecen yea s. The main
objec i e o his esea ch is o design a decen al-
ized sma cha ging coo dina ion amewo k o
EVs based on ede a ed lea ning (FL) algo i hms
in o de o p o ide an accep able collabo a i ely
lea ning model wi h p i acy p ese a ion o EVs,
imp o e cha ging scena ios, con ibu e o sma
g id s abiliza ion, mee EVs ene gy equi emen s
whe e e and whene e hey eques , and gain
wel a e o EV owne s. Mo eo e , FL is in oduced
wi h he goal o b inging machine lea ning (ML)
down o he edge le el in ehicula ne wo ks.
Ul ima ely, a mul ime ic ou ing p o ocol is also
used o p edic he bes ou e o ansmi ing
messages among EVs, in as uc u es, cha ging
s a ions (CSs), and cen al se e s.
Key wo ds—FL, Edge Compu ing, VANET, V2X, Mul i-
me ic Rou ing P o ocol
I. INTRODUCTION
In ecen yea s, he anspo indus y accoun s o
he bulk o g eenhouse gas emissions and pollu ion o
he en i onmen [1]. To ackle his issue, go e nmen s
a e implemen ing policies o suppo use o enewable
ene gies, dec ease he dependency on c ude oil, educe
CO2and pollu an emissions, and p omo e ansi ion o
mo e sus ainable mobili y. Cu en elec ic ehicles (EVs)
as a pa o he in elligen anspo sys ems (ITS) ha e
no iceably a ac ed subs an ial a en ion ecen ly because
hey a e e y iendly o he en i onmen as well as
pollu ion- ee ehicles. The ecen p ominen p og ess in
he cons uc ion o cha ging in as uc u e accele a es he
pene a ion o EVs in he ma ke . In 2018, he global EV
lee in he wo ld exceeded 5.1 million uni s and i is
expec ed o ise o 250 million uni s by 2030 [2].
Comme cializa ion o he i h-gene a ion (5G) com-
munica ion echnologies and he eme gence o ehicula
ne wo ks and edge compu ing can p omo e a supe io
pe o mance o he cha ging managemen [3]. In he
mean ime, u u e in elligen ehicles, which a e a he
hea o high mobili y ne wo ks, a e inc easingly equipped
wi h a wide a ie y o senso s o help he ehicle pe -
cei e he su ounding en i onmen as well as moni o
i s own ope a ional s a us in eal ime. Toge he wi h
high pe o mance compu ing and s o age de ices, hese
sensing echnologies a e ans o ming ehicles om a
simple anspo a ion acili y o a powe ul compu ing and
ne wo king hub wi h in elligen p ocessing capabili ies [4].
Howe e , p i acy conce ns a ise om he exchange o
da a wi h di e en pa ies. P i acy, in ecen yea s, has
been one o he mos impo an conce ns in ehicula
en i onmen s. Thus, EVs may ail o exchange da a among
hemsel es and o he pa ies due o p i acy es ic ions
in si ua ions whe e d i e s a e unwilling o p o ide hei
pe sonal da a due o he isk o da a misuse and leakage
[5]. In o de o sol e he a o emen ioned issue, ede a ed
lea ning (FL), which is a echnique ha enables dis ibu ed
ehicles o collabo a i ely lea n a sha ed machine lea ning
(ML) model wi hou sha ing hei aw da a, would be a
good solu ion o p ese e he p i acy o he local da a [6].
Vehicula ne wo ks, e.g., ehicula ad hoc ne wo ks
(VANETs) and cellula V2X (C-V2X), ha e gained popu-
la i y in ecen yea s. In such ne wo ks ehicles equipped
wi h wi eless communica ion de ices o m ehicula ne -
wo ks [7]. Howe e , wi h he e olu ion o echnology
and sudden g ow h in he numbe o sma ehicles, his
will esul in unp eceden ed p essu e on communica ion
in as uc u es. Mo eo e , adi ional VANET aces se e al
echnical challenges in deploymen and managemen due
o less lexibili y, less esou ces, scalabili y, poo con-
nec i i y, and inadequa e in elligence. Me ging ehicula
ne wo ks wi h edge compu ing, an eme ging pa adigm
which mo es compu ing asks and se ices om he
co e o he ne wo k edge, i.e. close o end use s, is
an app op ia e solu ion o hese ypes o challenging
ne wo ks. Vehicula ne wo ks show special ea u es such
as high ehicles’ mobili y, di icul ne wo k connec i i y,
which is specially challenging o eal- ime applica ions
212
Al-Zuhai i, Kannan, Aguila , 2021.
