Digital Communications and Networks 8 (2022) 303–313

Contents lists available at ScienceDirect

Digital Communications and Networks

journal homepage: www.keaipublishing.com/dcan

Simultaneous wireless information and power transfer with fixed and

adaptive modulation
Jie Hu a, Guangming Liang b, Qin Yu a, *, Kun Yang b, Xiaofeng Lu c
a
Yangtze Delta Region Institute (Huzhou) & School of Information and Communication Engineering, University of Electronic Science and Technology of China, Huzhou,
313001, China
b
School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu, 611731, China
c
State Key Laboratory of Integrated Services Networks, Xidian University, Xi'an, 710071, China

A R T I C L E I N F O A B S T R A C T

Keywords: Activating Wireless Power Transfer (WPT) in Radio-Frequency (RF) to provide on-demand energy supply to
Simultaneous wireless information and power widely deployed Internet of Everything devices is a key to the next-generation energy self-sustainable 6G network.
transfer (SWIPT) However, Simultaneous Wireless Information and Power Transfer (SWIPT) in the same RF bands is challenging.
Fixed/adaptive modulation
The majority of previous studies compared SWIPT performance to Gaussian signaling with an infinite alphabet,
Rate-energy-reliability trade-off
Transceiver design
which is impossible to implement in any realistic communication system. In contrast, we study the SWIPT system
Energy self-sustainability in a well-known Nakagami-m wireless fading channel using practical modulation techniques with finite alphabet.
The attainable rate-energy-reliability tradeoff and the corresponding rationale are revealed for fixed modulation
schemes. Furthermore, an adaptive modulation-based transceiver is provided for further expanding the attainable
rate-energy-reliability region based on various SWIPT performances of different modulation schemes. The
modulation switching thresholds and transmit power allocation at the SWIPT transmitter and the power splitting
ratios at the SWIPT receiver are jointly optimized to maximize the attainable spectrum efficiency of wireless
information transfer while satisfying the WPT requirement and the instantaneous and average BER constraints.
Numerical results demonstrate the SWIPT performance of various fixed modulation schemes in different fading
conditions. The advantage of the adaptive modulation-based SWIPT transceiver is validated.

1. Introduction battery or batteryless powered IoE devices. Radio Frequency (RF) signal-
based WPT is more suitable in the scenario of IoE than inductive coupling
1.1. Background [3] and magnetic resonance [4] based methods for the following reasons:

Internet of Everything (IoE) is one of the key 6G applications for smart
IoE devices are deployed in the far field of transmitters. Inductive
cities [1]. Many IoE devices for sensing and communications are being coupling and magnetic resonance based WPT can only deliver energy
massively deployed around every corner. However, the limited battery to devices centimeters away from transmitters. However, by radiating
capacity of these miniature devices substantially constrains their lifetime, RF signals, wireless power can be delivered more than tens of meters
which may heavily impair network functions and substantially increase the away.
maintenance cost for replacing drained batteries. Harvesting ambient en-
IoE devices are extensively deployed in a certain region. RF signals
ergy from the environment may relieve IoE devices from the constraints of transmitted by multiple antennas may flexibly form either narrow
limited battery capacity [2]. However, it primarily relies upon the unpre- beams for high-energy point-to-point WPT, or wide beams for wire-
dictable environment. The intermittent energy supply cannot ensure the less power multicast toward multiple devices. However, inductive
Quality of Service (QoS) of the sensing and communications of IoE devices. coupling and magnetic resonance are limited for point-to-point WPT.
Consequently, active Wireless Power Transfer (WPT) is needed for
There are many RF transmitters such as cellular Base Stations (BSs),
providing controllable and on-demand energy supply to a rechargeable WiFi Access Points (APs), and TV towers. By exploiting legacy

* Corresponding author.
E-mail addresses: hujie@uestc.edu.cn (J. Hu), lianggm@std.uestc.edu.cn (G. Liang), yuqin@uestc.edu.cn (Q. Yu), kunyang@uestc.edu.cn (K. Yang), luxf@xidian.
edu.cn (X. Lu).

https://doi.org/10.1016/j.dcan.2022.01.001
Received 20 January 2021; Received in revised form 29 December 2021; Accepted 12 January 2022
Available online 19 January 2022
2352-8648/© 2022 Chongqing University of Posts and Telecommunications. Publishing Services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

