Age of Information-Aware Cognitive Shared Access Networks with Energy Harvesting

Age o In o ma ion-Awa e Cogni i e Sha ed Access
Ne wo ks wi h Ene gy Ha es ing
Geo gios Smpokos1, Dionysis Xenakis2, Ma ios Koun ou is3, and Nikolaos Pappas1
1Link¨
oping Uni e si y, Sweden
2Na ional and Kapodis ian Uni e si y o A hens, G eece
3Uni e si y o G anada, Spain
Abs ac —This s udy in es iga es a cogni i e sha ed access
ne wo k wi h ene gy ha es ing capabili ies ope a ing unde Age
o In o ma ion (AoI) cons ain s o he p ima y use . Seconda y
ansmi e s a e spa ially dis ibu ed acco ding o a homogeneous
Poisson Poin P ocess (PPP), while he p ima y use is loca ed a
a ixed posi ion. The p ima y ansmi e handles bu s y packe
a i als, whe eas seconda y use s ope a e unde sa u a ed a ic
condi ions. To manage in e e ence and ene gy, wo dis inc zones
a e in oduced: an ene gy ha es ing zone a ound he p ima y
ansmi e and a gua d zone a ound he p ima y ecei e ,
wi hin which seconda y ansmissions a e p ohibi ed. Seconda y
use s access he channel p obabilis ically, wi h access decisions
depending on hei cu en ba e y s a e (cha ged o emp y) and
hei loca ion ela i e o he gua d zone. Ou objec i e is o
analyze he p ima y use ’s AoI pe o mance unde h ee dis inc
packe managemen policies.
Index Te ms — Age o In o ma ion, Mul iple Access, Ene gy
Ha es ing.
I. INTRODUCTION
Age o In o ma ion (AoI) has eme ged as a c i ical pe o -
mance me ic in sha ed access ne wo k en i onmen s, quan-
i ying he imeliness ( eshness) o he ecei ed da a [1].
In oduced in ounda ional s udies [2]–[4], AoI measu es he
ime elapsed om he ime a imes amped s a us upda e is
gene a ed a i s sou ce un il i is success ully ecei ed and
decoded a i s des ina ion. Main aining a low a e age AoI
ensu es ha he deli e ed in o ma ion emains up- o-da e and
ele an . The concep has since b oadened o inco po a e
upda e cos s, in o ma ion- alue me ics, and non-linea AoI
e alua ions [5]–[8]. The impo ance o p ese ing in o ma ion
eshness, as cap u ed by AoI, spans a wide ange o appli-
ca ion domains, including unmanned ae ial ehicle (UAV)-
based In e ne o Things (IoT) ne wo ks [9]–[11], mul icas
and ad hoc wi eless communica ions [12]–[17], wi eless
senso ne wo ks o en i onmen al and heal hca e moni o ing,
da a wa ehousing sys ems [18]–[24], and web caching ech-
nologies [25]–[28]. Ene gy Ha es ing (EH) enables de ices,
such as senso s, o echa ge hei ba e ies by cap u ing
elec omagne ic ene gy, making EH pa icula ly sui able o
deploymen in en i onmen s whe e ba e y eplacemen o
a wi ed powe supply is imp ac ical. Ano he ele an use
case o EH a ises in de ice- o-de ice (D2D) communica ion
sys ems wi hin cellula ne wo ks, whe e D2D use s ha es
ene gy om cellula ansmissions. Pe o mance e alua ion
and op imiza ion o such sys ems ha e been s udied in [29],
[30], whe e seconda y nodes access he medium unde speci ic
cons ain s and p ima y nodes a e ypically assumed o ope a e
unde sa u a ed a ic. These wo ks analyze he join impac o
andom packe a i als a p ima y nodes and oppo unis ic EH-
enabled channel access by seconda y use s, wi h a e age delay
as a key pe o mance me ic [31]. Addi ional s udies, such as
[32]–[35], apply queueing heo y o in es iga e EH-enabled
ne wo ks and quan i y how ene gy ha es ing in luences he
s abili y egion in small-scale sys ems.
II. SYSTEM MODEL
A p ima y ansmi e (PT) is loca ed a he cen e o a
ci cula , omnidi ec ional co e age a ea modeled as a disk C
o adius R. The associa ed p ima y ecei e (PR) is ixed
a (0, y0), a dis ance dp om he PT. Seconda y ansmi -
e s (STs) a e spa ially dis ibu ed o e Cacco ding o a
homogeneous Poisson poin p ocess (PPP) Φswi h in ensi y
λs. Each ST is pai ed wi h a designa ed seconda y ecei e
(SR) placed a a ixed dis ance dsin a uni o mly andom
di ec ion om i s co esponding ST. Each ST is equipped wi h
a adio equency (RF) ene gy ha es ing module ha con e s
ambien RF ene gy in o di ec cu en (DC) powe o ba e y
s o age. A key no ion in EH sys ems is he ene gy ha es ing
zone: a disk o adius eh cen e ed a an RF sou ce wi hin
which he ecei ed powe exceeds he minimum con e sion
h eshold. Any ST loca ed inside his zone can accumula e
su icien ene gy wi hin a single ime slo [36]–[38].
