Journal of Signal Processing Systems (2023) 95:435–457

https://doi.org/10.1007/s11265-022-01827-7

Signal Processing Techniques for 6G

Lorenzo Mucchi1 · Shahriar Shahabuddin2,5 · Mahmoud A. M. Albreem3 · Saeed Abdallah3 · Stefano Caputo1 ·
Erdal Panayirci4 · Markku Juntti5

Received: 17 February 2022 / Revised: 7 August 2022 / Accepted: 12 December 2022 / Published online: 2 February 2023
© The Author(s) 2023

Abstract
6G networks have the burden to provide not only higher performance compared to 5G, but also to enable new service domains
as well as to open the door over a new paradigm of mobile communication. This paper presents an overview on the role and
key challenges of signal processing (SP) in future 6G systems and networks from the conditioning of the signal at transmis-
sion to MIMO precoding and detection, from channel coding to channel estimation, from multicarrier and non-orthogonal
multiple access (NOMA) to optical wireless communications and physical layer security (PLS). We describe also the core
future research challenges on technologies including machine learning based 6G design, integrated communications and
sensing (ISAC), and the internet of bio-nano-things.

Keywords 6G · Signal processing · MIMO · Optical wireless communications · Internet of bio nano things

1 Introduction The global data traffic is foreseen to be more than

5000 EB/month in 2030. The value may even increase due
Wireless communications networks have evolved to per- to the increasing worldwide use of remote digital services
vasive and ubiquitous enablers for modern societies. The driven by the Covid-19 pandemic. The future society will
first generation (1G) networks were introduced in 1980s and require fully automated and connected systems. Those will
since then a new generation has emerged every ten years. use distributed artificial intelligence (AI) and machine
Each generation from the analog 1G to the current 5G has learning (ML), ultra-dense sensors, fast computing, fully
provided new service features via new technology enablers. integrated heterogeneous connectivity, etc. Those services
The service and quality metrics have focused on improving will consume huge amounts of data, which needs to be
data rate, reliability, quality, security, and more. While 5G transferred and to large extent to or from mobile sources.
has introduced some initial versions of distributed or edge Emerging Internet of Everything (IoE) applications will
intelligence to the system design, their actual breakthrough require the convergence of communications, sensing, con-
is expected with 6G systems. The technology for 6G net- trol, and computing functionalities. Some attempts toward
works is now under intensive research and the networks are this direction have been made in 5G networks already.
expected be enrolled in 2030s.

* Lorenzo Mucchi Markku Juntti

[email protected] [email protected]
Shahriar Shahabuddin 1
Dept. of Information Engineering, University of Florence,
[email protected]
Florence 50139, Italy
Mahmoud A. M. Albreem 2
Mobile Networks, Nokia, Dallas 75019, TX, USA
[email protected]
3
Dept. of Electrical Engineering, University of Sharjah,
Saeed Abdallah
Sharjah 27272, UAE
[email protected]
4
Dept. of Electrical and Electronics Engineering, Kadir Has
Stefano Caputo
Univ., Istanbul 34083, Turkey
[email protected]
5
Centre for Wireless Communications, University of Oulu,
Erdal Panayirci
Oulu 90014, Finland
[email protected]

13
Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

436 Journal of Signal Processing Systems (2023) 95:435–457

The stronger demand of high data rates with low latency augmented reality) to interactive robotics and AI-based
and low energy consumption, will require the use of THz autonomous systems, from wireless brain-computer inter-
spectrum in the radio frequency (RF) domain as well as action to haptic communications, from in-body communica-
the use of optical wireless communications (OWC), and in tions to human-bond communications, from massive Internet
particular of visible light communications (VLC). of everything to umanned mobility.
Although 5G has started the journey to address the
challenges described above, many open problems remain. 1.1 Main 6G System Technologies and Architectures
We need to investigate, e.g., the following: 1) How to uti-
lize higher frequencies with larger bandwidths and lower Since 6G research and projects emerged, papers started to
power communications? 2) How to satisfy the new require- appear in scientific literature, in particular overview papers.
ments: ultra low latency and ultra reliable communica- In [1], a survey of 6G from the point of view of energy con-
tions? 3) How to guarantee security, also in low-resourced sumption and green architectures and technologies is reported.
(but fully connected) devices? 4) How to integrate het- The first 6G technology white paper was published in 2019 by
erogeneous technologies (e.g., radio-optical)? 5) How to the Finnish 6G Flagship Program [2]. In [3], a general survey
use ML/AI and deep learning (DL) for networks and over over the technologies envisioned for 6G networks and ser-
networks? 6) How to guarantee the power/energy effi- vices is described. A holistic and forward-looking vision that
ciency and material sustainability of the new networks and defines the tenets of a 6G system can be found in [4]. In [5],
user devices? 7) How to design the systems and business the role of AI in designing and optimizing 6G architectures,
environment so that it is profitable while the services and protocols, and operations is analyzed. In [6], the potential use
applications serve the green transition and environmental cases enable by new technologies of 6G systems are taken
sustainability? into account, while [7] focused on 6G architecture, describing
The future 6G networks will be designed so that users’ the usage scenarios and requirements for multi-terabyte per
intelligence and needs will be further explored and satisfied. second and intelligent 6G networks. The role of deep learning
In other words, the various applications of run by the users to enhance 6G networks is discussed in [8]. In [9], signal pro-
will form the 6G network operating as a networked computer cessing is indicated as important factor for the integration of
or inference machine. This calls for power and energy effi- different frequency bands and dynamic resource management.
cient, reliable wireless connectivity and networks. In [10], the use of optical signal processing together with AI
Potential use cases and challenges (Fig. 1) for 6G con- is envisioned to revolutionize the next generation mobile net-
nectivity could span from extended reality (enhanced virtual/ works, while [11] envisions new signal processing methods as

Figure 1  6G challenges.

SLA vs. "best-effort" Area expansion Frequency
Freq
quency
y utilization
Service Level High Altitude Maximize
Agreement (SLA) Platform radio wave
& Low latency
y Systems use frequency
y
API-driven network
customization Frequency expansion Network security
y
Terahertz/
T
Service Based Optical Quantum
Architecture communication encryption
Radio wave usage Resilience, Redundacy,y
AI networks expansion Recoveryy

AI Networks
function Sensing and and servers
Positioning
100% coverage Net-zero
Non
Terrestrial
T Charging/
Network Power supply
y Carbon-free

Architectural Challenges Technological Challenges

T Social Challenges

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 437

mandatory for dealing with future massime MIMO networks

as well as for security in 6G networks. In [12], the signal pro-
cessing is seen as fundamental, together with AI, for network
orchestration of future intelligent IoT networks. In [13], the
role of signal processing in the 6G era, along with the associ-
ated challenges, is briefly discussed, although this is not the
main goal of the paper.
While 6G architectural/technological/societal challenges
are reported in Fig. 1, core technologies can be classified as

• Wireless communication systems: THz communica-

tions; visible light communications; nanoscale commu-
nications;
• Next generation antenna and materials: massive mul-
tiple antennas; cell-free MIMO; reflecting intelligent
surfaces; radio-reconfigurable antennas;
• Coding and modulation: channel coding; non-orthogonal
wave; multiple access systems;
• Spectrum sharing: free duplex; full duplex; dynamic
spectrum sharing;
• Full integration of: artificial intelligence; Internet of
things; blockchain.

From architectures point of view, the 6G era will show

several novelties Figure 2  Massive MIMO networks.

• Ubiquitous 3D coverage: non-terrestrial networks

(NTN); high amplitude platform systems (HAPS); inte- 1.2 Paper Outline
gration of space, aerial, terrestrial, and underwater net-
works; Although the general literature on 6G is getting rich, very
• Intelligence inside networks and systems: AI-based few overview papers on signal processing aspects for 6G
networking; real-time intelligence (edge computing); have appeared. This paper aims at discussing the role and
intelligent spectrum adaptation; key challenges of signal processing in future 6G systems
• New network protocols: next Internet protocols. and networks. In [8] the signal processing for 6G systems
is touched but its main interest is towards the deep learning
The current 5G networks already increase the data rate and techniques. Conversely, we discuss the whole signal process-
decrease the data connections latency. Further performance ing chain in future 6G systems, from the conditioning of
gains are expected from cell-free massive MIMO illustrated the signal at transmission to MIMO detection, from chan-
in Fig. 2. Massive MIMO is a key enabler already in the 5G nel coding to channel estimation, from multicarrier (radio)
networks, while the cell-free extension is largely expected to modulation to optical wireless communications.
be introduced to practical use in the 6G era. In cell-free mas- The paper is organized as follows. We cover frontend
sive MIMO, all multiple-antennas access points (AP)s are signal processing, transmit precoding and beamforming,
connected to a CPU which operates all as a massive MIMO multiantenna detection, channel coding and decoding, chan-
network with no cell boundaries where all users are served nel estimation, as well as non-orthogonal multiple access
by coherent transmission and reception [14]. The APs and the (NOMA) signal processing, optical wireless communica-
CPU exchange the users’ estimated channels between them, tions (OWC), and physical layer security (PLS) in Sec-
hence, the burden on the fronthaul network is increased pro- tions 2–9. In Section 10, the future research challenges on
viding opportunities for performance enhancements. Cell- technologies are discussed. Those include machine learning
free massive MIMO is first introduced by Yang in [15]. It based 6G design, integrated communications and sensing
implies that there are no boundaries between cells. Cell-free (ISAC), and the internet of bio-nano-things. Paper is briefly
massive MIMO is a combination of three concepts: massive summarized and conclusions are drawn in Section 11.
MIMO, distributed MIMO, and cell without boundaries [14].

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

438 Journal of Signal Processing Systems (2023) 95:435–457

2 Front‑End Signal Processing and Digital high volume. Some literature already exists on tri-band
Pre‑distortion DPDs. In [17], the authors presented a DPD for concurrent
tri-band PAs. The PA model is based on a pruned Volterra
The non-linear distortion of radio frequency components model which takes both phase distortion in multiband PAs
can severely degrade the performance of an entire com- and compound amplitude distortion. Conventional least-
munication system. The primary source of the non-linear square (LS) has been used to estimate model coefficients
distortion in radios are typically the high-power amplifier in [17]. A simplified dynamic deviation reduction model and
(PA). The non-linearity problem can be circumvented by LS is used for modeling the PAs and estimating the model
using linear class A PAs or operating any PA far from their coefficients, respectively, for a tri-band DPD.
saturation point. However, this leads to a bulky, expensive The cell-free massive MIMO system requires relatively
and inefficient PA which is far from an ideal solution to non- smaller antennas and analog circuitry for radio-frequency
linear distortion. A popular solution is to apply a non-linear (RF) modules so that the access point (AP) could be placed
filter at the digital domain of the transmitter that applies in any geographical location. This poses an interesting chal-
an inverse of the PA response. This technique to linearize lenge to DPD implementation because the pre-distortion fil-
PA response by pre-distorting the digital signal is known ter is typically placed closer to the RF circuits. Due to the
as digital pre-distortion or DPD. In Fig. 3, a communica- small size of the AP hardware unit close to the user, it might
tion system in the presense of a DPD is illustrated. Here, be challenging to implement complex algorithms for the
the baseband signal is denoted by a(KT0 ) which traverses DPD. As the AP distance from the central processing unit
through a pre-distortion filter which applies an inverse of (CPU) will be different, it will be challenging to synchronize
the PA response. The coefficients of the pre-distortion filter different feedback signals with a single feedback path. Due
gets repeatedly updated by the adaptation or training block. to the size of the APs and challenges with the feedback, a
The adaptation block compares the output of the PA at the non-adaptive simpler DPD solution will be ideal for a cell-
baseband, c(KT0 ) and the output of the pre-distortion filter, free massive MIMO system. A simple look-up table (LUT)
b(KT0 ) to generate the updated coefficients. based DPD can be used in this case which is typically used
DPD for dual-band PAs is a common feature for 5G radios for mobile phones. In [18], one such classical LUT DPD has
these days. It is an alternative to ultra-wideband DPDs by been presented. The AM-AM and AM-PM non-linearities
only compensating for the nonlinear distortions around the are characterized to calculate complex coefficients which
signal bands of interest. The dual-band DPD is derived from are placed in the LUT.
the wideband memory polynomial DPD model [16]. The 6G communication systems will use very high carrier
individual terms located around each band can be grouped to frequencies and thus, beamforming is essential to circum-
define the coefficient mapping between the wideband model vent the high attenuation and path loss. Phased array based
and its corresponding dual-band model. We envision that the beamforming transmitters typically apply multiple PAs for
trend to develop more sophisticated DPDs for multi-band a single digital stream. However, this introduces significant
will continue throughout the decade. To be more specific, challenges for DPD because multiple parallel PAs have to
the tri-band DPD products will be more popular by the time be linearized with a single DPD. According to [19], the two
6G arrives. For dual-band case, the terms centered around methods to design such a DPD would be to linearize indi-
the signal band are sufficient because the out-of-band inter- vidual PA as a LS problem or to linearize array response to
modulations are far from the band of interest. However, the a desired spatial direction. In [20], the authors presented a
out-of-band intermodulation terms might be located around DPD scheme that can linearize multiple PAs of a hybrid sys-
the band of interest in concurrent tri-band PAs due to their tem with antenna subarrays. The authors used LS to identify
the PA parameters. The identified parameters and the input
signal were used to design the DPD block which minimizes
the expected sum of squared error. To accelerate the con-
a(kT0) Pre-Distortion b(kT0) u(t) v(t)
Filter
Modulation PA vergence, the least mean-square (LMS)-Newton algorithm
was used for iteratively minimizing the expected sum error.
However, this scheme is less effective because the errors are
Weight measured independently and they can add up constructively
Update
Adaptation/ or destructively over the air. Therefore, minimizing the array
Demodulation
Training c(kT0) error in the desired beam direction provides better results.
In [21], the authors formed the cost function with the theo-
retical nonlinear behavior in the array far-field and input
Figure 3  A communication system with DPD. signal, which was solved by LS.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 439