equi ing low la ency. The no ion o exploi ing ehicles as
in as uc u es could make he bes use o se e al unused
esou ces o ehicles o mee he e e inc easing equi e-
men in communica ion and compu a ional capabili ies [8].
FL can u ilize ehicula big da a gene a ed om a la ge
numbe o ehicles, and build a global ML model ha
could be used by any ehicle. FL can also be easily
inco po a ed wi h edge compu ing whe e he edge ehicles
p o ide an unde lying in as uc u e o FL [5][9].
Vehicula communica ions a e c ucial o exchanging
da a ei he be ween ehicles h ough ehicle- o- ehicle
(V2V) o ia ehicle- o-in as uc u e (V2I) connec ions
[10]. Al hough oad side uni s (RSUs) enla ge he ne -
wo k communica ion capaci y, hey a e eally expensi e
and di icul o ully deploy along oads, pa icula ly on
a la ge scale such as o e a whole ci y. Fu he mo e,
u u e ehicles will need o communica e wi h e e y hing
a ound hem in wha is known as ehicle- o-e e y hing
(V2X) [10]. To enable hyb id ehicula communica ions,
dedica ed sho - ange communica ion (DSRC), which is
based on IEEE 802.11p, and cellula V2X (C-V2X) can
be adop ed.
In his pape , we p esen he opics ha will be s udied in
he i s au ho s’ doc o al hesis. In sec ion II we highligh
some ela ed wo ks. Sec ion III summa ies he basics o
FL amewo k o ehicula ne wo ks. Since we conside
he p esence o EVs in ou p oposals, sec ion IV depic s
inno a i e ways o cha ge EVs, which will be aken in o
accoun in he design o ou FL amewo k. Sec ion V
lis s ou ongoing wo k, whe eas sec ion VI shows which
simula o s a e we going o use o e alua e he pe o mance
o ou p oposals. Finally, sec ion VII concludes he pape .
II. LITERATURE SURVEY
The au ho s in [11] p opose a scheme o cha ging
mo ing EVs in a wi eless ashion. An unsupe ised ML
algo i hm is used o es ima e he cu en cha ging s a us
o each EV. Mo eo e , a ou ing p o ocol using dis ance
ec o in o ma ion was used o ad e ise pa icipa ing EVs
abou hei s a e o cha ge (SoC), ehicle ID, dis ance
in o ma ion, and loca ion. Au ho s men ioned ha sha ing
such in o ma ion among EVs causes conce ns ela ed o
p i acy educ ion. Resul s in his wo k show o be eliable
in e ms o dynamic wi eless cha ging in bo h s a ic and
dynamic scena ios.
In [9], a su ey o echnical challenges, possible solu-
ions, open p oblems, and u u e esea ch di ec ions o
applying FL in ehicula ne wo ks a e discussed.
In [6], an FL app oach o lea ning on edge de ices
is s udied. The pape analyzes he e ec o pa icipa ing
clien s and he impo ance o clien selec ion s a egies in
FL models. The au ho s in es iga e he pe o mance o he
FL model ocusing on he e ec o a ious pa ame e s on
i s accu acy and aining ime, as well as compa ing he
pe o mance o a adi ional ML echnique.
In he pape [12], a new p oposal o a no el p obabilis ic
mul ime ic ou ing p o ocol is p esen ed. The p oposal
akes be e o wa ding decisions ha gua an ee he packe
deli e ing o des ina ion wi h he highes p obabili y, while
keeping he a e age packe delay low. Fou designed
me ics a e conside ed in his a icle (dis ance o des ina-
ion, ehicles’ densi y, posi ion o ehicle, and a ailable
bandwid h). OMNeT++, VEINS, and SUMO a e used o
conduc simula ions in a ealis ic u ban scena io.