infrastructure, we do not require dedicated equipment for providing non-Gaussian distributed RF signals. Some pioneering works have
WPT services. initially discussed the influence of practical modulation schemes on the
SWIPT performance [16–21]. For example, the rate-energy tradeoff was
Coordinating WPT services along with conventional Wireless Infor- first discussed in Ref. [16] by considering discrete modulated symbols.
mation Transfer (WIT) services results in many challenges in the design The optimal precoder was optimally designed by Zhu et al. for SWIPT
of the physical layer, medium-access-control layer, and networking layer, with practical modulation schemes in both [17,18], where the WIT
which stimulate substantial interest in relevant research of Simultaneous performance was evaluated by the mutual information of
Wireless Information and Power Transfer (SWIPT) and Integrated Data discrete-input-continuous-output channels. Hu et al. [19] delivered a
and Energy Networks (IDENs) [5]. Moreover, the future 6G network is comprehensive guideline for modulation and coding design for SWIPT.
envisioned to be energetically self-sustainable [6]. WPT will be consid- Then, asymmetric phase-shift-keying was proposed in Ref. [20] for
ered a new service class as important as communication, computing, and achieving the optimal SWIPT performance. Moreover, Zhao et al. [22]
sensing, while infrastructure and IoE devices in energy self-sustainable studied the SWIPT performance by invoking a receive spatial modula-
networks will operate in a super energy-efficient manner. Integrating tion, which substantially reduced the demodulation complexity at
WPT services in future 6G networks requires a holistic system design in low-power IoE devices. Furthermore, a constellation rotation-aided
every layer. Our study makes the first attempt to integrate both WIT and modulation design was proposed for a non-orthogonal-multiple-access
WPT services by designing practically modulated RF signals. aided downlink SWIPT system in Ref. [21]. However, none of the
Although RF signals are capable of carrying wireless power to far-field above mentioned references revealed the rationale behind the SWIPT
IoE devices, the WPT still suffers from low efficiency, which results in a service delivery with practical modulation schemes. They all ignored the
limited transmission range (15 m) [5]. Therefore, SWIPT should be rate-energy-reliability tradeoff in the modulation design for SWIPT. To
exploited in some typical scenarios and integrated with other potential fill this gap, our previous work [23] originally investigated the
techniques. For example, it can be exploited for powering miniature IoE rate-energy-reliability tradeoff by jointly optimizing a fixed
devices in some indoor scenarios, such as sensors in smart homes and modulation-based SWIPT transceiver.
smart factories. Therefore, SWIPT can be integrated with WiFi access Moreover, different modulation schemes have different Bit-Error-
points and indoor base stations for broadcasting downlink information Ratio (BER) and Spectrum Efficiency (SE) performances. As a result,
and remotely charging these battery-powered and batteryless devices adaptively choosing a modulation scheme in a certain channel condition
[7]. Moreover, SWIPT can also be integrated with Unmanned Aerial may achieve a better BER/SE performance overall. This is well known as
Vehicles (UAVs) to simultaneously collect data from IoE devices an adaptive modulation [24]. According to Refs. [19,21], various WPT
deployed in wide areas for environment monitoring, as UAVs may flex- performances can be achieved by different modulation schemes in
ibly reduce their distance to batteryless devices to increase the efficiency various channel conditions when non-linear energy harvesters are
of WPT [8]. considered. However, none of the existing works consider the adaptive
modulation for SWIPT.
1.2. Related works
1.3. Novel contributions
By assuming that the transmit signal is also Gaussian distributed, the
classic Shannon-Hartely capacity was calculated for characterizing the Against this background, our novel contributions are summarized as
performance limit of WIT services in Additive-White-Gaussian-Noise follows:
(AWGN) channels. It is a common assumption in SWIPT [9–15].
Clerckx et al. [9] investigated a multi-sinusoidal based waveform
We study a SWIPT system. The SWIPT transmitter conceives a prac-
design for SWIPT at the physical layer, which could be easily integrated tical digital modulator by adopting either a fixed or adaptive modu-
with traditional Orthogonal-Frequency-Division-Multiplexing (OFDM)-- lation scheme, and a buffer for storing the modulated symbols. With
based waveforms. Buckley et al. [10] exploited the cyclic prefix for the aid of the buffer, reliable WIT can be guaranteed. The SWIPT
dedicated WPT, while the other OFDM symbols were invoked for WIT. receiver is batteryless. Its information demodulation is powered by
The Shannon-Hartely capacity was determined to evaluate the WIT the energy harvested from the downlink WPT service. It also harvests
performance of both [9,10]. Dai et al. [11] studied a millimeter-wave additional energy to power its other functions.
massive MIMO-based Non-Orthogonal-Multiple-Access (NOMA) system
We reveal the rationale behind the SWIPT service delivery with
for SWIPT. Hybrid precoding was optimally designed by considering practical modulation schemes. We characterize the influence of
Gaussian signalling. Yue et al. [12] investigated the optimal transceiver different modulation schemes on the SWIPT performance in a
design for simultaneous wireless information and power multicast in a Nakagami-m fading channel while evaluating the rate-energy-
multi-user millimeter-wave MIMO system by considering a limited reliability tradeoff in various wireless fading conditions.
number of RF chains and limited resolution of analog phase shifters. By
For a fixed modulation-based SWIPT transceiver, we jointly optimize
assuming Gaussian signaling, the wireless information multicast perfor- the transmission power and the transmission switching threshold at
mance was measured as the minimum spectral efficiency across all in- the SWIPT transmitter and the power splitting ratio at the SWIPT
formation users. receiver to maximize the spectrum efficiency of the WIT service while
In the medium-access-control layer, Lv et al. [13] optimized downlink satisfying the receiver's WIT reliability and WPT requirements.
and uplink resources in the time-domain for supporting multiple bat-
For an adaptive modulation-based SWIPT transceiver, we jointly
teryless devices in another energy-self-sustainable communication sys- optimize the modulation switching thresholds at the SWIPT trans-
tem, where both the sum-throughput and the fair-throughput were mitter and the power splitting ratio at the SWIPT receiver to maxi-
evaluated by assuming Gaussian signalling. Furthermore, Zhao et al. [14] mize the average spectrum efficiency of the WIT while satisfying the
proposed an enhanced carrier-sensing-multiple-access protocol for a receiver's instantaneous and average WIT reliability, as well as the
SWIPT-based wireless local area network, and they studied an optimal energy-harvesting requirement.
deployment of a hybrid access point for delivering SWIPT services in
Ref. [15]. The rest of the paper is organized as follows. Our system model is
Unfortunately, Gaussian-distributed signals cannot be actually introduced in Section 2, while a fixed modulation-based SWIPT trans-
generated, when a practical modulation scheme possessing finite alpha- ceiver is optimized in Section 3, followed by the optimization of an
bet is adopted. Therefore, we investigated the SWIPT performance of adaptive modulation-based SWIPT transceiver in Section 4. After

304
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

providing numerical results in Section 5, we finally provide the conclu- shown in Fig. 1(a), the Quadrature Amplitude Modulation (QAM) with an
sion in Section 6. order of four (4-QAM) has a lower spectrum efficiency than 16-QAM.
However, 4-QAM is much more reliable than 16-QAM. As shown in
2. System model Fig. 1(b), 16-QAM, 16-Phase-Shift-Keying (16-PSK), and 16-Pulse-Ampli-
tude-Modulation (16-PAM) have the same order. However, 16-QAM is
In this section, we establish the influence of different modulation the most reliable modulation scheme.
schemes on the SWIPT performance followed by the description of the Furthermore, different modulation schemes have diverse WPT per-
transceiver architecture. formances. As shown in Fig. 2, a SWIPT receiver has an information
demodulator and an energy harvester. However, the energy harvested
has to power the information demodulator. We assume that the minimum
2.1. Modulation for SWIPT energy required by all functional modules of the SWIPT receiver is Pth.
This indicates that only when the total energy harvested is higher than
Modulation is essential for practical communication systems, which Pth, the battery can be effectively recharged. The WPT capability of a
modulates information RF signal amplitudes, phases, and frequencies. In modulation scheme depends on how many modulated symbols carrying
this study, we mainly focus on amplitude- and phase-based modulation power are higher than Pth. As shown in Fig. 1(a), when we have a lower
schemes. As shown in Fig. 1, the WIT capability of a modulation scheme threshold Pth,1, the WPT performance of 4-QAM is higher than that of 16-
is normally described by its constellation in an in-phase-quadrature co- QAM. However, as we increase the threshold to Pth,3, 4-QAM is out-
ordinate. Nodes in a constellation represent all possible modulated performed by 16-QAM in terms of WPT performance. Furthermore, as
symbols. The square of the Euclidean distance from the origin to a spe- shown in Fig. 1(b), given a common threshold to Pth, 16-PAM has the
cific node represents the power carried by the corresponding modulated most modulated symbols carrying higher power than Pth. Therefore, it
symbols. Having more nodes in a constellation represents that a single has the highest WPT performance. All the modulated symbols of 16-PSK
modulated symbol may carry more information bits, which results in carry lower power than Pth. As a result, they cannot be exploited for WPT.
higher spectrum efficiency. Furthermore, the minimum Euclidean dis- Since different modulation schemes have various SWIPT perfor-
tance between all pairs of nodes in a constellation represents the reli- mances, we have to design the SWIPT transceiver by considering these
ability of a modulation scheme. A higher minimum Euclidean distance of distinctive characteristics.
a specific modulation scheme relates to higher reliability. Fig. 1 outlines
several classic modulation schemes, where we assume that all modulated
symbols have equal probabilities to be sent, and the average power of all 2.2. Transceiver architecture
the modulated symbols is the same for different modulation schemes. As
Our SWIPT transceiver is illustrated in Fig. 2. The following modules
constitute the SWIPT transmitter equipped with a single antenna:

Information Source generates information bits requested by the SWIPT

receiver.

