This article has been accepted for publication in IEEE Communications Letters.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2024.3475569

JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 1

A Generalized STAR-RIS-Empowered Ambient Backscatter

Short-Packet Communication Systems With Partial NOMA
Tien-Hoa Nguyen, Member, IEEE, and Thai-Hoc Vu Graduate Student Member, IEEE,

Abstract—This paper proposes a simultaneously transmitting radio systems [11] and, when combined with reconfigurable
and reflecting reconfigurable intelligent surface (STAR-RIS)- intelligent surfaces (RISs), creating diverse signal paths and
empowered ambient backscatter short-packet paradigm with par- a robust, energy-efficient communication environment [12].
tial non-orthogonal multiple access (p–NOMA). In this paradigm,
a multi-antenna base station communicates with users using Thus, exploring STAR-RISs with ambient backscatter com-
finite blocklength schemes to achieve low latency transmission munication is a natural and remains unexplored.
while flexibly exploiting the spectrum utilization via p–NOMA. To bridge the mentioned research gap above, this work
Considering Nakagami–m fading channels, discrete phase-shift analyzes the performance of active, passive, and hybrid STAR-
alignment, and imperfect successive interference cancellation, RIS models in ambient backscatter SPC systems with a focus
we provide a generalized information-theoretic framework that
characterizes passive, active, and hybrid STAR-RIS types, to on the partial NOMA (p-NOMA) technique to achieve higher
measure the block-error rate (BLER) and goodput. To gain throughput like NOMA and low inter-user interference as
useful insights into system designs, an upper-bound BLER at high OMA [13]. To the best of the authors’ knowledge, this ap-
transmit power has been derived. Numerical results demonstrate proach was studied with passive RIS [14] but failed to consider
the BLER superiority of p–NOMA over its orthogonal multiple STAR-RIS scenarios in general and SPCs in particular. The
access (OMA) and NOMA counterparts, as well as the respective
twofold and fourfold enhancements in terms of goodput. novelty of this work includes: 1) A generalized STAR-RIS-
empowered ambient backscatter SPC paradigm with p-NOMA
Index Terms—Hybrid active and passive simultaneously trans- is proposed, where multi-antenna sources communicate with
mitting and reflecting reconfigurable intelligent surface, short-
packet communication, partial non-orthogonal multiple access. two users in the light of discrete phase-shift alignment at
STAR-RISs, Nakagami–m fading channels, and imperfect
SIC; 2) A generalized information-theoretic framework has
been developed to efficiently measure the BLER and goodput
I. I NTRODUCTION
in types of passive, active, and hybrid STAR-RISs; 3) An
Simultaneously transmitting and reflecting reconfigurable asymptotic analysis of the average BLER and goodput in high
intelligent surfaces (STAR-RIS) have recently emerged as transmit power regime has been conducted to guide useful
a potential solution for achieving 360-degree coverage [1]. insights related to the diversity gain and the goodput floor;
Notably, integrating STAR-RISs into non-orthogonal multiple and 4) A deep analysis of the trade-off between p-NOMA and
access (NOMA) systems achieves better performance than conventional OMA and NOMA has been provided.
orthogonal multiple access (OMA) ones in terms of effective Notations: CN (0, σ 2 ) is the complex Gaussian with zero-
capacity [2], coverage region [3], and block-error rate (BLER) mean and variance σ 2 , E{·} denotes the expectation operator,
in short-packet communication (SPC) [4]–[6]. However, the FX (·) denotes the cumulative density function (CDF), fX (·)
double-fading effect significantly degrades STAR-RIS’s com- is the probability density function (PDF), Γ(·) and Γ(·, ·) are
munications quality. This prompts the emergence of active respectively the Gamma and upper incomplete Gamma func-
STAR-RIS capable of amplifying the incident signals [7], tions [15, eqs. (8.310.1) and (8.350.2)], H (z) is the Heaviside
·,· ... ;... ... ... ...
followed by early investigations of STAR-RIS-enabled NOMA step function, and H·,· [ ... |· ] and H...
... ;... [ ... | ... | ... |· ] are the
with imperfect successive-interference cancellation (SIC) [8]. respective Fox-H and bi-variate Fox-H functions, with their
However, employing a full amplifier function significantly details in [16, Appendixes A and B].
increases energy consumption and becomes ineffective when
users are close to the STAR-RISs. This underscores the need II. S YSTEM M ODEL
for hybrid active and passive models, which, to the best of the Fig. 1 shows SPC p-NOMA systems assisted by ambient
author’s knowledge, have not been studied in the literature. backscatter and STAR-RISs1 , including an L-antennas base
Besides, ambient backscatter communication is transform- station (BS), one tag (BD), one hybrid STAR-RIS (RIS), with
ing the Internet of Things (IoT) by enabling devices to commu- Ka -active and Kp -passive elements such that Ka + Kp = K,
nicate using existing radio-frequency signals without requiring and two single-antenna users located in RIS’s reflective zones
batteries [9], [10], making it a promising solution for symbiotic (i.e., RU) and refractive ones (i.e., TU). Direct links between
BS and UE ∈ {RU, TU} are blocked by obstacles. UE can only
Tien-Hoa Nguyen (Corresponding author) is with School of Electrical and
Electronic Engineering, Hanoi University of Science and Technology, Hanoi receive their messages from BS though the aid of BD and RIS
100000, Vietnam (e-mail: [email protected]). Thai-Hoc Vu is with and decode them by signal detection and SIC [9], [10].
Department of Electrical, Electronic, and Computer Engineering, University
of Ulsan, Ulsan 44610, Korea (e-mail: [email protected]). 1 The considered system can be realized in various scenarios, such as
This research is funded by Hanoi University of Science and Technology Internet-of-Vehicles networks, geological hazard/agricultural/bridge health
(HUST) under project number T2023-PC-032. monitoring, smart factories, logistics, and warehousing [12, Section VIII].

