Received 22 May 2024; accepted 21 June 2024. Date of publication 3 July 2024; date of current version 1 August 2024.

Digital Object Identifier 10.1109/OJCOMS.2024.3422801

A Survey on Orthogonal Time Frequency

Space Modulation
MAHMOUD ALDABABSA 1 (Senior Member, IEEE), SERDAR ÖZYURT 2 ,
GÜNEŞ KARABULUT KURT 3 (Senior Member, IEEE), AND OĞUZ KUCUR4
1 Department of Electrical and Electronics Engineering, Nisantasi University, 34481742 Istanbul, Türkiye

2 Department of Electrical and Electronics Engineering, Ankara Yıldırım Beyazıt University, 06010 Ankara, Türkiye

3 Department of Electrical Engineering, Polytechnique Montréal, Montreal, QC H3T 1J4, Canada

4 Department of Electronics Engineering, Gebze Technical University, 41400 Gebze, Türkiye

CORRESPONDING AUTHOR: G. KARABULUT KURT (e-mail: [email protected])

This work was supported in part by the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Program.

ABSTRACT Orthogonal time frequency space (OTFS) modulation is strongly considered as a promising
solution for high-mobility communications. In contrast to conventional modulation techniques, wherein
information symbols are multiplexed in a one-dimensional time or frequency domain, OTFS employs a
two-dimensional modulation scheme by multiplexing information symbols in the delay-Doppler domain.
This paper presents a comprehensive survey of OTFS. It starts with an overview of OTFS, its advantages
over conventional air interface techniques, general block diagrams, and implementations. Subsequently,
the paper explores the potential integration of multiple-input multiple-output and OTFS techniques. The
paper further discusses the feasibility of integrating OTFS into multiple access techniques as a solution
for maintaining acceptable performance in high-mobility scenarios. Then, widespread applications of
OTFS in satellite communications are highlighted. Also, the potential utilization of OTFS modulation
in integrated sensing and communications paradigm is thoroughly treated. In addition, the survey covers
further applications of OTFS in deep learning, index modulation, underwater acoustic, and unmanned
aerial vehicle communications. The paper concludes by pointing out numerous challenging and promising
directions for future OTFS research.

INDEX TERMS Orthogonal time frequency space (OTFS), delay-Doppler domain, multiple-input multiple-
output, multiple access techniques, satellite communications, integrated sensing and communications, deep
learning, index modulation, underwater acoustic, unmanned aerial vehicle.

I. INTRODUCTION would have extraordinary requirements to support the afore-

I N THE next few years, sixth-generation (6G) will bring

a substantial revolution in communication technologies.
It will enable the Internet of Everything (IoE) and pro-
mentioned great applications. Particularly, 6G has been
expected to offer a peak data rate of at least one terabit per
second (≥ 1 Tbps), a user-experienced data rate of more than
vide significant development in performance superior to one gigabit per second (≥ 1 Gbps), a connectivity density
the emerging 5G [1], [2], [3]. The 6G has been envi- to reach millions of devices/users per km2 , a reduction in
sioned to support different usage scenarios, summarized communication latency to small fractions of milliseconds
and depicted in Fig. 1, such as further enhanced mobile (0.01-0.1 ms), an increase in the mobility speed to a
broadband (FeMBB), ultra-massive machine-type communi- minimum of 1000 kilometers per hour (≥ 1000 kmph),
cations (umMTC), enhanced ultra-reliable and low-latency a network energy efficiency of 10-100 times, a spectrum
communications (eURLLC), long-distance and high-mobility efficiency of 5-10 times, and reliability of 10 times those of
communications (LDHMC), and extremely low power com- 5G [1]. Notably, these requirements are critical challenges
munications (ELPC) [1]. It has been anticipated that 6G of their own. However, the most challenging 6G requirement


c 2024 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
VOLUME 5, 2024 4483
ALDABABSA et al.: SURVEY ON OTFS MODULATION

FIGURE 1. Potential usage scenarios for 6G (inspired by Fig. 1 of [1]).

is driven by the usage scenarios, in which it is expected to means that the Doppler impact is inevitable in the 6G, and
provide reliable communications in high-mobility environ- if it is not taken into account for system design, the 6G
ments. will suffer from significant performance degradation over
Wireless communications operating at the high carrier high-mobility channels. Fortunately, an increasing amount
frequency (fc ) in high-mobility environments experience of research has recently been dedicated to high-mobility
fast time-variation of fading channels caused by the large communications of next-generation wireless networks. To
Doppler spread1 [4]. This will significantly deteriorate the cope with the high-mobility problem, the researchers think
performance of the system. Generally speaking, the wireless more deeply about the radio resource element, which carries
channel is doubly dispersive and has many paths. For each information signals from source to destination over the
path, there are associated delay and Doppler values. Thus, air. They have raised a crucial question, do the existing
the wireless channel can be easily depicted as a plurality air interface techniques and associated waveforms support
of reflectors in the two-dimensional delay-Doppler (DD) diverse challenging requirements and usage scenarios of the
domain. The wireless channels are categorized as linear 6G? or is it time to introduce a new air interface?
time-invariant (LTI) channels, in which Doppler values are The air interface has undergone notable changes over
considered zeros, and linear frequency-invariant channels, the generations of mobile communication systems. It was
where delay values are assumed to be zero. The system frequency division multiple access (FDMA), time division
performance will significantly degrade once both delay and multiple access (TDMA), and code division multiple access
Doppler effects are present. It means that a waveform that is (CDMA) in the 1G, 2G, and 3G networks, respectively.
DD invariant is highly desired. Nevertheless, it can be noted In 4G and 5G networks, respectively, the air interface was
that the spectrum of the 6G will be in the multi-THz range, orthogonal frequency division multiple access (OFDMA),
and its mobility will be as high as thousands of kmph. This which exploited the orthogonality principle, and is the mod-
ified version of orthogonal frequency division multiplexing
1 Doppler spread is defined as the difference in shift frequencies associated (OFDM), where the subcarrier bandwidth and cyclic prefix
with the multiple propagation paths when there is relative motion between (CP) interval are made flexible. Unfortunately, the OFDM
the transmitter and the receiver. used in the emerging 5G networks will not satisfy the 6G

4484 VOLUME 5, 2024

requirements with high system performance. Additionally, of time, the CFO is time-varying, which means that the
it will not be a well-suited solution for offering reliable OFDM finds it challenging to track and compensate
communications in high-mobility environments. The reasons the CFO. However, due to Doppler-resilience benefits,
behind that can be summarized as follows: the OTFS is more robust against the CFO than the
• In the OFDM, the channel is assumed to be quasi-static OFDM.
frequency-selective fading, which means that the chan- 3) Low idle time: The guard intervals are merely needed
nel remains constant over one OFDM symbol period. between consecutive OTFS frames rather than between
Nevertheless, the channel will change inside one OFDM time slots. Thus, the associated idle time is signifi-
symbol in high-mobility environments. Thus, the fading cantly reduced.
time-variation will destroy the orthogonality among sub- 4) Low number of pilot symbols for channel estimation
carriers and introduce inter-carrier interference (ICI) due training overhead: In high-mobility environments, the
to carrier frequency offset (CFO). This will significantly channel envelope fluctuates violently, even for a short
deteriorate the system’s performance. time. However, the DD response of wireless channels
• In the OFDM, a significant channel estimation overhead is much more sparse than the TF response, and the rate
is necessary due to the limited coherence time of of change of the channel will be slower. In addition,
the time-frequency (TF) domain channel. For example, multiplexing information in the DD-domain turns a
consider an OFDM system with a carrier frequency of time-varying channel into a quasi-static channel in
fc = 3.5 GHz, a subcarrier spacing of f = 15 KHz, the DD-domain. This might significantly reduce the
and a relative velocity of v = 300 kmph. OFDM symbol amount of signaling overhead required for channel
duration (including a 20% CP) is 80 ms. Accordingly, estimation. Hence, for channel estimation, the OFDM
the maximum Doppler shift is νmax ≈ 0.97 KHz, and the needs to send more times of pilots than the OTFS.
channel’s coherence time is around 257 ms. Therefore, 5) Compatibility with the existing communication
one channel coherence interval can only accommodate systems: The OTFS can be implemented as a pre-
at most three OFDM symbols, which means a pilot and post-processing block to the OFDM systems.
symbol is inserted every three OFDM symbols, leading This enables the OTFS to have architectural back-
to a large overhead [8]. compatibility with existing communication systems.
6) Full diversity and reliability: In contrast to the
Recently, orthogonal time frequency space (OTFS) mod- existing one-dimensional time-domain or frequency-
ulation has been suggested as a promising candidate domain modulation techniques, the OTFS is a
for high-mobility communications [5], [6]. Compared to two-dimensional modulation technique. The data sym-
the conventional modulation techniques, in which the bols are multiplexed in the DD-domain, and each
information symbols are multiplexed in a one-dimensional symbol is spread across the entire TF-domain. This
time-domain or frequency-domain, the OTFS is a leads to attaining maximum achievable diversity or
two-dimensional modulation scheme, where the information transmission over doubly dispersive channels, provided
symbols are multiplexed in the DD-domain. With special that each time-domain and frequency-domain sample
features, the OTFS has been recognized as a strong candidate experiences independent fading. Note that the number
among all modulation interface techniques to overcome of independently fading resolvable paths in the DD-
high-mobility communications challenges and to achieve the domain determines the maximum attainable diversity
requirements of the next mobile communication systems. The order. Therefore, the OTFS has the potential to offer
essential benefits of the OTFS can be remarked as follows. full diversity, which is the key to support reliable
1) Delay- and Doppler-resilience: The OTFS can radi- communications.
cally transform a random fading channel within the 7) Useful choice for different communications systems:
TF frame into a stationary, deterministic, and non-
fading channel. All information symbols in one OTFS • Massive multiple-input multiple-output (MIMO)
frame experience the same static channel response. systems: As known, multiantenna systems suffer
As a result, the OTFS can easily track time-varying from the problem of channel estimation, which
fading, especially in high-speed mobile communica- results in significant performance degradation.
tions, which means that OTFS is resilient to both delay However, since the OTFS operates in the DD-
and Doppler shifts. domain, where the channel can be characterized in
2) Robust against CFO: In high-mobility environments, a very compact form, the OTFS can enable dense
the Doppler shift leads to CFO, which means that and flexible packing of reference signals. This
the oscillators’ frequencies at the source and desti- means that the OTFS can provide accurate channel
nation are mismatched. The CFO will destroy the estimation. Therefore, the OTFS can support large
orthogonality among the subcarriers and introduce ICI antenna arrays in massive MIMO applications.
in multi-carrier systems such as OFDM. Since the • Millimeter-Wave (mm-Wave) systems: Regarding
Doppler shifts in high-mobility systems are a function that mm-Wave systems utilize high carrier

VOLUME 5, 2024 4485

ALDABABSA et al.: SURVEY ON OTFS MODULATION

frequencies, Doppler effects will be significantly recent advancements, and future directions. The authors
amplified. However, the OTFS can easily handle in [13] provided a comprehensive summary and analysis
the Doppler effect. This means the OTFS can be of OTFS techniques and also illustrated the advantages of
a beneficial choice for mm-Wave systems. OTFS modulation from the level of multicarrier modulation
Although the advantages mentioned above render the comparison.
OTFS eminently suitable for high-mobility and high-carrier Differing from the aforementioned survey papers, this
scenarios, the OTFS has different disadvantages, which can paper provides a more detailed guide for the research
be summarized as follows. on OTFS. The main contributions of this paper can be
highlighted as follows:
1) Latency: The OTFS modulation spreads information
over a more extended block than the OFDM to combat • provides an overview of OTFS, its advantages over
channel fluctuations. This can increase the overall conventional air interface techniques, general block
latency. In particular, each OTFS signal’s duration diagrams, and implementations. Then, it reviews the
is N times that of an OFDM symbol, with the literature on standard OTFS in terms of low-complexity
same number of subcarriers and the same subcarrier signal detection, channel estimation, peak-to-average
spacing. However, by increasing the subcarrier spacing power ratio (PAPR), system design, and performance
or reducing N at the expense of a reduced Doppler analysis.
resolution, the OTFS’s latency can be identical to that • explores the potential integration of MIMO and OTFS
of the OFDM. techniques. It starts by demonstrating the MIMO-OTFS
2) Fractional Doppler: In the case of fractional Doppler, transceiver block diagram. Then, it highlights the related
the DD-domain requires particular channel estimation literature on MIMO-OTFS.
and detection algorithms. Otherwise, the performance • discusses the feasibility of integrating OTFS into
may significantly degrade. multiple access techniques as a solution for maintaining
3) Fractional delay: Fractional delay can cause symbols acceptable performance in high-mobility scenarios. It
to overlap in time, leading to inter symbol interference focuses on related works that are interested in the
(ISI). Moreover, fractional delay complicates chan- combination of OTFS with orthogonal multiple access
nel estimation and equalization processes in OTFS (OMA) and non-orthogonal multiple access (NOMA)
systems. It introduces additional phase shifts and techniques.
distortions, making it challenging to estimate the • shows widespread applications of OTFS in satellite
channel response and apply appropriate equaliza- communications.
tion techniques accurately. As a result, the receiver • considers the potential of OTFS modulation in ISAC
may struggle to recover the transmitted symbols transmissions.
correctly, leading to increased error rates. Overall, • presents further applications of OTFS in deep learning
fractional delay can degrade the performance of OTFS (DL), index modulation (IM), underwater acoustic
systems in terms of data rate, error rate, and spectral (UWA), and unmanned aerial vehicle (UAV) communi-
efficiency. cations.
• The paper concludes by pointing out numerous chal-
In [7], the authors introduced the concept of OTFS,
lenging and promising directions for future OTFS
including the key characteristics of the DD-domain channel,
research. For example, when the DD domain channel
DD-domain multiplexing, and OTFS transceiver archi-
parameters are time-variant within a frame duration,
tecture. They then discussed the significant issues of
the adaptation of OTFS to such a scenario poses a
OTFS, such as channel estimation, efficient data detec-
challenging future research direction. Moreover, with
tion, and coding/decoding, and presented relevant initial
the new perspectives OTFS offer, there is likely an
results. Finally, they discussed the potential applications
increasing interest in OTFS from a security point of
of OTFS and proposed several promising research direc-
view. Furthermore, a comprehensive performance eval-
tions. In [8], [9], [10], the authors created a three-part
uation of OTFS-NOMA for any number of users under
series of tutorials. The first paper [8] focused on two
practical impairments over generalized fading channels
implementations of OTFS modulation. The second [9]
constitutes a potential future research topic. In addition,
provided a comprehensive overview of OTFS transceiver
the exploitation of machine learning approaches in
design. Lastly, [10] discussed OTFS-based integrated sensing
OTFS systems will attract intense interest when they
and communications (ISAC), which is expected to be a
are utilized in different scenarios with distinct goals.
powerful technology. In [11], the authors introduced an
Also, the synergy between OTFS and ISAC is expected
overview of the OTFS modulation for the Internet of Things
to be explored further in the future.
(IoT). They discussed the OTFS’s transceiver design, its
advantages, challenges, and future design principles. In [12], The rest of this paper is organized as follows: Section II
the authors presented an in-depth review of OTFS technology introduces OTFS, including its general block diagram, imple-
in the context of the 6G era, encompassing fundamentals, mentations, and comparison with OFDM. Section III presents

4486 VOLUME 5, 2024

TABLE 1. List of symbols frequently used in the text. TABLE 2. List of abbreviations frequently used in the text.

and reviews the combination of MIMO and OTFS techniques

in the relevant literature. Section IV introduces the poten-
tial of integrating OTFS into multiple access techniques,
such as NOMA, as a solution for maintaining acceptable
performance in high-mobility scenarios. Section V introduces
OTFS as a promising waveform technology for satellite
communications and reviews related literature on the topic.
Sections VI and VII present the application of OTFS-ISAC
and other application scenarios, respectively. Challenges and
future perspectives in OTFS are given in Section VIII.
Finally, the paper is concluded in Section IX.
Kindly note that the lists of symbols and abbreviations
frequently used in the text are given, respectively, in
Tables 1 and 2.

II. OTFS FUNDAMENTALS

In this section, we begin by providing an overview of
the wireless channel models. We then introduce OTFS,
its general block diagram, its implementations and its
comparison with OFDM. Lastly, we present a review of the
literature on OTFS modulation.

A. WIRELESS CHANNEL
In general, the wireless channel is linear time-varying (LTV).
When a signal goes through a typical wireless channel, it
gets reflected and refracted from multiple scatterers. This
multipath results in time-dispersion of the signal. In addition, consider three scenarios, as depicted in Fig. 2, that clearly
these refractors and transceivers can be mobile, resulting illustrate time-dispersive, frequency-dispersive, and doubly-
in frequency-dispersion of the signal. In the following, we dispersive channel models.

VOLUME 5, 2024 4487

ALDABABSA et al.: SURVEY ON OTFS MODULATION

effects, as shown in Fig. 2 (c). Here, the LTV channels are

doubly-dispersive due to the joint presence of multipath prop-
agation and Doppler effects. The transmitted signals suffer
from dispersions both in the time-domain and frequency-
domain. Hence, two-dimensional CIR in the time-delay
domain h(τ, t) is required for characterizing the time-
dispersive and frequency-dispersive channels. The impulse
response of the LTV channels can be written as

P
h(τ, t) = hi (t)δ(τ − τi (t)), (1)
i=1
where P is the number of paths, and hi (t) and τi (t) are the
complex channel coefficient and propagation delay of the
i-th path, respectively.
The LTV channels can be equivalently depicted in either
the TF- or DD-domain. The TF-domain channel, hTF (f , t),
can be obtained by taking the Fourier transform of h(τ, t)
with respect to (w.r.t.) delay τ . The TF-domain channel can
be written as
 ∞
hTF (f , t) = h(τ, t)e−j2π f τ dτ
−∞
 ∞ 
P
= hi (t)δ(τ − τi (t))e−j2π f τ dτ
−∞ i=1

FIGURE 2. Illustration of wireless channel (a) Time-dispersive 

P
(b) Frequency-dispersive (c) Doubly-dispersive. = hi (t)e−j2π f τi (t) . (2)
i=1
We can observe from (2) that hTF (f , t) is a complex channel
Consider the first scenario, where the transmitter and transfer function coefficient at frequency f and time instant
receiver are stationary, as shown in Fig. 2 (a). The difference t. On the other hand, the DD-domain channel, hDD (τ, ν) can
in the propagation delay of the direct and reflected paths be obtained by taking the Fourier transform of h(τ, t) w.r.t.
causes multiple copies of the transmitted signal to arrive at t. The DD-domain channel can be expressed as
the receiver at different times. In this case, an LTI system  ∞
can model the wireless channel. Hence, one-dimensional hDD (τ, ν) = h(τ, t)e−j2π ft dt
channel impulse response (CIR) in the delay-domain h(τ ) −∞
is sufficient for characterizing the time-dispersive channel.  ∞ 
P
The frequency response of this CIR is a frequency-selective = hi (t)δ(τ − τi (t))e−j2π ft dt. (3)
−∞ i=1
channel transfer function. The frequency selectivity becomes
more severe when the delay spread2 increases significantly With some mathematical calculations, the complex channel
since the frequency-domain fades’ separation is increasingly response in the DD-domain can be stated as
proportional to the CIR length.

P
Now consider the second scenario, where the receiver hDD (τ, ν) = hi δ(τ − τi )δ(ν − νi ), (4)
is a car moving towards the transmitter with a relative i=1
velocity v, as shown in Fig. 2 (b). The Doppler shift of
where τi ∈ [0, τmax ] and νi ∈ [−νmax , νmax ] denote the delay
the line-of-sight (LoS) due to the relative velocity is given
and Doppler shifts of the i-th path, respectively. Here, τmax
by ν = vfcc , where c denotes the light speed. In this
and νmax denote the maximum delay and Doppler shifts
case, an LTV system can model the wireless channel. The
in the considered high-mobility propagation environment,
LTV channels give rise to frequency shifts due to the
respectively. We can notice from (4) that the DD-domain
Doppler effect, yielding frequency-dispersive (time-selective)
channel hDD (τ, ν) characterizes the intensity of scatterers
channels. Here, the separation of the channels’ time-domain
having a propagation delay of τ and Doppler frequency shift
fades is increasingly proportional to the Doppler spread.
of ν.
Finally, consider the last scenario (high-mobility scenario),
Based on the TF-domain and DD-domain, we have the
which includes the multipath propagation and Doppler
following observations:
2 The delay spread is defined as the difference between the propagation 1) With the limited coherence time and bandwidth of
delays of the longest and shortest paths, i.e., τmax − τmin . LTV channels in high-mobility scenarios, channel

4488 VOLUME 5, 2024

 acquisition in the TF-domain can be challenging,
which may result in significant signaling overhead.
On the other hand, the DD-domain channel responses
fluctuate much slower than time-delay-domain or
TF-domain channel responses since the DD-domain
channel response changes only in the cases of remark-
able changes in propagation path length or very high
moving speed. This will reduce the signaling overhead
and give an advantage to DD-domain over TF-domain
FIGURE 3. OTFS transceiver block diagram.
to represent LTV channels in high-mobility scenarios.
2) The LTV channel in the DD-domain has valuable
features that significantly facilitate channel estimation
and data detection. These features can be highlighted
below.
• Separability: The DD-domain of wireless channels
can separate the propagation paths with the same
delay, thereby revealing the total number of
channel degrees-of-freedom (DoF) available.
• Stability: The DD-domain channel response is
much more stable than that of time-delay-domain
or TF-domain channel responses, as it only
changes significantly when there is a significant
difference in the propagation path length or when
the speed of the moving object is very high.
• Compactness: Generally, the maximum delay and
Doppler values of a wireless channel are such
that τmax νmax ≤ 14 . Accordingly, hDD (τ, ν) has a
compact DD-domain support within the intervals
[0, τmax ] and [−νmax , νmax ] along the delay and
Doppler dimensions, respectively.
• Potential Sparsity: When there are few moving
FIGURE 4. SFFT-based OTFS modulation and demodulation block diagrams.
scatterers in the propagation environment, the DD-
domain channel will have a sparse response.