The PA models are generally not perfect as the response better system design than saving logic area within unstable
of a PA is a non-linear continuous function. The PA has to and risky solutions. Several precoders have been proposed
support different types of signals and thus, the PA mod- based on QR and Cholesky decompositions [26].
eling can be challenging for a rapidly changing environment. The second class of precoders is the non-linear such as
Neural networks can be very efficient for non-linear map- the dirty-paper-coding (DPC) [27], Tomlinson-Harashima
pings. A feedforward neural network with sufficient neu- (TH) [28], and vector perturbation (VP) [29] based precod-
rons is known as universal approximators for an arbitrary ers. The DPC algorithm is not hardware friendly because
continuous function [22]. Therefore, the neural network infinite length of codewords and sophisticated signal pro-
has been a popular choice for DPD modeling. For example, cessing are required. The TH precoder is a suboptimal
in [23], the authors proposed a neural network DPD model implementation of the DPC algorithm and decomposition.
to jointly mitigate crosstalk, I/Q imbalance and nonlinearity In comparison with the DPC based precoders, TH precoder
in MIMO transmitters. During the feedforward computation, is more hardware friendly. The generalized TH algorithm,
the authors initially set the weights between [−0.8, 0.8] and also known as VP algorithm, obtains a much lower com-
during backward propagation weights are adjusted to reduce plexity compared with the DPC algorithm. In the VP algo-
the error. The Levenberg-Marquardt algorithm, which is an rithm, the data is aligned to the eigenvalues of the chan-
approximation of the Gauss-Newton method, is used for nel matrix inverse on an instantaneous basis. It performs a
updating the coefficients. We expect deep neural networks sphere search out of several candidate perturbation vectors
to play an important role for 6G DPD modeling. to reduce the precoded signal norm [29].
In order to improve the achievable downlink data rates
of a cell-free massive MIMO, conjugate beamforming (CB)
3 Transmit Precoding and Beamforming precoding [30], ZF precoding [31] and MMSE precoding
[32] have been utilized. It is shown that the centralized
In a wireless propagation environment, it is usually hard to MMSE significantly achieves higher DL data rates compared
obtain a reliable channel state information (CSI) where the to conventional CB precoding. In centralized approach, the
performance of a downlink (DL) transmission system could APs and the CPU exchange the users’ estimated channels
largely be affected. Precoding algorithms can be exploited between them, hence, the burden on the fronthaul network
to deal with imperfect CSI where the effects of interference is increased. Therefore, several advanced local precoding
and path-loss are reduced. Precoding can also be exploited techniques are proposed to eliminate such burden such as
at the MIMO’s base station (BS) to increase the spectral effi- the local full-point ZF (FZF) [33], partial ZF (PZF), and
ciency, and enhance the throughput and the capacity when protective partial ZF (PPZF) [34].
the number of antennas approaches infinity [24]. In litera- An important recent extension of the conventional beam-
ture, linear, non-linear, and machine learning based precod- forming or precoding is presented by the reconfigurable
ing algorithms were proposed. Usually linear precoding intelligent surface (RIS) technology, known also as intel-
techniques, such as the zero-forcing (ZF) and the MMSE, ligent reflecting surface (IRS) [35–38] illustrated in Fig. 4).
depend on multiplying the transmitted signal by the precod- A RIS typically consists of a large number of low-cost pas-
ing matrix [25]. However, such methods incur a high com- sive elements, and can support different functional modes,
plexity due to a matrix inversion. Linear precoders based on e.g., reflection, polarization, refraction, and absorption. A
approximate matrix inversion methods such as the Neumann RIS can be constructed, e.g., by varactor diodes or crystal
series approximation (NSA), Newton iteration (NI), Cheby- liquid [39, 40]. A ray tracing based RIS channel model was
shev iteration (CI), Gauss-Seidel (GS), conjugate gradient proposed in [41] for both indoor and outdoor environments.
(CG) and successive overrelaxation (SOR), were proposed.
Although approximate matrix inversion methods achieve a
satisfactory performance when the ratio between the num-
ber of BS antennas and user terminals is large, they have BS

a severe performance degradation when the ratio is small. RIS

They also need to calculate an initial value which could slow MS

the convergence. In addition, many methods need additional

BS

calculations to find suitable relaxation/optimization factors. MS

Matrix decomposition algorithms provide better numerical MS

stability over approximate matrix inversion methods at the

MS

RIS
cost of a high computational complexity [26]. However,
semiconductor technology has also matured greatly over the
last ten years where the interest has been shifted towards Figure 4  Reconfigurable intelligent surfaces.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

440 Journal of Signal Processing Systems (2023) 95:435–457

The objective of the RIS is to control the wireless propaga- first at each AP and then at the CPU. Most of the detection
tion environment so as to enable improved connectivity [35, techniques for cell-free massive MIMO are centralized as
42, 43]. What is more, the RIS can also be considered for they rely on a single CPU to do most digital signal process-
localization together with mmWave communications tech- ing (DSP) tasks with the help of irregular distributed APs.
nology networks [44–46]. The first centralized massive MIMO detector has used the
One key limitation of the passive RIS is the fact that the likelihood ascent search (LAS) because of its linear aver-
passive beamforming limits the beamforming gains or the age per bit complexity in number of users and its ability to
available degrees of freedom. Large numbers of RIS ele- achieve a near-maximum likelihood performance. Unfor-
ments are needed to outperform the decode-and-forward tunately, the bit-error-rate (BER) is significantly deterio-
(DF) half duplex relaying with moderate numbers of anten- rated in the scenario of high modulation order and realistic
nas [47, 48]. What is more, a practical RIS often applies environment. In addition, a computation of the initial vector
limited-resolution phase shifters resulting in further perfor- includes a matrix inversion which increases the computa-
mance penalty [49, 50]. Therefore, the RIS performance can tional complexity. Reactive tabu search (RTS) is another
be improved by so called hybrid relay-reflecting (HR-RIS), local search method where more restrictions are introduced
which combines the functionalities of a RIS and a relay [51]. to avoid an early termination, and hence, higher accuracy
The basic idea of the HR-RIS is to replace few elements of is achieved. Unfortunately, the RTS detector suffers from a
the RIS by simple reflect amplifiers. high computational complexity and a performance loss when
a high modulation order is used. In order to reduce the com-
plexity, most of the proposed detectors during 2008 - 2013
4 MIMO Detection had used local search algorithms and belief propagation (BP)
algorithms. The BP algorithms, such as the message pass-
In the last few years, there is a debate to choose the most ing and the Bayesian belief networks, iteratively search for
scalable massive MIMO scheme: centralized vs. decentral- the optimum solution in a space where the damping factor
ized. In centralized massive MIMO, the central processing (DF) is carefully optimized. The BP is very sensitive to both
unit (CPU) collects the CSI from all antenna elements. The the message update rules and prior information. However, it
signal processing tasks (demodulation, decoding, etc) are achieves a high performance when the correlation between
performed at the CPU which require extra radio frequency the channel elements is relatively small. In years after, due to
(RF) and analog-to-digital converter (ADC) components fail to guarantee convergence and implementation difficul-
(Fig. 5). In the last few years, cell-free massive MIMO has ties, a research on linear and nonlinear detectors based on
gained a lot of attention due to its potential to improve the free-matrix-inverse methods has been conducted. Unfortu-
energy and spectral efficiencies of wireless communication nately, these detectors suffer from a high performance loss
systems. In cell-free massive MIMO, data detection is per- and a high computational complexity when the massive
formed locally at each AP, centrally at the CPU, or partially MIMO size is large, the ratio between the BS antennas and
user antennas is small, and an existence of a high correlation
between channel elements. Researchers in the telecommuni-
cation industry intend to improve the system design to avoid
TX unstable and risky solutions for their products. Therefore,
ion
in [52], we present the computational complexity of linear
at
orm
Inf (bits)
Multi-
Antenna
Transmitted detection mechanisms based on the QR, Cholesky and LDL
Encoding
Signal decomposition algorithms for different massive MIMO con-
Model figurations. Other detectors, such as the sphere decoding
(SD), suffer from a high computational complexity. There-
Wireless fore, most existing detectors need a refinement to meet the
Loss Function Parameter
MIMO implementation demands of a low computational complex-
Reconstruction Optimization
Channel ity and high performance, in particular under complicated
environments.
ted
ruc
nst ation Multi-
Since 2017, there is a substantial trend in the research
o
community to exploit machine learning, artificial intelli-
c
Re form s) Antenna
In (bit Received
Decoding
Signal gence (AI), and deep learning (DL) in data detection. The
Model
deep network in the massive MIMO detector’s design based
RX on projected gradient descent method is utilized and called
DetNet [53]. It performs well in i.i.d. Gaussian channel and
Figure 5  SP MIMO processing scheme. low-order modulation schemes (i.e., BPSK and 4-QAM). A

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 441

modified DetNet [54] is proposed where a relatively small free-matrix-inversion methods in different channel condi-
number of parameters is required to optimize. Unfortunately, tions [60]. The FD [61] architectures based on the coordi-
the training is unstable in realistic and correlated channels. nate descent (CD) method and FD [61] data detector based
In addition, scalability of the DetNet algorithm is poor on recursive least square (RLS), stochastic gradient descent
because of a relatively large number of training parameters. (SGD), and averaged stochastic GD (ASGD) have also been
In 2018-2022, there is a notable trend in a research com- proposed.
munity to exploit the DL to build a robust massive MIMO DL architectures and (AI) could be exploited in decen-
detector. A model-driven DL network is proposed based tralized and cell-free massive MIMO. However, the lit-
on the orthogonal approximate message passing network erature has shown a paucity of employing artificial intel-
(OAMP-Net) [53]. It adds some adjustable parameters to ligence for data detection in decentralized and cell-free
the existing OAMP method. Unfortunately, it is very restric- massive MIMO.
tive where a strict assumption has to exist. The performance
of the OAMP-Net is dominated by the matrix inverse in
each layer. Therefore, it is not feasible for implementation 5 Channel Coding
because of a high complexity. OAMP-Net2 algorithm is an
extension of the OAMP-Net where new training parameters Channel coding is a technique to control errors in data com-
are utilized [53]. Unlike the OAMP-Net, imperfect channels munication over noisy channels. The key idea of channel
are considered. However, like the OAMP-Net, it is domi- coding is to add redundancy to the messages in the trans-
nated by the matrix inverse. The MMNet [55] algorithm is mitter for encoding. These redundant parts are used on the
proposed to overcome the challenges in the DetNet and the receiver side to detect the errors. Channel coding is an inte-
OAMPNet. It is designed to be trained online for each where gral part of wireless communication systems since the intro-
an iterative soft thresholding algorithm is used. Although it duction of the convolutional codes in 1955 by Elias [62].
achieves a good performance when implemented in a real- In 5G new radio (NR) standard, low-density parity-check
istic channel simulator, it incurs latency due to the sequen- (LDPC) and polar coding are adopted data and control chan-
tial online training. In addition, the performance degrades nels, respectively [63]. Therefore, LDPC and polar coding
significantly in a high modulation order. A HyperMIMO schemes will continue to evolve throughout this decade.
[56] based detector replaces the training process required by We envision that improved versions of LDPC and polar
the MMNet for each channel realization by a single infer- will be major candidates for 6G channel coding. The first
ence through a trained hyper-network. It also reduces the reason is we have seen in the past that popular channel cod-
number of parameters of the MMNet. In comparison with ing schemes have been adopted for more than one genera-
the MMNet, HyperMIMO performs slightly worse. It also tion of communication systems. For example, turbo coding
needs to be re-trained when the channel statistics change has been adopted for both 3G and 4G systems. The second
significantly. reason is LDPC and polar are already very good channel
Many testbeds, such as the Argos testbed, the LuMaMi coding schemes. LDPC are capacity-approaching codes
testbed, and the BigStation testbed, are available to support and polar are the first capacity achieving codes with low
the decentralized channel estimation and data detec- decoding complexity. The third reason is the existing hard-
tion at antenna elements. Unfortunately, they rely on ware implementations in the literature can already provide
the maximum ratio combining (MRC) that significantly hundreds of Gbps [64, 65]. Therefore, we can assume that
reduces spectral efficiency, and hence, prevents the use with improved algorithm and hardware architectures LDPC
of high-rate modulation and coding schemes. Therefore, or polar schemes will be able to provide Tbps throughput
alternative (BS) architectures based on a decentralized required for 6G communication systems.
approach are proposed. A decentralized data detection LDPC introduces more complexity in the encoding
method based on the (CG) is proposed where the BS phase, but the decoding phase is simpler than turbo codes
antenna array is partitioned into clusters and each clus- and thus, ideal for large block lengths. However, to reach
ter is associated with independent local (RF) elements the Tbps goals of 6G, the LDPC algorithms will require
and computing circuitry [57]. Another decentralized further modification and optimization. In [66], the authors
data detection based on alternating direction method proposed a Check Node Self-Update (CNSU) algorithm
of multipliers (ADMM) [58], partially decentralized (PD) for LDPC decoding which reduces the memory and power
and fully decentralized (FD) data detectors based on requirements. The hardware architecture based on CNSU
approximate message passing (AMP) [59] are proposed. algorithm was able to provide very high throughput which
Unfortunately, the proposed decentralized based processing makes it suitable for beyond 5G systems. In [67], the authors
(DBP) is not tested in different system configurations and studied the convolutional code LDPC (CC-LDPC). The
realistic channel conditions. The DBP is studied based on authors compared the CC-LDPC to the conventional block

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

442 Journal of Signal Processing Systems (2023) 95:435–457

code LDPC schemes. CC-LDPC has the advantage in terms algorithm. The improvements were demonstrated by differ-
of lower error floor, faster decoding convergence and lower ent LDPC codes. The polar decoder is enhanced by applying
decoding complexity. The authors concluded that CC-LDPC deep learning in [75]. The authors partitioned the encoding
has enormous potential for 6G communications due to its graph and train them individually which results in a non-
high reliability and low latency. There have been reinvigor- iterative and highly parallel decoder. In [76], a trained deep
ated interest in protograph-based LDPC as protographs pro- neural network is concatenated with a standard belief propa-
vide an efficient way to construct LDPC codes. In [68], the gation decoder. Iterating between the neural network and
authors proposed an improved protograph LDPC algorithm belief propagation resulted in better decoding performance.
suitable for 1-bit ADC-based massive MIMO systems. This
new method overcomes the error-floor issue of conventional
LDPC codes and is a suitable candidate for low resolution 6 Channel Estimation
6G base stations. NOMA schemes have great potential to be
an integral part of 6G which opens a new area of multi-user Channel estimation continues to be an essential receiver
oriented channel coding [69]. Interleave-division multiple- functionality in 6G systems. Several of the key technologies
access (IDMA), a capacity-approaching NOMA scheme, envisioned for 6G impose new channel estimation problems
with turbo joint decoding at the receiver can leave a gap of that cannot be solved efficiently using conventional methods.
1.4 dB to Shannon limit at the sum spectral efficiency with For instance, THz communication experiences significantly
16 users [70]. Recently, LDPC codes have been proposed long channel responses as well as low SNR due to the srong
for multiuser channels due to their lower complexity and noise [77]. Moreover, minor variations in the environment
flexibility. For example, a raptor-like quasi-cyclic LDPC has can cause significant channel estimation errors [78]. Further-
been constructed for IDMA based random access in [71]. more, massive MIMO systems operating in the THz range
Polar codes, which was introduced in 2009, is the first experience the beam split effect where the path components
capacity achieving codes with low encoding and decoding split into different spatial directions at different subcarrier
complexities. However, sequential coding (SC) is required frequencies, leading to serious array gain losses [79]. RISs
for the polar coding schemes to reach Tbps throughput. SC also introduce new challenges in channel estimation since
decoding traverses through a polar factor tree in a sequen- RIS elements are passive and cannot transmit, receive, or
tial manner which can be unrolled for high throughput [72]. process any pilot signals to realize channel estimation. It is
Even though SC decoding enables very high throughput, therefore important to develop channel estimation algorithms
they suffer from error correcting performance. Successive that can handle such challenges. These algorithms should
cancellation list (SCL) processes only a subset of candidates combine low computational complexity with high spectral
among the polar factor tree nodes. At leaf nodes, the less efficiency (low pilot overhead), without sacrificing the esti-
reliable candidates are sorted out. The sorting process intro- mation accuracy. In what follows, we will summarize the
duces marginal complexity which is negligible due to an state of the art in channel estimation for 6G systems.
improved error correction performance. We would also like The enormous bandwidth available for THz communica-
to mention that 6G systems might replace LDPC codes and tion enables the achievement of data rates in the order of
adopt polar codes also for the data channels. 1 Tbps. The unprecedented potential of THz communica-
Most channel codes are designed for a specific set of cod- tion also comes with major practical challenges for imple-
ing rates. Even though LDPC provides a large choice of mentation, including high propagation losses due to severe
coding rates for 5G, they are not truly rate-less. In [73], the signal attenuation and molecular absorption, as well as the
authors presented a novel rate-less code which they named frequency selectivity of the channel. Considering the large
as spinal codes. This novel coding scheme uses a hash func- number of channel parameters and the unsuitability of con-
tion over the message to generate pseudo-random bits which ventional estimation techniques, most works in the litera-
can be mapped directly to the constellations. The simulation ture either exploit the inherent sparsity characteristics of the
results show that spinal codes achieve Shannon capacity and channel through compressed sensing (CS), or leverage the
outperform the best-known fixed rate block codes. There- power of deep learning to reduce the computational com-
fore, spinal codes will enable a rate-less 6G system where plexity and improve the estimation accuracy and spectral
any coding rate can be used for transmission based on the efficiency.
receiver capacity and channel condition. CS was applied to estimate indoor THz channels in [77].
Deep learning will play a key part in 6G systems. Chan- In addition to proposing a compressive sampling matching
nel coding methods based on deep learning have also gained pursuit (CoSaMP) algorithm, the authors of [77] also consid-
a lot of attention in recent years. Deep learning has been ered the Dantzig selector (DS), a computationally tractable
used to decode linear codes in [74]. The application of deep CS-based approach that formulates the channel estimation as
learning improved the performance of the belief propagation a convex optimization problem. It was shown that the both