Authorized licensed use limited to: University of Ulsan. Downloaded on October 15,2024 at 13:07:29 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Communications Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2024.3475569

JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 2

Backscatter device (BD) Active where P is the transmit power of BS, ϖ ∈ (0, 1) is the
Passive reflection coefficient of BD to normalize c(t), ℘ ∈ {0, 1} refers
Hybrid STAR-RIS to RIS’s amplifier operation, w = hH BD /∥hBD ∥ is the beam-
Kp
forming weight, hBD ≜ [h1BD , . . . , hL 1
BD ], hUE ≜ [hUE , . . . , hUE ],
Kp e UE ≜ [h1 , . . . , hK
Base station hRIS ≜ [h1RIS , . . . , hRIS ], h UE UE ], nUE ∼
a

(BS) CN (0, N0 ) is the additive white Gaussian noise, where N0

Reflective user Transmission user
(RU) (TU) is the density noise power per bandwidth BW, and e eRIS is the
Ka × 1 thermal noise vector caused by Ka -active elements,
Fig. 1. Illustration of the proposed system. where each entry follows the complex Gaussian distribution
with zero-mean and variance σe2 , i.e., eeRIS ∼ CN (0, σe2 ).
From (3), the signal-to-interference-plus-noise ratio (SINR)
Suppose that the system operates in Nakagami-m block fad-
to detect the messages of xRU at UE ∈ {RU, TU} and xTU at
ing mediums, where the channel hX , with X ∈ {BD, RIS, UE},
TU over sSU can be, respectively, expressed as in [9], [10] as
of the respective link Y ∈ {BS → BD, BD → RIS, RIS →
xRU
UE} is static/quasi-static within each block. The n-th moment γ
eUE = αRU ϖP τBD τUE /[αTU ϖP τBD τUE + ℘e
τUE + βSU N0 ], (4)
of |hX | with the envelop mX and severity ΩX is given by γ xTU
eTU = αTU ϖP τBD τTU /[ζαRU ϖP τBD τTU + ℘e
τTU + βSU N0 ], (5)
(n) n/2
µ|hX | = E{|hX |n } = Γ(mX + n/2) (ΩX /mX ) /Γ(mX ). (1) K
where τBD ≜ ∥hBD ∥2 , τUE ≜ |hTUE ΦUEp hRIS + he T ΦKa h 2
UE UE RIS | ,
e
√ T
e ΦKa e
τeUE ≜ |h 2
UE UE eRIS | , and ζ ∈ (0, 1) represents the imperfect
1 k K