The OTFS modulation/demodulation can be implemented

B. OTFS AND ITS IMPLEMENTATIONS commonly via two typical implementations named symplec-
We assume an OTFS frame with a bandwidth of BOTFS tic finite Fourier transform (SFFT)-based OTFS [8], and Zak
and a time duration of TOTFS , defined in the TF-domain. transform (ZT) (discrete ZT (DZT) [8] and continuous ZT
The bandwidth is divided into L subcarriers with subcarrier (CZT) [14], [15], [16])-based OTFS. In general, the SFFT-
spacing f = BOTFS L and time duration is divided into K time based OTFS modulation maps the DD-domain information
slots with slot duration T = TOTFSK . As illustrated in Fig. 3,
symbols first to the TF-domain and then to the time-
at the transmitter side of the OTFS, we assume that a set domain. On the other hand, the ZT-based OTFS modulation
of LK information symbols {XDD [l, k]} are mapped into the directly transforms the DD-domain information symbols
two-dimensional DD-domain from the modulation alphabet into the time-domain. In the following, we will review
to be transmitted over the OTFS time duration TOTFS and both implementations in more detail and finally show their
OTFS bandwidth BOTFS . Here, l ∈ {0, 1, . . . , L − 1} and similarities and differences.
k ∈ {0, 1, . . . , K − 1} are the delay and Doppler indices,
respectively. In addition, we define s(t) as the time-domain 1) SFFT-BASED OTFS
transmit signal of the OTFS. On the other hand, we define
r(t) and YDD [l, k] as the received signal at the OTFS receiver As shown in Fig. 4, the SFFT-based OTFS modulation
in the time-domain and DD-domain, respectively. and demodulation are produced by a cascade of a pair
We notice from Fig. 3 that the key function of the OTFS of two-dimensional transforms at both the transmitter and
modulation is to transform the DD-domain signals, XDD [l, k], receiver. The following steps can summarize this:
into a time-domain signal, s(t), for transmission. On the • At the transmitter, the modulator first utilizes the ISFFT
other hand, the OTFS demodulation transforms the received to map the information symbols in the DD-domain to
time-domain signal, r(t), into a DD-domain signal YDD [l, k]. samples in the TF-domain.

VOLUME 5, 2024 4489

ALDABABSA et al.: SURVEY ON OTFS MODULATION

• Next, the modulator applies the Heisenberg transform Then, the signal s(t) will be transmitted through LTV
to create the time-domain signal to be transmitted over wireless channel. The time-domain received signal can be
the wireless channel. expressed as
• At the receiver, the demodulator uses Wigner transform  ∞ ∞
(the inverse of the Heisenberg transform) to map the r(t) = hDD (τ, ν)ej2π ν(t−τ ) s(t − τ )dτ dν + z(t),
−∞ −∞
received time-domain signal to the TF-domain.
(8)
• Finally, the demodulator exploits the SFFT to transform
the TF-domain signal to the DD-domain signal. where z(t) denotes the additive white Gaussian noise
Now, let us explain each part in SFFT-based OTFS modu- (AWGN) in the time-domain. Substituting (4) into (8) and
lation and demodulation deeply. using
By using the ISFFT, the DD-domain symbol XDD [l, k] is  ∞ ∞
transformed into TF-domain signal XTF [m, n], which can be s(t − τ )δ(τ − τi )δ(ν − νi )ej2π ν(t−τ ) dτ dν
−∞ −∞
written as
  = s(t − τi )ej2π νi (t−τi ) , (9)
1 
L−1 K−1
j2π Kk m− Ll n
XTF [m, n] = √ XDD [l, k]e , (5) Then
LK l=0 k=0

P
where m ∈ {0, 1, . . . , K − 1} and n ∈ {0, 1, . . . , L − 1} are r(t) = hi s(t − τi )ej2π νi (t−τi ) + z(t). (10)
the time slot and subcarrier indices, respectively. Note that i=1
XTF [m, n] in (5) can be rewritten as At the OTFS receiver, Wigner transform is performed to
1 
K−1   j2π k map the received time-domain signal r(t) to the TF-domain.
XTF [m, n] = √ XDD l̃, k e K m The TF-domain received signal for the mth time slot and nth
K k=0 subcarrier can be written as
  ∞
r(t)g∗rx (t − mT)e−j2π nf (t−mT) dt, (11)
IDFT
YTF [m, n] =
1 
L−1   j2π l −∞
×√ XDD l, k̃ e− L n , (6)
L where grx (t) is the receiver filter. This can be similar to
l=0
 performing multicarrier demodulation on the time-domain
DFT signal r(t) to obtain YTF [m, n], as shown in Fig. 6. Then,
where l̃ and k̃ refer to the fixed l-th delay and k-th Doppler, the TF-domain received signal is transformed into the DD-
respectively. We can observe from (6) that the XTF [m, n] can domain by the SFFT. The resulting DD-domain received
be calculated by taking the IDFT of the Doppler and DFT signal is given by
of the delay of XDD [l, k], respectively. Therefore, we can
1 
K−1 L−1
YTF [m, n]ej2π (− K k+ L l) . (12)
m n
produce a set of LK symbols of XTF by taking the K-point YDD [l, k] = √
IDFT of the rows of XDD and L-point DFT of the columns of LK m=0 n=0
XDD , respectively. In other words, the ISFFT can be produced Note that YDD [l, k] in (12) can be rewritten as
by taking the IDFT of the Doppler, which produces the
1 
K−1
time-domain, and the DFT of the delay, which produces m
the frequency-domain. To sum up, as illustrated in Fig. 5, YDD [l, k] = √ YTF m, ñ e−j2π K k
K m=0
the ISFFT corresponds to a two-dimensional transformation 
which takes a K-point IDFT of the rows of XDD and an L- DFT

1 
point DFT of the columns of XDD . Accordingly, the ISFFT L−1
n
transforms the signal from the DD-domain to the TF-domain. ×√ YTF m̃, n ej2π L l , (13)
L n=0
Next, the SFFT-based OTFS applies Heisenberg transform 
to create the time-domain transmit signal s(t), which can be IDFT
stated as where m̃ and ñ refer to the fixed mth slot and nth subcarrier,

K−1 L−1 respectively. We can observe from (13) that the YDD [l, k] can
s(t) = XTF [m, n]gtx (t − mT)ej2π nf (t−mT) , (7) be calculated by taking the DFT of the time and IDFT of
m=0 n=0 the frequency of YTF [m, n], respectively. Therefore, we can
where gtx (t) is the transmitter pulse shaping filter. We can obtain a set of LK symbols of YDD by taking the L-point
notice from (7) that the Heisenberg transform converts the IDFT of the rows and K-point DFT of the columns of YTF
TF domain signal XTF [m, n] to a continuous time-domain respectively. In other words, the SFFT can be produced by
waveform s(t) using a transmit waveform gtx (t). This can taking the DFT of the time, which produces the Doppler-
be similar to performing multicarrier modulation on the TF domain, and the IDFT of the frequency, which produces the
domain signal XTF [m, n] to obtain s(t), which is illustrated delay-domain. To sum up, as illustrated in Fig. 5, the SFFT
in Fig. 6. corresponds to a two-dimensional transformation which takes

4490 VOLUME 5, 2024

FIGURE 5. ISFFT and SFFT block diagrams.

a K-point DFT of the columns of YTF and an L-point IDFT and

of the rows of YTF . Accordingly, the SFFT transforms the
1 
K−1
signal from the TF-domain to the DD-domain. n
χ [l + nL] = √ ZAKχ [l, k]ej2π K k , (15)
K k=0
2) DZT-BASED OTFS
where l ∈ {0, 1, . . . , L − 1}, k ∈ {0, 1, . . . , K − 1} and n ∈
As shown in Fig. 7, in the DZT-based OTFS modulation, the {0, 1, . . . , K − 1}. At the transmitter, the time-delay (TD)-
DD-domain information symbols are transformed directly domain sequence, which can be obtained by applying the
into the time-domain signal by utilizing the IDZT. On the IDZT to the DD-domain symbols, can be written as
other hand, in the DZT-based OTFS demodulation, the time-
domain signal is transformed directly into the DD-domain 
√ K−1 n
symbols by applying the DZT. Given that χ is a periodic XTD [l + nL] = L XDD [l, k]ZAKgtx [l, k]ej2π K k , (16)
sequence with period LK. Then, the DZT and IDZT can be k=0
defined, respectively as
where ZAKgtx [l, k] is the DZT of the periodically extended

K−1 sampled transmitter pulse shape gtx ( [k]LLK T), where [ · ]W
1
χ [l + nL]e−j2π K k ,
n
ZAKχ [l, k] = √ (14) denotes the modulo operation w.r.t. W and T is the periodic
K n=0 time. Taking the digital-to-analog conversion (DAC) of

VOLUME 5, 2024 4491

ALDABABSA et al.: SURVEY ON OTFS MODULATION

Then, by conducting the DZT on the TD-domain received

signal, the DD-domain received signal is expressed as
√
YDD [l, k] = LKZAKYTD [l, k]ZAK∗grx [l, k], (19)
where ZAKYTD [l, k] is the DZT of YTD [l + nL] and
ZAKgrx [l, k] is the DZT of the periodically extended
sampled receiver filter grx ( [k]LLK T).

3) CZT-BASED OTFS
Due to the discrete nature of DZT, the DZT-based OTFS
exhibits a degraded performance when the DD resolutions
are not sufficient (fractional delay and Doppler). As a result,
CZT-based implementation of OTFS has been proposed
in [14], [15], [16]. Let x(t) be a complex-valued time-
continuous function. Then, for −∞ < τ < ∞ and −∞ <
ν < ∞, the CZT and ICZT of x(t) are defined, respectively,
as
√  ∞
Zx (τ, ν) = T x(τ + iT)e−j2π iνT , (20)
i=−∞

and
FIGURE 6. Heisenberg and Wigner transforms block diagrams. √  1
T
x(t) = T Zx (τ, ν)dν. (21)
0
In the CZT-based OTFS, shown in Fig. 8, the information
symbols are arranged as a two-dimensional array X[l, k], and
encoded as a discrete DD-domain information signal, i.e.,
k
XDD [l + nL, k + mK] = X[l, k]e2π n K . (22)
Note that XDD [l, k] is a quasi-periodic function on the
information grid with period L along the delay axis and
period K along the Doppler axis, i.e.,
k
XDD [l + nL, k + mK] = XDD [l, k]ej2π n K . (23)
The discrete DD-domain signal is then lifted to a continuous
DD-domain signal, i.e.,
  τp   νp 
XDD (τ, ν) = XDD [l, k]δ τ − l δ ν−k , (24)
L K
l,k

where τp and νp are the Delay and Doppler period of CZT-

based OTFS modulation, respectively. Note that XDD (τ, ν)
FIGURE 7. DZT-based OTFS modulation and demodulation block diagrams. is quasi-periodic, i.e.
 
XDD τ + nτp , ν + mνp = XDD (τ, ν)ej2π nντp . (25)
XTD [l + nL], the OTFS transmit signal s(t) in time-domain The twisted convolution of the transmit pulse wtx (τ, ν) with
is obtained as XDD (τ, ν) gives the DD-domain transmit signal, i.e.,
wtx
s(t) = XTD [l + nL], (17) XDD (τ, ν) = wtx (τ, ν) ∗ XDD (τ, ν), (26)

where t ∈ [ l−0.5 where ∗ denotes the twisted convolution operation.

L T + nT, L T + nT].
l+0.5
wtx
At the receiver, TD-domain received signal, which is an Using (21), the ICZT of XDD (τ, ν) gives the time domain
output of the analog-to-digital conversion (ADC) of time- signal sTD (t) which is then transmitted. Without any additive
domain received signal r(t), can be written as noise, the received time domain signal rTD (t) is given by
 
YTD [l + nL] = r(t)|t= l T+nT . (18) rTD (t) = hphy (τ, ν)sTD (t − τ )ej2π ν(t−τ ) dτ dν, (27)
L

4492 VOLUME 5, 2024

 CZT-based OTFS modulation

DD-domain
Quasi-periodic Twisted
ICZT
impulse train convolution

DD-domain Transmit pulse

Information shaping filter
symbol

CZT-based OTFS demodulation

DD-domain Twisted
CZT
sampling convolution

Receive pulse
shaping filter

FIGURE 8. CZT-based OTFS modulation and demodulation block diagrams.

where hphy (τ, ν) is the DD representation/spreading function (1) Although the OTFS has two different implementations,
of the underlying physical channel. The received time both of them have the same DD-domain input-output rela-
domain signal is converted to its DD-domain representation tionship in the case of rectangular transmitter and receiver
wtx
YDD (τ, ν) via the CZT in (20). The channel acts on XDD (τ, ν) filters, integer delays and Doppler shifts. This can be clearly
by twisted convolution so that displayed as follows.
wtx For the SFFT-based OTFS, the DD-domain input-output
YDD (τ, ν) = hphy (τ, ν) ∗ XDD (τ, ν). (28) relationship can be written as
After twisted convolution of YDD (τ, ν) with a receive DD  
L−1 K−1
pulse wrx (τ, ν), we obtain YDD [l, k] = XDD l , k HDD l, k, l , k + ZDD [l, k],
l =0 k =0
wrx
YDD (τ, ν) = wrx (τ, ν) ∗ YDD (τ, ν). (29) (31)
Finally, we sample this continuous output signal to obtain a where HDD [l, k, l , k ] and ZDD [l, k] are the discrete DD-
discrete DD domain output signal, i.e., domain channel response and noise, respectively. The
 τp νp  discrete DD-domain channel response can be expressed as
wrx
YDD [l, k] = YDD τ = l ,ν = k . (30)
L K 
P
 
HDD l, k, l , k = hi w l, k, l , k , li , ki e−j2π νi τi , (32)
4) COMPARISON BETWEEN SFFT AND DZT-BASED OTFS i=1
Based on the above analyses for both SFFT-based OTFS and where ki = νi LT and li = τi Kf . Here, the values of
DZT-based OTFS, we have the following observations. ki and li determine the integer and fractional Doppler,

VOLUME 5, 2024 4493

ALDABABSA et al.: SURVEY ON OTFS MODULATION

and delay scenarios. Additionally, w(l, k, l , k , li , ki ) in (32)

refers to the sampling function for analog DD-domain
channel response, which captures joint effects of delay and
Doppler shifts, transmitter pulse shaping filter, and receiver
filter. Now, assume that ki and li are integers, and the
transmitter and receiver filters are rectangular filters, i.e.,

1, 0 ≤ t ≤ T
gtx (t) = grx (t) = (33)
0, otherwise.
Then, (31) will be simplified to two dimensional circular
convolution input-output relationship, which can be stated
as

P
YDD [l, k] = hi XDD [[l − li ]L , [k − ki ]K ] + ZDD [l, k]. (34)
i=1
Now, for the DZT-based OTFS, assume that the transmitter FIGURE 9. Comparison between OTFS and OFDM.

and receiver filters are rectangular filters. Then, TABLE 3. Simulation parameters.
1
ZAKgtx [l, k] = ZAKgrx [l, k] = √ . (35)
LK
With the integer delays and Doppler shifts, the DD-domain
input-output relationship can be obtained as (34). This means
that though the OTFS has two different implementations,
both of them have the same DD-domain input-output
relationship.
(2) The SFFT-based OTFS can be more compatible with
the conventional OFDM. On the other hand, the DZT-based
OTFS has a lower computational complexity. This can be
interpreted as follows.
If we assume that K = 1 in (6) and (13), then the OFDM
modulation and demodulation are obtained, respectively.
Therefore, the OFDM is a special case of SFFT-based OTFS, the OTFS outperforms the OFDM. During the simulation,
which means that SFFT-based OTFS is compatible with the we use the parameters labeled in Table 3 for both OFDM
OFDM. Based on (6) and (13), LK-point IDFT and DFT and OTFS. Note that the OFDM is a special case of OTFS
are required at both the OTFS transmitter and receiver. On when K = 1.
the other hand, based on (16) and (19), only L-point IDFT
and DFT are required at the OTFS transmitter and receiver,
D. LITERATURE REVIEW
respectively. Therefore, the DZT-based OTFS has a lower
1) LOW-COMPLEXITY SIGNAL DETECTION
computational complexity than the SFFT-based OTFS.
In [17], the authors presented a design of low-complexity
modulator and demodulator structures for the OFDM-based
C. COMPARISON BETWEEN OTFS AND OFDM
OTFS system. The complexity analysis revealed that com-
Recall from (34) that the input-output relation in the DD bining the ISFFT/SFFT and OFDM modulator/demodulator
domain is a two dimensional convolution and it can be blocks in OTFS systems can save an incredible amount of
rewritten as computational complexity. In [18], the authors compared
y = Hx + z, (36) the performance of OTFS systems with ideal pulse-shaping
waveforms that satisfy the bi-orthogonality conditions and
where y ∈ CLK×1 ,x ∈ CLK×1
and z ∈ CLK×1
are rectangular waveforms which do not. They found that
respectively YDD , XDD and ZDD in vectorized form and H ∈ the ideal pulse-shaping waveforms caused inter-Doppler
CLK×LK is the fading channel coefficient matrix. Using the interference (IDI). On the other hand, the rectangular
zero forcing (ZF) detection algorithm, the x can be detected. waveforms caused additional ICI and ISI. To address this, the
The detected signal can be expressed as authors proposed a low-complexity message passing (MP)
x̂ = H−1 y. (37) algorithm for joint interference cancellation and symbol
detection. This algorithm uses phase shifting to eliminate
Fig. 9 plots the bit error rate (BER) versus signal-to-noise the ICI and ISI, and adapts the MP algorithm to mitigate
ratio (SNR) for OTFS and OFDM. It can be noticed that the IDI. The results showed that the OTFS with rectangular

4494 VOLUME 5, 2024

waveforms was superior to the OTFS with ideal pulse- decoder. This equalizer is not affected by short loops and
shaping waveforms. In [19], the authors studied a low can reach the same performance as the near-optimal symbol-
complexity linear minimum mean square error (LMMSE) wise maximum a posteriori (MAP) algorithm. In [28], the
receiver for OTFS. The LMMSE takes advantage of the authors proposed a matched filtering-based message-passing
sparsity and quasi-banded structure of the matrices involved detector (MF-MPD) with a novel probability clipping (PC)
in the demodulation process. This resulted in a log-linear solution. The proposed algorithm has low complexity and
order of complexity without any reduction in the BER reduced storage requirements due to the matched filtering
performance. In [20], the authors studied the transmission processing.
of OTFS signals over a continuous-Doppler spread chan- The low-complexity detection techniques used in OTFS
nel (CoDSC). To address the interference caused by the systems can be summarized in Table 4.
CoDSC, they proposed a low-complexity equalizer using
the mathematical least squares minimum residual (LSMR)
2) CHANNEL ESTIMATION
algorithm. They then developed an iterative OTFS receiver,
which incorporated a reliability-based dynamic detector and In [29], the authors proposed embedded pilot-aided OTFS
block-wise interference eliminator, based on the LSMR channel estimation techniques. They organized pilot, guard,
equalization. In [21], the authors explored a joint radar and information symbols in the DD grids to prevent
parameter estimation and communication system that used interference between pilot and data symbols. These arrange-
OTFS modulation. They developed a simplified maximum ments were designed for OTFS with ideal and rectangular
likelihood (ML) parameter estimation approach and provided pulses over channels with integer or fractional Doppler paths.
bounds and close approximations for the radar detector’s At the receiver, channel estimation is carried out using a
mean square error (MSE). Additionally, they proposed a threshold method and the estimated channel information
practical low-complexity soft-output detector for OTFS- is used for data detection via the ML algorithm. In [30],
separated detection and decoding based on MP. In [22], the the authors introduced a time-domain channel estimation
authors proposed a low-complexity linear equalizer for OTFS technique for CP-OTFS systems with DD-domain embedded
modulation. This approach takes advantage of the block pilots. This method has a lower complexity than DD-
circulant structure of the OTFS channel matrix to reduce domain channel estimation due to the increased sparsity
complexity. The proposed technique provides exact minimum of the channel matrix when residual synchronization errors
mean square error (MMSE) and ZF solutions with lower are present. In [31], the authors studied a superimposed
complexity than the traditional matrix inversion approach. pilot (SP)-based channel estimation and data detection
In [23], the authors suggested to use a low-complexity time- framework for OTFS systems. This framework superimposes
domain matrix inversion-based low-complexity expectation low-powered pilots onto data symbols in the DD-domain.
propagation (EP) detection to enhance the error performance Two channel estimation and data detection designs were
of the OTFS and reduce the computational load. In [24], proposed for SP-OTFS systems. The first design estimates
the authors proposed a new cross-domain iterative detection the channel by treating data as interference, which reduces
technique for OTFS modulation. They analyzed the state its performance at the high SNR. The second design
evolution of the proposed algorithm and studied its detection improves this issue by iteratively alternating between channel
performance, including the MMSE performance and the DD- estimation and data detection. Both designs detect data
domain effective SNR. They demonstrated that the proposed using a MP algorithm which takes advantage of the OTFS
algorithm can achieve the same error performance as channel sparsity and thus has low computational complexity.
maximum-likelihood sequence estimation (MLSE) detection, In [32], the authors proposed a data-aided channel estimation
even in complex fractional Doppler shifts, but requires much algorithm for OTFS with the SP and data transmission
less detection complexity. In [25], the authors proposed a new scheme. First, they coarsely estimated the channel based on
type of OTFS detector based on the approximate message the pilot symbol. Then they used an iterative process to
passing (AMP) and the unitary AMP (UAMP). This detector detect the data symbols and refine the channel estimates.
takes advantage of the structure of the channel matrix, In [33], the authors introduced a Bayesian learning (BL)
resulting in an efficient implementation. Furthermore, the framework for channel state information (CSI) acquisition.
UAMP-based detector includes the estimation of noise vari- This frame utilizes the sparsity of the DD-domain to enhance
ance. In [26], the authors proposed a novel low-complexity the estimation accuracy compared to the traditional MMSE-
parallel message passing (PMP) detection scheme for OTFS based approach. In [34], the authors suggested a new channel
modulation. This scheme synchronizes the detection of all estimation technique (smoothness regularized channel main
symbols and optimizes the structure of the symbol detection diagonal estimation (SR-CMDE)) with the capability of
algorithm, resulting in a significant reduction in detection reducing signal leakage for OTFS modulation. This approach
time. The authors of [27] suggested a turbo equalizer increases channel estimation accuracy while decreasing the
called doubly iterative sparsified MMSE. It can iteratively amount of signaling required.
exchange the extrinsic information between a soft-input- The channel estimation techniques used in OTFS systems
soft-output MMSE estimator and a soft-input-soft-output can be summarized in Table 5.

VOLUME 5, 2024 4495

ALDABABSA et al.: SURVEY ON OTFS MODULATION

TABLE 4. A summary of low-complexity detection techniques in OTFS systems.

TABLE 5. A summary for channel estimation techniques used in OTFS systems.

