6336 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO.

7, 1 APRIL 2023

Performance Analysis and Power Allocation for

Cooperative ISAC Networks
Meng Liu , Graduate Student Member, IEEE, Minglei Yang , Member, IEEE, Huifang Li ,
Kun Zeng , Member, IEEE, Zhaoming Zhang , Arumugam Nallanathan , Fellow, IEEE,
Guangjian Wang , Member, IEEE, and Lajos Hanzo , Life Fellow, IEEE

Abstract—To mitigate the overlapping of the radar and com- Index Terms—Cooperative communication, full-duplex (FD),
munication frequency bands caused by large-scale devices access, integrated sensing and communication (ISAC), nonorthogonal
we propose a novel integrated sensing and communication (ISAC) downlink transmission (NO-DLT).
system, where a micro base station (MiBS) simultaneously car-
ries out both target sensing and cooperative communication.
Concretely, the MiBS, acting as the sensing equipment, can I. I NTRODUCTION
also serve as a full-duplex decode-and-forward relay to assist
end-to-end communication. Moreover, nonorthogonal downlink A. Background
transmission (NO-DLT) is adopted between the macro base sta- RIGGERED by the roll-out of the Internet of Things
tion and the Internet of Things devices, so that the spectrum
utilization can be further improved. To facilitate the performance
evaluation, both the exact and asymptotic outage probabilities,
T (IoT), a large number of devices are connected to the
mobile communication networks, resulting in escalating spec-
the ergodic rates associated communication, and the probability trum demand in support of superior user experience [1],
of successful sensing detection are characterized. Subsequently, a [2], [3]. In order to alleviate the potential spectrum con-
pair of problems of maximizing the receive signal-to-interference- gestion, diverse methods have been proposed, among which
plus-noise ratio of the sensing signal and maximizing the sum
rate of communication are formulated that are solved by the clas- the sensing-communication coexistence (SCC) [4] and the
sic Lagrangian method while exploiting the associated function integrated sensing and communication (ISAC) [5] techniques
monotonicity. Our simulation results demonstrate that: 1) The are appealing, exhibiting promising system performance. As
proposed ISAC NO-DLT system improves both the communica- for the SCC scheme, the communication and sensing (C&S)
tion and sensing performance under the same power consumption devices operate independently of each other in the same
as noncooperative NO-DLT and 2) the proposed power allocation
(PA) schemes are superior to the random PA scheme. frequency band, hence resulting in interfering C&S signals.
In contrast, for the ISAC scheme, the C&S devices operate by
cooperating or sharing a physical site in support of the C&S
functions [6].
With the global introduction of the fifth-generation (5G)
Manuscript received 13 October 2022; revised 20 November 2022; accepted new radio (NR), reliability, delay, and user experience have
23 November 2022. Date of publication 28 November 2022; date of current reached unprecedented levels. Recently, looking forward to
version 24 March 2023. This work was supported in part by the National
Natural Science Foundation of China under Grant 62231020; in part by the the future multifunctional network application scenarios of
111 Project; in part by the Higher Education Discipline Innovation Project; smart cities, smart transportation, smart environments, and
in part by the National Natural Science Foundation of Shaanxi Province smart healthcare, purely relying on mobile communications
under Grant 2017CGZH-RGXQ-02; and in part by the Huawei Technologies
Company Ltd., under Grant YBN2019085064. The work of Lajos Hanzo was for data transmission has been shown insufficient to cater
supported in part by the Engineering and Physical Sciences Research Council for these demanding requirements [7]. These emerging needs
under Project EP/W016605/1 and Project EP/P003990/1 (COALESCE), and in have fueled the fusion of C&S paving the way for the com-
part by the European Research Council’s Advanced Fellow Grant QuantCom
under Grant 789028. (Corresponding author: Minglei Yang.) pleting concept of ISAC [8]. Indeed, heterogeneous wireless
Meng Liu, Minglei Yang, and Zhaoming Zhang are with the National applications are omni-present and the IoT has been widely
Laboratory of Radar Signal Processing, Xidian University, Xi’an 710071, used in various domains, which jointly pave the way for the
Shaanxi, China (e-mail: [email protected]; [email protected];
[email protected]). implementation of ISAC [9]. For instance, Zhang et al. [10]
Huifang Li is with the State Key Laboratory of Integrated Services proposed an ISAC system, which is capable of reducing
Networks, Xidian University, Xi’an 710071, Shaanxi, China (e-mail: the delay of information sharing among vehicles and of
[email protected]).
Kun Zeng and Guangjian Wang are with the Wireless Technology improving the date rate, while guaranteeing the safety of
Laboratory, Huawei Technologies Company Ltd., Chengdu 611731, Sichuan, autonomous driving. Furthermore, in [11], an ISAC system
China (e-mail: [email protected]; [email protected]). of unmanned aerial vehicle (UAV) based on beam sharing
Arumugam Nallanathan is with the School of Electronic Engineering and
Computer Science, Queen Mary University of London, E1 4NS London, U.K. was studied, in which UAVs could accomplish the func-
(e-mail: [email protected]). tions of C&S. The above-mentioned papers have shown
Lajos Hanzo is with the School of Electronics and Computer Science, that ISAC has the potential of further improving the sens-
University of Southampton, SO17 1BJ Southampton, U.K. (e-mail: lh@
ecs.soton.ac.uk). ing function of wireless communication networks to new
Digital Object Identifier 10.1109/JIOT.2022.3225281 levels.
2327-4662 
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: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6337

In practice, the success of ISAC requires the implemen- users [24], the authors further confirmed that the FD mode
tation of massive end-to-end data transmission. However, it succeeds in improving the system’s throughput and enable a
is challenging to realize information transmission with lim- fairer resource allocation than the conventional HD mode.
ited spectral resources. To address this issue, nonorthogonal
downlink transmission (NO-DLT) is capable of delivering
significant improvements in delay, energy efficiency, and B. Motivation and Contribution
throughput in contrast to the conventional orthogonal down- The authors of [6], [10], [11], [19], [20], [21], [22], [23],
link transmission (O-DLT). Specifically, O-DLT can only serve and [24] have respectively, laid the theoretical foundations of
a single user by exploiting the time/frequency/code domain C&S and of cooperative relaying systems. In this context, there
orthogonally, while NO-DLT can simultaneously provide a are serval attempts in the literature for investigating ISAC
balanced service for multiple devices while guaranteeing and SCC systems. To elaborate, Chiriyath et al. [25] assumed
the required Quality-of-Service (QoS) of the devices [12], that a radar can simultaneously perform target tracking and
[13], [14]. The principle of NO-DLT is to allow controlled cooperative communication. In particular, the bounds of radar
multiple access interference, while relying on successive and communication as well as the optimal Fisher information
interference cancellation (SIC) to decode the desired signals were discussed to evaluate the system performance attained.
at the receiver [15]. For instance, in [16], a traffic offloading Furthermore, the performance bounds of ISAC system’s were
scheme of a device-to-device (D2D) network was investigated developed in [26]. However, the detailed features of relay were
by using NO-DLT and the resources were optimized for ensur- not provided in [25] and [26]. As a further advance, single-
ing that the system capacity was maximized. Furthermore, the carrier and multicarrier SCC systems were devised in [27]
multiconstraint optimization algorithm was proposed in [17] and [28], respectively, and sophisticated PA schemes were
for maximizing the throughput in wireless-cached NO-DLT proposed for maximizing the communication sum rate or
UAV systems. The aforementioned contributions demonstrate receive signal-to-interference-plus-noise ratio (SINR) of the
that NO-DLT yields significant performance gains over O-DLT radar. However, [25], [26], [27], and [28] were based on
in various communication scenarios, serving as a promis- the conventional O-DLT philosophy without determining their
ing candidate in support of future mobile communication optimal spectrum sharing.
systems. As a remedy, Xu et al. [29] conceived a cognitive OFDM-
In order to further enhance the system’s robustness and to NO-DLT ISAC system, in which the cognitive source was
extend the coverage area, cooperative relaying has been intro- capable of sensing the state of the primary user for oppor-
duced into NO-DLT systems in [18]. For example, the authors tunistically accessing the primary network’s frequency band.
employed a large intelligent reflecting surface as a relay to In particular, a PA algorithm was designed for maximizing
assist the end-to-end transmission of the NO-DLT system the system capacity under the premise of a specific maximum
in [19], while the pairwise error probability expressions of tolerable interference caused by the cognitive source. In the
the users were derived for quantifying the system performance cognitive Industrial IoT system of [30], multiple sensing sta-
attained. Furthermore, in [20], the outage probabilities (OPs) tions worked together and transmitted the sensing information
and the ergodic sum rate (ESR) of users were analyzed both to the data center by relying on nonorthogonal communica-
in cooperative and noncooperative NO-DLT scenarios, which tion for improving the data transmission efficiency. Another
demonstrated that cooperative NO-DLT is indeed capable of NO-DLT ISAC framework was proposed in [31], in which the
enhancing the overall system performance. In addition, a novel transmit and receive beamformers were designed for minimiz-
cooperative NO-DLT system was explored in [21], where the ing the computational error and maximizing weighted the sum
near user acted as a relay in support of a far user to commu- rate. Indeed, [29], [30], [31] provided embryonic recipes for
nicate with the source and the closed-form expressions of the the development of NO-DLT ISAC systems, but no cooper-
associated power allocation (PA) were obtained to maximize ative ISAC schemes were considered. From a practical point
the sum rate at a low computational complexity. of view, the integration of ISAC and NO-DLT arrangements
In contrast to the half-duplex (HD) relaying schemes are capable of improving the spectrum efficiency, enhance the
adopted in [18], [19], [20], and [21], which requires two communication reliability, and ensure resource allocation fair-
orthogonal time slots for two-way signal transmission, full- ness. As for the cooperative ISAC systems of [25], [26], [32],
duplex (FD) relaying requires only a single time slot to and [33], the introduction of FD NO-DLT is capable of reduc-
accomplish this task that potentially doubles the system’s spec- ing the time duration of communication, thus reserving more
tral efficiency. Nevertheless, the inherent loop self-interference time for sensing and further improving the exploitation of the
(LSI) is unavoidable in FD systems. To address this issue, limited resources [34], [35], [36]. Therefore, the performance
Riihonen et al. [22] adopted the natural isolation, the time- analysis and PA schemes of FD NO-DLT ISAC networks is of
domain elimination, and the spatial suppression techniques for pivotal significance. Against the above background, we con-
substantially reducing the LSI, hence facilitating the practical sider an FD NO-DLT ISAC system, where a micro base station
implementation of FD systems. Furthermore, Zhang et al. [23] (MiBS) is capable of simultaneously realizing the functions of
studied an FD NO-DLT vehicle-to-everything system and cooperative relaying and target sensing. A bold and definitive
demonstrated that FD systems are capable of improving the comparison of our contributions to the prior art is presented
latency, despite the negative impacts of LSI. Moreover, in an in Table I. Concretely, the contributions of this article are
FD NO-DLT system supporting multiple downlink and uplink summarized as follows.