III. IMPLEMENTATION OF AN FL FRAMEWORK
Ins ead o sending aw da a o a cen al se e , which
is common in he adi ional cen alized ML app oach,
FL can be bene icial in ehicula en i onmen s o suppo
coope a ion o mul iple EVs wi h he cen al se e . This
way, dis ibu ed EVs will ain a pa ial model using own
local da a o ensu e p i acy p o ec ion. The global model
in he cen al se e will be upda ed om he pa ame e s’
agg ega ion o hose pa ial models.
The basic s eps o an FL amewo k a e as ollows:
•Clien selec ion: Ini ially, he cen al se e mus
speci y hose EVs ha should be in ol ed in he
model aining.
•Model dissemina ion: Once he EVs a e selec ed, he
cen al se e b oadcas s an ini ial lea ning model o
he aining o he selec ed EVs.
•Dis ibu ed lea ning: Each EV ains he model based
on i s own local da ase , and hen calcula es i s local
upda es o he global model.
•Global model agg ega ion: A e a p ede ined aining
pe iod o agg ega e he new e sion o he global
model, all EVs send only hei upda es o he com-
mon global model o ei he he cen al se e o
o an agg ega o . An EV can be designa ed as he
agg ega o looking o be close o pa icipa ing EVs.
The agg ega o EV will manage he aining p ocess
and ul ima ely send he esul o he cen al se e
[13].
•Model es ing: The cen al se e es s he agg ega ed
global model. Acco ding o he es ing esul s, he
cen al se e could une some hype -pa ame e s o
epea he aining p ocess, o con inue wi h he nex
s ep which is he model upda e.
•Model upda e: The se e upda es he sha ed model
acco ding o he agg ega ed esul om he EVs.
Then, he se e sends he upda ed global model o
he EVs.
These s eps a e epea ed un il he cen al se e achie es
a sa is ac o y global model. While designing an FL ame-
wo k, i is signi ican o selec a a ie y o pa ame e s and
compa e mul iple scena ios o a ain he ade-o s be ween
p i acy, e iciency, and accu acy in his en i onmen [5].
Fu he mo e, di e en om con en ional decen alized
ML app oaches, FL in ehicula en i onmen s is expec ed
o achie e many ad an ages, such as low communica ion
o e head, be e p i acy, la ge da a o aining, be e
e iciency, be e u ili y, and sho e esponse ime.
IV. FUTURE WAYS TO CHARGE ELECTRIC
VEHICLES
Cu en ly, he mos common o m o cha ging EVs is
ia a plug-in cha ging s a ion. Howe e , he e a e many
This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0)
213
Fede a ed lea ning o sma cha ging o connec ed elec ic ehicles
issues educing d i e s’ willingness o use EVs ins ead o
adi ional ehicles, such as limi ed numbe o cha ging
s a ions (CSs). Fu he mo e, he nea es CS is pe haps a
an incon enien si e in ela ion o an EV’s ou e. In his
poin , he d i e may be eeling anxie y o e he need o
echa ge he EV du ing a jou ney. A u he issue may
happen a e he EV eaches he plug-in CS, he d i e
may ind he cha ging slo s al eady aken by o he EVs.
These issues can e ec i ely be sol ed by adop ing a new
s uc u e o wi eless cha ging echnology o EVs based
on he magne ic esonan coupling wi eless powe ans e
echnique, which was in oduced in he las decade [11].
This echnique is used o exchange cha ge among pai ed
EVs in a s a ic posi ion as well as in dynamic posi ion
(mo ion). This s uc u e can wo k oge he wi h plug-in
cha ge EVs o ope a e independen ly in o de o inc ease
cha ging oppo uni ies o EVs.
The e a e mainly wo ways o wi eless cha ging: [14]
•Induc i e cha ging: I includes a ansmi e coil,
which is embedded in he loo o he cha ging a ea
and connec ed o he powe supply (g id), while a
ecei e coil which is embedded in he EV’s chassis
and connec ed o he EV’s ba e y.
