2021 IEEE/CIC International Conference on Communications in China (ICCC Workshops)

A Tutorial to Orthogonal Time Frequency Space

2021 IEEE/CIC International Conference on Communications in China (ICCC Workshops) | 978-1-6654-3944-2/21/$31.00 ©2021 IEEE | DOI: 10.1109/ICCCWorkshops52231.2021.9538891

Modulation for Future Wireless Communications

Shuangyang Li∗¶ , Weijie Yuan† , Zhiqiang Wei‡ , Ruisi He§ , Bo Ai§ , Baoming Bai¶ , and Jinhong Yuan∗
∗ School of Electrical Engineering and Telecommunications, University of New South Wales, Sydney, Australia
† Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen, China
‡ Institute for Digtal Communications (IDC), Friedrich-Alexander University Erlangen-Nuremberg, Germany
§ State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China
¶ State Key Lab. of ISN, Xidian University, Xi’an, China

Abstract—The recently proposed orthogonal time frequency DD domain signal to the TF domain. Thanks to the ISFFT,
space (OTFS) modulation has demonstrated superior perfor- each DD domain symbol spreads onto the whole TF domain
mances over the conventional orthogonal frequency-division mul- and thus, principally experiences the whole fluctuations of
tiplexing (OFDM) modulation. In this paper, we provide a tutorial
to the OTFS modulation for future wireless communications. the TF channel response over an OTFS frame. Therefore,
We first revisit the properties of the wireless channels and then OTFS modulation enjoys a strong resilience to the Doppler
introduce the OTFS system models. In particular, we also discuss spread and offers the potential of achieving the full channel
the characteristics of OTFS signals over different domains. Based diversity [4], [5].
on the OTFS characteristics, we briefly discuss the performances The unique properties of OTFS modulation have made it
of OTFS modulation, including the performances of diversity
and coding gain, channel estimation, and detection. Finally, we a promising waveform candidate for B5G wireless networks.
provide a discussion regarding the potential applications of OTFS Other than facilitating the applications with a high mobility,
modulation for wireless communications, where the research B5G networks also have stringent data rate requirements in
challenges are also highlighted. order to support the transmission of massive amount of data.
Considering the severe spectrum congestion under 6 GHz
I. I NTRODUCTION
frequency bands in current wireless networks, the ultimate
With the rapid development of technologies, the existing solution to meet the ultra-high data rate requirement is to
wireless communication networks based on the conventional utilize higher frequency bands, such as the millimeter wave
orthogonal frequency-division multiplexing (OFDM) modula- (mmWave) bands, and to allow frequent spectrum reuse at the
tion have encountered unprecedented challenges in meeting same time. However, operating on a higher carrier frequency
the requirements of various emerging applications. In partic- can inevitably aggravate the Doppler spread, while frequent
ular, the beyond 5G (B5G) wireless networks are expected reusing same spectrum resource can cause severe interference.
to provide reliable communications for a wide range of Fortunately, OTFS modulation is insensitive to the Doppler
terminals with a high mobility, including autonomous cars, spread. Furthermore, the symbol spreading from DD domain
high speed trains (HSTs), unmanned aerial vehicles (UAVs), to the TF domain produces a strong resilience to interference
and low-earth-orbit (LEO) satellites. In those scenarios, the and is thus eminently suitable for frequency reuse. Although
conventional OFDM modulation may break down due to the OTFS modulation has demonstrated various advantages for
severe Doppler spread introduced by the high mobility [1]. To wireless communications, the research of its relevant topics is
facilitate the design of B5G wireless networks, the recently only at the early stage. To fully unleash the great potentials of
proposed orthogonal time frequency space (OTFS) modulation OTFS modulation, challenging fundamental research problems
has attracted substantial attention [1], [2]. need to be addressed and potential applications need to be
OTFS modulation is a novel two-dimensional (2D) modula- investigated.
tion scheme that was proposed by Hadani et. al [2]. In contrast In this paper, we aim to provide a brief tutorial to OTFS
to the conventional OFDM modulation, OTFS modulation modulation with a focus on its concept, performances, and
adopts the delay-Doppler (DD) domain signal representation potential applications for future wireless communications.
instead of the time-frequency (TF) domain, where the channel We firstly study the channel characteristics among different
responses are relatively sparse and robust [2], [3]. The DD domains and then introduce the concept of OTFS modulation,
domain channel property gives rise to the symbol placement in where the connections to the conventional OFDM modulation
the DD domain. By doing so, the information symbols directly are also emphasized. Afterwards, we investigate the perfor-
interact with the DD domain channel response, resulting in mances of OTFS modulation, including the performances of
a much simpler input-output relationship compared to that diversity and coding gain, channel estimation, and detection. In
of the OFDM modulation, especially in the high-mobility the end, we provide a detailed discussion on the potential ap-
channels [4]. Furthermore, OTFS modulation invokes the 2D plications of OTFS modulation for wireless communications,
symplectic finite Fourier transform (ISFFT) to transform the where some potential research problems are also highlighted.

Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:17:47 UTC from IEEE Xplore. Restrictions apply.
978-1-6654-3944-2/21/$31.00 ©2021 IEEE 439
 2021 IEEE/CIC International Conference on Communications in China (ICCC Workshops)

II. C HANNEL C HARACTERISTICS AND OTFS C ONCEPT

A. Underspread WSSUS Channels

Without loss of generality, we consider the general

Rayleigh-fading underspread wide-sense stationary uncorre- ;
lated scattering (WSSUS) channels [3]. The underspread K W Q
WSSUS channel model is a commonly adopted model for real-
world wireless channels [3], [6], whose stochastic properties <
match well with the actual channel measurement results [3].
The underspread WSSUS channel model has several advan-
tageous properties that are important for practical designs of
wireless communication systems.
For any given linear time-varying channel, its second-
order statistics can in general be represented by four differ-
ent variables that are time, frequency, delay, and Doppler, Fig. 1. The block diagram of an OTFS system.
respectively. With the assumption of WSSUS, this complex
statistical dependence can be reduced to only two variables [7]. delay and Doppler values among all resolvable paths1 . If a
For example, we can fully describe a WSSUS channel by channel satisfies 4τmax νmax ≤ 1, then it is said to be under-
its response in either the TF domain or in DD domain. The spread [3]. For an underspread channel, there exists a TF re-
rationale of the WSSUS assumption is that in practical wireless gion that is no smaller than one, where the channel response is
channels, the channel impairments are introduced by physical roughly constant. Specifically, the corresponding time domain
scatterers with different reflectivities and any two distinct and frequency domain intervals are commonly referred to as
scatterers generally have uncorrelated reflectivities. In specific, the coherence time and coherence bandwidth, respectively. The
the US assumption implies that the different delay shifts concept of the underspread channel is commonly considered
associated with different resolvable paths are uncorrelated. for wireless communication designs, including both OFDM
On the other hand, the WSS assumption indicates that the and OTFS systems [3]. However, it also should be noted that
different Doppler shifts associated with different resolvable the underspread WSSUS channel model does not suitable for
paths are uncorrelated such that the time domain channel some specific communication scenarios, such as underwater
response is jointly WSS with respect to the time. Without acoustic communications and HST communications. Potential
loss of generality, the DD domain representation of a linear applications of OTFS modulation in those scenarios will be
time-varying channel is given by [2], [3], discussed in the coming section.

