Professional Documents
Culture Documents
Abstract—In this paper, transmission through large intelligent belong to the vast IM family [7], use the variations in the
surfaces (LIS) that intentionally modify the phases of incident signatures of received signals by exploiting reconfigurable
waves to improve the signal quality at the receiver, is put forward antennas or scatterers to transmit additional information bits
as a promising candidate for future wireless communication
systems and standards. For the considered LIS-assisted system, a in rich scattering environments. On the other hand, large
general mathematical framework is presented for the calculation intelligent surfaces/walls/reflect-arrays/metasurfaces are smart
of symbol error probability (SEP) by deriving the distribution of devices that control the propagation environment with the aim
the received signal-to-noise ratio (SNR). Next, the new concept of improving the coverage and signal quality [8].
of using the LIS itself as an access point (AP) is proposed. It is worth noting that the large intelligent surface (LIS)-
Extensive computer simulation results are provided to assess
the potential of LIS-based transmission, in which the LIS acts based transmission concept is completely different from ex-
either as an intelligent reflector or an AP with or without the isting MIMO, beamforming, amplify-and-forward relaying,
knowledge of channel phases. Our findings reveal that LIS- and backscatter communication paradigms, where the large
based communications can become a game-changing paradigm number of small, low-cost, and passive elements on a LIS
for future wireless systems. only reflect the incident signal with an adjustable phase
Index Terms—Beyond massive MIMO, error probability anal-
ysis, large intelligent surface (LIS), signal-to-noise ratio, smart shift without requiring a dedicated energy source for RF
reflect-array, software-defined surface. processing, decoding, encoding, or retransmission. Inspired
by the definition of software-defined radio, which is given as
I. I NTRODUCTION “radio in which some or all of the physical layer functions
The first commercial fifth generation (5G) wireless networks are software defined” and considering the interaction of the
have been already deployed in certain countries while the first intelligent surface with incoming waves in a software-defined
5G compatible handsets are expected to be available during fashion, we may also use the term of software-defined surface
2019. Although the initial stand-alone 5G standard, which (SDS) for these intelligent surfaces.
brings more flexibility into the system design by exploiting The concept of intelligent walls is proposed in one of the
millimeter-waves and multiple orthogonal frequency division early works by utilizing active frequency selective surfaces to
multiplexing numerologies, has been completed during 2018, control the signal coverage [9]. Alternative to beamforming
researchers are relentlessly exploring the potential of emerging techniques that require large number of antennas to focus the
technologies for later releases of 5G. These potential technolo- transmitted or received signals, the concept of smart reflect-
gies include non-orthogonal multiple access, optical wireless arrays is proposed in [10]. It has been also demonstrated that
communications and hybrid optical/radio frequency (RF) solu- reflect-arrays can be used effectively to change the phase of
tions, alternative waveforms, low-cost massive multiple-input reflected signals without buffering or processing the incoming
multiple-output (MIMO) systems, terahertz communications, signals and the received signal quality can be enhanced by
and new antenna technologies. Even though future 6G tech- adjusting the phase shift of each element on the reflect-
nologies look like as the extension of their 5G counterparts array. As an evolution of massive MIMO systems, the LIS
at this time [1], new user requirements, completely new concept is proposed in [11] by exploiting the whole contiguous
applications/use-cases, and new networking trends of 2030 and surface for transmitting and receiving. The authors of [12]–
beyond may bring more challenging communication engineer- [14] focused on a downlink transmission scenario through
ing problems, which necessitate radically new communication a LIS to support multiple users and investigated sum-rate
paradigms in the physical layer. and energy efficiency maximization problems. Low complexity
Within this context, there has been a growing interest in algorithms are also considered for the encountered non-convex
controlling the propagation environment in order to increase optimization problems to obtain the optimum reflector phases.
the quality of service for wireless communications. Schemes Recently, a joint active and passive beamforming problem is
such as media-based modulation [2]–[4], spatial scattering investigated in [15] and [16], and the user’s average received
modulation [5], and beam index modulation (IM) [6], which power is investigated.
