IEEE WIRELESS COMMUNICATIONS LETTERS, VOL. 11, NO.

10, OCTOBER 2022 2195

Bi-Objective Optimization of Information Rate and Harvested Power

in RIS-Aided SWIPT Systems
Abdelhamed Mohamed , A. Zappone , Senior Member, IEEE, and Marco Di Renzo , Fellow, IEEE

Abstract—The problem of simultaneously optimizing the to jointly optimize the transmit beamforming vector, the
information rate and the harvested power in a reconfigurable power-split ratio, and the RIS phase shifts under QoS con-
intelligent surface (RIS)-aided multiple-input single-output down- straints. Semidefinite relaxation coupled with alternating and
link multiuser wireless network with simultaneous wireless
sequential optimization methods are employed. In [6], the
information and power transfer (SWIPT) is addressed. The
beamforming vectors, RIS reflection coefficients, and power problem of maximizing the weighted sum-rate is investigated
split ratios are jointly optimized subject to maximum power in a SWIPT-based multi-user multiple-input multiple-output
constraints, minimum harvested power constraints, and real- (MIMO) downlink system, subject to minimum harvested
istic constraints on the RIS reflection coefficients. A practical energy constraints. Alternating optimization is used in con-
algorithm is developed through an interplay of alternating junction with sequential optimization and pricing methods.
optimization, sequential optimization, and pricing-based meth- In [7], the authors study the problem of resource allocation in
ods. Numerical results show that the deployment of RISs jointly
RIS-aided SWIPT-based systems, in which a large RIS is split
improves the information rate and the amount of harvested
power. into several tiles that are designed for reducing the computa-
tional complexity. A globally optimal algorithm and a practical
Index Terms—RIS, SWIPT, multi-objective optimization. approach are developed by means of branch-and-bound and
sequential methods. In [8], the trade-off between the sum-rate
I. I NTRODUCTION maximization and total harvested power is investigated. The
ECONFIGURABLE intelligent surfaces (RISs) have −method coupled with alternating optimization is used to
R emerged as a promising technology for sustainable sixth-
generation (6G) networks [1], [2]. Thanks to their ability of
tackle the resulting multi-objective problem. In [9], the data
rate maximization problem in an RIS-aided system in which
reflecting and refracting electromagnetic signals in a recon- multiple receivers perform information decoding and wireless
figurable fashion and with limited energy requirements, RISs power reception is analyzed. The problem is tackled by alter-
can drastically reduce the energy consumption in wireless nating optimization, sequential optimization, and sub-gradient
networks [3]. In this context, RISs have also been studied in searches.
conjunction with simultaneous wireless information and power This letter considers a network in which a multiple-antenna
transfer (SWIPT), which is another key technology to improve access point (AP) serves single-antenna users with the aid of
the energy sustainability of future wireless networks. an RIS. Each receiver jointly performs information decoding
Several studies show that the deployment of RISs can and wireless power harvesting by means of power splitting.
improve both the information and power transfer. In [4], the Unlike previous works, the following contributions are made.
problem of transmit power minimization subject to quality 1) We consider the novel bi-objective simultaneous
of service (QoS) constraints and minimum energy harvest- maximization of the information rate and harvested power,
ing requirements is addressed. The optimization problem subject to minimum downlink rate, minimum harvested power,
is tackled by means of a penalty-based algorithm cou- and maximum transmit power constraints. The resulting
pled with the alternating optimization technique. In [5], the non-convex problem is tackled by an interplay of alternat-
problem of transmit power minimization for an RIS-assisted ing maximization, sequential optimization, and penalty-based
SWIPT non-orthogonal multiple-access (NOMA) network is methods. Since the receivers perform both information decod-
investigated. A two-stage optimization algorithm is proposed ing and wireless power harvesting, the optimization of the
power split ratio is needed, which is a novel feature com-
Manuscript received 17 June 2022; revised 30 July 2022; accepted 3 pared to most related works on RIS-aided and SWIPT-based
August 2022. Date of publication 8 August 2022; date of current version
7 October 2022. The work of Abdelhamed Mohamed and Marco Di Renzo systems.
was supported in part by the European Commission through the H2020 2) We consider the realistic case in which the phases and
PAINLESS Project under Grant 812991; in part by the H2020 ARIADNE moduli of the RIS reflection coefficients are not independent of
Project under Grant 871464; and in part by the H2020 RISE-6G Project
under Grant 101017011. The work of A. Zappone was supported by the
one another, but are coupled by a deterministic function. This
Italian Ministry of University and Research through the Project Dipartimenti further complicates the scenario, making it challenging to find
di Eccellenza 2018–2022. The associate editor coordinating the review of this performing resource allocations at an affordable complexity.
article and approving it for publication was M. Derakhshani. (Corresponding 3) Numerical results confirm the effectiveness of the
author: Marco Di Renzo.)
Abdelhamed Mohamed and Marco Di Renzo are with the Université proposed algorithm compared to traditional approaches. It is
Paris-Saclay, CNRS and CentraleSupélec, Laboratoire des Signaux et found, in particular, that increasing the number of RIS ele-
Systèmes, 91192 Gif-sur-Yvette, France (e-mail: marco.di-renzo@universite- ments allows harvesting enough power, while at the same time
paris-saclay.fr).
A. Zappone is with the Department of Electrical and Information
supporting satisfactory sum-rate levels.
Engineering, University of Cassino and Southern Lazio, 03043 Cassino, Italy. Among previous works, [8] and [9] are the most closely-
Digital Object Identifier 10.1109/LWC.2022.3196906 related to our work. However, [8] investigates the rate and
2162-2345 
c 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: M.KUMARSAMY COLLEGE OF ENGINEERING. Downloaded on February 14,2023 at 05:55:06 UTC from IEEE Xplore. Restrictions apply.
 2196 IEEE WIRELESS COMMUNICATIONS LETTERS, VOL. 11, NO. 10, OCTOBER 2022