In each ime slo , he subse o STs ha can be ully cha ged
by he p ima y ansmi e ’s (PT’s) RF ansmission is gi en
by:
S= Φs∩ Ceh(0, eh),(1)
whe e C(a, b)deno es a disk cen e ed a poin awi h adius
b. In his con ex , Ceh(0, eh) ep esen s he ene gy ha es ing
zone cen e ed a he PT, while Φsdeno es he spa ial dis i-
bu ion o STs, modeled as a homogeneous PPP wi h in ensi y
λs.
We assume ha he p ima y ansmi e ope a es a powe
le el Pp, while each seconda y ansmi e ansmi s a a
lowe powe le el Ps, whe e Ps≪Pp. This powe dispa i y
e lec s he ex ended ansmission ange o he PT and he
limi ed communica ion dis ances o he STs. Al hough he STs
ope a e a educed indi idual powe le els, hei simul aneous
a Xi :2510.11198 1 [cs.NI] 13 Oc 2025
ansmissions can p oduce subs an ial cumula i e in e e ence
a he PR, he eby deg ading he pe o mance o he PT–PR
link. To mi iga e his in e e ence, a p o ec ion zone is de ined
a ound he PR, p ohibi ing ST ansmissions wi hin a disk o
adius gz cen e ed a he PR’s loca ion (0, y0)[39].
Inco po a ing his in e e ence-awa e cons ain , he se o
STs eligible o ansmi in a gi en ime slo mus sa is y wo
condi ions:
1) The ST mus eside wi hin he EH zone and ha e
success ully ha es ed su icien ene gy, i.e., i mus
belong o he se S;
2) The ST mus lie ou side he gua d zone su ounding he
PR.
Acco dingly, he se o ac i e STs in a gi en ime slo is
de ined as:
SA=S (Φs∩ Cgz(y0, gz)) ,(2)
whe e Cgz(y0, gz)deno es he gua d zone cen e ed a he PR’s
loca ion (0, y0)wi h adius gz.
Fo he ne wo k con igu a ion unde conside a ion, we de-
ine he ollowing p obabili ies ela ed o he beha io and
s a us o STs:
•p : he p obabili y ha an ST ansmi s in a gi en ime
slo ,
•pch: he p obabili y ha an ST’s ba e y is ully cha ged
a he beginning o a ime slo ,
•peh: he p obabili y ha an ST is loca ed wi hin he EH
zone o he PT,
•pgz: he p obabili y ha an ST is loca ed wi hin he
gua d zone o he PR and is he e o e p ohibi ed om
ansmi ing.
We conside a dynamic en i onmen in which STs a e
mobile ac oss ime slo s. Consequen ly, he p obabili y ha
an ST lies ou side he gua d zone and he p obabili y ha
i was ully cha ged in he p e ious ime slo a e ea ed as
independen e en s. Acco ding o he adop ed andom access
p o ocol, an eligible ST a emp s ansmission wi h p obabili y
psin any gi en ime slo .
Thus, he o e all p obabili y ha an ST becomes ac i e, i.e.,
ansmi s in a gi en ime slo , is gi en by:
p =pch ·(1 −pgz)·ps.(3)
This exp ession ep esen s he p opo ion o ac i e ST
ansmissions in each ne wo k con igu a ion, cap u ing he
combined e ec s o ene gy a ailabili y, he spa ial es ic ions
imposed by he gua d zone, and he p obabilis ic na u e o
access con ol.
The abili y o he PR o success ully decode signals om
he PT The abili y o he PR o success ully decode signals
om he PT depends on he signal- o-in e e ence-plus-noise
a io (SINR) measu ed a he PR. This SINR accoun s o
he cumula i e in e e ence gene a ed by he STs ha a e
simul aneously ac i e du ing he same ime slo . I is exp essed
as:
SINRp=Pp|hP|2d−α
p
σ2+Pi∈ S Ps|hi|2d−α
i
,(4)
whe e S ⊂ SAdeno es he se o STs ansmi ing in he
cu en ime slo . The e m |hP|2 ep esen s he small-scale
Rayleigh ading powe gain on he PT–PR link, and dpis he
dis ance be ween he PT and PR. Simila ly, |hi|2deno es he
ading powe gain be ween he i- h ST and he PR, wi h dias
he co esponding dis ance. The pa ame e α > 2is he pa h
loss exponen , and σ2deno es he backg ound he mal noise
powe , modeled as addi i e whi e Gaussian noise (AWGN).