•Wi eless ehicle- o- ehicle (V2V) cha ging: I is
based on wi eless powe ans e echnique which has
high powe ans e e iciency wi h a long ansmi -
ing dis ance. This echnology can achie e cha ging
be ween mo ing EVs.
In ou wo k, we plan o use se e al cha ging echnolo-
gies o mee equi emen s o EVs wi h a low cha ging
s a e. We will design a FL-based amewo k o assis s
he d i e in inding a sui able powe sou ce based on
he cu en ci cums ances o he en i onmen , e.g. ene gy
equi emen , dis ance o des ina ion and o possible CSs,
p icing, wai ing ime, ways o cha ging such as (mobile
CS, plug-in CS, V2V cha ging, o induc i e cha ging), and
o he ac o s. Ou FL-based amewo k will be assis ed by
V2X communica ion, SoC de ec ion, mul ime ic ou ing
p o ocol, and edge compu ing. Figu e (1) demons a es he
scena io o decen alized sma cha ging coo dina ion o
EVs based on FL.
This cha ging coo dina ion can o e solu ions o many
p oblems, such as:
1) O e ing solu ions o he issue o ha ing a limi ed
numbe o CSs.
2) Sol ing he p oblem o inc easing numbe s o EVs
in a plug-in EV ne wo k.
3) Reducing cha ging delay, and minimizing o e all
EV ene gy cos s.
4) A oiding he g id conges ion, especially in he
e enings, ha esul s om simul aneously cha ging
oo many EVs.
5) Fo ecas ing o EVs ene gy equi emen s.
6) Exchanging da a among EVs o wi h he cen al
se e while ensu ing EV’s p i acy p ese a ion.
7) Gene a ing mone a y bene i s o EV owne s.
8) Finding he cheapes cha ging op ions by gi ing an
EV d i e he op imal choice o cha ging.
Fig. 1. Decen alized sma cha ging coo dina ion amewo k o EVs
based on ede a ed lea ning (FL) algo i hms.
9) Encou aging consume s o pu chase EVs.
V. ONGOING WORK
The main con ibu ions ha we ha e planned in his
Ph.D. hesis, can be summa ized as ollows:
1) Employing mobile EVs as in as uc u es o com-
munica ion and compu a ion o achie e a be e
u iliza ion o edge esou ces.
2) P i acy o each EV can be p ese ed by using FL.
3) E icien collabo a ion amongs EVs wi h FL could
each he le el o collabo a i e in elligence. Such in-
elligen collabo a ion can con ibu e o ema kably
minimize he wai ing ime o cha ging.
4) Due o he po en ially high mobili y and
spa se/dense en i onmen s, a mul ime ic ou ing
p o ocol wi h FL can be employed o add ess he
issue o p o iding an e icien o wa ding scheme
in V2X communica ions.
5) SoC, which indica es he le el o cha ge a ailable
in each EV, is used o p edic he esidual ene gy o
EVs based on he designed FL amewo k.
6) As o cha ging ime a ailable, de e mine he bes
powe sou ce ei he ia wi eless o plug-in h ough:
•Conside ing he cha ging cos and du a ion
when selec ing ei he he bes (e.g. he nea es )
CSs o hose EVs ha ing high le el o powe o
sha e, among o he s a egies.
•Using an e icien mul ime ic ou ing p o ocol
o imp o e he da a ansmission needed in he
cha ging se ice ega ding he chosen powe
sou ce.
7) Es ima ing a el ime o sa e ene gy acco ding o
he cha ging a ailable ime and he EV’s des ina ion,
and boos he ene gy economy o EVs in cha ging
p ocesses.
8) Based on ou FL-based cha ging amewo k, educ-
ing load on CSs while keeping he g id s able can
be achie ed.
VI. IMPLEMENTATION AND SIMULATION
Because o he complexi y and he high deploymen
cos s o ehicula applica ions, i is ecommended ha
This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0)
214
Al-Zuhai i, Kannan, Aguila , 2021.
p oposals a e ex ensi ely es ed and e alua ed in possible
ways be o e being pu in o p ac ice.
Tenso Flow [15] is an open-sou ce FL lib a y ha o e s
an ex ensi e ange o algo i hms. I also wo ks in con-
junc ion wi h Ke as and Py hon. The cons uc ion o ML
models and he design o pa ame e s can be e alua ed by
compu ing me ics such as accu acy o con usion ma ix.