Against this background, this paper first provides a mathe-
This work was supported in part by the Scientific and Technological matical framework for the error performance analysis of LIS-
Research Council of Turkey (TUBITAK) under Grant 117E869, the Turkish
Academy of Sciences (TUBA) GEBIP Programme, and the Science Academy based communication systems. For the first time in the litera-
BAGEP Programme. Codes available at https://corelab.ku.edu.tr/tools. ture, we investigate the effect of number of reflecting elements,
113
100 10-1
P e (exact) N=4
N=8
P e (upper-bound)
10-2 N=16
P e (AWGN) 10-2 N=32
N=64
-4
10 N=128
N=256
10-3 Theo.
10-6
BEP
BER
N=16
10-8
10-4
10-10
10-5
10-12 N=32
10-14 10-6
-30 -20 -10 0 10 20 30 -40 -30 -20 -10 0 10
SNR(dB) SNR(dB)
Fig. 2. Theoretical average BEP of the LIS-based scheme for N = 16 and Fig. 3. Simulated BER performance of the LIS-based scheme with varying
N = 32 with BPSK. number of reflecting elements for BPSK with theoretical results of (6).
constellations as [18]
N = 32 with respect to Es /N0 . As seen from Fig. 2, the LIS- Z π/2
based scheme achieves significantly better BEP performance 4 1 −3
Pe = 1− √ Mγ dη
compared to the classical BPSK scheme operating over the π M 0 2(M − 1) sin2 η
pure AWGN channel. In other words, a LIS can convert a 4
1
2 Z π/4
−3
hostile wireless fading environment into a super communica- − 1− √ Mγ dη. (10)
π M 0 2(M − 1) sin2 η
tion channel that provides very low BEP at extremely low
SNR values through the smart adjustment of reflector phases. Removing the integrals by letting η = π/2 and η = π/4
The following remark explains this phenomenon. in the first and second terms of (10), we can obtain a tight
Remark: As seen from Fig. 2, the average BEP curves have upper-bound on the average SEP. Under the assumption of
N Es
a waterfall region and a saturation region. We observe that for N0 10 (at the SNR region of interest), the average SEP
N Es can be expressed as
N0 10, from (7), Pe becomes proportional to
3N 2 π 2 Es
Pe ∝ exp − (11)
N 2 π 2 Es 32(M − 1)N0
Pe ∝ exp − (8)
16N0 where we ignored the second exponential term coming from
(10) due to its relatively larger exponent. Since M appears in
which explains the superior BEP performance of the LIS-based the exponent of (11), the LIS-based scheme also suffers from
scheme. In this region, although the SNR (Es /N0 ) is relatively a degradation in error performance with increasing modulation
low, due to the N 2 term in the exponent, considerably low orders although benefiting from the N 2 term.
BEP values are possible, particularly with increasing N . On
B. Blind Transmission Through LIS
the other hand, for NNE0s 1, (7) can be approximated as
In this case, the LIS given in Fig. 1 does not have the
− 12 knowledge of channel phases θi and ψi , and consequently,
N (16 − π 2 )Es N π2
Pe ∝ exp − (9) cannot eliminate these phase terms to maximize the received
8N0 2(16 − π 2 ) SNR. Without loss of generality, assuming φi = 0 for i =
1, 2, . . . , N , the received signal becomes1
which explains the saturated BEP performance for high SNR "N #
values due to − 12 exponent of the SNR. However, the average r=
X
hi gi x + n = Hx + n. (12)
BEP still decays exponentially with respect to N and signifi- i=1
cant reductions are possible in Pe by increasing N .