we consider the associated rate function:


K
ξk ηk (1 − ρk )  
K
2
R PH = log 1 + (hd,k + vH Hr ,k )wi  (2)
σk2
k =1 i=1

where ξk ∈ [0, 1] is the efficiency of the conversion from

baseband power to RF power.
The goal of this letter is to maximize a weighted sum of
R ID and R PH , namely:
Eq
Rsum (ρ, w, v, {θn }) = R ID + λ̄R PH (3)

Fig. 1. Illustration of the MISO RIS-assisted SWIPT system model. where λ̄ is a parameter to be tuned by the network operator
according to the priorities granted to ID and PH [11].
Defining hk = hd,k + vH Hr ,k , the problem to be tackled
harvested energy trade-off for separated information and in the rest of this letter is formulated as:
power receivers, i.e., each receiver performs either information
decoding or wireless power harvesting. Also, independent PA : max Eq
Rsum (ρ, w, v, {θn }) (4)
phases and moduli are assumed for the RIS reflection ρ,w,v,{θn }

coefficients. In addition, [9] considers integrated information |hk wk |2

s.t. C1:  ≥ γmin , k = 1, . . . , K
and power receivers, but it focuses on maximizing only the δk2
K
i=k |hk wi |2 + σk2 + ρk
information rate, without considering power harvesting and the
(5)
relation between the phase and amplitude of the RIS reflection