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
6338 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO. 7, 1 APRIL 2023

TABLE I
B OLDLY AND E XPLICITLY C ONTRASTING ON C ONTRIBUTION TO THE S TART- OF - THE -A RT

1) For the proposed FD NO-DLT ISAC system, both the Notation: E[ · ] represents the expectation operation,
exact and asymptotic lower bound expressions of the CN(μ, σ 2 ) denotes a complex-valued Gaussian random vari-
OP of IoT devices are derived along with the diversity able with mean μ and variance σ 2 ,  denotes the definition
orders attained in the high signal-to-noise ratio (SNR) operation, Ei(·) is the exponential integral function, C rep-
regime. Furthermore, the ESR of the system consid- resents the set of complex numbers, fX (·) and FX (·) are the
ered is also investigated. To shed light on the results probability density function (PDF) and cumulative distribution
obtained, we further approximate the expression of the function (CDF) of the random variable X, respectively.
ESR. The results indicate that on the basis of realizing
the extra sensing function, the ESR of the proposed FD
NO-DLT ISAC system is better than that of the tradi- II. S YSTEM M ODEL
tional noncooperative NO-DLT system, despite requiring We consider the cooperative FD NO-DLT ISAC network
no additional power. of Fig. 1(a), where a MaBS, S, is in charge of communica-
2) The probability of successful detection (PoD) and prob- tion with two IoT devices,1 while a MiBS, R, is responsible for
ability of false alarm (PoFA) of the MiBS are derived sensing the surrounding targets and for facilitating cooperative
for characterizing the sensing performance of the FD communication between the MaBS and the IoT devices located
NO-DLT ISAC system. Also, the distance between the at different distances from the MaBS, namely, the far IoT
MiBS and the target and the radial velocity of the tar- device Df and the near IoT device Dn . Specifically, in order
get are analyzed. The results obtained show that the to improve the communication performance of the system,
proposed NO-DLT ISAC system outperforms its O-DLT the MiBS can also act as an FD decode-and-forward (DF)
counterpart and the sensing capability is mainly deter- relay to assist the end-to-end communication while sensing
mined by the transmit power of the macro base station a target T.2 In order to reduce both the power consump-
(MaBS). tion and the hardware resources, we assume furthermore that
3) To further improve the performance of the system the MiBS is equipped with two antennas, one for transmit-
considered, we formulate a pair of PA problems: ting and the other for receiving signals, respectively. The
a) a sensing-centric design (SCD), for maximizing MaBS and IoT devices are single-antenna devices. For the
the receive SINR of the sensing signal at the MiBS, convenience of the presentation, the channel coefficients of
while the IoT devices satisfy the minimum required the links S → R, S → Df , S → Dn , R → T, T → R,
SINR constraint and b) a communication-centric design R → Df , and R → Dn of Fig. 1 are denoted by hi ∈ C,
(CCD), for maximizing the sum rate of the IoT devices, i ∈ {SR, SDf , SDn , RT, TR, RDf , RDn }, respectively. We also
while the MiBS meets the minimum SINR require- assumed that all the links are subject to Rayleigh fading, i.e.,
ment. Finally, we analytically derive the closed-form the channel gains ρi = |hi |2 all obey an exponential distri-
expressions of the PA coefficients. bution, capturing both the path-loss and small-scale fading.3
Next, we will discuss the signal transmission process of the
C. Organization and Notations FD NO-DLT ISAC system.
The remainder of this article is organized as follows.
Section II discusses the system model of the FD NO-DLT 1 Although only two IoT devices are considered in this article, the network
ISAC system. In Section III, the performance of the system can be readily extended to multiple IoT devices. A feasible method is to
reaggregate the IoT devices into different clusters, where the devices use
considered is evaluated by deriving the exact and asymp- NO-DLT within each cluster, while the O-DLT philosophy is adopted by the
totic OPs and the corresponding diversity orders. The ESR different clusters [37].
of the IoT devices is derived in Section IV, while Section V 2 We should point out that the essence of the multitarget system is based

analyzes the sensing performance of the system. Two PA on the aggregate of multiple single-target echo signal, which will be set aside
for our future work.
problems are formulated in Section VI. Finally, the numer- 3 Explicitly, the PDF and CDF of ρ are, respectively, given by f (x) =
i  ρi
ical results are provided to demonstrate the correctness of exp(−x/βi )/βi and Fρi (x) = 1−exp(−x/βi ), where βi = h/ 1 + diα denotes
our theoretical analysis in Section VII, before concluding in the channel’s variance, h ∼ CN (0, ) is the complex channel coefficient, α
Section VIII. denotes the path loss exponent, and di is the distance between the nodes [38].

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6339

Fig. 1. Cooperative FD NO-DLT ISAC system model. (a) Illustration of cooperative FD NO-DLT ISAC network. (b) Diagram of the transmit and receive
power at the MiBS.

As for the direct links,

√ during the t-th  time slot, S intends to channel at R and η ∈ [0, 1] denotes the LSI cancellation capa-
send the signal yc = an Pcom xn (t) + af Pcom xf (t) to Df and bility. Specifically, η = 0 indicates that the LSI is perfectly
Dn concurrently, where Pcom denotes the total transmit power cancelled, while η = 1 represents that no interference cancel-
of S, xf ∈ C and xn ∈ C are the desired signals of Df and lation is employed between the transmit and receive antennas.
Dn , respectively, with E[|xf |2 ] = E[|xn |2 ] = 1, and af and an Furthermore, τl and τs represent the time delay of the LSI and
represent the PA coefficients of S, so that an + af = 1 and the target echo, respectively. Additionally, hRR ∈ C is the new
af > an [39], [40]. Thus, the signals received at Df and Dn channel coefficient which captures the joint impact of hRT ∈ C
can be expressed as and hTR ∈ C as well as δ ∈ [0, 1] denotes the reflection factor
   of the target [41]. In the sequel, we introduce the CDF and
yi1 = hi1 an Pcom xn (t) + af Pcom xf (t) + ni1 (1) PDF of ρRR , respectively.
Lemma 1: Upon assuming that the target is moving slowly,
where i1 ∈ {SDf , SDn }, ni1 ∼ CN(0, N0 ) is the additive white the channel environment experienced by the signals of R → T
Gaussian noise (AWGN), and N0 is the noise power. and T → R remains approximately time-invariant, i.e., ρRT =
√ As for the cooperative links, the MiBS transmits yr = ρTR . Then, the CDF and PDF of ρRR are expressed as
Psen xr (t − τr ) for sensing the target, then it receives the sig- √
x
−
nals from S and the target echo, where Psen is the power budget FρRR (x) = 1 − e βRR (3)
√
of R, xr ∈ C denotes the transmit signal of the MiBS with 1 −
x

E[|xr |2 ] = 1, and τr is the sensing signal duration. Fig. 1(b) fρRR (x) = √ e βRR . (4)
2βRR x
shows the transmit and receive power at the MiBS during the
t-th time slot. The signals received at the MiBS can be divided Proof: By using the CDF of ρi , we have
into five parts: 1) the communication signal transmitted form FρRR (x) = Pr(ρRR = ρRT ρTR ≤ x)
the MaBS; 2) the target echo reflected by the target; 3) the  √x  √x
communication LSI (CLSI) transmitting the communication = fρRT (t)dt = fρTR (t)dt
signal; 4) the sensing LSI (SLSI) transmitting the sensing sig- 0 √
0 √
x x
nal; and 5) the AWGN. Therefore, the signal received at R can −β −β
= 1−e RT =1−e TR . (5)
be expressed as4
 Using the basic variable substitution, (3) can be obtained.
ySR = hSR yc + hRR δPsen xr (t − τl − τs ) Then, taking the derivative of FρRR (x) with respect to x, (4)
  c 
+ hLI ηPsen xLI (t − τl ) + xLI
r
(t − τl ) + nSR (2) can be obtained.
After receiving the signals, the MiBS, serving as an FD DF
where nSR ∼ CN(0, N0 ) is the AWGN at R, xLI c ∈ C and relay for communication, decodes and forwards the receive
xLI ∈ C represent the CLSI and SLSI at R by transmitting
r signals to Df and Dn . The signals received by the IoT devices
the C&S signals associated with E[|xLI c |2 ] = E[|xr |2 ] = 1,
LI can be expressed as5
respectively. Furthermore, hLI ∈ C is the coefficient of the LSI   
yi2 = hi2 bn Psen xn (t − τc ) + bf Psen xf (t − τc ) + ni2 (6)
4 The secondary reflection from the target received by the MiBS due to
the illumination of the MaBS is ignored in the data transmission, because 5 It should be emphasized that if R can decode the communication signals
the direct-path communication interference arriving from the MaBS is more successfully, then it will forward them to the IoT devices. Otherwise R remains
serious than the secondary reflection [28]. silent and the cooperative links will be in a state of outage.

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
6340 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO. 7, 1 APRIL 2023

TABLE II
where i2 ∈ {RDf , RDn }, bf and bn denote the PA coefficients S UMMARY OF THE N OTATIONS
of the MiBS associated with bf + bn = 1 and bf > bn 6 [39],
ni2 ∼ CN(0, N0 ) denotes the AWGN.
The receive SINRs of the communication signals can be
divided into direct and cooperative links as well.

A. Direct Links
As for the direct links, the MaBS communicates with the
IoT devices directly. According to the NO-DLT protocol, xf is
first decoded at Dn , and then xn is decoded by the application
of SIC. Thus, the SINRs of Dn associated with decoding xf
In the following, the communication performance of the
and xn are respectively expressed as
system considered is evaluated by investigating both the exact
D af ρSDn γc and asymptotic OPs, diversity orders, and the ESR of the IoT
γSDf n = (7)
an ρSDn γc + 1 devices. As a further development, the PoFA and PoD of the
Dn MiBS are, respectively, derived for characterizing the sens-
γSDn
= an ρSDn γc (8)
ing performance. For the sake of convenience, the significant
where we have γc = Pcom /N0 . performance analysis identifiers are summarized in Table II.
The receive SINR of Df decoding its own signal over the
direct link can be expressed as III. OP A NALYSIS
D af ρSDf γc The OP is an important performance metric of a system,
γSDf f = . (9)
an ρSDf γc + 1 which is related to the average SINRs of the links and the
channel fading parameters. Therefore, in this section, we
B. Cooperative Links derive the exact OP lower bound of the IoT devices. In order to
As for the cooperative links, R decodes the IoT devices’ glean deeper insights concerning the OP, both the asymptotic
desired signals in turn by employing SIC and treats the echo, OPs as well as the diversity orders are analyzed for further
CLSI, and SLSI as interferences imposed on the communica- characterizing the performance of the proposed FD NO-DLT
tion signals. Thus, the receive SINRs of xf and xn at R can be ISAC system.
respectively expressed as
A. Exact OP
D af ρSR γc
γSRf = (10) An outage event is encountered at Df , when one of the
an ρSR γc + ρRR δγr + 2ηρLI γr + 1
Dn an ρSR γc following two events occurs. Case 1) Neither R nor Df can
γSR = (11) successfully decode xf from the S → R and S → Df links
ρRR δγr + 2ηρLI γr + 1
and case 2) R decodes and forwards xf to Df , while Df fails
where we have γr = Psen /N0 . to decode the desired signal from the R → Df and S → Df
If xf is decoded successfully at R, i.e., the receive SINR of links. Thus, the OP of Df is expressed as7
xf exceeds the outage threshold, xf will be sent to Df . Then,
D
  D D
 
the receive SINR of xf at Df can be expressed as Poutf = Pr max γSRf , γSDf f < γthf
D bf ρRDf γr I1
γRDf f = . (12)  D  D  
bn ρRDf γr + 1 D
+ Pr γSR ≥ γthf , max γRDf f , γSDf f < γthf (15)
f