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 443

CS methods significantly outperform the least squares (LS) systems. Considering an array-of-subarrays configura-
approach in terms of MSE, and that the DS method is prefer- tion, [84] develops a deep convolutional neural network
able to the CoSaMP method which exhibits some degree of (DCNN) channel estimation technique that learns the param-
instability for low number of observations. The work in [80] eters of the spherical wave channel model, including azi-
applies CS to estimate dynamic MIMO THz channels by muth and elevation angles, amplitude of the channel gain
exploiting sparsity in the angular/delay domain. An algo- and phase shift matrix. The work in [85] addresses the inac-
rithm is proposed based on accelerated gradient descent with curacies of the planar wave model and the limitations of the
adaptive restart (AGDAR), which is shown to be fast and spherical wave model by proposing a hybrid planar-spherical
effective. Moreover, two further improved CS algorithms model. The planar wave model is adopted within sub-arrays
are introduced, namely the selective AGDAR (S-AGDAR) and the spherical model among subarrays. A combination of
and the adaptive AGDAR (A-AGDAR). Substantial gains in DCNN and geometric relationships is employed to estimate
MSE, computational complexity and latency are observed the channel parameters over two stages. The work in [86]
over the LS method. A joint activity detection and channel exploits both machine learning techniques and the sparsity
estimation (JADCE) technique is proposed in [81] for wide- structure of the channel matrix by designing a hybrid trans-
band THz IoT systems to address the large pilot overhead ceiver where estimation is performed via a combination of
and the large dimensionality of the signal processing, by Bayesian learning and orthogonal matching pursuit (OMP).
exploiting both the sparsity pattern in the angular domain Generative adversarial networks (GANs) are trained in [87]
and the low-rank structure of the channel matrix. to generate samples from the unknown channel distribution.
While [77] and [81] consider SISO systems, and [80] The trained network is then used as a prior to estimate the
considers a MIMO system, massive MIMO THz channel current channel using the received signal.
estimation is more complicated due to the very large number CSI acquisition for RIS-assisted networks is another chal-
of channel parameters. Accurate channel estimation is essen- lenging problem due to the compound nature of the propa-
tial to enable hybrid precoding and to reduce the number gation. However, it is necessary for the RIS phase control,
of RF chains. Furthermore, massive MIMO THz systems beamforming, resource allocation, and interference man-
experience the beam-split effect, where the large number agement [88]. Several channel estimation algorithms have
of antennas and the wide bandwidth result in frequency- been proposed. The RIS is often assumed to be used in the
dependent sparse channel supports and make the spatial mmWave or (sub-)THz communications systems to enable
channel directions different from each other in the angular line-of-sight (LOS)-like connectivity even with the non-
domain for different subcarriers. Channel estimation for this line-of-sight (NLOS) conditions. The mmWave channels
scenario is studied in [82], where beam split patter detection are typically very directive and sparse with a small number
is first performed, and then the sparse channel supports at of propagation paths. Numerous compressive sensing (CS)
different subcarriers are estimated using a support detection based approaches, e.g., the atomic norm minimization, basis
window. The procedure is repeated until all path components pursuit, approximate message passing (AMP), and mixed
are considered, and the wideband channel is recovered by norm minimization have been proposed for RIS channel
considering the total sparse channel support containing the estimation [89–94]. Joint channel estimation and data-
channel supports for the different path components. rate maximization for THz-based RISs is proposed in [95]
The work in [83] also addresses the beam split effect through an iterative atom pruning based subspace pursuit
in THz massive MIMO systems by using uniform planar (IAP-SP) scheme, which is noted to exhibit lower compu-
arrays. Channel sparsity in the angular domain is exploited tational complexity than the classical subspace pursuit (SP)
to formulate the channel estimation problem as a CS prob- scheme. A two-stage algorithm that includes a sparse matrix
lem, which is solved using the orthogonal matching pur- factorization stage and a matrix completion stage is devel-
suit (OMP) algorithm. Contrary to the existing works, the oped in [96], and a novel message-passing based algorithm
authors employ a wideband dictionary and show that the is proposed to solve the matrix-calibration based matrix fac-
channels across different OFDM subcarriers share a com- torization problem in [97].
mon support in this case. This enables applying a variant In [98], the channel is estimated for the downlink MISO
of the simultaneous OMP algorithm, coined as generalized RIS system, with the assistance of active elements that are
simultaneous OMP (GSOMP), which exploits the informa- randomly distributed in the RIS, which successfully reduces
tion of multiple subcarriers to increase the probability of the pilot overhead. On the other hand, an uplink multi-user
successfully recovering the common support. It is reported MISO RIS system is considered in [99], and the channel is
that the proposed GSOPM outperforms the OMP in the low estimated based on parallel factor decomposition to unfold
and moderate SNR regimes. the cascaded channel model. The channel is then estimated
Other works have sought to leverage the power of deep using alternating least squares and vector approximate mes-
learning for channel estimation in THz massive MIMO sage passing.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

444 Journal of Signal Processing Systems (2023) 95:435–457

Geometric channel models explicitly couple the chan- limitations on the improvement in spectral efficiency. Non-
nel parameters and node locations leading naturally to joint orthogonal multiple access is a promising technology that
channel estimation and mobile positioning [100]. In addi- can effectively solve this problem. At the cost of increased
tion to conventional model-based approaches, data-driven receiver complexity, NOMA allows multiple users to use the
approaches, for instance, deep learning can also be employed same time/frequency resources, separating them in power
for channel estimation, positioning, RIS phase control, and or code domains [107]. The most common form is power-
symbol detection. domain NOMA, which multiplexes users by superposing
Cell-free massive MIMO is another promising 6G them at the transmitter side using different power levels. At
technology where channel estimation plays a critical role. the receiver side, successive interference cancellation (SIC)
Channel estimation enables the evaluation of the precod- is used to recover the transmissions of the different users.
ing/detection vectors used for DL/UL data transmission. NOMA exhibits both improved throughout and fairness in
An overview of channel estimation techniques for cell-free comparison to OMA, and it is expected to play a key role
massive MIMO is provided in [101]. As noted in [101], in 6G systems.
most techniques are based on pilot transmission where both The integration of NOMA and MIMO technologies
orthogonal and non-orthogonal pilot schemes have been has been highlighted as a powerful approach to achieving
studied. Orthogonal schemes are more suitable scenarios high spectral efficiency and better wireless services [108]
with low mobility and a small number of users, whereas (Fig. 6). In contrast to SISO NOMA, where the focus is to
non-orthogonal schemes are preferable for high mobility optimize power allocation among users, MIMO-NOMA pro-
scenarios. Most of the works adopt the minimum-mean vides additional degrees of freedom through beamforming in
squared error (MMSE) estimation scheme, and significant the spatial domain. As noted in [107], the beamforming and
effort has been expended to limit the impact of pilot con- SIC problems become coupled in the MIMO-NOMA setup,
tamination, for instance by proposing greedy-based pilot since the design of the beamformer has a direct impact on
assignment methods [102, 103]. A graph coloring-based both the signal power and the interference power of the dif-
pilot allocation scheme is also proposed in [104] to reduce ferent users. Since the SIC performance of MIMO-NOMA
the impact of pilot contamination assuming that only a lim- largely depends on the decoding order of the users, it needs
ited number of APs serves each user. A different approach to be designed jointly with the beamformer, which opens the
is taken in [105] by focusing on reducing the pilot overhead way to a new class of joint optimization problems.
through joint channel estimation and data detection, which A single-cluster MIMO-NOMA system is investigated
is formulated as a biconvex optimization problem and solved in [109], where the authors aim to optimize the power
using a forward-backward splitting algorithm. allocation and beamforming in order to maximize the sum
As evidenced by the above works, progress has been made rate of single-antenna users for a given SIC order. The
in developing channel estimation techniques to enable 6G optimization considers both a total transmit power constraint
systems. Yet there are still open problems and more effort as well as an additional constraint to protect weak users.
needs to be expended to find the most optimal solutions. It
is expected that future solutions will continue to build on Power at BS

the sparsity properties of the channel matrix and the power

of deep learning to produce algorithms that combine high
spectral efficiency with low computational complexity and
high estimation accuracy. Only few works have considered Resource blocks
the impact of various RF impairments in channel estima-
tion thus far [106], although these impairments are expected
to affect the performance of 6G systems. Furthermore, 6G Receiver 2

systems support high levels of mobility, while most of the Receiver 1

developed works focus on stationary channels.

BS Receiver 3
7 Non‑Orthogonal Multiple Access
Receiver 4
As 6G systems are expected to support extremely high data
rates for numerous users and devices, orthogonal multiple
access (OMA) schemes, which have been the mainstay of Received Signal
previous generations of wireless systems, may not be able
to cope with the increasing demand, resulting in undesired Figure 6  NOMA-MIMO scheme.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 445

The resulting problem is non-convex and is solved through optimization problem under imperfect CSI is investigated
a successive convex optimization approach based on in [117] using successive convex optimization and semi-
minorization-maximization. The simulation results indicate definite programming.
that MIMO-NOMA may be superior to traditional Zero- The previous works investigate the combination of
Forcing (ZF) beamforming when the number of users is NOMA with conventional MIMO. As massive MIMO
significantly higher than the number of transmit antennas is expected to play a leading role in 6G systems, massive
at the BS. A two-user downlink MIMO-NOMA system is MIMO-NOMA is considered an attractive research area,
considered in [110], where the ergodic capacity is maximized given the large number of spatial degrees of freedom avail-
for a given decoding order based on statistical CSI and able at the BS. Massive MIMO has traditionally focused on
optimizing the transmit covariance matrix. Inspired by the underloaded systems where the number of users is smaller
H-BLAST scheme, a MIMO-NOMA system with layered than the number of BS antennas. Hence, the spatial degrees
transmission is proposed in [111], and the power allocation is of freedom provided by massive MIMO may not be enough
optimized to maximize the sumrate. Furthermore, the authors to efficiently handle overloaded systems with an excessive
of [112] identify a “quasi-degraded” channel condition for number of users. The power domain multiplexing enabled
the two-user MISO channel, and accordingly optimize the by NOMA can facilitate serving more users. The authors
beamforming for MISO-NOMA by minimizing the transmit of [118] consider the overloaded massive MIMO-NOMA
power under user rate constraints. scenario where the number of users is larger than the num-
In the above scenarios, all the users are grouped into the ber of antennas, and propose a Gaussian message passing
same cluster, and hence each user interferes with all other (GMP) multiuser detection scheme. The GMP exhibits
users in the network. It can be prohibitively complex to opti- a complexity that is linear in the number of users. A user
mize both the beamformer and the decoding order in such clustering scheme is proposed for cell-free massive MIMO-
cases, especially when the number of users is large. To over- NOMA in [119] and the resulting sum rates are derived
come this problem, multi-cluster MIMO-NOMA is proposed considering intra-cluster pilot contamination, inter-cluster
in [113], where each cluster consists of several users that interference and imperfect SIC.
share the same beamformer. This allows grouping users with As massive MIMO-NOMA requires accurate CSI to
similar spatial characteristics into the same cluser to mini- realize its potential, the authors of [120] consider two pilot
mize inter-cluster interference. Furthermore, it is sufficient schemes, one of orthogonal pilots, and the other where pilots
to perform SIC only for users within the same cluster, which are superimposed with the data. A data-aided channel esti-
reduces the decoding complexity. Using ZF beamforming mation scheme is investigated, where partially decoded data
to eliminate inter-cluster interference, the authors in [113] are used to improve channel estimation. The use of NOMA
develop clustering algorithms to maximize the fairness for is shown to mitigate the impact of pilot contamination.
downlink MIMO-NOMA. The authors of [114] propose a Channel estimation for uplink massive MIMO-NOMA is
general MIMO-NOMA framework applicable to both uplink also studied in [121], using semi-blind estimation strategies.
and downlinke transmission, by employing the concept of Group successive interference cancellation is employed in
signal alignment. Using signal alignment, the multi-user conjunction with semi-blind estimation in a multi-cell sce-
MIMO-NOMA scenario is decomposed into several single- nario. After dividing the users into multiple groups accord-
antenna NOMA channels. Both fixed power allocation and ing to their large scale fading, eigenvalue decomposition is
cognitive-radio inspired power allocation are considered. A applied to separate the signal subspaces of different groups
precoding/detection vector selection scheme is also devel- using the same pilot sequences. The proposed method is
oped in order to efficiently exploit the available degrees of shown to outperform conventional estimation techniques.
freedom. The authors of [122] propose a method to alleviate the
A two-stage beamforming scheme is proposed for two- impact of channel estimation and SIC imperfections by
user downlink MISO-NOMA in [115] where the first stage employing a successive sub-array activation (SSAA) diver-
eliminates the inter-cluster interference through ZF beam- sity scheme, resulting in better performance.
forming, while the second stage employs intra-cluster beam- The application of NOMA at mmWave and THz bands
formers to minimize the transmit power. The authors of [116] is another avenue to combine high data rates with increased
consider a beamforming design for downlink MIMO-NOMA connectivity. As noted in [107], the number of users that can
to cancel a substantial part of the inter-cluster interference be supported at such high frequencies is limited by the num-
when the number of transmit antennas of the BS is smaller ber of available RF chains. NOMA can resolve this limitation
than the total number of user antennas. User clustering is also by increasing the number of users through power domain
considered in [116], where a method is proposed that assigns multiplexing. The authors of [123] propose the integration
users with maximally distinct channel gains to each clus- of NOMA with beamspace MIMO systems operating in
ter in order to optimize SIC performance. The beamformer mmWave settings. A ZF precoding scheme is also developed