Let ΦK UE = ξηUE diag ejϕUE , . . . , ejϕUE , . . . , ejϕUE be the
diagonal matrix of UE’s communication at the RIS, where SIC indicator. Meanwhile, the SINR to decode xUE at UE over
k ∈ {1, . . . , K}, ξ ≥ 1 is the amplification factor2 [7], ηUE is the sub-carrier sUE can be written as
the amplitude coefficient subject to ηRU + ηTU = 1 due to the xUE
γUE = ϖP τBD τUE /[℘e
τUE + βUE N0 ]. (6)
law of energy conversion [8], and ϕUE,k is the k-th phase-shift.
Akin to [2], [17], to simplify STAR-RISs’ hardware design III. I NFORMATION -T HEORETIC F RAMEWORK
and phase-shift computational complexity, ϕUE,k is considered Given a packet is delivered over wireless channels with the
with discrete phase-shift models, i.e., SINR γ ∈ {γUE xUE
,γ xUE
eUE }, the ABLER can be expressed as [4]
ϕkUE = − arg(hkRIS ) − arg(hkUE ) + φkUE , (2) Z ∞  
C(x) + OM − NUE /M
ϵxUEUE (γ) ≃ Q p fγ (x)dx, (7)
where arg(hkRIS ), arg(hkUE ) ∈ [−π, π), and φkUE is the error 0 V (x)/M
phase-shift, following uniform distribution within [−2q π, 2q π], where C(γ) = log2 (1 + γ), V (γ) = (log2 (e))2 (1 − 1/(1 −
with q being a finite number of control bits. γ)2R), Q−1 (·) is the inverse of Gaussian Q-function Q(x) =
In p-NOMA, the BS uses three sub-carriers to communicate 1 ∞ 2
2π x exp(−z /2)dz , and OM = log2 (M)/M. Provided
with UE, where the sub-carrier sSU delivers a short packet that M ≥ 100, ϵxUEUE (γ) in (7) is approximated as [5, eq. (8)]
(including NTU and NRU transmission bits and the size of the
channel blocklength (channel use) M) to TU and RU via the ϵxUEUE (γ) ≈ Fγ (uUE ) /4 + 7Fγ (νUE ) /12 + Fγ (vUE ) /6, (8)
√ √
superimposed signal x[t] = αTU xTU + αRU xRU over the where uUE = ϱUE − 5
vUE = ϱUE + νUE = ϱUE + 1 1
8χUE , 2χUE , 24χUE ,
bandwidth portion βSU , where βSU = 1 − βTU − βRU and αUE NUE /M
p
ϱUE = 2 − 1, and χUE = M/[2π(22NUE /M − 1)].
is the power allocation coefficient of xRU , with E{|xRU |2 } =
E{|xTU |2 } = 1. Suppose that RU has a priority higher than
A. CDF of SINR Analysis
TU; thus, αRU > αTU subject to αRU + αTU = 1. Thus, while
RU directly decodes xRU , TU uses SIC to first decode xRU and As τBD is expressed by a sum of L gamma random variables
then recover xTU . Meanwhile, the remaining two sub-carriers |hlBD |2 , its PDFis given by
sUE conveys individual short-packet of UE over the bandwidth z LmBD −1 mBD
 LmBD 
mBD

portion βUE by letting αRU = 0 or αTU = 0. fτBD (z) = exp − z . (9)
Γ(LmBD ) ΩBD ΩBD
At the t-th time slot, BD backscatters the signal x[t] to UE
with its signal c[t] such that E{|c[t]|2 } = 1. The composite Lemma 1: The CDF of τUE can be derived as
signal received at UE from RIS can be expressed as FτUE (z) = 1 − Γ κUE , θUE z ℓ /Γ(κUE ),


(10)
√ √ K
yUE [t] = P x[t] ϖc[t]whBD hT ΦUEp hRIS + h e T ΦKa h


UE
e RIS
UE UE where ℓ ∈ {1, 1/2} represents the hybrid and active modes.
(1) (2) (1)
e T ΦKa e When ℓ = 1/2 (hybrid), θUE = µκUE /[µκUE − (µκUE )2 ] and
+ ℘h UE UE eRIS + nUE , (3) (1) 2 (2) (1) 2
κUE = (µκUE ) /[µκUE − (µκUE ) ]. When ℓ = 1 (active), θUE =
(n)
2 In terms of the architecture design, RIS acts in full passive mode (FPM) ma /[Ωa ξηUE Ka2 ] and κUE = ma . Herein, µκUE , Ωw , and mw ,
if: i) Kp ̸= 0 and Ka = 0; ii) Kp , Ka ̸= 0, and ξ = 1; and iii) Kp = 0, with w ∈ {a, p}, can be determined as
Ka ̸= 0, and ξ = 1. In contrast, RIS operates in full active mode (FAM) if
Kp = 0 and ξ > 1 and hybrid mode (HM) with Kp , Ka ̸= 0 and ξ > 1. In
Xn
i

p i i (i) √
µ(n)
κUE =
n n−i (n−i)
n ( ξ) Ka µτa ( ηUE ) Kp µτp , (11)
light of its application, FPM offers simplicity and low costs but struggles with i=0
n
weak signals and long distances. Though costly, FAM overcomes these issues
with amplifiers effectively but typically faces thermal noise. HM balances
µ(n) n
τw = Γ(mw + /2)/Γ(mw ) (Ωw /mw ) ,
2
(12)
energy efficiency and signal enhancement by switching between FAM and (1) (1) 2
Ωw = µ|hUE | µ|hRIS | 2q sin(π/2q )/π ,


(13)
FPM, but managing transitions and optimizing performance is challenging.