Similarly, if both xf and xn are decoded successfully at R,

I2
the signals will be sent to the Dn . At Dn , the high-power
signal xf will be decoded first and then it will be remodu- where γthf denotes the SNR threshold of xf , while I1 refers to
lated and subtracted from Dn to cancel its effect until xn is case 1 and I2 refers to case 2. The exact OP expression of Df
detected. Therefore, the receive SINRs of xf and xn at Dn are, for the FD NO-DLT ISAC system considered is given in the
respectively, expressed as following theorem.
Theorem 1: The exact OP of Df device is expressed as
D bf ρRDn γr  
γRDf n = (13) θ1
bn ρRDn γr + 1 D −β βSR π βSR
Poutf = 1−e SDf
1−
Dn
γRD = bn ρRDn γr . (14) 2βRR (βSR + 2ηβLI γr θ1 ) δγr θ1
n
βSR θ1 ϕ1
  
2 δγ θ − βSR − βRD 1 βSR
6 As stated in [37], [39], [42], [43], and [44], we assume that the IoT ×e 4βRR r 1 f 1 − erfc (16)
2βRR δγr θ1
devices which are far/near from the MaBS, are far/near from the MiBS as
well. In practice, an easily overlooked scenario exists, where the IoT devices
are far/near from the MaBS, while near/far from the MiBS. In this case, the 7 The selection combining (SC) scheme is adopted to calculate the OPs of
PA coefficients of the MiBS satisfy bf < bn and bf + bn = 1. The expressions the IoT devices, since it has low implementation cost and does not reduce
of the SINRs, OP, and ESR also have to be changed according to the NO-DLT the diversity gain. Our analysis may also be reality extended to the maximal
protocol. ratio combining (MRC) scheme, which will be set aside for our future work.

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6341

where θ1 = γthf /(af γc − an γc γthf ) with af > an γthf , ϕ1 = high-SNR regime are analyzed in this section. The diversity
D order is defined as [45]
γthf /(bf γr − bn γr γthf ) with bf > bn γthf ; otherwise, = 1 Poutf  
and erfc(·) denotes the complementary error function which log P∞out
is expressed as DO = − lim (20)
ϒ→∞ log ϒ
 ∞
2 where ϒ ∈ {γc , γr } and P∞out is the asymptotic OP of the IoT
e−t dt.
2
erfc(z) = √ (17)
π z devices in the high-SNR regime.
The asymptotic OPs of Df and Dn in the high-SNR regime
Proof: See Appendix A. are summarized in the following corollaries.
Similarly, an outage event occurs at Dn when: case 3) Either Corollary 1: The asymptotic OP of Df is expressed as
xf or xn cannot be decoded successfully from the S → Dn and  
S → R links; case 4) R decodes and forwards xf and xn to Df ,∞ θ1 ϕ1 βSR
Pout ≈ 1− 1−  
Dn , while Dn fails to decode either of them from the S → Dn βSDf βRDf 2βRR βSR + 2ηβLI θ1∗
and R → Dn links. Thus, the OP of Dn is expressed as (18),  β
  
shown at the bottom of the page, where γthn denotes the SNR π βSR 4β 2SRδθ ∗ 1 βSR
× e RR 1 1 − erfc (21)
threshold of xn , I3 refers to case 3) and I4 refers to case 4). δθ1∗ 2βRR δθ1∗
The exact OP expression of Dn for the FD NO-DLT ISAC where we have θ1∗ = γthf /(af − an γthf ).
considered system is given in the following theorem. Proof: See Appendix B.
Theorem 2: The exact OP of Dn device is expressed as By substituting (21) into (20), we can show that the diversity
   order of Df in the high-SNR regime is DO = 1.
f
Dn −β θ β SR π βSR
Pout = 1 − e SDn 1− Corollary 2: The asymptotic OP of Dn is expressed as
2βRR (βSR + 2ηβLI γr θ ) δγr θ   
   Dn ,∞ θ ϕ βSR
βSR θ ϕ Pout ≈ 1− 1−
2 δγ θ − βSR − βRDn 1 βSR βSDn β 2βRR (βSR + 2ηβLI θ ∗ )
×e 4βRR r 1 − erfc  RDn  
2βRR δγr θ βSR
π βSR 4β 2 δθ ∗ 1 βSR
× e RR 1 − erfc (22)
(19) δθ ∗ 2βRR δθ ∗
where θ  max(θ1 , θ2 ), θ2 = γthn /(an γc ), ϕ  max(ϕ1 , ϕ2 ), where we have θ ∗  max(θ1∗ , θ2∗ ) and θ2∗ = γthn /an .
and ϕ2 = γthn /(bn γr ). Proof: Similar to Appendix upon, substituting (7), (8)
Proof: Similar to Appendix A, substituting (7), (8), and (10)–(14) into (18), with the PDF and CDF of ρi , (22)
and (10)–(14) into (18), and then using the PDFs and CDFs can be obtained after some mathematical manipulations.
of ρi , (19) can be obtained through a series of mathematical By substituting (22) into (20), the diversity order of Dn in
manipulations. the high-SNR regime is DnO = 1.
Remark 1: We can observe from Theorems 1 and 2 that the Remark 2: We can observe from (21) and (22), that the OPs
OP of Df depends on the decoding capability of xf at R and Df , of cooperative links are constant in the high-SNR regime due
while the OP of Dn is related to the decoding capability of both to the existence of the CLSI, SLSI, and echo, which further
xf and xn at R and Df . More particularly, when considering leads to having a zero diversity order for the cooperative links.
the direct links, the OPs of the far and near IoT devices are In this case, the diversity orders of the IoT devices are pro-
limited by the channel fading, which are, respectively, given vided by those of the direct links. Specifically, for the direct
D
by Poutf = 1 − exp(−θ1 /βSDf ) and PD out = 1 − exp(−θ/βSDn ).
n
links, the asymptotic OPs of the far and near IoT devices are,
D ,∞ Dn ,∞
When considering the cooperative links, the CLSI, SLSI, and respectively, given by Poutf ≈ θ1 /βSDf and Pout ≈ θ/βSDn ,
echo have a significant impact on the OPs of the IoT devices. f
while the corresponding diversity orders are DO = DnO = 1.

B. Diversity Order IV. ESR A NALYSIS

To obtain further insights concerning on the FD NO-DLT The ESR represents the sum rate of the IoT devices per unit
ISAC system, the diversity orders of the IoT devices in the time, which characterize the fading performance of the system

⎛  ⎛ ⎞ ⎞
D D Dn
γ f γ Dn γSDf n γSD
PDn
out = Pr⎝min SR , SR < 1, min⎝ , n⎠
< 1⎠
γthf γthn γthf γthn

I3
⎛  ⎛ ⎞ ⎛ ⎞ ⎞
D D Dn Df Dn
γSRf γ Dn γRDf n γRD γ γ
+ Pr⎝min ≥ 1, min⎝ n⎠
< 1, min⎝ n⎠
< 1⎠
SD SD
, SR , n
, (18)
γthf γthn γthf γthn γthf γthn

I4

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
6342 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO. 7, 1 APRIL 2023

from a different perspective than the OP. Thus, in this section, Thus, the ESR of the IoT devices is expressed as
the ESR of the IoT devices is explored for the FD NO-DLT
ISAC system considered. f
ave = Rave + Rave .
Rsum n
(29)
The achievable rates of xf and xn are, respectively, expressed
as [39]
  D  From (27) and (28), we can observe that it is challenging
D D
Rf = log2 1 + min γSRf , γRDf f , γRDf n to obtain the exact ESR of the IoT devices, if not impossi-
ble. As a compromise, we analyze the ESR by adopting some
R1f approximations. The approximate ERs of Df and Dn are given
  D D
 in the following corollary.
+ log2 1 + min γSDf f , γSDf n (23) Corollary 3: The approximated ER of the far IoT device is
expressed as (30), shown at the bottom of the page.
R2f
     The approximate ER of the near IoT device is formulated
Dn Dn Dn
Rn = log2 1 + min γSR , γRD n
+ log 2 1 + γ SDn
. (24) as (31), shown at the bottom of the page.
Based on (30) and (31), the ESR of the IoT devices can be
R1n R2n approximated as
Then, the ergodic rates (ERs) of Df and Dn are, respectively,
ap,sum ap,f ap,n
given by Rave ≈ Rave + Rave . (32)
   
f f ,1 f ,2
Rave = Rave + Rave = E R1f + E R2f (25)
    Proof: See Appendix D.
Rnave = Rn,1 Remark 3: From Theorem 3 and Corollary 3, we can
ave + Rave = E Rn + E Rn .
n,2 1 2
(26)
observe that the sensing signal has a negative impact on the
Theorem 3: The ER of the far IoT device is expressed ESR of the IoT devices. For the cooperative links, the ERs of
as (27), shown at the bottom of the page, where gc (w) = Df and Dn , respectively, tend to be log2 [1 + af βSR /(an βSR +
(af γc − an γc w)/w and gr (w) = (bf γr − bn γr w)/w. 2 δ + 2ηβ )] and log [1 + a β /(2β 2 δ + 2ηβ )] in the
2βRR LI 2 n SR RR LI
The ER of the near IoT device is expressed as (28), shown high-SNR regime. For the noncooperative scheme, the sum
at the bottom of the page, where qc (w) = an γc /w and qr (w) = rate reduces to Rsum = R2f + R2n . Moreover, it is noteworthy
bn γr /w that if the MiBS operates in the HD mode, the rates of the
Proof: See Appendix C. IoT devices can be increased due to the absence of CLSI.

