AIOTI Strategic Reseach and Innovation Agenda - Advancing Next-Generation IoT and Edge Computing Research and Innovation

© AIOTI. All igh s ese ed.
© AIOTI. All igh s ese ed.
2
Execu i e Summa y
The Alliance o IoT and Edge Compu ing Inno a ion (AIOTI) S a egic Resea ch and Inno a ion
Agenda (SRIA) aims bo h a iden i ying key In e ne o Things (IoT) and edge compu ing
echnologies and applica ions esea ch p io i ies and a p o iding a ision on how he u u e o
he IoT domain will look like, up un il he 2030 ime ame.
The AIOTI SRIA is a oadmap (2023-2030) o u u e IoT and edge compu ing esea ch and
inno a ion ac ions in Eu ope, p oposing speci ic hemes, sub- hemes, and p io i ies, which help
iden i ying gaps and a eas whe e esea ch and inno a ion ad ancemen s a e mos needed.
The iden i ied esea ch p io i ies o m he e e ence o conc e e ac ions o be implemen ed in
di e en esea ch p og ammes by a ious s akeholde s, such as indus y, esea che s, small-and-
medium en e p ises, academia, en ep eneu s, he public sec o , and he o e all socie y.
The SRIA deli e s a di ec con ibu ion in aligning wi h he Uni ed Na ions Sus ainable
De elopmen Goals (SDG) needs and he Eu opean G een Deal objec i es, while de eloping
in e na ional coope a ion o sol e global challenges using IoT and cloud compu ing
echnologies.
The esea ch and inno a ion e o s in IoT and edge compu ing equi e o be based on a long-
e m p og amming app oach ha p o ides con inui y ac oss echnology and applica ions
e o s o e se e al yea s. The AIOTI SRIA in oduces a mission-o ien ed esea ch and inno a ion
app oach ha answe s socie al and ma ke needs, main ains, and ex ends indus ial leade ship,
p o ec s he en i onmen , ensu es secu i y, p i acy, sa e y, and ene gy e iciency solu ions, while
p io i ising esea ch and inno a ion capabili ies and educa ion.
The AIOTI SRIA add esses esea ch and inno a ion p io i ies o u u e IoT and edge compu ing
echnologies and applica ions ha will d i e changes ac oss indus ial sec o s, he Eu opean
economy, and socie y. The p io i ies include con e gence wi h nex -gene a ion Tac ile IoT,
decen alised and dis ibu ed a chi ec u es, IoT knowledge-d i en edge p ocessing, a i icial
in elligence (AI) and us wo hiness.
The IoT and edge compu ing echnologies ha e e ol ed apidly in add essing he g and
challenge o de eloping human cen ed IoT echnologies and applica ions, which equi e
inc eased communica ion and coo dina ion amongs policy make s, end-use s, and expe s in
he IoT and edge compu ing ields.
The ex ensi e use o IoT and edge compu ing in di e en indus ial sec o s and he mo e om
cloud o edge p ocessing mus be accompanied by new dis ibu ed a chi ec u es and end- o-
end (E2E) IoT secu i y. The con e gence o connec i i y, IoT, edge compu ing, AI, and
Dis ibu ed Ledge Technologies (DLT) will be essen ial o nex -gene a ion In e ne applica ions
and ad ancemen s.
The opics p esen ed in he AIOTI SRIA a e aligned wi h he issues add essed by o he Eu opean
pa ne ships o imp o e he Eu opean IoT ecosys em's sus ainabili y and manage human well-
being, pa icula ly o de eloping sa e, secu e, and us wo hy IoT and edge compu ing
echnologies.
© AIOTI. All igh s ese ed.
3
Table o Con en
3.1 Technological de elopmen s ....................................................................................................................... 16
3.2 Main T ends, Issues and Challenges............................................................................................................. 19
3.3 Resea ch P io i ies Timeline............................................................................................................................ 20
4.1 Technological de elopmen s ....................................................................................................................... 21
4.2 Main T ends, Issues and Challenges............................................................................................................. 22
4.3 Resea ch P io i ies Timeline............................................................................................................................ 23
5.1 Technological de elopmen s ....................................................................................................................... 24
5.2 Main T ends, Issues and Challenges............................................................................................................. 25
5.3 Resea ch P io i ies Timeline............................................................................................................................ 28
6.1 Technological de elopmen s ....................................................................................................................... 29
6.2 Main T ends, Issues and Challenges............................................................................................................. 30
6.3 Resea ch P io i ies Timeline............................................................................................................................ 31
7.1 Technological de elopmen s ....................................................................................................................... 32
7.2 Main T ends, Issues and Challenges............................................................................................................. 33
7.3 Resea ch P io i ies Timeline............................................................................................................................ 35
8.1 Technological de elopmen s ....................................................................................................................... 36
8.2 Main T ends, Issues and Challenges............................................................................................................. 38
8.3 Resea ch P io i ies Timeline............................................................................................................................ 38
© AIOTI. All igh s ese ed.
4
9.1 Technological de elopmen s ....................................................................................................................... 39
9.2 Main T ends, Issues and Challenges............................................................................................................. 40
9.3 Resea ch P io i ies Timeline............................................................................................................................ 42
10.1 Technological de elopmen s ....................................................................................................................... 43
10.2 Main T ends, Issues and Challenges............................................................................................................. 45
10.3 Resea ch P io i ies Timeline............................................................................................................................ 46
11.1 Technological de elopmen s ....................................................................................................................... 47
11.2 Main T ends, Issues and Challenges............................................................................................................. 48
11.3 Resea ch P io i ies Timeline............................................................................................................................ 49
12.1 Technological de elopmen s ....................................................................................................................... 50
12.2 Main T ends, Issues and Challenges............................................................................................................. 51
12.3 Resea ch P io i ies Timeline............................................................................................................................ 53
13.1 Technological de elopmen s ....................................................................................................................... 54
13.2 Main T ends, Issues and Challenges............................................................................................................. 55
13.3 Resea ch P io i ies Timeline............................................................................................................................ 56
14.1 Technological de elopmen s ....................................................................................................................... 57
14.2 Main T ends, Issues and Challenges............................................................................................................. 57
14.3 Resea ch P io i ies Timeline............................................................................................................................ 59
15.1 Technological de elopmen s ....................................................................................................................... 60
15.2 Main T ends, Issues and Challenges............................................................................................................. 61
15.3 Resea ch P io i ies Timeline............................................................................................................................ 62
16.1 Technological de elopmen s ....................................................................................................................... 63
16.2 Main T ends, Issues and Challenges............................................................................................................. 64
16.3 Resea ch P io i ies Timeline............................................................................................................................ 65
© AIOTI. All igh s ese ed.
5
17.1 Technological de elopmen s ....................................................................................................................... 66
17.2 Main T ends, Issues and Challenges............................................................................................................. 66
17.3 Resea ch P io i ies Timeline............................................................................................................................ 67
18.1 Technological de elopmen s ....................................................................................................................... 68
18.2 Main T ends, Issues and Challenges............................................................................................................. 69
18.3 Resea ch P io i ies Timeline............................................................................................................................ 71

© AIOTI. All igh s ese ed.
6
Table o Figu es
Figu e 1 Edge IoT esea ch and inno a ion opics ....................................................................................................... 12
Figu e 2 Edge compu ing g anula i y ac oss he compu ing con inuum .................................................................. 17
Figu e 3 Edge cloud con inuum ep esen a ion ........................................................................................................... 18
Figu e 4 X- con inuum pa adigm ................................................................................................................................... 21
Figu e 5 Ma e and Th ead edge IoT connec i i y ...................................................................................................... 27
Figu e 6 Edge IoT digi al win echnology e olu ion ..................................................................................................... 37
Figu e 7 In e ne o hings senses .................................................................................................................................... 44
Figu e 8 Edge IoT sys em dependabili y cha ac e is ics.............................................................................................. 68
© AIOTI. All igh s ese ed.
7
Lis o Tables
Table 1 IoT and edge compu ing esea ch p io i ies ................................................................................................... 20
Table 2 IoT edge and X-Con inuum esea ch p io i ies ............................................................................................... 23
Table 3 In elligen connec i i y esea ch p io i ies ....................................................................................................... 28
Table 4 Ene gy-e icien in elligen IoT and edge compu ing sys ems esea ch p io i ies ...................................... 31
Table 5 He e ogeneous cogni i e edge IoT mesh esea ch p io i ies ........................................................................ 35
Table 6 IoT Digi al Twins, Modelling and Simula ion En i onmen s esea ch p io i ies ............................................. 38
Table 7 Swa m sys ems esea ch p io i ies .................................................................................................................... 42
Table 8 In e ne o hings senses esea ch p io i ies ..................................................................................................... 46
Table 9 Decen alised and dis ibu ed edge IoT sys ems esea ch p io i ies ............................................................ 49
Table 10 Fede a ed lea ning and AI o edge IoT sys ems esea ch p io i ies .......................................................... 53
Table 11 Ope a ing sys ems and o ches a ion concep s o edge IoT sys ems esea ch p io i ies ....................... 56
Table 12 Dynamic p og amming ools and en i onmen s o edge IoT sys ems esea ch p io i ies ...................... 59
Table 13 He e ogeneous edge IoT sys ems in eg a ion esea ch p io i ies ............................................................... 62
Table 14 Edge IoT sec o ial and C oss-Sec o ial Open Pla o ms esea ch p io i ies ............................................... 65
Table 15 IoT Ve i ica ion, Valida ion and Tes ing Me hods esea ch p io i ies .......................................................... 67
Table 16 IoT T us wo hiness and Edge Compu ing Sys ems Dependabili y esea ch p io i ies .............................. 71
© AIOTI. All igh s ese ed.
8
Lis o Ac onyms
AI A i icial In elligence
AIOp A i icial In elligence Ope a ions
AIOTI Alliance o IoT and Edge Compu ing Inno a ion
AIOTI SRIA AIOTI S a egic Resea ch and Inno a ion Agenda
ANDSF Access Ne wo k Disco e y and Selec ion Func ion
ANN A i icial Neu al Ne wo k
API Applica ion P og amming In e ace
AR Augmen ed Reali y
ASIC Applica ion-Speci ic In eg a ed Ci cui
ASSP Applica ion-Speci ic S anda d P oduc
AV Audio Video
BS Base S a ion
CBOR Concise Bina y Objec Rep esen a ion
CN Cogni i e Ne wo k
CPU Cen al P ocessing Uni
D2D De ice-De ice
DL Deep Lea ning
DLT Dis ibu ed Ledge Technologies
DMO Di ec Mode Ope a ion
DSP Digi al Signal P ocessing
DT Digi al Twin
E2E End o End
ENISA Eu opean Union Agency o Ne wo k and In o ma ion Secu i y
FL Func ion Le el
FPGA Field-P og ammable Ga e A ay
FW Fi mwa e
GPU G aphics P ocessing Uni
H2H Hos o Hos , Human o Human
HE Ho izon Eu ope
HMI Human-Machine In e ace
HW Ha dwa e
IaaS In as uc u e as a Se ice
ID Iden i ie s De ini ion
IETF OSCORE In e ne Enginee ing Task Fo ce Objec Secu i y o Cons ained REST ul En i onmen s
IoT In e ne o Things
IoT DT IoT Digi al Twin
IoTS In e ne o Things Senses
IT/OT In o ma ion/Ope a ion Technology
JSON-LD Ja aSc ip Objec No a ion o Linked Da a
© AIOTI. All igh s ese ed.
9
KPI Key Pe o mance Indica o ()
LDS Lase -Di ec S uc u ing
LoRa Long Range Radio
LPWAN Low-Powe Wide A ea Ne wo k
M2H Machine o Human
M2M Machine o Machine
MEC Mul i-access Edge Compu ing
ML Machine Lea ning
MLOp Machine Lea ning Ope a ions
mMTC Machine Type Communica ion
MR Machine Reasoning
NPN Non-Public-Ne wo k
NPU Ne wo k P ocessing Uni
O-RAN Open Radio Access Ne wo k
OEM O iginal Equipmen Manu ac u e
OS Ope a ing Sys em
OTA O e - he-Ai
PaaS Pla o m as a Se ice
PCI Pe iphe al Componen In e connec
PLC P og ammable Logic Con olle
QoI Quali y o In o ma ion
QoS Quali y o Se ice
RAN Radio Access Ne wo k
RAT Radio Access Technology
RF Radio F equency
RISC-V educed Ins uc ion Se Compu e p inciples 5 h gene a ion
RU (Mul i-)Resou ce Uni
SCADA Supe iso y Con ol and Da a Acquisi ion
SDG Sus ainable De elopmen Goals
SDO Se ice Da a Objec s
SLO Se ice Le el Objec i e
SoC Sys em-on-a-Chip
SRE Si e Reliabili y Enginee s
SRIA S a egic Resea ch and Inno a ion Agenda
SW So wa e
TPU Tenso P ocessing Uni
TSN Time-Sensi i e Ne wo king
VR Vi ual Reali y
Wi-Fi Wi eless Fideli y
XAI Explainable AI
XR eX ended Reali y
VV&T Ve i ica ion, Valida ion and Tes ing
© AIOTI. All igh s ese ed.
16
3. IoT and Edge Compu ing G anula i y
The concep o da a p ocessing, analysis and s o age using he cen alised cloud compu ing
pa adigm, comp ising o a se o echnologies, in as uc u e, se ices, and applica ions, is no
longe aligned wi h he always inc easing demand o cellula , e.g., massi e Machine Type
Communica ion (mMTC), se ices and wi eless in elligen connec i i y usage scena ios
9
. Such
scena ios aim a ensu ing ha he always g owing numbe o IoT de ices connec ed o he
ne wo k can ope a e acco ding o hei speci ica ion, do no ope a e unde condi ions limi ing
he capabili ies, and ul il he expec ed Quali y o Se ice (QoS). A new app oach is he e o e
looked o , so o a oid c ea ing silos and issues ega ding bo h connec i i y and eal- ime da a
p ocessing and s o age.
Edge compu ing p o ides he mechanisms o dis ibu ing da a p ocessing and ede ines he IoT
landscape by mo ing da a p ocessing and analy ics a he edge by using AI/ML echniques
and ensu ing an ad anced le el o embedded secu i y.
Edge compu ing allow an e ec i e deploymen o eal- ime applica ions, conside ing ha he
p ocessing is pe o med close o he da a sou ce. I also educes he o de o magni ude o
ansmi ed da a, by no ansmi ing he ex ensi e amoun o aw da a c ea ed by IoT de ices,
a he jus sending smalle amoun o da a, hanks o s o age and local p ocessing capabili ies.
Se e al bene i s can be de i ed by his app oach, e.g., a dec ease in needed communica ion
bandwid h and da a s o age equi emen s, as well as in he da a a ack su ace, hus Imp o ing
secu i y, p i acy da a p o ec ion, and inally also educing he o e all ene gy consump ion.
Edge compu ing solu ions a e implemen ed h ough di e en deploymen o ms such as
cloudle
10
, dew
11
, mobile edge
12
, og compu ing
13
, e c., ha ha e been de eloped o e he
las ew yea s.
3.1 Technological de elopmen s
Edge compu ing is de ined as a pa adigm ha can be implemen ed using di e en
a chi ec u es buil o suppo an IoT dis ibu ed in as uc u e o da a p ocessing (signals, image,
oice, e c.) wi h edge IoT de ices ope a ing close o he poin s o collec ion (da a sou ces) and
u ilisa ion. The edge compu ing dis ibu ed pa adigm p o ides compu ing capabili ies o he IoT
nodes and de ices o he edge o he ne wo k (o edge domain) o imp o e he pe o mance
(ene gy e iciency, la ency, e c.), ope a ing cos , secu i y and eliabili y o applica ions and
se ices. Edge compu ing pe o ms da a analysis by minimising he dis ance be ween IoT nodes
and de ices and educing he dependence on cen alised esou ces ha se e hem while
minimising ne wo k hops.
9
B. Raa e al., “Key echnology ad ancemen s d i ing mobile communica ions om gene a ion o gene a ion”, in In el Technology Jou nal 18 (1), 2014.
10
D. Bha a and L. Mashayekhy, "Physics-Inspi ed Mobile Cloudle Placemen in Nex -Gene a ion Edge Ne wo ks," 2022 IEEE In e na ional Con e ence on
Edge Compu ing and Communica ions (EDGE), Ba celona, Spain, 2022, pp. 159-168, doi: 10.1109/EDGE55608.2022.00031.
11
Z. S. Ageed, S. R. M. Zeeba ee, M. A. M. Sadeeq, R. K. Ib ahim, H. M. Shuku and A. Alkhayya , "Comp ehensi e S udy o Mo ing om G id and Cloud
Compu ing Th ough Fog and Edge Compu ing owa ds Dew Compu ing," 2021 4 h In e na ional I aqi Con e ence on Enginee ing Technology and Thei
Applica ions (IICETA), Naja , I aq, 2021, pp. 68-74, doi: 10.1109/IICETA51758.2021.9717894.