For this blind scheme, the CLT can be also applied for large
In Fig. 3, we show the bit error rate (BER) performance N , and considering H ∼ CN (0, N ), the MGF of the received
of the LIS-based scheme for different number of reflecting SNR is obtained as Mγ (s) = (1 − sNNE s −1
) . Following the
0
elements (N ) and BPSK signaling. As seen from Fig. 3, our same steps above, BEP of the blind LIS-based scheme can be
theoretical approximation in (6) using the CLT is considerably expressed for binary signaling as
accurate for increasing N values. Furthermore, we observe that v
Z π/2 ! u N Es
doubling N provides approximately 6 dB improvement (four- 1 1 1 u
fold decrease) in the required SNR at the waterfall region to Pe = dη = 1− t NN0 Es (13)
π 0 1+ sinN2EηN
s 2 1+ N0
achieve a target BER, which can be easily verified from (8). 0
Using the MGF of the received instantaneous SNR (Mγ (s)), 1 It is worth noting that the case of N = 1 is equivalent to the well-known
we can also obtain the average SEP for square M -QAM cascaded Rayleigh fading.
114
LIS adjusting reflector phases as φi = ψi + wm , where wm , m ∈
{1, 2, . . . , M } is the common additional phase term induced
by the LIS to carry the information of the mth message. In
gi light of this, the received signal can be expressed as
"N #
p X p
r = Es βi ejwm + n = Es Bejwm + n. (15)
i=1
115
Two main results can be inferred from (21). First, the pro- binary and M -ary signaling respectively yields the following
posed concept in which the LIS acts as an AP, can convey average BEP and SEP expressions:
information in an ultra-reliable manner as the dual-hop (DH) !
1 π/2
Z
LIS scheme given in Fig. 1. Second, by comparing (8) and 1
Pe = dη (26)
(21), to achieve a target BEP, around 1 dB improvement in π 0 1 + sinN2EηN
s
0
the required SNR can be obtained by the proposed concept
and
compared to the LIS-based DH scheme for binary signaling. !
Z (M −1)π/M
For M -ary signaling, substituting (19) in (5), we obtain the 1 1
Pe = dη. (27)
average SEP in the form of a definite integral as follows: π 0 1+ N sin2 (π/M )Es
sin2 ηN0
!21
1 (M −1)π/M It is worth noting that similar to the blind LIS-assisted DH
Z
1
Pe = sin2 (π/M )Es scheme, only an N times SNR gain can be obtained compared
π 0 1 + N (4−π) 2 sin2 ηN0
2 2
to point-to-point transmission over Rayleigh fading channels.
− N π4sin 2
(π/M )Es
sin ηN0
This proves that a LIS can be used as an AP as well by only
× exp sin2 (π/M )Es
dη. (22) adjusting reflector-induced phases according to the data.
1 + N (4−π)2 sin2 ηN0
IV. S IMULATION R ESULTS
Upper bounding this result by letting η = π/2 and focusing
on the SNR range of interest, we obtain In this section, we provide computer simulation results for
the LIS-based new (LIS-AP) scheme and make comparisons
N 2 πEs
2
Pe ∝ exp − sin (π/M ) . (23) with the LIS-assisted DH (LIS-DH) scheme. In all simula-
4N0 tions, we assume uncorrelated Rayleigh fading channels and
Comparing this result with (11), we conclude that a loss can be consider Es /N0 as the SNR, similar to the classical diversity
expected in the required SNR for higher order signaling (M ≥ combining schemes.
16) due to the SNR loss of M -PSK compared to M -QAM. In Fig. 5, we present the BER performance of LIS-DH and
However, as will be shown in next section, this loss becomes LIS-AP schemes for different number of reflecting elements
insignificant considering the potential of the new approach and (N ) and BPSK signaling along with theoretical curves of (20).
the relatively low SNR ranges of interest with increasing N . As seen from Fig. 5, a LIS can be effectively used as an AP
by providing ultra-reliable communications. Furthermore, as
B. Blind Access Point-LIS verified from (21), around 1 dB SNR improvement can be
In this worst-case scenario, the LIS does not have the obtained compared to the LIS-DH scheme when M = 2.
knowledge of channel phases ψi and plays the role of a data In Fig. 6, we evaluate the symbol error rate (SER) perfor-
source by simply adjusting its reflector-induced phase terms mance of LIS-DH and LIS-AP schemes for varying signaling
in a similar fashion to PSK. To be generalized later, let us orders M ∈ {4, 16, 64} with 64 reflectors. Theoretical SEP
focus on the simplest case of binary signaling, in which the curves obtained from (10) and (22) are also shown in the
reflector-induced phases of the LIS are adjusted for messages same figure to check the accuracy of our theoretical findings.