3
βSR gc (w1 ) 1 1 1
2 δγ − βSR gc (w1 ) − βRD gr (w1 ) − βRD gr (w1 )
   
 ∞ gc2 (w1 )e 4βRR r f n
1 − erfc 2β1RR gc (wδγ1 )βSR
f r
Rave = dw1
(1 + w1 )(βSR gc (w1 ) + 2ηβLI γr )
0
  ∞
βSR π βSR 1 1 − β g1 w − β g1 w
c( ) SDn c ( 2 )
× + e SDf 2 dw2 (27)
2βRR ln 2 δγr ln 2 0 1 + w2

3 βSR qc (w3 ) 1 1    
2 δγ − βSR qc (w3 ) − βRD qr (w3 )
 q 2 (w3 )
∞ (β q (wc )+2ηβ e 4βRR r n 1 − erfc 2β1RR qc (wδγ3 )βSR
SR c 3 LI γr ) r
Rnave = dw3
1 + w3
0
  
βSR π βSR 1 βSD an γc
1 1
× − e n Ei − (28)
2βRR ln 2 δγr ln 2 βSDn an γc

  
ap,f 1 af βSDf γc af βSDn γc
Rave ≈ ln 1 + min ,
ln 2 an βSDf γc + 1 an βSDn γc + 1
 
af βSR γc bf βRDf γr bf βRDn γr
+ ln 1 + min ,
2 δγ + 2ηβ γ + 1 b β
, (30)
an βSR γc + 2βRR r LI r n RDf γr + 1 bn βRDn γr + 1

  
ap,n 1   an βSR γc
Rave ≈ ln 1 + an βSDn γc + ln 1 + min ,b β γ
2 δγ + 2ηβ γ + 1 n RDn r
(31)
ln 2 2βRR r LI r

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6343

V. P O D A NALYSIS where c is the speed of light, and (τl + τs ) denotes the time
As for sensing, the fundamental task is to find the target delay of the target.
and make the correct identification, which is the premise of As a further step, the radial velocity of the target can be
target parameter estimation. Generally, the receive echo is a written as [49]
mixture of the desired signal, interference, and noise. The pres- c
vT = |fe − fs | (39)
ence/absence of the target can be determined by the power 2fs
of the receive signals. Assuming that H0 represents the null where |fe − fs | denotes Doppler frequency, while fs and fe ,
hypothesis and H1 includes echoes, interference, and AWGN, respectively, represent the frequency of the transmitted signal
they can be defined as and the received echo, which can be observed from the MiBS.
  c 
H0 : ySR = hSR yc + hLI ηPsen xLI + xLI
r
+ nSR Remark 4: It can be observed from (34) that for a given
√ PoFA, there exists a unique detection threshold corresponding
H1 : ySR = hSR yc + hRR δyr
  c  to it, and then the PoD can be obtained. In addition, we reiter-
+ hLI ηPsen xLI + xLI
r
+ nSR (33) ate that if the target is outside the range of sensing, the result
where yc and yr denote the signals transmitted by the MaBS obtained through (37) may be generated by the randomness of
and the MiBS, respectively. Indeed, PoFA and PoD are adopted noise, thus further signal processing is essential to determine
for characterizing the sensing performance. Specifically, if the whether the target exists.
receive power of the MiBS exceeds the preset threshold under Remark 5: It should be noted that as the basis of target
H0 , the MiBS concludes that a target is present and hence detection, it is essential to derive the PoD expression for
makes the wrong inference. In contrast, the target can be characterizing the sensing performance of the ISAC system
sensed successfully. Hence, it can be inferred from (33) that considered. In contrast to the mature cognitive radio networks,
the receive power of the MiBS is a noncentral Chi-square dis- in the ISAC systems, the target to be detected is noncooper-
tributed random variable having a Degree of Freedom (DoF) ative, which is challenging for the MiBS. Therefore, in the
of four under H0 and five under H1 [46]. Therefore, accord- following, we design a PA scheme for maximizing the received
ing to H0 , the false-alarm probability of the MiBS is expressed SINR of the sensing signal, so that the PoD can be further
as [47] improved.
  
2(ρSR Pcom + 2ηρLI Psen ) 2ζ VI. P OWER A LLOCATION S CHEMES
Pfa = Q2 , (34)
N0 N0 Having accurate PA is essential to enhance the system
performance and the resource utilization. Therefore, in this
where ζ denotes the sensing threshold and Q(·, ·) represents
section, a pair of PA designs are proposed for the FD NO-DLT
the Marcum Q-function [48]. The PoD of the MiBS according
ISAC system considered, namely, SCD and CCD. Specifically,
to H1 can be expressed as [33]
   SCD aims for maximizing the receive SINR of the sensing
2(ρSR Pcom + ηρLI Psen + ρRR δPsen ) 2ζ signal at the MiBS under the IoT devices’ QoS constraints.
Pd = Q 5 , . In contrast, CCD aims for maximizing the IoT devices’ sum
2 N0 N0
rate under the constraints of minimum SINR requirements.
(35) It is worth stating that both SCD and CCD are of practi-
In the following, we adopt the receive signal of the MiBS for cal interest [27], [28]. According to (34) and (35), it can
inferring the presence/absence of the target. According to (2), be observed that increasing the SINR of the sensing signal
the power received at the MiBS can be expressed as improves the receive echo power and enhances the sensing
  performance. This means that for SCD, increasing the receive
E |ySR |2 = ρSR Pcom + ρRR δPsen + 2ηρLI Psen + N0 . (36) SINR of sensing signal improves the PoD and guarantees the
decoding requirements for the cooperative links. Moreover,
Naturally, E[|ySR |2 ], ρSR , Pcom , Psen , and N0 are known for the proportion of the sensing signal is fixed at a prescribed
the MiBS. Then, the information related to the target can be PoD, which indicates that for CCD, the sum rate is improved,
expressed as while meeting the sensing performance target. Indeed, these
  PA problems are meaningful for ISAC systems in practical
E |ySR |2 − 2ηρLI Psen − ρSR Pcom − N0
ρRR δ = . (37) scenarios from a range of different perspectives. To this end,
Psen the receive SINR of xr at the MiBS, which according to (2)
Apart from the target detection, the measurements of target can be expressed as
parameters are also significant. In this context, the distance
δρRR γr
between the target and the MiBS as well as the velocity of γSR
R
= . (40)
the target are analyzed according to the echo at the receiver. ρSR γc + 2ηρLI γr + 1
According to the time difference between transmitting the
A. SCD Scheme
sensing signal and receiving the target echo, the distance
between the MiBS and target may be expressed as [49] On the basis of (40), the SCD scheme can be optimized as
c maximize γSR
R
(41)
dRT = (τl + τs ) (38) Pcom ,Psen ,an ,bn
2
Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
6344 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO. 7, 1 APRIL 2023

 D D D

s.t. min γSRf , γRDf f , γRDf n ≥ γthf (41a) The problem can be solved by employing the subgradient
  method, and the subgradient at (λ11 , λ12 , λ13 ) is written as
Dn Dn
min γSR , γRD n
≥ γthn (41b)
d = [d11 , d12 , d13 ] (47)
0 ≤ Pcom ≤ Pmax , 0 ≤ Psen ≤ Pmax (41c)
an + af = 1, 0 < an < af < 1 where d11 = P∗com − (δρRR + 2ηρLI )θ P∗sen − N0 θ ,
d12 = P∗sen − C1 , and d13 = Pmax − P∗sen .
bn + bf = 1, 0 < bn < bf < 1 (41d)
Finally, let us turn our attention to the PA coefficients of an
where (41a) and (41b) indicate that xf and xn , respectively, and bn at fixed Pcom and Psen . Observe from (42) that γSR R is
can be decoded successfully from the cooperative links, (41c) ∗ ∗
independent of an and bn , while Pcom and Psen are related to
characterizes the power budget of the MaBS and MiBS, and an and bn upon combining with (40), (45), and (46), we can
(41d) guarantees the enforceability of NO-DLT. observe that γSRR is decreasing with respect to θ ∗ . By exploiting
Substituting (10)–(12) into (41a) and (41b), the PA problem this property, the PA problem can be reformulated to find the
can be rewritten as minimum θ ∗ . Observe from Corollaries 1 and 2 that θ1∗ is
maximize γSR
R
(42) monotonically increasing, while θ2∗ is decreasing with respect
Pcom ,Psen ,an ,bn to an . Therefore, θ ∗ can be expressed as
 ∗  
∗ θ2 , if an ∈ 0, a†n 
s.t. Pcom ≥ (δρRR + 2ηρLI )θ Psen + N0 θ θ = (48)
(42a) θ1∗ , if an ∈ a†n , 0.5
Psen ≥ C1 , (41c), (41d) (42b)
where a†n = γthn /(γthf + γthn + γthf γthn ). According to the
where θ = θ ∗ /ρSR , C1  max(C11 , C12 , C13 ), C11 = characteristics of θ1∗ and θ2∗ , θ ∗ can be minimized if and only
ϕ1∗ /ρRDf , C12 = ϕ1∗ /ρRDn , C13 = N0 γthn /bn ρRDn , and ϕ1∗ = if an = a†n .
N0 γthf /(bf − bn γthf ), while θ1∗ and θ ∗ can be obtained from Furthermore, we can obtain b†n = ρRDf γthn /(ρRDf γthn +
Corollaries 1 and 2, respectively. ρRDn γthf +ρRDf γthn γthf ) by using the same method. Therefore,
Observe from (42) that the PA coefficients are coupled, thus a∗n and b∗n are expressed as
it is difficult to obtain them simultaneously. In the following,  
we apply the popular alternating optimization (AO) technique a∗n = min a†n , 0.5 − σ (49)
 
and the expressions of Psen , Pcom , an , and bn are derived b∗n = min b†n , 0.5 − σ (50)
iteratively.
Substituting (40) into (42), we can observe that γSR R is a
where σ > 0 and σ → 0.
convex function, since as ∂ γSR /∂Psen < 0. Therefore, the
2 R 2
The AO algorithm of the SCD scheme is summarized in
Lagrangian function of (42) can be expressed as Algorithm 1, which iteratively determines the PA coefficients.
δρRR Psen Remark 6: For the SCD scheme, the computational com-
L1 (Psen , λ11 , λ12 , λ13 ) = plexity of the algorithm is on the order of O[(K1 + 2)K2 ],
ρSR Pcom + 2ηρLI Psen + 1
  where K1 is the number of iterations required for finding the
+ λ11 Pcom − (δρRR + 2ηρLI )θ Psen − N0 θ
optimal Lagrangian and K2 is the number of iterations in the
+ λ12 (Psen − C1 ) + λ13 (Pmax − Psen ) (43) main loop.
where λ11 , λ12 , and λ13 are the non-negative Lagrange
multipliers with respect to the constraints (42a)–(41c), respec- B. CCD Scheme
tively. Examining the Karush-Kuhn-Tucker (KKT) condition Now, we switch our focus to the CCD scheme. Based
for L(·) yields on (23) and (24), the CCD scheme can be optimized as
∂L1 δρRR (ρSR Pcom + 1) follows:
=
∂Psen (ρSR Pcom + 2ηρLI Psen + 1)2 maximize Rsum (51)
Pcom ,Psen ,an ,bn
− λ11 (δρRR + 2ηρLI )θ + λ12 − λ13 = 0. (44)
After some mathematical manipulations, the PA coefficient  D D D D D

P∗senwith the fixed multipliers is expressed as (45), shown at s.t. min γSRf , γRDf f , γRDf n , γSDf f , γSDf n ≥ γthf (51a)
 
the bottom of the page, where ()+  max(0, ) min γSRDn Dn
, γRD , γ Dn
Next, we optimize Pcom at fixed Psen , an , and bn , where the n SDn ≥ γthn (51b)
objective function in (42) is monotonically decreasing with γSR
R
≥ κ, (41c), (41d) (51c)
respect to Pcom . Herewith, according to (41c), (42a), and (45),
the PA coefficient P∗com is expressed as
  where (51a) and (51b) indicate that the communication signals
P∗com = min (δρRR + 2ηρLI )θ Psen + N0 θ , Pmax . (46) can be decoded successfully from the cooperative links, (51c)

 +
1 δρRR (ρSR Pcom + 1)
P∗sen = − (ρSR Pcom + 1) (45)
2ηρLI λ11 (δρRR + 2ηρLI )θ − λ12 + λ13