12
J. Lee and W. Na, "A Su ey on Mobile Edge Compu ing A chi ec u es o Deep Lea ning Models," 2022 13 h In e na ional Con e ence on In o ma ion
and Communica ion Technology Con e gence (ICTC), Jeju Island, Ko ea, Republic o , 2022, pp. 2346-2348, doi: 10.1109/ICTC55196.2022.9952954.
13
I. Ma inez, A. S. Ha id and A. Ja ay, "Design, Resou ce Managemen , and E alua ion o Fog Compu ing Sys ems: A Su ey," in IEEE In e ne o Things
Jou nal, ol. 8, no. 4, pp. 2494-2516, 15 Feb.15, 2021, doi: 10.1109/JIOT.2020.3022699.

© AIOTI. All igh s ese ed.
17
IoT edge compu ing capabili ies include a s eady ope a ing p ocedu e ac oss di e en
pla o m in as uc u es o deli e p ocessing se ices o emo e IoT de ices, applica ion
in eg a ion, o ches a ion, and se ice deli e y equi emen s. The edge compu ing echnologies
a e mean o p ope ly conside ha dwa e (HW) limi a ions and cos cons ain s, o e ec i ely
handle limi ed o in e mi en ne wo k connec ions, and o implemen me hods o sa is y he
mos di e se equi emen se s, e.g., IoT applica ions equi ing low la ency o g ea ly di e ing in
da a a es.
Fo in elligen IoT applica ions, he edge compu ing concep is mi o ed in he de elopmen o
di e en edge compu ing le els (mic o, deep, me a), ha inco po a e he compu ing and
in elligence con inuum om he senso s/ac ua o s, p ocessing uni s, con olle s, ga eways, on-
p emises se e s o he in e ace wi h mul i-access, og, and cloud compu ing.
Edge IoT de ices and hei unc ions co e he edge compu ing, communica ion, and da a
analy ics capabili ies, so o make i possible o call such de ices sma o in elligen . An edge IoT
de ice is designed a ound he compu ing uni s (CPUs, GPUs/FPGAs, ASICs pla o ms, AI
accele a o s/p ocessing), communica ion ne wo k, s o age in as uc u e and he applica ions
(wo kloads) ha un on i .
The edge domain can scale om a ew IoT de ices o ens o housands o IoT de ices,
dis ibu ed in di e en loca ions wi h a unique iden i y. The IoT de ices in he edge compu ing
en i onmen a e physically sepa a ed and connec ed using wi eless o wi ed connec ions in
di e en kinds o opologies, e.g., by using a mesh ne wo k
14
. The IoT edge de ices can e en
ope a e semi-au onomously, especially in he case o senso s sp ead in emo e loca ions, using
emo e managemen adminis a ion ools.
The edge de ices can be op imised based on di e en aspec s, like p ocessing, memo y,
ene gy, connec i i y, size, cos . Thei pe o mance and capabili ies a e o cou se cons ained
by hese pa ame e s.
A desc ip ion o he mic o-, deep- and me a-edge concep s is p o ided in he ollowing sec ions
and aligned wi h o he Eu opean pa ne ships
15
. The g anula i y o he edge compu ing
en i onmen s is illus a ed in Figu e 2.
Figu e 2 Edge compu ing g anula i y ac oss he compu ing con inuum
The mic o-edge desc ibes he in elligen senso s, machine ision, and IoT de ices ha gene a e
insigh da a and a e implemen ed using mic ocon olle s buil a ound p ocesso s ha can well
14
Ka hika K.C, "Wi eless mesh ne wo k: A su ey," 2016 In e na ional Con e ence on Wi eless Communica ions, Signal P ocessing and Ne wo king (WiSPNET),
Chennai, India, 2016, pp. 1966-1970, doi: 10.1109/WiSPNET.2016.7566486.
15
Ve mesan, O., Pé o , F., Coppola, M., Schneide , M. and Höß, A. (Au ho s). Indus ial AI Technologies o Nex -Gene a ion Au onomous Ope a ions wi h
Sus ainable Pe o mance, in "In elligen Edge-Embedded Technologies o Digi ising Indus y" (Chap e 1), Ri e Publishe s Se ies in Communica ions, June
2022. ISBN: 9788770226110, e-ISBN: 9788770226103. Online a : h ps://www. i e publishe s.com/pd /ebook/chap e /RP_9788770226103C1.pd
© AIOTI. All igh s ese ed.
18
cope wi h cos s and powe consump ion cons ain s. The dis ance om he da a sou ce
gene a ed by he senso s is minimised. The compu e esou ces p ocess his aw da a in line and
p oduce insigh da a wi h minimal la ency. The HW de ices o he mic o-edge physical
senso s/ac ua o s gene a e om aw da a insigh da a and/o ac ua e based on physical
objec s by in eg a ing AI-based elemen s in o hese de ices and unning AI-based echniques
o in e ence and sel - aining. In elligen mic o-edge allows IoT eal- ime applica ions o
become ubiqui ous and me ged in o he en i onmen whe e a ious IoT de ices can sense hei
en i onmen s and eac as and in elligen ly wi h a e y low powe budge ene gy-e icien
gain. AI capabili ies in eg a ed in o IoT de ices signi ican ly enhance hei capabili ies
( unc ionali y, pe o mances, low la ency, low powe consump ion, high p ocessing powe ) and
use ulness, especially when he ull powe o hese ne wo ked de ices is ha nessed – a end
called AI on edge
16
,
17
,
18
.
The deep-edge comp ises in elligen con olle s like P og ammable Logic Con olle s (PLC),
Supe iso y con ol and da a acquisi ion (SCADA) elemen s, machine ision connec ed
embedded sys ems, ne wo king equipmen , ga eways and compu ing uni s ha agg ega e
da a om he senso s/ac ua o s o he IoT de ices ha gene a e da a. Deep edge p ocessing
esou ces a e implemen ed wi h pe o man p ocesso s and mic ocon olle s such as In el i-
se ies, A om, ARM M7+, e c., including CPUs, GPUs, Tenso P ocessing Uni s (TPU), and ASICs. The
sys em a chi ec u e, including he deep edge, depends on he en isioned unc ionali y and
deploymen op ions conside ing ha hese de ices' co es a e con olle s: PLCs, ga eways wi h
cogni i e capabili ies ha can acqui e, agg ega e, unde s and, eac o da a, exchange, and
dis ibu e in o ma ion
19
.
The me a-edge in eg a es p ocessing uni s, ypically loca ed on-p emises, implemen ed wi h
high-pe o mance embedded compu ing uni s, edge machine ision sys ems, edge se e s
(e.g., high-pe o mance CPUs, GPUs, FPGAs, e c.) ha a e designed o handle compu e-
in ensi e asks, such as p ocessing, da a analy ics, AI-based unc ions, ne wo king, and da a
s o age
20
.
This classi ica ion is closely ela ed o he dis ance be ween he da a sou ce and p ocessing,
impac ing o e all la ency. A high-le el ep esen a ion o he edge cloud con inuum is
ep esen ed in Figu e 3 and a ough es ima ion o he communica ion la ency and he dis ance
om he da a sou ces a e p esen ed below.
Figu e 3 Edge cloud con inuum ep esen a ion
• Mic o-edge la ency below 1 ms, ange om mm o 15 m.
• Deep-edge la ency below 2-5 ms, ange up o 1 km.
16
Ve mesan, O., Pé o , F., Coppola, M., Schneide , M. and Höß, A. (Au ho s). Indus ial AI Technologies o Nex -Gene a ion Au onomous Ope a ions wi h
Sus ainable Pe o mance, in "In elligen Edge-Embedded Technologies o Digi ising Indus y" (Chap e 1), Ri e Publishe s Se ies in Communica ions, June
2022. ISBN: 9788770226110, e-ISBN: 9788770226103. Online a : h ps://www. i e publishe s.com/pd /ebook/chap e /RP_9788770226103C1.pd
17
Dhi aj Joshi; Ni mi Desai; Shyama P osad Chowdhu y; Wei‐Han Lee; Luis Ba hen; Shiqiang Wang; Dinesh Ve ma, "AI a he Edge: Challenges, Applica ions,
and Di ec ions," in IoT o De ense and Na ional Secu i y, IEEE, 2023, pp.133-160, doi: 10.1002/9781119892199.ch9.
18
A. Muni , E. Blasch, J. Kwon, J. Kong and A. A ed, "A i icial In elligence and Da a Fusion a he Edge," in IEEE Ae ospace and Elec onic Sys ems Magazine,
ol. 36, no. 7, pp. 62-78, 1 July 2021, doi: 10.1109/MAES.2020.3043072.
19
Ve mesan, O., Pé o , F., Coppola, M., Schneide , M. and Höß, A. (Au ho s). Indus ial AI Technologies o Nex -Gene a ion Au onomous Ope a ions wi h
Sus ainable Pe o mance, in "In elligen Edge-Embedded Technologies o Digi ising Indus y" (Chap e 1), Ri e Publishe s Se ies in Communica ions, June
2022. ISBN: 9788770226110, e-ISBN: 9788770226103. Online a : h ps://www. i e publishe s.com/pd /ebook/chap e /RP_9788770226103C1.pd
20
Ve mesan, O., Pé o , F., Coppola, M., Schneide , M. and Höß, A. (Au ho s). Indus ial AI Technologies o Nex -Gene a ion Au onomous Ope a ions wi h
Sus ainable Pe o mance, in "In elligen Edge-Embedded Technologies o Digi ising Indus y" (Chap e 1), Ri e Publishe s Se ies in Communica ions, June
2022. ISBN: 9788770226110, e-ISBN: 9788770226103. Online a : h ps://www. i e publishe s.com/pd /ebook/chap e /RP_9788770226103C1.pd
© AIOTI. All igh s ese ed.
19
• Me a-edge la ency below 10 ms, ange up o 50 km.
• MEC la ency 10-5 ms, ange up o 75 km.
• Fog la ency 10-20 ms, ange up o 100 km.
• Fa -edge la ency 20-50 ms, ange up o 200 km.Cloud and da a cen es la ency 50-100
ms, ange up o 1000 km.
IoT applica ion deploymen s a he edge can p o ide mo e ene gy-e icien p ocessing
solu ions by in eg a ing a ious compu ing a chi ec u es a he edge (e.g., neu omo phic
chips, CPUs, GPUs, ASICs, FPGAs), educing da a a ic and da a s o age.
Edge compu ing educes he la ency and bandwid h cons ain s o he communica ion
ne wo k and In e ne connec i i y by p ocessing locally and dis ibu ing compu ing esou ces,
in elligence, and so wa e s acks among he compu ing ne wo k IoT de ices.
3.2 Main T ends, Issues and Challenges
Edge compu ing mo es se ice p o isioning close o p oduce s and use s o such se ices. I
p o ides low-la ency, mobili y suppo , da a analy ics close o he da a sou ce, and educed
ene gy consump ion.
In IoT, he e a e many expec a ions o AI-based applica ions. AI p ocessing happens mainly in
he cloud.
Wi h he imp o emen o AI enabling echnologies, AI p ocessing is mo ing in o IoT de ices.
In elligen edge de ices will in eg a e so wa e o ain he AI model o di e en applica ions
and execu able so wa e ha uns he AI algo i hms on he IoT de ices.
Machine Lea ning (ML) and Deep Lea ning (DL) will ha e unique oles o in elligen IoT de ices
a he edge.
As a subse o AI, ML enables machines o ecognise pa e ns and make p edic ions by analysing
da a ins ead o using explici p og amming. Fo a highe le el o accu acy, ML can be
upg aded o DL.
DL is a class o ML algo i hms ha p og essi ely uses a hie a chy o mul iple laye s o ex ac
highe -le el ea u es om he aw inpu .
Wi h DL, a compu e can ain i sel wi h an ex ensi e da a se collec ed o his pu pose. The
esul is an A i icial Neu al Ne wo k (ANN) ha con ains all he in o ma ion o ca y ou he ask.
The ANN uses he knowledge acqui ed in aining o in e da a ea u es om new incoming
da a.
© AIOTI. All igh s ese ed.
20
Achie ing a middlewa e a chi ec u e ha in eg a es all he le els o he edge g anula i y and
manage o e ec i ely handle he e ogeneous IoT esou ces, including ne wo king and
compu ing, is a new challenge o IoT/edge compu ing, leading o a uni ied ne wo king and
compu ing a chi ec u e.
Deploymen o DL a he IoT edge has se e al challenges such as iden i ica ion o leading
pe o mance indica o s o edge DL algo i hms, exploi a ion o ade-o s be ween he indica o s
o imp o e he e iciency o he algo i hms, e ec i e combina ion o he simpli ied DL models
a he IoT edge wi h he DL model in he cloud, coo dina ion be ween aining and in e ence
and ene gy-e icien deploymen o DL a he IoT edge.
3.3 Resea ch P io i ies Timeline
Table 1 IoT and edge compu ing esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
AI
De elop e ec i e AI-based
solu ions ha can exploi he
bene i o DL and Fede a ed
Lea ning In di e en ways o
he di e en needs o he Edge
G anula i y
How o deli e an IoT/edge-
cloud con inuum ha akes
ca e o all he speci ici ies o all
he In ol ed laye s in a smoo h
way.
Dis inc ion o he se e al Edge
G anula i y o ally anspa en
o bo h sys ems and humans
( inal use s).
Accele a o s
Adap , by using dedica ed
accele a o s like ASICs o
econ igu able ones like FPGAs
(depending on he cons ain s
and equi emen s in ocus), he
di e en needs o he di e en
edge g anula i y.
Exploi he smalle echnology
node so o embed in edge
de ices ad anced
accele a o s unc ionali ies.
No dis inc ion possible anymo e
be ween FPGA and dedica ed
HW.
New HW
a chi ec u es
P omising new p e-comme cial
a chi ec u es (like
neu omo phic compu ing ones)
will be assessed agains exis ing
adi ional a chi ec u es based
on CPU/GPUs o ec o
p ocesso s.
Edge de ices con aining no el
HW a chi ec u es ha can un AI
algo i hms so o achie e be e
esul s han he cu en ly exis ing
HW a chi ec u e based on CPU
and GPU.
Syne gy be ween legacy and
no el a chi ec u es and usage
o one o he o he acco ding o
he eal- ime needs o he use
case in ocus, aking in o
conside a ion also eal- ime sel -
adap a ion o he di e en
wo kloads.
Secu i y
Design a secu e-by-
cons uc ion dis ibu ed
a chi ec u e ha can impac all
he di e en g anula i ies o he
Edge.
Deploy such a chi ec u e ac oss
di e en e icals and indus ial
domains.
Take on boa d quan um and
pos -quan um no el app oach
o ensu e he highes possible
secu i y le el a all laye s and
kind o de ices in a anspa en
way o inal use s.
© AIOTI. All igh s ese ed.
21
4. IoT Edge and X-Con inuum Pa adigm
Dis ibu ed IoT digi al pla o ms and in as uc u es o collec ing, p ocessing, compu ing,
communica ing, and unning analy ics a e e ol ing owa ds an in e connec ed ecosys em
allowing complex applica ions o be execu ed om IoT edge o high-pe o mance compu ing
capabili ies.
The IoT digi al con inuum includes compu ing esou ces placed a op imal p ocessing poin s in
he IoT sys em om he cloud da a cen e o edge IoT sys ems and endpoin de ices ha
in eg a e E2E capabili ies such as sensing, p ocessing, connec i i y, compu ing, s o age,
in elligence, secu i y, sa e y, and p i acy as illus a ed in Figu e 4.
Figu e 4 X- con inuum pa adigm
Implemen ing IoT E2E capabili ies in such a dis ibu ed con inuum is challenging and equi es
econciling di e en applica ion equi emen s and cons ain s wi h IoT in as uc u e design
choices including he ene gy consump ion con inuum.
One main challenge in his domain is how o accu a ely ep oduce ele an beha iou s o IoT
applica ions wo k low and ep esen a i e se ings o he IoT physical in as uc u e, including AI-
based lea ning/ aining and in e encing unde lying he complex dis ibu ed con inuum om
mic o-, deep-, me a-edge o cloud.
4.1 Technological de elopmen s
IoT da a p ocessing, compu ing and communica ion no longe lean only on adi ional
app oaches ha send all da a o cen alised cloud acili ies o p ocessing, a he le e age on
he dis ibu ed esou ces close o whe e he da a is gene a ed, so o ex ac insigh s in eal- ime
while keeping e icien esou ce usage and p ese ing secu i y and p i acy cons ain s.