B. OTFS System Model

P

h (τ, ν) = hi δ (τ − τi ) δ (ν − νi ). (1) OTFS systems has shown a great compatibility with the
i=1
conventional OFDM systems, where the transceivers of con-
ventional OFDM systems can be adopted for OTFS systems
with simple pre- and post- processing units [8]. Without loss
In (1), P is the number of resolvable paths, while hi , τi , and
of generality, we consider the OFDM-based OTFS transceiver
νi are the channel coefficients, delay shifts, and Doppler shifts
design as shown in Fig. 1. Let X be the 2D information
corresponding to the i-th path, respectively. According to the
symbol matrix in the DD domain of size M × N , where
WSSUS assumption, it can be shown that the delay shift τi ,
M denotes the number of delay intervals, corresponding to
Doppler shift νi , and channel coefficient hi associated with the
the number of subcarriers, while N denotes the number of
i-th path are uncorrelated to those from the j-th path, given
Doppler intervals, corresponding to the number of time slots
i = j. Furthermore, the delay and Doppler shifts are usually
(OFDM symbols), respectively. As shown in Fig. 1, the TF
modeled as uncorrelated variables for each resolvable path,
domain OTFS symbols are obtained based on X via the
because they are corresponding to the relative distance and
ISFFT and are then transformed into the time delay (TD)
relative velocity that are two different physical attributes of the
domain2 via the IFFT. Given the time domain OTFS symbols,
scatterer [3]. However, the channel coefficient hi dependents
the OTFS signal s(t) can be obtained based on the pulse
on the statistical characterization of the reflectivity strength
shaping filter gtx (t). According to Fig. 1, the DD domain
of the underlying path, which is usually related to the corre-
sponding delay shift τi and Doppler shift νi . The underlying
max = c fc , where v is
1 The maximum Doppler value is given by ν v
relationship between hi , τi , and νi are usually characterized the relative speed, c is the speed of light, and fc is the carrier frequency,
by the power delay profile and power Doppler profile [3]. respectively. The maximum delay shift is given by τmax = dmax c
, where
dmax is the maximum distance difference among the P channel paths.
It is also important to characterize whether the WSSUS 2 The TD domain has often been referred to as the time domain in the
channel is underspread. Let τmax and νmax be the maximum related literature. Therefore, we will use the term time domain henceforth.

Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:17:47 UTC from IEEE Xplore. Restrictions apply.
440
 2021 IEEE/CIC International Conference on Communications in China (ICCC Workshops)