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6345

Algorithm 1: Alternate Optimization Algorithm C2 ≤ Pcom ≤ Pmax , C3 ≤ Psen ≤ Pmax

input : The maximum number of iterations L,  = 1, and (41d) (53d)
-λ > 0.
output: the optimal PA coefficients Psen , Pcom , an , and bn . where l2 = ρRR δ + 2ηρLI , l3 = af ρSR − an ρSR γthf , l4 =
1 Initialize P1sen , P1com , a1n , and b1n according to (39c) and (39d). δρRR − 2ηκρLI , C2 = max(C21 , C22 , C23 ) with C21 =
2 repeat N0 γthf /(af ρSDf −an ρSDf γthf ), C22 = N0 γthf / (af ρSDn − an ρSDn
3 =+1 γthf ), and C23 = N0 γthn /(an ρSDn ), C3 = max(C31 , C32 , C33 )
−1
4 Fixed an = a−1 −1
n , bn = bn , Pcom = Pcom , λ11  0, with C31 = N0 γthf /(bf ρRDf − bn ρRDf γthf ), C32 = N0 γthn /
λ12  0, and λ13  0.
5 repeat (bf ρRDn − bn ρRDn γthn ), and C32 = N0 γthn /(bn ρRDn ).
6 Set t = 1. Similar to (42), it is hard to simultaneously optimize Pcom ,
7 Derive the optimal λ∗11 , λ∗12 , and λ∗13 until convergence Psen , an , and bn , therefore we adopt AO to derive the PA
using the subgradient-based method. coefficients as well. By introducing the slack variable V and
8 Solve (43) for given λ∗11 , λ∗12 , and λ∗13 and denote the according to (52), we can rewrite (53) as (54), shown at the
solution as Pt+1sen . bottom of the page.
9 Update t = t + 1.
10 until The objective value of (43) reaches convergence;
Since (54) is a convex problem, the Lagrangian function
11 Update Psen = Ptsen . can be expressed as (55), shown at the bottom of the page.
12 Fixed Psen = Psen , update Pcom according to (44). It is easy to obtain that λ21 + λ22 = 1 and the KKT
13 Fixed Psen = Psen and Pcom = Pcom , update an and bn conditions are expressed as
according to (46) and (47) Update  =  + 1.
until γSRR () − γ R ( − 1)  — λ or  = L; ∂L λ21 an ρSDn
14
SR = − λ23 κρSR + λ24 l3
∂Pcom f3
+ λ25 an ρSR − λ26 + λ28 = 0 (56)
∂L (λ21 + λ22 )an ρRDn
is the receive SINR of xr constraint at the MiBS, κ denotes = + λ23 l4 − λ24 l2 γthf
∂Psen f2
the lower bound of the receive SINR and Rsum can be further
− λ25 l2 γthn − λ27 + λ29 = 0. (57)
expressed as (52), shown at the bottom of the page,8 where
l1 = ρSR + ρRR δ + 2ηρLI , f1 = an ρSR + ρRR δ + 2ηρLI , f2 = Based on (56) and (57), the PA coefficients P∗com and P∗sen
an ρRDn Psen + N0 , and f3 = an ρSDn Pcom + N0 . are, respectively, expressed as
Substituting (7)–(12) and (40) into (51a)–(51c), after some
!+
mathematical manipulations, (51) can be rewritten as λ21 N0
P∗com = − (58)
λ23 κρSR − λ24 l3 − λ25 an ρSR + λ26 − λ28 an ρSDn
maximize Rsum (53)  +
Pcom ,Psen ,an ,bn λ21 + λ22 N0
P∗sen = − .
λ24 l2 γthf − λ23 l4 + λ25 l2 γthn + λ27 − λ29 an ρSDn
s.t. an ρSR Pcom − l2 γthn Psen − N0 γthn ≥ 0 (53a) (59)
l3 Pcom − l2 γthf Psen − N0 γthf ≥ 0 (53b)
Similarly, the problem can be solved by adopting the
l4 Psen − κρSR Pcom − κN0 ≥ 0 (53c) subgradient method, and the subgradient at (λ21 , λ23 ,
8 It should be noted that due to the intractability of the sum rate, it is λ24 , λ25 , λ26 , λ27 , λ28 , λ29 ) is written as
extremely difficult to obtain the exact PA factors, thus we focus on the
approximate sum rate in the high-SNR regimes as in [37] and [50]. d = [d21 , d23 , d24 , d25 , d26 , d27 , d28 , d29 ] (60)

1  
Rsum = min(ln(l1 ) − ln(f1 ), − ln(bn )) + ln(f2 ) + ln(f3 ) − ln(an ) − 2 ln(N0 ) (52)
ln 2

maximize V (54)
Pcom ,Psen ,V

s.t. V ≤ ln(l1 ) − ln(f1 ) + ln(f2 ) + ln(f3 ) − ln(an ) − 2 ln(N0 ) (54a)

V ≤ − ln(bn ) + ln(f2 ) + ln(f3 ) − ln(an ) − 2 ln(N0 ), (53a)−(53d) (54b)

L(Pcom , Psen , V, λ21 , λ22 , λ23 , λ24 , λ25 , λ26 , λ27 , λ28 , λ29 ) = V + λ21 (ln(l1 ) − ln(f1 ) + ln(f2 ) + ln(f3 )
− ln(an ) − 2 ln(N0 ) − V) + λ22 (− ln(bn ) + ln(f2 ) + ln(f3 ) − ln(an ) − 2 ln(N0 ) − V)
  
+ λ23 (l4 Psen − κρSR Pcom − κN0 ) + λ24 l3 Pcom − l2 γthf Psen − N0 γthf + λ25 an ρSR Pcom − l2 γthn Psen

− N0 γthf + λ26 (Pmax − Pcom ) + λ27 (Pmax − Psen ) + λ28 (Pcom − C2 ) + λ29 (Psen − C3 ) (55)

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
6346 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO. 7, 1 APRIL 2023

where we have d21 = ln(l1 ) − ln(f1 ) + ln(bn ), d23 = l4 P∗sen − Algorithm 2: Alternate Optimization Algorithm
κρSR P∗com − κN0 , d24 = l3 P∗com − l2 γthf P∗sen − N0 γthf , d25 = input : The maximum number of iterations , ψ = 1, and
an ρSR P∗com − l2 γthn P∗sen − N0 γthf , d26 = Pmax − P∗com , d27 = † > 0.
ψ ψ ψ ψ
Pmax − P∗sen , d28 = P∗com − C2 , and d29 = P∗sen − C3 . output: the optimal PA coefficients Psen , Pcom , an , and bn .
Next, we optimize an and bn at a fixed Pcom and Psen . 1 Initialize P1sen , P1com , a1n , and b1n according to (39d) and (50d).
Substituting (7)–(12) into (53a) and (53b), after some mathe- 2 repeat
matical manipulations, the PA problem can be written as 3 ψ =ψ +1
ψ−1 ψ−1
4 Fixed an = an , bn = bn , 0  λ21  1, λ23  0,
maximize Rsum (61) λ24  0, λ25  0, λ26  0, λ27  0, λ28  0, and λ29  0.
an ,bn 5 repeat
6 Set q = 1.
s.t. C4 ≤ an ≤ C5 , C6 ≤ bn ≤ C7 , (41d) (61a) 7 Derive the optimal λ∗21 , λ∗23 , λ∗24 , λ∗25 , λ∗26 , λ∗27 , λ∗28 ,
and λ∗29 until convergence using the subgradient-based
where C4 = max(C41 , C42 ), C41 (l2 Psen γthn +N0 method.
γthn )/(ρSR Pcom ), C42 = N0 γthn /(ρSDn Pcom ), C5 = min 8 Solve (55) for given λ∗21 , λ∗23 , λ∗24 , λ∗25 , λ∗26 , λ∗28 and
(C51 , C52 , C53 ), C51 = (Pcom ρSR − l2 Psen γthf − N0 denote the solution as Pcom .
q+1
γthf )/[Pcom ρSR (1 + γthf )], C52 = g1 (ρSDf ), C53 = g1 (ρSDn ), 9 Solve (56) for given λ∗21 , λ∗23 , λ∗24 , λ∗25 , λ∗27 , λ∗29 and
C6 = N0 γthn /(ρRDn Psen ), C7 = min(C71 , C72 ), C71 = q+1
denote the solution as Psen .
g2 (ρRDf ), and C72 = g2 (ρRDn ). Then, g1 (ρ) and g2 (ρ) are, 10 Update q = q + 1.
respectively, given by 11 until The objective value of (55) and (56) reaches
convergence;
Pcom ρ − N0 γthf ψ q ψ q
g1 (ρ) =   (62) 12 Update Pcom = Pcom and Psen = Psen .
Pcom ρ 1 + γthf 13
ψ q ψ q
Fixed Pcom = Pcom and Psen = Psen , update an and bn
ψ ψ

Psen ρ − N0 γthf according to (61) and (62).

g2 (ρ) =  . (63) 14 Update ψ = ψ + 1.
Psen ρ 1 + γthf 15 until Rsum (ψ) − Rsum (ψ − 1)  † or ψ = ;
For the sake of further analysis, Rsum is divided into two
cases as (64), shown at the bottom of the page.
From (64), it can be inferred that the first-order derivative of
Rsum with respect of an in the high-SNR regime always satis-
fies ∂Rsum /∂an > 0. Furthermore, if ln(l1 ) − ln(f1 ) ≤ ln(b−1n ),
we can observe that ∂Rsum /∂bn = 0, and we can further
observe that ∂Rsum /∂bn < 0, if ln(l1 ) − ln(f1 ) > ln(b−1 n ).
Thus, Rsum can be maximized if and only if bn satisfies
ln(l1 ) − ln(f1 ) = ln(b−1
n ). Therefore, the PA coefficients an
∗
∗
and bn are, respectively, given by
a∗n = 0.5 − σ (65)
a∗ ρSR + ρRR δ + 2ηρLI
b∗n = n . (66)
ρSR + ρRR δ + 2ηρLI
Fig. 2. OPs of the IoT devices versus the transmit power of S.
The AO algorithm of the CCD scheme is summarized in
Algorithm 2, which iteratively determines the PA coefficients.
Remark 7: For the CCD scheme, the computational com-
dRDf = 200 m, dRDn = 150 m, dSDn = 200 m, dSDf = 250 m,
plexity of the algorithm is O[(K3 + 1)K4 ], where K3 denotes
and dRT = dTR = 75 m [38], [39], [40].
the number of iterations to find the optimal Lagrangian and
Fig. 2 presents the OPs of Dn and Df versus the transmit
K4 is the number of iterations in the main loop.
power of S with Pcom = Psen . For comparison, the OP of
the traditional noncooperative NO-DLT and cooperative HD
VII. N UMERICAL R ESULTS NO-DLT schemes are presented as well at the same power
In this section, some numerical results are provided for sup- consumption. It can be seen from Fig. 2 that the Monte Carlo
porting the theoretical analysis presented in Sections III– IV by simulation perfectly matches the analytical results and the
using Monte-Carlo computer simulations. Hereinafter, unless convergence between the curves in the high-SNR regime ver-
otherwise mentioned, the simulation parameters are as follows: ifies the accuracy of the results derived in (16), (19), (21),
af = bf = 0.7, an = bn = 0.3, γthf = 1, γthn = 2, δ = 0.2, and (22). We can also observe from Fig. 2 that the OPs of the
N0 = 1,  = 5, α = 2, η = 0.01, dLI = 0.5 m dSR = 100 m, IoT devices for the proposed FD NO-DLT ISAC scheme are

    −1 
ln 2 ln(l1 ) − ln(f1 ) + ln(f2 ) + ln(f3 ) − ln(an ) − 2 ln(N
 0 ) , if ln(l1 ) − ln(f1 ) ≤−1ln
1
bn
Rsum = (64)
ln 2 − ln(bn ) + ln(f2 ) + ln(f3 ) − ln(an ) − 2 ln(N0 ) , if ln(l1 ) − ln(f1 ) > ln bn
1

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6347

Fig. 3. ERs of the IoT devices versus the transmit power of S. Fig. 5. PoD versus the transmit power of the MiBS.