Authorized licensed use limited to: University of Ulsan. Downloaded on October 15,2024 at 13:07:29 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Communications Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2024.3475569

JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 3

Next, leveraging the first-order Riemann integral [5] for

Kw Ωw large P introduces the following asymptotic ABLER results
mw = (2) (2)
. (14) (
1


µ|hUE | µ|hRIS | 2 + sin (21−q π)/[22−q π] − Ωw ∞ Fγ∞ (ϱRU , RU, ϖP , βRU N0 ), sUE ,
ϵRU = (24)
Proof: Please find the details in Appendix A. ■ Fγ∞ (ϱRU , RU, (αRU −ϱRU αTU )ϖP , βSU N0 ), sSU ,

∞
Lemma 2: The PDF of τeUE can be derived as Fγ (ϱTU , TU, ϖP , βTU N0 ),
 sTU ,
∞
ϵTU = Fγ∞ (ϱRU , TU, (αRU − ϱRU αTU )ϖP , βSU N0 ) +
fτeUE (z) = exp (−z/λUE ) /λUE , λUE = σe2 ΩUE ξηUE Ka . (15)  ∞
 , sSU .
Fγ (ϱTU , TU, (αTU − ζϱTU αRU )ϖP , βSU N0 )
Proof: The result of (15) can be achieved by using the method (25)
of moments first to fit τeUE into circularly symmetric complex Corollary 1: The diversity of UE is LmBD .
Gaussian and then convert it into exponential distribution. ■
−LmBD
xUE
Proof: Since ϵ∞ ∞
UE ∝ Fγ (z, UE, A1 , A2 ) ∝ A1 , where
Knowing that Fγ (z) = Pr [γ < z], the CDF of γUE in (6), e2e
A1 ∝ P , one has that ϵUE ∝ 1/P LmBD
. ■
xRU xTU
eUE (4) and γ
γ eTU in (5) can be derived as
xUE Corollary 2: Provided that 2NUE bits are required to be
Pr [τBD τUE < z(℘eτUE + βUE N0 )/[ϖP ]] , γUE , (16)


   transmitted to UE, p-NOMA users achieve error performance
 Pr τ τ < z(℘e τUE + βSU N0 )

xRU lower than conventional OMA3 by approximately βUE times

BD UE , γ
eUE , (17)
Fγ (z) = (αRU − zαTU )ϖP on the sub-carrier sUE . While for the sub-carrier sSU , the
  

 z(℘eτUE + βSU N0 ) xTU performance gap equals to the ratio βSU /(αRU −ϱRU αTU ) for RU
 Pr τBD τUE <
 , γ
eTU . (18)
(αTU − zζαRU )ϖP and βSU / max{ϱTU (αRU − ϱRU αTU ), αTU − ϱTU ζαRU } for TU.
Let A1 ∈ {ϖP , (αRU − zαTU )ϖP , (αTU − zζαRU )ϖP } and Proof: As for sUE , we have ϵ∞ ∞
UE ∝ Fγ (z, UE, A1 , A2 ) ∝
A2 ∈ {βUE N0 , βSU N0 }. Fγ (z) is rewritten by a unique form: A2ℓκUE
and A2 ∝ βUE N0 . Thus, ϵe2e UE ∝ (βUE N0 )
ℓκUE
, while
ℓκUE
that of OMA is ϵOMA ∝ N . As for s , one can deduce
Fγ (z, UE, A1 , A2 ) = Pr [τBD τUE < z(℘e
τUE + A2 )/A1 ] . (19) e2e
UE 0
SU
ℓκUE e2e
that ϵRU ∝ βSU N0 /(αRU − ϱRU αTU ) and ϵTU ∝
Proposition 1: An analytical solution for Fγ (z, UE, A1 , A2 ) βSU N0 / max{ϱTU (αRU − ϱRU αTU ), αTU − ϱTU ζαRU }