received symbols Y are obtained by firstly matched-filtering An interesting fact of OTFS signals is that each DD domain
the received signal r(t) with grx (t) and then performing FFT symbol corresponds to a delayed, phase-rotated pulse train in
and SFFT. Note that due to the doubly selectivity of general the time domain [1], where the location of the symbol on
wireless channels, the DD domain received symbols Y usually DD domain grids determines how the pulse train is delayed
suffer from inter-Doppler/delay interference (IDI). Therefore, and phase-rotated in the time domain. In fact, the time domain
channel equalization is important for OTFS systems in order to OTFS signal behaves locally like time-division multiple access
achieve a good error performance. This issue will be discussed (TDMA) (localized pulses in the time domain), globally like
in the later section. OFDM (localized pulses in the Doppler domain) and spreading
To design the OTFS system, it is necessary to select suitable like code-division multiple access (CDMA) (2D spreading in
values of M and N according to the channel conditions. the DD domain), thus inheriting their beneficial properties [1].
According to the Cohere’s website [9], the typical delay and
III. P ERFORMANCE A NALYSIS
Doppler periods are 50 μs and 20 KHz, respectively, while the
typical delay and Doppler resolutions are 100 ns and 1 KHz, A. Diversity Gain and Coding Gain
respectively. In particular, the delay and Doppler periods The special DD domain symbol placement of OTFS modu-
correspond to the OFDM symbol duration and subcarrier lation offers the potential of achieving full channel diversity,
spacing under the conventional OFDM settings, respectively. i.e., diversity order P . Indeed, the diversity advantages of
On the other hand, the delay and Doppler resolutions denote OTFS modulation has been demonstrated in recent literatures.
the minimum widths of the DD domain shaping pulse along It is reported in [4] that the full diversity can be achieved
the delay and Doppler axes, respectively. Given those con- almost surely for the case of P = 2 when the frame size is
50μs
ditions, it is easy to calculate that M = 100ns = 500 and sufficiently large, even for uncoded OTFS modulation systems.
20KHz
N = 1KHz = 20, yielding the total bandwidth of 10 MHz The potential of achieving full channel diversity has introduc-
and frame duration of 1 ms [9]. ing new design criterion for channel coding. In particular, it
However, in practical systems, there may not be sufficient has been shown in [5] that there exists an interesting trade-
resources to support the 1 KHz Doppler resolution3 . In this off between the coding gain and diversity gain for OTFS
case, the OTFS system will suffer from the fractional Doppler modulation. In specific, the diversity gain of OTFS systems
shifts. Consequently, the effective DD domain channel will no improves with the number of independent resolvable paths,
longer be sparse, thus imposing challenges for channel esti- while the coding gain declines [5]. Intuitively speaking, this
mation and equalization. A potential solution to the fractional is because the transmit power of the OTFS signal is equally
Doppler shifts is to apply TF domain windows to the OTFS divided to all the resolvable paths and therefore, the signal-
systems, in order to retrieve the channel sparsity [12], [13]. to-noise ratio (SNR) improvement from channel coding, i.e.,
the coding gain, is also scaled by with respect to the number
C. OTFS Characteristics of resolvable paths. On the other hand, it is also suggested
Next, we will briefly study the OTFS signals in different in [5] that increasing the Euclidean distance between codeword
domains. In Fig. 1, we consider an example, where there is pairs can generally improve the error performance for coded
only one information symbol in the DD domain. As illustrated OTFS systems. Therefore, a preliminary code design criterion
in Fig. 1, by performing ISFFT, each DD domain symbol is is to maximize the minimum Euclidean distance between all
spread onto the whole TF domain, and then transferred to a codeword pairs, which is similar to the code design for additive
pulse train in the time domain. Assuming that the wireless white Gaussian noise (AWGN) channels. We demonstrate the
channel is noiseless and only has two paths with different diversity and coding gain trade-off in Fig. 2 via the frame error
delay and Doppler indices. The received time domain OTFS rate (FER) analysis, where the rate-half convolutional code
signal is then the superposition of the delayed, phase-rotated with generator matrix [1 + D, D] is applied. It is clearly shown
versions of the transmitted OTFS signal. After performing the in the figure that compared to the case where P = 2, both
FFT and the SFFT, we observe that the DD domain received QPSK and 16-QAM mapped OTFS systems with P = 4 enjoy
symbols are the delay- and Doppler-shifted versions of the a better diversity gain but the coding gain is much smaller.
original DD domain information symbol corresponding to the B. Channel Estimation Performance
channel coefficients and delay, Doppler shifts. By observing
Different from the conventional estimation methods in the
the connections between the transmitted and received OTFS
TF domain, OTFS systems can apply simple channel esti-
signals in the DD and TF domains, it is obvious that the DD
mation methods in the DD domain with a much smaller
domain input-output relationship is much simpler compared
signaling overhead. Perhaps, the most famous channel esti-
to that of the TF domain. Specifically, this example demon-
mation method for OTFS systems is the one proposed in [14],
strates the sparsity, separability properties of the DD domain
where the authors smartly embed a DD domain impulse (pilot
channel representation, thus validates the motivation of OTFS
symbol) together with a guard space to estimate the channel.
modulation.
With such a simple frame format, the embedded DD domain
3 The value of delay resolution is usually sufficiently small in typical wide- pilot symbol is circularly convolved with the DD domain
band systems [10], [11] channel response. With integer delay and Doppler indices,

Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:17:47 UTC from IEEE Xplore. Restrictions apply.
441
 2021 IEEE/CIC International Conference on Communications in China (ICCC Workshops)

10 0 10 0

16-QAM, P = 4

16-QAM, P = 2 10 -1

10 -1
QPSK, P = 2
QPSK, P = 4 10 -2

BER
FER

10 -2 10 -3

QPSK, Uncoded, P = 2
QPSK, Conv coded, P = 2
10 -4
QPSK, Uncoded, P = 4 DD domain MMSE
10 -3 QPSK, Conv coded, P = 4 Message passing algorithm
16-QAM, Uncoded, P = 2 Cross domain iterative detection 1 iter
16-QAM, Conv coded, P = 2 10 -5 Cross domain iterative detection 2 iter
16-QAM, Conv coded, P = 4 Cross domain iterative detection 3 iter
16-QAM, Uncoded, P = 4 Cross domain iterative detection 5 iter
10 -4 10 -6
0 5 10 15 20 25 0 2 4 6 8 10 12 14 16
Eb/N0 (dB) Es/N0 (dB)