The buffer stores the information bits that need to be transmitted in
the wireless medium. The information bits may output to the digital
modulator when a certain channel condition can be achieved to
guarantee the WIT's Quality of Service (QoS). Otherwise, they are
pending to be transmitted in the buffer.

Digital Modulator modulates information bits onto amplitudes and
phases of RF signals by adopting either a fixed modulation scheme or
an adaptive one. For a fixed modulation scheme, the digital modu-
lator only operates when a certain channel condition satisfies the
basic Bit-to-Error-Ratio (BER) requirement. For an adaptive modula-
tion scheme, the digital modulator may adaptively switch among
different modulation schemes to increase the attainable spectrum
efficiency while satisfying both the BER requirement on the WIT and
the energy harvesting requirement on the WPT.

Baseband to Passband converts the modulated baseband signals to
passband to transmit in wireless fading channels.

We consider an uncorrelated block fading wireless channel. The channel

Fig. 1. Influence of different modulation schemes on SWIPT performance. Fig. 2. SWIPT transceiver architecture.

305
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

amplitude coefficient, h, does not change within a symbol duration, but it represent the statistical expectation on a random variable and the abso-
varies from one symbol duration to another. We assume that the channel lute value of a complex number, respectively. We assume that the
amplitude coefficient, h, follows a Nakagami-m distribution because it is average transmit power of the SWIPT transmitter is Pt. All modulated
feasible for modelling various fading conditions. Therefore, the corre- symbols in a constellation of a specific modulation scheme constitute a
sponding channel power coefficient v ¼ |h|2 follows a Gamma distribution set X . The i-th modulated symbol xi 2 X has a probability of pi to be sent
with parameter m, whose probability density function (pdf) is given by by the SWIPT transmitter, while its transmit power is Pi. Both pi and Pi
satisfy the following equalities:
vm1 mm expðmvÞ
f ðvÞ ¼ (1) XM XM
ΓðmÞ pi ¼ 1; pi Pi ¼ Pt (2)
i¼1 i¼1

for 8m ¼ 1, 2, ⋯ [25]. In Eq. (1), m represents the channel fading factor. Moreover, by assuming pi ¼ 1
M
for 8i ¼ 1, …, M, we have
When m ¼ 1, it is equivalent to a Rayleigh fading channel without any
Line of Sight (LoS). A higher m indicates a less fading channel with a 1 XM
Pi ¼ Pt (3)
stronger LoS. For the channel attenuation incurred by the path loss, we M i¼1

have the distance, d, between the SWIPT transmitter and receiver while In Eqs. (2) and (3), M ¼ jXj represents the total number of possible
the path loss exponent is λ. The reference distance for path loss is d0 ¼ 1 modulated symbols in the set X , which is the order of a specific modu-
m. lation scheme.
The SWIPT receiver comprises the following modules: We assume that the power splitting ratio of the SWIPT receiver is ρPS,
pffiffiffiffiffiffiffi
which means that a portion ρPS of the received RF signal is used for

Signal Splitter splits the received RF signals into two portions in the pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
energy harvesting, and a portion 1  ρPS is used for information
power domain. The first portion is used for energy harvesting and the
demodulation. In the k-th symbol duration, the received baseband signal
second portion is used for information demodulation.
for information demodulation is expressed as

RF-DC Converter converts the alternative current carried by a portion
of RF signal to the Direct Current (DC), which can be used for rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1  ρPS ÞPt
recharging batteries and powering information demodulators. yID ½k ¼ h½k x½k þ nID ½k (4)
dλ

Battery is recharged by the DC, which completes the energy harvest-
ing process. The actual energy harvested in the battery should exclude where nID ½k  N ð0; σ 2ID Þ is the additive Gaussian distributed noise. For
energy consumed for powering other functional modules. convenience, we will omit the index k in the rest of the derivations.

Passband to Baseband converts a portion of received passband RF We focus on the spectrum efficiency and BER performance of the WIT.
signal to the baseband for processing in the digital domain. As we consider a modulation scheme having an order of M, we cannot

Digital Demodulator demodulates the baseband signal for extracting exploit the classic Shannon-Hartley's capacity for characterizing the
the information. spectrum efficiency of the WIT. This is because Shannon-Hartley capacity

Information Destination finally receives the demodulated information. is derived by conceiving Gaussian distributed signals having infinite al-
phabets. Since the input signal is finite and discrete, we turn to the
We assume that the SWIPT transmitter has sufficient knowledge of mutual information of discrete-input-continuous-output channels for
channel state information. characterizing the spectrum efficiency of the WIT [26]. The mutual in-
formation between the discrete finite channel input, X, and the contin-
3. Transceiver optimization for fixed modulation uous infinite channel output, YID, can be expressed as
Z þ∞
In this section, we firstly analyze the spectrum efficiency, BER, and P pðyID jxÞ
IðX; YID Þ ¼ pðyID jxÞpðxÞlog X 0 0 dyID
energy harvesting performance of a fixed modulation-based SWIPT x2X ∞ pðyID jx Þpðx Þ (5)
0
transceiver. Then, an optimal transceiver design is obtained for the sake x 2X

of achieving a balance among these conflicted performance metrics. To

guarantee the reliability requirement, the fixed modulation-based SWIPT where we have the conditional pdf p(yID|x) expressed as
transmitter adopts the following transmission strategy: !
1 ðyID  xÞ2
pðyID jxÞ ¼ pffiffiffiffiffiffiffiffiffiffiffi
ffi exp  (6)

When instantaneous Signal-to-Noise-Ratio (SNR) γ is higher than 2πσ ID 2 2σ 2ID
threshold γ 0, the SWIPT transmitter sends the modulated symbols to
the SWIPT receiver to simultaneously satisfy its information and en- for 8x 2 X. According to Ref. [27], the mutual information I(X; YID) can
ergy requests. be maximized when pðxÞ ¼ M1 for 8x 2 X . However, by substituting

When the instantaneous SNR, γ, is lower than the threshold, γ 0, the pðxÞ ¼ M1 and p(yID|x) of Eq. (6) into Eq. (5), we cannot obtain the
SWIPT transmitter does not send any modulated symbols to the closed-form mutual information I(X; YID) due to the existence of the
SWIPT receiver. Therefore, in these instances, the SWIPT system is in integration. Fortunately, a tight upper-bound for I(X; YID) has been found
the outage. The SWIPT receiver can neither receive any information by Baccarelli et al. [28], which is expressed as
nor harvest any energy. " !#
1 X γ X 0 2
CM ðγÞ ¼ log2 M   log 1 þ ðM  1Þexp  jx  x j
Therefore, we can guarantee that either the instantaneous BER or the M x2X 2 M 1 0
x 2X
average BER is lower than the corresponding requirement. Moreover,
(7)
γ 0 is defined as the transmission switching threshold.
In Eq. (7), the SNR, γ, of the information demodulator is formulated as