ℓκUE
.
is in (20), where ∆UE ≜ 1/[Γ(κUE )Γ(LmBD )]. Diving ϵe2e to ϵ OMA
yields the outcome. ■
UE UE
Proof: Please find the details in Appendix B. ■ Corollary 3: Provided that 2NUE bits are required to be trans-
mitted to UE, p-NOMA users obtain the error performance
lower than conventional NOMA using two unit bandwidth.
B. Analysis of Average BLER (ABLER) and Goodput As for sSU , the performance gap is βSU . As for sUE , the
Based on (20) and (8), the respective end-to-end (e2e) performance gap fall into βRU (αRU − ϱRU αTU ) < 1 for RU and
ABLERs of RU and TU under SIC can be generalized as βTU max{ϱTU (αRU − ϱRU αTU ), αTU − ϱTU ζαRU } < 1 for TU.
xRU xRU xRU xRU
ϵe2e
RU = {ϵRU (γRU ) for sUE , ϵRU (γRU ) for sSU }, (21) Proof: This proof can be proven similar to Corollary 2. ■
(
ϵxTUTU (γTU
xTU
), sTU , The next focus is to study goodput, which refers to the num-
ϵe2e
TU = xRU xRU xRU xRU xTU xTU (22)
ϵTU (e γTU ) + [1− ϵTU (eγTU )]ϵTU (e
γTU ), sSU . ber of actual data bits successfully delivered to destinations
within a certain time. Goodput is a lower bound to throughput,
The Fox-H function’s complexity in (20) prevents extracting as it excludes retransmission and data overhead.
useful insights from ϵe2e
UE , requiring asymptotic BLER analysis.
Proposition 3: Given the transmission bits NUE = nUE + nt ,
Proposition 2: The asymptotic CDF of γ can be derived as where nUE and nt are the number of actual data bits and

z LmBD

mBD A2 /ΩBD
LmBD training bits, the network goodput for UE can be written as

 Γ(Lm
 BD +1) Ψ(UE,ℓ)A1 ,℘ = 0  
 LmBD nt h
i
RUE 1 − ϵe2e e2e


∞ GUE = 1 − UE sUE + 1 − ϵUE sSU , (26)

Fγ (z, UE, A1 , A2 ) = z LmBD mBD λUE /ΩBD
Γ(LmBD +1) Ψ(UE,ℓ)A1 NUE
 , ℘ = 1.


 
A
 
A2
Γ LmBD + 1, λ 2 / exp − λUE

UE where RUE = NUE /M is the target rate in bits per channel
(23) use (bpcu) and (·)s is the reliability of the carrier s.
1/ℓ
where Ψ(UE, ℓ) = Γ(κUE + 1/ℓ)/[Γ(κUE )θUE ].
3 To achieve this goal, conventional OMA must exploit two sub-carriers to
Proof: Please find the details in Appendix C. ■ serve users and each needs a unit bandwidth.

   ℓ 
 1 − ∆UE H3,0 (1,1) A2 mBD
θ z , ℘ = 0,

1,3 (LmBD ,ℓ),(κUE ,1),(0,1) UE A1 ΩBD
Fγ (z, UE, A1 , A2 ) =   h ℓ
i (20)
 1 − ∆UE exp A2 H0,1;0,1;3,0 0;1,ℓ (1,1)
 (1,1) λUE
; (ΩθUE (zmBD )
, ℘ = 1.
λUE 1,0;1,1;1,3 − (0,1) (LmBD ,ℓ),(κUE ,1),(0,1) A2 BD A1 /λUE )
ℓ

Authorized licensed use limited to: University of Ulsan. Downloaded on October 15,2024 at 13:07:29 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Communications Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2024.3475569

JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 4

Corollary 4: At high transmit
P , the ABLER e2eapproaches A PPENDIX A: P ROOF OF L EMMA 1

zero, which means 1 − ϵe2e


UE sUE ≃ 1 and 1 − ϵUE sSU ≃ 1.
PKw k
To begin, let τw = K1w k=1 |hUE ||hkRIS | exp(−jφUE,k ).
∞
Thus, the goodput floor of UE would be GUE = 2nUE RUE /NUE . √ √ 2
Thus, one has that τUE =  | ξηUE Ka τ a + ηUE Kp τp | . On this
k k
Remark 1: When NUE and RUE are fixed, Proposition 3 basis, one can derive E ℜ[τ  w ] = E |hUE ||hRIS | cos(φUE,k )
shows that increasing Nt decreases nUE , thereby degrading the and Var {ℜ[τw ]} = Var |hkUE ||hkRIS | cos(φUE,k ) /Kw . Mak-
spectrum efficiency. This yields a trade-off between nt and nUE ing
 use of [2,(1)eq. (1) (A.2) and (A.3)] enables us to get
on the goodput maximization. In the meanwhile, Corollary 4 E ℜ[τw ] = µ|hUE | µ|hRIS | 2q sin(π/2q )/π and Var {ℜ[τw ]} =
 (2) (2) 2
reveals that the goodput proposed by the p-NOMA is twice µ|hUE | µ|hRIS | 21 + sin ( 2π 2−q