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

446 Journal of Signal Processing Systems (2023) 95:435–457

to reduce the inter-beam interference, while a dynamic power end-to-end CSI in [131]. In particular, a long short-term
allocation scheme is developed to optimize the sum rate, memory (LSTM) network is integrated into the NOMA sys-
which considers both inter-beam and intra-beam power allo- tem, to obtain channel characteristics automatically. User
cation. The proposed system is shown to provide superior clustering through deep learning is proposed in [132]. A
energy and spectrum efficiency compared to systems that deep learning framework is proposed in [133] to maximize
do not utilize NOMA. To guarantee the rate performance for the sum rate and energy efficiency for MIMO-NOMA. A
all users,the authors of [124] maximize the minimal rate of deep convolutional neural network, aided by training algo-
the system using max-min fairness, assuming that NOMA rithms is used to address the power allocation problem.
users in the same beam share the same precoding vector. Deep learning is used in [134] to maximize the sum rate
The minimal rate maximization problem is non-convex due for a downlink NOMA system by optimizing the power
to the inter-beam and intra-beam interferences. Hence, alter- allocation.
nating optimization is used to solve the power allocation and It is obvious form the above works that NOMA will play
precoding problems. The integration of simultaneous wire- an important role in the evoluation of 6G systems due to
less information and power transfer (SWIPT) with mmWave its attractive spectral efficiency. Furthermore, NOMA can
massive MIMO-NOMA is proposed in [125], where hybrid be effectively integrated with other emerging technologies
precoding is considered to reduce the number of RF chains. such as massive MIMO, mmWave communication, and
A power splitting receiver is proposed to allow each user RIS. NOMA also lends itself to the application of machine
to extract both information and energy. Joint optimization learning and deep learning techniques to improve perfor-
of the user power allocation and power splitting factors for mance and/or reduce computational complexity. Important
SWIPT is solved through an iterative optimization algorithm. challenges remain to be addressed, however. While most
Furthermore, a novel cluster grouping scheme is proposed works assume perfect channel knowledge, accurate channel
in [126] to reduce the inter-cluster interference for mmWave estimation requires a significant training overhead, which
MIMO-NOMA with hybrid precoding. MIMO-NOMA may have a non-negligible impact on spectral efficiency.
using the THz band is studied in [127], where user cluster- Furthermore, while most works consider perfect SIC, error
ing, hybrid precoding and power allocation are optimized to propagation remains a importance consideration in practical
maximize energy efficiency. SIC for NOMA systems [107]. Finally, the efficient design
The integration of RIS and NOMA technologies is also of modulation and detection schemes for NOMA remains an
considered an attractive option since the careful deploy- open problem since most works base their analysis on the
ment and selection of reflection coefficients of RISs can ideal Gaussian signaling [135].
increase the channel disparity among users, which would
lead to higher NOMA gains [107]. Furthermore, this inte-
gration can aid in satisfying the QoS constraints of users, 8 Optical Wireless Communications
since the QoS constraints are not necessarily of the same
order as the decoding order, which is dictated by channel Optical wireless communications (OWC) is an efficient
conditions. The ability to modify the channels through RIS and mature technology that has been developed alongside
can help in satisfying the constraints. Hence, there has been cellular technology, which has only used radio spectrum.
a significant interest in combining the two technologies. OWC can potentially satisfy the demanding requirements
The work in [128] considers a downlink MISO RIS-aided at the backhaul and access network levels beyond 5G net-
NOMA system, where the active beamforming of the BS works. As the 6G development gains momentum, compre-
and the passive beamforming of the RIS are jointly opti- hensive research activities are being carried out on develop-
mized to maximize the sum rate of all users subject to vari- ing OWC-based solutions capable of delivering ubiquitous,
ous constraints. Alternating optimization is used to solve ultra-high-speed, low-power consumption, highly secure,
the resulting non-convex optimization problems. Further- and low-cost wireless access in diverse application sce-
more, low-complexity user ordering schemes are proposed narios [136, 137]. In particular, this includes using hybrid
in [129], which achieve close performance to the exhaustive networks that combine OWC with radio frequency or wired/
search used in [128]. A signal cancellation design is devel- fiber-based technologies. Solutions for IoT connectivity in
oped in [130], where the reflection coefficients are selected smart environments are being investigated for developing
to reduce inter-cluster interference of the different NOMA flexible and efficient backhaul/fronthaul OWC links with
clusters. The signal cancellation enabled by RIS relaxes the low latency and support for access traffic growth [138].
constraints on the number of transmit and receive antennas. The OWC technology covers the three optical bands of
Given the growing appreciation for the power of deep infrared (IR: 187-400 THz, 750-1600 nm wavelength), vis-
learning, it has also been applied to solve various NOMA- ible light (VL: 400-770 THz, 390-750 nm) and ultraviolet
related research problems. Deep learning is used to acquire (UV: 1000-1500 THz, 200-280 nm). Free space optics (FSO)

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 447

and visible light communications (VLC) are commonly used is performed in the electrical domain, and intensity modu-
terms to describe various forms of OWC technology [139]. lation/direct detection is the most practical scheme. LEDs
FSO mainly refers to the use of long-range, high-speed with large FoV or laser diodes with a small FoV encode
point-to-point outdoor/space laser links in the IR band [140], and transmit data over the line-of-sight (LOS)/NLOS opti-
while VLC relies on LEDs operating in the VL band, mostly cal channel. Photo-detectors at the receiver convert data,
in indoor vehicular environments [141]. carrying light intensity back to electrical signals for base-
In comparison to RF, OWC systems offer significant band processing. A VLC-enabled device inside a pocket
technical and operational advantages including, but not or briefcase cannot be connected optically, which is one
limited to i) huge bandwidth, which leads to high data example of why a hybrid optical-radio wireless network is
rates; e.g., a recent FSO system achieved a world record needed. A reconfigurable optical-radio network is a high
data rate of 13.16 Tbps over a distance of 10 km [139], and performance and highly flexible communications system
multiple Gbps in indoor VLC setups [142]; ii) operation in that can be adapted for changing situations and different
the unregulated spectrum, thus no license fees and associ- scenarios [143].
ated costs; iii) immunity to the RF electromagnetic inter- Performance-wise, data throughput below 100 Mbps
ference; iv) a high degree of spatial confinement, offering can be achieved with relatively simple optical transceivers
virtually unlimited frequency reuse capability, inherent and off-the-shelf components. Data rates of up to hundreds
security at the physical layer, and no interference with other of Gbps have been demonstrated in laboratory conditions,
devices; v) a green technology with high energy efficiency and it is expected that even Tbps-communications will be
due to low power consumption and reduced interference. achieved in the future.
With such features, OWC is well-positioned to be a prevail- Open research directions in OWC and in VLC toward
ing complement to RF wireless solutions from micro- to 6G include:
macro-scale applications, including intra/inter-chip con-
nections and indoor wireless access (WA) localization, • Accurate VLC channel modeling and characterization for
underwater, outdoor and space point-to-point links, etc. various deployment scenarios with a particular emphasis
Beyond the state-of-the-art, however, the dominance of on user-dense environments. Incorporating user mobility
RF-based WA technologies will be challenged. LiFi [142] and device orientation into the VLC channel models and
is a promising technology to provide local broadband con- combining VLC and RF systems [144, 145].
nectivity [141]. As shown in Fig. 7, VLC provides high- • New non-coherent physical-layer transmission schemes
speed, bi-directional, networked data delivery through the such as spatial modulation and its variations can be used,
lighting infrastructure. When a device moves out of the as well as non-orthogonal communication schemes such
light cone of one light source, the services can be handed as MIMO [144, 145].
over to the next light source, or eventually, the device can • Exploiting R-G-B LEDs, developing new materials and
be connected and handed over to an RF-based system if optoelectronic devices (e.g., fast non phosphorous LEDs,
optical access is no longer provided. In VLC, all the base- micro-LEDs), very fast switching mechanisms between
band signal processing at the transmitter and the receiver optical and radio systems, etc. [146].

Figure 7  Bi-directional, point-

to-point and mobile communi-
cations with networked wireless
access.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

448 Journal of Signal Processing Systems (2023) 95:435–457

• Use of OWC to provide secure and safe connectivity in in- perpetual data upload and download, caching, and inter-net-
body communications applications, including communica- working. Hence, security should be considered an essential
tions to and from the body [147], communications between performance requirement in 6G systems, and signal process-
sensors inside the body, etc. Recent results have shown that ing can strongly support it.
near-infrared light can be used for this purpose [148, 149]. Physical layer security (PLS) can play a vital role in
• Design of new and novel optical IoT, new devices, and enhancing cyber-security in 6G wireless networks. It
interfaces to tackle the efficient generation, routing, detec- refers to transmission schemes that exploit dissimilarities
tion, and processing of optical signals [150]. among the channels of different receivers to hide informa-
• For ultra-dense IoT scenarios, there are a number of open tion from unauthorized users without reliance on upper
challenges that appeal for a radical rethinking of network network layer encryption techniques. The secrecy capacity
topologies and the design of media access control and net- is used as a performance measure to determine the maxi-
work layers in OWC [151]. mum communication rate that guarantees the authorized
• In VLC, to account for multi-user scenarios and user mobil- receiver’s reliable reception of the secret message. PLS
ity, robust low-complexity multiple access techniques need mechanisms, that are mainly based on advanced signal
to be designed, together with efficient cellular architectures processing techniques, will also help reduce the latency
with user-scheduling and intra-room handover capability, and the complexity of novel security algorithms.
achieving high capacity, low latency, and fairness [138, Two well-known PLS techniques are based on either
152]. applying beamforming in the direction of the legitimate
• At the MAC layer, robust link quality estimators will be user or generating a friendly jamming signal that creates
developed due to the small packet sizes used in machine- an artificial noise, which lies in the null space of the legiti-
to-machine (M2M) applications and constraints on sensor mate user. After combining the confidential information
devices. Routing algorithms will be devised taking into with the jamming signal at the transmitter side, only the
account the optimal trade-off between the link capacity, eavesdropper will experience destructive effects from the
connectivity, latency, and energy consumption [141, 153, jamming signal [157–159]. These techniques are anyway
154]. based on having knowledge of the location of the eaves-
• In medium-range OWC, the effects of weather and environ- dropper or at least an estimation of its channel state infor-
mental conditions, ambient noise, and link misalignment mation, which is often hard to get. In [160, 161] a noise-
need to be investigated to enable connectivity between dis- loop modulation is proposed, which uses equipment noise
tant vehicles. Physical-layer designs need to be built upon to modulate the information bits to provide confidentiality
multi-hop transmission to reduce the delay in transmission without any knowledge about the eavesdropper.
of road safety-related information [138, 141, 154].
• For long-range links, extensive research should be carried
out to minimize the terminal size to enable the technology 9.2 PLS Through Optical Wireless Communications
to be integrated into small satellites, e.g., CubeSats, with
data rates up to 10 Gbps and for the investigation of how to In VLC, PLS is especially important when a large physical
deal with cloud obstruction. Site diversity techniques and indoor environment is accessible to or shared by multiple
smart route selection algorithms should be devised for satel- users and potential eavesdroppers. Some examples include
lite links and airborne networks, respectively. Also, hybrid meeting rooms, public libraries, airplanes, hospitals, etc.
RF/FSO and optimized multi-hop transmission techniques Light does not propagate through opaque objects (e.g.
should be investigated to improve link reliability between walls), it is directional and light beams can be formed with
satellites or high altitude platforms (HAPs) [155, 156]. signal processing efforts. It is, therefore, possible to sig-
nificantly reduce the possibilities of man-in-middle attacks
in LiFi compared to WiFi (Fig. 8). It has been shown that
9 Physical‑Layer Security the secrecy capacity of a LiFi network is 20 times higher
than that of a WiFi network.
9.1 PLS Through Wireless Communications PLS methods employing signal processing techniques in
MIMO-VLC have been proposed recently. In MIMO-based
Given the intrinsic nature of future 6G services, with the index-modulation (IM) techniques [157, 162], a random
increase in traffic volumes over wireless networks, data pri- switching among the antennas (LEDs) is exploited to gener-
vacy and security are a predominant concern for users and ate a strong and friendly jamming signal, which is invaluable
network administrators. Providing wireless networks both in for PLS applications. In precoding approaches [163–165],
RF and optical domains with trusted communications is a the channel state information at the transmitter (CSIT) of
crucial objective for successfully deploying services, such as the legitimate user is used to construct the precoding matrix

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 449

compare the performance of ML models and algorithms

[166]. Training at wide ranges of SNR should be consid-
ered as it severely impacts training time. Moreover, wireless
communication systems usually deal with complex baseband
representation while the most used neural networks (NN)
utilize real arithmetic. Therefore, to meet the requirements
of 6G, there is a need for NNs to operate with complex
numbers. ML could be exploited at the MAC layer to have
an adaptive control channel based on the traffic and other
requirements of the connected devices. ML also could be
exploited at the network layer to develop routing protocols.

10.2 Integrated Communications and Sensing

Figure 8  WiFi-LiFi hybrid communications.