3.1. WIT performance vð1  ρPS ÞPt

γ¼ (8)
d λ σ 2ID
At the SWIPT transmitter, the modulated symbol is denoted as x(t),
Based on Eqs. (1) and (8), the pdf of the SNR, γ, is obtained as
having a zero mean and a variance of E[|x(t)|2] ¼ 1, where E[⋅] and |⋅|

306
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

ðγ=aÞm1 mm expðmγ=aÞ As the power of nEH is negligibly small, we do not consider it in Eq.
f ðγÞ ¼ (9) (17). According to Ref. [30], the transmit power, Pi, carried by a specific
aΓðmÞ
modulated symbol is expressed as
ð1ρPS ÞPt
where m is a positive integer, a ¼ dλ σ 2ID
. Therefore, the average spec- 8    2
>
> 3Pt  M þ 1
trum efficiency can be formulated as >
>
> 2 3⌈i  ⌉  1
 M  PAM
2
>M  1
>
>
Z ∞
>
>
>
> P M  PSK
~M ¼
C CM ðγÞf ðγÞ dγ (10) >
>
<
t
" 
0
pffiffiffiffiffi  2
Pi ¼ 3Pt  i
 M  1  (18)
>
> 2⌈ ⌈
 M p ffiffiffiffi
ffi ⌉  ⌉
The actual WIT spectrum efficiency by conceiving the above trans- >
> 2ðM  1Þ 2
>
>
mission strategy is reformulated as >
>  pffiffiffiffiffi 2 #
>
> pffiffiffiffiffi
>
> M 1
: þ ⌈modði; M Þ  ⌉1 M  QAM
Z ∞ >
>
2
CM ¼ CM f ðγÞ dγ (11)
γ0

Apart from spectrum efficiency, we should also evaluate the BER for 8i ¼ 1, …, M, where ⌈⋅⌉ finds the minimum integer higher than the
performance, when a specific modulation scheme is adopted. According input parameter and mod(⋅, ⋅) represents operation of the first input
to Ref. [29], the instantaneous BER with respect to the SNR, γ, is parameter modulo the second. With the aid of Eqs. (17) and (18), and by
formulated as considering the transmission strategy of the SWIPT transmitter, the
average energy harvested by the SWIPT receiver is formulated as
8 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
> 2ðM  1Þ 6γ Z ∞
>
>
> Q for M  PAM 1 XM
>
> Mlog 2 M M2  1 PEH;M ¼ PEH ðxi ; γÞf ðγÞ dγ (19)
>
> M i¼1
γ0
>
>
< 2 pffiffiffi  
BERM ðγÞ ¼ Q γ sin π M for M  PSK (12) which is derived by assuming equal probabilities for all the modulated
>
> log 2 M
>
>
>
> rffiffiffiffiffiffiffiffiffiffiffiffiffi ! symbols, namely, pi ¼ M1 for 8i.
>
> 4 3γ
>
> Q for M  QAM
: log M M1
2
3.3. Transceiver optimization

where the classic Q-function is expressed as

To achieve a balance among the WIT spectrum efficiency, CPS , of Eq.
Z þ∞ (11), the average BER, B ERM , of Eq. (14), and the average energy har-
expð0:5t2 Þ
QðxÞ ¼ pffiffiffiffiffi dt (13)
x 2π vested, PEH;M , of Eq. (19), the SWIPT transceiver optimization problem is
formulated as
As a result, the average BER over all possible values of random SNR γ
is formulated as ðP1Þ: max CM (20)
ρPS ;Pt ;γ 0
R∞
γ0
BERM ðγÞf ðγÞ dγ
B ERM ¼ R∞ (14)
γ0
f ðγÞ dγ s: t: PEH;M  PEH;0 (20a)

B ERM  BER0 (20b)

3.2. WPT performance

Let Pr represent the power of the RF signal for energy harvesting. We 0  Pt  Pave (20c)
consider a linear RF-DC converter, where a constant η represents the
linear RF-DC energy conversion efficiency. Without any loss of general- 0  ρPS  1 (20d)
ity, we assume normalized symbol duration. Hence, by considering the
This problem maximizes the effective spectrum efficiency CM by
minimum energy Pth required for information demodulation, the effec-
optimizing the power splitting ratio, ρPS, the transmit power, Pt, of the
tive energy harvested can be expressed as
SWIPT transmitter, and the transmission switching threshold, γ 0. More-
PEH ¼ ηðPr  Pth Þþ (15) over, the constraint (20a) represents that the average energy harvested,
PEH;M , should be higher than a threshold PEH,0; however, the constraint
where (⋅)þ represents the operation of taking the maximum value be- (20b) represents that the average BER, B ER, should be lower than a
tween 0 and the input parameter. minimum requirement BER0. Furthermore, as expressed in the constraint
After being split by the signal splitter in the power domain, as por- (20c), the average transmit power, Pt, should be lower than a threshold
trayed in Fig. 2, the RF signal for energy harvesting is expressed as Pave, while the power splitting ratio, ρPS, is a real number between 0 and
rffiffiffiffiffiffiffiffiffiffiffi 1.
ρ Pt By substituting Eq. (14) into the constraint (20b), we obtain a func-
yEH ¼ h PSλ x þ nEH (16)
d tion φ(γ 0) with respect to γ 0 as

where nEH is the remaining noise after the signal splitter. Z ∞ Z ∞

By assuming that a modulated symbol xi 2 X is sent by the SWIPT φðγ 0 Þ ¼ BERðγÞf ðγÞ dγ  BER0 f ðγÞ dγ  0 (21)
γ0 γ0
transmitter according to Eqs. (8), (15), and (16), the energy harvested by
the SWIPT receiver can be formulated as whose derivative is expressed as
 þ
ρPS Pi γ σ
2
dφðγ 0 Þ
PEH ðxi ; γÞ ¼ η ID
 Pth (17) ¼ ½BERðγ 0 Þ  BER0 φðγ 0 Þ (22)
ð1  ρPS ÞPt dγ 0