The IoT digi al con inuum seamlessly combines esou ces, p ocessing, and s o age capabili ies,
and se ices a he cen alised cloud, IoT edge, and in- ansi , along he dis ibu ed da a pa h.

© AIOTI. All igh s ese ed.
22
The de elopmen o di e en dedica ed sys ems o da a p ocessing on each componen o
he con inuum lacks a holis ic app oach and a us wo hy a chi ec u e o implemen ing u u e
IoT ubiqui ous compu ing sys ems. One o he main challenges in his con ex is ela ed o he
complexi y o deploying la ge-scale, eal-li e IoT applica ions on such he e ogeneous
in as uc u es, which b eaks down o con igu ing many sys em-speci ic pa ame e s and
ha monising many equi emen s o cons ain s ega ding p ocessing, communica ion la ency,
s o age, ene gy, ne wo k e iciency, in e ope abili y, mobili y, secu i y, and da a p i acy.
The IoT digi al con inuum equi es an IoT digi al in as uc u e join ly used by complex applica ion
wo k lows, commonly combining eal- ime da a gene a ion, p ocessing, compu a ion, and
analy ics.
The combined edge-cloud in as uc u es deli e IoT applica ions' i ual dis ibu ed-cen alised
compu ing and s o age esou ces o pe o m s eam and ba ch analy ics on comp ehensi e
his o ical da a, AI model aining, and complex simula ions. IoT edge in as uc u es complemen
hese i ual esou ces wi h dis ibu ed compu ing and s o age capabili ies close o IoT in elligen
de ices ha sense he en i onmen .
The IoT elemen s include nodes, p og amable logic con olle s, ga eways, ou e s, mic o-se e s,
and embedded high-pe o mance uni s dis ibu ed ac oss he edge con inuum and used o
inal o in- ansi p ocessing on da a agg ega ed om mul iple IoT edge de ices o educe
u he da a olumes ha need o be ans e ed and p ocessed in he clouds and da a cen es.
Ligh weigh amewo ks, based on message b oke s using IoT ligh weigh p o ocols, enable
dis ibu ed and hie a chical p ocessing and in elligen agg ega ion, minimising la ency, and
bandwid h usage.
The IoT digi al con inuum equi es an e icien a chi ec u e ha ocuses on bo h ho izon al and
e ical esou ce dis ibu ion in he IoT edge- o-cloud con inuum o secu e seamless E2E se ices
ac oss he whole con inuum. As a esul , IoT sys em a chi ec u es and managemen
mechanisms a e needed o seamlessly encompass compu ing, s o age, and ne wo king.
4.2 Main T ends, Issues and Challenges
The implemen a ion o IoT con inuum-X equi es esea ch ha in eg a es ideally open-sou ce
in elligence ools, HW and SW pla o ms, and sys ems, hus add essing he non- unc ional
aspec s o IoT sys ems wi h mul iple elemen s as pa o he con inuum-X.
Ano he issue is how o gua an ee embedded in e ope abili y in o con inuum-X wi h anspa en
agg ega ion, dis ibu ion and logging o pa icipan s and sys em ac i i ies when using IoT
dis ibu ed edge compu ing se ices.
The esea ch e o mus add ess how o collec simul aneously a huge amoun o da a om a
e y dispe se, he e ogeneous and nume ous numbe s o sou ces when eal- ime decisions a e
o be aken.
Fu u e esea ch ac i i ies include add essing he esou ce managemen o con inuum-X
capabili ies by managing he esou ces on he p ocessing poin s and iden i ying he op imal E2E
capabili ies and hei ele ance, depending on he IoT applica ion con ex .
Wo k is equi ed on p o iding solu ions o dynamic placemen o edge compu ing pla o ms
o minimise he ne wo k access delay depending on he con ex , IoT ac i i y a he edge and
IoT applica ion scalabili y equi emen s.
© AIOTI. All igh s ese ed.
23
Fu he esea ch is needed o p o ide collabo a i e and hie a chical compu ing o dis ibu e
he IoT de ices' wo kload among a ious dis ibu ed p ocessing/compu ing poin s and alloca e
op imal esou ces o each de ice based on he con inuum-X capabili ies and equi emen s.
The esea ch in he ollowing yea s should add ess he in as uc u e edge and he de ice edge
oge he and p o ide solu ions o op imising he in eg a ion o he wo.
Resea ch in con ex -awa e, scalable he e ogeneous con inuum-X op imisa ion o IoT
applica ions is equi ed due o he emendous di e si y o wo kloads and applica ions being
un ac oss a a ie y o di e en IoT edge de ices.
Secu i y, sa e y, and moni o ing o c i ical E2E capabili ies o IoT sys ems a e essen ial o many
IoT applica ions. The in eg a ion o solu ions o add essing hese elemen s o he con inuum-X
holis ically in o IoT pla o ms should be p io i ised.
To p o ide con inuum-X implemen a ion lexibili y on di e en IoT pla o ms, new esea ch
di ec ions should add ess de eloping and deploying con aine s a he edge ha add ess
challenges ela ed o connec i i y, dis ibu ion, and synch onisa ion, and le e age he di e en
IoT a chi ec u e o deploy new solu ions and applica ions. De eloping edge AI o IoT
applica ions equi es o loading AI models lea ning/ aining o IoT de ices o e ain he AI
models locally wi h nea eal- ime da a collec ed by he IoT de ices o e icien da a
p ocessing.
Fu u e esea ch is equi ed o iden i y how he compu a ional asks, AI models, inpu s,
pa ame e s, and weigh s, a e ins an ia ed on IoT de ices conside ing he speci ic equi emen s
o he con inuum-X elemen s and how hei alloca ion changes du ing execu ion.
Add essing IoT dis ibu ed AI and ede a ed lea ning/ aining conside ing he con inuum-X
elemen s equi es de ining a new gene a ion o ools and mechanisms ha enable ine-
g anula i y compu a ion, p ocessing, communica ion, coo dina ion, and mobili y managemen
ac oss he edge.
Finally, new esea ch app oaches a e equi ed o add ess he agmen a ion in IoT o ackle he
E2E challenges, by p o iding se e al in e ope able, in eg a ed HW, so wa e, and se ices
app oaches o add ess he con inuum-X.
4.3 Resea ch P io i ies Timeline
Table 2 IoT edge and X-Con inuum esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
Con inuum
e inemen
High-le el IoT-edge-Cloud
con inuum solu ions o basic
se ices, p o ocols, and
esou ce alloca ion.
Mo e e ined con inuum suppo
o unc ionali ies and lowe
laye s managemen .
Fully suppo ed and smoo h
managemen o all he edge
g anula i ies, ypes, se ices
esou ces, and unc ions, in a
ully he e ogeneous ecosys em.
Common sha ed
da a space
De ini ion o domain-speci ic
edge IoT applica ions aligned
o common Eu opean da a
spaces.
De ini ion and implemen a ion
o common Eu opean da a
spaces.
Fully compa ible in e wo k o
da a spaces wi h all
geog aphical a eas.
Ins an ia ion
Piecewise and e ical-speci ic
ins an ia ion o all kinds o
esou ces.
Cohe en and homogeneous
ins an ia ion o all kinds o
esou ces.
No dis inc ion be ween he
kinds o esou ces, all a e
ea ed as I hey we e
homogeneous and a ailable
e e ywhe e.
© AIOTI. All igh s ese ed.
24
5. In elligen Connec i i y
Ne wo ks and connec i i y a e becoming mo e he e ogeneous and IoT de ices use mo e and
mo e a a ie y o equipmen wi h di e en wi eless access echnologies. Wi h he cons an
inc ease o he numbe s o use s (new con ac s) and especially o hings (senso s, me e ing
sys ems, mobile/s a ic obo s, ehicles, d ones, e c.) joining each day he communica ion
ne wo k, he need o a sma e and mo e e ec i e way o handling he ela ed g owing da a
c ea ed by hose de ices is becoming mo e and mo e an issue ha needs he a en ion o he
IoT communi y.
To handle such complexi y and o manage in a sma way how he communica ion sys em can
p ope ly manage such an inc easing amoun o da a (scalabili y p oblem), he e’s he need o
de ine and elabo a e on he in elligen connec i i y concep , o op imise he usage o he
ne wo ks bu also o dec ease he o e all ene gy consump ion o he o e all IoT sys em.
In elligen connec i i y can be seen as he smoo h syne gy o di e en echnologies and
domains, so o o e inal use s o a communica ion sys em he bes possible QoS, acco ding o
he gi en esou ce, a ailable echnology, and ene gy cons ain s.
5.1 Technological de elopmen s
Wi h he echnology minia u iza ion ha s ill p oceeds a almos yea ly cadence, and ecen
ad ancemen s in ba e y echnology, and he mo e e ec i e use o ou -o -de ice p ocessing
capabili y se ices and applica ions, i is possible o inc ease he numbe o unc ions and he
communica ion capabili ies o e minals, so o p ope ly handle mos o he issues desc ibed
abo e and make in elligen connec i i y inally a eali y.
The nex wa e o IoT de ices a he edge will he e o e use a kind o in elligen connec i i y ha
elies on he consolida ion and syne gy o di e en communica ion echnologies. Whe eas he
In e ne is oday he la ges wo ldwide communica ion ne wo k, such in elligen connec i i y o
IoT s ill equi es u he de elopmen o allow high demand in bandwid h and quali y in signals
and p o ocols, in suppo o mo e c i ical con en and da a exchange.
The equi emen s o eal- ime esponse may be, amongs o he s, sa e y and mission-c i ical (like
in elemedicine, o example) and, consequen ly, IoT communica ions solu ions a e nume ous
and di e se:
• A i icial In elligence, especially when declined in dis ibu ed AI a he edge o he ne wo k,
bu also as a se o me hods o mo e in elligen ly manage he huge and di e se amoun o
da a c ea ed and consumed by e minals.
• Ad anced communica ion p o ocols: o accommoda e he di e en needs o he di e en
e icals ha make use o he communica ion ne wo k, dedica ed and always imp o ing
communica ion p o ocols a e o be de ined and deployed. S anda ds bodies cons an ly
upda e and add new ea u es o each new gene a ion o he ne wo k hey de ine.
• Edge P ocessing: he need o low la ency o se e al applica ions and he complexi y o he
da a o be managed eques ha always mo e da a shall be handled and p ocessed as
close as possible o he sou ce and des ina ion o ha da a, i.e., a he edge. The Mul i-
Access Edge Compu ing (MEC) concep , in cons an e olu ion, and i s ela ed ea u es can
be seen as a co ne s one in he ques o a mo e e icien da a managemen a he Edge.
© AIOTI. All igh s ese ed.
25
• Ene gy Consump ion: his is a sys em pa ame e ha becomes mo e and mo e impo an a
each new gene a ion o he communica ion ne wo k. Reason o ha is no only he need
o CO2 educ ion o ame he dis up i e clima e change consequences, bu also shee
economic easons, as inc easing he numbe o se e s and chips needed o handle he
aising amoun o da a exchanged in he wo ld, implies a cons an and s eep aise o he
ene gy demand o accommoda e he needed QoS and eliabili y o such communica ions.
The e o e, i is key ha in he o hcoming In elligen Connec i i y pa adigm low ene gy
consump ion is aken as a key design cons ain o all new chips, pla o ms, and SW sui es.
• Spec um managemen : wi h he in oduc ion o se e al access echnologies (LoRa, Wi-Fi,
cellula , e c.) and he comme cial a ailabili y o new spec um bands o new applica ion
(mmWa e bands unde and abo e 100 GHz in addi ion o he s anda ds sub-6 GHz bands),
he e is he need o make e ec i e use o all he di e en bands now made a ailable. Tha
e ec i e use has he ad an age no only o inc ease he co e age, he esiliency, and he
pe o mance, bu also, o enable o e ing he lowes ene gy consuming access echnology
in a speci ic e ical use case, also educing in a dynamic way he ene gy consumed o
deli e a speci ic se ice. Dynamic usage o di e en kind o spec um, o non-adjacen
bands, and e en mixing di e en bands coming om di e en access echnologies, based
on local a ailabili y, will change he way se ices can be deli e ed om he cos and om
he quali y poin s o iew. Finally, i is wo h men ioning ha also d ones and sa elli e accesses
and ela ed bands will soon be b oadly a ailable o comme cial use, and he a ailabili y o
co e age e e ywhe e will change he way we hink o connec i i y and ela ed se ices.
• Mul i-Access capable de ices: o deli e highe QoS, o handle he communica ion
bo lenecks (bo h in he on - and he back-haul) in highly densely popula ed a eas, and o
ul il Key Pe o mance Indica o s (KPI) o he applica ions eques ing a huge amoun o
bandwid h, e.g., o Augmen ed Reali y / Vi ual Reali y (AR/VR) se ices, he need o a
handse capable o making use o he mos sui able a ailable access echnologies and
bands is becoming mo e and mo e impo an .
• Seamless connec i i y b oke : ha is needed o op imise he connec i i y o a speci ic
ne wo k, acco ding o se e al echnical and business pa ame e s. The seamless oaming
should a oid ha da a ge los du ing he swi ching p ocedu e; AI/ML echnologies should
also con ibu e o make he mos ele an choice.
An essen ial cha ac e is ic o he nex gene a ion IoT is he equi emen s ange om high
eliabili y and esilience in he communica ion ne wo k o ul a-low la ency and inc eased
capaci y a he communica ion channel. IoT In elligen Connec i i y solu ions a e also e y
dependen on he con ex in which hey a e applied and whe he i is necessa y o espond o
s ic ene gy e iciency cons ain s o co e la ge ou doo a eas, deep indoo en i onmen s o
ehicles mo ing a high speeds.
5.2 Main T ends, Issues and Challenges
The nex -gene a ion IoT mus add ess he con e gence be ween di e en cells and adia ion
and de elop new managemen models o con ol oaming while exploi ing he coexis ence o
many di e en cells and adio access echnology (RAT). New managemen p o ocols o handle
use assignmen s ega ding cells and echnology will ha e o be deployed in he mobile co e
ne wo k o access ne wo k esou ces mo e e icien ly.
© AIOTI. All igh s ese ed.
32
7. He e ogeneous Cogni i e Edge IoT Mesh
IoT edge can be o med by a mesh ne wo k o in elligen IoT de ices using edge IoT pla o ms,
including AI model aining and ML in e ence ha p ocess, analyse, s o e in o ma ion locally
close o he da a sou ces, and communica e and exchange in o ma ion wi h o he edge
de ices, compu ing uni s and ac oss he compu ing con inuum, made o cloud pla o ms and
da a cen es.
Cogni i e compu e con inuum applied in decen alised en i onmen s, including edge mesh
in as uc u es, can ope a e e icien ly and eliably wi hou a cen al en i y o ull knowledge
and managemen abou a ailable esou ces and cu en wo kloads.
Au onomous compu ing echniques and ML should be b oadly applied, so o be able o adap
o un o eseen si ua ions and an icipa e u u e ci cums ances, such ha he cogni i e edge IoT
de ices become sel -awa e, sel -managed, sel -p o ec ed, sel -healing, and sel -op imising.
7.1 Technological de elopmen s
IoT and edge compu ing sys ems e ol e a ound he undamen al con ex managemen
p ocesses: acquisi ion, modelling, easoning, and dis ibu ion, while dis ibu ing he p ocesses
based on he con ex in o ma ion.
No el IoT and edge compu ing designs b ing inno a i e adap i e and beha iou al-awa eness
capabili ies o he IoT, so o se he ounda ions o de eloping he nex -gene a ion cogni i e
and sel -adap i e IoT sys ems.
The huge numbe o he e ogeneous edge IoT de ices need o be in eg a ed in o mesh
ne wo ks, in which he in as uc u e nodes connec di ec ly, dynamically, and non-
hie a chically o o he nodes and coope a e wi h one ano he o e icien ly ou e da a o and
om edge IoT de ices.
The mesh ne wo k opologies allow mul iple ou es o exchanging in o ma ion among
connec ed nodes. The mesh app oach inc eases he ne wo k's esilience in case o a node o
connec ion ailu e.
Mo e ex ensi e edge IoT mesh ne wo ks may include mul iple ou e s, swi ches and o he IoT
de ices ope a ing as nodes. In he e ogenous cogni i e IoT ne wo ks, con ex -awa eness is
becoming c i ical o IoT sys ems as he p ocessing is mo ing owa ds pe asi e, swa m
e olu iona y compu ing.
Resou ces and se ices a e emo ely accessed by edge IoT de ices om anywhe e in he
ne wo k a any ime. The en i onmen in which da a is exchanged and p ocessed can change
dynamically as di e en ad-hoc he e ogenous ne wo ks a e c ea ed.
C ea ing sel -con igu ed mesh IoT ne wo ks, whe e hings in eg a e sel -pe cep ion and
au oma ed esponse, can suppo his.