K
coefficients. C2: ηk (1 − ρk ) |hk wi |2 ≥ Pmin , k = 1, . . . , K (6)
i=1

II. S YSTEM M ODEL AND P ROBLEM F ORMULATION 

K
C3: wk 2 ≤ PT ,, 0 ≤ ρk ≤ 1 (7)
We consider an RIS-based multi-user multiple-input single- k =1
output (MISO) downlink indoor scenario in which an access C4: vn = fn (θn )e j θn , n = 1, . . . , N (8)
point (AP) equipped with M antennas serves K single- C5: − π ≤ θn ≤ π, n = 1, . . . , N (9)
antenna user equipment (UEs) employing SWIPT. The k-th UE
employs a fraction ρk of the received power for information It can be seen that PA is a non-convex problem due to
decoding (ID), while the rest is used for power harvesting the non-concavity of both the objective function and the
(PH). The channels from the AP to the RIS, from the AP to constraints C1, C2, C4. Thus, traditional methods do not apply.
the k-th user, and from the RIS to the k-th user are denoted
by G ∈ CN ×M , hd,k ∈ C1×M , and hr ,k ∈ C1×N , respec-
tively, and are assumed to follow the Rician fading model. III. P ROPOSED A PPROACH
The reflection coefficient vector of the RIS is denoted by
To tackle PA , we first reformulate the sum of logarithms
v = [v1 , . . . , vn ] ∈ CN ×1 , where vn = fn (θn )e j θn is the
into a more tractable form by applying the method from [12]
reflection coefficient of the n-th reflecting element of the
to each sum in the objective function. This yields:
RIS, with −π ≤ θn ≤ π and, e.g., fn (θn ) = fmin + (1 −
sin(θn −φ)+1 α
fmin )( 2 ) is a function that relates the phase of PA : max fA (αI , βI , αE , βE , ρ, w, v, {θn })
the reflection coefficient to its modulus, where fmin ≥ 0, αI ,βI ,αE ,βE ,ρ,w,v,{θn }
α ≥ 0, φ ≥ 0 are circuit implementation constants [10]. The s. t . (C1), (C2), (C3), (C4), (C5) (10)
proposed approach can be applied to any continuous and dif-
ferentiable function fn (θn ). Given this notation and defining wherein fA is shown in (11), shown at the bottom of the next
Hr ,k = diag(hr,k )G, the sum-rate is: page, with η̄k = ξk ηk and  being the real part operator. In
⎛  ⎞ order to tackle (10), the first step is to embed (C4) into the
 

K  hd,k + vH Hr ,k wk 2 objective, resorting to a penalty-based approach, which yields:
R ID = log⎝1 +    
⎠
δk2
k =1
K  hd,k + vH Hr ,k wi 2 + σ 2 +
i=1,i=k k ρk
(1) N 
 2
 
PB : max fA − Γ vn − fn (θn )e j θn 
wherein wk ∈ CM ×1 is the transmit beamforming vec- αI ,βI ,αE ,βE ,ρ,w,v,{θn }
n=1
tor, and σk2 and δk2 model the power of the thermal noise s. t . (C1), (C2), (C3), (C5) (12)
and of the noise due to the conversion of the RF signal
to the baseband. Furthermore, considering a linear harvest- wherein Γ represents the penalty coefficient used for penaliz-
ing model, the power harvested by the k-th UE is PH ,k = ing the violation of the equality constraint (C4). If Γ → ∞,
 2
ηk (1 − ρk ) K i=1 |(hd,k + v Hr ,k )wi | , where ηk ∈ [0, 1] is
H
the solution of the original problem is obtained. Problem (12)
the efficiency of the power harvesting circuit. Similar to [11], is tackled by alternating optimization, as explained next.

Authorized licensed use limited to: M.KUMARSAMY COLLEGE OF ENGINEERING. Downloaded on February 14,2023 at 05:55:06 UTC from IEEE Xplore. Restrictions apply.
 MOHAMED et al.: BI-OBJECTIVE OPTIMIZATION OF INFORMATION RATE AND HARVESTED POWER 2197

A. Optimization of αI ,k , αE ,k , βIk , βEk , ρk C. Optimization of v

The optimal αI ,k , αE ,k , βIk , βEk are found by setting the The approach is similar to that used for the optimization
gradient of the objective to zero, which yields: of wk . In fact, the objective is concave in v, while the con-
√ straints (C1) and (C2) can be handled by the SCA method.
r2 + r r2 + 4 ρk (1 + αIk )(hk wk )
αI ,k = , βI ,k = K 2
(13) Specifically, (C1) can be replaced by the convex constraint:
2 2
i=1 k |hk wi | + ρk σk + δk
ρ 2
√  ⎛ ⎞
r̃ 2 + r̃ r̃ 2 + 4 η̄k (1 − ρk )(1 + αEk ) K i=1 hk wi