307
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

0
By letting BER0 ¼ BER(γ 0), we may find the corresponding γ 0 ¼ γ 0 . mance measures, an optimal transceiver design is created. The adaptive
According to Eq. (12), BER(γ 0) is a decreasing function with respect to its modulation-based SWIPT transmitter uses the following transmission
input parameter γ 0. Therefore, by observing the derivative (22), when approach to increase the attainable spectrum efficiency:
0
γ 0 2 ð0; γ 0 Þ, φ(γ 0) is an increasing function, whereas it is a decreasing
0 0
function if γ 0 2 ½γ 0 ; þ ∞Þ. Hence, if we let γ 0 ¼ γ 0 , φ(γ 0) achieves its
When the instantaneous SNR, γ, is lower than the threshold, γ 0, the
maximum value 0. If we require the instantaneous BER, BER(γ), lower SWIPT transmitter does not send any modulated symbols to the
0 SWIPT receiver. Therefore, the SWIPT system is in the outage state in
than BER0 for 8γ, we choose the transmission threshold, γ 0, to be γ 0 .
these instances, as portrayed in Fig. 3. The SWIPT receiver can neither
However, in (P1), we consider the average BER constraint, as shown in
receive any information nor harvest any energy.
(20b), and the optimal switching threshold γ *0 should be within the region
0

According to their various SWIPT performances, the SWIPT trans-
γ 0 2 ð0; γ 0 Þ, which can be found by a one-dimensional search. mitter is capable of adaptively switching among QAM family of \{4-
According to Eqs. (7) and (17)-(19), the average spectrum efficiencies, QAM, 16-QAM, 64-QAM, 256-QAM, ⋯ \} without any loss of gener-
CM and PEH;M , are both increasing functions with respect to the average ality. As illustrated in Fig. 3, when the effective SNR for information
transmit power, Pt, of the SWIPT transmitter. Moreover, PEH;M is an demodulation falls in a specific region, the digital modulator of the
increasing function with respect to ρPS, while CM is a decreasing function SWIPT transmitter switches to a certain modulation scheme. For
with respect to ρPS. Therefore, based on the above analysis, letting Pt ¼ Pave instance, M-QAM (simply represented by its order M) is adopted by
the digital modulator when the effective SNR is within a region of
and PEH;M ¼ PEH;0 may return us the optimal transmit power P*t ¼ Pave and
½γ ðlog 2 MÞ=21 ; γ ðlog 2 MÞ=2  for 8M 2 f4; 16; 64; 256; …; Mmax g ≜ M.
the optimal power splitting ratio ρ*PS . Substituting P*t ¼ Pave and ρ*PS into

Since the modulation schemes available for the digital modulator are
(P1), we may obtain the optimal transmission switching threshold γ *0 for
finite, the highest order of the QAM is Mmax. When the effective SNR,
maximizing the average spectrum efficiency. The pseudo code for solving
γ, is higher than γ ðlog 2 Mmax Þ=21 , the Mmax-QAM is always adopted for
(P1) is detailed in Algorithm 1, whose complexity is Oðlog 2 ðnÞÞ.
modulating the information bits, which means γ ðlog 2 Mmax Þ=2 ¼ þ ∞.
Algorithm 1. Fixed modulation based SWIPT transceiver design with
an average BER constraint
4.1. WIT performance

When M-QAM is selected at the SWIPT transmitter and the power

splitter with a ratio ρPS is adopted at the SWIPT receiver, the attainable
spectrum efficiency, CM, can be then evaluated by Eq. (7). However, this
evaluation is only valid when the effective SNR γ satisfies γ 2
½γ ðlog 2 MÞ=21 ; γ ðlog 2 MÞ=2 . Hence, the average spectrum efficiency can be
formulated as
XZ γ ðlog
2 MÞ=2
^¼
C CM f ðγÞdγ (23)
γ ðlog
M2M 2 MÞ=21

where f(γ) is the PDF of the effective SNR, γ, expressed as Eq. (9).
Apart from the average spectrum efficiency, the reliability of the WIT
with adaptive modulation needs to be characterized. Given an instanta-
neous effective SNR γ, the BER BERM(γ) of M-QAM can be calculated by
the third line of Eq. (12). Therefore, by considering the adaptive modu-
lation, the average BER can be formulated as
P R γðlog
M2M γ ðlog
2 MÞ=2 BERM ðγÞf ðγÞdγ
^ ¼ 2 MÞ=21
BER R γ0 (24)
1 0
f ðγÞdγ

The denominator in Eq. (24) ensures that the outage state is not
included in the calculation of the average BER.

4.2. WPT performance

As different modulation schemes are selected within different effective

SNR regions, the energy harvesting performance of the SWIPT receiver
varies. When the effective SNR, γ, falls in the region of ½γ ðlog 2 MÞ=21 ;
γ ðlog 2 MÞ=2 , M-QAM is adopted by the SWIPT transmitter, and the average
amount of energy harvested by the SWIPT receiver is formulated as
4. Transceiver optimization for adaptive modulation
XM Z γðlog2 MÞ=2
^ EH;M ¼ 1
P PEH ðxi ; γÞf ðγÞdγ (25)
The spectrum efficiency, BER, and energy-harvesting performance of M i¼1
γ ðlog MÞ=21
2
an adaptive modulation-based SWIPT transceiver will be formulated in
this section. Then, to achieve a balance among these conflicting perfor- where PEH(xi, γ) given by Eq. (17) represents the amount of energy

308
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

harvested, when the symbol xi(8i ¼ 1, …, M) is transmitted by the SWIPT performance becomes poorer, given a specific instantaneous effective
^ EH;M is obtained by assuming that all the modu-
transmitter. Note that P SNR γ. Similarly, we may get the boundary threshold, γ yðlog 2 MÞ=2 , for the
lated symbols have equal probabilities to be transmitted. Moreover, the
(M þ 1)-QAM. If the instantaneous effective SNR γ exceeds γ yðlog 2 MÞ=2 , the
total average amount of energy harvested by the SWIPT receiver is
(M þ 1)-QAM satisfies the instantaneous BER constraint, BER0. More-
expressed as
over, the (M þ 1)-QAM has a higher spectrum efficiency than M-QAM.
X X 1 XM Z γðlog2 MÞ=2 The digital modulator may adaptively select (M þ 1)-QAM for maxi-
^ EH ¼
P ^ EH;M ¼
P PEH ðxi ; γÞf ðγÞdγ (26)
M i¼1
γ ðlog MÞ=21
mizing the average spectrum efficiency of Eq. (28). Therefore, a range of
M2M M2M
feasible boundary thresholds γy can be decided, which characterize the
2

operational SNR region of the QAM schemes.