This exp ession cap u es he impac o concu en ST ans-
missions on he SINR a he PR and, consequen ly, on he
eliabili y o he p ima y communica ion link.
III. PERFORMANCE ANALYSIS
Fo seconda y use s, a key me ic e lec ing o e all ne wo k
pe o mance is he a e age seconda y h oughpu . To e alu-
a e his me ic, wo undamen al p obabili ies mus i s be
de ined: he ac i e p obabili y, which deno es he likelihood
ha a seconda y use is eady o ansmi in a gi en ime slo ,
and he success ul ansmission p obabili y, which ep esen s
he likelihood ha a seconda y use ’s signal is success ully
decoded a i s in ended ecei e .
Figu e 1: Two-s a e Disc e e-Time Ma ko Chain (DTMC) modeling
he ST ba e y s a e e olu ion.
In ou analysis, we employ a single-slo cha ging and
discha ging model, whe e each ST’s ba e y is ep esen ed as a
wo-s a e disc e e- ime Ma ko chain (DTMC), as depic ed in
Figu e 1. In his model, s a e Fco esponds o a ully cha ged
ba e y, allowing he ST o ansmi , whe eas s a e Edeno es
an emp y ba e y, p e en ing ansmission.
Using his DTMC app oach, we compu e he s a iona y
p obabili y ha a seconda y use has su icien ene gy o
ansmi , which cons i u es a key componen in e alua ing he
o e all a e age seconda y h oughpu .
The p obabili y ha an ST is ully cha ged and hus able o
ansmi in he nex ime slo is gi en by:
pch =peh
peh +ps−pgzps
.(5)
The p obabili ies ha an ST is loca ed wi hin he EH zone
and wi hin he PR gua d zone a e de i ed based on geome ic
conside a ions and he law o cosines, espec i ely:
peh = 2
eh
R2,(6)
pgz =






2
gz
R2, gz ≤R−dp,
ϕ 2
gz
πR2+φ
π−dp
πR sin φ, gz > R −dp,
(7)
whe e ϕ= a ccosd2
p+ 2
gz −R2
2dp gz and φ=
a ccosR2+d2
p− 2
gz
2dpR.
In ou analysis, we de ine he seconda y use h oughpu
as he a e age numbe o success ul packe ansmissions pe
ime slo pe uni a ea, measu ed in packe s/slo /m2. Le Ts
deno e he a e age seconda y h oughpu , exp essed as
Ts=λsp psx =λs
peh
peh +ps−pgzps
(1 −pgz)pspsx,(8)
whe e psx deno es he p obabili y o a success ul ansmission
om an ST o i s designa ed SR. To e alua e psx, we i s
analyze he SINR, assuming ha a ansmission is deemed
success ul i he ecei ed SINR exceeds a p ede ined h eshold
θ. Conside ing ha he SR o he i- h seconda y ansmi -
e – ecei e pai is loca ed a he o igin, we can de i e he
ollowing exp ession:
SINRi=Ps|hi,i|2ds−α
σ2+Pj∈S {i}Ps|hj,i|2d−α
j,i +Pp|hp,i|2d−α
p,i
,
(9)
whe e S deno es he se o ac i e STs, |hj,i|2 ep esen s he
small-scale Rayleigh ading gain om ansmi e j o ecei e
i, assumed o ha e uni mean, and dj,i is he dis ance be ween
ansmi e jand he SR i. As in equa ion (4), he e m d−α
models he dis ance-dependen pa h-loss a enua ion, whe e
α > 2is he pa h-loss exponen , and σ2deno es he he mal
noise powe .
Due o he p esence o a gua d zone a ound he PR, he
in e e ence ield expe ienced by SRs becomes non-iso opic.
Consequen ly, STs loca ed nea he bounda y o he gua d zone
end o expe ience highe ansmission success p obabili ies
because o he educed in e e ence in ha egion.
psx =P[SINRi> θ]
(a)
=exp(−θdα
sIn)Edp,i 

1
1 + θPpda
s
Psda
p,i

exp −θσ2dα
s
Ps
(b)
≃ LIn(da
sIn)
exp −θσ2dα
s
Ps
1 + d2
s
E[dp,i]2θPp
Ps
2
a
(c)
≈exp −πλspchpsd2
sθ2
aexp −θσ2da
s
Ps
1 + d2
s
E[dp,i]2θPp
Ps
2
a
,
(10)
whe e In=P
j∈S
|hj,i|2d−α
j,i deno es he no malized agg ega e
in e e ence a he SR, and LIn(s) = Ee−sIn ep esen s he
Laplace ans o m o he in e e ence. In equa ion (10), s ep (a)
ollows om he exponen ial p obabili y densi y unc ion o
|hi,i|2; s ep (b) esul s om applying he Laplace ans o m
o he in e e ence and he app oxima ion in oduced in [40];
and s ep (c) app oxima es psx using he p obabili y gene a ing
unc ional (p.g. l) o a PPP. Gi en he andom placemen o
STs ac oss he ne wo k co e age a ea, each SR is uni o mly
dis ibu ed wi hin he disk C(0, R). The expec ed dis ance
be ween he PT and an SR is hus gi en by E[dp,i] = 2R
3.