6G is envisioned to continue to transform from connected

coefficients so that the confidential message is perceived by people and connected things, to connected “intelligences”.
the legitimate user clearly while the eavesdropper’s bit error The mobile network will connect large numbers of intel-
rate (BER) performance is degraded substantially. ligent devices, equipped with not only communication but
also sensing capabilities at the same time. These intelligent
nodes will have the capability to sense their surroundings,
10 Research Directions and Challenges and exchange their observations through communication:
the cognitive circle (sense-communicate-think-act) applied
In addition to the core signal processing technologies to intelligent networks. AI can be combined with sens-
described in the previous section, there are several emerg- ing capability such that the network will have human-like
ing technologies possibly having a significant impact on 6G cognition capabilities [167]. Communications and sensing
systems. A few possible ones are briefly discussed below. can be integrated and work jointly to give benefits to each
other [168]. When those are integrated to a single device to
10.1 Machine Learning for 6G Design coexist based on different system designs and waveforms
(possibly operating on separate frequency bands), the term
The availability of a large amount of data, advanced tech- ICAS is often used. When the systems and waveforms are
nological progress, the revolution in optimization tools, the jointly designed to serve both functionalities, the solution is
availability of powerful processing units, a huge amount of often called joint communications and sensing (JCS). Both
available memory, and systematic data mining and extrac- approaches can provide the system with accurate localiza-
tion techniques may jointly lead the basis to the AI utiliza- tion, imaging, and, in general, high-resolution environment
tion to achieve further improvements to the physical, MAC, map. Those can improve communication performance, open-
and network layers in B5G networks. However, the avail- ing the way to a broad range of new services [169].
able resources and the technological development remain The use of higher frequency spectrum enables a high-
limited, hence, the ML paradigms look promising in 6G as accuracy localization and tracking, together with the capa-
the communication performance requirements keep aggres- bility of imaging and 3D mapping of the environment where
sively increasing. Although a plethora of machine learning communication subjects (humans, machines, things) are
(ML) based 5G can be seen in literature, the concepts are immersed. Similarly, an accurate sensing capability in/on
taken from mature technologies, such as computer vision the human body could open the way to augmented human
and natural language processing domains, and exploited in communications and very accurate gesture/activity recogni-
communication systems, and hence, many challenges are tion [170].
raised such as the selection of optimal data representations, In such scenario, the role of signal processing is crucial to
loss-functions, and training strategies. In other words, 6G model the environment and adapt the communication: from
should have its own definitions, algorithms, techniques, one side the model-based methods (geometric optics, statis-
and tools of ML for wireless communication scenarios. The tical signal processing, optimization theory, etc.) [171] and,
training complexity and the generalization capabilities of the from the other side, model-free methods (data-driven ML
trained models in wireless communications are really chal- and AI) [172]. Although the former methods are rigorous
lenges. Most existing models are not adaptable to changes and provide certain performance and design optimization,
in channel statistics, realizations, and modulation orders. the latter can be used when the modelling is too complex to
In addition, there is a lack of datasets to benchmark and be obtained or to quickly adapt to changes.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

450 Journal of Signal Processing Systems (2023) 95:435–457

conditioning of the signal at transmission to MIMO detec-

tion, from channel coding to channel estimation, from
multicarrier (radio) modulation to optical wireless com-
munications. Physical-layer security and bio-nano in-body
communications are also considered as an important part
of next generation networks.
DPD in the THz band large array transmitters is a sig-
nificant challenge together with the design of efficient
transmit and receive processing chains providing energy
and power efficient solutions given the large numbers of
ADCs and digital-to-analog converters. Channel estima-
tion is challenging also due to the rapid channel varia-
tions and phase noise in the THz band. The large antenna
arrays and the introduction of RIS technology further
complicates the processing. Efficient channel coding and
Figure 9  In-body bio nano networks as part of the Net. practical implementation of Tb/s decoders is by no means
trivial and requires both algorithmic and architectural
innovations. The multicarrier OFDM based air interfaces
10.3 Internet of Bio‑Nano‑Things and the design of NOMA solutions together with realistic
transceivers is an important challenge determining to large
The Internet of Things (IoT) is one of the most important ele- extent the power and energy efficiency characteristics of
ment of 6G systems. Things refer to interconnected machines the emerging 6G networks. OWC is a promising solution
and objects with embedded computing capabilities employed for special indoor use cases providing inherent security
to extend the Internet to many application domains such as and avoiding the electromagnetic interference problems
health, home, office, transportation, food, space, ocean, typical for the microwave and THz bands. PLS may also
etc. 6G should be the connectivity “tissue” that makes all solve elegantly some of the security and privacy concerns,
these different resourced and sized devices to connect each which are very significant in the data avalanche enabled in
other. There are many application domains where Things are part by 5G and even more by 6G.
required to be tiny, concealable, and non-intrusive. Nanoma- All the above mentioned signal processing challenges still
terials (graphene, etc.) have stimulated the recent concept of require significant research efforts. In addition, some more
Internet of NanoThings (IoNT), referring to interconnection future directions were also identified and discussed. There
of nano-scale (artificial) devices. More recently bio-nano- is a room for fundamental research to exploit the artificial
devices have been investigated, as a potential disrupting intelligence and deep learning in the design of DPD, pre-
method to take and deliver information inside the human coders and detectors. What is more, ML is expected to play
body. The concept of Internet of Bio-Nano Things (IoBNT) some role in the design of actual air interfaces, modulation,
was firstly introduced in [173], envisioning the use of natu- waveforms, channel decoders etc. The benefit of such an
ral or engineered bio-devices with embedded computing and approach is still debatable and more research is necessary.
communication capabilities (by using bio-chemical signals). The merger of communications and sensing will be one key
6G networks should support the exchange of information from feature of 6G and AI/ML will play a role therein, too. The
inside the body to Internet (Fig. 9), following the concept that internet of bio-nano-things was also discussed as one poten-
the human body will be part of the global Network [174]. tial direction for future networks. It may, however, be more
Bio-Nano-Things, together with the 6G support, could enable a technology for 7G than 6G. Wireless quantum computing
applications such as intra-body sensing and actuation systems and communications is also an important emerging area,
for a new era of health monitoring and treatment [175]. which may also mature for practical realization in the 7G
systems expected to emerge in 2040s.

11 Summary and Conclusions Author Contributions L.M., S.S. and M.J. contributed to the conceptu-
alization and organization of this work; all authors contributed equally
In this paper an overview of the signal processing tech- to the contents of this article.
niques for future 6G networks is provided. The sig- Funding Open access funding provided by Università degli Studi di
nal processing chain is considered as a whole, from the Firenze within the CRUI-CARE Agreement.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 451

Availability of Data and Material No data or other material was pro- 9. Xiao, Z., & Zeng, Y. (2021). An overview on integrated localiza-
duced in this study. tion and communication towards 6G. Science China Information
Sciences, 65(3). https://​doi.​org/​10.​1007/​s11432-​020-​3218-8
Declarations 10. Zong, B., Fan, C., Wang, X., Duan, X., Wang, B., & Wang, J.
(2019). 6G technologies: Key drivers, core requirements, system
Ethics Approval Research involved no Human Participants and/or architectures, and enabling technologies. IEEE Vehicular Tech-
Animals. nology Magazine, 14(3), 18–27. https://​doi.​org/​10.​1109/​MVT.​
2019.​29213​98
11. Viswanathan, H., & Mogensen, P. E. (2020). Communications
Conflicts of Interest The authors have no conflict of interest/compet- in the 6G era. IEEE Access, 8, 57063–57074. https://​doi.​org/​10.​
ing interests to declare that are relevant to the content of this article. 1109/​ACCESS.​2020.​29817​45
12. Tariq, F., Khandaker, M. R. A., Wong, K. K., Imran, M. A.,
Open Access This article is licensed under a Creative Commons Attri- Bennis, M., & Debbah, M. (2020). A speculative study on 6G.
bution 4.0 International License, which permits use, sharing, adapta- IEEE Wireless Communications, 27(4), 118–125. https://d​ oi.o​ rg/​
tion, distribution and reproduction in any medium or format, as long 10.​1109/​MWC.​001.​19004​88
as you give appropriate credit to the original author(s) and the source, 13. Matthaiou, M., Yurduseven, O., Ngo, H. Q., Morales-Jimenez,
provide a link to the Creative Commons licence, and indicate if changes D., Cotton, S. L., & Fusco, V. F. (2021). The road to 6G: Ten
were made. The images or other third party material in this article are physical layer challenges for communications engineers. IEEE
included in the article's Creative Commons licence, unless indicated Communications Magazine, 59(1), 64–69. https://​doi.​org/​10.​
otherwise in a credit line to the material. If material is not included in 1109/​MCOM.​001.​20002​08
the article's Creative Commons licence and your intended use is not 14. Elhoushy, S., Ibrahim, M., & Hamouda, W. (2021). Cell-free
permitted by statutory regulation or exceeds the permitted use, you will massive MIMO: A survey. IEEE Communications Surveys Tuto-
need to obtain permission directly from the copyright holder. To view a rials, 1–1. https://​doi.​org/​10.​1109/​COMST.​2021.​31232​67
copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. 15. Yang, H., & Marzetta, T. L. (2013). Capacity performance of mul-
ticell large-scale antenna systems. In 2013 51st Annual Allerton
Conference on Communication, Control, and Computing (Aller-
ton) (pp. 668–675). IEEE. https://​doi.​org/​10.​1109/​Aller​ton.​2013.​
67365​89
References 16. Ding, L., Yang, Z., & Gandhi, H. (2012). Concurrent dual-
band digital predistortion. In 2012 IEEE/MTT-S International
1. Huang, T., Yang, W., Wu, J., Ma, J., Zhang, X., & Zhang, D. Microwave Symposium Digest (pp. 1–3). https://d​ oi.o​ rg/1​ 0.1​ 109/​
(2019). A survey on green 6G network: Architecture and tech- MWSYM.​2012.​62595​32
nologies. IEEE Access, 7, 175758–175768. https://​doi.​org/​10.​ 17. Younes, M., Kwan, A., Rawat, M., & Ghannouchi, F. M. (2013).
1109/​ACCESS.​2019.​29576​48 Linearization of concurrent tri-band transmitters using 3-D
2. Latva-aho, M., & Leppanen, L. (Eds.). (2019). Key drivers and phase-aligned pruned Volterra model. IEEE Transactions on
research challenges for 6G ubiquitous wireless intelligence. Microwave Theory and Techniques, 61(12), 4569–4578. https://​
White paper. Retrieved February 15, 2022, from http://​urn.​fi/​ doi.​org/​10.​1109/​TMTT.​2013.​22871​76
urn:​isbn:​97895​26223​544 18. Ceylan, N., Mueller, J., Pittorino, T., & Weigel, R. (2003). Mobile
3. Chowdhury, M. Z., Shahjalal, M., Ahmed, S., & Jang, Y. M. phone power amplifier linearity and efficiency enhancement
(2020). 6G wireless communication systems: Applications, using digital predistortion. In 33rd European Microwave Con-
requirements, technologies, challenges, and research directions. ference Proceedings (IEEE Cat. No.03EX723C) (vol. 1, pp.
IEEE Open Journal of the Communications Society, 1, 957–975. 269–272). https://​doi.​org/​10.​1109/​EUMC.​2003.​12622​71
https://​doi.​org/​10.​1109/​OJCOMS.​2020.​30102​70 19. Tervo, N., Jokinen, M., Leinonen, M. E., Aikio, J., Kursu, O.,
4. Saad, W., Bennis, M., & Chen, M. (2020). A vision of 6G Rahkonen, T. & Pärssinen, A. (2019). Digital predistortion con-
wireless systems: Applications, trends, technologies, and open cepts for linearization of mmW phased array transmitters. In
research problems. IEEE Network, 34(3), 134–142. https://​doi.​ 2019 16th International Symposium on Wireless Communication
org/​10.​1109/​MNET.​001.​19002​87 Systems (ISWCS) (pp. 325–329). IEEE. https://​doi.​org/​10.​1109/​
5. Letaief, K. B., Chen, W., Shi, Y., Zhang, J., & Zhang, Y. J. A. ISWCS.​2019.​88772​85
(2019). The roadmap to 6G: AI empowered wireless networks. 20. Lee, S., Kim, M., Sirl, Y., Jeong, E. R., Hong, S., Kim, S., &
IEEE Communications Magazine, 57(8), 84–90. https://​doi.​org/​ Lee, Y. H. (2015). Digital predistortion for power amplifi-
10.​1109/​MCOM.​2019.​19002​71 ers in hybrid MIMO systems with antenna subarrays. In 2015
6. Giordani, M., Polese, M., Mezzavilla, M., Rangan, S., & Zorzi, IEEE 81st Vehicular Technology Conference (VTC Spring) (pp.
M. (2020). Toward 6G networks: Use cases and technologies. 1–5). https://​doi.​org/​10.​1109/​VTCSp​ring.​2015.​71457​77
IEEE Communications Magazine, 58(3), 55–61. https://​doi.​org/​ 21. Tervo, N., Aikio, J., Tuovinen, T., Rahkonen, T., & Parssinen, A.
10.​1109/​MCOM.​001.​19004​11 (2017). Digital predistortion of amplitude varying phased array
7. Zhang, Z., Xiao, Y., Ma, Z., Xiao, M., Ding, Z., Lei, X., Karagi- utilising over-the-air combining. In 2017 IEEE MTT-S Interna-
annidis, G. K., & Fan, P. (2019). 6G wireless networks: Vision, tional Microwave Symposium (IMS) (pp. 1165–1168). https://d​ oi.​
requirements, architecture, and key technologies. IEEE Vehicu- org/​10.​1109/​MWSYM.​2017.​80588​09
lar Technology Magazine, 14(3), 28–41. https://​doi.​org/​10.​1109/​ 22. Haykin, S. (2007). Neural networks: A comprehensive founda-
MVT.​2019.​29212​08 tion. Prentice-Hall, Inc.
8. Jagannath, A., Jagannath, J., & Melodia, T. (2021). Redefining 23. Jaraut, P., Rawat, M., & Ghannouchi, F. M. (2018). Composite
wireless communication for 6G: Signal processing meets deep neural network digital predistortion model for joint mitigation
learning with deep unfolding. IEEE Transactions on Artificial of crosstalk, I∕Q imbalance, nonlinearity in MIMO transmitters.
Intelligence, 2(6), 528–536. https://​doi.​org/​10.​1109/​TAI.​2021.​ IEEE Transactions on Microwave Theory and Techniques, 66(11),
31081​29 5011–5020. https://​doi.​org/​10.​1109/​TMTT.​2018.​28696​02