3) PEAK-TO-AVERAGE POWER RATIO maximum PAPR increases linearly with the number of
Doppler bins. In [36], an efficient PAPR reduction method
In [35], the authors investigated the PAPR of the OTFS
modulation waveform. They derived an upper bound on based on the iterative clipping and filtering framework was
the PAPR of the OTFS signal and demonstrated that the proposed for pilot-embedded OTFS modulation. This method

4496 VOLUME 5, 2024

allowed the guard symbols to be padded with signals of small the system’s performance. In [45], waveform design is
magnitudes. The clipping noise other than the clipped signal considered along with channel estimation based on unique-
was then filtered. The filtering coefficients for the clipping word and CP-free transmission. It is shown that for highly
noises in the data-plus pilot region and guard region were time-varying channels, it is advantageous to suppress the CP
set to different values to achieve a good trade-off between so that the channel estimation can be improved. A framework
PAPR and BER performance. In [37], the authors studied is developed to choose the waveform and frame parameters,
the PAPR of OTFS signals when the amount of data is large such as the number of subcarriers and subblocks. For the
in the DD-domain. They found that when the number of CP-free transmission, it is shown that the waveforms that
data points is very large, the complex-valued OTFS signals spread symbols energy in the time-domain per sub-block are
tend to follow a Gaussian distribution due to the central the most resilient ones due to equal interference levels of the
limit theorem (CLT). Furthermore, the extremal theory of the symbols. The authors in [46] presented a generic modulation
Chi-squared process for stationary OTFS signals was used framework for robust wireless communications over doubly
to derive an accurate expression for the PAPR distribution. dispersive channels. This robustness is achieved by spreading
the symbols over time and frequency so that they experience
4) SYSTEM DESIGN the same channel gain. By exploiting an additional degree
The authors of [38] studied the input and output relation of freedom enabled by the generic framework, the authors
of an OTFS system when using a variety of pulse- proposed a low-complex waveform based on the sparse
shaping waveforms. They demonstrated that the OTFS has Walsh-Hadamard transform.
a direct sparse input-output relation, which allows to use
low-complexity detection algorithms. In [39], the authors 5) PERFORMANCE ANALYSIS
proposed an OTFS system design developed on the Discrete The authors of [47] derived a closed-form BER expression
Fractional Fourier Transform (DFrFT) based OFDM system. for the ZF-OTFS receiver in the presence of both perfect and
It was designed to be more efficient than the traditional imperfect receive channel information. In [48], the authors
OTFS system while having the same design complexity. obtained a closed-form expression for the outage probability
In [40], the authors investigated the design of receivers (OP) of the OTFS-based low earth orbit satellite (LEO-
for OTFS with fractionally spaced sampling (FSS). They Sat) transmission. In [49], the authors investigated the error
derived a general channel input-output relationship for OTFS performance of rectangular pulse-shaped OTFS modulation
in DD-domain without relying on unrealistic assumptions with a practical receiver. In [50], the error performance of
such as ideal bi-orthogonal pulses. Next, they proposed coded OTFS modulation over high-mobility channels was
two equalization algorithms, iterative combining MP and inspected. In [51], the authors proposed precoders for OTFS
turbo MP, to detect symbols by exploiting the DD channel systems with linear equalizers to improve the BER. These
sparsity and the channel diversity gain through FSS. In [41], precoders are designed to optimally distribute the power
the authors inspected a receiver design based on the basis among the sub-channels with different gains in the DD-
expansion model (BEM) OTFS. A low-order generalized domain. In [52], the authors studied the security performance
complex exponential BEM (GCE-BEM) was initially used of an uplink LEO-Sat system, in which a cooperative UAV
to estimate the channel with minimal pilot overhead. The is used to send interference signals to a reconnaissance
accuracy of the estimation was then improved by iteratively satellite. The OTFS technique was employed in the uplink
refining the channel estimation and equalization, using a transmission to counter the severe Doppler effect caused by
high-resolution GCE-BEM model with a large BEM order the high-mobility of the LEO-Sat. In [53], the authors studied
and exploiting the detected data symbols as pseudo-pilots. the OTFS performance over static multipath channels.
In [42], the authors studied the effects of using a window
at the transmitter or receiver for OTFS modulation. They III. MULTIANTENNA OTFS SYSTEMS
found that any window can provide the same performance As compared to the conventional single-input single-output
in effective channel estimation. Additionally, the transmitter (SISO) approaches, MIMO technology paves the way
window can be seen as a power allocation in the TF- for dramatically improved spectral and energy efficiency
domain, while the receiver window can create colored noise. performance at a manageable cost by counting on the avail-
In [43], the authors proposed a two-stage equalizer system able spatial degrees of freedom. Specifically, MIMO enables
that processes the received signals in two domains. Initially, linear scaling of the channel capacity with the rank of the
a sliding window-assisted minimum mean square error (SW- channel matrix. Due to its versatility and ease of integration,
MMSE) equalizer is used in the TF-domain to reduce the ICI. MIMO methods have been widely embraced in contemporary
Then, the DD-domain equalizer is applied to the output of and emerging transmission systems under distinct scenarios.
the first stage to reduce the residual ISI. The authors in [44] The existence of multiple antennas can mainly be leveraged
introduced a new OTFS structure, which adapts the full-CP in two distinct ways: spatial multiplexing for boosting the
(FCP) structure used in OTFS modulation. This approach spectral efficiency and spatial diversity for improving the
utilizes the FCP structure in the DD-domain to decode the transmission reliability. Single and multiuser MIMO, massive
data in the CP block instead of discarding it, thus improving (large-scale) MIMO, cooperative MIMO, distributed MIMO,

VOLUME 5, 2024 4497

ALDABABSA et al.: SURVEY ON OTFS MODULATION

FIGURE 10. MIMO-OTFS transceiver block diagram.

transmit/receiver beamforming, space-time coding, physical model allows to employ a number of signal detection
layer security, mmWave MIMO, MIMO radar (for joint algorithms. The authors present a low-complexity iterative
communications and sensing), and interference alignment are MP algorithm for the detection process and show that
among some such applications of smart antenna technologies. it remarkably outperforms the traditional MIMO-OFDM
As a recently introduced DD domain transmission scheme, scheme in terms of BER performance for a 2 × 2 MIMO
OTFS is proven to bring about promising performance over system over doubly-dispersive channels. The introduced
high-mobility frequency-selective fading channels. In order algorithm attains solid BER performance even at high
to attain robust resilient performance over doubly-dispersive Doppler frequencies (e.g., 1880 Hz resulting from a speed
channels, the combination of MIMO and OTFS techniques of 507.6 km/h at a carrier frequency of 4 GHz). In addition,
has attracted significant attention with the objective of reap- a channel estimation scheme where DD impulses are utilized
ing the advantages of both approaches. A transceiver block as pilots in the DD domain is presented. A discrete-
diagram of MIMO-OTFS is sketched in Fig. 10 where nt and time analysis of MIMO OFDM-based OTFS modulation
nr represent the number of transmit and receive antennas, is performed in [55]. A precise description of the ergodic
respectively. Appropriate space-time encoding (enc.) and capacity is presented and it is shown that both OFDM-
decoding (dec.) processes are invoked at the transmitter and and OTFS-based approaches accomplish the same ergodic
receiver, respectively. OTFS modulator is comprised of a capacity. In [56], the diversity level of OTFS is investigated
respective application of ISFFT, transmitter windowing (Tx and it is demonstrated that the asymptotic diversity order
Window.), and Heisenberg transform operations. In addition, of the single-antenna OTFS (as SNR → ∞) equals unity.
Wigner transform, receiver windowing (Rx Window.), and Also, the authors introduce a phase rotation (PR) method
SFFT operations are implemented in the given order within that extracts full diversity of P in the DD domain. Here, P
OTFS demodulator. When the transmit antennas are used for stands for the number of resolvable multipath components
the purpose of spatial multiplexing, OTFS modulated signal in either the delay or Doppler dimension. The idea relies
vectors are independently transmitted from the transmit upon that the minimum rank of the symbol difference matrix
antennas in a parallel and simultaneous fashion. Assuming is augmented to P by means of multiplying the OTFS
a far-field scenario, the channel gain between the kth transmit vector with a diagonal PR matrix with different
transmit antenna and lth receive antenna corresponding transcendental numbers. In addition, it is proven that the
to delay τ and Doppler ν in the DD domain can be asymptotic diversity order is equal to nr for an nt × nr
written as MIMO-OTFS scenario where nt independent OTFS signal
vectors are simultaneously transmitted in a parallel manner.

P
By generalizing the PR method, it is also shown that the
hlk (τ, ν) = hlki δ(τ − τi )δ(ν − νi ) (38)
presented MIMO-OTFS system can attain the full diversity
i=1
of Pnr . The use of space-time coding (STC) for a MIMO-
for k ∈ {1, 2, . . . , nt } and l ∈ {1, 2, . . . , nr } where P denotes OTFS setup is initially investigated in [57]. By adopting
the number of resolvable multipath components in either the well-known Alamouti code, the authors show that full
the delay or Doppler dimension. Also, we further assume spatial and DD diversity of 2P can be achieved in a 2 × 1
that the adjacent antennas at either side are placed half a STC-OTFS system with the PR operation. The work in [58]
wavelength apart. The probability density function of the evaluates the spectral efficiency of a 2 × 2 MIMO-OTFS
random variable hlki is determined by the underlying small- scheme and confirms that the capacity can be doubled in this
scale fading type. For instance, it has a zero-mean complex case as compared to a SISO-OTFS scenario. The diversity
Gaussian distribution under Rayleigh fading. In this section, performance of receive antenna selection (RAS) (based on
a review of the relevant studies on MIMO-OTFS is provided. the maximum channel Frobenius norm in the DD domain)
In [54], the input/output relation is formulated in a vec- for a number of MIMO-OTFS arrangements is studied
torized fashion for the MIMO-OTFS system. The obtained in [59]. The findings indicate that when combined with

4498 VOLUME 5, 2024

a PR scheme, single-input multiple-output OTFS (SIMO- scheme is demonstrated to attain notably better spectral
OTFS) and STC-OTFS systems with RAS can extract the efficiency as compared the OFDM-based approach over
full spatial and DD diversity. On the other hand, MIMO- high-mobility channels. The authors design a low-overhead
OTFS system with RAS is shown to extract the full DD and low-complexity receiver structure for the OTFS system
diversity only (not the full receive diversity) as a result of in [68] by deploying a large-scale antenna array at the
rank deficiency even under the PR approach. In [60], the receive side to decouple the received signals from distinct
diversity performance of transmit antenna selection (TAS) is angles of arrival into multiple parallel signal branches. The
inspected for an nt × nr MIMO-OTFS system with only one proposed scheme encompasses the pilot pattern design, the
resolvable path in the DD domain, i.e., P = 1. By assuming channel estimation, the symbol detection, and the carrier
a limited feedback from the receiver to the transmitter, TAS frequency offset compensation. In [69], a downlink MIMO-
is performed based on the maximum channel Frobenius OTFS system is considered where a BS transmits information
norm in the DD domain. It is proven that full spatial to multiple users at the same time. Assuming that each
diversity of nt nr is accomplished when only a single transmit node is equipped with multiple antennas, two low-complexity
antenna is selected. On the other hand, the diversity order precoding algorithms are introduced and shown to have
equals nr when TAS activates more than one transmit significantly reduced computational complexity with a lim-
antenna. An OTFS system with multiple transmit antennas ited performance loss as compared to the high-complexity
is investigated in [61] by precluding the bi-orthogonal block diagonalization precoding. The authors in [70] devise
assumption in the ideal pulse shaping. The rectangular low-complexity ZF and MMSE receivers for MIMO-OTFS
pulse shaping is considered and the guard symbols, the systems by considering both perfect and imperfect CSI
pilot symbols, and the information symbols are accordingly scenarios. The complexity reduction is performed by utilizing
allocated to achieve the transmit diversity. In [62], the the doubly-circulant structure of the MIMO-OTFS channel
authors present cyclic delay-Doppler shift technique (in time- matrix. In [71], a tight signal-to-noise-plus-interference-ratio
domain and modulation-domain forms) for MIMO-OTFS (SINR) approximation for the receiver structures in [70] is
systems to extract transmit diversity over doubly-selective analytically determined under imperfect receive CSI. Based
channels. The main idea relies on introducing DD shifts at on the inherent channel sparsity and the channel-agnostic
distinct transmit antennas such that the number of effective form of matrices involved in the detection process, the work
propagation paths in the wireless channel is increased. It is in [72] presents a low-complexity linear minimum mean
shown that the introduced scheme can yield full transmit squared error (LMMSE) receiver structure for practical-
diversity gain in certain cases. Analytical BER upper bounds pulse-shaped MIMO-OTFS systems. It is shown that the
for MIMO-OTFS systems are obtained in [63] with ideal proposed scheme with only a log-linear complexity yields
and generalized waveforms by utilizing miscellaneous kinds precisely the same solution, and therefore the same BER
of guard intervals. performance as compared to the traditional LMMSE receiver
The spectral efficiency of an uplink MIMO-OTFS which has a cubic order of complexity. A low-complexity
system is inspected in [64] with delay division multiple deep neural networks (DNN) model is proposed in [73]
access (DDMA) and Doppler division multiple access for OTFS signal detection under both single-antenna and
(DoDMA) schemes where MMSE combining is adopted multiantenna cases. It is demonstrated that when the additive
at the receive side. The authors numerically demon- noise at the receive side differs from the conventional
strate the performance enhancement of MIMO-OTFS independent and identically distributed Gaussian model, the
over MIMO-OFDMA systems. In [65], the DD domain DNN-based detector can outperform the standard ML and
Tomlinson-Harashima precoding is studied for a downlink MMSE detection approaches. In [74], the authors present a
MIMO-OTFS system where a base station (BS) with multiple linear-complexity detection scheme for MIMO-OTFS with
antennas simultaneously communicates with various single- rectangular pulse-shaping waveform. The introduced method
antenna users. The presented technique does not necessitate exploits maximum-ratio combining (MRC) at the receiver
any matrix decomposition or inversion. The authors conclude and is shown to accomplish enhanced performance as com-
that the sum-rate increases logarithmically with the number pared to LMMSE and MP detectors. Relying on the sparsity
of antennas and linearly with the number of users. By of the channel matrix, a LSMR channel equalizer and an
utilizing the block circulant structure of the effective DD enhanced data detection technique are introduced for MIMO-
MIMO channel matrix, ZF and MMSE equalizers with OTFS in [75]. It is demonstrated that the presented scheme
low-complexity are proposed in [66] for a 2 × 2 MIMO- can effectively and robustly demodulate the superimposed
OTFS system. In [67], a downlink MIMO-OTFS system high-order quadrature amplitude modulation (QAM) symbols
is inspected where a BS with a large number of antennas even under high-mobility. In [76], the authors propose a low-
simultaneously communicates with multiple single-antenna complexity MMSE receiver for the uplink of a rectangular
users. It is shown that separate demodulation of each DD pulse-shaped massive MIMO-OTFS system. The presented
domain information symbol at the users is possible by approach makes use of the sparsity of the channel matrix
using a multiuser precoder at the BS and a low-complexity and the MMSE expression matrix structure. It is shown that
detection algorithm at the users. The proposed MIMO-OTFS as compared to its traditional counterparts, the recommended

VOLUME 5, 2024 4499

ALDABABSA et al.: SURVEY ON OTFS MODULATION

receiver has a considerably lower computational complexity method is developed. It is shown that when the maximum
without any degradation in the BER performance. Doppler shift is within OTFS limits, doubly-dispersive
In [77], [78], the BER performance of MIMO-OTFS mod- channel effects even out in the large antenna regime. A joint
ulation is investigated with decode-and-forward relaying in radar sensing and data transmission system is introduced
high-mobility environments where all the nodes are equipped in [86] where a multiantenna BS equipped with a reduced
with multiple antennas. The communication between the number of RF chains and a radar receiver broadcasts data
transmit and receive ends takes place in two hops through a utilizing OTFS in mmWave frequency bands. Two modes of
single relay and through multiple selected relays respectively operation are separately inspected. The proposed approaches
in [77] and [78]. The achievable asymptotic diversity order are shown to perform effectively in terms of detection
is determined and a PR approach on the transmitted OTFS and estimation quality. In [87], the authors make use of
frames is proposed to increase the diversity level under both the radar sensing information about the users and the
cases. surrounding environment to assist the channel estimation
In [79], a new path division multiple access tech- for MIMO-OTFS systems. Numerical findings reveal that
nique is proposed for massive MIMO-OTFS networks for the introduced method performs better than the traditional
both downlink and uplink scenarios. The efficacy of the orthogonal matching pursuit (OMP)-based approach regard-
introduced scheme is established through a number of ing the normalized MSE for the channel estimation and the
simulation results. An uplink massive MIMO-OTFS system pilot overhead. The authors of [88] carry out a thorough
is considered in [80] where multiple single-antenna users study on signal modeling, performance analysis, algorithm
simultaneously transmit their symbols to a multiantenna BS, design, and numerical evaluations for massive MIMO-
which applies ZF beamforming to decode the users’ signals. OTFS-based simultaneous localization and communications.
A rigorous performance analysis is also provided. In [81], Initially, a massive MIMO-OTFS channel model is built by
the spectral efficiency of cell-free massive MIMO-OTFS is assuming the ultra-wide bandwidth mmWave signals and
studied by considering the impact of channel estimation. large antenna arrays. Cramer-Rao lower bounds are then
Analytical expressions are derived for the spectral efficiency obtained for channel estimation and localization results based
under both downlink and uplink cases. The numerical on the new multipath wideband channel model. The joint
results exhibit the advantage of the MIMO-OTFS system radar and communication problem is studied in [89] by
in yielding enhanced data rates over high-mobility channels adopting multiuser MIMO scenarios. A new scheme named
as compared to the traditional MIMO-OFDM approach. as non-uniform OTFS is introduced and its performance
The downlink spectral efficiency of a cell-free massive is compared with other related OTFS-based methods for
MIMO-OTFS is inspected in [82] where every receiver relies monostatic MIMO radar applications and under standard
on MMSE-based successive interference cancellation (SIC) 3GPP vehicular channel models. In [90], a MIMO-OTFS
operation. The authors present a max-min fairness resource system with 19.3 bits/s/Hz spectral efficiency at a carrier
allocation scheme based on per-user spectral efficiency. It frequency of 28 GHz is experimentally investigated in over-
is demonstrated by simulation results that as compared to a the-air and mobility environments. The superior robustness of
full power transmission case, the introduced approach can MIMO-OTFS as compared to MIMO-OFDM is empirically
considerably enhance the spectral efficiency at each user. demonstrated over time-variant channels. A LEO satellite
The authors of [83] investigate the grant-free random access communication system based on MIMO-OTFS is studied
in LEO satellite communications by embracing massive in [91] by taking spatial antenna correlation into account.
MIMO-OTFS where numerous Internet-of-Things devices In order to cope with the spatial correlation, the authors
desire to communicate with a LEO satellite in a sporadic propose whitening transformations at the transmit and receive
manner. In order to estimate the channel and to detect active sides. These decorrelation operations are numerically shown
devices, a novel method relying on the channel sparsity in to bring about important error performance enhancement.
the DD-angle domain is introduced. Assuming an URLLC However, the whitening transform technique in [91] cannot
scenario with ideal pulse shape functions and fractional completely cancel spatial correlation at the receiver. In order
delay and Doppler shifts, an uplink MIMO-OTFS system to remedy this problem, an alternative whitening technique
is inspected in [84] where a set of single-antenna users is developed in [92] over a time-varying spatially correlated
communicate with a BS with multiple antennas. The BS MIMO-OTFS system. An upper bound on the pairwise
initially applies MRC and subsequently invokes the ML error probability (PEP) is obtained as a function of the
decoding rule to decode the desired symbols. A number of diversity and coding gains and SNR. In [93], a high-Doppler
conclusions are drawn on the achievable rate and diversity of airborne communication ad-hoc network case is investigated
the system. In [85], a scenario where a single-antenna user where relative mobile node speeds can be as high as 1200
transmits towards a BS with a massive number of antennas m/s. It is shown that MIMO-OTFS has the potential to
is treated for OTFS modulation. In order to entirely remove offer sufficiently reliable transmission under the consid-
the CP and to break the corresponding limitation on the ered scenario. A summary of the prominent MIMO-OTFS
maximum Doppler frequency that can be tolerated, a time- studies with a focus on the diversity level is provided in
reversal MRC with residual Doppler correction windowing Table 6.

4500 VOLUME 5, 2024

TABLE 6. A summary of the diversity-oriented multiantenna OTFS studies.