In his pa o he pe o mance e alua ion, we analyze he
p ima y use ’s pe o mance in e ms o i s a e age AoI unde
h ee dis inc packe managemen s a egies. These s a egies
de ine how s a us upda e packe s a e handled a he PT, pa ic-
ula ly unde a ying packe a i al and se ice p obabili ies.
The conside ed scena ios a e as ollows:
•Fi s -Come Fi s -Se ed (FCFS): A s anda d queueing
model wi hou packe managemen , whe e packe s a e
se ed s ic ly in he o de o hei a i al.
•Queue wi h Replacemen (QR): A single-packe bu e in
which any newly a i ed packe eplaces he one cu en ly
s o ed (i any), ensu ing ha only he mos ecen s a us
upda e is e ained o ansmission.
•Gene a e-a -Will (GW): A model in which s a us upda e
packe s a e gene a ed only when he channel is imme-
dia ely a ailable o ansmission; i he ansmission
a emp ails, he packe is disca ded.
In he conside ed se up, µpdeno es he success p obabili y
o a p ima y packe ansmission wi hin a gi en ime slo .
Using he SINR de ini ion in equa ion (4) in Sec ion II, and o
a gi en h eshold θ, we p oceed analogously o he analysis in
Sec ion III-A o he seconda y use success p obabili y. This
yields he PT’s se ice a e as ollows:
µp=P[SINRp> θ]
=exp −θdα
pPs
Pp
Isexp −θσ2dα
p
Pp!
=LIsθdα
pPs
Ppexp −θσ2dα
p
Pp!
=L∗
Iθdα
pPs
Pp
, λsp , gzexp −θσ2dα
p
Pp!,
(11)
whe e Is=P
i∈S
Ps|hi|2d−α
ideno es he agg ega e in e e -
ence om seconda y use s a he PR, and L∗
I( , λa, )is
de ined as in [41] and ep esen s he Laplace ans o m o
he in e e ence, gi en by
L∗
I( , λa, ) = exp−2πλaZ∞
υ−α
1+ υ−αυ dυ.
Figu e 2: DTMC ep esen ing he e olu ion o he Geo/Geo/1 queue
a he PT.
1) FCFS - Geo/Geo/1 Queue:The queue a he PT is
modeled as a DTMC, illus a ed in Figu e 2, whe e λdeno es
he packe a i al a e and µp ep esen s he se ice a e. Each
s a e in he DTMC co esponds o he numbe o packe s in
he PT’s queue.
Lemma 3.1: F om he DTMC p esen ed in Figu e 2, he
s eady-s a e p obabili ies a e gi en by
πn=ρn−1π1, n ≥1,(12)
π0=µp(1 −λ)
λπ1,(13)
whe e ρ=λ(1−µp)
µp(1−λ),π1=λ(1−ρ)
µp, =λ(1 −µp),s=µp(1 −
λ), and µpis gi en in equa ion (11).
The queue a he PT is s able as long as he a i al a e
sa is ies λ < µp[42]. Following he analysis p esen ed in [43],
he a e age AoI ∆pis exp essed as
∆p=1
λ+1−λ
µp−λ−λ
µ2
p
+λ
µp
.(14)
2) Queue wi h Replacemen (QR):The second packe
managemen s a egy conside ed is he queue wi h eplace-
men , in which newly a i ing packe s eplace any exis ing
packe wai ing in he queue, p o ided i has no ye been
ansmi ed. This model assumes a single-packe bu e a he
PT, allowing a mos one packe o be held in addi ion o he
one cu en ly being se ed. I a new s a us upda e a i es while
he bu e is occupied and he queued packe has no ye been
ansmi ed, he exis ing packe is disca ded and eplaced by
he new a i al. The queue wi h a eplacemen policy a he
PT can be ep esen ed using a h ee-s a e DTMC [44]. The
s a es a e de ined as ollows:
1) The sys em is emp y (no packe s a e p esen a he PT).
2) One packe is cu en ly in se ice.
3) One packe is in se ice and ano he is wai ing in he
queue ( he bu e is ull).
Al hough his policy does no al e he maximum numbe
o packe s ha can eside in he sys em a any gi en ime,
i allows newly a i ing upda es o o e w i e olde ones s ill
wai ing in he queue. Consequen ly, i p io i izes in o ma ion
eshness o e deli e y eliabili y.