Fig. 2. FER performance of convolutionally coded OTFS systems with Fig. 3. BER performance for uncoded OTFS modulation with P = 10 paths
different number of paths and the relative speed of 250 km/h, where N = 8 and fractional Doppler shifts, where M = 64 and N = 32.
and M = 16.

more attractive solution [19]. The rationale behind this idea

there will be in total P received pilot symbols that are confined
is that the unitary transformation between different domains
within the area with respect to τmax and νmax . Therefore,
allows the detection/estimation errors in one domain to be
by comparing each received pilot symbol and transmitted
principally orthogonal to the detection/estimation errors in
pilot symbol, the DD domain channel parameters can be
the other domain, which can suppress the correlation between
simply obtained. On the other hand, the channel estimation
the input errors and the output errors for each domain de-
performance for OTFS systems is generally better than that of
tection/estimation. Therefore, it is possible to apply iterative
the OFDM systems, thanks to the channel sparsity and quasi-
signal processing approaches to jointly detect the OTFS signals
stationarity in the DD domain [1], [15]. Particularly, one can
from different domains in order to fully exploit the channel
further improve the channel sparsity by applying TF domain
characteristics. For example, a cross domain iterative detection
windows to facilitate the estimation performance [13].
algorithm is proposed for OTFS modulation in [19], where a
C. Detection Performance linear minimum mean squared error (L-MMSE) estimator is
applied in the time domain while a symbol-by-symbol detector
Channel detection is an important part for fully exploiting is utilized in the DD domain. Interestingly, by combining
the OTFS potentials. To this end, various detection methods such two basic methods, this algorithm shows promising
have been proposed in the literature. A message passing algo- error performance even in very severe and complex fractional
rithm (MPA) is introduced in [10], where the interference from Doppler cases. We show the bit error rate (BER) performance
other information symbols is treated as a Gaussian variable to in Fig. 3, where a complex P = 10 case is considered with
reduce the detection complexity. However, due to the short fractional Doppler shifts. In this case, conventional OTFS
cycles on probabilistic graphical model, the proposed MPA detection algorithms, such as the MMSE detection, and MPA,
may fail to converge and results in performance degradation. cannot achieve good error performance. However, the cross
To solve this issue, the authors of [16] proposed a convergence domain iterative detection algorithm can still work very well.
guaranteed receiver based on the variational Bayes framework.
The basic idea of this detector is to approximate the cor-
responding a posteriori distribution of the optimal detection IV. P OTENTIAL A PPLICATIONS
by exploiting the Kullback-Leibler (KL) divergence such that
A. Integrated Sensing and Communication
the MPA can be implemented based on a simpler graphical
model. A hybrid detection scheme is proposed in [17], where The integrated sensing and communication (ISAC) technol-
both maximum a posteriori (MAP) detection and parallel ogy, which allows the share of hardware architectures, signal
interference cancellation (PIC) detection are considered ac- processing algorithms, and frequency resources, is capable of
cording to the channel coefficients. Simulation results show improving both power and spectral efficiency while fulfilling
that this hybrid detection can approach the error performance sensing and communication functionalities. As for the radar
of near-optimal symbol-wise MAP detection, especially for sensing, the estimation of delays and Dopplers associated
coded OTFS systems, but only requires a reduced complexity. with targets is an essential task. From the communication
In [18], an unitary approximate message passing (UAMP)- perspective, the OTFS modulation also places symbols in the
based approach is proposed, which not only enjoys a strong DD domain. Therefore, it is natural to consider the adoption
resilience to the error propagation but also enables an efficient of OTFS waveform for ISAC. Some early contributions on
implementation. OTFS-ISAC can be found in [20], [21] and references therein.
Recent development of OTFS transmission has shown that Nevertheless, the inherent relationship between the OTFS-
detecting OTFS signals crossing different domains could be a based sensing and communication still needs to be unveiled.

Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:17:47 UTC from IEEE Xplore. Restrictions apply.
442
 2021 IEEE/CIC International Conference on Communications in China (ICCC Workshops)

B. Underwater Communications [3] F. Hlawatsch and G. Matz, Wireless Communications over Rapidly Time-
varying Channels. Academic Press, 2011.
Different to wireless radio channels, underwater acoustic [4] P. Raviteja, Y. Hong, E. Viterbo, and E. Biglieri, “Effective diversity
channels (UACs) suffer from long delay spreads up to hun- of OTFS modulation,” IEEE Wireless Commun. Lett., vol. 9, no. 2, pp.
dreds of milliseconds due to the low speed of sound (about 249–253, Feb. 2020.
[5] S. Li, J. Yuan, W. Yuan, Z. Wei, B. Bai, and D. W. K. Ng, “Performance
1500m/s) and severe Doppler spread caused by internal waves, analysis of coded OTFS systems over high-mobility channels,” to appear
platform and sea-surface motions [22]. Compared with the in IEEE Trans. Wireless Commun., 2021.
conventional TF domain modulation schemes, OTFS modu- [6] A. F. Molisch, Wireless Communications. John Wiley & Sons, 2012,
vol. 34.
lation might be beneficial for efficient channel estimation and [7] P. Bello, “Characterization of randomly time-variant linear channels,”
equalization by exploiting the DD domain channel sparsity. IEEE Trans. Commun. Syst., vol. 11, no. 4, pp. 360–393, Dec. 1963.
Moreover, UAC have inherent wideband characteristics where [8] A. Farhang, A. RezazadehReyhani, L. E. Doyle, and B. Farhang-
Boroujeny, “Low complexity modem structure for OFDM-based orthog-
the Doppler effect is actually a Doppler-scale effect instead of onal time frequency space modulation,” IEEE Wireless Commun. Lett.,
a simple Doppler shift. In fact, in UACs, the received signal vol. 7, no. 3, pp. 344–347, Jun. 2018.
is a time-delayed and time-scaled version of the transmitted [9] Cohere, “6G technology,” https://www.cohere-technologies.com/
technology/6g/.
signal caused by a moving scatterer [1]. Hence, how to exploit [10] P. Raviteja, K. T. Phan, Y. Hong, and E. Viterbo, “Interference can-
the potential multipath-scale diversity of UAC channels with cellation and iterative detection for orthogonal time frequency space
OTFS modulation is a potential research direction. modulation,” IEEE Trans. Wireless Commun., vol. 17, no. 10, pp. 6501–
6515, Oct. 2018.
C. Satellite and Space Communications [11] D. Tse and P. Viswanath, Fundamentals of Wireless Communication.
Cambridge university press, 2005.
Extremely high-mobility is the principal characteristic for [12] Z. Wei, W. Yuan, S. Li, J. Yuan, and D. W. K. Ng, “Performance analysis
the satellite and space communications. For example, the low- and window design for channel estimation of OTFS modulation,” in
Proc. IEEE Int. Commun. Conf., 2021, pp. 1–6.
earth-orbit (LEO) satellite system, which is recognized as the [13] Z. Wei, W. Yuan, S. Li, J. Yuan, and D. W. K. Ng, “Transmitter
enabling technology for ubiquitous access in the future mobile and receiver window designs for orthogonal time frequency space
networks, will have a speed of 10, 000 km/h. Such a high modulation,” IEEE Trans. Commun., vol. 69, no. 4, pp. 2207–2223,
Apr. 2021.
speed will inevitably introduce very high Doppler shifts, which [14] P. Raviteja, K. T. Phan, and Y. Hong, “Embedded pilot-aided channel
is challenging for reliable information transmission. A recent estimation for OTFS in delay-Doppler channels,” IEEE Trans. Veh.
work [23] has analyzed the performance of OTFS modulation Technol., vol. 68, no. 5, pp. 4906–4917, May 2019.
[15] W. Yuan, S. Li, Z. Wei, J. Yuan, and D. W. K. Ng, “Data-aided
under high-mobility satellite-to-ground channels. However, the channel estimation for OTFS systems with a superimposed pilot and