The work in [94] treats the joint channel estimation and the superior BER performance of the introduced scheme
data detection problem in hybrid reconfigurable intelligent as compared to the other related approaches. MIMO/SISO-
surface (HRIS) aided mmWave MIMO-OTFS systems. The OTFS systems with intelligent reflecting surface (IRS)
authors determine a new transmission framework where assistance are investigated in [97] with MMSE detection. The
some HRIS elements are alternatively activated throughout authors offer a method to conclude a set of IRS phase shifts
the pilot durations and HRIS stays in passive mode through- such that the received SNR is maximized. It is demonstrated
out each OTFS block duration such that the impinging that the presented technique outperforms its OFDM-based
signal is reflected without any absorption. A number of IRS-assisted counterpart in terms of BER performance.
simulation results are drawn to confirm the strength of In [98], massive MIMO-OTFS downlink channel estima-
the introduced scheme under different scenarios. A point- tion problem is studied. It is proven that the time-variant
to-point MIMO-OTFS scenario is covered in [95] and a massive MIMO channel has three dimensional sparsity given
low-complexity hybrid digital-analog beamforming transmis- by normal sparsity along the delay dimension, block sparsity
sion framework is proposed. It is confirmed by numerical along the Doppler dimension, and burst sparsity along the
results that the presented technique can attain near-optimal angle dimension. Building upon this fact, the estimation
rate performance. In [96], the authors adopt an air-to-ground problem is converted into a sparse signal recovery problem
communications scenario and promote a low-complexity which is then solved by resorting to a structured OMP
massive MIMO-OTFS hybrid precoding method with rect- algorithm. The work in [99] considers an uplink-aided high-
angular waveforms. Numerical results are used to show mobility downlink channel estimation scheme for massive

VOLUME 5, 2024 4501

ALDABABSA et al.: SURVEY ON OTFS MODULATION

MIMO-OTFS networks. The expectation-maximization (EM) scenarios are studied and a new two-step procedure is
based variational Bayesian framework is utilized to deter- presented for transmit beamformer/precoder and receiver
mine the uplink channel parameters including the angle, the combiner design, and for estimating the DD-domain CSI.
delay, the Doppler frequency, and the channel gain for each It is shown that the corresponding CSI exhibits sparsity
physical scattering path. Then, using the angle, delay, and and block sparsity under analog and hybrid beamforming
Doppler reciprocity between the uplink and the downlink cases, respectively. Bayesian learning is then resorted for
(even under the frequency division duplex mode), downlink CSI estimation. Simulation results affirm the improved CSI
channel parameters as the angles, the delays, and the Doppler estimation performance of the introduced algorithm against
frequencies are recovered at BS. The authors in [100] other relevant sparse signal recovery techniques. In order
address the channel estimation problem for mmWave massive to eliminate the channel estimation performance degradation
MIMO-OTFS systems where the large OTFS symbol size and due to the off-grid channel parameters, the authors of [107]
the massive number of antennas may seriously escalate the establish a cross-domain channel estimation method for
involved computational complexity. As a remedy, a tensor- OTFS-based hybrid beamforming systems. By using the
based OMP method is presented by utilizing the channel time-domain and DD-domain signal models, two parallel
sparsity in the DD-angle domain. Numerical results are used sparse recovery problems are built and solved. It is shown
to confirm the superiority and robustness of the introduced that the presented technique can attain superior performance
algorithm. A CSI acquisition method is presented in [101] as compared to the relevant schemes in the presence of the
for downlink massive MIMO-OTFS systems under the effect off-grid channel parameters. In [108], a cubature Kalman
of fractional Doppler. Deterministic pilot design and the filter is used to address channel estimation problem in
determination of the channel estimation algorithm problems mmWave MIMO-OTFS systems. The authors adopt a beam-
are covered. The simulation results verify the efficacy of the switching approach to minimize the tracking errors and show
introduced technique in acquiring precise downlink CSI. The by simulations that the proposed method outperforms the
authors of [102] present a channel estimation technique for prior relevant techniques. A CSI estimation framework based
massive MIMO-OTFS systems with the aim of lessening ISI on an orthogonal affine-precoded superimposed pilot struc-
when the subtraction of the two delay paths is less than the ture in DD domain is presented in [109] for CP-aided SISO-
resolution of the system. A 3D structured OMP approach is and MIMO-OTFS systems with arbitrary transmitter-receiver
adopted by utilizing the fact that the inner product descent pulse shaping. The authors derive Bayesian Cramer-Rao
ratio difference of the traditional OMP algorithm has a bounds on the system performance and provide a number of
maximum value near sparsity. Simulation results verify that BER simulation results to expose the enhanced performances
the proposed method outperforms the related conventional of the proposed strategies in various settings. In [110],
algorithms under distinct scenarios. In [103], a sparse CSI online Bayesian learning-assisted CSI estimation methods
estimation model for MIMO-OTFS systems is proposed. In are developed for both SISO- and MIMO-OTFS modulated
order to achieve a reduction in terms of the training overhead, systems. Under the MIMO-OTFS scenario, the authors
the pilot signals are directly transmitted over the time- offer an online row and group sparse Bayesian learning
frequency-domain grid to estimate the DD-domain CSI. It is method for channel estimation and a low-complexity detector
shown that the corresponding DD-domain CSI turns out to relying on an iterative block matrix inversion approach.
be simultaneously row and group sparse. The authors present The work in [111] introduces a channel estimation method
an OMP algorithm complemented by an enhanced Bayesian for MIMO-OTFS. The objective function is first expressed
learning framework and a low-complexity linear detector. as a block sparse signal recovery problem and then a
The performance advantages of the introduced techniques solution is determined by a block sparse Bayesian learning
against the other related approaches are illustrated through technique with block reorganization. Simulation results are
simulation results. A DD-angular domain representation of used to show the superiority of the presented scheme in
the wireless channel is developed in [104] for MIMO- comparison with the other relevant methods. In [112], the
OTFS systems. By utilizing the intrinsic four dimensional pilot pattern construction problem is treated for channel
sparsity, the authors devise a time-domain pilot aided estimation in MIMO-OTFS systems. As compared to the
channel estimation model which functions based on an OMP conventional approach, the authors’ proposed pilot scheme
framework. A novel channel estimation method based on is shown to exhibit similar BER performance with a
compressed sensing is introduced for MIMO-OTFS systems lower pilot overhead. The authors of [113] study channel
in [105] by taking advantage of the structured sparsity of the estimation problem in an asymmetrical massive MIMO-
DD domain MIMO channel. The authors devise a row-block OTFS system where antenna array sizes differ in the uplink
OMP algorithm and show that their approach yields better and downlink. A parametric channel estimation method
performance as compared to the similar methods relying on called proximal gradient OMP is introduced and shown to
compressed sensing. In [106], the authors initially develop enhance the precision of channel estimation as compared
the DD-domain input-output relationship considering a to the other methods that perform estimation directly from
DD-angular domain channel model for mmWave MIMO- delay, Doppler, and angular domains. In [114], an iterative
OTFS systems. Then, both analog and hybrid beamforming algorithm is introduced for channel estimation and data

4502 VOLUME 5, 2024

detection in the DD domain for MIMO-OTFS systems. The spread over the whole TF grid in OTFS. Hence, there exists
presented scheme iterates between MP-aided data detection an analogy between the two ideas and it is highly projected
and data-aided channel estimation where two methods are that the upcoming communication systems will promote
recommended for the latter process. It is shown that the both approaches. Recently, a great deal of interest has been
proposed approach outperforms the related schemes in [103] concentrated on an adequate and effective combination of
and [111] with regard to the channel estimation quality and OTFS and multiple access techniques by exploiting TF, DD,
BER performance. space, power, and code domains. Fig. 11 exemplifies one
such application of user multiplexing in DD domain with
IV. MULTIPLE ACCESS TECHNIQUES AND OTFS OTFS for downlink and uplink cases, respectively. In this
5G and beyond (5G +) systems are envisaged to support section, a survey of prominent studies on OTFS with various
high data rate and URLLC as well as massive and ubiquitous orthogonal and non-orthogonal multiple access methods is
connectivity. Some of these systems are further expected provided.
to efficiently operate even under high-mobility conditions, In [116], an uplink multiple access technique is proposed
e.g., under a speed of 500 km/h. However, in such cases, for OTFS-based communication systems over doubly-
a corresponding Doppler shift is also coupled with each dispersive fading channels. By subdividing delay and
of the multipath components over the wireless channel Doppler axes into equal-length portions in DD domain,
which thereupon exhibits selective fading both in time and a certain number of DD resource blocks are obtained.
frequency domains. The transmitted signal is consequently These are accordingly allocated to distinct users in DD
exposed to Doppler spread together with delay spread. domain such that non-overlapping localized portions of TF
In response to stringent quality of service requirements, domain are assigned to different users without any interuser
multiple access techniques are designed with the purpose interference. The introduced scheme is shown to attain
of maintaining high data rate and reliable communications high sum spectral efficiency over doubly-dispersive fading
to a multitude of users/equipments/terminals over a mutual channels with no guard band. The problem of devising
channel. When a common entity such as a BS commu- a multiple access plan for an uplink OTFS system is
nicates with a number of users over a joint medium, the studied in [117]. Depending on how DD resource blocks are
corresponding channel and system are respectively called distributed to users, three distinct techniques are considered.
a broadcast channel and a downlink system. On the other In the first and second schemes, users are respectively
hand, the opposite scenario in which multiple users transmit multiplexed along the delay and Doppler axes such that
data to a common receiver forms an uplink system over a single user’s signal spans the whole TF plane. In the
a multiple access channel. Multiple access methods and third scheme, DD resource blocks are allocated similar
associated waveforms have traditionally played key roles to [116] where any user’s signal is constrained to span
during transition from one communications standard to only a subdivision of TF plane. It is numerically shown
another. Orthogonal multiple access (OMA) techniques have that the first scheme outperforms the others as wells as
principally been adopted from the early generations to the OFDMA and single-carrier FDMA techniques in terms of
currently deployed 5G systems which implement OFDMA. BER. Also, a pilot-based channel estimation method with
Orthogonal TF resource blocks are allocated to the users the first scheme is demonstrated to provide a performance
under OFDMA, which features some important assets given close to that with perfect channel knowledge. In [118],
by its resilience against frequency-selective fading and the authors study DD channel estimation for an uplink
its ease of implementation and integration in terms of system with OTFS where user multiplexing is applied in
backward compatibility. As a result of the first advantage, DD domain. In order to exploit the inherent sparsity of
orthogonality among users can be retained at the receive the channel in DD domain, a compressed sensing-based
side(s) even under the presence of delay spread due to estimation method is implemented. The results indicate the
the multipath propagation. However, as the mobility of the superior performance of the proposed technique against an
active terminals rises and Doppler spread becomes more impulse-based channel estimation scheme reported for uplink
pronounced, OFDMA’s error probability performance starts OTFS systems. A novel path division multiple access scheme
declining and preserving the orthogonality among distinct for both downlink and uplink massive MIMO-OTFS systems
signals at the receiver entails computationally complex is introduced in [79], [119]. Modeling the received signal
equalization schemes. in the DD-angle domain, a path scheduling algorithm is
Thanks to its robustness under both time- and frequency- conceived. In addition, MRC-based and beamforming-based
selective fading, OTFS combined with an appropriate detection schemes are designed respectively for uplink and
multiple access plan has the potential to offer a solution downlink scenarios. Simulation results verify the efficiency
for maintaining acceptable performance over high-mobility of the presented technique. DDMA and DoDMA methods
scenarios. In contrast to OMA practice, NOMA approaches are studied for an uplink OTFS system in terms of
allocate more than one user in a unique resource unit achievable rate under MIMO and SISO scenarios in [64]
such that a group of users simultaneously exploit the same and [120], respectively. It is asserted that a robust SINR
frequency band [115]. Similarly, the modulated symbols are performance can be attained thanks to a favorable channel

VOLUME 5, 2024 4503

ALDABABSA et al.: SURVEY ON OTFS MODULATION

FIGURE 11. An example of user multiplexing in DD domain with OTFS for downlink and uplink scenarios, respectively.

hardening effect appearing as a result of DD domain access is noticeably higher than those of the other considered
signal processing. In addition, numerical results reveal a schemes.
noteworthy enhancement in the users’ achievable rates in A combination of OTFS and NOMA is presented in [124]
comparison with the conventional OFDMA-based approach. for both uplink and downlink scenarios with heterogeneous
The outage performance of OTFS modulation is analyzed user mobility profiles. User grouping is applied such that the
in [121] for doubly-dispersive channels. Both the single- ones with high-mobility get service in the DD plane with
user transmission and the uplink multiuser transmission their signals being modulated by OTFS. The low mobility
(relying on DDMA introduced in [64], [120]) scenarios are users, on the other hand, are served in TF plane such that
covered. A number of numerical results are provided to their signals are modulated in a fashion analogous to con-
show that OTFS-based systems achieve considerably superior ventional OFDM. It is demonstrated that both types of users
outage performance as compared to traditional OFDM-based enjoy from the introduced OTFS-NOMA method in terms
approaches. of enhanced spectral efficiency and reduced latency. The
In [122], the authors focus on a multiuser uplink scenario beamforming design for an OTFS-NOMA downlink scenario
with OTFS and introduce a user multiplexing scheme (called is investigated in [125] by assuming that a BS simultaneously
interleaved TF multiple access) where TF resource blocks communicates with multiple low-mobility users along with
allocated to distinct users are non-contiguous and interleaved. one high-mobility user over the same frequency band. The
The simulation results reveal that as compared to the relevant findings assert that the presented beamforming scheme can
method of [116], the proposed technique accomplishes better yield an important performance advantage as compared to
performance in terms of BER and PAPR. Several OMA a scenario with random beamforming. In [126], an uplink
methods are integrated with an uplink OTFS system in [123] DD-angle domain NOMA approach is inspected for massive
where non-overlapping resource blocks in DD and/or TF MIMO-OTFS systems. The authors suggest to schedule
domains are allocated to distinct users. Specifically, guard multiple users with overlapped angle signatures within the
band based multiple access, interleaved DD multiple access, same DD domain resource block. User clustering and optimal
and interleaved TF multiple access techniques are investi- transmission strategy problems are also investigated. The
gated in terms of their spectral efficiency performances under efficiency of the presented scheme is confirmed by means
practical rectangular pulses. Simulation results indicate that of simulation results. An OTFS-NOMA system is treated
the sum spectral efficiency of the interleaved DD multiple in [127] by assuming that multiple high-mobility users all

4504 VOLUME 5, 2024

served by OTFS are multiplexed in power-domain for both present power allocation and relay selection methods to
downlink and uplink scenarios. Various power allocation provide reliability and security. Numerical results signify
policies are devised to maximize the sum spectral efficiency that a notable enhancement can be achieved as compared
under distinct cases. Performance comparisons reveal that the to the relevant OFDM-NOMA techniques. In [134], [135],
presented OTFS-NOMA scheme achieves a larger spectral an iterative low-complexity method for equalization and
efficiency as compared to OTFS-OMA. On the other hand, in detection in downlink OTFS-NOMA systems is introduced.
comparison with OFDM-NOMA, outage spectral efficiency The proposed approach exploits the least-squares with QR
of the introduced OTFS-NOMA approach is enhanced at factorization algorithm for equalization and a novel reliability
the expense of a reduction in average spectral efficiency. zone detection technique to detect the users’ symbols.
The authors of [128] present a coded uplink OTFS-NOMA In addition, multiuser interference is eliminated in each
architecture where sub-vector resources of Doppler and step by interference cancellation. The presented method
delay dimensions are respectively allocated to the stationary is numerically shown to outperform the existing MMSE
and mobile users. An iterative SIC receiver is devised for SIC benchmark scheme in terms of symbol error rate
multiuser detection and decoding under intensive co-channel performance. An uplink heterogeneous NOMA system is
interference. The introduced iterative SIC turbo receiver investigated in [136] where an iterative SIC-based detection
is demonstrated to accomplish performance enhancement algorithm is implemented with realizable rectangular pulses
over other relevant receivers and to attain resilience against and without the assumption of ideal SIC. Users are divided
imperfect SIC and CSI uncertainty. By utilizing LEO satellite into two groups where OFDM and OTFS are adopted by
communications, grant-free NOMA is integrated with OTFS the low-mobility and high-mobility user groups, respectively.
in [129], [130] to enable massive IoT connectivity in a An extrinsic information transfer chart analysis is used
terrestrial-satellite link scenario. The authors introduce a two- to confirm that the introduced algorithm manifests fast
step successive active terminal identification and channel convergence and provides the same performance as the
estimation algorithm along with a low-complexity multiuser detection with ideal SIC for sufficiently large SNR values. As
signal detection technique. A number of numerical results are a low-complexity extension of [83], the work in [137] relies
provided to show the superiority of the proposed approaches on MIMO-OTFS for the grant-free random access in LEO
as compared to the other related schemes in terms of satellite communications with large differential delay and
error probability performance and channel estimation quality. Doppler shift. In order to estimate the channel and to detect
In [83], the authors study the grant-free random access active devices, a deep learning-based generalized approxi-
in LEO satellite communications by adopting a massive mate MP algorithm with cross-correlation-based Gaussian
MIMO-OTFS scenario where multiple IoT devices wish to prior is utilized. In addition, the EM rule is used to learn
communicate with a LEO satellite in a sporadic fashion. the corresponding hyperparameters. Simulation results are
The input-output relationship is initially analyzed. Then, a provided to show that the presented scheme outperforms
novel method to estimate the channel and to detect active other related works in terms of channel estimation quality
devices is introduced by relying on the channel sparsity in the and device activity detection performance. In [138], a
DD-angle domain. Numerical results are used to verify that downlink OTFS-NOMA system is discussed by assuming
the proposed technique can obtain precise CSI and device the existence of both high-mobility and low-mobility user
activity status. A two-user downlink OTFS-NOMA system profiles. The authors focus on a resource allocation and user
is addressed in [131] where a slow-mobile user with a better pairing problem to minimize average BER. The resulting
channel condition than a fast-mobile user performs SIC. mixed integer programming problem is initially solved for a
This process is carried out as follows: the slow-mobile user case with two users and the solution is then generalized for
initially decodes the fast-mobile user’s signal in DD domain the general scenario. The presented scheme is demonstrated
and cancels it from the aggregate signal. Subsequently, the to perform similarly to the optimal exhaustive approach. An
slow-mobile user decodes its own signal in TF domain. The OTFS-NOMA downlink scheme is investigated in [139] for
fast-mobile user, on the other hand, directly decodes its own coordinated direct and relay transmission system. Here, a
signal in DD domain. The presented system is asserted to source directly transmits to a close-by mobile terminal and
provide advantages in terms of overall latency and spectral sends information to a remote terminal with high-mobility
efficiency. In [132], OTFS is combined with filter bank multi- by getting assistance from the nearby terminal as a relay.
carrier and NOMA techniques to attain massive connectivity A number of closed-form expressions are derived on the
and larger spectral efficiency over a mmWave channel. outage probability and on the outage sum-rate assuming
Novel user-pairing techniques are proposed by assuming a both perfect and imperfect SIC. Numerical results confirm
downlink scenario where a BS with massive number of the performance advantage of the introduced approach. A
antennas communicates with multiple single-antenna users. downlink OTFS-NOMA scenario is inspected in [140] where
The introduced scheme is shown to outperform similar a BS communicates with multiple high-mobility users. The
related works in terms of error probability performance. K-means clustering algorithm is implemented to classify
An OTFS-NOMA assisted cooperative multicast-unicast the users into a number of distinct clusters while power-
streaming scenario is covered in [133] where the authors domain NOMA is applied on the cluster centers. In [141],

VOLUME 5, 2024 4505

ALDABABSA et al.: SURVEY ON OTFS MODULATION

a novel OTFS-NOMA-assisted downlink ISAC method is idea is shown to provide BER performance enhancement
presented where an UAV acts as an air BS to support multiple under jamming as compared to conventional OTFS-SCMA
users. Initially, the UAV acquires the users’ position and systems. In [149], it is stated that the performance of con-
velocity information by utilizing the signal echo spread in ventional OTFS-based multiple access approaches heavily
the LoS channel and a 3D motion prediction topology is depends on the characteristics of the channel spreading
then executed to direct the NOMA transmission towards the function. In this sense, the traditional schemes with a limited
multiple users. The proposed scheme is shown to accomplish spectrum resource and non-sparse non-compact channel
superior achievable spectral efficiency as compared to the spreading function suffer from a lack of capacity. The authors
other related approaches. correspondingly propose a grant-free multiple access method
A multiuser OTFS system relying on a code-domain called orthogonal time frequency code space modulation
NOMA method called sparse code multiple access (SCMA) by exploiting orthogonal code domain resources. A number
is proposed in [142] for both downlink and uplink cases of simulation results are illustrated to validate the superi-
by judiciously positioning SCMA codewords in DD plane. ority of the introduced technique in terms of false alarm
The authors present proper detection techniques for both probability and BER performances. The performance of the
scenarios and demonstrate that the introduced OTFS-SCMA preceding technique is evaluated for joint grant-free access
scheme attains better BER performance as compared to and positioning of massive terminals with high-mobility
the combinations of OTFS with OMA and power-domain over Internet of Vehicles networks in [150]. The numerical
NOMA. In [143], the authors offer a two-dimensional down- evaluations are performed for positioning error and BER
link OTFS-NOMA approach where power-domain NOMA of terminals under distinct scenarios. The proposed scheme
is adopted between one high-mobility user and multiple is shown to outperform the conventional OFDM systems
low-mobility users. In addition, SCMA is implemented for with orthogonal spreading. An uplink MIMO-OTFS SCMA
the users under the subgroup of low-mobility. The superior system is investigated in [151] by introducing a memory
spectral efficiency performance of the presented scheme as AMP detector. The proposed approach is shown to bring
compared to the other existing power-domain NOMA and about implementation advantage in terms of computational
OTFS combinations is verified by means of simulations. complexity with desired BER performance. In [152], an
The work in [144] presents a channel estimation method IRS-assisted OTFS-based uplink SCMA communications
for the uplink OTFS-SCMA scenario introduced in [142]. system is studied. Using an upper bound on the word error
Exploiting the convolutional sparse coding approach, a probability, the authors develop two IRS phase shift design
minimal overhead equivalent to that of a single-user case algorithms based on the semidefinite relaxation and gradient
is accomplished without compromising on the estimation ascent approaches. The performance of the presented scheme
error. The efficiency of the presented method is demonstrated is inspected in terms of diversity and SNR gains.
via simulation results in terms of BER, normalized MSE, In [153], tandem spreading multiple access (TSMA),
and spectral efficiency. The works in [145], [146] consider which is a novel grant-free code-domain NOMA approach
a downlink OTFS-SCMA scheme and introduce a cross- to yield a trade-off between the reliability and the user data
domain receiver architecture where OTFS symbol estimation rate [154], is combined with OTFS for smart high-speed
and SCMA decoding operations are performed in a joint railway massive IoT networks. Simulation results confirm
fashion by iteratively passing the extrinsic information that desired levels of high user connectivity and transmission
between time and DD domains. It is demonstrated that the reliability can be accomplished by the presented scheme.
detector structure in [142] exhibits poor BER performance OTFS-TSMA approach in [153] is extended for a LEO
due to the fact that fractional Doppler shifts result in inferior satellite system based on differential Doppler shift in [155].
channel equalization causing performance deterioration in It is demonstrated that the presented scheme brings about
the subsequent SCMA decoding. The proposed method enhanced connectivity as well as a reduction in the system
in [145], [146] is shown to outperform the approach in [142] resource consumption. In order to satisfy the low-latency,
in terms of BER performance by bypassing the preceding high-reliability, and massive connection requirements of 6G
effect. An uplink coordinated multipoint coverage case is networks in high-mobility environments, a novel OTFS-
addressed in [147] where numerous mobile user terminals based pattern division multiple access (PDMA) method is
are congregated for SCMA. These users simultaneously get introduced in [156]. As a variant of NOMA, PDMA enables
service from remote radio heads before and behind them multiple number of users to commonly use orthogonal
by exploiting OTFS. The authors offer practical receiver resources (time, frequency, space, DD domain resources) in
algorithms for centralized and decentralized detectors to a non-orthogonal fashion by a joint design of transmitter
reap the inherent channel diversity. The presented detector and receiver. Relying on the EP algorithm, the authors
structures are demonstrated to yield effective and robust propose a combined detection scheme for the OTFS-PDMA
performance under CSI uncertainty. A simple yet functional system. Simulation results indicate that the presented method
resource hopping technique over delay or Doppler axis is outperforms the traditional OTFS-OMA approaches.
proposed in [148] to suppress interference due to common The study in [157] considers an uplink OTFS scenario
jamming signals in OTFS-SCMA systems. The introduced and presents a low-complexity detection method based on