Embedding app op ia e plug-and-play in elligen and au onomous edge IoT de ices acili a e
he ansi ion in o IoT ne wo ks wi h sel -X capabili ies (sel -con igu ing, sel -healing, sel -
op imising, sel -p o ec ing, e c.).
Such sel -con igu ed mesh IoT ne wo ks o m a lexible a chi ec u e whe e hings a e ac i e and
locally in eg a e almos e e y unc ional and ope a ional equi emen .

© AIOTI. All igh s ese ed.
33
The IoT end-de ices can manage c i ical aspec s such as secu i y, sa e y, and us wo hiness
and embed ede a ed, decen alised ML and decision-making mechanisms, enabling
dis up i e c oss-domain applica ions wi h high complexi y and scale.
The secu i y echniques used by edge IoT de ices can be c ucial o ob aining con ex ual
in o ma ion om o he end de ices using a ligh weigh and E2E app oach based on cu en
s anda ds (e.g., IETF OSCORE
24
). The p ocess is linked o an ini ial boo s apping by which a
legi ima e de ice can be success ully and secu ely deployed in he ne wo k.
The modelling o con ex in o ma ion should be based on cu en s anda ds ha gua an ee he
ep esen a ion o ma 's in e ope abili y using lexible app oaches (e.g., JSON-LD
25
, o addi ional
ep esen a ions based on CBOR
26
). In his case, he p ima y pu pose is o achie e a ade-o
be ween ligh ness and exp essi eness o ep esen he dependencies among de ices'
con ex ual in o ma ion.
7.2 Main T ends, Issues and Challenges
The e olu ion o IoT and edge compu ing sys em will be based on he in eg a ion o AI-based
mechanisms. Mo ing pa o he sys em in elligence down o he end de ices b ings a ange o
ad an ages and allows a widesp ead collec i e-in elligence web.
The use o in elligence on he de ice (e.g., TinyML
27
and o he simila embedded ML
applica ions, algo i hms, HW, and SW) boos s he e olu ion o end de ices o in elligen IoT
objec s ha can o e , among o he ad an ages, a much mo e au onomous beha iou .
In eg a ing AI wi hin edge IoT de ices is one s ep owa ds de eloping cogni i e and e olu i e
IoT sys ems. The in eg a ion o cogni i e sel -e olu ion capabili ies in he en i e E2E ne wo k chain
(edge, dew, og, and cloud) goes beyond he cu en limi s o AI by pe mi ing end de ices
and, by ex ension, he whole sys em o g ow and e ine i s in elligence.
The de elopmen o e olu i e a i icial cogni i e mechanisms and hei in eg a ion wi h he sel -
x concep s pa e he way o designing IoT de ices and sys ems wi h inc eased in elligence ha
will pe mi hem o me a-lea n om new si ua ions, scena ios, and en i onmen al changes.
The coope a ion among indi idual me a-sma IoT objec s can c ea e a highe in elligence laye
o ackling la ge-scale issues. The decision p ocess is no isola ed bu s a s om a comple e
iew o he scena io and ga he s collabo a i e e o s.
The he e ogeneous cogni i e and mobile IoT edge mesh solu ions u ilise low-le el IoT de ices o
he decision-making p ocess, he ga eway de ices and he mic o se e s dis ibu ed ac oss he
edge con inuum. The da a collec ion, p ocessing, and decision-making asks a e dis ibu ed
among hese edge de ices wi hin he e ogeneous ne wo ks.
The compu a ion asks and da a a e sha ed using a cogni i e mesh ne wo k o edge IoT
de ices, ga eways, and mic o se e s. He e ogeneous cogni i e and mobile IoT edge mesh
sys ems o e dis ibu ed p ocessing, low la ency, aul ole ance, be e scalabili y, secu i y, and
p i acy. These ea u es a e essen ial o c i ical applica ions ha need highe eliabili y, eal- ime
p ocessing, mobili y suppo , and con ex awa eness.
24
h ps://da a acke .ie .o g/doc/ c8613/
25
h ps://json-ld.o g/
26
h ps://cbo .io/
27
h ps://www. inyml.o g/
© AIOTI. All igh s ese ed.
34
Dis ibu ed compu a ions on edge IoT sys ems imply add essing he e ogeneous cogni i e
ea u es o IoT applica ions and de eloping a aul - ole an obus collec i e compu ing
amewo k sui ed o mul i-IoT de ice sys ems ha include decen alised ope a ions applied o
dynamic en i onmen s.
Mobile IoT edge mesh cha ac e is ics, such as scalabili y, he e ogenei y, mobili y, and
ubiqui ous ne wo king, equi e a obus coope a i e and cogni i e compu ing amewo k ha
add esses he edge IoT sys ems cha ac e is ics, including new secu i y solu ions adequa e o
hese new IoT-based sys ems.
The esea ch needs o p o ide IoT pla o m-agnos ic solu ions wi h scalable edge IoT de ices in
a sel -healing and sel -o ganising ne wo k. In a he e ogeneous mesh ne wo k en i onmen ,
communica ion and cogni i e coexis ence a e c i ical o allow maximum obus ness o RF
dis u bances and minimize nega i e e ec s due o he co-exis ence o a ious wi eless and
cellula ne wo ks.
The coope a ion be ween edge IoT de ices and he dynamic communica ion in as uc u es o
he edge and swa m sys ems c ea e he he e ogeneous cogni i e and dis ibu ed in elligence
amewo k needed o op imally suppo edge IoT mesh applica ions.
The he e ogeneous cogni i e and mobile IoT edge mesh esea ch mus add ess opics a he
in e sec ion be ween IoT, AI, connec i i y, and edge compu ing. These opics ela e o con ex
awa eness, au onomous con ol, ambien in elligence, seman ic easoning, and cogni i e IoT
o enable dis ibu ed in elligence in IoT.
The edge IoT dis ibu ed in elligence mus be add essed holis ically by de eloping cogni i e
pla o ms in eg a ing ea u es such as dis ibu ed da a collec ion, da a analy ics, ne wo king,
da a managemen , edge IoT de ice managemen , esou ce managemen , se ice
managemen , o ches a ion, and ede a ed lea ning.
Se e al esea ch ques ions a e ela ed o de ining he dis ibu ed mesh ne wo k and he
cogni i e compu ing model, he dis ibu ion o da a p ocessing, and he global op imisa ion
(ene gy, p ocessing, ime, e c.) o communica ion and compu a ion. New compu a ion
algo i hms mus be de eloped o dis ibu ed compu ing o be pe o med by he a ious
he e ogeneous, esou ce-cons ain edge IoT de ices ope a ing wi h dynamic communica ion
and in e mi en connec i i y.
© AIOTI. All igh s ese ed.
35
7.3 Resea ch P io i ies Timeline
Table 5 He e ogeneous cogni i e edge IoT mesh esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
A chi ec u e
A chi ec u al models and me a-
embedded ope a ing sys ems
wi h in eg a ed s ack o wi eless
mesh ne wo king.
AI-based cogni i e mesh
a chi ec u es ha in eg a e
componen s and modules
add essing con ex awa eness,
au onomous con ol, ambien
in elligence, seman ic
easoning, and ede a ed
lea ning.
Dynamic cogni i e mesh
a chi ec u es wi h AI and
con ex -based con igu a ion
capabili ies in eg a ing ea u es
o swa m in elligence and
dis ibu ed p ocessing
capabili ies.
Cogni i e
Capabili ies
E olu i e a i icial cogni i e
mechanisms o edge IoT
sys ems and he in eg a ion wi h
mesh ne wo ks opologies.
De elopmen o new algo i hms
and SW/HW sel -X capabili ies.
In eg a ion o cogni i e sel -
e olu ion capabili ies in he
en i e mesh ne wo k and ac oss
he edge g anula i y including
scalable AI capabili ies ac oss
he con inuum.
Compu ing
Models
Compu a ion algo i hms o
dis ibu ed compu ing applied
o di e en he e ogeneous,
esou ce-cons ain edge IoT
de ices ac oss he edge
con inuum.
He e ogeneous cogni i e edge
IoT mesh amewo ks in eg a ing
ea u es such as dis ibu ed
da a collec ion, da a analy ics,
ne wo king, da a
managemen , edge IoT de ice
managemen , esou ce
managemen , se ice
managemen , o ches a ion,
and ede a ed lea ning.
Me a-mesh compu ing models
o global op imisa ion (ene gy,
p ocessing, ime, e c.) o
dis ibu ed sys ems.
© AIOTI. All igh s ese ed.
36
8. IoT Digi al Twins, Modelling and Simula ion En i onmen s
An IoT DT is a i ual ep esen a ion o an IoT de ice ha models he de ice's cha ac e is ics,
p ope ies, en i onmen al condi ions, beha iou s, and unc ions o e he ope a ional li e ime,
based on eal- ime da a and in o ma ion synch onised au oma ically and bi-di ec ionally a a
speci ied equency and accu acy. An IoT DT uses simula ion, ML, and easoning o simula e
a ious scena ios in di e en IoT applica ions and help op imise and imp o e he o e all IoT
sys em unc ionali ies and se ices.
The eal- ime ea u e ep esen s a i al cha ac e is ic o de ine IoT DTs, conside ing ha he eal-
ime ins ances a y acco ding o IoT applica ions. In many IoT applica ions, ime alues a e no
de ined iden ically, and such issue should be ca e ully conside ed when designing IoT DT
ins ances. The synch onisa ion be ween he physical IoT de ice and i s i ual ep esen a ion in
he simula ion en i onmen and he synch onisa ion o he e en s and scena ios in he simula ion
pla o m is c i ical o he pe o mance o he whole IoT sys em.
8.1 Technological de elopmen s
IoT DTs suppo op imal decision-making and e ec i e and e icien ac ions o he IoT de ices in
IoT applica ions, using eal- ime and his o ical da a o ep esen he pas and p esen and
simula e p edic ed ac i i ies ailo ed o use cases based on domain knowledge, and
implemen ed in In o ma ion Technology / Ope a ion Technology (IT/OT) sys ems. The IoT DTs can
expose a se o se ices o execu e ce ain ope a ions and p oduce da a desc ibing he
physical ac i i y o he IoT de ices ha hey i ualise.
How o choose and op imize cha ac e is ics, p ope ies, en i onmen al condi ions, beha iou s,
and unc ions o e he ope a ional li e ime o an IoT de ice ha a e mapped o he IoT DT is a
ma e ha needs u he esea ch; mo eo e , u he analysis is needed also o p ope ly de elop
IoT DT use in e aces ha can connec he IoT DTs wi h o he de ices, wi h humans, and wi h
o he DTs coming om di e en domains (e.g., elco DTs o connec i i y DTs).
The li ecycle e olu ion o he IoT DTs mus ake in o conside a ion he upda abili y and
upg adabili y o he IoT de ices, including new ea u es add essing he dependabili y
cha ac e is ics.
Ope a ional in elligence is used o build IoT DTs as i suppo s digi ising he IoT and edge
compu ing in as uc u e, moni o ing ope a ions in eal- ime, p edic ing e en s, aking ac ions
based on in elligence and engaging wi h di e en s akeholde s.
The i ual ep esen a ion o an IoT DT e lec s all he ele an dynamics, cha ac e is ics, c i ical
componen s, and essen ial p ope ies o he IoT physical de ice h oughou i s li e cycle. The
c ea ion and upda e o IoT DTs ely on imely and eliable mul i-sense wi eless sensing, while he
cybe -physical in e ac ion elies on imely and dependable wi eless con ol o e many
in e ac ion poin s, whe e wi eless in e aces o he IoT de ice a e embedded.
In u u e ne wo ks, IoT DTs will be a aluable ool o c ea e no el and dis up i e solu ions,
especially o e ical indus ies, ha a e enabled by a la ge scale o eal- ime, obus , and
seamless in e ac ions among, o example, machines, humans, and en i onmen s.
IoT DTs need o be scaled up in IoT applica ions, hus enabling o he b oades possible
popula ion a sus ainable li ing wi h sys ema ic clima e mi iga ion measu es, imp o ing socie y's
esilience in c ises by ac i ely moni o ing and simula ing a huge numbe o u u e scena ios, and
po en ially helping ans o m he whole socie al s uc u e, so o make i mo e obus and
capable o add essing he en i onmen al (bu no only) challenges o he u u e.
© AIOTI. All igh s ese ed.
37
IoT DTs mus possess a minimum se o a ibu es o be in eg a ed in o IoT applica ions and
pla o ms o op imise he unc ions and se ices o hese applica ions. These IoT DT a ibu es a e
summa ised below:
• Abs ac ness – ee om de ails which a e speci ic o implemen a ions.
• Co ec ness – gi e a co ec eplica ion o he IoT ecosys em and i s de ices.
• Comple eness – upda ed is a is he unc ionali y in he eal-wo ld sys em.
• Expandabili y – adap easily o eme ging echnologies and applica ions.
• Pa ame e ised – accessible o analysis, design, and implemen a ion.
• Rep oducible – be able o eplica e he same esul o he same inpu as he eal sys em.
• Scalabili y – mus be able o ope a e a any scale.
• Soundness – exhibi only he unc ionali y a ailable in he eal-wo ld sys em.
IoT DTs will con inue o g ow in indus ial and p oduc ion en i onmen s, leading o he new
designing app oach called massi e winning. I will enable o go beyond he cu en le els o
agili y o p oduc ion, hus allowing mo e e icien in e ac ion o p oduc ion means o
encompass a mo e signi ican ex en o he espec i e p ocesses, and achie e he ans e o
massi e olumes o da a, as well as, o en, each unp eceden ed pe o mance and eliabili y
le els. The e olu ion o edge IoT digi al win echnology is illus a ed in Figu e 6.
Figu e 6 Edge IoT digi al win echnology e olu ion
IoT DTs, as pa o IoT echnologies and applica ions, a e being expanded o suppo mo e
applica ions, use cases and e ical indus ies, as well as combined wi h mo e echnologies,
such as speech, augmen ed eali y o an imme si e expe ience and AI capabili ies, enabling
o look inside he IoT DT by elimina ing he need o go and check he physical IoT de ice.

© AIOTI. All igh s ese ed.
38
8.2 Main T ends, Issues and Challenges
The esea ch oday ocuses on de eloping he i ual model ep esen a ion o an IoT de ice,
he e ol ing da a se s ela ing o he IoT domain, he mechanisms o dynamically synch onise
and adjus he i ual ep esen a ion ollowing he changes in o he physical IoT de ice, and he
simula ion en i onmen in which he IoT DTs will ope a e.
The IoT DT mapping o he physical en i onmen s in o he digi al wo ld is acili a ed by IoT
simula ion pla o ms and SW le e aged o c ea e a digi al model and a i ual ep esen a ion o
he physical IoT de ice. IoT digi al i ual ep esen a ion can be used o manipula e and con ol
he eal-wo ld IoT de ice h ough a eleope a ion DT modelling solu ion.
The esea ch a eas o IoT DTs mus conside he scope and augmen a ions o he IoT DT and he
ope a ional en i onmen combined wi h he unc ions needed o ealise he IoT DTs'
communica ion capabili ies, and he upda e equency equi ed o p o iding he op imal
p ecision o he IoT DT, based on da a measu ed and acqui ed du ing he ope a ion and use o
physical IoT de ices.
Fu he esea ch needs o in es iga e he di e en le els o IoT and edge compu ing in elligence
h ough cogni i e unc ions, implemen ed in he physical/digi al and i ual de ices.
Unde s anding, de ining, and designing he simula ion capabili ies o he u u e IoT pla o ms,
which will be able o p o ide di e en ideli y le els o simula ion uned by inpu pa ame e s,
ime dependency, beha iou , and p edic ion aspec s, in elligence, and IoT de ice complexi y,
a e e y challenging asks, which equi e u he in es iga ion.
8.3 Resea ch P io i ies Timeline
Table 6 IoT Digi al Twins, Modelling and Simula ion En i onmen s esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
IoT DT Models
Agg ega ion o he e ogeneous
IoT DT models.
Ene gy-e icien models.
E2E ea u es and op imisa ion.
Ho izon al and e ical
in eg a ion o IoT DT models.
IoT DT ha is capable o
modelling and simula ing he
u u e s a e and beha iou o
he IoT de ice.
IoT DT Modelling
and Simula ion
Pla o ms
IoT DT pla o ms a he edge.
Vi ual sensing and ac ua ion
unc ions and simula ions.
P edic i e modelling pla o ms.
Modelling and simula ion o
ene gy e iciency.
In eg a ed IoT pla o ms wi h
i ual simula ion en i onmen s
including XR.
IoT DT Secu i y
IoT DT secu i y ea u es
in eg a ed.
IoT DTs coun e ei ing
iden i ica ion and mi iga ion.