4.3. Transceiver optimization Based on a similar analysis of the SWIPT transceiver design for the fixed
modulation, we may obtain the optimal average transmit power as Pyt ¼
According to the transmission strategy of the adaptive modulation-
Pave . A feasible power splitting ratio ρyPS is derived by solving the equation
based SWIPT transmitter, the modulation switching thresholds can be
^EH ¼ PEH;0 . Therefore, (P2) with a instantaneous BER constraint can be
of P
represented by the following vector:
solved by a bisection method, as detailed in Algorithm 2.
γ ¼ γ 0 ; γ 1 ; …; γ ðlog2 MÞ=21 ; …; γ ðlog2 Mmax Þ=21 (27) Algorithm 2. Adaptive modulation-based transceiver design with an
instantaneous BER constraint
Therefore, the adaptive modulation-based transceiver design is
formulated by the following optimization problem:

^ PS
ðP2Þ: max C (28)
ρPS ;Pt ;γ

^ EH  PEH;0
s: t: P (28a)

^  BER0
BERðγÞ  BER0 or BER (28b)

0  Pt  Pave (28)

0  ρPS  1 (28d)
As shown in Eq. (28), the objective of (P2) is to maximize the average
spectrum efficiency of the SWIPT system by optimizing the power split-
ting ratio, ρPS, of the SWIPT receiver, as well as the modulation-switching
thresholds, γ, and the transmit power, Pt, of the SWIPT transmitter. The
constraint (28a) represents that the average amount of energy harvested
by the SWIPT receiver has to exceed a minimum requirement, PEH,0.
Moreover, BER(γ)  BER0 in the constraint (28b) requires that the
instantaneous BER(γ) has to be lower than the BER requirement BER0. It
can be relaxed to an average BER constraint of BER ^  BER0 , which means
the BER constraint, BER0, can be occasionally violated, but the long-term
average BER constraint has to be satisfied.
4.3.2. Average BER constraint
4.3.1. Instantaneous BER constraint
We consider the average BER constraint BER d  BER0 in (28b). Ac-
First, we consider the instantaneous BER constraint BER(γ)  BER0 in
cording to Eqs. (23), (24) and (26), transmitting the RF signals with the
(28b). According to Eq. (12), the BER is a decreasing function with
respect to the instantaneous effective SNR, γ. Therefore, we investigate highest average power Pzt ¼ Pave may achieve the optimal spectral effi-
the boundary SNR threshold for every member of the QAM family \{4- ciency, BER, and total energy harvested. However, given non-closed-
QAM, 16-QAM, 64-QAM, 256-QAM, ⋯ \}, above which the correspond- form expressions on all of these performance metrics and given their
ing QAM is capable of satisfying the instantaneous BER constraint, BER0. non-monotonicity with respect to the power splitting ratio, ρPS, and the
modulation switching thresholds, γ, the optimal transceiver design
Specifically, by letting BERM ðγ yðlog 2 MÞ=21 Þ ¼ BER0 of Eq. (12), a feasible
problem (P2) with the average BER constraint cannot be solved effi-
boundary SNR threshold γ ðlog 2 MÞ=21 y for the M-QAM is derived as
ciently with polynomial time complexity. Therefore, a genetic algorithm
  2 is invoked for solving this generalized optimization problem, which is
M  1 1 BER0 log 2 M
γ yðlog2 MÞ=21 ¼ Q (29) detailed in Algorithm 3.
3 4
Algorithm 3. Adaptive modulation-based transceiver design with an
where Q1(⋅) is the inverse Q-function. As the order M increases, the BER instantaneous BER constraint

309
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

Fig. 3. Different modulation schemes with different orders.

outperforms M-PSK and M-QAM when the minimum energy-harvesting

requirement, PEH,0, is very high. Furthermore, a modulation scheme
having a higher order achieves a higher spectrum efficiency, when the
minimum energy-harvesting requirement, PEH,0, is in a low region.
However, if we continuously increase PEH,0, the modulation schemes
having M ¼ 4 achieve the highest spectrum efficiency. We then investi-
gate the SWIPT performance of a high-order modulation scheme. Fig. 4
shows that 256-QAM has the poorest SWIPT performance. This is because
256-QAM has the worst reliability performance. To satisfy the BER
requirement, the SWIPT transmitter has to stay in the outage state, which
substantially reduces the spectrum efficiency and energy-harvesting
performance.
Second, we study the effect of fading channel on the SWIPT perfor-
mance in Fig. 5, where the fading factor is set to m ¼ {1, 5, 10}. Fig. 5
shows that when m ¼ 10, which represents a less fading channel, the
spectrum efficiency of all the modulation schemes is higher than that of
m ¼ 1, which represents a more fading channel. However, in a less fading
channel associated with m ¼ 10, the energy harvested by the SWIPT
receiver is much lower than a fading channel having m ¼ 1. This
observation provides meaningful insight for deploying SWIPT in
different channel conditions. The WIT performance is better in a less
fading channel. By contrast, since a more fading channel provides plenty
of multi-path transmissions, the accumulated signal received at the
SWIPT receiver has a higher chance of exceeding the energy-harvesting
threshold, Pth, which results in a better WPT performance.
Third, we investigate the effect of energy harvesting threshold, Pth, on
the SWIPT performance of different modulation schemes with a fading
factor of m ¼ 1, where we have Pth ¼ 0 (no threshold) and Pth ¼  20
dBm. Fig. 6 shows that when Pth ¼ 0, 16-QAM performs best in terms of
SWIPT performance, followed by 16-PSK and 16-PAM in descending
order. In contrast, when we have Pth ¼  20 dBm, 16-QAM performs best.
However, 16-PSK may achieve a higher spectrum efficiency than 16-PAM
in a low region of the minimum energy harvesting requirement, PEH,0. In
5. Numerical results a high region of PEH,0, 16-PAM performs better than 16-PSK. Further-
more, when we do not have any threshold effect in energy harvesting, all
The fixed and adaptive modulation-based SWIPT transceiver gener- modulation schemes have a substantially higher SWIPT performance
ally obeys the following parameter settings without further descriptions. than a non-zero threshold Pth.
The transmit distance is d ¼ 10 m, while the path loss exponent is set to λ Finally, we study the effect of different BER requirements on the
¼ 3. The fading parameter is m ¼ 1 and the RF-DC conversion efficiency SWIPT performance in Fig. 7, where we have BER0 ¼ 104 as the black
is set to η ¼ 0.5. The average transmit power of the SWIPT transmitter is curves and BER0 ¼ 102 as the red curves. Fig. 7 shows that when we
set to Pave ¼ 10 dBm, while the noise power of the information demod- have BER0 ¼ 102, the SWIPT performance of all the modulation schemes
ulator is σ 2ID ¼ 50 dBm. The energy consumption of the SWIPT receiver is higher than that of BER0 ¼ 104. This is because when we have a strict
is Pth ¼  20 dBm. The maximum BER that can be tolerated by the SWIPT BER requirement, the SWIPT transmitter may frequently be in the outage
receiver is set to BER0 ¼ 104. state, which substantially reduces the spectrum efficiency and energy-
harvesting performance. Furthermore, we compare the SWIPT perfor-
5.1. Fixed modulation mance of practical modulation schemes associated with a finite alphabet
to the Gaussian distributed signals. The big gap in terms of the SWIPT
First, we study the SWIPT performance of different modulation performance between Gaussian signals and practical modulations tells us
schemes with a fading factor of m ¼ 1. Fig. 4 shows that the M-PAM that Gaussian signals are meaningless in the future deployment of the
scheme performs better when M is low. However, if we increase M to 16, SWIPT system. Instead, we have to consider actual signalling with
M-QAM outperforms M-PAM and M-PSK. As portrayed in Fig. 4, M-PAM practical modulations.