2q )/[2 π] − E ℜ[τw ] /Kw . Us-
that of NOMA and four times that of OMA. ing [17, Theorem 1], τw can be mapped  into Nakagami-m dis-
IV. N UMERICAL R ESULTS AND D ISCUSSIONS tribution with
 the severity Ωw = E2 ℜ[τw ] and the envelope
mw = E2 ℜ[τw ] /(4Var {ℜ[τ√ w ]}). Provided that Kw ̸= 0
This section presents some numerical results to verify the √
and τUE = |κUE |2 with κUE = ξηUE Ka τa + ηUE Kp τp , we
theoretical analyses in Section III via Monte-Carlo simulation, obtain the n-moment of κUE in (11) via raw moment method.
where parameters are set as follows: L = 4, mBD = mRU = Applying the method of moments to √fit κUE into Gamma
2.5, mRIS = 1.5, mTU = 2, and ΩX = Ld−ς X , with the path- distribution and using FτUE (u) = FκUE ( z) yields the results
loss factors L = −30 dB and ς = 2 [7], and the average 2
distance dX ∈ {dBD = dTU = 10, dRIS = dRU = 12} (m).
for ℓ = 1/2. Given p Kp = 0, τUE = ξηUE Ka |τa |. Applying
FτUE (u) = Fτa ( z/[ξηUE Ka2 ]) yields the results of ℓ = 1.
ϖ = 0.8, ζ = 0.01, Ka = 24, Kp = 32, ηUE = 0.5 [4], ξ = 5,
σe2 = −70 dBm [7], q = 2 [17], N0 = −174 + 10 log10 (BW) A PPENDIX B: P ROOF OF P ROPOSITION 1
dBm for BW = 100 MHz [16], βUE = 0.25, αRU = 0.7 [7],
Using (9) and (10) yields the joint CDF of τUE and τBD as
NRU = 150 and NTU = 100 [6], and M = 200 [4]. Z ∞ z
Figs. 2(a) and (b) plots the e2e ABLER of UE under active, FτUE τBD (z) = FτUE fτBD (x)dx
passive, and hybrid STAR-RISs. As expected, the developed 0 x
framework (black solid lines) shows a small gap compared to = 1 − ∆UE H3,0
 (1,1) ℓ
1,3 (LmBD ,ℓ),(κUE ,1),(0,1) θUE (mBD z/ΩBD ) , (27)
simulation (markers) results and well predicts the asymptotic
 (1−κ,1),(1,1) 
outcome at high P , confirming the correctness of (21), (22), which is obtained using Γ (κ, 1/x) = H0,2 2,1 (0,1)
x ,
(24), and (25). As for the focus of hybrid STAR-RISs, we can −
exp(−x) = H1,0
 
0,1 (0,1) x , and [18, eqs. (1.2.2) and (2.68)].
observer from Fig. 2(c) that p-NOMA outperforms NOMA in When ℘ = 0, Fγ (z, UE, A1 , A2 ) in (19) can be derived as
all sUE and sSU . Compared to OMA, p-NOMA shows superior
ABLER enhancement with sUE and a small ABLER reduction Fγ (z, UE, A1 , A2 ) = FτUE τBD (zA2 /A1 ) . (28)
with sSU . These verify the conclusions in Corollaries 2 and 3. When ℘ = 1, Fγ (z, UE, A1 , A2 ) in (19) can be computed as
Considering hybrid STAR-RISs, Fig. 3(a) depicts that users’ Z ∞  
goodput significantly improves with low and moderate P z
Fγ (z, UE, A1 , A2 ) = FτUE τBD (x + A2 ) fτeUE (x)dx
but saturates at high P . Besides, utilizing flexible bandwidth A
Z ∞ 0  1  
resources makes p-NOMA’s goodput improvement stand out ∆UE 1,0 − y 0,1 (1,1) y
=1− H H1,1 (0,1) (29)
compared to OMA and NOMA, verifying Corollary 4 and λUE 0 0,1 (0,1) λUE A2
Remark 1. In Fig. 3(b), when ξ increases from 0 (perfect     ℓ 
A2 (1,1) zmBD y
SIC) to 0.35 (imperfect SIC), the goodput of TU decreases exp H3,0
1,3 (LmBD ,ℓ),(κUE ,1),(0,1) θUE dy,
significantly at P around 20 and 25 dBm. Finally, Fig. 3(c) λUE ΩBD A1
plots the users’ goodput as a function of transmissions bits where the above result is done by three steps: 1) chang-
NUE with different blocklength setting M = 128, 256 channel R ∞ variable y =
ing R ∞ x + A2 ; 2) using integral exchange
uses (c.u) at P = 20 dBm. As observed, the goodput c
f (y)dy = 0
f (y)H (y/c − 1) dy; and 3) applying
1,0  −  (1,1) 
exp(−x) = H0,1 (0,1) x and H (|x| − 1) = H0,1