Au oma ic ecogni ion o ake
DTs and hei isola ion o
elimina ion.
IoT DT
Connec i i y
Simula ion and modelling o he
communica ion channels.
De ine in luence o he
en i onmen s on he
communica ion pa ame e s o
he IoT DTs.
Vi ual pla o ms o he
connec i i y o IoT de ices.
© AIOTI. All igh s ese ed.
39
9. IoT Swa m Sys ems
Swa m in elligence o edge IoT sys ems deals wi h a ious IoT de ices coo dina ed using sel -
o ganisa ion, decen alised and dis ibu ed con ol. I consis s o simple au onomous edge IoT
de ices and agen s ha come as eme gen collec i e in elligence.
Swa m in elligence add esses he a ea o IoT sys em collec i e beha iou and is inspi ed by
social swa ms in na u e such as bi d locks, an colonies and honeybees.
Swa m and edge compu ing combined wi h IoT a e used o implemen he collec i e beha iou
o physical and i ual AI decen alised and dis ibu ed IoT sys ems.
F om ha concep new e ms such as "a i icial in elligence o hings", "in e ne o in elligen
hings" a e de i ed and a e eme ging o add ess collec i e in elligence in IoT sys ems wi h no
cen alised con ol in as uc u e.
These new de elopmen s c ea e new "In e ne o Robo ic Things", "In e ne o In elligen Mobile
Things" and "In elligen IoT De ices Colonies" applica ions ha apply he p inciple o swa m and
edge compu ing o lee s and g oups o edge IoT de ices, suppo ing an always inc easing
numbe o i ems as he echnology imp o es.
9.1 Technological de elopmen s
The u u e edge IoT swa m sys ems b ing undamen al esea ch challenges in c oss-domain IoT
esou ce o ches a ion. Fo ins ance, in designing new mul i-in elligen -agen -based edge IoT
amewo ks. Each IoT de ice could be associa ed in he u u e wi h an agen o a DT a he
edge. The agen s could u ilise swa m in elligence o join ly op imise he ope a ions and
in o ma ion low o hei espec i e IoT de ices.
The mo e owa ds collec i e in elligence in edge IoT sys ems equi es in elligen o ches a o s
o he e ogeneous sys ems leading o swa m compu ing concep s.
To achie e ha , new in elligen p og amming o ches a ions o dis ibu ed and decen alised
open a chi ec u es and me hods o upda ing/upg ading o e - he-ai (OTA) IoT sma de ices
a e equi ed.
Collabo a i e unc ions ask o le e aging edge IoT swa m sys ems o imp o e ne wo k
connec i i y, enhancing in o ma ion collec ion abili y h ough each edge IoT de ice and in a-
ne wo king issues o a swa m o edge IoT de ices.
The ene gy spen on edge IoT de ices du ing swa m ask execu ion is a key aspec o be
conside ed. To al ene gy consump ion includes s a ic ene gy consump ion and dynamic
ene gy consump ion du ing di e en asks and he connec i i y o o he IoT de ices and edge
p ocessing.
The de elopmen o collabo a i e edge IoT swa m sys ems aise he issue o ulne abili y o
a acks by hacke s om o he lee s o ia he in e ace wi h he cloud. The inhe en dis ibu ed
a chi ec u e o edge IoT swa m sys ems makes hem mo e obus and secu e and implies ha
he in o ma ion is sha ed, s o ed, exchanged, and analysed locally and inside he swa m lee ,
making i ha de o hacke s o access sensible da a.
© AIOTI. All igh s ese ed.
40
The eal- ime p ocessing and esponse o edge IoT swa m sys ems make i ha d o malicious
a acke s o de ec de ices' sensi i e in o ma ion. '
The edge IoT sys ems in e aces o o he edge-cloud in as uc u e de i e se e al secu i y
p oblems ha ha e o be conside ed (swa m iden i y (ID) leakage, o ge y, ampe ing, spam,
jamming, swa m IoT de ice impe sona ion).
The dis ibu ed na u e o edge IoT swa m sys ems equi e new e icien se ice disco e y
p o ocols o design such ha IoT de ices and use s can iden i y and loca e he ele an swa m
se ices and p o ide s o sa is y he applica ion equi emen s.
Real- ime op imisa ion p o ocols o edge IoT swa m sys ems a e equi ed as u u e mobile IoT
applica ions a e dynamic and in ol e using online edge esou ce o ches a ion capabili ies and
p o isioning o con inuously handle dynamic edge IoT swa m de ices wo kloads and asks.
The dis ibu ed edge compu ing a chi ec u e o edge IoT swa m sys ems shi s he esea ch on
how us and secu i y a e add essed. The de elopmen o decen alised us solu ions wi h
se ices p o ided by di e en secu ed and us wo hy edge IoT en i ies is pa o he new E2E
implemen a ions.
New e icien secu i y mechanisms o edge IoT swa m sys ems a e equi ed o ensu e IoT de ice
au hen ica ion, da a in eg i y, and mu ual IoT swa m pla o m e i ica ion and alida ion. New
secu e ou ing schemes and us ne wo k opologies a e key o edge IoT swa m sys ems.
The de elopmen edge IoT swa m sys ems mus conside he balance be ween he HW and SW
cons ain s o edge IoT swa m de ices combined wi h he le el o in elligence implemen ed in
he edge de ices and hei connec i i y capabili ies o pe o m in a swa m lee adequa ely. As
a esul , de eloping SW, HW, and AI-based algo i hms o handling compu a ion o loading om
edge o he IoT swa m de ice is a c i ical issue o be add essed.
The inhe en dis ibu ed a chi ec u e o u u e edge IoT swa m sys ems conside in e ope abili y,
collabo a ion as pa o he sys em, and he pla o ms ope a ing hese sys ems include e icien
algo i hms o acili a e collabo a ion.
Edge IoT swa m sys ems may display a conside able le el o he e ogenei y in e ms o IoT
de ices, le el o in elligence, connec i i y, and p ocessing capabili ies and, he e o e, e icien
IoT pla o ms ha deal wi h his he e ogenei y a e highly desi ed.
The collabo a i e in elligen in e ac ion among he IoT swa ms equi es u he esea ch on
ede a ed lea ning o execu e o ain ML models in edge IoT de ices as pa o a swa m lee .
9.2 Main T ends, Issues and Challenges
The esea ch in IoT edge swa m sys ems complemen s he adi ional AI app oach o cogni ion
and combines indi idual easoning, collabo a i e in elligence, and mobile and sensing
in elligence.
The IoT swa m compu ing mus add ess he in e ac ion wi h he en i onmen in which he swa m
IoT lee s ope a e, including esea ch conce ning ede a ed lea ning p ocesses and simple
pe cep ions o o ms, ecogni ion o o he IoT de ices, and he swa m lee s' in e ac ion wi h
o he lee s and humans.
© AIOTI. All igh s ese ed.
41
The e olu ion o mobile edge IoT lee s ha mimic complex beha iou s mus conside he
de elopmen o echniques o model and simula e he in e ac ion o simple edge IoT de ices,
o explo e psychological ideas and he na u e o collabo a i e in elligence in mobile IoT lee s.
Resea ch on expanding he concep o in elligen agen s ecei ing eal- ime da a combined
wi h mesh compu ing applied o IoT edge swa m sys ems can suppo he ad ancemen in
collabo a i e in elligence o edge IoT.
Edge IoT swa m-based in elligence assumes ha a g oup o IoT swa m de ices o in elligen
agen s can pe o m asks wi hou explici ep esen a ions o he en i onmen , and he IoT swa m
de ices o agen s ope a e combining planning wi h eac i i y.
The sel -o ganisa ion o low pa e ns in edge IoT swa ms esea ch is a challenge, especially in
p o iding solu ions o in elligen and e icien beha iou o he whole IoT swa m lee when we
combine he limi ed in elligence o he indi idual IoT de ices.
Fu he esea ch needs o s udy he essen ial elemen s o edge IoT swa m dynamics p o iding
mechanisms o implemen ing such beha iou s and HW, SW, algo i hms o modelling,
simula ing, and in eg a ing sel -o ganisa ion agen s in o edge IoT de ices.
One o he majo challenges o edge IoT swa m sys ems is o encode and deploy he collec i e
beha iou al cha ac e is ics o a lee o edge IoT swa m-based lee s in o he beha iou o he
indi idual edge IoT de ices.
Resea ch on complex adap i e beha iou is needed o p o ide HW, SW, and connec i i y
mechanisms o implemen in e ac ions be ween edge IoT de ices as dis inc om beha iou
ha di ec ly esul s om he ope a ions o indi idual edge IoT de ices.
The ad ances in edge IoT swa m sys ems a e di ec ly linked wi h he esea ch in p o iding
op imised mechanisms and echniques o p o iding he abili y o mobile IoT swa m de ices o
sepa a e, cohe e, align, and a oid obs acles.
Sepa a ion is he abili y o ope a e wi h o he IoT de ices and keep a ce ain sepa a ion
dis ance om di e en de ices and a oid c owding oo closely oge he in a lee . The cohesion
unc ion allows he IoT de ices o app oach o o m a g oup o edge IoT de ices.
The alignmen ea u e suppo s he de ices o iden i y he expec ed nex mo emen and
synch onise he mo emen s o IoT de ices ela i e o each o he .
Na iga ing and a oiding obs acles (e.g., pe cep ion and p ocessing e icien collision de ec ion
algo i hms) is ano he c i ical challenge o be add essed by esea ch in IoT swa m-based lee s.
The wo k in edge IoT swa m sys ems mus be combined wi h he la es ad ances in
neu omo phic compu ing wi h edge IoT de ices using neu omo phic HW and collabo a i e
neu omo phic ope a ing sys ems (OS).
The connec i i y solu ions in edge IoT swa m sys ems becomes c i ical in implemen ing
collabo a i e in elligence. New ul a-low-powe mesh communica ion p o ocols mus be
combined wi h neu omo phic a chi ec u e o p o ide sel -X unc ions o edge IoT swa m
sys ems.
© AIOTI. All igh s ese ed.
48
He e, mic o-ne wo ks o po en ial di e en owne ship and wi h a po en ially ex e nal secu i y
managemen migh sha e pa s o he in as uc u e wi h wide a ea ne wo ks, pa ing he way
o a se -up whe e a p i a e ne wo k wi h pa ly owned in as uc u e and a p i a e us policy
a e in eg a ed in a public ne wo k.
11.2 Main T ends, Issues and Challenges
Decision making and op imisa ion: aking in o conside a ion an E2E sys em made o h ee
elemen s, he edge, he ne wo k, and he cloud (whe e in o ma ion lows om he edge h ough
he ne wo k o he cloud and ice e sa), c i ical issues a e wha a e hese decisions, and
equally impo an , whe e do hose decisions ake place, wi h which implica ions on he o e all
ne wo k.
The main conside a ion is ha any ne wo k is a complex en i y, since i is dis ibu ed,
he e ogeneous in di e en aspec s (da a s eams, compu a ion, and s o age capabili ies,
equi ed QoS, e c.), ime a ying (in some bu no all aspec s), and esou ce limi ed.
The applica ion o he decision heo y o he mo e and mo e impo an p oblem o dis ibu ed
esou ce alloca ion is s ill a ho esea ch ques ion ha needs o be add essed
36
.
Wha is looked o is a amewo k o model ha would allow o decide whe e in a ne wo k (edge,
cloud, in e media e s ages?), bo h p ocessing and decisions abou esou ce alloca ion a e
made, and ha does ha in a way ha is esponsi e in nea eal- ime.
Resea ch is needed in no el IoT dis ibu ed a chi ec u es o add ess he con e gence o low
la ency, Tac ile In e ne , edge p ocessing, AI and dis ibu ed secu i y based on ledge o o he
echnologies, and an e ec i e deploymen o mul i-access edge compu ing (MEC).
De eloping speci ic a chi ec u al equi emen s o dis ibu ed in elligence and con ex
awa eness a he edge is a u u e esea ch opic, especially when conside ing he in eg a ion
wi h mesh ne wo k a chi ec u es, so o o m knowledge-cen ic ne wo ks o IoT, capable o
se ing many di e en applica ions coming om nume ous e ical sec o s.
Resea ch on o ches a ion o IoT he e ogeneous ne wo ks, adap a ion o so wa e de ined adio
and ne wo king echnologies o IoT, conside ing buil -in E2E dis ibu ed secu i y as well as
ha dwa e-based secu i y solu ions
37
, us wo hiness, and p i acy issues in edge compu ing
en i onmen mus be ex ended, add essing he ede a ion and c oss-pla o m In eg a ion o
edge IoT applica ions.
36
A. Mukhe jee, P. Goswami, M. A. Khan, L. Manman, L. Yang and P. Pillai, "Ene gy-E icien Resou ce Alloca ion S a egy in Massi e IoT o Indus ial 6G
Applica ions," in IEEE In e ne o Things Jou nal, ol. 8, no. 7, pp. 5194-5201, 1 Ap il1, 2021, doi: 10.1109/JIOT.2020.3035608.
37
J. Gopika Rajan. and R. S. Ganesh, "Ha dwa e Based Da a Secu i y Techniques in IOT: A Re iew," 2022 3 d In e na ional Con e ence on Sma Elec onics
and Communica ion (ICOSEC), T ichy, India, 2022, pp. 408-413, doi: 10.1109/ICOSEC54921.2022.9952021.

© AIOTI. All igh s ese ed.
49
11.3 Resea ch P io i ies Timeline
Table 9 Decen alised and dis ibu ed edge IoT sys ems esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
Decision making
Pa ially dis ibu ed decision-
making mechanisms and
echniques.
Fede a ed sub-sys em sha ing a
common decision mechanism
Fully dis ibu ed decision-
making me hods.
Fede a ed sys ems o sys ems
sha ing he same decision
mechanisms
Fully dis ibu ed and ede a ed
sys ems, using he e ogeneous
decision mechanisms a ge ed
o speci ic QoS o e ical
sec o s.
Secu i y aspec s
Es ablished P i acy p ese ing
echniques.
Block-chain based secu i y
me hods applied o selec ed
e ical sec o s.
Scalable block-chain based
secu i y mechanisms add essing
some e ical sec o s.
T us wo hy and au o-adap ing
communica ion among se e al
e ical sec o s.
Secu e-by-design sys em o
sys ems, endowed wi h swa m
in elligence and ha dwa e-
based secu i y measu es, ully
Decen alised secu i y
p ocedu es, sel -adap ing o
eal- ime demands o
he e ogenous ac o s.
Lea ning
mechanisms
Dis ibu ed lea ning.
Fede a ed lea ning.
Dis ibu ed and ede a ed
lea ning.
Con inuous lea ning, sel -
adap ing o he dynamic
changing en i onmen o a
he e ogeneous ne wo k
suppo ing se e al e ical
sec o s.
© AIOTI. All igh s ese ed.
50
12. Fede a ed Lea ning, A i icial In elligence echnologies and
lea ning o edge IoT Sys ems
To pe o m acco ding o he de ised expec a ions
38
, he new dis ibu ed IoT a chi ec u es o
compu ing op imisa ion ac oss he edge con inuum need o imp o e esponsi eness by
educing decision-making la ency, o inc ease da a secu i y and p i acy, o dec ease powe
consump ion, using less ne wo k bandwid h, hus maximizing e iciencies, eliabili y, and
au onomy.
The IoT edge con ains compu ing capabili ies scaled ac oss he mic o-, deep- and me a-edge
o p ocess wo kloads, including he la es echnology like AI model aining and ML in e ence
and signal p ocessing, using signal condi ioning
39
ollowed by neu al ne wo ks
40
compu ing. The
neu al ne wo k compu ing and memo y equi emen s a e signi ican ly educed by using signal
condi ioning on he aw da a.
In his con ex , ede a ed lea ning as a dis ibu ed ML echnique, which c ea es a global model
by lea ning om mul iple decen alised edge clien s, is a signi ican echnological de elopmen
ha can be implemen ed in dis ibu ed IoT a chi ec u es ac oss he edge con inuum.
Fede a ed lea ning uses complex me hods o handling dis ibu ed da a aining by enabling
he coope a i e aining o common AI models, by combining and a e aging locally calcula ed
upda es submi ed by edge IoT de ices. Fede a ed lea ning pe mi s aining new models on
mul iple edge IoT de ices simul aneously wi hou he need o ha e da a s o ed in a cen al
cloud.
12.1 Technological de elopmen s
In eg a ing AI-based echniques ac oss he edge con inuum equi es a new laye o edge
p ocessing in as uc u e and scalable, ene gy-e icien modules o AI-based p ocessing
41
.