K
δk2
αE ,k = , βE ,k = K 2 γmin ⎝ |(hd,k + v Hr ,k )wi | +
H 2
σk2 + ⎠
2 η̄k (1 − ρk ) i=1 |hk wi | + σk 2 ρk
i=k
(14)
 + (hd,k + v(t)H Hr ,k )wk wkH (hd,k + v(t)H Hr ,k )H
∗ 
with r̃ = η̄(1 − ρk ){βE ,k i=1 hk wi }, − 2{(hd,k + v(t)H Hr ,k )wk wkH (hd,k + vH Hr ,k )H } ≤ 0 (17)
√
r = ρk {βI∗,k hk wk }.
The optimization with respect to the coefficients {ρk } is per- and (C2) by the convex constraint:
formed as follows. With respect to {ρk }, the objective function 
K

is strictly concave and the constraints (C2), (C3) are affine. ηk (1 − ρk ) (hd,k + v(t)H Hr ,k )wi wiH (hd,k + v(t)H Hr ,k )H −
i=1
Moreover, (C1) can be rewritten in a linear form as follows,
2{(hd,k + v(t)H Hr ,k )wi wiH (hd,k + vH Hr ,k )H } ≥ Pmin (18)
for any k = 1, . . . , K :
⎛ ⎞

K By replacing the constraints (C1) and (C2) with (17)
ρk |hk wk |2 − γmin ⎝ρk |hk wi |2 + ρk σk2 + δk2 ⎠ ≥ 0 (15) and (18), we obtain the convex surrogate problem to be solved
i=1,i=k in each iteration of the SCA method for optimizing v.
Thus, with respect to {ρk }, the problem is convex and can
be solved by standard convex optimization algorithms [13]. D. Updating θn
The RIS phase shifts are the solutions of the problem:
B. Optimization of wk N  2
  
When all the other variables are fixed, the objective function max − vn − fn (θn )e j θn  , s.t. − π ≤ θn ≤ π (19)
is a concave function in the transmit beamforming vectors wk . {θn }
n=1
However, constraints (C1) and (C2) are still not convex. To It can be seen that the problem is separable over n, i.e.,
deal with them, we use the successive convex approximation each summand can be optimized separately. Thus, defining
(SCA) method [6], [14]. Specifically, the convex term |hk wk |2 ϕn = arg(vn ), the optimal θn is the solution of the problem:
is upper-bounded by its first-order Taylor expansion as follows:

(t)H H
 
(t)H H (t)
 max 2fn (θn )|vn | cos(ϕn − θn ) − fn2 (θn ) (20)
wkH hH
k hk w k ≥ 2 wk hk hk w k − w k hk hk w k (16) θn ∈[−π,π]

(t) which can be solved by standard numerical methods.

wherein wk , ∀k is the solution from the previous iteration.
(w ) (t)H (t)
Thus, defining dk = wk hH k hk wk , exploiting (16) and E. Convergence and Complexity
elaborating, (C1) can be recast as: Finally, the overall algorithm to solve the optimization
⎛ ⎞
 δk2   problem is obtained by iteratively optimizing the different
γmin ⎝ 2
|hk wi | + σk2 + ⎠ + d (w ) − 2 w(t)H hH
k hk wk ≤ 0 optimization variables. Each iteration monotonically increases
ρk k k
i=k the objective value, which guarantees convergence. Moreover,
Similarly, (C2) can be reformulated as follows: the computational complexity is polynomial in the number
of variables, since only the solution of convex problems

K  
ηk (1 − ρk )
(t)H H
2 wi
(t)H H
hk hk w i − w i
(t)
hk hk wi ≥ Pmin is required.1 Thus, the complexity of optimizing {ρk } is
k =1
O(K ηk ), while the complexity of optimizing {wk } and v
are O(Iw (MK )ηw ), and O(Iv N ηv ), respectively, with Iw
By replacing (C1) and (C2) with their reformulated versions,
we obtain the convex surrogate problem to be solved in each 1 We recall that a convex problem can be solved with a complexity C =
iteration of the SCA method for optimizing wk . O(Lη ), where L is the number of variables and 1 ≤ η ≤ 4 [15].