Lemma 3.2: F om he DTMC p esen ed in Figu e 3, he
s eady-s a e p obabili ies a e gi en by
πn=λn(1 −µp)n−1
µn
p(1 −λ)nπ0, n ∈ {1,2},(15)
π0=λ−µp
λρ2−µp
,(16)
whe e ρ=λ(1−µp)
µp(1−λ), =λ(1 −µp),s=µp(1 −λ), and µpis
gi en in equa ion (11).
We e alua e he p obabili y ha a packe in he PT’s queue
is d opped based on he s eady-s a e p obabili ies de i ed om
he DTMC in Figu e 3, using Lemma 2, as ollows:
pd=π1λ(1 −µp)+π2(1 −s)
=λ2(1 −µp)
µ2
p(1 −λ)1 + λ
µp(1 −λ)+λ2(1 −µp)
µ2
p(1 −λ)−1
.(17)
Figu e 3: DTMC modeling he queue e olu ion a he PT unde he
eplacemen policy. He e, =λ(1 −µp)and s=µp(1 −λ).
The e ec i e a i al a e, accoun ing o he packe d op
p obabili y, is exp essed as ollows:
λe=λ(1 −pd)
=λ−λ3(1 −µp)
λ2(1 −µp)+λ(1 −µp)µp+µ2
p
.(18)
To e alua e he a e age AoI o he PT unde he queue-wi h-
eplacemen policy, we ollow he analysis in [43]:
∆p=1
δ λµpϵδ
2λµpϵ+
λλ(3µp−2) −2µp+ 1
λ2(µp−1)2+λµp(1 −2µp)+µ2
p
+λ3(µp−2)(µp−1) + λ2(µp−2)(µp−1)µp
2λ2µ2
pϵ
+λµ2
p(2 −3µp)+2µ3
p
2λ2µ2
pϵ+1−λ
λµp
+2λ+ 1
ϵ
−λ+ 1
ϵ2+1
µ2
p!
(19)
whe e δ=λ2(1−µp)+λ(1−µp)µp+µ2
pand ϵ=λ+µp−λµp.
3) Gene a e-a -Will policy (GW):In his subsec ion, we
de i e he exp ession o he a e age AoI unde he gene a e-
a -will packe managemen s a egy. In his scheme, he PT
gene a es and ansmi s a s a us upda e only when he channel
is a ailable. I he channel is busy, he upda e is disca ded,
and a new one is gene a ed in he nex ime slo acco ding o
he sampling p ocess.
The s a us upda e gene a ion ollows a Be noulli p ocess
wi h success p obabili y λ, which also ep esen s he sampling
a e o he PT. This a e co esponds o he equency a which
upda es a e gene a ed unde his policy.
The a e age AoI a he p ima y ecei e unde he gene a e-
a -will policy is exp essed as ollows:
∆p=1
µpq,(20)
whe e µpdeno es he se ice a e (i.e., he p obabili y o a
success ul ansmission in a ime slo ), and q ep esen s he
e ec i e packe gene a ion (o sampling) a e a he PT.
This model cap u es he ade-o be ween agg essi e sam-
pling, which aims o main ain a low AoI, and he inc eased
isk o packe d ops due o channel una ailabili y.
IV. NUMERICAL ANALYSIS
In his sec ion, we e alua e he AoI pe o mance o all h ee
packe managemen s a egies by a ying mul iple sys em
pa ame e s. Figu e 4 compa es he AoI unde wo di e en
seconda y use densi y scena ios (λs= 10−3and 2×10−3
poin s/m2) as a unc ion o he access p obabili y, wi h ixed
alues o eh =80m and gz = 120 m. Among he h ee
s a egies, he GW policy consis en ly achie es he lowes AoI.
When compa ing FCFS and QR, he la e achie es be e
AoI pe o mance as he access p obabili y pso he seconda y
use s inc eases.
Figu e 4: A e age AoI compa ison among FCFS, QR, and GW as a
unc ion o he access p obabili y ps.
Figu e 5 p esen s 3D su ace plo s o he a e age AoI
pe o mance o he h ee queueing s a egies, e alua ed as
unc ions o (ps, eh) and (ps, gz), espec i ely.
In he FCFS scena io shown in Figu e 5(a), he a e age
AoI inc eases sha ply as he ene gy ha es ing zone adius ex-
pands. This beha io aligns wi h p e ious obse a ions, whe e
he AoI ises wi h highe seconda y use access p obabili ies.
The inc ease can be a ibu ed o he ac ha a la ge numbe
o seconda y use s no only ha e su icien ene gy o be
ac i e bu a e also allowed o ansmi , he eby in ensi ying
in e e ence and causing longe se ice delays o he p ima y
use . In he QR scena io (Figu e 5(b)), he inc ease in a e age
AoI is mo e mode a e compa ed o he FCFS case, owing
o he eplacemen mechanism ha helps main ain eshe
upda es in he queue. Finally, he GW policy (Figu e 5(c))
consis en ly ou pe o ms he o he wo s a egies, ollowing
a end simila o ha o he QR policy bu yielding o e all
lowe a e age AoI alues.