Df is dRDf = dSR 2 + d 2 − 2d d
SDf SR SDf cos φ. Observe from
Fig. 4 that for the cooperative NO-DLT schemes, the ESR of
the IoT devices improves first and then degrades upon increas-
ing dSR . The reason for this phenomenon can be attributed to
the fact that the increase of dSR improves the channel gains
between the relay and the IoT devices. However, at a shorter
dRDn , the path loss between the MaBS and the relay increases,
thus degrading the data transmission. Additionally, it can also
be seen that if R and Dn are particularly close to each other,
the ERs of Dn for the FD and HD NO-DLT become inferior
to that of noncooperative NO-DLT, since the cooperative NO-
DLT schemes have to share a portion of the total power for
target sensing.
Fig. 4. ERs of the IoT devices versus the relay location.
Fig. 5 shows the PoD of the MiBS versus the transmit
power of R at Pfa = 10−5 . Intuitively, the PoD can be
lower than the traditional noncooperative NO-DLT at the same improved by increasing the transmit power of R. It can be
energy consumption. This is due to the fact that cooperative observed that the PoD of the proposed NO-DLT ISAC system
NO-DLT transmissions are capable of significantly improving is superior to that of the traditional O-DLT, since NO-DLT
the frequency band exploitation. Furthermore, we notice that can serve multiple IoT devices at the same time, whereas
the OP of the proposed FD NO-DLT scheme is worse than O-DLT requires repeated communication signal transmission,
that of HD NO-DLT, mainly due to the additional CLSI. which further increases the interference during sensing. We
Fig. 3 demonstrates the ERs of Dn and Df versus the can further see from Fig. 5 that the sensing behavior can be
transmit power of S with Pcom = Psen . We reiterate that improved by reducing Pcom , which is due to the fact that the
the total power consumption of the three schemes is iden- signal transmitted by the MaBS is regarded as interference.
tical to ensure a fair comparison. It can be observed from Furthermore, due to the existence of supererogatory CLSI,
Fig. 3 that the ER of Dn keeps on increasing, while for the sensing performance of FD NO-DLT scheme lags behind
Df , the ER tends to be saturate eventually. In fact, the that of HD NO-DLT at the same parameter settings, which is
ER of the far IoT device is numerically expressed as consistent with the result of (35).
log2 (1 + af /an ) + log2 [1 + af βSR /(an βSR + 2βRR
2 δ + 2ηβ )]
LI Fig. 6 plots the receive SINR of xr at the MiBS versus the
for the FD NO-DLT, log2 (1 + af /an ) + power budget of the MaBS and the MiBS Pmax for both the
0.5log2 [1 + af βSR /(an βSR + 2βRR2 δ + 2ηβ )] for the HD
LI FD and HD modes. For comparison, we plot the curves of
NO-DLT, and log2 (1 + af /an ) for the noncooperative NO- the proposed PA scheme and three other random PA (RPA)
DLT in the high SNR regime. Moreover, it can be seen that baseline schemes, i.e.: 1) jointly optimizing Psen , an , and bn
the ER of the near IoT device would eventually govern the at fixed Pcom = 15 dBm; 2) jointly optimizing Pcom , an , and
ESR in the high-SNR regime. bn at fixed Psen = 15 dBm; and 3) jointly optimizing Pcom
Fig. 4 illustrates the ERs of Dn and Df versus the relay loca- and Psen at fixed an = bn = 0.2. For a fixed Pcom , the receive
tion for the FD, HD, and noncooperative schemes. Since the SINR of xr first increases with Pmax and then gradually sat-
near IoT device makes a significant contribution to the ESR, urating with a diminishing small return for an exceedingly
for tangibly characterizing the system performance, we assume large Pmax . This is because the increase of Pmax also increases
that S, R, and Dn are positioned along a straight line, and R Psen and further improves the receive SINR of xr . However,
is located between S and Dn . We set Pcom = Psen = 20 dBm, if the critical communication decoding capability is reached,
dSDn = 200 m, dSDf = 250 m, and the angle between S → Df no additional sensing power is allocated so as to cater for the
and S → Dn is set to φ = 30◦ , thus the distance between R and lowest signals decoding of communication, and the receive

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
6348 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO. 7, 1 APRIL 2023

and bn , restricted by the characteristics of the objective func-

tion and the decoding of communication signals, the sum rate
becomes inferior to that of the proposed PA scheme. Finally, as
expected, the sum rate of the FD NO-DLT system considered
is reduced by the CLSI compared to that of the HD mode.

VIII. C ONCLUSION
In this article, we studied an FD NO-DLT ISAC system,
in which the MiBS supports simultaneous target sensing and
cooperative relaying. We characterized the performance of
C&S by deriving both the exact and asymptotic OPs, the
diversity orders, the approximate ESR expressions for com-
Fig. 6. Receive SINR of xr versus the maximum power budget of the MaBS munication, the PoFA and PoD as well as the velocity and
and MiBS. distance between the MiBS and target for sensing. To fur-
ther characterize the performance of the system considered,
both the SCD and CCD, PA problems were solved. Then,
the maximization problems of the sensing signal’s receive
SINR at the MiBS and the IoT devices’ sum rate were deter-
mined by using the Lagrangian method. The simulation results
demonstrated that the FD NO-DLT system outperformed its
traditional noncooperative counterpart in terms of its OP at a
constant total power, while it was inferior to the HD NO-DLT
system. In terms of the ESR, the location of the MiBS had
a substantial influence on it and the FD NO-DLT revealed
its benefits. Furthermore, the sensing performance of the
proposed FD NO-DLT ISAC system is improved over that of
the traditional O-DLT scheme as well. Finally, we confirmed
Fig. 7. Sum rate versus the maximum power budget of the MaBS and MiBS. that the proposed PA scheme is superior to the RPA schemes.

SINR of xr remains unchanged. For a fixed Psen , the sensing A PPENDIX A

performance is degraded as Pcom increases, once the required P ROOF OF T HEOREM 1
decoding requirements are achieved, the receive SINR of xr Upon substituting (9) and (10) into (15), I1 can be rewritten
saturates. For the fixed an and bn , limited by the decod- as (67), shown at the bottom of the page.
ing requirements, the receive SINR of xr observed for the By using PDFs and CDFs of ρi , I1,I can be expressed as
RPA scheme lags behind that of the proposed PA scheme. βSR θ
− 1
Additionally, it can be observed from Fig. 5 that due to the I1,I = 1 − e βSR
2β (β + 2ηβ γr θ1 )
existence of CLSI, the receive SINR of the FD mode becomes  ∞RR SR √x xδγLI
1 − − r θ1
lower than that of the HD mode. × √ e βRR βSR dx . (68)
Fig. 7 shows the advantages of the proposed PA scheme 0 x
over other RPA baseline schemes in terms of the sum rate 
I1,I
attained in both the FD and HD modes. As seen from Fig. 7, √  can
the proposed PA scheme is superior to the schemes of the By defining t  x and using [51, eq. (3.322.2)], I1,I
fixed Pcom , Psen , or an . For a fixed Pcom , due to associated be further expressed as
the sensing constraint and the requirements of communication  ∞ δγ θ t2
 − t − r 1
signals decoding, the sum rate remains constant. Furthermore, I1,I =2 e βRR βSR dt
for a given Psen , the sum rate increases with Pmax . This phe- 0 β
  
nomenon continues until the communication scheme reaches π βSR 4β 2 SR
δγ θ
1 βSR
= e RR r 1 1 − erfc . (69)
its peak power at a fixed sensing demand. For the fixed an δγr θ1 2βRR δγr θ1

  
af ρSR γc af ρSDf γc
I1 = Pr < γthf Pr < γthf
an ρSR γc + ρRR δγr + 2ηρLI γr + 1 an ρSDf γc + 1
 
= Pr(ρSR < ρRR δγr θ1 + 2ηρLI γr θ1 + θ1 ) Pr ρSDf < θ1
 ∞  ∞  θ1 (xδγr +2yγr +1)  θ1
= fρSR (x)fρLI (y)fρRR (z)dxdydz fρSDf (w)dw (67)
0 0 0 0
I1,I I1,II

Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6349

Substituting the PDF of ρi into the second part of (67), after expressed as
some straightforward mathematical manipulations, I1 can be 
δγr ρRR 2ηγr ρLI 1
finally expressed as FW1 (w1 ) = 1 − Pr ρSR ≥ + +
 gc (w1 ) gc (w1 ) gc (w1 )
θ1 β θ 
−β βSR 2
SR −β1
I1 = 1−e SDf
1− e 4βRR δγr θ1 SR ρRDf ≥ gr (w1 ), ρRDn ≥ gr (w1 )
2βRR (βSR + 2ηβLI γr θ1 )
    