Th ough simula ions we will ep esen he whole sys em
us wo hy as close o eali y as possible, and we will
assess he pe o mance o he p oposed sys em.
To simula e a ehicula ne wo k in u ban scena ios,
OMNeT++, SUMO, VEINS, and A e y-C a e used.
•OMNeT++: I is a modula , objec o ien ed and
disc e e e en simula o based on C++ [16].
•Simula ion o U ban Mobili y (SUMO) ool: I is a
highly po able and open-sou ce so wa e o simula e
he mo emen o he ehicles [17].
•Vehicles in ne wo ks simula ions (VEINS) ool: I
is a amewo k o ehicula ne wo k simula ion. I
acili a es he bidi ec ional in e ac ion among SUMO
and OMNeT++ simula o s [18].
•A e y-C amewo k: A e y-C [19] is a simula ion
amewo k o he pe o mance e alua ion o Cellula
V2X p o ocols and V2X applica ions. I is an ex-
ension o he simula ion amewo k SimuLTE [20],
de eloped unde he OMNeT++ pla o m.
VII. CONCLUSIONS
In ehicula en i onmen s, i is no always easible
o ans e massi e amoun s o da a o p ocessing o
a emo e cen al se e due o limi ing ac o s, such as
uns able and limi ed wi eless connec i i y, unaccep able
la ency, and ne wo k bandwid h. Mo eo e , he in e es s
on da a a ailabili y ha e igge ed he discussion o da a
owne ship, and he conce n on da a p i acy and con i-
den iali y especially when sha ing such da a wi h cen al
se e s. In his esea ch wo k we will design a decen-
alized sma cha ging coo dina ion amewo k o EVs
based on FL aking ad an age o he dis ibu ed powe o
edge compu ing. Ou goal is o encou age collabo a ion
amongs pa icipa ing EVs in aining a global model
locally while ensu ing p i acy p ese a ion o each EV,
and hen sending i o a cen al se e o ob ain a global
model. The aim is o op imize cha ging decisions, o e
e icien EV cha ging se ice o u ban a eas when EVs a e
in a c i ical SoC, a oid g id o e load, and acqui e d i e s’
com o and sa is ac ion. As a u u e wo k, an incen i e
mechanism will be used o mo i a e EVs wi h hei p i a e
da a o pa icipa e in FL aining in o de o imp o e he
global FL model accu acy. E en ually, a d i e can also
maximize hei own u ili y h ough con ibu ion o ei he
he aining p ocess o cha ging ano he EV ha ing low
le el SoC in wi eless way.
VIII. ACKNOWLEDGEMENTS
This wo k was suppo ed by he Spanish Go e n-
men unde esea ch p ojec “Enhancing Communica-
ion P o ocols wi h Machine Lea ning while P o ec ing
Sensi i e Da a (COMPROMISE)” PID2020-113795RB-
C31/AEI/10.13039/501100011033.
REFERENCES
[1] A. Ghosh, “Possibili ies and challenges o he inclusion o he
elec ic ehicle (e ) o educe he ca bon oo p in in he anspo
sec o : A e iew,” Ene gies, ol. 13, no. 10, 2020. [Online].
A ailable: h ps://www.mdpi.com/1996-1073/13/10/2602
[2] Global EV ou look 2019 - scaling-up he ansi ion o elec ic
mobili y, 2019.
[3] S. Zhang, W. Dou, Y. Zhang, W. Hao, Z. Chen, and Y. Liu,
“A ehicle-en i onmen coope a i e con ol based eloci y p o ile
p edic ion me hod and case s udy in ene gy managemen o plug-in
hyb id elec ic ehicles,” IEEE Access, ol. 7, pp. 75 965–75 975,
2019.
[4] W. Xu, H. Zhou, N. Cheng, F. Lyu, W. Shi, J. Chen, and X. Shen,
“In e ne o ehicles in big da a e a,” IEEE/CAA Jou nal o Au o-
ma ica Sinica, ol. 5, no. 1, pp. 19–35, 2018.