K
 
fA (αI , βI , αE , βE , ρ, w, v, {θn }) = log(1 + αI ,k ) + λ̄ log(1 + αE ,k ) − αI ,k + λ̄αE ,k
k =1
K 
 
K
  K
+ 2 ρk (1 + αI ,k )(βI∗,k hk wk ) − βI ,k 2 ρk |hk wi |2 + ρk σk 2 + δk2
k =1 k =1 i=1
K K K
 2 
K
+ 2λ̄ ∗
η̄k (1 − ρk )(1 + αE ,k )(βE ,k hk wi ) − λ̄βE ,k  ξk ηk (1 − ρk ) |hk wi |2 + σk 2 (11)
k =1 i=1 k =1 i=1

Authorized licensed use limited to: M.KUMARSAMY COLLEGE OF ENGINEERING. Downloaded on February 14,2023 at 05:55:06 UTC from IEEE Xplore. Restrictions apply.
 2198 IEEE WIRELESS COMMUNICATIONS LETTERS, VOL. 11, NO. 10, OCTOBER 2022

TABLE I
and Iv being the number of iterations of the SCA meth- S IMULATION PARAMETERS
ods used to optimize w and v. On the other hand, the
optimal {αI ,k , αE ,k , βIk , βEk } are available in closed-form
in (13), (14) and thus the complexity required for their com-
putation can be neglected. Similarly, the complexity of the
problem in (19) can also be neglected, as it is linear in N. In
fact, the problem can be decoupled over the N optimization
variables, and, for each N, the optimal θn is obtain by solv-
ing (20). Thus, the overall complexity of the proposed method
is given by C = I (O(K ηk ) + O(Iw (MK )ηw ) + O(Iv N ηv )),
where I is the number of alternating optimization iterations
to be run until convergence. The exponents of the polynomial
are not available in closed-form, but it is known that they are
upper-bounded by 4 [15]. A typical value is 3.5, which comes
up when interior-point methods are used [13].

IV. N UMERICAL R ESULTS

We consider an RIS-assisted MISO communication system
in a typical indoor scenario, in which M = 32 transmit anten-
nas are arranged in a uniform linear array, and each antenna
has 6 dBi gain. Unless stated otherwise, we set K = 4 UEs
and N = 400 RIS elements. The UEs are randomly and uni-
formly distributed within a disk of 2 m radius centered at
(10 m, 10 m). An N-element RIS is located at (0 m,5 m).
All the channels are modeled as X = Lx ( 1+ X̄ LOS +
1 X̄ NLOS ), where X̄ LOS and X̄ NLOS are the line-
1+
of-sight (LOS) and non-LOS (NLOS) components, and X
is either G, hr ,k , or hd,k . The NLOS component follows
the Rayleigh fading model, while the LOS component is Fig. 2. Sum-rate and harvested power versus the number of RIS elements
X̄ LOS = aN (θAoA )aH M (θ
AoD ), with: for different values of λ̄ (K = 4).