4506 VOLUME 5, 2024

decomposing the original large channel into parallel small for high-mobility scenarios, the authors consider sub-6-GHz
sub-channels. The decomposition process is applied initially and millimeter-wave bands for satellite-to-ground channels
in space domain and then in time domain. The introduced with high-mobility [161]. Both LoS and non-LoS conditions
scheme requires partial CSI feedbacks in terms of delay and are investigated. To improve the error performance, MMSE
Doppler shifts from the receiver to the transmitters and is with successive detection is used instead of solely MMSE.
shown to provide enhanced BER performance with lower The authors in [164] investigate the potential performance
complexity as compared to the existing related techniques. improvement in a dual satellite transmission system. The
In [158], a new uplink multiple access method is presented flexibility introduced by the OTFS is used to counteract
where each user forms a coded QAM-modulated OFDM the residual frequency offset from the varying inter-satellite
signal and then transmits it after upsampling and circular distances. They make use of precoding the symbols of the
shifting operations, respectively, over a doubly-dispersive distant satellite to compensate for differential delays and
channel. It is proven that the aggregate received signal at frequency shifts with respect to the closest satellite. Hence
the BS has the same form as the received signal in an the proposed scheme allows the ground-based receiver to
equivalent single-user OTFS system. In addition, the authors lock on to a single satellite.
propose multiuser SIC and turbo receivers to obtain diversity The extension of the OTFS to MIMO scenarios for LEO
gain. The superior performance of the proposed scheme is satellites is proposed in [91], where the authors investigate
numerically shown in comparison with the corresponding the performance of the system in the presence of spatially
multiuser OFDM system. The authors of [159] inspect an correlated channels. Whitening transformations are proposed
OTFS-based uplink multiuser MIMO system by embracing at the transmitter and the receiver to combat the spatial
practical rectangular pulses along with fractional delay and correlation and to enhance the error performance. In [165],
Doppler shifts. In the considered setting, multiple single- the authors make use of the properties of the OTFS waveform
antenna users desire to simultaneously communicate with a to jointly estimate the DD channel vector and detect the
multiantenna receiver under ML and nearest-neighbor decod- transmitted data. Unknown symbols in an OTFS frame are
ing rules. It is demonstrated that the severe performance used as ‘virtual pilots’ with the goal of improving the
degradation in terms of PEP is induced with the adoption estimation performance. The use of IRS is also considered
of the rectangular pulses. in space-air-ground integrated networks in [166], where the
authors proposed to create “virtual” Doppler frequencies to
V. APPLICATIONS OF OTFS IN SATELLITE facilitate the IRS configuration. It is also shown that the
COMMUNICATIONS proposed IRS configuration helps overcome the error floors.
In recent years, the space networks have been evolving rather Another line of research focused on the use of OTFS
rapidly, with the deployments of LEO mega-constellations between satellite-to-ground channels for IoT networks.
and lunar missions [160]. These networks introduce a Instead of classical grant-free random access schemes (that
number of unknowns regarding channel characteristics, such are not suitable for rapidly varying channel conditions), the
as the ambient noise conditions. In addition, the large authors in [129] proposed to use NOMA-OTFS. The inves-
delays associated with long propagation distances and high tigated methodology offers multi-functionality by jointly
Doppler shifts due to the high speeds of LEO satellites performing active user detection and channel estimation
(5-10 km/s [161]) render space networks, in particular LEO based on the proposed OTFS data frame structure. The
mega-constellations, very suitable for the use of OTFS authors in [83] extended the work to MIMO settings.
modulation. The unique features of OTFS make it a promis- To exploit the sparsity of the channel, a sparse Bayesian
ing waveform technology for satellite communications, learning-based channel estimation approach is presented.
providing robustness to time-varying channels, multipath A novel grant-free random access scheme, TSMA is con-
fading, and high spectral efficiency with low computational sidered for OTFS-based satellite communication systems
complexity. This synergy is well-elaborated in the recent in [155]. The authors also consider smart railways as the
work [162], where the authors also use a case study to application scenario where the ground stations can also be
demonstrate the potential performance improvement that can high-mobility nodes. They propose a differential Doppler
be provided by OTFS. In [163], the authors also highlight shift-based TSMA-OTFS transmission scenario to enable
the Doppler mitigation capabilities of the OTFS modulation. service continuity and to control multiuser interference. The
The use of OTFS in varying mobility scenarios is first extended application of the grant-free access to scenarios
discussed in [124]. Inspired by the superior performance where the differential delay is more than one symbol duration
of OTFS in doubly dispersive channels, the authors use jointly with the differential Doppler shift exceeding one
OTFS with NOMA to jointly harvest multipath and multiuser subcarrier spacing is considered in [137]. In this work,
diversities in the presence of heterogeneous mobility pro- the authors estimate the channel in the delay dimension
files. The benefits of the proposed approach in terms and decompose the 3D tensor for parallel computation. A
of increased spectral efficiency and reduced latency are DL-based generalized approximate MP is proposed, and
demonstrated numerically for both high-mobility and low- the EM algorithm is used to learn the hyperparameters.
mobility users. Motivated by the suitability of the OTFS The combination of grant-free NOMA and OTFS is also

VOLUME 5, 2024 4507

ALDABABSA et al.: SURVEY ON OTFS MODULATION

TABLE 7. A summary of OTFS applications in satellite communications.

studied in [167] where the authors compare different con-
figurations, including the traditional frame structures, virtual
oversampling lattices, and a proposed training sequence to
demonstrate various configurations that provide a wide range
of flexibility according to the target use case. Finally in [130]
a two-stage successive active terminal identification and
channel estimation is proposed for multiuser transmission
that exploits the sparsity of satellite-to-ground channels.
A least-squares-based parallel detection technique is also
proposed to alleviate the detection complexity.
The performance of OTFS in satellite systems is also
considered in [48], where the authors derive the closed-
form expression for the OP. In addition, UAV cooperation is
considered for increased performance. In [168], the authors
investigate a mobile multi-UAV cluster that facilitates the
communication between the satellite and the ground terminal
by compensating the path loss and improving the coverage.
The system OP is presented based on the analysis using
stochastic geometry. The physical layer security performance
of the OTFS-based satellite system is also investigated
in [52]. The authors derive the secrecy OP and compare
the performance of OTFS and OFDM. A cooperative-
jamming UAV node is also considered to improve the secrecy
performance.
Overall, the current literature solidifies the use of OTFS for
satellite communication systems (See Table 7). Their use for
inter-satellite communications and communication systems
between spacecrafts remains an open issue.

VI. OTFS-ISAC
International Telecommunications Union (ITU) has clearly
stated the expectation about the integration of sensing and
communication systems to become a part of the IMT
2030 [169]. In line with this expectation, the recent literature
has also concentrated on identifying OTFS’s advantages
for ISAC systems [10]. The preliminary work OTFS-
ISAC is [170], which focuses on the vehicular network constant false alarm rate target detection module, targeting to
applications. The authors have shown that the slow time- significantly enhance the detection performance and accuracy
varying delay-Doppler domain channel coefficients provide in target range and velocity estimation, as demonstrated by
a significant advantage in accurately predicting the channel simulation results. Another recent work exploring a user
parameters. Following this, in [171], the authors propose state refinement algorithm within OTFS-ISAC systems par-
a spatially spread OTFS to obtain an angular domain ticularly suitable for high-mobility networks where reliable
discretization. A beam tracking, angle estimation, and power data transmission is critical is [174], whereby leveraging a
allocation scheme is proposed for radar sensing by making hybrid digital-analog architecture and a nested array-based
use of the effective radar sensing matrix. In [172], the technique, the authors propose an algorithm that refines
authors propose a matched filter-Fibonacci (MF-F) algorithm initial coarse angle estimates, subsequently enhancing joint
to improve the estimation of sensing parameters in an delay and Doppler shift estimations, as validated by numer-
ISAC OTFS system. The MF-F method is verified through ical results, thus improving both radar and communication
a software-defined radio implementation, and the authors performance metrics.
show high precision and robustness in estimating parameters, [175] proposes the amplitude barycenter calibration (ABC)
potentially enhancing the efficiency and speed of the sensing algorithm to address the challenge of fractional delay and
and communication system. Reference [173] introduces a Doppler effects in OTFS-ISAC systems, which may result
framework for OTFS-ISAC systems targeting high-mobility in estimation errors due to non-integer multiples in channel
environments while maintaining high spectral efficiency. resolutions. The proposed algorithm utilizes integer DD
The proposed framework consists of sequential coarse and grid observations to calibrate estimates effectively, and its
fine-tuning spectrum matching modules coupled with a performance analysis, including error estimation and the

4508 VOLUME 5, 2024

Cramer–Rao lower bound, shows significant improvements algorithm with alternating iterative minimization effectively
in sensing resolution for distance and velocity while main- reduces WISL and interference, enhancing data rates and
taining robust communication performance. In [176], the detection performance as demonstrated by simulation results.
authors offer a high-order spectra barycenter calibration
algorithm by targeting some of the shortcomings in sensing VII. OTHER APPLICATIONS AND SCENARIOS
resolution caused by limited bandwidth and frame time. A. UWA
The introduced technique outperforms the ABC approach Due to the low speed of acoustic waves and wideband
by providing higher sensing accuracy even at low pilot characteristics, UWA communications undergo large time
SNR values. A method for joint radar sensing and com- delay, Doppler shift and scale effects such that UWA
munications using OTFS modulation is proposed in [177] communications is much more challenging than RF wireless
precisely for detecting moving targets during digital wireless communications. Since the UWA channels are both time-
transmissions. The technique employs random pilot signals selective and frequency-selective, i.e., doubly dispersive,
for channel impulse response estimation, which proves OFDM fails with ICI and thus OTFS being Doppler resilient
superior in identifying vehicle velocities in noisy conditions and converting worse TF domain channels into better DD
compared to the traditional methods. Simulations and field domain channels emerges as a good alternative. OTFS based
tests validate the effectiveness of this OTFS-based sensing UWA system has first been proposed in [182]. The authors
approach, demonstrating its advantages over the conventional showed by simulations that OTFS outperforms OFDM in
cross-ambiguity function radar processing in terms of target terms of BER and spectral efficiency in different UWA
detection accuracy. Reference [141] introduces a network channel conditions. In [183], OTFS is shown to provide
design that integrates NOMA for OTFS-ISAC, utilizing better performance than OFDM when simulated in the
UAVs as air BSs to serve multiple users. The system Bellhop-based statistical UWA channel model for AWGN,
leverages ISAC capabilities to extract user positional and complex noise, and real channel noise. References [184]
velocity data from echo signals, optimizing power allocation and [185] also show the superiority of OTFS over OFDM
for enhanced data rates, and incorporates a 3D motion with channel estimation errors and IM, respectively. To
prediction topology for effective NOMA transmission. The handle time scaling effect of UWA channels, OTFS-like
authors explore the enhancement of uplink transmission in a orthogonal delay scale space modulation, outperforming
vehicular network using OTFS-ISAC in [178], focusing on OTFS and OFDM is developed with a low complexity
the challenge of associating sensing parameters with trans- receiver based on subcarrier-by-subcarrier equalization [186].
mitters to improve communication. A scheme that jointly In [187], a two-step Doppler compensation to alleviate
handles parameter association, channel estimation, and signal Doppler effects and a practical channel estimation method to
detection is proposed as a constrained bilinear recovery avoid mathematical problem of fractional Doppler shifts are
problem, utilizing a bilinear unitary approximate mes- proposed. The simulation results obtained in different UWA
sage passing (Bilinear-UAMP) algorithm. Reference [179] channel environments verify the performance improvement
introduces an OTFS-ISAC system that provides precise as compared to OFDM. Reference [188] proposed a learned
range-velocity profiles without requiring large bandwidths denoising based sparse adaptive channel estimation method
or extended transmission durations. The proposed system enhancing the estimation performance significantly. In [189],
uses a single OTFS carrier with rectangular pulse shaping a structured sparsity-based generalized AMP channel esti-
to estimate delays and Doppler shifts, thus determining mation algorithm, providing lower computational complexity
range and velocity, while an innovative algorithm leverages and superior performance is developed for quantized OTFS
the sidelobes of the pilot signal’s physical pulse shape to systems. Several detection techniques such as DNN based
detect radar targets beyond conventional resolution limits. signal detection [190], passive time reversal [191], LMMSE
MIMO extension of OTFS-ISAC is considered in [180] for SIC detection [192] and unitary AMP based detection [193]
the first time. The authors present a generalized likelihood are also studied in OTFS UWA communications.
ratio test based multi-target detection algorithm and delay-
Doppler-angle estimation for MIMO-OTFS radar, which B. UAV
effectively handles ISI and ICI to enhance target detection. Since UAVs are flexible in terms of maneuverability, altitude,
Additionally, they propose adaptive transmission strategies deployment and of low cost, and provide LOS connectivity
for OTFS ISAC in discovery and track modes, utilizing DD to ground users and establish ubiquitous and seamless con-
multiplexing for wide-ranging environmental probing and nectivity of different communication devices and networks,
directional transmissions for optimal radar and communica- UAV communications will be an indispensable part of
tion performance. Finally, in [181], the authors present a joint 6G aiming to unify all terrestrial, non-terrestrial, RF and
design approach for an OTFS signal and a receiving filter, free space optical (FSO) communications. However, UAV
formulating the optimization problem as a minimization applications require high mobility causing severe Doppler
of the weighted integrated sidelobe level (WISL) and effects. Thus, OTFS appears to provide Doppler resilience in
interference from the ISAC waveform, while maintain- such non-terrestrial UAV communications. For such Doppler
ing SNR targets. The proposed majorization-minimization compensation, OTFS-based UAV communication system

VOLUME 5, 2024 4509

ALDABABSA et al.: SURVEY ON OTFS MODULATION

has been proposed in [194]. The authors optimized UAV DNN. It is numerically demonstrated that as compared to
trajectory by minimizing the energy consumption under the the conventional detectors, the devised scheme has a better
constraints of BER and transmission rate, and showed the error performance when fractional Doppler effect exists. A
superiority of the system over OFDM based UAV system. neural network-based structure is examined in [203] for
Reference [195] proposes an emergency localization tech- OTFS equalization problem. It is shown that ISI can be
nique for UAV equipped with cellular BS to establish OTFS decreased noticeably by exploiting a reservoir computing-
modulated physical random access channel (PRACH) trans- based neural network architecture with a particularly tailored
mission and reception scheme performing time-of-arrival training dataset. In [204], a Gaussian AMP based signal
measurements. The simulation results show that the proposed detection approach is designed for OTFS by expanding the
solution outperforms standard PRACH-based localization standard Gaussian AMP into a corresponding DNN structure.
techniques. In [196], an iterative algorithm based on belief The simulation results affirm that the presented model-driven
propagation and gradient descent optimization for localizing detection scheme performs better than the standard Gaussian
a swarm of UAVs communicating with each other and an AMP by around 2 dB at a BER of 10−4 . The authors
edge server has been presented by using time difference of [205] inspect an EP aided model-driven DL architecture
of arrival measurements obtained through OTFS-modulated for OTFS signal detection. In order to increase the speed of
signals. The proposed algorithm increases the localization the convergence and to enhance the detection performance,
accuracy of the limited bandwidth systems. Reference [197] every iteration is unfolded into a layer-wise network through
and also aforementioned works [48], [52], [168] deal with embedding trainable parameters which can be optimized
OTFS performance of satellite communication systems by offline training. Numerical results confirm that the
including UAV cooperation. introduced scheme can indeed enhance the signal detection
performance with similar computational complexity in com-
C. DL parison with traditional EP approaches. The work in [206]
Deep/machine learning (DL/ML) algorithms have been considers a DL-based signal detector approach for RIS-
taking their roles in everyday daily routine with a plethora aided OTFS systems. It is shown that the studied DL-based
of diverse application scenarios. Their popularity can be detector can achieve infinitesimally low BER at a lower
credited to their ability in performing precise predictions SNR as compared to the traditional detectors. A DNN based
in an iteratively improving fashion by learning from new fractional Doppler channel estimation method is conceived
data. Stimulated by their advantages, DL/ML techniques in [207] for an air-to-ground OTFS communication system.
are currently being inspected for physical-layer of wireless It is demonstrated that as compared to the conventional
communications as well. To that end, there exists a con- approach, a similar performance can be accomplished by
tinuously increasing interest on the combination of DL/ML the proposed technique with a lower pilot power. In [208],
and OTFS schemes in order to collectively reap the benefits an OTFS-enabled URLLC scenario is inspected and a
of both approaches. In [198], a damped generalized AMP predictive precoder scheme with an unsupervised learning
algorithm is proposed to enhance the detection performance based DL mechanism is recommended with the purpose of
of OTFS systems with reduced complexity where the enhancing the system reliability performance in terms of
damping factors are optimized relying on DL techniques. In frame error rate. The performance of a DL-based OTFS
order to alleviate the PAPR problem in OTFS, an autoencoder system is investigated in [209] under the presence of practical
based DL scheme is introduced in [199]. Using a two- hardware impairments such as in-phase and quadrature-
stage training procedure, the presented technique reduces the phase component mismatch and DC offset. By applying
PAPR considerably by yielding a reasonable compromise data augmentation, the adopted DL-OTFS model is trained
between BER and PAPR. The authors of [200] treat an in an offline manner and subsequently implemented for
autoencoder-based OTFS system within a DL framework. online signal detection in the DD domain. The simulation
Considering a feed-forward neural network architecture for results verify that the presented approach can attain enhanced
both encoding and decoding, it is shown that the detection BER performance in comparison with the traditional MP-
performance of the recommended method outperforms the and MMSE-based receiver structures. In [210], a graph
conventional OTFS with the MP-based detection at high neural network detection architecture is designed for OTFS
SNR levels. In [201], a two-dimensional convolutional systems. The authors utilize the Markov random field theory
neural network based detection is studied for single-antenna to effectively extract the hidden features of input data and
OTFS systems. It is revealed by simulation results that to obtain more precise detection. An automatic modulation
the suggested technique brings about an enhanced BER classification technique is developed for OTFS in [211]
performance over the traditional MP detector with a very low by utilizing a hybrid convolutional neural network and
time complexity. As an extension of [201] for multiantenna long short-term memory network with a residual stack.
scenarios, a low-complexity detection approach based on The numerical results show that the introduced approach
two-dimensional convolutional neural network is presented accomplishes a classification precision over 98.5% for an
for MIMO-OTFS in [202] by only taking into account the SNR value of 20 dB and performs adequately at low SNR
received signal and an augmented dataset as input to the levels as well. In order to enhance the detection performance

4510 VOLUME 5, 2024

in the presence of channel estimation errors for OTFS index of OTFS block in in-phase and quadrature dimensions.
systems, the study in [212] investigates the performance The presented scheme is shown to perform better than the
of a model-driven DL based orthogonal AMP detection other related OTFS-IM techniques in terms of both BER and
architecture. It is revealed by the simulation results that PAPR. In [219], it is articulated that the studies in [214]
with much reduced complexity and much less parameters, and [216] only activate independent delay-Doppler resources
the presented detector can attain better error performance by adopting impractical bi-orthogonal OTFS pulses. Thus,
under channel estimation errors as compared to other related they cannot perform well under ISI caused by the channel
nonlinear detectors. In [213], a learning-based channel delay and Doppler shifts. As a solution, the authors of [219]
estimation scheme for fractional DD domain channel for offer two new block-wise OTFS-IM designs under a practical
OTFS systems is proposed where the learning process is scenario by activating a block of delay/Doppler resource
implemented in TF domain instead of DD domain. This bins at the same time. Simulation results affirm the BER
is driven by the fact that the range of values in the TF advantage of the investigated approaches against the other
channel matrix is more favorable for the training purposes related techniques. A physical layer anti-eavesdropping idea
as compared to the DD channel matrix values which may is inspected in [220] for OTFS-IM systems under frequency-
have large variance values and thus may not be suitable division duplexing mode. As no channel reciprocity exists
for the training process. Numerical results reveal that with under such a scenario, a secure mapping technique is
a significantly lower complexity, the introduced idea can crafted relying on a chaos sequence generated by utilizing
acquire almost the same performance as that of the best the angular reciprocity of the legitimate link. Numerical
performing scheme available in the literature. findings reveal that the eavesdropper’s BER can be increased
significantly without seriously affecting the BER of the
D. IM legitimate user. The authors in [221] indicate that the existing
Index modulation with its various forms enables an inno- OFDM-IM methods implement a fixed-sized constellation
vative low-complexity practice for the transmission of on a number of activated grids in DD domain to avoid
extra data bits by means of the indices of the available ambiguous detection. They accordingly offer a new OTFS-
transmit resource blocks. With the purpose of attaining IM technique where the number of active grids transmitting
enhanced spectral efficiency over high-mobility channels, modulation symbols from any cardinality of constellations is
the integration of IM into OTFS systems is addressed in permitted to be altered. The ambiguous detection is prevented
several works. As one of the pioneering studies on the by transmitting a fixed length of information bits via joint-
application of IM to OTFS, the work in [214] aims to mapping. Simulation results show that the introduced idea
enhance BER performance while guaranteeing a certain enhances the BER performance of the standard OTFS-
spectrum efficiency level over doubly-dispersive channels. In IM scheme. In [222], a block-wise OTFS-IM method is
addition to the constellation symbols, a number of extra bits designed where, unlike [219], an OTFS frame can include
are transmitted via indices bits on DD domain. Simulation subblocks that transmit information symbols in delay or
results verify that the introduced OTFS-IM idea leads to Doppler bins at the same time and active resource blocks may
superior BER performance as compared to the conventional implement different constellations. The presented approach
OTFS approach. It is stated in [215] that OTFS-IM signal is numerically demonstrated to outperform other related
of [214] activates all the subcarriers when converted into TF schemes in terms of BER performance. Two new detectors
domain. This, on the other hand, may result in an escalated based on MMSE and vector-by-vector-aided MP approaches
BER especially with large numbers of active lattices. As a are introduced in [223] for OTFS-IM systems and are shown
remedy, the authors of [215] offer a new approach called to bring about BER performance gain as compared to the
OTFS with improved IM where every index bit is transmitted standard OTFS-IM scheme. In [224], the performance of
two times. Numerical results confirm the performance gain OTFS-IM is inspected in a two-hop decode-and-forward
in terms of BER as compared to the standard OTFS-IM. relay communication system with a single relay where each
In [216], OTFS with dual-mode IM is introduced to attain node has a single antenna. It is shown that the utilization
an enhanced trade-off between the spectral efficiency and of indexing enhances the performance of OTFS with decode
BER. Relying on the minimum Hamming distance criterion, and forward relaying and the diversity order of the system
the authors devise a modified log-likelihood ratio detector becomes equal to the minimum of the number of resolvable
with low complexity, which is shown to perform almost DD domain paths between the source-to-relay and the relay-
identically as the ML approach. A symbol-by-symbol aided to-destination links with a PR operation at the transmitter.
EP detection procedure is proposed in [217] for OTFS- In [225], it is shown that the introduction of spatial
IM. It is numerically demonstrated that the introduced idea modulation results in an increase in the overall spectral
can yield important performance gains beyond the other efficiency of MIMO-OTFS by the base-two logarithm
detection approaches such as MMSE [214] for OTFS-IM. of the number of transmit antennas. Similarly, a spatial
A novel transmission method, called OTFS with in-phase multiplexing aided OTFS scheme with IM is devised
and quadrature IM, is conceived in [218] where a number in [226] by adopting vertical Bell Labs layered space-time
of extra information bits are conveyed by means of the grid approach for simultaneous and parallel transmission over