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

452 Journal of Signal Processing Systems (2023) 95:435–457

24. Wang, Z., Zhang, J., Björnson, E., & Ai, B. (2021). Uplink per- 39. Hum, S. V., Okoniewski, M., & Davies, R. J. (2007). Modeling
formance of cell-free massive MIMO over spatially correlated and design of electronically tunable reflectarrays. IEEE Transac-
Rician fading channels. IEEE Communications Letters, 25(4), tions on Antennas and Propagation, 55(8), 2200–2210. https://​
1348–1352. https://​doi.​org/​10.​1109/​LCOMM.​2020.​30418​99 doi.​org/​10.​1109/​TAP.​2007.​902002
25. Prabhu, H. Rodrigues, J. Edfors, O., & Rusek, F. 40. Perez-Palomino, G., et al. (2015). Design and demonstration of
(2013). Approximative matrix inverse computations for very- an electronically scanned reflectarray antenna at 100 GHz using
large MIMO and applications to linear pre-coding systems. In multiresonant cells based on liquid crystals. IEEE Transactions
2013 IEEE Wireless Communications and Networking Con- on Antennas and Propagation, 63(8), 3722–3727. https://d​ oi.o​ rg/​
ference (WCNC) (pp. 2710–2715). https://​d oi.​o rg/​1 0.​1 109/​ 10.​1109/​TAP.​2015.​24344​21
WCNC.​2013.​65549​90 41. Basar, E., & Yildirim, I. (2020). Indoor and outdoor physical
26. Wu, J., Fang, S., Li, L., & Yang, Y. (2014). QR decomposition channel modeling and efficient positioning for reconfigur-
and Gram Schmidt orthogonalization based low-complexity able intelligent surfaces in mmWave bands. Preprint retrieved
multi-user MIMO precoding. In 10th International Conference from https://​arxiv.​org/​abs/​2006.​02240
on Wireless Communications, Networking and Mobile Computing 42. Huang, C., Zappone, A., Alexandropoulos, G. C., Debbah, M., &
(WiCOM 2014) (pp. 61–66). https://d​ oi.o​ rg/1​ 0.1​ 049/i​ c.2​ 014.0​ 074 Yuen, C. (2019). Reconfigurable intelligent surfaces for energy
27. Jacobsson, S., Durisi, G., Coldrey, M., Goldstein, T., & Studer, efficiency in wireless communication. IEEE Transactions on
C. (2017). Quantized precoding for massive MU-MIMO. IEEE Wireless Communications, 18(8), 4157–4170. https://​doi.​org/​
Transactions on Communications, 65(11), 4670–4684. https://​ 10.​1109/​TWC.​2019.​29226​09
doi.​org/​10.​1109/​TCOMM.​2017.​27230​00 43. Shlezinger, N., Alexandropoulos, G. C., Imani, M. F., Eldar, Y.
28. Yan, H., Tian, T., Chen, L., & Qiu, J. (2012). A physical layer C., & Smith, D. R. (2021). Dynamic metasurface antennas for 6G
solution for Tomlinson-Harashima precoding in the framework extreme massive MIMO communications. IEEE Wireless Com-
of LTE-advanced. In 2012 IEEE Globecom Workshops (pp. muications (pp. 1–8). https://​doi.​org/​10.​1109/​MWC.​001.​20002​67
291–296). https://​doi.​org/​10.​1109/​GLOCO​MW.​2012.​64775​85 44. Wymeersch, H., He, J., Denis, B., Clemente, A., & Juntti, M.
29. Chen, R., Moretti, M., & Wang, X. (2017). Hybrid TH-VP pre- (2020). Radio localization and mapping with reconfigurable
coding for multiuser MIMO. IEEE Transactions on Vehicular intelligent surfaces: Challenges, opportunities, and research
Technology, 66(12), 11399–11403. https://d​ oi.o​ rg/1​ 0.1​ 109/T ​ VT.​ directions. IEEE Vehicular Technology Magazine, 15(4), 52–61.
2017.​27259​15 https://​doi.​org/​10.​1109/​MVT.​2020.​30236​82
30. Ngo, H. Q., Ashikhmin, A., Yang, H., Larsson, E. G., & Marzetta, 45. He, J., Wymeersch, H., Kong, L., Silvén, O., & Juntti, M. (2020). Large
T. L. (2017). Cell-free massive MIMO versus small cells. IEEE intelligent surface for positioning in millimeter wave MIMO sys-
Transactions on Wireless Communications, 16(3), 1834–1850. tems. In Proceedings of IEEE VTC2020-Spring (pp. 1–5). https://​
https://​doi.​org/​10.​1109/​TWC.​2017.​26555​15 doi.​org/​10.​1109/​VTC20​20-​Sprin​g48590.​2020.​91290​75
31. Nayebi, E., Ashikhmin, A., Marzetta, T. L., Yang, H., & Rao, B. 46. He, J., Wymeersch, H., Sanguanpuak, T., Silvén, O., & Juntti, M.
D. (2017). Precoding and power optimization in cell-free mas- (2020). Adaptive beamforming design for mmWave RIS-aided
sive MIMO systems. IEEE Transactions on Wireless Communi- joint localization and communication. In Proceedings of IEEE
cations, 16(7), 4445–4459. https://​doi.​org/​10.​1109/​TWC.​2017.​ WCNC Workshops (WCNCW) (pp. 1–6). https://d​ oi.o​ rg/1​ 0.1​ 109/​
26984​49 WCNCW​48565.​2020.​91248​48
32. Björnson, E., & Sanguinetti, L. (2020). Scalable cell-free massive 47. Wu, Q., & Zhang, R. (2019). Intelligent reflecting surface
MIMO systems. IEEE Transactions on Communications, 68(7), enhanced wireless network via joint active and passive beam-
4247–4261. https://​doi.​org/​10.​1109/​TCOMM.​2020.​29873​11 forming. IEEE Transactions on Wireless Communications,
33. Interdonato, G., Karlsson, M., Bjornson, E., & Larsson, E. G. 18(11), 5394–5409. https://d​ oi.o​ rg/1​ 0.1​ 109/T ​ WC.2​ 019.2​ 93602​ 5
(2018). Downlink spectral efficiency of cell-free massive MIMO 48. Björnson, E., Özdogan, O., & Larsson, E. G. (2019). Intelligent
with full-pilot zero-forcing. In 2018 IEEE Global Conference reflecting surface vs. decode-and-forward: How large surfaces
on Signal and Information Processing (GlobalSIP) (pp. 1003– are needed to beat relaying? IEEE Wireless Communications Let-
1007). https://​doi.​org/​10.​1109/​Globa​lSIP.​2018.​86466​66 ters, 1–1. https://​doi.​org/​10.​1109/​LWC.​2019.​29506​24
34. Interdonato, G., Karlsson, M., Björnson, E., & Larsson, E. G. 49. Guo, H., Liang, Y. C., Chen, J., & Larsson, E. G. (2019). Weighted
(2020). Local partial zero-forcing precoding for cell-free massive sum-rate maximization for intelligent reflecting surface enhanced
MIMO. IEEE Transactions on Wireless Communications, 19(7), wireless networks. In IEEE Global Communications Confer-
4758–4774. https://​doi.​org/​10.​1109/​TWC.​2020.​29870​27 ence (GLOBECOM) (pp. 1–6). IEEE. https://​doi.​org/​10.​1109/​
35. Huang, C., et al. (2020). Holographic MIMO surfaces for 6G GLOBE​COM38​437.​2019.​90132​88
wireless networks: Opportunities, challenges, and trends. IEEE 50. Zhang, Y., Zhong, C., Zhang, Z., & Lu, W. (2020). Sum rate opti-
Wireless Communications, 27(5), 118–125. https://​doi.​org/​10.​ mization for two way communications with intelligent reflect-
1109/​MWC.​001.​19005​34 ing surface. IEEE Communications Letters, 24(5), 1090–1094.
36. Letaief, K. B., Chen, W., Shi, Y., Zhang, J., & Zhang, Y. A. https://​doi.​org/​10.​1109/​LCOMM.​2020.​29783​94
(2019). The roadmap to 6G: AI empowered wireless networks. 51. Nguyen, N. T., Vu, Q. D., Lee, K., & Juntti, M. (2021). Hybrid
IEEE Communications Magazine, 57(8), 84–90. https://​doi.​org/​ relay-reflecting intelligent surface-assisted wireless communica-
10.​1109/​MCOM.​2019.​19002​71 tion. Preprint retrieved from https://​arxiv.​org/​abs/​2103.​03900
37. Di Renzo, M., et al. (2020). Smart radio environments empow- 52. Shahabuddin, S., Islam, M. H., Shahabuddin, M. S., Albreem, M.
ered by reconfigurable intelligent surfaces: How it works, state A., & Juntti, M. (2020). Matrix decomposition for massive MIMO
of research, and the road ahead. IEEE Journal on Selected Areas detection. In 2020 IEEE Nordic Circuits and Systems Conference
in Communications, 38(11), 2450–2525. https://d​ oi.o​ rg/1​ 0.1​ 109/​ (NorCAS) (pp. 1–6). https://d​ oi.o​ rg/1​ 0.1​ 109/N
​ orCAS ​ 51424.2​ 020.​
JSAC.​2020.​30072​11 92649​98
38. Wu, Q., Zhang, S., Zheng, B., You, C., & Zhang, R. (2021). Intel- 53. He, H., Wen, C., Jin, S., & Li, G. Y. (2020). Model-driven deep
ligent reflecting surface aided wireless communications: A tuto- learning for MIMO detection. IEEE Transactions on Signal
rial. IEEE Transactions on Communications, 1–1. https://d​ oi.o​ rg/​ Processing, 68, 1702–1715. https://​doi.​org/​10.​1109/​TSP.​2020.​
10.​1109/​TCOMM.​2021.​30518​97 29765​85

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 453

54. Samuel, N., Diskin, T., & Wiesel, A. (2019). Learning to detect. 70. Ping, L., Liu, L., Wu, K., & Leung, W. K. (2006). Interleave
IEEE Transactions on Signal Processing, 67(10), 2554–2564. division multiple-access. IEEE Transactions on Wireless Com-
https://​doi.​org/​10.​1109/​TSP.​2019.​28998​05 munications, 5(4), 938–947. https://​doi.​org/​10.​1109/​TWC.​2006.​
55. Khani, M., Alizadeh, M., Hoydis, J., & Fleming, P. (2020). Adap- 16189​43
tive neural signal detection for massive MIMO. IEEE Transac- 71. Zhang, Y., Peng, K., Chen, Z., & Song, J. (2019). Construc-
tions on Wireless Communications, 1–1. https://d​ oi.o​ rg/1​ 0.1​ 109/​ tion of rate-compatible raptor-like Quasi-Cyclic LDPC code
TWC.​2020.​29961​44 with edge classification for IDMA based random access. IEEE
56. Goutay, M., Aoudia, F. A., & Hoydis, J. (2020). Deep HyperNetwork- Access, 7, 30818–30830. https://​d oi.​o rg/​1 0.​1 109/​ACCESS.​
based MIMO detection. In 2020 IEEE 21st International Work- 2019.​29028​84
shop on Signal Processing Advances in Wireless Communications 72. Wehn, N., Sahin, O., & Herrmann, M. (2021). Forward-error-
(SPAWC). IEEE. https://​doi.​org/​10.​1109/​spawc​48557.​2020.​91542​83 correction for Beyond-5G ultra-high throughput communica-
57. Li, K., Chen, Y., Sharan, R., Goldstein, T., Cavallaro, J. R., & tions. In 2021 11th International Symposium on Topics in Cod-
Studer, C. (2016). Decentralized data detection for massive MU- ing (ISTC) (pp. 1–5). https://​doi.​org/​10.​1109/​ISTC4​9272.​2021.​
MIMO on a Xeon Phi cluster. In Proceedings Asilomar Confer- 95941​26
ence on Signals, Systems and Computers (pp. 468–472). https://​ 73. Perry, J., Balakrishnan, H., & Shah, D. (2011). Rateless spinal
doi.​org/​10.​1109/​ACSSC.​2016.​78690​83 codes. In Proceedings of the 10th ACM Workshop on Hot Topics
58. Li, K., Sharan, R. R., Chen, Y., Goldstein, T., Cavallaro, J. R., in Networks (pp. 1–6). https://d​ oi.o​ rg/1​ 0.1​ 145/2​ 07056​ 2.2​ 07056​ 8
& Studer, C. (2017). Decentralized baseband processing for 74. Nachmani, E., Be’ery, Y., & Burshtein, D. (2016). Learning to
massive MU-MIMO systems. IEEE Journal on Emerging and decode linear codes using deep learning. In 2016 54th Annual
Selected Topics in Circuits System, 7(4), 491–507. https://​doi.​ Allerton Conference on Communication, Control, and Comput-
org/​10.​1109/​JETCAS.​2017.​27751​51 ing (Allerton) (pp. 341–346). https://​doi.​org/​10.​1109/​ALLER​
59. Jeon, C., Li, K., Cavallaro, J. R., & Studer, C. (2019). Decen- TON.​2016.​78522​51
tralized equalization with feedforward architectures for massive 75. Cammerer, S., Gruber, T., Hoydis, J., & ten Brink, S.
MU-MIMO. IEEE Transactions on Signal Processing, 67(17), (2017). Scaling deep learning-based decoding of polar codes
4418–4432. https://​doi.​org/​10.​1109/​TSP.​2019.​29289​47 via partitioning. In GLOBECOM 2017 - 2017 IEEE Global
60. Li, K., McNaney, J., Tarver, C., Castañeda, O., Jeon, C., Cavallaro, J. Communications Conference (pp. 1–6). https://​doi.​org/​10.​1109/​
R., & Studer, C. (2019). Design trade-offs for decentralized base- GLOCOM.​2017.​82548​11
band processing in massive MU-MIMO systems. In Proceedings of 76. Liang, F., Shen, C., & Wu, F. (2018). An Iterative BP-CNN
the Asilomar Conference on Signals, Systems and Computers (pp. Architecture for Channel Decoding. IEEE Journal of Selected
906–912). https://​doi.​org/​10.​1109/​IEEEC​ONF44​664.​2019.​90487​27 Topics in Signal Processing, 12(1), 144–159. https://​doi.​org/​10.​
61. Sanchez, J. R., Rusek, F., Sarajlic, M., Edfors, O., & Liu, L. 1109/​JSTSP.​2018.​27940​62
(2018). Fully decentralized massive MIMO detection based on 77. Schram, V., Moldovan, A., & Gerstacker, W. H. (2018). Com-
recursive methods. In Proceedings of the IEEE International pressive sensing for indoor THz channel estimation. In 2018
Workshop on Signal Processing Systems (pp. 53–58). https://​ 52nd Asilomar Conference on Signals, Systems, and Comput-
doi.​org/​10.​1109/​SiPS.​2018.​85983​21 ers (pp. 1539–1546). https://​doi.​org/​10.​1109/​ACSSC.​2018.​
62. Elias, P. (1955). Coding for noisy channels. IRE Convention Record, 86454​79
3, 37–46. 78. Sarieddeen, H., Alouini, M. S., & Al-Naffouri, T. Y. (2021).
63. 3GPP TS 38.212. (2021). NR; Multiplexing and channel coding. An overview of signal processing techniques for terahertz com-
V16.7.0 (2021-09), Release 16. munications. Proceedings of the IEEE. https://​doi.​org/​10.​1109/​
64. Shahabuddin, S., Janhunen, J., & Juntti, M. (2013). Design of a JPROC.​2021.​31008​11
transport triggered architecture processor for flexible iterative 79. Tan, J., & Dai, L. (2019). Delay-phase precoding for THz mas-
turbo decoder. In Proceedings of Wireless Innovation Forum sive MIMO with beam split. In 2019 IEEE Global Communica-
Conference on Wireless Communications Technologies and tions Conference (GLOBECOM) (pp. 1–6). https://​doi.​org/​10.​
Software Radio (SDR WINCOMM). 1109/​GLOBE​COM38​437.​2019.​90143​04
65. Shahabuddin, S., Janhunen, J., Bayramoglu, M. F., Juntti, M., 80. Brighente, A., Cerutti, M., Nicoli, M., Tomasin, S., & Spagnolini,
Ghazi, A., & Silvén, O. (2013). Design of a unified transport trig- U. (2020). Estimation of wideband dynamic mmWave and THz
gered processor for LDPC/turbo decoder. In 2013 International channels for 5G systems and beyond. IEEE Journal on Selected
Conference on Embedded Computer Systems: Architectures, Areas in Communications, 38(9), 2026–2040. https://​doi.​org/​10.​
Modeling, and Simulation (SAMOS) (pp. 288–295). 1109/​JSAC.​2020.​30008​89
66. Li, H., Yu, Z., Tongwei, L., Zheng, W., Feng, H., Ma, Z., & 81. Shao, X., Chen, X., Zhong, C., & Zhang, Z. (2020). Joint activity
Zhu, F. (2021). Novel memory efficient LDPC decoders for detection and channel estimation for MMW/THz wideband mas-
beyond 5G. Physical Communication, 101538. https://​doi.​org/​ sive access. In ICC 2020-2020 IEEE International Conference
10.​1016/j.​phycom.​2021.​101538 on Communications (ICC) (pp. 1–6). IEEE. https://​doi.​org/​10.​
67. Zhu, K., & Wu, Z. (2020). Comprehensive study on CC-LDPC, 1109/​ICC40​277.​2020.​91493​36
BC-LDPC and Polar code. In IEEE Wireless Communications 82. Tan, J., & Dai, L. (2021). Wideband channel estimation for THz
and Networking Conference Workshops (pp. 1–6). https://​doi.​ massive MIMO. China Communications, 18(5), 66–80. https://​
org/​10.​1109/​WCNCW​48565.​2020.​91248​97 doi.​org/​10.​23919/​JCC.​2021.​05.​005
68. Dang, H. N., & Van Nguyen, T. (2021). Protograph LDPC code 83. Dovelos, K., Matthaiou, M., Ngo, H. Q., & Bellalta, B. (2021).
design for LS-MIMO 1-bit ADC systems. REV Journal on Elec- Channel estimation and hybrid combining for wideband tera-
tronics and Communications, 11(1–2). https://​doi.​org/​10.​21553/​ hertz massive MIMO systems. IEEE Journal on Selected Areas
rev-​jec.​266 in Communications, 39(6), 1604–1620. https://​doi.​org/​10.​1109/​
69. Yuan, Y., Zhao, Y., Zong, B., & Parolari, S. (2020). Poten- JSAC.​2021.​30718​51
tial key technologies for 6G mobile communications. Science 84. Chen, Y., & Han, C. (2020). Deep CNN-based spherical-wave
China Information Sciences, 63, 1–19. https://​doi.​org/​10.​1007/​ channel estimation for terahertz ultra-massive MIMO systems.
s11432-​019-​2789-y In GLOBECOM 2020-2020 IEEE Global Communications