Finally, Figu e 5 illus a es he impac o he gua d zone
adius gz and he seconda y use access p obabili y pson
he a e age AoI. In he FCFS scena io (Figu e 5(a)), he
a e age AoI dec eases apidly as he gua d zone expands.
This beha io occu s because a la ge gua d zone p e en s
mo e seconda y use s om ansmi ing, he eby educing
in e e ence a he p ima y ecei e and consequen ly lowe ing
he AoI. Fo bo h he QR (Figu e 5(b)) and GW (Figu e 5(c))
policies, a simila dec easing end in a e age AoI is obse ed
wi h inc easing gz, highligh ing he consis en in e e ence-
mi iga ion bene i s ac oss all h ee queueing s a egies.
(a) Fi s Come Fi s Se e queu-
ing policy.
(b) Queue wi h Replacemen
queuing policy.
(c) Gene a e-a -Will policy.
(d) Fi s Come Fi s Se e queu-
ing policy.
(e) Queue wi h Replacemen
queuing policy.
( ) Gene a e-a -Will policy.
Figu e 5: A e age AoI beha io o he conside ed policies.
Le : as a unc ion o he access p obabili y psand he ene gy
ha es ing zone adius eh. Righ : as a unc ion o he access
p obabili y psand he gua d zone adius gz.
V. CONCLUSIONS
The esul s highligh he c i ical in luence o he ene gy ha -
es ing zone adius ( eh) and he gua d zone adius a ound he
p ima y ecei e ( gz) on o e all sys em pe o mance. While
inc easing eh ini ially allows mo e STs o ha es ene gy and
become ac i e, excessi e expansion ul ima ely deg ades pe -
o mance due o highe in e e ence le els. Simila ly, enla ging
he gua d zone gz e ec i ely mi iga es in e e ence a he PR
and can signi ican ly enhance AoI pe o mance; howe e , his
imp o emen comes a he expense o educed ansmission
oppo uni ies o seconda y use s, he eby dec easing hei
h oughpu .
Fu u e wo k could in es iga e ad anced EH models, adap-
i e access con ol, and ex ensions o he e ogeneous ne wo k
se ings. Fu he mo e, in eg a ing machine lea ning echniques
o dynamically adjus zone adii and access policies ep esen s
a p omising di ec ion o op imizing he ade-o be ween
h oughpu and in o ma ion eshness in cogni i e ne wo ks.

ACKNOWLEDGMENT
This wo k has been suppo ed in pa by he Swedish
Resea ch Council (VR), ELLIIT, he Eu opean Union (6G-
LEADER) unde G an 101192080; he Eu opean Union’s
Ho izon Eu ope Resea ch and Inno a ion P og amme un-
de he Ma ie Skłodowska-Cu ie G an unde Ag eemen
101131481 (SOVEREIGN).
REFERENCES
[1] A. Kos a, N. Pappas, and V. Angelakis, “Age o in o ma ion: A new
concep , me ic, and ool,” Founda ions and T ends® in Ne wo king,
ol. 12, no. 3, pp. 162–259, 2017.
[2] E. Al man, R. El-Azouzi, D. S. Menansche, and Y. Xu, “Fo e e young:
Aging con ol o sma phones in hyb id ne wo ks,” in 20 h ACM
Mobihoc, 2019.
[3] S. Kaul, M. G u ese , V. Rai, and J. Kenney, “Minimizing age o
in o ma ion in ehicula ne wo ks,” in 8 h Annual IEEE Commun.
Socie y Con . on Senso , Mesh and Ad Hoc Commun. and Ne wo ks,
June 2011, pp. 350–358.
[4] S. Kaul, R. Ya es, and M. G u ese , “Real- ime s a us: How o en should
one upda e?” in IEEE INFOCOM, Ma ch 2012, pp. 2731–2735.
[5] A. Kos a, N. Pappas, A. Eph emides, and V. Angelakis, “Age and alue
o in o ma ion: Non-linea age case,” in IEEE ISIT, June 2011.
[6] Y. Sun and B. Cy , “Sampling o da a eshness op imiza ion: Non-
linea age unc ions,” Jou nal o Commun. and Ne w., ol. 21, no. 3, p.
204–219, 2019.
[7] Y. Sun, I. Kado a, R. Talak, and E. Modiano, “Age o in o ma ion: A
new me ic o in o ma ion eshness,” Mo gan & Claypool Publishe s,
2019.
[8] R. D. Ya es, Y. Sun, D. R. B own, S. K. Kaul, E. Modiano, and
S. Ulukus, “Age o in o ma ion: An in oduc ion and su ey,” IEEE
Jou nal on Sel. A eas in Commun., ol. 39, no. 5, pp. 1183–1210, May
2021.