Fede a ed lea ning me hods o e se e al ad an ages, including scalabili y and da a p i acy,
wi h ML and DL algo i hms ha can be execu ed on edge IoT de ices, deli e ing as e eal-
ime insigh s o inc eased IoT applica ion e iciency. B inging AI o he edge inc ease he
e iciency o p ocessing he da a locally and educes la ency and he cos o connec i i y o
many IoT applica ions.
In he dis ibu ed da a exchange en i onmen , ede a ed lea ning/ aining combined wi h he
IoT he e ogeneous compu e based on a ious unde lying p ocessing a chi ec u es (CPUs,
GPUs, NPUs, neu omo phic, e c.) can p o ide he solu ion o u u e IoT edge in elligen
he e ogeneous sys ems.
38
J. Lin, W. Yu, N. Zhang, X. Yang, H. Zhang and W. Zhao, "A Su ey on In e ne o Things: A chi ec u e, Enabling Technologies, Secu i y and P i acy, and
Applica ions," in IEEE In e ne o Things Jou nal, ol. 4, no. 5, pp. 1125-1142, Oc . 2017, doi: 10.1109/JIOT.2017.2683200.
39
R. Ti upa hi and S. K. Ka , "Design and analysis o signal condi ioning ci cui o capaci i e senso in e acing," 2017 IEEE In e na ional Con e ence on
Powe , Con ol, Signals and Ins umen a ion Enginee ing (ICPCSI), Chennai, India, 2017, pp. 1717-1721, doi: 10.1109/ICPCSI.2017.8392007.
40
Z. Li, F. Liu, W. Yang, S. Peng and J. Zhou, "A Su ey o Con olu ional Neu al Ne wo ks: Analysis, Applica ions, and P ospec s," in IEEE T ansac ions on Neu al
Ne wo ks and Lea ning Sys ems, ol. 33, no. 12, pp. 6999-7019, Dec. 2022, doi: 10.1109/TNNLS.2021.3084827.
41
Ve mesan, O. and Na a, M.D. (Eds.). In elligen Edge-Embedded Technologies o Digi ising Indus y. Ri e Publishe s Se ies in Communica ions, June
2022. ISBN: 9788770226110, e-ISBN: 9788770226103. Online a : h ps://www. i e publishe s.com/pd /ebook/RP_E9788770226103.pd
© AIOTI. All igh s ese ed.
51
The ede a ed lea ning/ aining mus conside he unde lying connec i i y sys ems, including
Blue oo h, Wi-Fi, LPWAN, mesh, 5G and beyond applied o he IoT applica ion opology, ange,
and powe equi emen s. The con e gence o connec i i y, AI/ML, and p ocessing can ease
he implemen a ion o di e en IoT ede a ed lea ning/ aining solu ions o indus ial and
consume applica ions.
The concep o IoT ede a ed lea ning/ aining is combined wi h he IoT edge mesh, which
alloca es he decision-making asks among edge de ices wi hin he ne wo k. The compu a ion
and p ocessing asks and da a a e sha ed using an IoT ne wo ks o edge de ices, ou e s, PLCs,
and mic o-se e s.
The IoT edge mesh combined wi h ede a ed lea ning/ aining p o ide dis ibu ed p ocessing,
low la ency, aul ole ance, scalabili y, secu i y, and p i acy, as equi ed by IoT applica ions,
which demand highe eliabili y, eal- ime p ocessing, mobili y suppo , and con ex awa eness.
Coope a i e IoT compu ing based on ede a ed lea ning/ aining p o ides be e usage o
esou ces, educed la ency, due o easy access o local esou ces, be e se ices, as IoT
de ices can coope a e o ge be e in o ma ion, educed communica ion wi h cen alised
en i ies, and imp o ed secu i y and p i acy as da a emain mos o he ime wi hin a local
ne wo k.
Enabling dis ibu ed in elligence in IoT/edge compu ing o swa m compu ing applica ions is
di icul due o a se o known p oblems, e.g., synch onisa ion, consensus, coope a ion e c. In
addi ion, due o scalabili y and complexi y issues o IoT sys ems, i is challenging o de e mine
how o gene a e, coo dina e and ede a e he in elligence, which edge IoT de ice p o ides
in elligen unc ionali y and how di e en edge IoT de ices coope a e, ans e , and acqui e
in elligence.
12.2 Main T ends, Issues and Challenges
The challenge o many IoT applica ions is ha ede a ed lea ning/ aining uses mul iple en i ies
collabo a ing o sol e ML p oblems unde he coo dina ion o a cen al se e . This app oach
o edge ne wo ks c ea es many issues conce ning secu i y and p i acy and he da a i sel .
Fede a ed lea ning aises se e al isks and weaknesses in e ms o compu a ional complexi y in
he case o he e ogeneous edge IoT de ices ha may ha e limi ed compu ing esou ces,
inadequa e wi eless connec i i y quali y, o may use di e en OSs.
Ano he edge IoT ede a ed lea ning challenge ela es o communica ion delays exp essed as
he la ency be ween edge IoT de ices and he ML sys em. Dec easing la ency is c i ical o
AI‐based edge IoT de ices ope a ing in eal- ime applica ions such as indus ial equipmen .
Replacing he clien -se e p ocess o he ede a ed lea ning/ aining model wi h ully
decen alised lea ning eplaces communica ion wi h he se e by pee - o-pee
communica ion be ween indi idual clien s.
Op imisa ion algo i hms a e necessa y o implemen ede a ed lea ning/ aining a he edge,
conside ing edge IoT de ices' cons ain s and esou ce limi a ions as pa o he edge
con inuum.
The IoT de ices limi ed bandwid h can es ain scalabili y, bu his is sol ed in IoT edge mesh
a chi ec u es, as da a is sen o mul iple edge IoT de ices ha sha e da a wi h o he de ices.
The communica ion bo leneck issue is esol ed due o he dis ibu ed na u e o he sys em.
© AIOTI. All igh s ese ed.
52
A he same ime, he compu a ion asks a e o loaded o di e en edge IoT de ices, ope a ion
which speeds up he p ocessing ime and inc eases he e iciency o ede a ed
lea ning/ aining, leading o be e esponse ime, educed make span, and highe h oughpu .
The dis ibu ion o loads d i es he edge IoT sys ems o be mo e lexible and obus , as, in he
case o a de ice ailu e, o he de ices can sha e he load o he ailed IoT de ice.
IoT sys ems a e dynamic, as de ices can be mobile, added, emo ed, o changed in
con igu a ion; all o he abo e equi e new con ex -awa e solu ions o dis ibu ed secu i y and
p i acy algo i hms.
The he e ogenei y o compu ing, communica ion and AI echnologies equi es AI-based
algo i hms ha a e po able ac oss di e en IoT edge en i onmen s. Communica ion
echnology is in insically he e ogeneous o wha conce ns da a a e, ansmission ange, and
bandwid h. The IoT SW solu ions depend on he HW, and p og amming models a e needed o
execu e wo kloads simul aneously a mul iple HW le els.
Resea ch on s anda d p o ocols and in e aces should add ess he in eg a ion o AI-based
algo i hms and ligh weigh p o ocols o communica ing wi h di e en de ices in a
he e ogeneous en i onmen .
Mo e esea ch is needed o de elop dis ibu ed lea ning/ aining algo i hms and o maximise
he a e age ime be ween e o s and op imise he a ailabili y by minimising he ailu e
p obabili y and a e age eco e y ime in he OTA lea ning/ aining p ocess.
The IoT ede a ed lea ning/ aining mus conside he hyb id compu ing me hod ha combines
HW/SW and AI echniques ac oss he edge con inuum. Hyb id compu ing implies he in eg a ion
o specialised ad anced AI p ocesso s a di e en compu ing le els o bo h high-le el and low-
le el ope a ions.
Fu he esea ch is needed o de elop op imised algo i hms o ede a ed lea ning/ aining o
complemen and e ec i ely le e age he compu ing capabili ies wi h he AI-based p ocesso s
and o he ypes o p ocesso a chi ec u es.
© AIOTI. All igh s ese ed.
53
12.3 Resea ch P io i ies Timeline
Table 10 Fede a ed lea ning and AI o edge IoT sys ems esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
Fede a e
lea ning
app oaches
Techniques and me hods o
in eg a e ede a ed lea ning
in o IoT/edge compu ing
sys ems.
Managemen o edge IoT
sys ems, by add essing mesh
ne wo k secu i y and
managemen by le e aging
ML.
Resea ch on cen al aining
da a se s and edge IoT local
da a se s.
Edge IoT in elligence
a chi ec u es and AI
amewo ks o ede a ed
lea ning.
De elopmen o ools and ool
chains o dedica ed edge IoT
ede a ed lea ning.
Me hods o p o iding
e e ence aining da ase s o
pe o ming s anda d ede a ed
lea ning applica ion uning.
Benchma king echniques and
me hods o edge IoT ede a ed
lea ning.
Scalabili y and po abili y o AI-
based models o ede a ed
lea ning ac oss he edge
g anula i y.
Fede a ed
lea ning
a chi ec u e and
amewo ks
Ad anced a chi ec u al
app oaches o he ede a ed
lea ning se e in eg a ed in o
mesh ne wo king en i onmen s.
Ex end he capabili ies o open-
sou ce ede a ed lea ning
amewo ks.
Communica ion and
compu a ion e iciency o he
ede a ed lea ning
a chi ec u es, synch onisa ion
op imisa ion among edge IoT
de ices.
Fede a ed lea ning
a chi ec u es add essing ask
scheduling, dynamic esou ce
alloca ion o achie e low-
la ency se ices.
Ha dwa e
pla o ms o
ede a ed
lea ning
HW equi emen s o
implemen ing ede a ed
lea ning in edge IoT compu ing
en i onmen s.
HW he e ogenous solu ions o
minimise memo y ans e ,
inc ease ene gy e icien and
imp o e compu a ional speed.
Compu a ion o loading and
con en caching using dynamic
cache alloca ion echniques,
con ex -awa e o loading
algo i hms adap ed o
esou ce-cons ained (e.g.,
limi ed s o age and capaci y)
edge IoT de ices.
HW/SW/AI algo i hms
he e ogenei y managemen .
Unde s anding o he e ec o
sys em he e ogenei y on he AI
model agg ega ion e iciency,
accu acy and he di e gence
o con e gence o op imisa ion
p ocesses.

© AIOTI. All igh s ese ed.
54
13. Ope a ing Sys ems and O ches a ion Concep s o edge IoT
Sys ems
The ex ension o he IoT edge and edge-cloud ede a ion is ede ining he use o OSs and
o ches a ion concep s ac oss he IoT-edge-cloud con inuum in eal- ime applica ions.
The inc eased use o embedded sys ems combined wi h p ocessing uni s a he edge equi es
new ways o i ualising he compu ing capabili ies and he OSs ac oss he di e si y o he HW
componen s a he IoT edge. This also equi es mo e equen upda es and upg ades o FW, SW,
algo i hms, and secu i y pa ches o edge IoT de ices. Challenges a ise wi h mo e unc ions and
se ices implemen ed in FW/SW/algo i hms and he inc easing complexi y o elec onic
componen in e ope abili y, including managing agile and ulne able edge IoT de ices.
13.1 Technological de elopmen s
The de elopmen o ubiqui ous me a-OSs
42
o p o ide an in eg a ion amewo k ha
implemen s p ocessing and a be e aceabili y, sa e y, and secu i y ac oss he IoT edge mus
conside se e al challenges. Ubiqui ous me a-OSs o IoT edge applica ions mus embed
ea u es and unc ions o allow he applica ions o be au onomous, coope a i e, si ua ional,
e ol able, and us wo hy.
As he he e ogenei y o edge IoT inc eases, edge IoT de ices need o ac wi h a alid secu e
me hod, bo h using cen alised and dis ibu ed in elligence. The de elopmen s o ubiqui ous
me a-OSs o IoT edge applica ions a e no ye eady o ensu e scalabili y and p o ide he
unc ionali ies and capabili ies ha would allow Billions o edge IoT de ices o wo k
independen ly o each o he and coope a e in eal- ime.
Among he bigges challenges o ubiqui ous me a-OSs o IoT edge applica ions one can
men ion how o e ec i ely p o ide eal- ime ea u es o applica ions and OS capabili ies o
he e ogeneous edge IoT de ices, and how o in e ac wi h ede a ed pla o ms and in e ace
wi h cloud p o ide agnos ic solu ions. Building suppo ing pla o ms would ha e a subs an ial
impac in p o iding solu ions o hose issues.
The challenges o au onomous o ches a ion o dis ibu ed edge sys ems a e ela ed o
maximising u ilisa ion while assu ing he Se ice Le el Objec i e (SLO) o wo kloads unning a
esou ce cons ain edges and dealing wi h inhomogeneous/he e ogeneous pla o ms a he
IoT edge.
The abo e-men ioned challenges need o add ess c i ical ea u es, like ne wo k slicing, ha
open o an en i ely new se o use cases. Fo example, gi en a se o ne wo k slices, how do we
assu e ha each one will hi I s pe o mance a ge s while a he same ime no o e -p o isioning
esou ces a la ge. O e all, he o ches a ion needs o become mo e au onomous and li he
bu den o Si e Reliabili y Enginee s (SRE) unning he sys ems o da e.
A i icial In elligence Ope a ions (AIOps) and Machine Lea ning Ope a ions (MLOps) echniques
and me hodologies could o m he basis o genuinely au onomous sys ems a he edge.
42
P T akadas, e Al., “A Re e ence A chi ec u e o Cloud–Edge Me a-Ope a ing Sys ems Enabling C oss-Domain, Da a-In ensi e, ML-Assis ed Applica ions:
A chi ec u al O e iew and Key Concep s”, in Senso s 2022, 22(22), 9003; h ps://doi.o g/10.3390/s22229003.
© AIOTI. All igh s ese ed.
55
The he e ogenei y o IoT edge b ings a pa adigm shi owa ds applica ion-d i en componen s
and sys ems, in e ope abili y, and open-sou ce HW, SW solu ions. The e is a s ong need o a
da a s a egy and es ablishing us ed da a exchange amewo ks ac oss OEMs, sys em
in eg a o s, and componen endo s.
Mission- and sa e y-c i ical eal- ime edge IoT applica ions equi e gua an eed la ency and
bandwid h, he in eg i y o da a, secu i y, esilience, and con olled mesh ne wo king. Ubiqui ous
me a-OSs o IoT edge applica ions can ac as o ches a o s ha agg ega e and compose
se ices acco ding o use s' equi emen s and coo dina e hei execu ion in a cohe en and
sma way.
Ubiqui ous me a-OS o IoT edge and o ches a ing me hods a e needed o simpli y he
de elopmen , o ches a ion, and secu i y o dis ibu ed IoT edge a chi ec u es and solu ions.
The de elopmen o new solu ions using a he edge con aine s, i ual machines, and
unike nels
43
p o ides a lexible ounda ion o expanding dis ibu ed edge compu ing
deploymen s wi h a choice o he e ogeneous HW, applica ions, and ede a ed edge-clouds
in as uc u es. This b ings u he challenges o ensu e dis ibu ed i ewall, open o ches a ion
APIs, suppo o i ual machines, con aine s, and unike nels applica ion deploymen models.
13.2 Main T ends, Issues and Challenges
New esea ch and key inno a ions a e needed o add ess o ches a ion in he u u e. Fo
ins ance, no el in e aces be ween he legacy IoT a chi ec u e and he new one will be
equi ed. This will allow use s o exploi inno a ions in he con ol planes (e.g., mo ing om
cen alised sys ems o mo e decen alised ones) and new HW ea u es in he e ogeneous ( om
he compu e capabili y poin o iew) pla o ms, so o be able o place mission-c i ical
wo kloads a he edge.
OSs and o ches a ion concep s o decen alised and dis ibu ed edge IoT sys ems a e needed
o p o ide an in eg a ed en i onmen o he nex -gene a ion in elligen IoT applica ions. Such
applica ions embed IoT in dis ibu ed compu ing sys ems ope a ing in a con inuum a he edge
ac oss mic o- deep-, me a-edge and a e in e aced and ede a ed wi h he cloud solu ions.
IoT/edge compu ing pla o ms and he new IoT applica ions can c ea e new business models
elying on me a-OS o ches a ion, dis ibu ed edge in elligence, s o age, and esou ces
he e ogenei y.
Ubiqui ous me a-OSs o IoT edge and o ches a ing mechanisms a e equi ed o
he e ogeneous sys ems wi h ligh weigh i ualisa ion, i ual machines, mic ose ices and
con aine s. Key issues o be add essed a e da a p e-p ocessing a he edge, edge analy ics,
scalabili y, e iciency, dependabili y, us wo hiness, adap abili y, and anspa ency. The issues
a e e en mo e c i ical conside ing he e olu ion o IoT owa ds Tac ile IoT and IoTS.