VOLUME 5, 2024 4511

ALDABABSA et al.: SURVEY ON OTFS MODULATION

transmit antennas. It is shown that the introduced method spectral efficiency and decreasing BER, an OTFS-NOMA
can provide an important SNR gain as compared to its scheme with IM is devised in [237] for a two-user downlink
counterpart in [225] under the same data rate with lower scenario over a doubly-selective channel. Numerical findings
computational complexity. The authors of [227] treat a indicate that the introduced system outperforms other related
MIMO-OTFS scheme with spatial modulation and provide OTFS-based approaches with IM.
a number of closed-form expressions on the error rate of
the system by means of union bounding together with the VIII. FUTURE CHALLENGES
moment-generating function approach. A space-time block In this section, we present several research directions of
coding aided MIMO-OTFS system with spatial modulation interest to the research community working on OTFS.
is investigated in [228]. The proposed scheme is demon- • It is conceivable that the prospective communica-
strated to outperform the corresponding similar techniques in tion systems may include transceiver nodes with
terms of BER performance. A combination of spatial-index extremely high velocities. Under such cases, Doppler
modulation and OTFS systems is inspected in [229], [230]. shift levels that are unprecedentedly large for the
The authors suggest to use a three dimensional IM approach contemporary wireless communication systems will be
including the transmit antenna index in the spatial domain experienced and the channel parameters associated with
along with the delay and Doppler indices in the DD domain. the multipath propagation such as delay, Doppler, and
The presented scheme is shown to yield superior BER complex gain will alter at an excessive rate. When
performance as compared to the traditional approaches. not dealt with appropriately, this can cause serious
In [231], a generalized spatial modulation combined MIMO- performance degradation in terms of BER. Hence, when
OTFS system is proposed together with a decision feedback the DD domain channel parameters are time-variant
detector. The introduced technique is shown to accomplish within a frame duration, the adaptation of OTFS to such
better BER performance as compared to the conventional a scenario poses a challenging future research direction.
MIMO-OTFS and OFDM-based approaches. The authors • A number of works have recently studied possible
of [232] develop a spatial modulation-aided OTFS technique utilization of the DD-domain signaling for security pur-
for high-Doppler cases. A low-complexity distance-based poses. The efficacy of the traditional physical layer key
ordering subspace check detector is proposed by utilizing a generation schemes strictly depends on the stability of
priori information of the transmit symbol vector. Simulation TF domain channel. Hence, these approaches severely
results are used to reveal the superior BER performance suffer from performance loss under high-mobility (low
of the introduced scheme as compared to the traditional coherence time) cases. Contrarily, under the same
SIMO-OTFS system. A transmission technique named as scenarios, a sparse and relatively stationary channel
generalized space-delay-Doppler index modulated OTFS is characterization can be accomplished in the DD domain
introduced in [233] by taking advantage of multi-domain by leveraging OTFS. Thus, with new perspectives OTFS
resources. Here, additional information bits are conveyed offers, it is quite likely that there will be an increasing
by means of the combined space-delay-Doppler resource interest on OTFS from a security point of view.
units. The adopted idea is demonstrated to outperform • The number of nodes to be connected in IoT networks
the existing spatial-index modulation based and generalized is expected to escalate in an exponential rate in
spatial modulation based OTFS approaches in terms of BER the upcoming years. Hence, efficiency in time and
performance. In [234], a low complexity two-step detector frequency utilization has become particularly crucial in
structure for the scheme in [233] is presented. the emerging communication systems and standards.
In [235], space-time shift keying-aided OTFS-based Non-orthogonal approaches for multiple access allow
multiple access is introduced for a single-cell uplink commu- cost-effective usage of these resources and it is highly
nication scenario. By collectively invoking the DD, space and anticipated that they will be featured in the future
time domain resource blocks, multiple users simultaneously standards. As such, the combination of OTFS and
transmit to a common BS. Here, additional information bits NOMA is being researched currently from distinct
are conveyed by the indices of the active dispersion matrices. viewpoints including DD domain user multiplexing.
The presented scheme is shown to enjoy enhanced diversity In this sense, a comprehensive performance evaluation
level and coding gain. The authors of [236] present an uplink of OTFS-NOMA for any number of users and under
multiple access scheme by relying on a combination of practical impairments over generalized fading channels
OTFS and IM over DD domain. All the users independently constitutes a potential future research topic.
implement IM by activating subgroups of the available DD • Cell-free massive MIMO technology is proven to
resource blocks. Here, it is possible that a certain resource support five- to ten-fold enhancement in 95%-likely
block can be used simultaneously by multiple users. It per-user throughput as compared to the traditional
is demonstrated that in spite of such probable collisions cellular operation [238]. From the network design point
among user signals, the introduced idea provides improved of view, the cell-free massive MIMO idea can sub-
BER performance as compared to other similar multiple stantially change the conventional complex centralized
access schemes with OTFS. With the purpose of augmenting processing approach. In addition, OTFS in a cell-free

4512 VOLUME 5, 2024

 massive MIMO network can respond to the stringent [9] S. Li, W. Yuan, Z. Wei, R. Schober, and G. Caire, “Orthogonal time
requirements in terms of spectral efficiency, coverage, frequency space modulation—Part II: Transceiver designs,” IEEE
Commun. Lett., vol. 27, no. 1, pp. 9–13, Jan. 2023.
connectivity, and mobility. On the other hand, such a [10] W. Yuan, Z. Wei, S. Li, R. Schober, and G. Caire, “Orthogonal time
system must deal with a number of emerging challenges, frequency space modulation—Part III: ISAC and potential applica-
such as downlink channel estimation under frequency- tions,” IEEE Commun. Lett., vol. 27, no. 1, pp. 14–18, Jan. 2023.
[11] L. Xiao, S. Li, Y. Qian, D. Chen, and T. Jiang, “An overview of OTFS
division duplexing mode, which may require excessive for Internet of Things: Concepts, benefits, and challenges,” IEEE Int.
overhead. The sparsity in the DD-domain channel Things J., vol. 9, no. 10, pp. 7596–7618, May 2022.
characterization can be exploited in this case. [12] W. Yuan et al., “New delay Doppler communication paradigm in 6G
era: A survey of orthogonal time frequency space (OTFS),” China
• Machine learning has been attracting intense interest Commun., vol. 20, no. 6, pp. 1–25, Jun. 2023.
and is being utilized in different scenarios with distinct [13] M. Li, W. Liu, J. Lei, “A review on orthogonal time–frequency space
goals. Thus, some very recent studies have investigated modulation: State-of-art, hotspots and challenges,” Comput. Netw.,
vol. 224, Apr. 2023, Art. no. 109597.
the exploitation of machine learning approaches in [14] S. K. Mohammed, R. Hadani, A. Chockalingam, and R. Calderbank,
OTFS systems, especially for channel estimation and “OTFS—A mathematical foundation for communication and radar
sensing in the delay-Doppler domain,” IEEE BITS Inf. Theory Mag.,
data decoding purposes. However, there still exist many vol. 2, no. 2, pp. 36–55, Nov. 2022.
pending issues to be resolved in terms of miscellaneous [15] S. K. Mohammed, R. Hadani, A. Chockalingam, and R. Calderbank,
applications and practical implementations. “OTFS—Predictability in the delay- Doppler domain and its value
to communication and radar sensing,” IEEE BITS Inf. Theory Mag.,
• OTFS is a promising candidate for future ISAC systems, early access, Sep. 27, 2023, doi: 10.1109/MBITS.2023.3319595.
offering a balanced approach to the dual demands of [16] S. Li, W. Yuan, Z. Wei, J. Yuan, B. Bai, and G. Caire, “On the
communication and sensing in dynamic and spectrum- pulse shaping for delay-Doppler communications,” in Proc. IEEE
Glob. Commun. Conf. (GlobeCom), Kuala Lumpur, Malaysia, 2023,
constrained environments. This makes it a suitable pp. 1–6.
choice for emerging applications in autonomous driving, [17] A. Farhang, A. RezazadehReyhani, L. E. Doyle, and B. Farhang-
UAV systems, and other areas requiring robust mobile Boroujeny, “Low complexity modem structure for OFDM-based
orthogonal time frequency space modulation,” IEEE Wireless
communication and precise sensing. Cooperative dis- Commun. Lett., vol. 7, no. 3, pp. 344–347, Jun. 2018.
tributed use of OTFS-ISAC still remains an open issue [18] P. Raviteja, K. T. Phan, Y. Hong, and E. Viterbo, “Interference
that should be addressed in the near future. The synergy cancellation and iterative detection for orthogonal time frequency
space modulation,” IEEE Trans. Wireless Commun., vol. 17, no. 10,
between OTFS and ISAC is expected to be explored pp. 6501–6515, Oct. 2018.
further in the future, as we discuss in Section VI. [19] S. Tiwari, S. S. Das, and V. Rangamgari, “Low complexity
LMMSE receiver for OTFS,” IEEE Commun. Lett., vol. 23, no. 12,
pp. 2205–2209, Dec. 2019.
IX. CONCLUSION [20] H. Qu, G. Liu, L. Zhang, S. Wen, and M. A. Imran, “Low-
complexity symbol detection and interference cancellation for OTFS
In this survey paper, OTFS, a promising solution for system,” IEEE Trans. Commun., vol. 69, no. 3, pp. 1524–1537,
high-mobility communications towards 6G, is reviewed Mar. 2021.
comprehensively. Its advantages over other competing [21] L. Gaudio, M. Kobayashi, G. Caire, and G. Colavolpe, “On the
effectiveness of OTFS for joint radar parameter estimation and
techniques, implementations, and integration with MIMO, communication,” IEEE Trans. Wireless Commun., vol. 19, no. 9,
multiple access techniques, satellite communications, ISAC, pp. 5951–5965, Sep. 2020.
DL, IM, UWA, and UAV are discussed in detail. Future [22] G. D. Surabhi and A. Chockalingam, “Low-complexity linear
equalization for OTFS modulation,” IEEE Commun. Lett., vol. 24,
challenges and open issues are also indicated. no. 2, pp. 330–334, Feb. 2020.
[23] Y. Shan, F. Wang, and Y. Hao, “Orthogonal time frequency space
detection via low-complexity expectation propagation,” IEEE Trans.
REFERENCES Wireless Commun., vol. 21, no. 12, pp. 10887–10901, Dec. 2022.
[1] Z. Zhang et al., “6G wireless networks: Vision, requirements, [24] S. Li, W. Yuan, Z. Wei, and J. Yuan, “Cross domain iterative
architecture, and key technologies,” IEEE Veh. Technol. Mag., detection for orthogonal time frequency space modulation,” IEEE
vol. 14, no. 3, pp. 28–41, Sep. 2019. Trans. Wireless Commun., vol. 21, no. 4, pp. 2227–2242, Apr. 2022.
[2] W. Saad, M. Bennis, and M. Chen, “A vision of 6G wireless [25] Z. Yuan, F. Liu, W. Yuan, Q. Guo, Z. Wang, and J. Yuan,
systems: Applications, trends, technologies, and open research “Iterative detection for orthogonal time frequency space modulation
problems,” IEEE Netw., vol. 34, no. 3, pp. 134–142, May 2020. with unitary approximate message passing,” IEEE Trans. Wireless
[3] I. F. Akyildiz, A. Kak, and S. Nie, “6G and beyond: The Commun., vol. 21, no. 2, pp. 714–725, Feb. 2022.
future of wireless communications systems,” IEEE Access, vol. 8, [26] Z. Yuan, M. Tang, J. Chen, T. Fang, H. Wang, and J. Yuan, “Low
pp. 133995–134030, 2020. complexity parallel symbol detection for OTFS modulation,” IEEE
[4] J. Wu and P. Fan, “A survey on high mobility wireless communica- Trans. Veh. Technol., vol. 72, no. 4, pp. 4904–4918, Apr. 2023.
tions: Challenges, opportunities and solutions,” IEEE Access, vol. 4, [27] H. Li and Q. Yu, “Doubly- iterative spasified MMSE turbo
pp. 450–476, 2016. equalization for OTFS modulation,” IEEE Trans. Commun., vol. 71,
[5] R. Hadani et al., “Orthogonal time frequency space modulation,” in no. 3, pp. 1336–1351, Mar. 2023.
Proc. IEEE Wireless Commun. Netw. Conf. (WCNC), San Francisco, [28] H. Zhang and T. Zhang, “A low-complexity message passing detector
CA, USA, 2017, pp. 1–6. for OTFS modulation with probability clipping,” IEEE Wireless
[6] R. Hadani and A. Monk, “OTFS: A new generation of modulation Commun. Lett., vol. 10, no. 6, pp. 1271–1275, Jun. 2021.
addressing the challenges of 5G,” 2018, arXiv:1802.02623. [29] P. Raviteja, K. T. Phan, and Y. Hong, “Embedded pilot-aided channel
[7] Z. Wei et al., “Orthogonal time-frequency space modulation: A estimation for OTFS in delay–Doppler channels,” IEEE Trans. Veh.
promising next-generation waveform,” IEEE Wireless Commun., Technol., vol. 68, no. 5, pp. 4906–4917, May 2019.
vol. 28, no. 4, pp. 136–144, Aug. 2021. [30] S. S. Das, V. Rangamgari, S. Tiwari, and S. C. Mondal, “Time domain
[8] Z. Wei, S. Li, W. Yuan, R. Schober, and G. Caire, “Orthogonal time channel estimation and equalization of CP-OTFS under multiple
frequency space modulation—Part I: Fundamentals and challenges fractional Dopplers and residual synchronization errors,” IEEE
ahead,” IEEE Commun. Lett., vol. 27, no. 1, pp. 4–8, Jan. 2023. Access, vol. 9, pp. 10561–10576, 2021.

VOLUME 5, 2024 4513

ALDABABSA et al.: SURVEY ON OTFS MODULATION

[31] H. B. Mishra, P. Singh, A. K. Prasad, and R. Budhiraja, [52] J. Hu, J. Shi, S. Ma, and Z. Li, “Secrecy analysis for orthogonal
“OTFS channel estimation and data detection designs with super- time frequency space scheme based uplink LEO satellite communica-
imposed pilots,” IEEE Trans. Wireless Commun., vol. 21, no. 4, tion,” IEEE Wireless Commun. Lett., vol. 10, no. 8, pp. 1623–1627,
pp. 2258–2274, Apr. 2022. Aug. 2021.
[32] W. Yuan, S. Li, Z. Wei, J. Yuan, and D. W. K. Ng, “Data-aided [53] P. Raviteja, E. Viterbo, and Y. Hong, “OTFS Performance on static
channel estimation for OTFS systems with a superimposed pilot and multipath channels,” IEEE Wireless Commun. Lett., vol. 8, no. 3,
data transmission scheme,” IEEE Wireless Commun. Lett., vol. 10, pp. 745–748, Jun. 2019.
no. 9, pp. 1954–1958, Sep. 2021. [54] M. K. Ramachandran and A. Chockalingam, “MIMO-OTFS in high-
[33] S. Srivastava, R. K. Singh, A. K. Jagannatham, and L. Hanzo, Doppler fading channels: Signal detection and channel estimation,” in
“Bayesian learning aided sparse channel estimation for orthogonal Proc. IEEE Glob. Commun. Conf. (GLOBECOM), Abu Dhabi, UAE,
time frequency space modulated systems,” IEEE Trans. Veh. Technol., 2018, pp. 206–212.
vol. 70, no. 8, pp. 8343–8348, Aug. 2021. [55] A. RezazadehReyhani, A. Farhang, M. Ji, R. R. Chen, and
[34] A. Pfadler, T. Szollmann, P. Jung, and S. Stanczak, “Leakage B. Farhang-Boroujeny, “Analysis of discrete-time MIMO OFDM-
suppression in pulse-shaped OTFS delay-Doppler-pilot chan- based orthogonal time frequency space modulation,” in Proc. IEEE
nel estimation,” IEEE Wireless Commun. Lett., vol. 11, no. 6, Int. Conf. Commun. (ICC), 2018, pp. 1–6.
pp. 1181–1185, Jun. 2022. [56] G. D. Surabhi, R. M. Augustine, and A. Chockalingam, “On the
[35] G. D. Surabhi, R. M. Augustine, and A. Chockalingam, “Peak-to- diversity of uncoded OTFS modulation in doubly-dispersive chan-
average power ratio of OTFS modulation,” IEEE Commun. Lett., nels,” IEEE Trans. Wireless Commun., vol. 18, no. 6, pp. 3049–3063,
vol. 23, no. 6, pp. 999–1002, Jun. 2019. Jun. 2019.
[57] R. M. Augustine, G. D. Surabhi, and A. Chockalingam, “Space-time
[36] S. Gao and J. Zheng, “Peak-to-average power ratio reduction in
coded OTFS modulation in high-Doppler channels,” in Proc. IEEE
pilot-embedded OTFS modulation through iterative clipping and
89th Veh. Technol. Conf. (VTC), Kuala Lumpur, Malaysia, 2019,
filtering,” IEEE Commun. Lett., vol. 24, no. 9, pp. 2055–2059, Sep.
pp. 1–6.
2020.
[58] C. An and H.-G. Ryu, “High throughput mobile communication
[37] P. Wei, Y. Xiao, W. Feng, N. Ge, and M. Xiao, “Charactering the based on OTFS system with the delay-Doppler compensa-
peak-to-average power ratio of OTFS signals: A large system analy- tion,” Wireless Pers. Commun., vol. 106, pp. 473–486, May 2019.
sis,” IEEE Trans. Wireless Commun., vol. 21, no. 6, pp. 3705–3720,
[59] V. S. Bhat, S. G. Dayanand, and A. Chockalingam, “Performance
Jun. 2022.
analysis of OTFS modulation with receive antenna selection,” IEEE
[38] P. Raviteja, Y. Hong, E. Viterbo, and E. Biglieri, “Practical pulse- Trans. Veh. Technol., vol. 70, no. 4, pp. 3382–3395, Apr. 2021.
shaping waveforms for reduced-cyclic-prefix OTFS,” IEEE Trans. [60] V. S. Bhat and A. Chockalingam, “Performance analysis of OTFS
Veh. Tech., vol. 68, no. 1, pp. 957–961, Jan. 2019. modulation with transmit antenna selection,” in Proc. IEEE 93rd Veh.
[39] R. Mallaiah and V. V. Mani, “A novel OTFS system based Technol. Conf. (VTC), Helsinki, Finland, 2021, pp. 1–6.
on DFrFT-OFDM,” IEEE Wireless Commun. Lett., vol. 11, no. 6, [61] D. Wang, B. Sun, F. Wang, X. Li, P. Yuan, and D. Jiang, “Transmit
pp. 1156–1160, Jun. 2022. diversity scheme design for rectangular pulse shaping based OTFS,”
[40] Y. Ge, Q. Deng, P. C. Ching, and Z. Ding, “Receiver design for China Commun., vol. 19, no. 3, pp. 116–128, Mar. 2022.
OTFS with a fractionally spaced sampling approach,” IEEE Trans. [62] H. Yin et al., “Cyclic delay-Doppler shift: A simple transmit
Wireless Commun., vol. 20, no. 7, pp. 4072–4086, Jul. 2021. diversity technique for delay-Doppler waveforms in doubly selective
[41] Y. Liu, Y. L. Guan, and G. David González, “Near-optimal BEM channels,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process.
OTFS receiver with low pilot overhead for high-mobility commu- Workshops (ICASSPW), Rhodes Island, Greece, 2023, pp. 1–5.
nications,” IEEE Trans. Commun., vol. 70, no. 5, pp. 3392–3406, [63] S. Li, L. Xiao, Y. Liu, G. Liu, P. Xiao, and T. Jiang, “Performance
May 2022. analysis for orthogonal time frequency space modulation systems
[42] Z. Wei, W. Yuan, S. Li, J. Yuan, and D. W. K. Ng, “Transmitter with generalized waveform,” China Commun., vol. 20, no. 4,
and receiver window designs for orthogonal time-frequency space pp. 57–72, Apr. 2023.
modulation,” IEEE Trans. Commun., vol. 69, no. 4, pp. 2207–2223, [64] R. Chong, M. Mohammadi, H. Q. Ngo, S. L. Cotton, and
Apr. 2021. M. Matthaiou, “On the spectral efficiency of MMSE-based MIMO
[43] C. Jin, Z. Bie, X. Lin, W. Xu, and H. Gao, “A simple two-stage OTFS systems,” in Proc. Int. Symp. Wireless Commun. Syst. (ISWCS),
equalizer for OTFS with rectangular windows,” IEEE Commun. Lett., Hangzhou, China, 2022, pp. 1–6.
vol. 25, no. 4, pp. 1158–1162, Apr. 2021. [65] S. Li, J. Yuan, P. Fitzpatrick, T. Sakurai, and G. Caire, “Delay-
[44] S. E. Zegrar and H. Arslan, “A novel cyclic prefix configuration for Doppler domain Tomlinson-Harashima precoding for OTFS-based
enhanced reliability and spectral efficiency in OTFS systems,” IEEE downlink MU-MIMO transmissions: Linear complexity implementa-
Wireless Commun. Lett., vol. 12, no. 5, pp. 888–892, May 2023. tion and scaling law analysis,” IEEE Trans. Commun., vol. 71, no. 4,
pp. 2153–2169, Apr. 2023.
[45] R. Bomfin, M. Chafii, A. Nimr, and G. Fettweis, “A robust baseband
transceiver design for doubly-dispersive channels,” IEEE Trans. [66] G. D. Surabhi and A. Chockalingam, “Low-complexity linear
Wireless Commun., vol. 20, no. 8, pp. 4781–4796, Aug. 2021. equalization for 2 × 2 MIMO-OTFS signals,” in Proc. IEEE 21st Int.
Workshop Signal Process. Adv. Wireless Commun. (SPAWC), Atlanta,
[46] R. Bomfin, A. Nimr, M. Chafii, and G. Fettweis, “A robust and GA, USA, 2020, pp. 1–5.
low-complexity Walsh-Hadamard modulation for doubly-dispersive
[67] B. C. Pandey, S. K. Mohammed, P. Raviteja, Y. Hong, and E. Viterbo,
channels,” IEEE Commun. Lett., vol. 25, no. 3, pp. 897–901,
“Low complexity precoding and detection in multi-user massive
Mar. 2021.
MIMO OTFS downlink,” IEEE Trans. Veh. Technol., vol. 70, no. 5,
[47] P. Singh, K. Yadav, H. B. Mishra, and R. Budhiraja, “BER analysis pp. 4389–4405, May 2021.
for OTFS zero forcing receiver,” IEEE Trans. Commun., vol. 70, [68] Y. Shan and F. Wang, “Low-complexity and low-overhead receiver
no. 4, pp. 2281–2297, Apr. 2022. for OTFS via large-scale antenna array,” IEEE Trans. Veh. Technol.,
[48] J. Shi, J. Hu, Y. Yue, X. Xue, W. Liang, and Z. Li, “Outage probabil- vol. 70, no. 6, pp. 5703–5718, Jun. 2021.
ity for OTFS based downlink LEO satellite communication,” IEEE [69] B. Cao, Z. Xiang, and P. Ren, “Low complexity transmitter precoding
Trans. Veh. Technol., vol. 71, no. 3, pp. 3355–3360, Mar. 2022. for MU MIMO-OTFS,” Digit. Signal Process., vol. 115, Aug. 2021,
[49] C. Shen, J. Yuan, and H. Lin, “Error performance of rectangu- Art. no. 103083.
lar pulse-shaped OTFS with practical receivers,” IEEE Wireless [70] P. Singh, H. B. Mishra, and R. Budhiraja, “Low-complexity lin-
Commun. Lett., vol. 11, no. 12, pp. 2690–2694, Dec. 2022. ear MIMO-OTFS receivers,” in Proc. IEEE Int. Conf. Commun.
[50] S. Li, J. Yuan, W. Yuan, Z. Wei, B. Bai, and D. W. K. Workshops (ICC Workshops), Montreal, QC, Canada, 2021, pp. 1–6.
Ng, “Performance analysis of coded OTFS systems over high- [71] P. Singh, A. Gupta, H. B. Mishra, and R. Budhiraja, “Low-complexity
mobility channels,” IEEE Trans. Wireless Commun., vol. 20, no. 9, ZF/MMSE MIMO-OTFS receivers for high-speed vehicular commu-
pp. 6033–6048, Sep. 2021. nication,” IEEE Open J. Commun. Soc., vol. 3, pp. 209–227, 2022.
[51] J. Sun, Z. Wang, and Q. Huang, “A closed-form minimum BER [72] P. Singh, S. Tiwari, and R. Budhiraja, “Low-complexity LMMSE
precoder for orthogonal time frequency space systems,” IEEE receiver design for practical-pulse-shaped MIMO-OTFS systems,”
Commun. Lett., vol. 26, no. 8, pp. 1898–1902, Aug. 2022. IEEE Trans. Commun., vol. 70, no. 12, pp. 8383–8399, Dec. 2022.