[9] M. A. Abd-Elmagid and H. S. Dhillon, “A e age peak age-o -
in o ma ion minimiza ion in UAV-assis ed IoT ne wo ks,” IEEE T ans.
on Veh. Tech., ol. 68, no. 2, pp. 2003–2008, Feb. 2019.
[10] M. A. Abd-Elmagid, N. Pappas, and H. S. Dhillon, “On he ole o
age-o -in o ma ion in in e ne o hings,” IEEE Commun. Mag., ol. 57,
no. 12, p. 72–77, 2019.
[11] P. D. Manka , Z. Chen, M. A. Abd-Elmagid, N. Pappas, and H. S.
Dhillon, “Th oughpu and age o in o ma ion in a cellula -based io
ne wo k,” IEEE T ans. on Wi eless Commun., ol. 20, no. 12, Dec. 2021.
[12] I. Kado a, E. Uysal-Biyikoglu, R. Singh, and E. Modiano, “Minimizing
he age o in o ma ion in b oadcas wi eless ne wo ks,” in P oc. Alle on,
Sep embe 2016, pp. 844–851.
[13] Y. Hsu, E. Modiano, and L. Duan, “Age o in o ma ion: Design and
analysis o op imal scheduling algo i hms,” in IEEE ISIT, June 2017.
[14] I. Kado a, A. Sinha, E. Uysal-Biyikoglu, R. Singh, and E. Modiano,
“Scheduling policies o minimizing age o ino ma ion in b oadcas
wi eless ne wo ks,” IEEE/ACM T ans. Ne w., ol. 26, no. 6, pp. 2637–
2650, Oc obe 2018.
[15] B. Buyuka es, A. Soyal, and S. Ulukus, “Age o in o ma ion in mul icas
ne wo ks wi h mul iple upda e s eams,” in 53 d Asiloma , 2019.
[16] Q. He, G. D`
an, and V. Fodo , “Age o in o ma ion: Design and analysis
o op imal scheduling algo i hms,” in P oc. IEEE INFOCOM, Ap il
2018, pp. 547–552.
[17] M. Salimnejad and N. Pappas, “On he age o in o ma ion in a wo-
use mul iple access se up,” En opy Special Issue ”Age o In o ma ion:
Concep , Me ic and Tool o Ne wo k Con ol”, Ap . 2022.
[18] X. Wu, J. Yang, and J. Wu, “Op imal s a us upda e o age o in o ma ion
minimiza ion wi h an ene gy ha es ing sou ce,” IEEE T ans. G een
Commun., ol. 2, no. 1, pp. 193–204, Ma ch 2018.
[19] A. A a a and S. Ulukus, “Timely upda es in ene gy ha es ing wo-
hop ne wo ks: O line and online policies,” IEEE T ans. on Wi eless
Commun., ol. 18, no. 8, pp. 4017–4030, Aug. 2019.
[20] A. A a a, J. Yang, S. Ulukus, and H. V. Poo , “Age-minimal ansmission
o ene gy ha es ing senso s wi h ini e ba e ies: Online policies,” IEEE
T ans. on In o m. Theo y, ol. 66, no. 1, pp. 534–556, Jan. 2020.
[21] S. Na h, J. Wu, and J. Yang, “Delay and ene gy e iciency adeo o
in o ma ion pushing sys em,” IEEE T ans. G een Commun., ol. 2, no. 4,
pp. 1027–1040, June 2018.
[22] Z. Chen, N. Pappas, E. Bj¨
o nson, and E. G. La sson, “Age o in o ma-
ion in mul iple access channel wi h he e ogeneous a ic and ene gy
ha es ing node,” in P oc. IEEE INFOCOM, Ap il 2019, pp. 1–6.
[23] Y. Gu, H. Chen, Y. Zhou, Y. Li, and B. Vuce ic, “Timely s a us upda e
in in e ne o hings moni o ing sys em: An age-ene gy adeo ,” IEEE
In e ne o Things Jou nal, ol. 6, no. 3, pp. 5324–5335, June 2019.
[24] E. Foun oulakis, T. Cha alambous, N. Nomikos, A. Eph emides, and
N. Pappas, “In o ma ion eshness and packe d op a e in e play in a
wo-use mul i-access channel,” IEEE/KICS Jou nal o Commun. and
Ne w. (JCN), ol. 24, no. 3, pp. 357–364, June 2022.
[25] H. Yu, L. Be slau, and S. Shenke , “A scalable web cache consis ency
a chi ec u e,” ACM SIGCOMM Compu e Commun. Re iew, ol. 29,
no. 4, pp. 163–174, Augus 1999.
[26] C. Kam, S. Kompella, G. D. Nguyen, J. E. Wiesel hie , and
A. Eph emides, “In o ma ion eshness and popula i y in mobile
caching,” in P oc. IEEE ISIT, June 2017, pp. 136–140.