[5] V. Mugun han, A. Pe ai e-Bueno, and l. Kagal, “P i acy l: A
simula o o p i acy-p ese ing and secu e ede a ed lea ning,” in
CIKM ’20: The 29 h ACM In e na ional Con e ence on In o ma ion
and Knowledge Managemen , 2020.
[6] S. F. Lameh, W. Noble, Y. Amannejad, and A. A sha , “Analysis
o ede a ed lea ning as a dis ibu ed solu ion o lea ning on edge
de ices,” in 2020 In e na ional Con e ence on In elligen Da a
Science Technologies and Applica ions (IDSTA), 2020, pp. 66–74.
[7] M. S. Sheikh, J. Liang, and W. Wang, “Secu i y and p i acy in
ehicula ad hoc ne wo k and ehicle cloud compu ing: A su ey,”
Wi eless Communica ions and Mobile Compu ing, ol. 2020, no.
5129620, 2020.
[8] X. Hou, Y. Li, M. Chen, D. Wu, D. Jin, and S. Chen, “Vehicula og
compu ing: A iewpoin o ehicles as he in as uc u es,” IEEE
T ansac ions on Vehicula Technology, ol. 65, no. 6, pp. 3860–
3873, 2016.
[9] Z. Du, C. Wu, T. Yoshinaga, K.-L. A. Yau, Y. Ji, and J. Li, “Fede -
a ed lea ning o ehicula in e ne o hings: Recen ad ances and
open issues,” IEEE Open Jou nal o he Compu e Socie y, ol. 1,
pp. 45–61, 2020.
[10] M. Boban, A. Kousa idas, K. Manolakis, J. Eichinge , and W. Xu,
“Connec ed oads o he u u e: Use cases, equi emen s, and design
conside a ions o ehicle- o-e e y hing communica ions,” IEEE
Vehicula Technology Magazine, ol. 13, no. 3, pp. 110–123, 2018.
[11] M. Adil, J. Ali, Q. T. H. Ta, M. A ique, and T.-S. Chung, “A
eliable senso ne wo k in as uc u e o elec ic ehicles o enable
dynamic wi eless cha ging based on machine lea ning echnique,”
IEEE Access, ol. 8, pp. 187 933–187 947, 2020.
[12] Le icia Lemus C´
a denas, Ahmad M. Mezhe , Pablo A. Ba becho
Bau is a, M´
onica Aguila Iga ua, “A p obabili y-based mul ime ic
ou ing p o ocol o ehicula ad hoc ne wo ks in u ban scena ios,”
IEEE Access, ol. 7, pp. 178 020–178 032, 2019.
[13] J. Posne , L. Tseng, M. Aloqaily, and Y. Ja a weh, “Fede a ed
lea ning in ehicula ne wo ks: Oppo uni ies and solu ions,” IEEE
Ne wo k, ol. 35, no. 2, pp. 152–159, 2021.
[14] X. Mou, R. Zhao, and D. T. Gladwin, “Vehicle o ehicle cha ging
(V2V) bases on wi eless powe ans e echnology,” in IECON
2018 - 44 h Annual Con e ence o he IEEE Indus ial Elec onics
Socie y, 2018, pp. 4862–4867.
[15] “Tenso Flow,” h ps://www. enso low.o g/ ede a ed/ ede a ed
lea ning.
[16] “OMNeT++,” h ps://omne pp.o g/in o, accessed: 2020-10-06.
[17] “SUMO,” h ps://www.eclipse.o g/sumo/.
[18] “Veins,” h p:// eins.ca 2x.o g/.
[19] A. Hegde and A. Fes ag, “A e y-C: An OMNeT++ based disc e e
e en simula ion amewo k o cellula V2X,” in MSWiM ’20:
23 d In e na ional ACM Con e ence on Modeling, Analysis and
Simula ion o Wi eless and Mobile Sys ems, 2020.
[20] A. Vi dis, G. Na dini, and G. S ea, “Modeling unicas de ice- o-
de ice communica ions wi h simuLTE,” in 2016 1s In e na ional
Wo kshop on Link- and Sys em Le el Simula ions (IWSLS), 2016,
pp. 1–6.
This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0)
215