4514 VOLUME 5, 2024

 [73] A. Naikoti and A. Chockalingam, “Low-complexity delay-Doppler [95] M. Liu, S. Li, Z. Wei, and B. Bai, “Near optimal hybrid digital-analog
symbol DNN for OTFS signal detection,” in Proc. IEEE 93rd Veh. beamforming for point-to-point MIMO-OTFS transmissions,” in
Technol. Conf. (VTC), Helsinki, Finland, 2021, pp. 1–6. Proc. IEEE Wireless Commun. Netw. Conf. (WCNC), 2023, pp. 1–6.
[74] T. Thaj and E. Viterbo, “Low-complexity linear diversity-combining [96] Y. Yan, C. Shan, J. Zhang, and H. Zhao, “Low-complexity OTFS-
detector for MIMO-OTFS,” IEEE Wireless Commun. Lett., vol. 11, based hybrid precoding in mmWave high mobility systems,” Digit.
no. 2, pp. 288–292, Feb. 2022. Signal Process., vol. 126, Jun. 2022, Art. no. 103496.
[75] H. Qu, G. Liu, M. A. Imran, S. Wen, and L. Zhang, “Efficient channel [97] A. S. Bora, K. T. Phan, and Y. Hong, “IRS-assisted high mobility
equalization and symbol detection for MIMO OTFS systems,” IEEE communications using OTFS modulation,” IEEE Wireless Commun.
Trans. Wireless Commun., vol. 21, no. 8, pp. 6672–6686, Aug. 2022. Lett., vol. 12, no. 2, pp. 376–380, Feb. 2023.
[76] M. A. Sheikh, P. Singh, and R. Budhiraja, “Low-complexity MMSE [98] W. Shen, L. Dai, J. An, P. Fan, and R. W. Heath, “Channel estimation
receiver design for massive MIMO OTFS systems,” IEEE Commun. for orthogonal time frequency space (OTFS) massive MIMO,” IEEE
Lett., vol. 26, no. 11, pp. 2759–2763, Nov. 2022. Trans. Signal Process., vol. 67, no. 16, pp. 4204–4217, Aug. 2019.
[77] V. S. Bhat and A. Chockalingam, “Performance analysis of MIMO- [99] Y. Liu, S. Zhang, F. Gao, J. Ma, and X. Wang, “Uplink-aided high
OTFS with decode and forward relaying,” in Proc. IEEE 98th Veh. mobility downlink channel estimation over massive MIMO-OTFS
Technol. Conf. (VTC), 2023, pp. 1–7. system,” IEEE J. Sel. Areas Commun., vol. 38, no. 9, pp. 1994–2009,
[78] V. S. Bhat and A. Chockalingam, “Performance analysis of MIMO- Sep. 2020.
OTFS with selective decode and forward relaying,” in Proc. IEEE [100] X. Wu, S. Ma, and X. Yang, “Tensor-based low-complexity channel
98th Veh. Technol. Conf. (VTC), 2023, pp. 1–6. estimation for mmWave massive MIMO-OTFS systems,” J. Commun.
[79] M. Li, S. Zhang, F. Gao, P. Fan, and O. A. Dobre, “A new path divi- Inf. Netw., vol. 5, no. 3, pp. 324–334, Sep. 2020.
sion multiple access for the massive MIMO-OTFS networks,” IEEE [101] D. Shi et al., “Deterministic pilot design and channel estimation
J. Sel. Areas Commun., vol. 39, no. 4, pp. 903–918, Apr. 2021. for downlink massive MIMO-OTFS systems in presence of the
[80] J. Feng, H. Q. Ngo, M. F. Flanagan, and M. Matthaiou, “Performance fractional Doppler,” IEEE Trans. Wireless Commun., vol. 20, no. 11,
analysis of OTFS-based uplink massive MIMO with ZF receivers,” in pp. 7151–7165, Nov. 2021.
Proc. IEEE Int. Conf. Commun. Workshops (ICC Workshops), [102] G.-R. Wang and Z.-X. Wang, “3D-IPRDSOMP algorithm for channel
Montreal, QC, Canada, 2021, pp. 1–7. estimation in massive MIMO with OTFS modulation,” IEEE Access,
[81] M. Mohammadi, H. Q. Ngo, and M. Matthaiou, “Cell-free massive vol. 11, pp. 94679–94690, 2023.
MIMO meets OTFS modulation,” IEEE Trans. Commun., vol. 70, [103] S. Srivastava, R. K. Singh, A. K. Jagannatham, and L. Hanzo,
no. 11, pp. 7728–7747, Nov. 2022. “Bayesian learning aided simultaneous row and group sparse
[82] M. Mohammadi, H. Q. Ngo, and M. Matthaiou, “Cell-free channel estimation in orthogonal time frequency space modulated
massive MIMO with OTFS modulation: Statistical CSI-based detec- MIMO systems,” IEEE Trans. Commun., vol. 70, no. 1, pp. 635–648,
tion,” IEEE Wireless Commun. Lett., vol. 12, no. 6, pp. 987–991, Jan. 2022.
Jun. 2023. [104] S. Srivastava, R. K. Singh, A. K. Jagannatham, and L. Hanzo, “Delay-
[83] B. Shen, Y. Wu, J. An, C. Xing, L. Zhao, and W. Zhang, “Random Doppler and angular domain 4D-sparse CSI estimation in OTFS
access with massive MIMO-OTFS in LEO satellite communica- aided MIMO systems,” IEEE Trans. Veh. Technol., vol. 71, no. 12,
tions,” IEEE J. Sel. Areas Commun., vol. 40, no. 10, pp. 2865–2881, pp. 13447–13452, Dec. 2022.
Oct. 2022. [105] A. Mohebbi, W.-P. Zhu, and M. O. Ahmad, “Compressive sensing-
[84] J. Feng, H. Q. Ngo, and M. Matthaiou, “OTFS-based massive MIMO based channel estimation for MIMO OTFS Systems,” in Proc. Bienn.
with fractional delay and Doppler Shift: The URLLC case,” in Proc. Symp. Commun. (BSC), Montreal, QC, Canada, 2023, pp. 71–76.
IEEE 33rd Annu. Int. Symp. Pers., Indoor Mobile Radio Commun. [106] S. Srivastava, R. K. Singh, A. K. Jagannatham, A. Chockalingam,
(PIMRC), Kyoto, Japan, 2022, pp. 522–528. and L. Hanzo, “OTFS transceiver design and sparse doubly-selective
[85] D. L. Li and A. Farhang, “OTFS without CP in massive MIMO: CSI estimation in analog and hybrid beamforming aided mmWave
Breaking Doppler limitations with TR-MRC and windowing,” in MIMO systems,” IEEE Trans. Wireless Commun., vol. 21, no. 12,
Proc. IEEE Wireless Commun. Netw. Conf. (WCNC), Austin, TX, pp. 10902–10917, Dec. 2022.
USA, 2022, pp. 2518–2523. [107] Y. Yan, C. Shan, J. Zhang, and H. Zhao, “Off-grid channel estimation
[86] S. K. Dehkordi, L. Gaudio, M. Kobayashi, G. Caire, and for OTFS-based mmWave hybrid beamforming systems,” IEEE
G. Colavolpe, “Beam-space MIMO radar for joint communication Commun. Lett., vol. 27, no. 8, pp. 2167–2171, Aug. 2023.
and sensing with OTFS modulation,” IEEE Trans. Wireless Commun., [108] X. Qi and J. Xie, “Cubature Kalman filter based millimeter wave
vol. 22, no. 10, pp. 6737–6749, Oct. 2023. beam tracking for OTFS systems,” China Commun., vol. 20, no. 7,
[87] S. Jiang and A. Alkhateeb, “Sensing aided OTFS massive MIMO pp. 233–240, Jul. 2023.
systems: Compressive channel estimation,” in Proc. IEEE Int. [109] A. Mehrotra, S. Srivastava, A. K. Jagannatham, and L. Hanzo, “Data-
Conf. Commun. Workshops (ICC Workshops), Rome, Italy, 2023, aided CSI estimation using affine-precoded superimposed pilots in
pp. 794–799. orthogonal time frequency space modulated MIMO systems,” IEEE
[88] Z. Gong, F. Jiang, C. Li, and X. Shen, “Simultaneous localization Trans. Commun., vol. 71, no. 8, pp. 4482–4498, Aug. 2023.
and communications with massive MIMO-OTFS,” IEEE J. Sel. Areas [110] A. Mehrotra, S. Srivastava, S. Asifa, A. K. Jagannatham, and
Commun., vol. 41, no. 12, pp. 3908–3924, Dec. 2023. L. Hanzo, “Online Bayesian learning-aided sparse CSI estimation
[89] A. C. Serrano, N. Petrov, M. G. Huici, and A. Yarovoy, “MIMO in OTFS modulated MIMO systems for ultra-high-Doppler scenar-
OTFS with arbitrary time-frequency allocation for joint radar and ios,” IEEE Trans. Commun., vol. 72, no. 4, pp. 2182–2200, Apr. 2024.
communications,” IEEE Trans. Radar Syst., vol. 1, pp. 707–718, [111] L. Zhao, J. Yang, Y. Liu, and W. Guo, “Block sparse Bayesian
Nov. 2023. learning-based channel estimation for MIMO-OTFS systems,” IEEE
[90] N. Tawa, T. Kuwabara, Y. Maruta, and T. Kaneko, “28 GHz over-the- Commun. Lett., vol. 26, no. 4, pp. 892–896, Apr. 2022.
air measurement using an OTFS multi-user distributed MIMO,” in [112] Q. Ren, Q. Li, Z. Xu, and Z. Pan, “Low overhead pilot design for
Proc. 51st Eur. Microw. Conf. (EuMC), 2022, pp. 450–453. channel estimation in MIMO-OTFS systems,” in Proc. IEEE/CIC Int.
[91] A. S. Bora, K. T. Phan, and Y. Hong, “Spatially correlated MIMO- Conf. Commun. (ICCC Workshops), Dalian, China, 2023, pp. 1–6.
OTFS for LEO satellite communication systems,” in Proc. IEEE Int. [113] C. Chen, J. Zhang, Y. Han, J. Lu, and S. Jin, “Channel estimation for
Conf. Commun. Workshops (ICC Workshops), 2022, pp. 723–728. massive MIMO-OTFS system in asymmetrical architecture,” IEEE
[92] A. S. Bora, K. T. Phan, and Y. Hong, “Mitigating spatial correlation in Signal Process. Lett., vol. 30, pp. 1412–1416, Oct. 2023.
MIMO-OTFS,” IEEE Trans. Veh. Technol., vol. 73, no. 3, pp. 3608– [114] R. Ouchikh, T. Chonavel, A. Aissa-El-Bey, and M. Djeddou,
3622, Mar. 2024. “Iterative channel estimation and data detection algorithm for
[93] T. M. C. Chu, H. J. Zepernick, A. Westerhagen, A. Hook, and MIMO-OTFS systems,” Digit. Signal Process., vol. 143, Nov. 2023,
B. Granbom, “Performance assessment of OTFS modulation in high Art. no. 104234.
Doppler airborne communication networks,” Mobile Netw. Appl., [115] Y. Liu et al., “Evolution of NOMA toward next generation multiple
vol. 27, pp. 1746–1756, Aug. 2022. access (NGMA) for 6G,” IEEE J. Sel. Areas Commun., vol. 40, no. 4,
[94] M. Li, S. Zhang, Y. Ge, F. Gao, and P. Fan, “Joint channel estimation pp. 1037–1071, Apr. 2022.
and data detection for hybrid RIS aided millimeter wave OTFS [116] V. Khammammetti and S. K. Mohammed, “OTFS-based multiple-
systems,” IEEE Trans. Commun., vol. 70, no. 10, pp. 6832–6848, access in high Doppler and delay spread wireless channels,” IEEE
Oct. 2022. Wireless Commun. Lett., vol. 8, no. 2, pp. 528–531, Apr. 2019.

VOLUME 5, 2024 4515

ALDABABSA et al.: SURVEY ON OTFS MODULATION

[117] G. D. Surabhi, R. M. Augustine, and A. Chockalingam, “Multiple [138] X. Lin et al., “Resource allocation for NOMA based OTFS
access in the delay-Doppler domain using OTFS modulation,” 2019, transmission with heterogeneous mobility users,” in Proc. IEEE 10th
arXiv:1902.03415. Int. Conf. Inf., Commun. Netw. (ICICN), Zhangye, China, 2022,
[118] O. K. Rasheed, G. D. Surabhi, and A. Chockalingam, “Sparse delay- pp. 91–98.
Doppler channel estimation in rapidly time-varying channels for [139] Y. Xu, Z. Du, W. Yuan, S. Jia, and V. C. M. Leung, “Performance of
multiuser OTFS on the uplink,” in Proc. IEEE 91st Veh. Technol. OTFS-NOMA scheme for coordinated direct and relay transmission
Conf. (VTC), Antwerp, Belgium, 2020, pp. 1–5. networks in high-mobility scenarios,” IEEE Wireless Commun. Lett.,
[119] M. Li, S. Zhang, P. Fan, and O. A. Dobre, “Multiple access for mas- vol. 12, no. 12, pp. 2268–2272, Dec. 2023.
sive MIMO-OTFS networks over angle-delay-Doppler domain,” in [140] T. M. C. Chu and H.-J. Zepernick, “Sum rate of OTFS-NOMA
Proc. IEEE Glob. Commun. Conf. (GLOBECOM), Taipei, Taiwan, systems with K-means clustering of user equipment,” in Proc.
2020, pp. 1–6. IEEE 34th Annu. Int. Symp. Pers., Indoor Mobile Radio Commun.
[120] R. Chong, S. Li, J. Yuan, and D. W. K. Ng, “Achievable rate upper- (PIMRC), Toronto, ON, Canada, 2023, pp. 1–6.
bounds of uplink multiuser OTFS transmissions,” IEEE Wireless [141] L. Xiang, K. Xu, J. Hu, C. Masouros, and K. Yang, “Robust NOMA-
Commun. Lett., vol. 11, no. 4, pp. 791–795, Apr. 2022. assisted OTFS-ISAC network design with 3-D motion prediction
[121] R. Chong, S. Li, W. Yuan, and J. Yuan, “Outage analysis for OTFS- topology,” IEEE Internet Things J., vol. 11, no. 9, pp. 15909–15918,
based single user and multi-user transmissions,” in Proc. IEEE Int. May 2024.
Conf. Commun. Workshops (ICC Workshops), 2022, pp. 746–751. [142] K. Deka, A. Thomas, and S. Sharma, “OTFS-SCMA: A code-
[122] R. M. Augustine and A. Chockalingam, “Interleaved time-frequency domain NOMA approach for orthogonal time frequency space
multiple access using OTFS modulation,” in Proc. IEEE 90th Veh. modulation,” IEEE Trans. Commun., vol. 69, no. 8, pp. 5043–5058,
Technol. Conf., Honolulu, HI, USA, 2019, pp. 1–5. Aug. 2021.
[123] V. Khammammetti and S. K. Mohammed, “Spectral efficiency [143] Z. Kang, H. Zhao, and H. Wang, “An efficient two-dimension
of OTFS based orthogonal multiple access with rectan- OTFS-NOMA scheme based on heterogenous mobility users group-
gular pulses,” IEEE Trans. Veh. Technol., vol. 71, no. 12, ing,” in Proc. IEEE 21st Int. Conf. Commun. Technol. (ICCT),
pp. 12989–13006, Dec. 2022. Tianjin, China, 2021, pp. 726–730.
[124] Z. Ding, R. Schober, P. Fan, and H. V. Poor, “OTFS-NOMA: [144] A. Thomas, K. Deka, P. Raviteja, and S. Sharma, “Convolutional
An efficient approach for exploiting heterogenous user mobility sparse coding based channel estimation for OTFS-SCMA in
profiles,” IEEE Trans. Commun., vol. 67, no. 11, pp. 7950–7965, uplink,” IEEE Trans. Commun., vol. 70, no. 8, pp. 5241–5257,
Nov. 2019. Aug. 2022.
[125] Z. Ding, “Robust beamforming design for OTFS-NOMA,” IEEE [145] H. Wen, W. Yuan, Z. Liu, and S. Li, “OTFS-SCMA: A downlink
Open J. Commun. Soc., vol. 1, pp. 33–40, 2020. NOMA scheme for massive connectivity in high mobility chan-
[126] W. Shao, S. Zhang, C. Zhong, X. Lei, and P. Fan, nels,” IEEE Trans. Wireless Commun., vol. 22, no. 9, pp. 5770–5784,
“Angle-delay-Doppler domain NOMA over massive MIMO-OTFS Sep. 2023.
networks,” in Proc. IEEE/CIC Int. Conf. Commun. China (ICCC), [146] H. Wen, W. Yuan, and S. Li, “Downlink OTFS non-orthogonal
Chongqing, China, 2020, pp. 74–79. multiple access receiver design based on cross-domain detec-
[127] A. Chatterjee, V. Rangamgari, S. Tiwari, and S. S. Das, tion,” in Proc. IEEE Int. Conf. Commun. Workshops (ICC
“Nonorthogonal multiple access with orthogonal time–frequency Workshops), 2022, pp. 928–933.
space signal transmission,” IEEE Syst. J., vol. 15, no. 1, [147] Y. Ge, Q. Deng, G. D. Gonzalez, Y. L. Guan, and Z. Ding, “OTFS
pp. 383–394, Mar. 2021. signaling for SCMA with coordinated multi-point vehicle communi-
[128] Y. Ge, Q. Deng, P. C. Ching, and Z. Ding, “OTFS signaling for uplink cations,” IEEE Trans. Veh. Technol., vol. 72, no. 7, pp. 9044–9057,
NOMA of heterogeneous mobility users,” IEEE Trans. Commun., Jul. 2023.
vol. 69, no. 5, pp. 3147–3161, May 2021. [148] Q. Deng, Y. Ge, and Z. Ding, “Jamming suppression via resource
[129] X. Zhou and Z. Gao, “Joint active user detection and chan- hopping in high-mobility OTFS-SCMA systems,” IEEE Wireless
nel estimation for grant-free NOMA-OTFS in LEO constellation Commun. Lett., vol. 12, no. 12, pp. 2138–2142, Dec. 2023.
Internet-of-Things,” in Proc. IEEE/CIC Int. Conf. Commun. China [149] Y. Ma, G. Ma, B. Ai, N. Wang, and Z. Zhong, “Orthogonal
(ICCC), Xiamen, China, 2021, pp. 735–740. time frequency code space modulation enabled multiple access
[130] X. Zhou et al., “Active terminal identification, channel estimation, under compactness-reduced channel spreading function,” IEEE Trans.
and signal detection for grant-free NOMA-OTFS in LEO satellite Wireless Commun., vol. 23, no. 3, pp. 1810–1826, Mar. 2024.
Internet-of-Things,” IEEE Trans. Wireless Commun., vol. 22, no. 4, [150] Y. Ma, G. Ma, B. Ai, J. Liu, N. Wang, and Z. Zhong, “OTFCS-
pp. 2847–2866, Apr. 2023. modulated waveform design for joint grant-free random access and
[131] Y. Liu, W. Yi, Z. Ding, X. Liu, O. A. Dobre, and N. Al-Dhahir, positioning in C-V2X,” IEEE J. Sel. Areas Commun., vol. 42, no. 1,
“Developing NOMA to next generation multiple access: Future vision pp. 103–119, Jan. 2024.
and research opportunities,” IEEE Wireless Commun., vol. 29, no. 6, [151] Y. Ge, L. Liu, S. Huang, D. Gonzalez, Y. L. Guan, and Z. Ding,
pp. 120–127, Dec. 2022. “Low-complexity memory AMP detector for high-mobility MIMO-
[132] S. M. Pishvaei and B. M. Tazehkand, “Error and outage performance OTFS SCMA systems,” in Proc. IEEE Int. Conf. Commun. (ICC),
of the FBMC-based OTFS using large-scale and massive MISO- Rome, Italy, 2023, pp. 1–6.
NOMA with millimeter-wave channel by LoS and NLoS paths [152] J. Li and Y. Hong, “Performance analysis and phase shift
impact,” Wireless Netw., vol. 28, no. 5, pp. 1835–1855, 2022. design of IRS-aided uplink OTFS-SCMA,” IEEE Access, vol. 11,
[133] J. Hu, J. Shi, X. Wang, X. Lu, Z. Li, and Z. Tie, “Secrecy trans- pp. 133059–133069, 2023.
mission of NOMA-OTFS based multicast-unicast streaming,” China [153] Y. Ma, G. Ma, N. Wang, Z. Zhong, and B. Ai, “OTFS-TSMA
Commun., vol. 20, no. 1, pp. 1–13, Jan. 2023. for massive Internet of Things in high-speed railway,” IEEE Trans.
[134] S. McWade, A. Farhang, and M. F. Flanagan, “Low-complexity Wireless Commun., vol. 21, no. 1, pp. 519–531, Jan. 2022.
reliability-based equalization and detection for OTFS-NOMA,” IEEE [154] G. Ma, B. Ai, F. Wang, and Z. Zhong, “Tandem spreading network-
Trans. Commun., vol. 71, no. 11, pp. 6779–6792, Nov. 2023. coded division multiple access,” IEEE Trans. Ind. Informat., vol. 13,
[135] S. McWade, M. F. Flanagan, and A. Farhang, “Low-Complexity no. 1, pp. 390–398, Feb. 2017.
Equalization and Detection for OTFS-NOMA,” in Proc. IEEE Int. [155] Y. Ma, G. Ma, N. Wang, Z. Zhong, J. Yuan, and B. Ai, “Enabling
Conf. Commun. Workshops (ICC Workshops), Rome, Italy, 2023, OTFS-TSMA for smart railways mMTC over LEO satellite: A
pp. 530–535. differential Doppler shift perspective,” IEEE Internet Things J.,
[136] H. Zhang, K. Niu, J. Xu, J. Dai, and J. Zhang, “Iterative vol. 10, no. 6, pp. 4799–4814, Mar. 2023.
SIC-based multiuser detection for uplink heterogeneous NOMA [156] H. Li, C. Gong, Q. Li, S. Hao, X. Wang, and X. Dai, “OTFS-
system,” in Proc. IEEE Globecom Workshops (GC Wkshps), Rio de PDMA scheme with EPA-based receivers for high-mobility IoT
Janeiro, Brazil, 2022, pp. 94–99. networks,” IEEE Trans. Wireless Commun., vol. 23, no. 5, pp. 4950–
[137] B. Shen, Y. Wu, W. Zhang, G. Y. Li, J. An, and C. Xing, 4963, May 2024.
“LEO satellite-enabled grant-free random access with MIMO- [157] Y. Zhang and P. Du, “Decomposition of uplink multi-user OTFS
OTFS,” in Proc. IEEE Global Commun. Conf. (GLOBECOM), Rio channel with partial CSI feedback,” Phys. Commun., vol. 55,
de Janeiro, Brazil, 2022, pp. 3308–3313. Dec. 2022, Art. no. 101911.