[27] R. D. Ya es, P. Cibla , A. Yene , and M. Wigge , “Age-op imal con-
s ained cache upda ing,” in P oc. IEEE ISIT, June 2017, pp. 141–145.
[28] J. Zhong, R. D. Ya es, and E. Soljanin, “Age-op imal cons ained cache
upda ing,” in P oc. IEEE ISIT, June 2018, pp. 436–441.
[29] H. H. Yang, J. Lee, and T. Q. S. Quek, “He e ogeneous cellula ne wo k
wi h ene gy ha es ing-based D2D communica ion,” IEEE T ans. on
Wi eless Commun., ol. 15, no. 2, pp. 1406–1419, Feb ua y 2016.
[30] A. H. Sak and E. Hossain, “Cogni i e and ene gy ha es ing-based
D2D communica ion in cellula ne wo ks: S ochas ic geome y modeling
analysis,” IEEE T ans. on Commun., ol. 63, no. 5, pp. 1867–1880, May
2015.
[31] Z. Chen, N. Pappas, and M. Koun ou is, “Ene gy ha es ing in delay-
awa e cogni i e sha ed access ne wo ks,” in IEEE ICC Wo kshops, May
2017.
[32] N. Pappas, M. Koun ou is, J. Jeon, A. Eph emides, and A. T agani is,
“Ne wo k le el coope a ion in ene gy ha es ing wi eless ne wo ks,” in
IEEE GLOBALSIP, Decembe 2013.
[33] J. Jeon and A. Eph emides, “On he s abili y o andom mul iple access
wi h s ochas ic ene gy ha es ing,” IEEE Jou nal on Sel. A eas in
Commun., ol. 33, no. 3, pp. 571–584, Ma ch 2015.
[34] N. Pappas, M. Koun ou is, J. Jeon, A. Eph emides, and A. T agani is,
“E ec o ene gy ha es ing on s able h oughpu in coope a i e elay
sys ems,” Jou nal o Commun. and Ne w., ol. 18, no. 2, pp. 261–269,
Ap il 2016.
[35] C. Xu, X. Zhang, H. Yang, X. Wang, N. Pappas, D. Niya o, and T. Quek,
“Op imal s a us upda es o minimizing age o co ela ed in o ma ion in
io ne wo ks wi h ene gy ha es ing senso s,” IEEE T ans. on Mobile
Compu ing, ol. 23, no. 6, pp. 6848 – 6864, June 2024.
[36] K. Huang, “Spa ial h oughpu o mobile ad hoc ne wo ks powe ed by
ene gy ha es ing,” IEEE T ans. on In o m. Theo y, ol. 59, no. 11, pp.
7597–7612, No embe 2013.
[37] S. Lee, R. Zhang, and K. Huang, “Oppo unis ic wi eless ene gy ha -
es ing in cogni i e adio ne wo ks,” IEEE T ans. on Wi eless Commun.,
ol. 12, no. 9, p. 4788–4799, Sep embe 2013.
[38] H. H. Yang, J. Lee, and T. Q. S. Quek, “He e ogeneous cellula ne wo k
wi h ene gy ha es ing-based D2D communica ion,” IEEE T ans. on
Wi eless Commun., ol. 15, no. 2, p. 1406–1419, Feb. 2016.
[39] A. Hasan and J. G. And ews, “The gua d zone in wi eless ad hoc
ne wo ks,” IEEE T ans. on Wi eless Commun., ol. 6, no. 3, p. 897–906,
Ma . 2007.
[40] N. Lee, X. Lin, J. G. And ews, and R. W. Hea h, “Powe con ol o
D2D unde laid cellula ne wo ks: Modeling, algo i hms, and analysis,”
IEEE Jou nal on Sel. A eas in Commun., ol. 33, no. 1, p. 1–13, Jan.
2015.
[41] Z. Chen and M. Koun ou is, “Gua d zone based D2D unde laid cellula
ne wo ks wi h wo- ie dependence,” IEEE ICC Wo kshops, June 2015.
[42] R. Loynes, “The s abili y o a queue wi h non-independen in e -a i al
and se ice imes,” in P oc. Camb. Philos. Soc, ol. 58, no. 3, 1962, p.
497–520.
[43] A. Kos a, N. Pappas, A. Eph emides, and V. Angelakis, “Age o in o -
ma ion pe o mance o mul iaccess s a egies wi h packe managemen ,”
Jou nal o Commun. and Ne w., ol. 21, no. 3, pp. 244–255, June 2019.
[44] M. Cos a, M. Cod eanu, and A. Eph emides, “On he age o in o ma ion
in s a us upda e sys ems wi h packe managemen ,” IEEE T ans. on
In o m. Theo y, ol. 62, no. 4, pp. 1897–1910, June 2016.