1 and 2 as follows: φi = ω1 and φi = ω2 for all i. For this As seen from Fig. 6, both schemes suffer from a degradation
scheme, again assuming that the RF source is close enough in error performance with increasing M , while this is more
to the LIS and its transmission is not affected by fading, the noticeable for the LIS-AP scheme due to the SNR loss of
received signal in the baseband becomes M -PSK over M -QAM for M ≥ 16.
"N # In Fig. 7, we show the BER performance of the blind LIS-
p X p DH and LIS-AP schemes for different number of reflectors
r = Es gi ejωm + n = Es Gejωm + n (24) and BPSK signaling. For comparison, theoretical curves ob-
i=1
tained from (13) are also shown. It is worth noting that both
where m ∈ {1, 2}. For this signal model, following a similar LIS-DH and LIS-AP schemes have the same received SNR
analysis, the CPEP can be obtained as follows: distribution for blind transmission, and provide N times SNR
s gain compared to point-to-point signaling over Rayleigh fading
2
E s |G| (1 − cos(ω2 − ω ))
1 channels. As seen from Fig. 7, doubling N provides a 3 dB im-
Pe|G = Q . (25)
N0 provement in the required SNR to achieve a target BER value.
We also note that although improvements are possible with
This CPEP expression also requires uniformly distributed increasing N , the clear advantage of using a LIS diminishes
phases around the unit circle for the minimization of the SEP. when the intelligence of the surface is not exploited through
For instance, selecting w1 = 0 and w2 = π as in BPSK will phase removal in this worst-case transmission scenario.
be optimum for binary signaling in terms of BEP.
Noting that G ∼ CN (0, N ) under the CLT, the MGF of V. C ONCLUSIONS
2
the instantaneous received SNR γ = |G| Es /N0 becomes In this study, we have evaluated the potential of LIS-assisted
Mγ (s) = (1 − sNNE 0
s −1
) . Then, substituting Mγ (s) in (5) for communications from an error performance perspective and
116
10-1 10-1
LIS-DH,N=32 N=4
LIS-DH,N=64 N=8
LIS-DH,N=128 N=16
10-2 LIS-DH,N=256 10-2 N=32
LIS-AP,N=32 N=64
LIS-AP,N=64 N=128
LIS-AP,N=128 N=256
10-3 LIS-AP,N=256 10-3 Theo.
LIS-AP,Theo.
BER
BER
10-4 10-4
10-5 10-5
10-6 10-6
-45 -40 -35 -30 -25 -20 -15 -5 0 5 10 15 20 25 30 35 40 45 50
SNR(dB) SNR(dB)
Fig. 5. BER performance of LIS-DH and LIS-AP schemes with varying Fig. 7. BER performance of LIS-DH and LIS-AP schemes with blind
number of reflectors for BPSK. transmission and varying number of reflectors for BPSK.
10-1
LIS-DH,M=4
LIS-DH,M=16
[4] E. Basar, “Media-based modulation for future wireless systems: A
LIS-DH,M=64 tutorial,” IEEE Wireless Commun. (to appear), Mar. 2019. [Online].
10-2 LIS-DH,Theo. Available: https://arxiv.org/abs/1811.08730
LIS-AP,M=4 [5] Y. Ding, K. J. Kim, T. Koike-Akino, M. Pajovic, P. Wang, and P. Orlik,
LIS-AP,M=16
LIS-AP,M=64
“Spatial scattering modulation for uplink millimeter-wave systems,”
10-3 LIS-AP,Theo. IEEE Commun. Lett., vol. 21, no. 7, pp. 1493–1496, July 2017.
[6] Y. Ding, V. Fusco, A. Shitvov, Y. Xiao, and H. Li, “Beam index
SER
117