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

454 Journal of Signal Processing Systems (2023) 95:435–457

Conference (pp. 1–6). IEEE. https://​doi.​org/​10.​1109/​GLOBE​ on Communications, 69(6). https://​doi.​org/​10.​1109/​TCOMM.​

COM42​002.​2020.​93221​74 2021.​30632​36
85. Chen, Y., Yan, L., & Han, C. (2021). Hybrid spherical- and pla- 100. He, J., Wymeersch, H., & Juntti, M. (2021). Leveraging loca-
nar-wave modeling and DCNN-powered estimation of terahertz tion information for RIS-aided mmWave MIMO communica-
ultra-massive MIMO channels. IEEE Transactions on Commu- tions. IEEE Wireless Communications Letters, 1–1. https://​doi.​
nications, 69(10), 7063–7076. https://d​ oi.o​ rg/1​ 0.1​ 109/T
​ COMM.​ org/​10.​1109/​LWC.​2021.​30674​74
2021.​30986​96 101. Elhoushy, S., Ibrahim, M., & Hamouda, W. (2021). Cell-free
86. Srivastava, S., Tripathi, A., Varshney, N., Jagannatham, A. K., & massive MIMO: A survey. IEEE Communications Surveys Tuto-
Hanzo, L. (2021). Hybrid transceiver design for tera-hertz MIMO rials. https://​doi.​org/​10.​1109/​COMST.​2021.​31232​67
systems relying on Bayesian learning aided sparse channel estima- 102. Zhang, Y., Cao, H., Zhong, P., Qi, C., & Yang, L. (2018). Loca-
tion. Preprint retrieved from https://​arxiv.​org/​abs/​2109.​09664 tion-based greedy pilot assignment for cell-free massive MIMO
87. Balevi, E., & Andrews, J. G. (2021). Wideband channel estima- systems. In 2018 IEEE 4th International Conference on Com-
tion with a generative adversarial network. IEEE Transactions puter and Communications (ICCC) (pp. 392–396). https://​doi.​
on Wireless Communications, 20(5), 3049–3060. https://​doi.​org/​ org/​10.​1109/​CompC​omm.​2018.​87807​56
10.​1109/​TWC.​2020.​30471​00 103. Interdonato, G., Ngo, H. Q., Frenger, P., & Larsson, E. G. (2019).
88. Abeywickrama, S., Zhang, R., Wu, Q., & Yuen, C. (2020). Intelli- Downlink training in cell-free massive MIMO: A blessing in dis-
gent reflecting surface: Practical phase shift model and beamform- guise. IEEE Transactions on Wireless Communications, 18(11),
ing optimization. IEEE Transactions on Communications, 68(9), 5153–5169. https://​doi.​org/​10.​1109/​TWC.​2019.​29338​31
5849–5863. https://​doi.​org/​10.​1109/​TCOMM.​2020.​30011​25 104. Liu, H., Zhang, J., Jin, S., & Ai, B. (2020). Graph coloring
89. Wang, P., Fang, J., Duan, H., & Li, H. (2020). Compressed chan- based pilot assignment for cell-free massive MIMO systems.
nel estimation for intelligent reflecting surface-assisted millim- IEEE Transactions on Vehicular Technology, 69(8), 9180–9184.
eter wave systems. IEEE Signal Processing Letters, 27, 905–909. https://​doi.​org/​10.​1109/​TVT.​2020.​30004​96
https://​doi.​org/​10.​1109/​LSP.​2020.​29983​57 105. Song, H., You, X., Zhang, C., Tirkkonen, O., & Studer, C.
90. He, J., Leinonen, M., Wymeersch, H., & Juntti, M. (2020). Chan- (2020). Minimizing pilot overhead in cell-free massive MIMO
nel estimation for RIS-aided mmWave MIMO channels. In Pro- systems via joint estimation and detection. In 2020 IEEE 21st
ceeding of the IEEE Global Communications Conference (pp. International Workshop on Signal Processing Advances in Wire-
1–6). https://d​ oi.o​ rg/1​ 0.1​ 109/G
​ LOBEC
​ OM420​ 02.2​ 020.9​ 34811​ 2 less Communications (SPAWC). IEEE. https://​doi.​org/​10.​1109/​
91. He, J., Wymeersch, H., & Juntti, M. (2021). Channel estimation SPAWC​48557.​2020.​91543​10
for RIS-aided mmWave MIMO systems via atomic norm mini- 106. Sha, Z., & Wang, Z. (2021). Channel estimation and equalization
mization. IEEE Wireless Communications Letters, 1–1. https://​ for terahertz receiver with RF impairments. IEEE Journal on
doi.​org/​10.​1109/​TWC.​2021.​30700​64 Selected Areas in Communications, 39(6), 1621–1635. https://​
92. Ardah, K., Gherekhloo, S., de Almeida, A. L. F., & Haardt, M. doi.​org/​10.​1109/​JSAC.​2021.​30718​24
(2021). TRICE: An efficient channel estimation framework for 107. Liu, Y., Zhang, S., Mu, X., Ding, Z., Schober, R., Al-Dhahir, N.,
RIS-aided MIMO communications. IEEE Signal Processing Let- Hossain, E., & Shen, X. (2021). Evolution of NOMA toward
ters, 28, 513–517. https://​doi.​org/​10.​1109/​LSP.​2021.​30593​63 next generation multiple access (NGMA). Preprint retrieved
93. Schroeder, R., He, J., & Juntti, M. (2020). Passive RIS vs. from https://​arxiv.​org/​abs/​2108.​04561
hybrid RIS: A comparative study on channel estimation. Preprint 108. Huang, Y., Zhang, C., Wang, J., Jing, Y., Yang, L., & You,
retrieved from https://​arxiv.​org/​abs/​2010.​06981 X. (2018). Signal processing for MIMO-NOMA: Present and
94. Wei, X., Shen, D., & Dai, L. (2021). Channel estimation for RIS future challenges. IEEE Wireless Communications, 25(2), 32–38.
assisted wireless communications-part i: Fundamentals, solutions, https://​doi.​org/​10.​1109/​MWC.​2018.​17001​08
and future opportunities. IEEE Communications Letters, 25(5), 109. Hanif, M. F., Ding, Z., Ratnarajah, T., & Karagiannidis, G. K.
1398–1402. https://​doi.​org/​10.​1109/​LCOMM.​2021.​30528​22 (2015). A minorization-maximization method for optimizing
95. Ma, X., Chen, Z., Chen, W., Li, Z., Chi, Y., Han, C., & Li, S. sum rate in the downlink of non-orthogonal multiple access sys-
(2020). Joint channel estimation and data rate maximization for tems. IEEE Transactions on Signal Processing, 64(1), 76–88.
intelligent reflecting surface assisted terahertz mimo communi- https://​doi.​org/​10.​1109/​TSP.​2015.​24800​42
cation systems. IEEE Access, 8, 99565–99581. https://​doi.​org/​ 110. Sun, Q., Han, S., Chin-Lin, I., & Pan, Z. (2015). On the ergodic
10.​1109/​ACCESS.​2020.​29941​00 capacity of MIMO NOMA systems. IEEE Wireless Communications
96. He, Z. Q., & Yuan, X. (2019). Cascaded channel estimation for Letters, 4(4), 405–408. https://​doi.​org/​10.​1109/​LWC.​2015.​24267​09
large intelligent metasurface assisted massive MIMO. IEEE 111. Choi, J. (2016). On the power allocation for MIMO-NOMA sys-
Wireless Communications Letters, 9(2), 210–214. https://​doi.​ tems with layered transmissions. IEEE Transactions on Wireless
org/​10.​1109/​LWC.​2019.​29486​32 Communications, 15(5), 3226–3237. https://​doi.​org/​10.​1109/​
97. Liu, H., Yuan, X., & Zhang, Y. J. A. (2020). Matrix-calibration- TWC.​2016.​25181​82
based cascaded channel estimation for reconfigurable intelligent 112. Chen, Z., Ding, Z., Dai, X., & Karagiannidis, G. K. (2016). On
surface assisted multiuser mimo. IEEE Journal on Selected Areas the application of quasi-degradation to MISO-NOMA downlink.
in Communications, 38(11), 2621–2636. https://d​ oi.o​ rg/1​ 0.1​ 109/​ IEEE Transactions on Signal Processing, 64(23), 6174–6189.
JSAC.​2020.​30070​57 https://​doi.​org/​10.​1109/​TSP.​2016.​26039​71
98. Mishra, D., & Johansson, H. (2019). Channel estimation and low- 113. Liu, Y., Elkashlan, M., Ding, Z., & Karagiannidis, G. K. (2016).
complexity beamforming design for passive intelligent surface Fairness of user clustering in MIMO non-orthogonal multiple
assisted miso wireless energy transfer. In ICASSP 2019-2019 access systems. IEEE Communications Letters, 20(7), 1465–
IEEE International Conference on Acoustics, Speech and Signal 1468. https://​doi.​org/​10.​1109/​LCOMM.​2016.​25594​59
Processing (ICASSP) (pp. 4659–4663). IEEE. https://​doi.​org/​10.​ 114. Ding, Z., Schober, R., & Poor, H. V. (2016). A general MIMO
1109/​ICASSP.​2019.​86836​63 framework for NOMA downlink and uplink transmission based
99. Wei, L., Huang, C., Alexandropoulos, G. C., Yuen, C., Zhang, Z., on signal alignment. IEEE Transactions on Wireless Communi-
& Debbah, M. (2021). Channel estimation for RIS-empowered cations, 15(6), 4438–4454. https://​doi.​org/​10.​1109/T ​ WC.​2016.​
multi-user MISO wireless communications. IEEE Transactions 25420​66