commonly used satellite and space communication schemes, data transmission scheme,” to appear in IEEE Wireless Commun. Lett.,
e.g., spectrum spreading are not well understood and the 2021.
[16] W. Yuan, Z. Wei, J. Yuan, and D. W. K. Ng, “A simple variational Bayes
application of OTFS modulation needs investigation. detector for orthogonal time frequency space (OTFS) modulation,” IEEE
Trans. Veh. Technol., vol. 69, no. 7, pp. 7976–7980, Apr. 2020.
D. High-Speed Train Communications [17] S. Li, W. Yuan, Z. Wei, J. Yuan, B. Bai, D. W. K. Ng, and Y. Xie,
“Hybrid MAP and PIC detection for OTFS modulation,” to appear in
High-mobility and complex transmission scenarios are two IEEE Trans. Veh. Technol., 2021.
of the important challenges for HST communications. In par- [18] Z. Yuan, F. Liu, W. Yuan, Q. Guo, Z. Wang, and J. Yuan, “Iterative
ticular, the frequent changes of the HST-specific environments, detection for orthogonal time frequency space modulation using ap-
proximate message passing with unitary transformation,” arXiv preprint
such as viaducts, cuttings, and tunnels, have imposed a great arXiv:2008.06688, 2020.
difficulty for robust communications. In those cases, OTFS [19] S. Li, W. Yuan, Z. Wei, and J. Yuan, “Cross domain iterative detec-
modulation may be a potential solution because it is highly tion for orthogonal time frequency space modulation,” arXiv preprint
arXiv:2101.03822, 2021.
symmetrical to any channel distortions caused by delay and [20] L. Gaudio, M. Kobayashi, G. Caire, and G. Colavolpe, “On the effective-
Doppler spreads. However, how to design OTFS parameters, ness of OTFS for joint radar parameter estimation and communication,”
such as delay and Doppler resolutions, requires extensive IEEE Trans. Wireless Commun., vol. 19, no. 9, pp. 5951–5965, Sep.
2020.
channel measurements and investigations [24]. [21] W. Yuan, Z. Wei, S. Li, J. Yuan, and D. W. K. Ng, “Integrated sensing
and communication-assisted orthogonal time frequency space transmis-
V. C ONCLUSIONS sion for vehicular networks,” IEEE J. Sel. Topics. Signal Process., arXiv
In this paper, we have discussed the related subjects of preprint arXiv:2105.03125, 2021.
[22] C. R. Berger, S. Zhou, J. C. Preisig, and P. Willett, “Sparse channel
OTFS modulation, including the channel representations, sym- estimation for multicarrier underwater acoustic communication: From
bol model, performance analysis, and potential applications subspace methods to compressed sensing,” IEEE Trans. Signal Process.,
for wireless communications. We have also summarized some vol. 58, no. 3, pp. 1708–1721, Dec. 2009.
[23] T. Li, R. He, B. Ai, M. Yang, Z. Zhong, and H. Zhang, “OTFS
potential research topics for OTFS modulation. modulation performance in a satellite-to-ground channel at sub-6-GHz
and millimeter-wave bands with high mobility,” Front. Inf. Technol. &
R EFERENCES Elec. Engineer., vol. 22, no. 4, pp. 517–526, Apr. 2021.
[1] Z. Wei, W. Yuan, S. Li, J. Yuan, G. Bharatula, R. Hadani, and [24] R. He, B. Ai, G. Wang, K. Guan, Z. Zhong, A. F. Molisch, C. Briso-
L. Hanzo, “Orthogonal time-frequency space modulation: A promising Rodriguez, and C. P. Oestges, “High-speed railway communications:
next generation waveform,” to appear in IEEE Wireless Commun., 2021. From GSM-R to LTE-R,” IEEE Veh. Technol. Mag., vol. 11, no. 3, pp.
[2] R. Hadani, S. Rakib, M. Tsatsanis, A. Monk, A. J. Goldsmith, A. F. 49–58, Sept. 2016.
Molisch, and R. Calderbank, “Orthogonal time frequency space mod-
ulation,” in Proc. IEEE Wireless Commun. Net. Conf., Mar. 2017, pp.
1–6.

Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:17:47 UTC from IEEE Xplore. Restrictions apply.
443