2πd sin(θ AoA ) 2πd (N −1) sin(θ AoA ) T

aN (θAoA ) = [1, e j λ , . . . , ej λ ]
j 2πd
λ sin(θ
AoD
) j 2πd
λ (M −1) sin(θ
AoD
) T
aM (θ AoD
) = [1, e ,...,e ] In Fig. 3, we explore the trade-off between the sum-rate and
harvested power as a function of the number of RIS elements.
where d and λ are the inter-antenna distance and the wave- The case without RIS is reported for comparison. We observe
length, respectively. We assume d /λ = 1/2. The path-loss is that employing more RIS elements enhances the sum-rate
L = C0 ( Dr0 )−xl , where r is the link distance and D0 = 1 and harvested power, and the proposed algorithm outperforms
is the reference distance at which the reference path-loss the “No RIS” scenario. Moreover, the figure highlights the
C0 = −30 dB is defined, xl is the path-loss exponent. Other impact of the parameter λ̄, which is utilized to determine the
simulation parameters are given in Table I and are in agree- service priority between optimizing the sum-rate (i.e., small
ment with [9]. For comparison, we evaluate the performance values of λ̄) or optimizing the harvested power (i.e., large val-
gain obtained by the proposed algorithm against the “No-RIS” ues of λ̄). When the system prioritizes power harvesting, the
scenario, where no RIS is deployed in the network. proposed algorithm allocates more power to the power har-
Figure 2 depicts the relation between the number of RIS vesting receiver, and thus the sum-rate decreases. Similarly,
elements versus the sum-rate and the harvested power. We reducing the value of λ̄ gives higher priority to the information
see that employing more RIS elements leads to a monotonic decoding receiver. The figure shows the trade-off between
growth of the amount of harvested power and sum-rate. The the sum-rate and harvested power for any values of λ̄. The
figure reveals the effectiveness of the proposed algorithm com- extreme points to the left and to the right of the curves corre-
pared to the “No RIS” case (denoted by “w/o RIS”). This spond to the single-objective sum-rate maximization and to the
monotonic gain is due to the appropriate design of the phase single-objective harvested power maximization, respectively.
shifts of the RIS elements, which results in strong virtual LOS In Fig. 4, we examine the trade-off between the sum-rate
paths between the AP and the UEs. Also, the figure shows and harvested power with the minimum value of the amplitude
two special cases in which we optimize only the achievable of the reflection coefficient fmin of the RIS elements. We see
rate, i.e., λ̄ = 0, and only the harvested power, i.e., λ̄ → ∞. that the sum-rate and harvested power increase with fmin , with
The proposed bi-objective optimization problem allows us to fmin = 1 corresponding to the ideal case for an RIS.
obtain the desired trade-off between sum-rate and harvested In Fig. 5, finally, we investigate the trade-off between the
power by varying λ̄, and it can be utilized to optimize only sum-rate and harvested power as a function of the number of
the sum-rate or the harvested power as two special cases. UEs. We see that the Pareto region is characterized by a larger

Authorized licensed use limited to: M.KUMARSAMY COLLEGE OF ENGINEERING. Downloaded on February 14,2023 at 05:55:06 UTC from IEEE Xplore. Restrictions apply.
 MOHAMED et al.: BI-OBJECTIVE OPTIMIZATION OF INFORMATION RATE AND HARVESTED POWER 2199

obtained by deploying a sufficient number of RIS elements.

Compared with the “No RIS” case, we see, e.g., that the har-
vested power increases from 0.11 mW to 0.19 mW while
guaranteeing the same sum-rate of 29 bps/Hz if K = 4 and
N = 400. The sum-rate and the harvested power can be further
increased by deploying larger RISs.

V. C ONCLUSION
We have investigated the trade-off between the sum-rate
and harvested power in a multi-user RIS-aided downlink
MISO system with SWIPT. Enforcing QoS constraints and
practical phase shift constraints, the beamforming vector, the
power splitting ratio, and the RIS reflection coefficients are
Fig. 3. Trade-off between the sum-rate and harvested power with the number jointly optimized by a two-layer penalty-based algorithm.
of RIS elements (K = 4).
Simulation results show that the proposed algorithm signifi-
cantly outperforms conventional approaches in the absence of
RISs.