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 455

115. Choi, J. (2015). Minimum power multicast beamforming with 130. Hou, T., Liu, Y., Song, Z., Sun, X., & Chen, Y. (2020). MIMO-
superposition coding for multiresolution broadcast and applica- NOMA networks relying on reconfigurable intelligent surface: A
tion to NOMA systems. IEEE Transactions on Communications, signal cancellation-based design. IEEE Transactions on Commu-
63(3), 791–800. https://d​ oi.o​ rg/1​ 0.1​ 109/T
​ COMM.2​ 015.2​ 39439​ 3 nications, 68(11), 6932–6944. https://d​ oi.o​ rg/1​ 0.1​ 109/T
​ COMM.​
116. Ali, S., Hossain, E., & Kim, D. I. (2016). Non-orthogonal multi- 2020.​30181​79
ple access (NOMA) for downlink multiuser MIMO systems: User 131. Gui, G., Huang, H., Song, Y., & Sari, H. (2018). Deep learning
clustering, beamforming, and power allocation. IEEE Access, 5, for an effective nonorthogonal multiple access scheme. IEEE
565–577. https://​doi.​org/​10.​1109/​ACCESS.​2016.​26461​83 Transactions on Vehicular Technology, 67(9), 8440–8450.
117. Cui, J., Ding, Z., & Fan, P. (2018). Outage probability con- https://​doi.​org/​10.​1109/​TVT.​2018.​28482​94
strained MIMO-NOMA designs under imperfect CSI. IEEE 132. Dejonghe, A., Antón-Haro, C., Mestre, X., Cardoso, L., & Goursaud,
Transactions on Wireless Communications, 17(12), 8239–8255. C. (2021). Deep learning-based user clustering for MIMO-NOMA
https://​doi.​org/​10.​1109/​TWC.​2018.​28754​90 networks. In 2021 IEEE Wireless Communications and Network-
118. Liu, L., Yuen, C., Guan, Y. L., Li, Y., & Huang, C. (2018). ing Conference (WCNC), (pp. 1–6). IEEE. https://​doi.​org/​10.​1109/​
Gaussian message passing for overloaded massive MIMO- WCNC4​9053.​2021.​94174​26
NOMA. IEEE Transactions on Wireless Communications, 18(1), 133. Huang, H., Yang, Y., Ding, Z., Wang, H., Sari, H., & Adachi, F.
210–226. https://​doi.​org/​10.​1109/​TWC.​2018.​28787​20 (2020). Deep learning-based sum data rate and energy efficiency
119. Rezaei, F., Heidarpour, A. R., Tellambura, C., & Tadaion, A. optimization for MIMO-NOMA systems. IEEE Transactions on
(2020). Underlaid spectrum sharing for cell-free massive MIMO- Wireless Communications, 19(8), 5373–5388. https://​doi.​org/​10.​
NOMA. IEEE Communications Letters, 24(4), 907–911. https://​ 1109/​TWC.​2020.​29927​86
doi.​org/​10.​1109/​LCOMM.​2020.​29661​95 134. Saetan, W., & Thipchaksurat, S. (2019). Power allocation for sum
120. Ma, J., Liang, C., Xu, C., & Ping, L. (2017). On orthogonal and rate maximization in 5G NOMA system with imperfect SIC: A
superimposed pilot schemes in massive MIMO NOMA systems. deep learning approach. In 2019 4th International Conference on
IEEE Journal on Selected Areas in Communications, 35(12), Information Technology (InCIT), (pp. 195–198). IEEE. https://​
2696–2707. https://​doi.​org/​10.​1109/​JSAC.​2017.​27260​19 doi.​org/​10.​1109/​INCIT.​2019.​89118​64
121. Hu, C., Wang, H., & Song, R. (2021). Group successive inter- 135. Pei, X., Chen, Y., Wen, M., Yu, H., Panayirci, E., & Poor, H. V.
ference cancellation assisted semi-blind channel estimation (2022). Next-generation multiple access based on NOMA with
in multi-cell massive MIMO-NOMA systems. IEEE Commu- power level modulation. IEEE Journal on Selected Areas in
nications Letters, 25(9), 3085–3089. https://​doi.​org/​10.​1109/​ Communications. https://​doi.​org/​10.​1109/​JSAC.​2022.​31432​40
LCOMM.​2021.​30951​19 136. Agiwal, M., Roy, A., & Saxena, N. (2016). Next generation 5G
122. de Sena, A. S., Lima, F. R. M., da Costa, D. B., Ding, Z., Nardelli, wireless networks: A comprehensive survey. IEEE Communi-
P. H., Dias, U. S., & Papadias, C. B. (2020). Massive MIMO- cations Surveys Tutorials, 18(3), 1617–1655. https://​doi.​org/​10.​
NOMA networks with imperfect SIC: Design and fairness 1109/​COMST.​2016.​25324​58
enhancement. IEEE Transactions on Wireless Communications, 137. Marabissi, D., Mucchi, L., Caputo, S., Nizzi, F., Pecorella, T.,
19(9), 6100–6115. https://​doi.​org/​10.​1109/​TWC.​2020.​30001​92 Fantacci, R., Nawaz, T., Seminara, M., & Catani, J. (2020).
123. Wang, B., Dai, L., Wang, Z., Ge, N., & Zhou, S. (2017). Spec- Experimental measurements of a joint 5g-VLC communication
trum and energy-efficient beamspace MIMO-NOMA for millim- for future vehicular networks. Journal of Sensor and Actuator
eter-wave communications using lens antenna array. IEEE Jour- Networks, 9(3), 32. https://​doi.​org/​10.​3390/​jsan9​030032
nal on Selected Areas in Communications, 35(10), 2370–2382. 138. Uysal, M., & Nouri, H. (2014) Optical wireless communications
https://​doi.​org/​10.​1109/​JSAC.​2017.​27258​78 – an emerging technology. In 2014 16th International Conference
124. Jiao, R., & Dai, L. (2020). On the max-min fairness of beamspace on Transparent Optical Networks (ICTON) (pp. 1–7). https://d​ oi.​
MIMO-NOMA. IEEE Transactions on Signal Processing, 68, org/​10.​1109/​ICTON.​2014.​68762​67
4919–4932. https://​doi.​org/​10.​1109/​TSP.​2020.​30185​59 139. Khalighi, M. A., & Uysal, M. (2014). Survey on free space opti-
125. Dai, L., Wang, B., Peng, M., & Chen, S. (2018). Hybrid precod- cal communication: A communication theory perspective. IEEE
ing-based millimeter-wave massive MIMO-NOMA with simul- Communications Surveys Tutorials, 16(4), 2231–2258. https://​
taneous wireless information and power transfer. IEEE Journal doi.​org/​10.​1109/​COMST.​2014.​23295​01
on Selected Areas in Communications, 37(1), 131–141. https://​ 140. Fawaz, W., Abou-Rjeily, C., & Assi, C. (2018). Uav-aided coop-
doi.​org/​10.​1109/​JSAC.​2018.​28723​64 eration for fso communication systems. IEEE Communications
126. Hu, X., Zhong, C., Chen, X., Xu, W., & Zhang, Z. (2019). Cluster Magazine, 56(1), 70–75. https://​doi.​org/​10.​1109/​MCOM.​2017.​
grouping and power control for angle-domain mmwave MIMO 17003​20
NOMA systems. IEEE Journal of Selected Topics in Signal Pro- 141. Pathak, P. H., Feng, X., Hu, P., & Mohapatra, P. (2015). Visible
cessing, 13(5), 1167–1180. https://​doi.​org/​10.​1109/​JSTSP.​2019.​ light communication, networking, and sensing: A survey, potential
29228​21 and challenges. IEEE Communications Surveys Tutorials, 17(4),
127. Zhang, H., Zhang, H., Liu, W., Long, K., Dong, J., & Leung, 2047–2077. https://​doi.​org/​10.​1109/​COMST.​2015.​24764​74
V. C. (2020). Energy efficient user clustering, hybrid precoding 142. Haas, H., Yin, L., Wang, Y., & Chen, C. (2016). What is lifi?
and power optimization in terahertz MIMO-NOMA systems. Journal of Lightwave Technology, 34(6), 1533–1544. https://d​ oi.​
IEEE Journal on Selected Areas in Communications, 38(9), org/​10.​1109/​JLT.​2015.​25100​21
2074–2085. https://​doi.​org/​10.​1109/​ICC40​277.​2020.​91491​88 143. Saud, M. S., & Katz, M. (2017). Implementation of a hybrid opti-
128. Mu, X., Liu, Y., Guo, L., Lin, J., & Al-Dhahir, N. (2020). cal-RF wireless network with fast network handover. In European
Exploiting intelligent reflecting surfaces in NOMA networks: Wireless 2017; 23th European Wireless Conference (pp. 1–6).
Joint beamforming optimization. IEEE Transactions on Wireless 144. Miramirkhani, F., Uysal, M., & Panayirci, E. (2015). Novel chan-
Communications, 19(10), 6884–6898. https://​doi.​org/​10.​1109/​ nel models for visible light communications. In Dingel, B. B., &
TWC.​2020.​30069​15 Tsukamoto, K. (Eds.), Broadband Access Communication Tech-
129. Yang, G., Xu, X., Liang, Y. C., & Di Renzo, M. (2021). Recon- nologies IX. International Society for Optics and Photonics (vol.
figurable intelligent surface-assisted non-orthogonal multiple 9387, pp. 150 – 162). SPIE. https://​doi.​org/​10.​1117/​12.​20775​65
access. IEEE Transactions on Wireless Communications, 20(5), 145. Uysal, M., Miramirkhani, F., Narmanlioglu, O., Baykas, T., &
3137–3151. https://​doi.​org/​10.​1109/​TWC.​2020.​30476​32 Panayirci, E. (2017). IEEE 802.15.7r1 reference channel models for

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

456 Journal of Signal Processing Systems (2023) 95:435–457

visible light communications. Communications Magazine, 55(1), Communications, 58(1), 95–109. https:// ​d oi. ​o rg/ ​1 0. ​1 007/​
212–217. https://​doi.​org/​10.​1109/​MCOM.​2017.​16008​72CM s11277-​011-​0290-y
146. Rajagopal, S., Roberts, R. D., & Lim, S. K. (2012). IEEE 162. Cogalan, T., Haas, H., & Panayirci, E. (2020). Optical spatial
802.15.7 visible light communication: modulation schemes modulation design. Philosophical Transactions of the Royal
and dimming support. IEEE Communications Magazine, 50(3), Society A: Mathematical, Physical and Engineering Sciences,
72–82. https://​doi.​org/​10.​1109/​MCOM.​2012.​61635​85 378(2169), 20190195. https://​doi.​org/​10.​1098/​rsta.​2019.​0195
147. Caputo, S., Biotti, L., & Mucchi, L. (2021). Bio-optical com- 163. Yesilkaya, A., Basar, E., Miramirkhani, F., Panayirci, E., Uysal, M.,
munication. In Proceedings of the Eight Annual ACM Interna- & Haas, H. (2017). Optical MIMO-OFDM with generalized LED
tional Conference on Nanoscale Computing and Communica- index modulation. IEEE Transactions on Communications, 65(8),
tion. ACM. https://​doi.​org/​10.​1145/​34772​06.​34774​70 3429–3441. https://​doi.​org/​10.​1109/​TCOMM.​2017.​26999​64
148. Ahmed, I., Bykov, A., Popov, A., Meglinski, I., & Katz, M. 164. Panayirci, E., Yesilkaya, A., Cogalan, T., Poor, H. V., & Haas,
(2019). In Lecture Notes of the Institute for Computer Sci- H. (2020). Physical-layer security with optical generalized space
ences, Social Informatics and Telecommunications Engineer- shift keying. IEEE Transactions on Communications, 68(5),
ing. Springer International Publishing (pp. 191–205). https://​ 3042–3056. https://​doi.​org/​10.​1109/​TCOMM.​2020.​29698​67
doi.​org/​10.​1007/​978-3-​030-​34833-5_​16 165. Chen, Y., Wang, L., Zhao, Z., Ma, M., & Jiao, B. (2016). Secure
149. Ahmed, I., Bykov, A., Popov, A., Meglinski, I., & Katz, M. multiuser MIMO downlink transmission via precoding-aided
(2020). Wireless data transfer through biological tissues using near- spatial modulation. IEEE Communications Letters, 20(6), 1116–
infrared light: Testing skull and skin phantoms. In Neural Imaging 1119. https://​doi.​org/​10.​1109/​LCOMM.​2016.​25490​14
and Sensing 2020. SPIE. https://​doi.​org/​10.​1117/​12.​25452​21 166. O’Shea, T., & Hoydis, J. (2017). An introduction to deep learning
150. Chowdhury, M. Z., Shahjalal, M., Hasan, M. K., & Jang, Y. M. for the physical layer. IEEE Transactions on Cognitive Communi-
(2019). The role of optical wireless communication technolo- cations and Networking, 3(4), 563–575. https://​doi.​org/​10.​1109/​
gies in 5G/6G and IoT solutions: Prospects, directions, and chal- TCCN.​2017.​27583​70
lenges. Applied Sciences, 9(20), 4367. https://​doi.​org/​10.​3390/​ 167. Pin Tan, D. K., He, J., Li, Y., Bayesteh, A., Chen, Y., Zhu, P.,
app92​04367 & Tong, W. (2021). Integrated sensing and communication in
151. Chen, S., Ma, R., Chen, H. H., Zhang, H., Meng, W., & Liu, J. 6G: Motivations, use cases, requirements, challenges and future
(2017). Machine-to-machine communications in ultra-dense net- directions. In 2021 1st IEEE International Online Symposium on
works-a survey. IEEE Communications Surveys Tutorials, 19(3), Joint Communications Sensing (JC S) (pp. 1–6). https://​doi.​org/​
1478–1503. https://​doi.​org/​10.​1109/​COMST.​2017.​26785​18 10.​1109/​JCS52​304.​2021.​93763​24
152. Cossu, G., Ali, W., Corsini, R., & Ciaramella, E. (2015). Gigabit- 168. De Lima, C., Belot, D., Berkvens, R., Bourdoux, A., Dardari, D.,
class optical wireless communication system at indoor distances Guillaud, M., Isomursu, M., Lohan, E. S., Miao, Y., Barreto, A.
(15–4 m). Optics Express, 23(12), 15700. https://​doi.​org/​10.​ N., Aziz, M. R. K., Saloranta, J., Sanguanpuak, T., Sarieddeen,
1364/​oe.​23.​015700 H., Seco-Granados, G., Suutala, J., Svensson, T., Valkama, M.,
153. Gomez, A., Shi, K., Quintana, C., Sato, M., Faulkner, G., Thomsen, Van Liempd, B., & Wymeersch, H. (2021). Convergent commu-
B. C., & O’Brien, D. (2015). Beyond 100-Gb/s indoor wide field-of- nication, sensing and localization in 6G systems: An overview
view optical wireless communications. IEEE Photonics Technology of technologies, opportunities and challenges. IEEE Access, 9,
Letters, 27(4), 367–370. https://​doi.​org/​10.​1109/​LPT.​2014.​23749​95 26902–26925. https://​doi.​org/​10.​1109/​ACCESS.​2021.​30534​86
154. Cǎilean, A. M., & Dimian, M. (2017). Current challenges for vis- 169. Wymeersch, H., Shrestha, D., de Lima, C. M., Yajnanarayana,
ible light communications usage in vehicle applications: A sur- V., Richerzhagen, B., Keskin, M. F., Schindhelm, K., Ramirez,
vey. IEEE Communications Surveys Tutorials, 19(4), 2681–2703. A., Wolfgang, A., de Guzman, M. F., Haneda, K., Svensson, T.,
https://​doi.​org/​10.​1109/​COMST.​2017.​27069​40 Baldemair, R., & Parkvall, S. (2021). Integration of communica-
155. Kaushal, H., & Kaddoum, G. (2017). Optical communication in tion and sensing in 6G: a joint industrial and academic perspec-
space: Challenges and mitigation techniques. IEEE Communica- tive. In 2021 IEEE 32nd Annual International Symposium on Per-
tions Surveys Tutorials, 19(1), 57–96. https://​doi.​org/​10.​1109/​ sonal, Indoor and Mobile Radio Communications (PIMRC) (pp.
COMST.​2016.​26035​18 1–7). https://​doi.​org/​10.​1109/​PIMRC​50174.​2021.​95693​64
156. Hughes, R. J., & Nordholt, J. E. (2017). Quantum space race 170. Re, E. D., Morosi, S., Mucchi, L., Ronga, L. S., & Jayousi, S.
heats up. Nature Photonics, 11(8), 456–458. https://​doi.​org/​10.​ (2016). Future wireless systems for human bond communica-
1038/​nphot​on.​2017.​124 tions. Wireless Personal Communications, 88(1), 39–52. https://​
157. Liu, Y., Chen, H. H., & Wang, L. (2017). Physical layer security doi.​org/​10.​1007/​s11277-​016-​3240-x
for next generation wireless networks: Theories, technologies, 171. Wang, G., Ho, K. C., & Chen, X. (2021). Bias reduced sem-
and challenges. IEEE Communications Surveys Tutorials, 19(1), idefinite relaxation method for 3-D rigid body localization using
347–376. https://​doi.​org/​10.​1109/​COMST.​2016.​25989​68 AOA. IEEE Transactions on Signal Processing, 69, 3415–3430.
158. Wu, F., Zhang, R., Yang, L. L., & Wang, W. (2016). Transmitter https://​doi.​org/​10.​1109/​TSP.​2021.​30863​60
precoding-aided spatial modulation for secrecy communications. 172. Wang, B., Xu, Q., Chen, C., Zhang, F., & Liu, K. R. (2018). The
IEEE Transactions on Vehicular Technology, 65(1), 467–471. promise of radio analytics: A future paradigm of wireless posi-
https://​doi.​org/​10.​1109/​TVT.​2015.​23954​57 tioning, tracking, and sensing. IEEE Signal Processing Maga-
159. Arafa, A., & Panayirci, E., & Poor, H. V. (2018). Relay-aided zine, 35(3), 59–80. https://​doi.​org/​10.​1109/​MSP.​2018.​28063​00
secure broadcasting for VLC. In 2018 IEEE Global Conference 173. Akyildiz, I. F., Pierobon, M., Balasubramaniam, S., & Kouch-
on Signal and Information Processing (GlobalSIP) (pp. 1286– eryavy, Y. (2015). The internet of bio-nano things. IEEE Com-
1290). https://​doi.​org/​10.​1109/​Globa​lSIP.​2018.​86465​09 munications Magazine, 53(3), 32–40. https://​doi.​org/​10.​1109/​
160. Mucchi, L., Ronga, L. S., & Re, E. D. (2010). A novel approach MCOM.​2015.​70605​16
for physical layer cryptography in wireless networks. Wireless 174. Mucchi, L., Jayousi, S., Caputo, S., Panayirci, E., Shahabuddin,
Personal Communications, 53(3), 329–347. https://​doi.​org/​10.​ S., Bechtold, J., Morales, I., Stoica, R. A., Abreu, G., & Haas,
1007/​s11277-​010-​9950-6 H. (2021). Physical-layer security in 6G networks. IEEE Open
161. Mucchi, L., Ronga, L. S., & Re, E. D. (2011). Physical layer Journal of the Communications Society, 2, 1901–1914. https://​
cryptography and cognitive networks. Wireless Personal doi.​org/​10.​1109/​OJCOMS.​2021.​31037​35

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Journal of Signal Processing Systems (2023) 95:435–457 457

175. Mucchi, L., Martinelli, A., Jayousi, S., Caputo, S., & Pierobon, Publisher’s Note Springer Nature remains neutral with regard to
M. (2019). Secrecy capacity and secure distance for diffusion- jurisdictional claims in published maps and institutional affiliations.
based molecular communication systems. IEEE Access, 7,
110687–110697. https://​doi.​org/​10.​1109/​access.​2019.​29325​67

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

[email protected]