πβSR 1 βSR βSR gc (w1 ) π βSR gc (w1 )
× 1 − erfc . (70) = 1−
δγr θ1 2βRR δγr θ1 2βRR (βSR gc (w1 ) + 2βLI γr ) δγr
βSR gc (w1 )
Upon substituting (9), (10), and (12) into (15), and simi- − β g1 w − β 1
− 1
2 SR c ( 1 ) RDf gr (w1 ) βRDn gr (w1 )
lar to I1 , after some mathematical manipulations, I2 can be × e 4βRR δγr 
 
expressed as 1 gc (w1 )βSR
  × 1 − erfc . (77)
−β
θ1
−β
ϕ1
βSR 2βRR δγr
I2 = 1 − e 1−e
SDf RDf
2βRR (βSR + 2ηβLI γr θ1 ) Next, we can further observe that
βSR θ1
     ∞
2 δγ θ − βSR πβSR 1 βSR f ,1
Rave = log2 (1 + w1 )fW1 (w1 )dw1
×e 4βRR r 1 1 − erfc . (71)
δγr θ1 2βRR δγr θ1 0 
1 ∞ 1 − F (w )
W1 1
Substituting (70) and (71) into (15), after some straightfor- = dw1 . (78)
ln 2 0 1 + w1
ward mathematical manipulations, (16) can be obtained.
Substituting (77) into (78), we can obtain the expression
f ,1
A PPENDIX B of Rave . The same procedure can be adopted for the proof
f ,2
P ROOF OF C OROLLARY 1 of Rave . Then, we can obtain (27) after some mathematical
manipulations.
In the high-SNR regime of γc , γr → ∞, we can obtain Likewise, we can derive Rn,1 ave following the same approach
γthf as in (77).
lim θ1 = lim ≈0 (72) Dn
γc →∞ γc →∞ af γc − an γc γth
f For Rn,2 ave , denoting U  γSDn first, the CDF of U can be
γthf written as
lim ϕ1 = lim ≈0 (73)   − u
γr →∞ γr →∞ af γr − an γr γth
f FU (u) = Pr an ρSDn γc ≤ u = 1 − e βSDn an γc . (79)
γr γthf
lim γr θ1 = lim ≈ θ1∗ . (74) Similar to (78), and with the aid of [51, eq. (3.352.4)], the
γc ,γr →∞ γr →∞ af γc − an γc γth
f
exact expression of Rn,2
ave can be formulated as
Upon substituting (74) into (67), I1,I can be expressed as in  
1 βSD 1an γc 1
the high-SNR regime Rn,2
ave = − e n Ei − . (80)
ln 2 βSDn an γc
 ∞  ∞  xδθ ∗ +2yθ ∗
1 1
∞
I1,I = fρRR (x)fρLI (y)fρSR (z)dxdydz. (75) After some straightforward mathematical
0 0 0 manipulations, (28) can be obtained.
Similar to (68), after some mathematical manipulations, we
∞.
can obtain the expression of I1,I A PPENDIX D
When x → 0, the approximation 1 − ex ≈ x holds and P ROOF OF C OROLLARY 3
substituting (72) into (67), I1,II in the high SNR regime is Substituting (10), (12), and (13) into (23), and using the
∞ = θ /β
expressed as I1,II inequality [39]
1 SDf .
Likewise, we can obtain I2∞ by using the same method.  ! 
x E[x]
Substituting I1∞ and I2∞ into (15), (21) can be obtained. E log2 1 + ≈ log2 1 +   (81)
y Ey
A PPENDIX C f ,1
Rave can be approximated as (83), shown at the bottom of the
P ROOF OF T HEOREM 3
D D D page, where E[ρSR ] is expressed as
Upon defining W1  min(γSRf , γRDf f , γRDf n ) and W2   ∞
D D
min(γSDf f , γSDf n ), the achievable rate of Df can be rewritten as E[ρSR ] = xfρSR (x)dx = βSR . (82)
0
Rf = log2 (1 + W1 ) + log2 (1 + W2 ) . (76) Then, we can obtain E[ρRDf ], E[ρRDn ], E[ρRR ], and E[ρLI ]
ap,f 1
with the same method. Rave can be obtained after some
R1f R2f
mathematical manipulations.
ap,f 2 ap,n1 ap,n2
Substituting (10), (12), and (13) into W1 , with the Similarly, Rave , Rave , and Rave can be derived. Then,
aid of probability theory, the CDF of W1 can be the approximate ESR of the IoT devices can be obtained.

     
ap,f 1 1 af E ρRDf γr af E ρRDn γr af E[ρSR ]γc
Rave ≈ ln 1 + min   ,   , (83)
ln 2 an E ρRDf γr + 1 an E ρRDn γr + 1 an E[ρSR ]γc + E[ρRR ]δγr + 2ηE[ρLI ]γr + 1
Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
6350 IEEE INTERNET OF THINGS JOURNAL, VOL. 10, NO. 7, 1 APRIL 2023

R EFERENCES [24] Y. Sun, D. W. K. Ng, Z. Ding, and R. Schober, “Optimal joint power and
subcarrier allocation for full-duplex multicarrier non-orthogonal multiple
[1] F. Liu, C. Masouros, A. P. Petropulu, H. Griffiths, and L. Hanzo, “Joint access systems,” IEEE Trans. Commun., vol. 65, no. 3, pp. 1077–1091,
radar and communication design: Applications, state-of-the-art, and the Mar. 2017.
road ahead,” IEEE Trans. Commun., vol. 68, no. 6, pp. 3834–3862, [25] A. R. Chiriyath, B. Paul, G. M. Jacyna, and D. W. Bliss, “Inner bounds
Jun. 2020. on performance of radar and communications co-existence,” IEEE Trans.
[2] H. Hong, J. Zhao, T. Hong, and T. Tang, “Radar-communication inte- Signal Process., vol. 64, no. 2, pp. 464–474, Jan. 2016.
gration for 6G massive IoT services,” IEEE Internet Things J., pp. 1–11, [26] A. R. Chiriyath, B. Paul, and D. W. Bliss, “Radar-communications con-
Mar. 2021. vergence: Coexistence, cooperation, and co-design,” IEEE Trans. Cogn.
[3] A. Hassanien, M. G. Amin, E. Aboutanios, and B. Himed, “Dual- Commun. Netw., vol. 3, no. 1, pp. 1–12, Mar. 2017.
function radar communication systems: A solution to the spectrum [27] F. Wang and H. Li, “Joint power allocation for radar and commu-
congestion problem,” IEEE Signal Process. Mag., vol. 36, no. 5, nication co-existence,” IEEE Signal Process. Lett., vol. 26, no. 11,
pp. 115–126, Sept. 2019. pp. 1608–1612, Nov. 2019.
[4] F. Wang and H. Li, “Joint power allocation for Multicarrier radar and [28] F. Wang, H. Li, and M. A. Govoni, “Power allocation and co-design of
communication coexistence,” Proc. IEEE Int. Radar Conf. (RADAR), Multicarrier communication and radar systems for spectral coexistence,”
pp. 141–145, Jun. 2020. IEEE Trans. Signal Process., vol. 67, no. 14, pp. 3818–3831, Jul. 2019.
[5] T. Wild, V. Braun, and H. Viswanathan, “Joint design of communica- [29] W. Xu, X. Li, C.-H. Lee, M. Pan, and Z. Feng, “Joint sensing duration
tion and sensing for beyond 5G and 6G systems,” IEEE Access, vol. 9, adaptation, user matching, and power allocation for cognitive OFDM-
pp. 30845–30857, Feb. 2021. NOMA systems,” IEEE Trans. Wireless Commun., vol. 17, no. 2,
[6] F. Liu, C. Masouros, A. Li, H. Sun, and L. Hanzo, “MU-MIMO commu- pp. 1269–1282, Feb. 2018.
nications with MIMO radar: From co-existence to joint transmission,” [30] X. Liu and X. Zhang, “NOMA-based resource allocation for cluster-
IEEE Trans. Wireless Commun., vol. 17, no. 4, pp. 2755–2770, 2018. based cognitive industrial Internet of Things,” IEEE Trans. Ind.
[7] N. Chen and M. Okada, “Toward 6G Internet of Things and the con- Informat., vol. 16, no. 8, pp. 5379–5388, Aug. 2020.
vergence with RoF system,” IEEE Internet Things J., vol. 8, no. 11, [31] Q. Qi, X. Chen, C. Zhong, and Z. Zhang, “Integrated sensing, com-
pp. 8719–8733, Jun. 2021. putation and communication in B5G cellular Internet of Things,” IEEE
[8] F. Liu et al., “Integrated sensing and communications: Toward dual- Trans. Wireless Commun., vol. 20, no. 1, pp. 332–344, Jan. 2021.
functional wireless networks for 6G and beyond,” IEEE J. Sel. Areas [32] A. R. Chiriyath and D. W. Bliss, “Joint radar-communications
Commun., vol. 40, no. 6, pp. 1728–1767, Jun. 2022. performance bounds: Data versus estimation information rates,”
[9] L. Chettri and R. Bera, “A comprehensive survey on Internet of Things MILCOM Proc. IEEE Mil. Commun. Conf., pp. 1491–1496, Oct. 2015.
(IoT) toward 5G wireless systems,” IEEE Internet Things J., vol. 7, [33] A. R. Chiriyath, B. Paul, and D. W. Bliss, “Simultaneous radar detection
no. 1, pp. 16–32, Jan. 2020. and communications performance with clutter mitigation,” Proc. IEEE
[10] Q. Zhang, H. Sun, Z. Wei, and Z. Feng, “Sensing and communication Radar Conf. (RadarConf), pp. 279–284, Jun. 2017.
integrated system for autonomous driving vehicles,” IEEE INFOCOM [34] Z. Xiao and Y. Zeng, “Waveform design and performance analysis for
2020 - IEEE Conf. Comp. Commun. Workshops (INFOCOM WKSHPS), full-duplex integrated sensing and communication,” IEEE J. Sel. Areas
pp. 1278–1279, Aug. 2020. Commun., vol. 40, no. 6, pp. 1823–1837, Jun. 2022.
[11] X. Chen, Z. Feng, Z. Wei, F. Gao, and X. Yuan, “Performance of joint [35] Z. Wang, Y. Liu, X. Mu, Z. Ding, and O. A. Dobre, “NOMA empowered
sensing-communication cooperative sensing UAV network,” IEEE Trans. integrated sensing and communication,” IEEE Commun. Lett., vol. 26,
Veh. Technol., vol. 69, no. 12, pp. 15545–15556, Dec. 2020. no. 3, pp. 677–681, Jan. 2022.
[12] Z. Ding, P. Fan, and H. V. Poor, “Impact of user pairing on 5G [36] C. Ouyang, Y. Liu, and H. Yang, “On the performance of uplink ISAC
nonorthogonal multiple-access Downlink transmissions,” IEEE Trans. systems,” IEEE Commun. Lett., pp. 1–5, May 2022.
Veh. Technol., vol. 65, no. 8, pp. 6010–6023, Sept. 2016. [37] X. Li et al., “Cooperative wireless-powered NOMA relaying for B5G
[13] X. Chen, R. Jia, and D. W. K. Ng, “On the design of massive IoT networks with hardware impairments and channel estimation errors,”
non-orthogonal multiple access with imperfect successive interference IEEE Internet Things J., vol. 8, no. 7, pp. 5453–5467, Apr. 2021.
cancellation,” IEEE Trans. Commun., vol. 67, no. 3, pp. 2539–2551, [38] X. Yue, Y. Liu, S. Kang, A. Nallanathan, and Z. Ding, “Spatially random
Mar. 2019. relay selection for full/half-duplex cooperative NOMA networks,” IEEE
[14] Y. Liu et al., “Evolution of NOMA toward next generation multiple Trans. Commun., vol. 66, no. 8, pp. 3294–3308, Aug. 2018.
access (NGMA) for 6G,” IEEE J. Sel. Areas Commun., vol. 40, no. 4, [39] X. Li, M. Liu, C. Deng, P. T. Mathiopoulos, Z. Ding, and Y. Liu, “Full-
pp. 1037–1071, Apr. 2022. duplex cooperative NOMA relaying systems with i/Q imbalance and
[15] M. Vaezi, Z. Ding, and H. V. Poor, Multiple Access Techniques for 5G imperfect SIC,” IEEE Wireless Commun. Lett., vol. 9, no. 1, pp. 17–20,
Wireless Networks and Beyond. Eds. Cham, Switzerland: Springer, 2019. Jan. 2020.
[16] M. Sun, X. Xu, X. Tao, P. Zhang, and V. C. M. Leung, “NOMA-based [40] T. M. Hoang, L. T. Dung, B. C. Nguyen, X. N. Tran, and T. Kim,
D2D-enabled traffic offloading for 5G and beyond networks employing “Secrecy outage performance of FD-NOMA relay system with multiple
licensed and unlicensed access,” IEEE Trans. Wireless Commun., vol. 19, non-colluding eavesdroppers,” IEEE Trans. Veh. Technol., vol. 70,
no. 6, pp. 4109–4124, Jun. 2020. no. 12, pp. 12985–12997, Oct. 2021.
[17] H. Li, J. Li, M. Liu, Z. Ding, and F. Gong, “Energy harvesting and [41] V. Liu, A. Parks, V. Talla, S. Gollakota, D. Wetherall, and J. R. Smith,
resource allocation for cache-enabled UAV based IoT NOMA networks,” “Ambient backscatter: Wireless communication out of thin air,” ACM
IEEE Trans. Veh. Technol., vol. 70, no. 9, pp. 9625–9630, Sept. 2021. SIGCOMM Comput. Commun. Rev., vol. 43, no. 4, pp. 39–50, Sep. 2013.
[18] X. Chen, R. Jia, and D. W. K. Ng, “The application of relay to massive [42] A. Jee, K. Agrawal, and S. Prakriya, “A coordinated direct AF/DF
non-orthogonal multiple access,” IEEE Trans. Commun., vol. 66, no. 11, relay-aided NOMA framework for low outage,” IEEE Trans. Commun.,
pp. 5168–5180, Nov. 2018. vol. 70, no. 3, pp. 1559–1579, Mar. 2022.
[19] L. Bariah, S. Muhaidat, P. C. Sofotasios, F. E. Bouanani, O. A. Dobre, [43] L. Yuan, N. Yang, F. Fang, Q. Du, and Z. Zheng, “Optimal power allo-
and W. Hamouda, “Large intelligent surface-assisted nonorthogonal cation for finite blocklength cooperative NOMA with coordinated direct
multiple access for 6G networks: Performance analysis,” IEEE Internet and relay transmission,” IEEE Wireless Commun. Lett., vol. 11, no. 3,
Things J., vol. 8, no. 7, pp. 5129–5140, Apr. 2021. pp. 523–527, Mar. 2022.
[20] X. Li, J. Li, Y. Liu, Z. Ding, and A. Nallanathan, “Residual transceiver [44] H. Liu, Z. Ding, K. J. Kim, K. S. Kwak, and H. V. Poor, “Decode-
hardware impairments on cooperative NOMA networks,” IEEE Trans. and-forward relaying for cooperative NOMA systems with direct links,”
Wireless Commun., vol. 19, no. 1, pp. 680–695, Jan. 2020. IEEE Trans. Wireless Commun., vol. 17, no. 12, pp. 8077–8093,
[21] L. Zhang, J. Liu, M. Xiao, G. Wu, Y.-C. Liang, and S. Li, “Performance Dec. 2018.
analysis and optimization in downlink NOMA systems with cooperative [45] X. Li, M. Liu, D. Deng, J. Li, C. Deng, and Q. Yu, “Power beacon
full-duplex relaying,” IEEE J. Sel. Areas Commun., vol. 35, no. 10, assisted wireless power cooperative relaying using NOMA with hard-
pp. 2398–2412, Oct. 2017. ware impairments and imperfect CSI,” AEU Int. J. Electron. Commun.,
[22] T. Riihonen, S. Werner, and R. Wichman, “Mitigation of loopback self- vol. 108, pp. 275–286, Aug. 2019.
interference in full-duplex MIMO relays,” IEEE Trans. Signal Process., [46] M. A. Richards, J. A. Sheer, and W. A. Holm, Principles of Modern
vol. 59, no. 12, pp. 5983–5993, Dec. 2011. Radar: Basic Principles. Raleigh, North Carolina: SciTech Publishing,
[23] D. Zhang, Y. Liu, L. Dai, A. K. Bashir, A. Nallanathan, and 2010.
B. Shim, “Performance analysis of FD-NOMA-based decentralized [47] Z. Zhang, B. Chen, and M. Yang, “Moving target detection based on
V2X systems,” IEEE Trans. Commun., vol. 67, no. 7, pp. 5024–5036, time reversal in a Multipath environment,” IEEE Trans. Aerosp. Electron.
Jul. 2019. Syst., vol. 57, no. 5, pp. 3221–3236, Apr. 2021.
Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: PERFORMANCE ANALYSIS AND POWER ALLOCATION FOR COOPERATIVE ISAC NETWORKS 6351