Resea ch ac i i ies mus sol e se e al c i ical echnical challenges, including he a ailabili y o
a dis ibu ed a chi ec u e o ubiqui ous me a-OSs, scalabili y, pe o mance and applicabili y
issues, sel -X ea u es, and au onomous conside a ions.
Ubiqui ous me a-OSs de elopmen s o IoT edge mus align wi h he eme ging new compu ing
pa adigms such as IoT swa m, including in elligen obo ics hings, o ganic compu ing, b oad
usage o AI, and neu omo phic concep s.
43
h p://unike nel.o g/
© AIOTI. All igh s ese ed.
56
The de elopmen o new SW-de ined abs ac ions and capabili ies o edge IoT de ices is
needed o suppo he managemen , applica ion de elopmen , and communica ions
be ween de ices ac oss he edge and cloud con inuum.
Con ex -awa eness in he edge IoT sys em is ela ed o he con ex in o ma ion, he eal- ime
ope a ion and he dynamici y o p ocessing, scalabili y and managing he in o ma ion wi h he
suppo o he ubiqui ous me a-OS.
Resea ch should add ess he issue o c ea ing a lexible con ex modelling amewo k o p o ide
means o p esen ing, main aining, sha ing, p o ec ing, easoning, and que ying con ex
in o ma ion. In his con ex , he IoT edge me a-OS mus include adap abili y and sel -X ea u es
o allow he IoT applica ion o adap i s beha iou acco ding o he con ex . The IoT sys em can
hen be econ igu ed o adap o hese con ex changes.
Sel -X unc ions, including sel -o ganising, sel -healing, sel -p o ec ing, sel -diagnosing, sel -
econ igu ing, equi e a o m o sel -awa eness o unde s and he s a e o he edge IoT sys em
and a e o be embedded in o he me a-OS, which mus ealise i s s a e and con igu a ion, and
he s a e and he con igu a ion o he esou ces i con ols and manages.
Resea ch in de eloping models o edge IoT HW and ubiqui ous me a-OS-le el SW con igu a ion
is necessa y o desc ibe he opologies and he a ionale o he designs o IoT de ices and hei
in eg a ion in o he dis ibu ed IoT edge applica ions a chi ec u e. The de elopmen o new
so wa e-de ined abs ac ions and capabili ies o edge IoT de ices is needed o suppo he
managemen , applica ion de elopmen , and communica ions be ween de ices ac oss he
edge and cloud con inuum.
The me a-OS ea u es o IoT edge equi e adap ing o he cha ac e is ics o hese u u e
applica ions ha will be con ex -sensi i e, adap i e, cus omised, and econ igu able. The
unc ional equi emen s behind he concep s o IoT sys ems econ igu abili y, con ex -
awa eness, adap abili y, and cus omisa ion mean ha he me a-OS suppo s and allows ha
he IoT sys em’s HW and SW con igu a ions can seamlessly change a un ime.
13.3 Resea ch P io i ies Timeline
Table 11 Ope a ing sys ems and o ches a ion concep s o edge IoT sys ems esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
Decen aliza ion
echniques
Ma u e ede a ed lea ning
echniques ha p o ide clea
ad an ages in selec ed
e icals and use cases
Ad anced ede a ed lea ning
echniques sui able o all
e icals
New pa adigm o dis ibu ed
managemen
Con ex
awa eness
Unde s anding he key
su ounding pa ame e s and
unc ionali ies ha can
gua an ee an ad ance
con ex awa eness
Con ex awa eness imp o ed
wi h seman ic capabili ies, i.e.,
abs ac ing om he shee
he e ogeneous da a and going
up in he abs ac ion laye
New con ex awa eness
pa adigm
Ope a ing
Sys ems
New me a-OS ha can p oo
ad an ages agains exis ing
adi ional OSs
Enhanced dis ibu ed me a-OSs
ha adap esou ce
compu a ion and s o age
alloca ion o he un- ime
en i onmen on a ms scale
Fully au onomous and
econ igu able managemen
o he sys em esou ces in he
IoT-edge-cloud con inuum wi h
ully dis ibu ed decision
capabili ies
© AIOTI. All igh s ese ed.
57
14. Dynamic P og amming Tools and En i onmen s o Decen alised
and Dis ibu ed IoT Sys ems
Cu en p og amming en i onmen s and ools o IoT
44
p o ide a cen alised app oach (ei he
on-p emises o cloud-based) whe e he main componen ans o ms and p ocesses mos o he
compu a ion on da a supplied by he edge and a edge de ices.
This app oach un o una ely implies se e al d awbacks, o ins ance unused edge
compu a ional powe , a single poin o ailu e, and iola ion o da a bounda ies (p i a e,
echnological, e c.).
In his con ex , new esea ch is needed in le e aging he esou ces a ailable in lowe - ie
de ices o imp o e o e all dependabili y, pe o mance, scalabili y, obse abili y, and
ep oducibili y o IoT sys ems using new p og amming and ools o dis ibu ed IoT.
14.1 Technological de elopmen s
Ex ensi e edge IoT dis ibu ed sys ems a e di icul o implemen . P og amming hese sys ems is
challenging due o he equi emen o asse nume ous con ol pa hs esul ing om he
innume able in e lea ing o messages and ailu es, p oduces by a huge numbe o
he e ogeneous de ices.
The edge IoT dis ibu ed sys ems de elopmen s need o conside he in eg a ion o ligh weigh
me a-OSs, dis ibu ed da abases, middlewa e, mesh ne wo king, and applica ion a chi ec u e
ypes. This means add essing he design and analysis o dis ibu ed algo i hms, p og amming
languages, compile s, SW ools and dynamic middlewa e en i onmen s.
Managing edge IoT dis ibu ed sys ems in as uc u e equi es dedica ed se s o p og amming
ools.
The dynamic p og amming ool o coding, modelling, simula ing, and isualising he cu en
ope a ional s a e o all edge IoT de ices, including e i ica ion and alida ion componen s and
debugging and es ing modules no i ying when he ailu e occu s in he dis ibu ed IoT sys ems.
The edge IoT dis ibu ed sys ems b ing inhe en challenges such as concu ency, pa ial ailu e,
node dynamism, o asynch ony in hei design and implemen a ion.
The inc easing complexi y o he concu en ac i i ies and eac i e beha iou s in edge IoT
dis ibu ed sys ems a e unmanageable by he exis ing p og amming models, ools, and
abs ac ion mechanisms.
14.2 Main T ends, Issues and Challenges
The esea ch should conside ans o ming and decomposing da a low and pa i ioning o da a
exchange be ween he dis ibu ed IoT nodes while de ec ing cu en condi ions in deciding
when o mo e compu a ions along he edge con inuum.
44
A ecen su ey o he mos used ools can be ound a : h ps://www.io o all.com/ op-io - ools-and-pla o ms- o -io -de elopmen -and-de elope s.
© AIOTI. All igh s ese ed.
64
The nex -gene a ion edge IoT pla o ms equi e a mo e om speci ic e ical indus ial sec o s
o ho izon al IoT pla o ms ha implemen dis ibu ed a chi ec u es based on edge compu ing
capabili ies ope a ing ac oss se e al indus ial sec o s.
Such pla o ms need o in eg a e se e al connec i i y and p ocessing echnologies, ad anced
analy ics and da a managemen capabili ies ha enable de elope s o easily design and
deploy IoT solu ions as e , accele a ing ime o insigh o applica ions ope a ing ac oss se e al
indus ial sec o s.
16.2 Main T ends, Issues and Challenges
Speci ic esea ch challenges o nex -gene a ion edge IoT pla o ms a e o add ess he issues o
he e ogenei y, scalabili y and ede a ed lea ning (FL) by implemen ing dis ibu ed a chi ec u es
ha gua an ee he secu i y and p i acy o he da a exchanged by an ex ensi e numbe o
in elligen edge IoT de ices while subduing in e ope abili y issues.
The esea ch ad ances in he a ea o cogni i e cloud pla o ms equi e new solu ions o he
ede a ion o edge IoT pla o ms and he o ches a ion o edge-cloud domains o op imising
he use o he esou ces, imp o ing se ice quali y, educing he ene gy, he ine icien low o
da a, and he cos s.
Fu he esea ch is needed o add ess he echnological and seman ic in e ope abili y issues
among he e ogeneous IoT de ices and pla o ms in he con ex o implemen ing dis ibu ed
a chi ec u es and he in eg a ion o new echnologies such as swa m compu ing.
The esea ch needs o ocus on minimising he complexi y o collec ing and p ocessing as
amoun s o eal- ime da a gene a ed by in elligen IoT de ices, add ess scalabili y and secu i y
issues.
The inc eased ocus on IoT solu ions buil using open-sou ce SW and HW, which a e based on
open speci ica ions allowing po abili y and educing IoT applica ions de elopmen , equi e
u he esea ch on open-sou ce echnologies.
The esea ch and inno a ion p io i ies mus add ess he con e gence o echnologies o edge
IoT pla o ms combining he echnological de elopmen in a long lis o a eas, e.g., sensing,
p ocessing, communica ion, compu ing, AI, and s o age echnologies.
The esea ch p io i ies need o suppo he de elopmen o new open, edge IoT ho izon al
pla o ms combining he dis ibu ed unc ions o mobile edge compu ing and he in e aces o
syne gize and ede a e wi h o he edge o cloud pla o ms.
The ad ances in in elligen indus ial IoT applica ions, Tac ile IoT and au onomous/ obo ic
sys ems solu ions equi e eal- ime esponse and compu ing a he edge o he ne wo ks and FL
ac oss he edge.
New esea ch is equi ed o p o ide AI algo i hms o ope a e in a he e ogeneous dis ibu ed IoT
applica ion con ex , including a ede a ion o edge IoT and cloud pla o ms.
AI and IoT FL ad ances equi e new de elopmen amewo ks o enable he e ec i e
de elopmen o edge IoT pla o ms embedding AI-based componen s a di e en IoT
a chi ec u al laye s.
Fu he de elopmen o edge IoT pla o ms needs o suppo he c ea ion o in elligen , sel -X
unc ions (e.g., sel -o ganising, sel -healing, sel -con igu ing, sel -managing) in eg a ed in o he
pla o ms o deal wi h he in elligen au onomous IoT and swa m sys ems.

© AIOTI. All igh s ese ed.
65
These unc ions include SW/HW and AI-based algo i hms ha enable au oma ed asks (e.g., o
c ea e, p o ision, con igu e, oubleshoo ) ha manage lee s o IoT de ices and ga eways
secu ely, emo ely, in olume o indi idually.
Va ious AI echniques, ule engines, e en s eam p ocessing, da a isualisa ion and ML need o
be de eloped o add ess he mic o-, deep, and me a-edge and in eg a ed in o he IoT
pla o ms.
The IT and OT secu i y measu es and ea u es mus be conside ed and de eloped in he con ex
o dis ibu ed a chi ec u e and ac oss he edge-cloud con inuum.
The ad ances in IoT/edge compu ing need o add ess secu e da a s o age, e icien da a
e ie al and dynamic da a collec ion as pa o a p ocessing amewo k o IoT along he
compu ing con inuum conside ing he unc ions o da a p e-p ocessing, s o age and analy ics
based on bo h edge and cloud compu ing in as uc u es.
16.3 Resea ch P io i ies Timeline
Table 14 Edge IoT sec o ial and C oss-Sec o ial Open Pla o ms esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
Simula ion
capabili ies
Edge IoT pla o ms can simula e
and un a IoT de ices in a
speci ic e ical domain
Edge IoT pla o ms ac oss
mul iple IoT e icals domains
and suppo ing pa ial winning.
Edge AI pla o ms wo king and
seamless simula ing all indus ial
domains and make use o
ad anced winning.
In e ope abili y
among pla o ms
Each edge IoT pla o m has i s
own API o expose i s se ices o
he ou side wo ld
Edge IoT pla o ms ha can
exchange da a among
hemsel es, making use o he
ecen ly de ined da a spaces a
Eu opean le el.
Edge IoT dis ibu ed pla o ms
wi h embedded AI capabili ies
applied o indus ial sec o s and
use Resea ch on ad anced
edge AI pla o ms o exchange
abs ac da a se s and make
use o all de ined da a spaces
a he Eu opean le el.
Con e gence o
echnologies
Edge IoT pla o ms ha
combine dis ibu ed
a chi ec u es con e ging mesh,
DLT and AI echnologies.
Ad anced edge IoT AI-based
pla o ms wi h in eg a ed
cogni i e unc ions o
ede a ed lea ning and o he
eme ging dis ibu ed lea ning
echnologies.
Ad anced edge IoT pla o ms
including digi al wining
echnologies o add ess he
me a e se con inuum o no el
imme si e applica ions.
© AIOTI. All igh s ese ed.
66
17. IoT Ve i ica ion, Valida ion and Tes ing (VV&T) Me hods
The b oad ange o exis ing secu i y ce i ica ion schemes o p oduc s, sys ems, domains,
solu ions, se ices, and o ganiza ions de i es on a he e ogeneous en i onmen o solu ions,
making i di icul o unde s and wha is needed o achie e a ce ain le el o secu i y in each
con ex o echnology. This he e ogenei y also makes compa ing ce i ied de ices mo e di icul ,
especially when di e en ce i ica ion app oaches, coun ies, and con ex s a e used. Cu en ly,
he e is no uni ied solu ion ha copes wi h hese p oblems, acili a ing he p ocess o compa ing
and assessing he secu i y le els.
17.1 Technological de elopmen s
Due o he dynamism o secu i y de ined by he new ulne abili ies disco e ed, a ec ing he
secu i y o he change o he domain ha ha e di e en secu i y equi emen s (e.g., IoT de ices
ha passes om a medical o a home domain), he ce i ica ion app oach mus adap o hese
changing condi ions. An agile ce i ica ion p ocess is equi ed o ensu e such secu i y le el is up
o da e du ing he li ecycle o a de ice. In addi ion, he app oach mus cope wi h he business
equi emen s and needs o he ma ke . I means ha secu i y ce i ica ion app oaches should
be e icien and cos -e ec i e, so he ma ke p oduc launch is no delayed.
17.2 Main T ends, Issues and Challenges
While pu e unc ional e i ica ion o IoT sys ems s ill poses challenges, e.g., due o HW/SW
in eg a ion issues, AI a he edge, and he dis ibu ed na u e o IoT sys ems, e i ying non-
unc ional p ope ies, such as cybe secu i y o esou ce consump ion, usually is a leas as
challenging. Also, in eg a ing un us ed hi d pa y IoT de ices in sensi i e o sa e y-c i ical
sys ems o en comes wi h he challenge o p o e he absence o hidden “unwan ed”
unc ionali y (e.g., backdoo s), which can go as deep as needing o employing echniques like
bina y code analysis.
The edge IoT dis ibu ed, he e ogeneous sys ems employ mul i-languages o implemen a ion o
o se ices edge IoT de ices p o ide. Dis ibu ed sys ems use di e en o ms o concu en
p og amming ha inc ease HW concu ency and sou ces o he e ogenei y.
De eloping eliable, sa e, and secu e edge IoT dis ibu ed sys ems is a ade-o p ocess ha
mus be able o also scale among he di e en le el o pe o mance ha a e needed in he
di e en con igu a ions.
The de elopmen equi es no el models, logical no a ions, e i ica ion echniques, ex ensions,
imp o emen s, and combina ions o exis ing ones o ca ch he beha iou o such sys ems and
ensu e ha hey mee a ious speci ic equi emen s. In his con ex , ad anced o mal me hods
applied o he e i ica ion and alida ion o edge IoT concu en and dis ibu ed he e ogeneous
sys ems can p o ide new unde s andings o he e iciency o hese me hods, when compa ed
wi h o he app oaches. Gi en he inancial p essu e on IoT de ice de elopmen , any
e i ica ion ac i i y mus be cos e ec i e and e icien , hence, he au oma ion o e i ica ion
asks is an impo an end. Cos and e iciency o e i ica ion also is hea ily linked o he p ope
usage o design me hodologies like “secu i y-by-design”, as igo ous design will signi ican ly cu
down es ing e o s.
Ve i ica ion o non- unc ional p ope ies like cybe secu i y lags behind unc ional e i ica ion,
e en i he o me can ha e e en bigge inancial isk a ached.
© AIOTI. All igh s ese ed.
67
Resea ch shall ocus on new and imp o ed me hods o he au oma ed e i ica ion o non-
unc ional HW/SW p ope ies o IoT sys ems. In addi ion, he unc ional e i ica ion o edge IoT
AI-based solu ions, which is a la gely open challenge, mus be add essed.
Ve i ica ion esea ch on edge IoT sys em should conside i ual en i onmen s (incl. digi al wins),
and design app oaches p omising g ea e e iciency a educed e o - and ulne abili y a e
in o accoun . This may include o mal de elopmen / igo ous design, low-code/model-based
and o he app oaches. Finally, in es iga ing au oma ing and imp o ing he scalabili y o
e i ica ion app oaches o hi d-pa y “black-box” sys ems, which look a he HW and bina y
SW le el o IoT sys ems, is equi ed.