310
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

5.2. Adaptive modulation

In Fig. 8, we show the SWIPT performance of the adaptive

modulation-based transceiver and its fixed modulation-based equivalents
with the immediate and average BER constraints. When the instanta-
neous and average BER limitations are equivalent to BER0 ¼ 104,
adaptive modulation can achieve a substantially bigger rate-energy re-
gion than other fixed modulation systems, as demonstrated in Fig. 8(a)
and (b).
Specifically, as shown in Fig. 8(a), we consider an instantaneous BER
constraint. When the minimum power requirement of the energy har-
vesting is 0.5 μW, the 64-QAM-based transceiver achieves a rate of 3.6
bits/channel use, which is the highest among all the fixed modulation
schemes. However, the adaptive modulation is capable of achieving a
rate of 5.1 bits/channel use, which is almost 50% higher than the 64-
QAM-based counterpart. Moreover, given a certain rate of 3 bits/chan-
nel use, the energy harvesting performance of the 16-QAM is 1.3 μW,
which is the highest among all the fixed counterparts. However, the
adaptive modulation is capable of achieving 1.7 μW, which is 31% higher
than the 16-QAM. Note that the performance of the fixed modulation-
based transceiver with an instantaneous BER constraint is obtained by Fig. 5. Different modulation schemes with different fading channel.
invoking Algorithm 2, when a single modulation scheme is conceived.
Similar performance gains of adaptive modulation over the fixed Shannon capacity by assuming Gaussian-distributed continuous
counterparts can be observed in Fig. 8(b), when an average BER channel input.
constraint is considered. For example, when a specific energy-harvesting
We consider a Nakagami-m fading channel to characterize different
power requirement 0.5 μW is considered, the adaptive modulation can fading conditions. For example, m ¼ 1 represents Rayleigh fading
achieve 6.5 bits/channel use, which is 51% higher than the 64-QAM. without LoS, while an increasing m represents an increasing portion
Furthermore, when a rate of 3 bits/channel is specified, adaptive mod- of LoS.
ulation can attain an energy-harvesting performance of 2.32 W, which is
55% greater than 16-QAM. Furthermore, by comparing the simulation Therefore, considering the above-mentioned practicality, we may
results in Fig. 8(a) and (b), we can see that imposing an average BER limit obtain the following insights from our analysis:
effectively expands the rate-energy region.

Different modulation schemes have distinct SWIPT performance, as
6. Practicality and insights introduced in Section 2.1. Specifically, a higher order of the modu-
lation scheme has a better WPT performance when we have a higher
Our model and analysis are based on a practical digital communica- energy-harvesting threshold Pth. Otherwise, a lower order of modu-
tion system for SWIPT that is reflected by the following aspects: lation scheme may achieve a better WPT performance. This analysis
motivates us to design an adaptive modulation scheme for achieving a

We consider practical modulation schemes with finite numbers, such better SWIPT performance.
as M-QAM, rather than theoretical Gaussian signalling.
The DICO mutual information of an M-QAM scheme converges to M

We consider practical Discrete Input Continuous Output (DICO) bits/channel use as we continuously increase the SNR of the infor-
mutual information to evaluate WIT performance, rather than mation decoder. This indicates that we do not need to let more
received RF signals flow to the information decoder, as this may not
linearly increase the WIT performance when we design the power

Fig. 4. Different modulation schemes with different orders. Fig. 6. Different modulation with two energy harvesting threshold.

311
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

(a)
Fig. 7. M-QAM with different BER requirements.

splitter. When the WIT performance converges, we should let more

received RF signals to the energy harvester to further increase the
WPT performance.

A lower m represents a more severe multipath fading condition,
which harms the WIT performance but improves the WPT perfor-
mance. This is because in a more severe multipath fading condition,
the signals propagating along these paths may have a good chance to
be constructively combined at the receiver. Therefore, the total
received signal has a higher Peak-to-Average Power Ratio (PAPR),
which is beneficial for increasing the WPT performance.

Imperfect channel estimation does affect the performance of the

proposed adaptive modulation scheme. According to Ref. [31], the
average BER of the M-QAM based adaptive modulation scheme can only
tolerate the channel estimation error with a Mean Square Error (MSE)
lower than  25 dB. However, in practical applications of SWIPT, the
distance between the SWIPT transmitter and receiver should not be (b)
higher than 15 m to ensure that sufficient energy can be harvested. By
considering the simulation settings in Section 5, the modulation Fig. 8. Adaptive modulation-based transceiver versus fixed modulation-based
counterpart with (a) instantaneous BER constraint and (b) average BER
switching thresholds are obtained as 12, 18, 24, and 30 dB, as shown in
constraint. The adaptive modulation-based SWIPT transmitter chooses a mod-
Fig. 8(b), when the minimum energy-harvesting requirement is 1.5 μW.
ulation scheme from 4-QAM, 16-QAM, 64-QAM, and 256-QAM.
According to Fig. 8 of [32], the MSE of the channel estimation is
approximately  36 dB by exploiting a linear minimum mean square
error-based channel estimation scheme when the SNR is 12 dB. When the
SNR increases, the MSE can be further reduced. When the MSE is lower between ideal Gaussian signaling and practical modulations indicates
than 36 dBm, the average BER and other SWIPT performance are no that practical modulation-based signaling should be considered in future
longer sensitive to the channel estimation error in practice. As a result, SWIPT system design; and 4) the adaptive modulation can substantially
for the practical application of SWIPT, we think the imperfect CSI might increase both the spectrum efficiency and the total energy harvested,
not be a problem for our proposed adaptive modulation scheme. given a specific BER constraint.

7. Conclusions Declaration of competing interest

In this study, we investigated the SWIPT performance of practical The authors declare that they have no known competing financial
modulation schemes with finite alphabets. We used fixed and adaptive interests or personal relationships that could have appeared to influence
modulation schemes to optimize the SWIPT transceiver's WIT spectrum the work reported in this paper.
efficiency while meeting energy-harvesting and reliability criteria. We
came to some crucial conclusions based on our numerical findings: 1) Acknowledgement
with a practical energy harvester, M-PAM outperforms M-PSK and M-
QAM in terms of WPT. M-QAM, on the contrary, outperforms both M- The authors would like to thank the financial support of National
PAM and M-PSK in terms of WIT performance; 2) a more fading channel Natural Science Foundation of China (NSFC), Grant No. 61 971 102,
is bad for WIT but good for WPT; 3) the large SWIPT performance gap 61 871 076 and the Key Research and Development Program of Zhejiang
Province under Grant No. 2022C01093.