4516 VOLUME 5, 2024

[158] B. V. S. Reddy, C. Velampalli, and S. S. Das, “Performance analysis [179] S. E. Zegrar, H. Haif, and H. Arslan, “OTFS-based ISAC for
of multi-user OTFS, OTSM, and single carrier in uplink,” IEEE super-resolution range-velocity profile,” IEEE Trans. Commun., early
Trans. Commun., vol. 72, no. 3, pp. 1428–1443, Mar. 2024. access, Feb. 26, 2024, doi: 10.1109/TCOMM.2024.3369672.
[159] J. Feng, H. Q. Ngo, and M. Matthaiou, “Rectangular pulse-shaped [180] M. F. Keskin, C. Marcus, O. Eriksson, A. Alvarado, J. Widmer,
OTFS with fractional delay and Doppler shift for MU-MIMO and H. Wymeersch, “Integrated sensing and communications
systems,” in Proc. IEEE Int. Conf. Commun. Workshops (ICC with MIMO-OTFS: ISI/ICI exploitation and delay-Doppler
Workshops), Rome, Italy, 2023, pp. 788–793. multiplexing,” IEEE Trans. Wireless Commun., early access, Mar. 8,
[160] M. Y. Abdelsadek et al., “Future space networks: Toward the next 2024, doi: 10.1109/TWC.2024.3370501.
giant leap for humankind,” IEEE Trans. Commun., vol. 71, no. 2, [181] K. Zhang, W. Yuan, P. Fan, and X. Wang, “Dual-functional
pp. 949–1007, Feb. 2023. waveform design with local sidelobe suppression via OTFS sig-
[161] Y. Su, Y. Liu, Y. Zhou, J. Yuan, H. Cao, and J. Shi, “Broadband LEO naling,” IEEE Trans. Veh. Technol., early access, May 6, 2024,
satellite communications: Architectures and key technologies,” IEEE doi: 10.1109/TVT.2024.3395373.
Wireless Commun., vol. 26, no. 2, pp. 55–61, Apr. 2019. [182] X. Feng et al., “Underwater acoustic communications based on
[162] J. Shi et al., “OTFS enabled LEO satellite communications: A OTFS,” in Proc. 15th IEEE Int. Conf. Signal Process. (ICSP), 2020,
promising solution to severe Doppler effects,” IEEE Netw., vol. 38, pp. 439–444.
no. 1, pp. 203–209, Jan. 2024 [183] D. Varshney, M. Agrawal, and A. S. Gupta, “Orthogonal time
[163] Z. Xiao et al., “LEO satellite access network (LEO-SAN) toward frequency space modulation: Analysis in noisy underwater acoustic
6G: Challenges and approaches,” IEEE Wireless Commun., vol. 31, channels,” in Proc. IEEE Int. Symp. Ocean Technol. (SYMPOL),
no. 2, pp. 89–96, Apr. 2024 2023, pp. 1–5.
[164] M. Caus, M. Shaat, A. I. Pérez-Neira, M. Schellmann, and H. Cao, [184] T. P. S. Babu, J. Francis, and R. D. Koilpillai, “OTFS and
“Reliability oriented OTFS-based LEO satellites joint transmission OCDM based underwater acoustic communication: System design
scheme,” in Proc. IEEE Globecom Workshops (GC Wkshps), Rio de and evaluation,” in Proc. OCEANS, pp. 1–6, 2022.
Janeiro, Brazil, 2022, pp. 1406–1412.
[165] X. Wang, W. Shen, C. Xing, J. An, and L. Hanzo, “Joint Bayesian [185] X. Guo, B. Wang, Y. Zhu, Z. Fang, and S. Zhang, “Underwater
channel estimation and data detection for OTFS systems in LEO acoustic communication based on OTFS-IM,” in Proc. 14th Int. Conf.
satellite communications,” IEEE Trans. Commun., vol. 70, no. 7, Signal Process. Syst. (ICSPS), 2022, pp. 216–219.
pp. 4386–4399, Jul. 2022. [186] K. P. Arunkumar and C. R. Murthy, “Orthogonal delay scale
[166] C. Xu et al., “OTFS-aided RIS-assisted SAGIN systems outperform space modulation: A new technique for wideband time-varying
their OFDM counterparts in doubly selective high-Doppler scenar- channels,” IEEE Trans. Signal Process., vol. 70, pp. 2625–2638,
ios,” IEEE Internet Things J., vol. 10, no. 1, pp. 682–703, Jan. 2023. May 2022.
[167] Z. Gao et al., “Grant-Free NOMA-OTFS paradigm: Enabling [187] S. Hanga and W. Li, “OTFS for underwater acoustic communications:
efficient ubiquitous access for LEO satellite Internet-of-Things,” Practical system design and channel estimation,” in Proc. OCEANS,
2022, arXiv:2209.12159. 2022, pp. 1–7.
[168] J. Hu, Y. Jin, J. Shi, X. Liu, Z. Dai, and Z. Li, “Reliability [188] L. Jing, Q. Wang, C. He, and X. Zhang, “A learned denoising-based
analysis of stochastic geometry-based multi-UAV-aided LEO-Satcom sparse adaptive channel estimation for OTFS underwater acoustic
under OTFS,” in Proc. Int. Symp. Wireless Commun. Syst. (ISWCS), communications,” IEEE Wireless Commun. Lett., vol. 13, no. 4,
Hangzhou, China, 2022, pp. 1–6. pp. 969–973, Apr. 2024.
[169] “Framework and overall objectives of the future development [189] Y. Zhang, Q. Zhang, C. He, and C. Long, “Channel estimation for
of IMT for 2030 and beyond,” ITU-Rec. M.2160-0, Int. OTFS system over doubly spread sparse acoustic channels,” China
Telecommun. Union, Geneva, Switzerland, 2023. [Online]. Commun., vol. 20, no. 1, pp. 50–65, Jan. 2023.
Available: https://www.itu.int/dms_pubrec/itu-r/rec/m/R-REC- [190] S. Zhang, Y. Zhang, J. Chang, B. Wang, and W. Bai, “DNN-based
M.2160-0-202311-I signal detection for underwater OTFS systems,” in Proc. IEEE/CIC
[170] W. Yuan, Z. Wei, S. Li, J. Yuan, and D. W. K. Ng, “Integrated Int. Conf. Commun. China (ICCC Workshops), 2022, pp. 348–352.
sensing and communication-assisted orthogonal time frequency space [191] L. Jing, N. Zhang, C. He, J. Shang, X. Liu, and H. Yin,
transmission for vehicular networks,” IEEE J. Sel. Topics Signal “OTFS underwater acoustic communications based on passive time
Process., vol. 15, no. 6, pp. 1515–1528, Nov. 2021. reversal,” Appl. Acoust., vol. 185, Jan. 2022, Art. no. 108386.
[171] S. Li et al., “A novel ISAC transmission framework based on [192] X. Wang, X. Shi, J. Wang, and Z. Sun, “Iterative LMMSE-
spatially-spread orthogonal time frequency space modulation,” IEEE SIC detector for DSE-aware underwater acoustic OTFS
J. Sel. Areas Commun., vol. 40, no. 6, pp. 1854–1872, Jun. 2022. systems,” IEEE Trans. Veh. Technol., early access, Feb. 6, 2024,
[172] Z. Tang, Z. Jiang, W. Pan, and L. Zeng, “The estimation method doi: 10.1109/TVT.2024.3362894.
of sensing parameters based on OTFS,” IEEE Access, vol. 11, [193] Z. Xue, Q. Wang, L. Jing, C. Zhou, and C. He, “Unitary approximate
pp. 66035–66049, 2023. message passing detection method for OTFS underwater acoustic
[173] X. Xia, K. Xu, Y. Wang, Y. Xu, and W. Xie, “Achieving better accu- communication system,” in Proc. IEEE Int. Conf. Signal Process.,
racy with less computations: A delay-Doppler spectrum matching Commun. Comput. (ICSPCC), 2023, pp. 1–5.
assisted active sensing framework for OTFS based ISAC systems,”
[194] R. Han, J. Ma, and L. Bai, “Trajectory planning for OTFS-based
IEEE Trans. Wireless Commun., vol. 23, no. 6, pp. 6204–6220,
UAV communications,” China Commun., vol. 20, no. 1, pp. 114–124,
Jun. 2024.
Jan. 2023.
[174] Y. Shi and Y. Huang, “Integrated sensing and communication-assisted
user state refinement for OTFS systems,” IEEE Trans. Wireless [195] F. Linsalata, A. Albanese, V. Sciancalepore, F. Roveda, M. Magarini,
Commun., vol. 23, no. 2, pp. 922–936, Feb. 2024. and X. Costa-Perez, “OTFS-superimposed PRACH-aided localization
[175] X. Liu, Y. Yang, J. Gong, N. Xia, J. Guo, and M. Peng, “Amplitude for UAV safety applications,” in Proc. IEEE Global Commun. Conf.
barycenter calibration of delay-Doppler spectrum for OTFS signal— (GLOBECOM), 2021, pp. 1–6.
An endeavor to integrated sensing and communication waveform [196] A. Nordio, C. F. Chiasserini, and E. Viterbo, “Robust localization
design,” IEEE Trans. Wireless Commun., vol. 23, no. 4, pp. 2622– of UAVs in OTFS-based networks,” in Proc. IEEE Global Commun.
2637, Apr. 2024. Conf. (GLOBECOM), 2023, pp. 7471–7477.
[176] Y. Yang, Y. Pan, X. Liu, and M. Peng, “Barycenter calibration [197] H. Niu, J. Hu, J. Shi, and Z. Li, “Secrecy performance analysis
with high order spectra of windowed delay-Doppler signals for for OTFS modulation based downlink LEO satellite communica-
OTFS based ISAC systems,” IEEE Trans. Signal Process., vol. 72, tion,” in Proc. IEEE Globecom Workshops (GC Wkshps), 2023,
pp. 2450–2466, 2024. pp. 2135–2139.
[177] T. P. Zieliński et al., “Wireless OTFS-based integrated sensing and [198] X. Xu, M.-M. Zhao, M. Lei, and M.-J. Zhao, “A damped
communication for moving vehicle detection,” IEEE Sensors J., GAMP detection algorithm for OTFS system based on deep
vol. 24, no. 5, pp. 6573–6583, Mar. 2024. learning,” in Proc. IEEE 92nd Veh. Technol. Conf., Victoria, BC,
[178] X. Yang, H. Li, Q. Guo, J. A. Zhang, X. Huan, and Z. Cheng, Canada, 2020, pp. 1–5.
“Sensing aided uplink transmission in OTFS ISAC with joint [199] M. Liu, M.-M. Zhao, M. Lei, and M.-J. Zhao, “Autoencoder based
parameter association, channel estimation and signal detection,” IEEE PAPR reduction for OTFS modulation,” in Proc. IEEE 94th Veh.
Trans. Veh. Technol., vol. 73, no. 6, pp. 9109–9114, Jun. 2024. Technol. Conf., Norman, OK, USA, 2021, pp. 1–5.

VOLUME 5, 2024 4517

ALDABABSA et al.: SURVEY ON OTFS MODULATION

[200] J. Park, J.-P. Hong, H. Kim, and B. J. Jeong, “Auto-encoder based [220] K. Ma, Z. Gao, J. Wang, and L. Cheng, “Physical layer security
orthogonal time frequency space modulation and detection with meta- design for FDD IM-OTFS transmissions based on secure map-
learning,” IEEE Access, vol. 11, pp. 43008–43018, 2023. ping,” IEEE Access, vol. 11, pp. 98293–98304, 2023.
[201] Y. K. Enku et al., “Two-dimensional convolutional neural network- [221] K. Zhang, X.-Q. Jiang, H. Hai, R. Qiu, and S. Mumtaz,
based signal detection for OTFS systems,” IEEE Wireless Commun. “Enhanced index modulation aided orthogonal time frequency
Lett., vol. 10, no. 11, pp. 2514–2518, Nov. 2021. space with variable active grids and multiple constella-
[202] Y. K. Enku, B. Bai, S. Li, M. Liu, and I. N. Tiba, “Deep-learning tions,” IEEE Trans. Veh. Technol., early access, Mar. 7, 2024,
based signal detection for MIMO-OTFS systems,” in Proc. IEEE Int. doi: 10.1109/TVT.2024.3373591.
Conf. Commun. Workshops (ICC Workshops), 2022, pp. 1–5. [222] Y. I. Tek and E. Basar, “Joint delay-Doppler index modulation for
[203] Z. Zhou, L. Liu, J. Xu, and R. Calderbank, “Learning to orthogonal time frequency space modulation,” IEEE Trans. Commun.,
equalize OTFS,” IEEE Trans. Wireless Commun., vol. 21, no. 9, early access, Feb. 26, 2024, doi: 10.1109/TCOMM.2024.3370827.
pp. 7723–7736, Sep. 2022. [223] Y. Wu, L. Xiao, Y. Xie, G. Liu, and T. Jiang, “Efficient signal
[204] X. Zhang, L. Xiao, S. Li, Q. Yuan, L. Xiang, and T. Jiang, “Gaussian detector design for OTFS with index modulation,” Digital Commun.
AMP aided model-driven learning for OTFS system,” IEEE Commun. Netw., to be published.
Lett., vol. 26, no. 12, pp. 2949–2953, Dec. 2022. [224] V. S. Bhat and A. Chockalingam, “OTFS-IM with decode and
[205] S. Li, C. Ding, L. Xiao, X. Zhang, G. Liu, and T. Jiang, “Expectation forward relaying,” in Proc. IEEE 98th Veh. Technol. Conf., 2023,
propagation aided model driven learning for OTFS signal detec- pp. 1–6.
tion,” IEEE Trans. Veh. Technol., vol. 72, no. 9, pp. 12407–12412, [225] C. Zhang, D. Feng, M. Liu, and B. Bai, “Spatial modulation based
Sep. 2023. MIMO-OTFS transmissions,” in Proc. IEEE/CIC Int. Conf. Commun.
[206] M. H. Abid, I. A. Talin, and M. I. Kadir, “Reconfigurable intelligent China (ICCC Workshops), Xiamen, China, 2021, pp. 427–432.
surface-aided orthogonal time frequency space and its deep learning- [226] S. Li, L. Xiao, X. Zhang, L. Li, and T. Jiang, “Spatial multiplexing
based signal detection,” IEEE Access, vol. 11, pp. 47321–47338, aided OTFS with index modulation,” IEEE Trans. Veh. Technol.,
2023. vol. 72, no. 6, pp. 8192–8197, Jun. 2023.
[207] L. Guo, P. Gu, J. Zou, G. Liu, and F. Shu, “DNN-based fractional [227] Y. Yang et al., “Design and analysis of spatial modulation based
doppler channel estimation for OTFS modulation,” IEEE Trans. Veh. orthogonal time frequency space system,” China Commun., vol. 18,
Technol., vol. 72, no. 11, pp. 15062–15067, Nov. 2023. no. 8, pp. 209–223, Aug. 2021.
[208] C. Liu, S. Li, W. Yuan, X. Liu, and D. W. K. Ng, “Predictive precoder [228] Y. Qian, L. Xiao, and T. Jiang, “SM-STBC aided orthogonal time
design for OTFS-enabled URLLC: A deep learning approach,” IEEE frequency space modulation,” in Proc. IEEE Wireless Commun. Netw.
J. Sel. Areas Commun., vol. 41, no. 7, pp. 2245–2260, Jul. 2023. Conf. (WCNC), Austin, TX, USA, 2022, pp. 2172–2177.
[209] A. Singh, S. Sharma, K. Deka, and V. Bhatia, “DL-based OTFS [229] Y. Yang, Z. Bai, H. Liu, K. Pang, X. Hao, and K. J. Kim, “Design and
signal detection in presence of hardware impairments,” IEEE Wireless performance analysis of spatial-index modulation based orthogonal
Commun. Lett., vol. 12, no. 9, pp. 1533–1537, Sep. 2023. time frequency space system,” in Proc. IEEE Int. Conf. Commun.
[210] X. Zhang, S. Zhang, L. Xiao, S. Li, and T. Jiang, “Graph neural Workshops (ICC Workshops), 2022, pp. 922–927.
network assisted efficient signal detection for OTFS systems,” IEEE [230] Y. Yang, Z. Bai, K. Pang, S. Guo, H. Zhang, and K. S. Kwak,
Commun. Lett., vol. 27, no. 8, pp. 2058–2062, Aug. 2023. “Spatial-index modulation based orthogonal time frequency space
[211] A. Kumar and U. Satija, “Residual stack-aided hybrid CNN- system in vehicular networks,” IEEE J. Intell. Transport. Syst.,
LSTM-based automatic modulation classification for orthogonal vol. 24, no. 6, pp. 6165–6177, Jun. 2023.
time-frequency space system,” IEEE Commun. Lett., vol. 27, no. 12, [231] X. Zou, S. Fan, H. Chen, Y. Xiao, C. Di, and J. Ji,
pp. 3255–3259, Dec. 2023. “Orthogonal time frequency space with generalized spatial modula-
[212] Y. Yue, J. Shi, Z. Li, J. Hu, and Z. Tie, “Model-driven deep learning tion,” in Proc. IEEE Veh. Technol. Conf., Helsinki, Finland, 2022,
assisted detector for OTFS With channel estimation error,” IEEE pp. 1–5.
Commun. Lett., vol. 28, no. 4, pp. 842–846, Apr. 2024. [232] Z. Sui, H. Zhang, Y. Xin, T. Bao, L.-L. Yang, and L. Hanzo, “Low
[213] S. R. Mattu and A. Chockalingam, “Learning in time- complexity detection of spatial modulation aided OTFS in doubly-
frequency domain for fractional delay-Doppler channel estimation selective channels,” IEEE Trans. Veh. Technol., vol. 72, no. 10,
in OTFS,” IEEE Wireless Commun. Lett., vol. 13, no. 5, pp. 13746–13751, Oct. 2023.
pp. 1245–1249, May 2024. [233] D. Feng, B. Bai, and F. Wan, “Generalised space-delay-Doppler index
[214] Y. Liang, L. Li, P. Fan, and Y. Guan, “Doppler resilient orthogonal modulated OTFS transmission,” in Proc. IEEE Wireless Commun.
time-frequency space (OTFS) systems based on index modula- Netw. Conf. (WCNC), 2023, pp. 1–5.
tion,” in Proc. IEEE 91st Veh. Technol. Conf., Antwerp, Belgium, [234] D. Feng, J. Zheng, B. Bai, J. Jiang, and H. Chu, “Generalized index
2020, pp. 1–5. modulation for MIMO-OTFS transmission,” IEEE Wireless Commun.
[215] H. Ren, W. Xu, and L. Wang, “Orthogonal time-frequency space Lett., vol. 12, no. 5, pp. 907–911, May 2023.
with improved index modulation,” in Proc. 15th Int. Conf. Signal [235] Z. Sui, H. Zhang, S. Sun, L.-L. Yang, and L. Hanzo, “Space-
Process. Commun. Syst. (ICSPCS), 2021, pp. 1–6. time shift keying aided OTFS modulation for orthogonal multiple
[216] H. Zhao, D. He, Z. Kang, and H. Wang, “Orthogonal time frequency access,” IEEE Trans. Commun., vol. 71, no. 12, pp. 7393–7408,
space (OTFS) with dual-mode index modulation,” IEEE Wireless Dec. 2023.
Commun. Lett., vol. 10, no. 5, pp. 991–995, May 2021. [236] S. Doğan-Tusha, A. Tusha, S. Althunibat, and K. Qaraqe,
[217] L. Li, L. Xiao, X. Zhang, G. Liu, S. Li, and T. Jiang, “An efficient “Orthogonal time frequency space multiple access using index modu-
symbol-by-symbol aided expectation propagation detector for OTFS lation,” IEEE Trans. Veh. Technol., vol. 72, no. 12, pp. 15858–15866,
with index modulation,” IEEE Wireless Commun. Lett., vol. 11, Dec. 2023.
no. 10, pp. 2046–2050, Oct. 2022. [237] A. Tariq, M. S. Sarwar, and S. Y. Shin, “Orthogonal time
[218] D. Feng, J. Zheng, B. Bai, J. Jiang, and L. Zheng, “In-phase frequency space index modulation based on non-orthogonal multiple
and quadrature index modulation aided OTFS transmission,” IEEE access,” in Proc. 14th Int. Conf. Inf. Commun. Technol. Converg.
Commun. Lett., vol. 26, no. 6, pp. 1318–1322, Jun. 2022. (ICTC), 2023, pp. 838–841.
[219] M. Qian, F. Ji, Y. Ge, M. Wen, X. Cheng, and H. V. Poor, [238] E. Nayebi, A. Ashikhmin, T. L. Marzetta, H. Yang, and
“Block-wise index modulation and receiver design for high-mobility B. D. Rao, “Precoding and power optimization in cell-free massive
OTFS communications,” IEEE Trans. Commun., vol. 71, no. 10, MIMO systems,” IEEE Trans. Wireless Commun., vol. 16, no. 7,
pp. 5726–5739, Oct. 2023. pp. 4445–4459, Jul. 2017.

4518 VOLUME 5, 2024