curves have indeed concave forms, meaning that there are 1,1 (0,1) x .
the optimal values of NUE at which the users’ goodputs are Now, invoking [19, eq. (2.3)] for (29) and injecting (27) into
maximized. However, such optimal values vary depending on (28), one can obtain the result in (20).
the blocklength M.
A PPENDIX C: P ROOF OF P ROPOSITION 2
V. C ONCLUSIONS
At high P , it is feasible to assume stable communication
This paper proposed and investigated the performance of between RIS and UE, which averages the value of τUE as
a hybrid STAR-RIS-empowered ambient backscatter short- Z ∞
packet system with p-NOMA. To characterize passive, ac- Ψ(UE, ℓ) ≜ E{τUE } = [1 − FτUE (x)]dx
tive, and hybrid STAR-RIS types, a generalized information- 0
Z ∞
theoretic framework was provided in terms of both approxi- u←xℓ 1 1

(1,1)

= u ℓ −1 H2,0
1,2 (κUE ,1),(0,1)
θ UE u du, (30)
mate and asymptotic, and its efficacy was corroborated through Γ(κUE )ℓ 0
simulation results. Additionally, the findings indicated the which can be solved using [19, eq. (1.5)]. On this basis, the
ABLER superiority of the proposed p-NOMA over its OMA result in (19) with P → ∞ can be approximated as
and NOMA counterparts and significant enhancement in the
goodput performance when using the proposed p-NOMA. τUE + A2 )/[Ψ(UE, ℓ)A1 ]] .
Fγ (z)≃ Pr [τBD < z(℘e (31)

Authorized licensed use limited to: University of Ulsan. Downloaded on October 15,2024 at 13:07:29 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Communications Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2024.3475569

JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 5

10 0 10 0 10 0

10 -1 10 -1 10 -1
Average BLER

Average BLER

Average BLER
NOMA - RU
10 -2 10 -2 10 -2
OMA - TU NOMA - TU

10 -3 10 -3 10 -3
OMA - RU
-4 -4 -4
10 10 10
10 15 20 25 30 35 40 10 15 20 25 30 35 40 10 15 20 25 30 35 40

(a) RU’s BLER (b) TU’s BLER (c) p-NOMA versus NOMA and OMA

Fig. 2. Average BLER evaluation.

1.5 1 1.5

0.8
Goodput [bpcu]

Goodput [bpcu]

Goodput [bpcu]
1 1
0.6

0.4
0.5 0.5

0.2

0 0 0
10 15 20 25 30 35 40 10 15 20 25 30 35 40 32 92 152 212 272 332 392 452 512

(a) Impact of P (b) Impact of ζ (c) Impact of NUE

Fig. 3. G
oodput evaluation with fixed training bits at nt = 32 (bits).