312
J. Hu et al. Digital Communications and Networks 8 (2022) 303–313

References [16] I.-M. Kim, D.I. Kim, Wireless information and power transfer: rate-energy tradeoff
for equi-probable arbitrary-shaped discrete inputs, IEEE Trans. Wireless Commun.
15 (6) (2016) 4393–4407.
[1] Toh C.K., Security for smart cities[J]. IET Smart Cities 2 (2) (2020) 95–104.
[17] X. Zhu, W. Zeng, C. Xiao, Precoder design for simultaneous wireless information
[2] Y. Zhao, J. Hu, K. Yang and S. Cui, “Deep Reinforcement Learning Aided Intelligent
and power transfer systems with finite-alphabet inputs, IEEE Trans. Veh. Technol.
Access Control in Energy Harvesting Based WLAN,” in IEEE Transactions on
66 (10) (2017) 9085–9097.
Vehicular Technology, vol. 69, no. 11, pp. 14078-14082, Nov. 2020.
[18] B. Chen, X. Zhu, X. Tu, Joint precoder design for SWIPT-enabled MIMO relay
[3] B.H. Choi, V.X. Thai, E.S. Lee, J.H. Kim, C.T. Rim, Dipole-coil-based wide-range
networks with finite-alphabet inputs, IEEE Access 8 (2020) 179105–179117.
inductive power transfer systems for wireless sensors, IEEE Trans. Ind. Electron. 63
[19] J. Hu, Y. Zhao, K. Yang, Modulation and coding design for simultaneous wireless
(5) (2016) 3158–3167.
information and power transfer, IEEE Commun. Mag. 57 (5) (2019) 124–130.
[4] V. Jiwariyavej, T. Imura, Y. Hori, Coupling coefficients estimation of wireless power
[20] E. Bayguzina, B. Clerckx, Asymmetric modulation design for wireless information
transfer system via magnetic resonance coupling using information from either side
and power transfer with nonlinear energy harvesting, IEEE Trans. Wireless
of the system, IEEE J. Emerg. Sel. Top. Power Electron. 3 (1) (2015) 191–200.
Commun. 18 (12) (2019) 5529–5541.
[5] J. Hu, K. Yang, G. Wen, L. Hanzo, Integrated data and energy communication
[21] Y. Zhao, J. Hu, Z. Ding, K. Yang, Joint interleaver and modulation design for multi-
network: a comprehensive survey, IEEE Commun. Surv. Tutorials 20 (4) (2018)
user SWIPT-NOMA, IEEE Trans. Commun. 67 (10) (2019) 7288–7301.
3169–3219.
[22] Y. Zhao, J. Hu, A. Xie, K. Yang, K.-K. Wong, Receive spatial modulation aided
[6] W. Saad, M. Bennis, M. Chen, A vision of 6G wireless systems: applications, trends,
simultaneous wireless information and power transfer with finite alphabet, IEEE
technologies, and open research problems, IEEE Netw. 34 (3) (2020) 134–142.
Trans. Wireless Commun. 19 (12) (2020) 8039–8053.
[7] Y. Zhao, J. Hu, R. Guo, K. Yang and S. Leng, “Enhanced CSMA/CA Protocol Design
[23] J. Hu, L. Zhang, Q. Yu, K. Yang, Wireless information and energy provision with
for Integrated Data and Energy Transfer in WLANs,” 2018 IEEE Global
practical modulation in energy self-sustainable wireless networks, in: Proceedings
Communications Conference (GLOBECOM), 2018, pp. 1-6.
of IEEE MSN, vol. 2020, 2020.
[8] J. Hu, X. Cai and K. Yang, “Joint Trajectory and Scheduling Design for UAV Aided
[24] A. Svensson, An introduction to adaptive QAM modulation schemes for known and
Secure Backscatter Communications,” in IEEE Wireless Communications Letters,
predicted channels, Proc. IEEE 95 (12) (2007) 2322–2336.
vol. 9, no. 12, pp. 2168-2172, Dec. 2020.
[25] J. Hu, L. Yang, L. Hanzo, Maximum average service rate and optimal queue
[9] B. Clerckx, Wireless information and power transfer: nonlinearity, waveform
scheduling of delay-constrained hybrid cognitive Radio in Nakagami fading
design, and rate-energy tradeoff, IEEE Trans. Signal Process. 66 (4) (2018)
channels, IEEE Trans. Veh. Technol. 62 (5) (2013) 2220–2229.
847–862.
[26] G. Xia, Y. Lin, X. Zhou, W. Zhang, F. Shu, J. Wang, Hybrid precoding design for
[10] R.F. Buckley, R.W. Heath, System and design for selective OFDM SWIPT
secure generalized spatial modulation with finite-alphabet inputs, IEEE Trans.
transmission, IEEE Trans. Green Commun. Netw. (2020), 1–1.
Commun. (2020), 1–1.
[11] L. Dai, B. Wang, M. Peng, S. Chen, Hybrid precoding-based millimeter-wave
[27] R. Zhang, L.-L. Yang, L. Hanzo, Error probability and capacity analysis of
massive MIMO-NOMA with simultaneous wireless information and power transfer,
generalised pre-coding aided spatial modulation, IEEE Trans. Wireless Commun. 14
IEEE J. Sel. Area. Commun. 37 (1) (2019) 131–141.
(1) (2015) 364–375.
[12] Q. Yue, J. Hu, K. Yang, C. Huang, Transceiver design for simultaneous wireless
[28] E. Baccarelli, A. Fasano, Some simple bounds on the symmetric capacity and outage
information and power multicast in multi-user mmWave MIMO system, IEEE Trans.
probability for QAM wireless channels with Rice and Nakagami fadings, IEEE J. Sel.
Veh. Technol. 69 (10) (2020) 11394–11407.
Area. Commun. 18 (3) (2000) 361–368.
[13] K. Lv, J. Hu, Q. Yu, K. Yang, Throughput maximization and fairness assurance in
[29] A. Goldsmith, Wireless Communications, Cambridge University Press, 2005.
data and energy integrated communication networks, IEEE Internet Things J. 5 (2)
[30] M.K. Simon, M.S. Alouini, Digital Communication over Fading Channels: A Unified
(2018) 636–644.
Approach to Performance Analysis, Wiley, 2000.
[14] Y. Zhao, J. Hu, Y. Diao, Q. Yu, K. Yang, Modelling and performance analysis of
[31] J. Paris, M. Aguayo-Torres, J. Entrambasaguas, Impact of channel estimation error
wireless LAN enabled by RF energy transfer, IEEE Trans. Commun. 66 (11) (2018)
on adaptive modulation performance in flat fading, IEEE Trans. Commun. 52 (5)
5756–5772.
(2004) 716–720.
[15] Y. Zhao, D. Wang, J. Hu, K. Yang, H-AP deployment for joint wireless information
[32] V. Savaux, F. Bader, Y. Lou€et, A joint MMSE channel and noise variance estimation
and energy transfer in smart cities, IEEE Trans. Veh. Technol. 67 (8) (2018)
for OFDM/OQAM modulation, IEEE Trans. Commun. 63 (11) (2015) 4254–4266.
7485–7496.

313