R EFERENCES
[1] M. Di Renzo et al., “Smart radio environments empowered by reconfig-
urable intelligent surfaces: How it works, state of research, and the road
ahead,” IEEE J. Sel. Areas Commun., vol. 38, no. 11, pp. 2450–2525,
Nov. 2020.
[2] Q. Wu and R. Zhang, “Towards smart and reconfigurable environment:
Intelligent reflecting surface aided wireless network,” IEEE Commun.
Mag., vol. 58, no. 1, pp. 106–112, Jan. 2020.
[3] C. Huang et al., “Holographic MIMO surfaces for 6G wireless networks:
Opportunities, challenges, and trends,” IEEE Wireless Commun., vol. 27,
no. 5, pp. 118–125, Oct. 2020.
[4] Q. Wu and R. Zhang, “Joint active and passive beamforming
optimization for intelligent reflecting surface assisted SWIPT under QoS
Fig. 4. Trade-off between the sum-rate and harvested power with fmin constraints,” IEEE J. Sel. Areas Commun., vol. 38, no. 8, pp. 1735–1748,
(N = 300). Aug. 2020.
[5] Z. Li, W. Chen, Q. Wu, K. Wang, and J. Li, “Joint beamform-
ing design and power splitting optimization in IRS-assisted SWIPT
NOMA networks,” IEEE Trans. Wireless Commun., vol. 21, no. 3,
pp. 2019–2033, Mar. 2022.
[6] C. Pan et al., “Intelligent reflecting surface aided MIMO broadcasting
for simultaneous wireless information and power transfer,” IEEE J. Sel.
Areas Commun., vol. 38, no. 8, pp. 1719–1734, Aug. 2020.
[7] D. Xu, V. Jamali, X. Yu, D. W. K. Ng, and R. Schober, “Optimal
resource allocation design for large IRS-assisted SWIPT systems: A
scalable optimization framework,” IEEE Trans. Commun, vol. 70, no. 2,
pp. 1423–1441, Feb. 2022.
[8] A. Khalili, S. Zargari, Q. Wu, D. W. K. Ng, and R. Zhang, “Multi-
objective resource allocation for IRS-aided SWIPT,” IEEE Commun.
Lett., vol. 10, no. 6, pp. 1324–1328, Jun. 2021.
[9] C. He, X. Xie, K. Yang, and Z. J. Wang, “A joint power splitting,
active and passive beamforming optimization framework for IRS assisted
MIMO SWIPT system,” 2021, arXiv:2105.14545.
[10] S. Abeywickrama, R. Zhang, Q. Wu, and C. Yuen, “Intelligent reflecting
surface: Practical phase shift model and beamforming optimization,”
Fig. 5. Trade-off between the sum-rate and harvested power with the number IEEE Trans. Commun, vol. 68, no. 9, pp. 5849–5863, Sep. 2020.
of UEs (N = 400). [11] J. Tang et al., “Joint power allocation and splitting control for SWIPT-
enabled NOMA systems,” IEEE Trans. Wireless Commun., vol. 19, no. 1,
pp. 120–133, Jan. 2020.
[12] K. Shen and W. Yu, “Fractional programming for communication
excursion of the harvested power as the priority parameter λ̄ systems—Part II: Uplink scheduling via matching,” IEEE Trans. Signal
varies, while the sum-rate shows a less significant variation. Process., vol. 66, no. 10, pp. 2631–2644, May 2018.
[13] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge, U.K.:
This is due to the stringent requirement that we imposed on Cambridge Univ. Press, 2004.
the minimum harvested power, which forces the optimization [14] C. Pan, H. Ren, M. Elkashlan, A. Nallanathan, and L. Hanzo, “Robust
algorithm to allocate more resources for power harvesting, beamforming design for ultra-dense user-centric C-RAN in the face of
which inevitably leads to a lower sum-rate. Higher values realistic pilot contamination and limited feedback,” IEEE Trans. Wireless
Commun., vol. 18, no. 2, pp. 780–795, Feb. 2019.
of sum-rate can be obtained by reducing the value of the [15] A. Ben-Tal and A. Nemirovski, Lectures on Modern Convex
minimum acceptable harvested power. Nevertheless, satisfac- Optimization: Analysis, Algorithms, Engineering Applications (MPS-
tory values of both the sum-rate and harvested power can be SIAM Series on Optimization). Philadelphia, PA, USA: SIAM, 2001.

Authorized licensed use limited to: M.KUMARSAMY COLLEGE OF ENGINEERING. Downloaded on February 14,2023 at 05:55:06 UTC from IEEE Xplore. Restrictions apply.