Next, invoking the first-order term in [15, eq. (8.31.2)] yields [7] J. Xu, J. Zuo, J. T. Zhou, and Y. Liu, “Active simultaneously transmitting
z→0 Lm and reflecting (STAR)-RISs: Modeling and analysis,” IEEE Commun.
Fτ∞
BD
(z) ≃ (mBD /ΩBD z) BD /Γ(LmBD + 1). Conditioned on Lett., vol. 27, no. 9, pp. 2466–2470, June 2023.
℘ = 0, the result of (31) can be further approximated as [8] X. Yue, J. Xie, C. Ouyang, Y. Liu, X. Shen et al., “Active simultaneously
transmitting and reflecting surface assisted NOMA networks,” IEEE
P →∞
Fγ (z, UE, A1 , A2 ) ≃ Fτ∞
BD
(zA2 /[Ψ(UE, ℓ)A1 ]) . (32) Trans. Wireless Commun., vol. 23, no. 8, pp. 9912–9926, Feb. 2024.
[9] Y. Xu, Z. Qin, G. Gui, H. Gacanin, H. Sari et al., “Energy efficiency
While for ℘ = 1, the result of (31) can be approximated as maximization in NOMA enabled backscatter communications with QoS
Z ∞   guarantee,” IEEE Wireless Commun. Lett., vol. 10, no. 2, pp. 353–357,
P →∞ z(x + A2 ) Oct. 2020.
Fγ (z, UE, A1 , A2 ) ≃ Fτ∞ fτeUE (x)dx [10] W. U. Khan, X. Li, M. Zeng, and O. A. Dobre, “Backscatter-enabled
0
BD
Ψ(UE, ℓ)A1 NOMA for future 6G systems: A new optimization framework under
Z ∞ Lm imperfect SIC,” IEEE Commun. Lett., vol. 25, no. 5, pp. 1669–1672,
(mBD /ΩBD z) BD (x + A2 )LmBD e−x/λUE
= dx, (33) Jan. 2021.
Lm
0 Γ(LmBD + 1) (Ψ(UE, ℓ)A1 ) BD λUE [11] X. Li, Q. Wang, M. Zeng, Y. Liu, S. Dang et al., “Physical-Layer
Authentication for Ambient Backscatter-Aided NOMA Symbiotic Sys-
where the above integral can be solved by [15, eq. (3.382.4)]. tems,” IEEE Trans. Commun., vol. 71, no. 4, pp. 2288–2303, Feb. 2023.
[12] Y.-C. Liang, Q. Zhang, J. Wang, R. Long, H. Zhou et al., “Backscatter
Communication Assisted by Reconfigurable Intelligent Surfaces,” Proc.
R EFERENCES IEEE, vol. 110, no. 9, pp. 1339–1357, May 2022.
[1] Y. Liu, X. Mu, J. Xu, R. Schober, Y. Hao et al., “STAR: Simultaneous [13] B. Kim, Y. Park, and D. Hong, “Partial non-orthogonal multiple access
Transmission and Reflection for 360° Coverage by Intelligent Surfaces,” (P-NOMA),” IEEE Wireless Commun. Lett., vol. 8, no. 5, pp. 1377–
IEEE Wireless Commun., vol. 28, no. 6, pp. 102–109, Dec. 2021. 1380, May 2019.
[2] T.-H. Vu, A. Jee, D. B. da Costa, and S. Kim, “STAR-RIS empowered [14] B. Zhuo, J. Gu, W. Duan, X. Gu, G. Zhang et al., “Partial non-orthogonal
NOMA systems with caching and SWIPT,” IEEE Open J. Commun. multiple access: A new perspective for RIS-aided downlink,” IEEE
Soc., vol. 5, pp. 379–396, Dec. 2023. Wireless Commun. Lett., vol. 11, no. 11, pp. 2395–2399, Sep. 2022.
[3] C. Wu, Y. Liu, X. Mu, X. Gu, and O. A. Dobre, “Coverage characteriza- [15] A. Jeffrey and D. Zwillinger, Table of integrals, series, and products.
tion of STAR-RIS networks: NOMA and OMA,” IEEE Commun. Lett., Elsevier, 2007.
vol. 25, no. 9, pp. 3036–3040, June 2021. [16] A. Soulimani, M. Benjillali, H. Chergui, and D. B. da Costa, “Multihop
[4] F. Karim, S. K. Singh, K. Singh, S. Prakriya, and M. F. Flanagan, weibull-fading communications: Performance analysis framework and
“On the performance of STAR-RIS-Aided NOMA at finite blocklength,” applications,” J. Franklin Inst., vol. 358, no. 15, pp. 8012–8044, Oct.
IEEE Wireless Commun. Lett., vol. 12, no. 5, pp. 868–872, Feb. 2023. 2021.
[5] J. Xu, L. Yuan, N. Yang, N. Yang, and Y. Guo, “Performance analysis of [17] M.-A. Badiu and J. P. Coon, “Communication through a large reflecting
STAR-IRS aided NOMA short-packet communications with statistical surface with phase errors,” IEEE Wireless Commun. Lett., vol. 9, no. 2,
CSI,” IEEE Trans. Veh. Technol., vol. 72, no. 9, pp. 12 385–12 390, Apr. pp. 184–188, Oct. 2019.
2023. [18] M. E. Ismail, The H-Function with Applications in Statistics and Other
[6] T.-H. Vu, T.-V. Nguyen, Q.-V. Pham, D. B. da Costa, and S. Kim, Disciplines. JSTOR, 1980.
“STAR-RIS-enabled short-packet NOMA systems,” IEEE Trans. Veh. [19] P. Mittal and K. Gupta, “An integral involving generalized function of
Technol., vol. 72, no. 10, pp. 13 764–13 769, May 2023. two variables,” Proc. Math. Sci., vol. 75, no. 3, pp. 117–123, Mar. 1972.

Authorized licensed use limited to: University of Ulsan. Downloaded on October 15,2024 at 13:07:29 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.