The he e ogenei y o edge IoT dis ibu ed sys ems makes deploying and managing hese
sys ems complex. The dis ibu ed a chi ec u e o hese sys ems adds o he complexi y, as he
IoT sys ems a e deployed ac oss ne wo ks and spa ial bounda ies.
Dis ibu ed sys ems display a high le el o concu ency, which is a challenge in es ima ing he
consis ency and co ec ness o hese sys ems, mainly when hei beha iou s a e no
app op ia ely de ined. In he edge dis ibu ed sys ems, IoT de ices become eac i e sys ems,
cons an ly in e ac ing wi h he en i onmen , and adjus ing hei s a es.
The dis ibu ed en i onmen and he changing beha iou o edge IoT de ices d i e he IoT
applica ions o be e y dynamic and, he e o e, exposed o unexpec ed beha iou . Iden i ying
unexpec ed beha iou and assu ing ha he demanded beha iou is ollowed can c ea e a
challenge in dynamic sys ems. In ended beha iou s in edge IoT dis ibu ed sys ems can be
de ined as p ope ies. Model-checking echniques a e used o see whe he hese p ope ies
hold on o he o mal speci ica ion o he sys em. In his con ex , p ope y speci ica ion,
e i ica ion and i ual alida ion can help de ec e o s in edge IoT dynamic and dis ibu ed
sys ems.
Fu he esea ch mus conside upcoming echnologies such as HW/SW neu omo phic and
swa m compu a ion. O u he in e es is he au oma ion and in eg a ion o he
design/ e i ica ion/debug cycle wi h ope a ional da a: au oma ed epai , diagnosis, and
upda es o sys ems in he ield based on obse ed ailu es.
17.3 Resea ch P io i ies Timeline
Table 15 IoT Ve i ica ion, Valida ion and Tes ing Me hods esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
Edge IoT AI-
based sys ems
VV&T
Vi ual alida ion o edge IoT
based sys em including AI-
based componen s.
VV&T o sel -X beha iou s (e.g.,
sel - con igu a ion, sel -healing,
and sel -adap ion, e c.) o edge
IoT sys ems.
De elopmen o VV&T me hods
and echniques o eal ime
and ime cons ained AI-based
edge IoT dis ibu ed sys ems.
Dis ibu ed
sys ems VV&T
Valida ion me hods o edge IoT
dis ibu ed sys ems ocusing on
bo h p oo -based e i ica ion
and sys ema ic es ing.
Mesh ne wo king sys ems
in eg a ed as pa o edge IoT
in as uc u e.
Edge IoT dis ibu ed sys ems
embedding digi al win
solu ions.
He e ogenous
sys ems VV&T
HW/SW neu omo phic and
swa m compu a ion applied o
edge IoT dis ibu ed sys ems.
He e ogenous IoT dis ibu ed
sys ems comp ising o DLT
pla o ms, edge AI amewo ks
and a ious ML
implemen a ions.
Fede a ion o mul i-blockchain-
based da a p ocessing, edge
IoT compu ing and swa m
in elligence.
He e ogenous edge IoT
in e ope abili y es ing.
© AIOTI. All igh s ese ed.
68
18. IoT T us wo hiness and Edge Compu ing Sys ems Dependabili y
T us wo hiness o he edge IoT sys ems can be de ined as he deg ee o con idence one has
ha he edge IoT sys em pe o ms as expec ed. Designing IoT edge us wo hiness in o edge
IoT sys ems equi es he iden i ica ion o cha ac e is ics and mechanisms o us ha can be
embedded in o he edge IoT sys em.
T us in edge IoT sys ems is ealised a he in e sec ion o se e al dependabili y cha ac e is ics,
as shown in Figu e 8. As a esul , us in an edge IoT sys em is a concep wi h mul iple dimensions,
combining dependabili y cha ac e is ics wi h human and machine beha iou . In his con ex ,
he e is a need o a g ea e unde s anding o how indi iduals in e ac wi h edge IoT de ices
and how edge de ices in e ac wi h o he de ices/ hings conce ning he ex ension o us .
18.1 Technological de elopmen s
The dependabili y
50
o a sys em e lec s he use 's deg ee o us in ha sys em. Dependabili y
cha ac e is ics applied o edge IoT sys ems a e sa e y, secu i y, eliabili y, esilience, a ailabili y,
connec abili y, and main ainabili y as illus a ed in Figu e 8.
Figu e 8 Edge IoT sys em dependabili y cha ac e is ics
The secu i y aspec s o IoT a e c i ical due o he amoun , di e si y, and po en ial impac o
secu i y h ea s on e e yday c i ical in as uc u es. Secu i y aspec s ep esen one o he mos
p essing ba ie s o he adop ion o la ge-scale IoT deploymen s.
50
Lap ie, JC. (1995). Dependabili y — I s A ibu es, Impai men s and Means. In: Randell, B., Lap ie, JC., Kope z, H., Li lewood, B. (eds) P edic ably
Dependable Compu ing Sys ems. ESPRIT Basic Resea ch Se ies. Sp inge , Be lin, Heidelbe g. h ps://doi.o g/10.1007/978-3-642-79789-7_1.
© AIOTI. All igh s ese ed.
69
Recen cybe a acks, such as he Mi ai
51
o B icke Bo
52
IoT Bo ne s
53
, highligh he need o design
app op ia e p o ec ion mechanisms o he nex gene a ion o IoT de ices.
To add ess cybe secu i y conce ns, one o he mos ambi ious ini ia i es in Eu ope is he de ini ion
o a cybe secu i y ce i ica ion amewo k. In his di ec ion, he ecen "Cybe secu i y Ac "
54
ep esen s he Eu opean Commission e o o s eng hen he Eu opean Union Agency o
Ne wo k and In o ma ion Secu i y (ENISA)
55
ole o he de ini ion o his amewo k by p o iding
addi ional guidelines and challenges o i s ealiza ion.
The de ini ion o a cybe secu i y amewo k equi es e o s om di e en knowledge domains
o sa is y he equi emen s and needs o a ious s akeholde s, such as manu ac u e s, ins i u ions,
legal i ms, and consume s.
On he one hand, he e y b oad ange o exis ing secu i y ce i ica ion schemes o p oduc s,
sys ems, domains, solu ions, se ices, and o ganiza ions de i es on a he e ogeneous
en i onmen o solu ions, making di icul unde s anding wha is needed o achie e a ce ain
le el o secu i y in each con ex o echnology.
This he e ogenei y also makes compa ing ce i ied de ices mo e di icul , especially when hese
de ices a e ce i ied wi h di e en ce i ica ion app oaches, coun ies, and con ex s. Cu en ly,
he e is no a uni ied solu ion ha copes wi h hese p oblems, acili a ing he p ocess o
compa ing and assessing he secu i y le el.
On he o he hand, due o he dynamism o secu i y (e.g., a new ulne abili y disco e ed
a ec ing he secu i y o a change o domain such as a de ice ha passes om a medical o a
home domain, implying di e en secu i y equi emen s), he ce i ica ion app oach mus
conside hese as -changing condi ions, which could a ec he de ice’s secu i y le el.
IoT T us wo hiness equi es in eg a ing s akeholde s o he c i ical sec o (ne wo k ope a o s,
echnology supplie s, cybe secu i y solu ion p o ide s, s anda d and ce i ica ion bodies, e c.)
o de ining cybe secu i y s anda ds and es p ocedu es.
18.2 Main T ends, Issues and Challenges
Wi h IoT de ices aking o e c i ical asks in sys em con ol and/o heal h a eas, dependabili y is
one majo end: da a p o ided mus be eliable and us wo hy.
Dependable edge IoT sys ems will inco po a e de ices and se ices o mul iple di e en endo s
which will only inc ease he challenge. Fo example, such sys ems need o cope wi h da a
p i acy, cybe secu i y, eliabili y, and sa e y a he same ime.
Wi hou p ope s anda ds and planning, gua an eeing all hese a ibu es o e a se o di e se
de ices om di e en endo s (being deployed in di e en ne wo ks and da a cen es) is nex
o impossible.
51
G. Ta eba ake and S. Yamaguchi, "Ma hema ical Modeling and Analysis o he Dic iona y A ack Mechanism in IoT Malwa e Mi ai," 2022 IEEE In e na ional
Con e ence on Consume Elec onics-Asia (ICCE-Asia), Yeosu, Ko ea, Republic o , 2022, pp. 1-5, doi: 10.1109/ICCE-Asia57006.2022.9954838.
52
C. Kolias, G. Kambou akis, A. S a ou and J. Voas, "DDoS in he IoT: Mi ai and O he Bo ne s," in Compu e , ol. 50, no. 7, pp. 80-84, 2017, doi:
10.1109/MC.2017.201.
53
R. Raman, "De ec ion o Malwa e A acks in an IoT based Ne wo ks," 2022 Six h In e na ional Con e ence on I-SMAC (IoT in Social, Mobile, Analy ics and
Cloud) (I-SMAC), Dha an, Nepal, 2022, pp. 430-433, doi: 10.1109/I-SMAC55078.2022.9987253.
54
h ps://digi al-s a egy.ec.eu opa.eu/en/policies/cybe secu i y-ac .
55
h ps://www.enisa.eu opa.eu/

© AIOTI. All igh s ese ed.
70
Dependabili y managemen o e he comple e p oduc li ecycle will become key and include
echniques such as AI-augmen ed De Ops, sa e y and secu i y-by design, dependabili y
a chi ec u es, low-code de elopmen , code- e-use, and p ope secu i y managemen o e he
whole p oduc li ecycle.
In a mo e echnology-dependen wo ld, cybe secu i y conce ns ha e a ac ed an inc easing
in e es om companies, egula o y bodies, and end use s. Wi h he IoT, such aspec s a e
exace ba ed due o he amoun , di e si y, and po en ial impac o secu i y h ea s on e e yday
c i ical in as uc u es. Indeed, nowadays, secu i y aspec s ep esen one o he mos impo an
ba ie s o he adop ion o la ge-scale IoT deploymen s.
The e o e, an agile ce i ica ion p ocess is equi ed o ensu e such secu i y le el is up o da e
du ing he li ecycle o a de ice. In addi ion, he app oach mus cope wi h he business
equi emen s and needs om he ma ke . I means ha secu i y ce i ica ion app oaches should
be e icien and cos e ec i e, so he p oduc launch in he ma ke is no delayed.
As a esul o his p ocess, a cybe secu i y label is looked o , which con ains he secu i y le el
achie ed by he de ice. I should p o ide a clea isibili y o he le el o secu i y achie ed bu
also gi e a non-ambiguous and comple e ep esen a ion o he esul s o he cybe secu i y
e alua ion p ocess.
This is a he di icul o achie e, because in compa ison o he ene gy label, which measu e a
physical quan i y, he measu emen o secu i y is a mo e complex, in ol ing se e al secu i y
p ope ies o dimensions. In addi ion, he label should be able o show in eal ime he secu i y
o e ed by he de ice.
The IoT and edge compu ing sys ems ha e challenges in add essing he au oma ed
cybe secu i y e alua ion owa ds a mo e agile and cos -e ec i e ce i ica ion, dealing wi h he
IoT dynamism.
Finally, he e is also he need o objec i es and e idence-based e alua ion me hodologies ha
would allow o a homogeneous e alua ion, including compa abili y aspec s. This limi s he
secu i y managemen mechanisms add essing he whole li ecycle o he IoT de ice, dealing
wi h he secu i y changes ha migh in alida e he ce i ica e.
© AIOTI. All igh s ese ed.
71
18.3 Resea ch P io i ies Timeline
Table 16 IoT T us wo hiness and Edge Compu ing Sys ems Dependabili y esea ch p io i ies
Topic
Sho Te m
Medium Te m
Long e m
2023-2024
2025-2027
2028-2030
T us wo hiness
Edge IoT us wo hiness o ML
models and explainabili y o AI
(XAI) models used in Fede a ed
Lea ning.
T us wo hiness models o edge
IoT sel -adap ing sys ems based
on digi al win and AI-based
echnologies.
F amewo ks p o iding
guidelines, good p ac ices and
s anda ds o ien ed o end- o-
end us in edge IoT sys ems.
Dependabili y
De ine me hods and ools o
suppo he dependabili y
sys em p ope ies de ini ion,
composi ion and alida ion and
ela e he p ope ies o di e en
s anda ds add essing di e en
echnologies.
Dependabili y p ope ies a
di e en laye s o he edge IoT
a chi ec u e by conside ing
scalable concep s o HW, SW,
connec i i y, Al algo i hms
(in e ence, lea ning) and he
design o lexible
he e ogeneous a chi ec u es
ha op imise he use o
compu ing esou ces and he
use o esou ce-cons ained
de ices.
Vi ualisa ion and simula ion
ools o managing he
e alua ion o edge IoT sys em
dependabili y.
Benchma king
De ine di e en benchma king
me hods and echniques o
us o an edge IoT sys em and
p o ide compliance o an
ag eed-upon s anda d ia
ce i ica ion schemes.
Benchma king he e ogenous
edge IoT sys ems based on DLT,
digi al wins, mesh ne wo king
and AI echnologies.
Benchma king e e ence da a
se s o aining edge IoT sys ems
o ede a ed lea ning.
© AIOTI. All igh s ese ed.
72
Con ibu o s
Edi o s:
O idiu Ve mesan, SINTEF
Vale io F ascolla, In el
Re iewe :
Dami Filipo ic, AIOTI Sec e a y Gene al
Con ibu o s:
Alessand o Liani, Video Sys ems, I aly
And é Bou doux, IMEC, Belgium
An onio Ska me a, Uni e sidad de Mu cia, Spain
An onio So iano Asensi, Uni e si a de Valencia, Spain
Asbjo n Ho s o, Ha ens om, No way
Cha les S u man, Huawei, UK
Cian O Mu chu, Tyndall Na ional Ins i u e, I eland
Daniela Buleand a, SIMAVI, Romania
F ançois Fische , FSCOM, F ance
Geo ge Suciu, BEIA Consul In e na ional SRL, Romania
Geo gios Ka agiannis, Huawei, Ge many
Ignacio Lacalle, Uni e si a Poli ecnica de Valencia, Spain
Ilia Pie i, In acom Telecom Solu ions, G eece
Ja a Pascual, Collabwi h G oup, Ne he lands
Jesus Angel Ga cia Sanchez, Ind a Sis emas
Joachim Hilleb and, Vi ual Vehicle Resea ch, Aus ia
Juho Pi skanen, Wi epas, Finland
Kons an inos Loupos, Inlecom, G eece
Kons an inos Thi aios, NETCOMPANY INTRASOFT, Luxembou g
© AIOTI. All igh s ese ed.
73
Laza os Ka agiannidis, Ins i u e o Communica ion and Compu e Sys ems, G eece
Luis Munoz, Uni e si y o Can ab ia, Spain
Luis Pe ez-F ei e, Fundacion Cen o Tecnoloxico de Telecomunicacions de Galicia, Spain
Ma cin Plociennik, Poznan Supe compu ing and Ne wo king Cen e , Poland
Ma co Gonzalez Hie o, IKERLAN, Spain
Ma io D obics, AIT Aus ian Ins i u e o Technology GmbH, Aus ia
Ma in Se ano, Insigh SFI esea ch Cen e o Da a Analy ics, I eland
Mau izio Spi i o, Fondazione LINKS, I aly
Mikel La anaga, Fundacion Teknike , Spain
Mi ko P esse , Aa hus Uni e si y, Denma k
Monica Flo ea, SIMAVI, Romania
Na alie Samo ich, Ene cou im, Po ugal
Ol ion Xhezo, Voda one, UK
O idiu Ve mesan, SINTEF, No way
Ped o Maló, UNPARALLEL Inno a ion, Po ugal
Ped o Ruiz, In eg asys, Spain
Philippe Sayegh, Ve ses Global, Ne he lands
Pie e Y es Dane , 48deg79min-Consul ing, F ance
Pie e Becue, IMEC, Belgium
Pie o Dionisio, Medea, I aly
Ranga Rao Venka esha P asad, Technical Uni e si y Del , Ne he lands
Rica do Vi o ino, Ubiwhe e, Po ugal
Roumen Nikolo , Vi ech, Bulga ia
Roy Bah , SINTEF, No way
Ru e So ia, o iss, Ge many
Sean McG a h, Uni e si y o Lime ick, I eland
Se gio Gusme oli, Poli ecnico di Milano, I aly
S djan K co, Duna NET doo, Se bia
Vale io F ascolla, In el, Ge many
Ve onica Quin una Rod iguez, O ange, F ance