[48] D. W. Bliss and S. Govindasamy, Adaptive Wireless Communications: Arumugam Nallanathan (Fellow, IEEE) received
MIMO Channels and Networks. New York, New York: Cambridge the Ph.D. degree from the University of Hong
University Press, 2013. Kong, Hong Kong, in 2000.
[49] M. A. Richards, Fundamentals of Radar Signal Processing. New York, He has been a Professor of Wireless
New York: McGraw-Hill, 2005. Communications and the Head of the
[50] O. Abbasi, H. Yanikomeroglu, A. Ebrahimi, and N. M. Yamchi, Communication Systems Research Group,
“Trajectory design and power allocation for drone-assisted NR-V2X School of Electronic Engineering and Computer
network with dynamic NOMA/OMA,” IEEE Trans. Wireless Commun., Science, Queen Mary University of London,
vol. 19, no. 11, pp. 7153–7168, Nov. 2020. London, U.K., since September 2017. He was with
[51] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and the Department of Informatics, King’s College
Products, 7th ed. New York, NY, USA: Academic, 2007. London, London, from December 2007 to August
2017, where he was a Professor of Wireless Communications from April
2013 to August 2017 and a Visiting Professor from September 2017. He
Meng Liu (Graduate Student Member, IEEE) was an Assistant Professor with the Department of Electrical and Computer
received the M.S. degree from Henan Polytechnic Engineering, National University of Singapore, Singapore, from August 2000
University, Jiaozuo, China, in 2020. He is cur- to December 2007. He published nearly 500 technical papers in scientific
rently pursuing the Ph.D. degree with the School journals and international conferences. His research interests include artificial
of Electronic Engineering, Xidian University, Xi’an, intelligence for wireless systems, beyond 5G wireless networks, Internet of
China. Things, and molecular communications.
His research interests include integrated sensing Prof. Nallanathan is a co-recipient of the best paper awards presented
and communication, nonorthogonal multiple access, at the IEEE International Conference on Communications 2016, IEEE
and cooperative communications. Global Communications Conference 2017, and IEEE Vehicular Technology
Conference 2018. He received the IEEE Communications Society SPCE
Outstanding Service Award 2012 and the IEEE Communications Society RCC
Outstanding Service Award 2014. He is an IEEE Distinguished Lecturer. He
Minglei Yang (Member, IEEE) received the B.E.
has been selected as a Web of Science Highly Cited Researcher, since 2016.
degree in electronic engineering and the Ph.D.
He is an Editor of the IEEE T RANSACTIONS ON C OMMUNICATIONS and a
degree in signal and information processing from
Senior Editor of the IEEE W IRELESS C OMMUNICATIONS L ETTERS. He was
Xidian University, Xi’an, China, in 2004 and 2009,
an Editor for the IEEE T RANSACTIONS ON W IRELESS C OMMUNICATIONS
respectively.
from 2006 to 2011, IEEE T RANSACTIONS ON V EHICULAR T ECHNOLOGY
He has been with the National Laboratory of
from 2006 to 2017, and IEEE Signal Processing Letters. He served as the
Radar Signal Processing, Xidian University, since
Chair for the Signal Processing and Communication Electronics Technical
2009, where he is currently a Full Professor. He
Committee of IEEE Communications Society and Technical Program
was a Visiting Scholar with the Elisha Yegal Bar-
Chair and member of Technical Program Committees in numerous IEEE
Ness Center for Wireless Communications and
conferences.
Signal Processing Research, New Jersey Institute
of Technology, Newark, NJ, USA, from 2014 to 2015. He has
authored/coauthored more than 100 peer-reviewed journals and conference
papers and more than 60 inventions. His research interests include integrated
sensing and communication, radar signal processing, array signal processing,
MIMO radar, and distributed array signal processing. Guangjian Wang (Member, IEEE) received the
M.Eng. degree from Dalian Maritime University,
Huifang Li received the B.S. degree from North Dalian, China, in 2007.
University of China, Taiyuan, China, in 2017. She is He joined Huawei Technologies Company Ltd.,
currently pursuing the Ph.D. degree in communica- Chengdu, China, in April 2007, where he is currently
tion and information system with Xidian University, a Senior Research Expert with the 2012 Laboratory.
Xi’an, China. Over the past 15 years, he has conducted various
Her research interests include nonorthogonal researches and prototype verifications in wireless
multiple access, cooperative communications, and communications, especially new system architecture
physical-layer security. and RF technologies for millimeter-wave/terahertz
frequency communication and sensing. He has led
the research and development of multiple millimeter-wave and terahertz com-
munication systems and prototypes with the highest performance in the
industry, as well as integrated sensing and communication systems and proto-
Kun Zeng (Member, IEEE) received the Ph.D. types. His current research interests include advanced wireless technologies,
degree in electrical engineering from the University millimeter-wave and terahertz communication, and sensing technologies.
of Electronic Science and Technology of China,
Chengdu, China, in 2011.
He is currently a Research Expert with
the Wireless Technology Laboratory, Huawei
Technologies Company Ltd., Chengdu, India.
Lajos Hanzo (Life Fellow, IEEE) received the
He has long been engaged in 5G/6G research
master’s and Ph.D. degrees from the Technical
and 3GPP/ITU standardization, and has more
University (TU) of Budapest, Budapest, Hungary, in
than 40 granted U.S. patents. His research
1976 and 1983, respectively, the Doctor of Sciences
interests include advanced wireless communication
degree from the University of Southampton,
technologies, networks, and standardization.
Southampton, U.K., in 2004, the first Ph.D. degree
(Hons.) from TU of Budapest in 2009, and the sec-
Zhaoming Zhang received the B.E. degree in elec- ond Ph.D. degree (Hons.) from the University of
tronic and information engineering from Xidian Edinburgh, Edinburgh, U.K., in 2015.
University, Xi’an, China, in 2017, and the Ph.D. He has published more than 2000 contributions at
degree from the School of Electronic Engineering, IEEE Xplore, 19 Wiley–IEEE Press books, and has
Xidian University in 2022. helped the fast-track career of 123 Ph.D. students. Over 40 of them are pro-
He is currently an Engineer with Xi’an Modern fessors at various stages of their careers in academia and many of them are
Control Technology Research Institute, Xi’an. His leading scientists in the wireless industry.
current research interests include joint radar and Dr. Hanzo is a Foreign Member of the Hungarian Academy of Sciences
communication, time-reversal technology, radar tar- and the former Editor-in-Chief of the IEEE Press. He is also a Fellow of the
get detection, multistatic radar, and automatic target Royal Academy of Engineering, IET, and EURASIP. He has served several
recognition. terms as a Governor of both IEEE ComSoc and VTS.
Authorized licensed use limited to: SHIV NADAR UNIVERSITY. Downloaded on October 09,2024 at 07:21:32 UTC from IEEE Xplore. Restrictions apply.