14 IEEE COMMUNICATIONS LETTERS, VOL. 27, NO.

1, JANUARY 2023

Orthogonal Time Frequency Space Modulation—

Part III: ISAC and Potential Applications
Weijie Yuan , Member, IEEE, Zhiqiang Wei , Member, IEEE, Shuangyang Li , Member, IEEE,
Robert Schober , Fellow, IEEE, and Giuseppe Caire , Fellow, IEEE
(Invited Paper)

Abstract— The first two parts of this tutorial on orthogonal mirrors the geometry of the scatterings comprising the wireless
time frequency space (OTFS) modulation have discussed the channel. In future wireless networks, both sensing and com-
fundamentals of delay-Doppler (DD) domain communications as munications functionalities are highly desired, leading to the
well as some advanced technologies for transceiver design. In this
letter, we will present an OTFS-based integrated sensing and
integrated sensing and communications (ISAC) technology
communications (ISAC) system, which is regarded as an enabling by using the same signaling waveform and hardware archi-
technology in next generation wireless communications. In par- tecture [2]. In contrast to the OFDM-based ISAC system,
ticular, we illustrate the sensing as well as the communication OTFS-ISAC system provides the direct interaction between
models for OTFS-ISAC systems. Next, we show that benefiting transmitted signals and the channel responses in a unified
from time-invariant DD channels, the sensing parameters can DD domain for both functionalities, on top of the resilience
be used for inferring the communication channels, leading to
an efficient transmission scheme. As both functionalities are to delay and Doppler spreads. To this end, OTFS has been
realized in the same DD domain, we briefly discuss several regarded as a promising signal waveform to fully unleash
promising benefits of OTFS-based ISAC systems, which have the potentials of ISAC system. The pioneering work [3]
not been completely unveiled yet. Finally, a range of potential showed that OTFS is more robust for velocity estimation
applications of OTFS for the future wireless networks will compared to OFDM scheme. Some recent works discussed
be highlighted. the ISAC performance limits, the beamforming designs, the
Index Terms— OTFS, integrated sensing and communications resource allocation schemes, and the receiver structures for
(ISAC), delay-Doppler domain. OTFS-based ISAC systems [4], [5], [6], [7]. Although all these
contributions have demonstrated the effectiveness of OTFS for
I. I NTRODUCTION ISAC system, several promising benefits of OTFS-based ISAC
systems have not been well discussed in the literature.

T HE orthogonal time frequency space (OTFS) modulation

has gained considerable attentions due to its capability
of providing more reliable communications compared to the
In this letter, we will first study the OTFS-based ISAC
model in the DD domain. It is noted that the delay and
Doppler associated with the DD domian communication
existing orthogonal frequency division multiplexing (OFDM) channel are half of the round-trip delay and Doppler para-
modulation, especially in high-mobility environments [1]. meters extracted from the sensing echoes. This motivates
As overviewed in Part I of this tutorial, OTFS utilizes the us that the sensing parameters can be used for predicting
delay-Doppler (DD) domain for data multiplexing, which the communication channels given the slow time-varying
property of the DD domain channel. Afterwards, we briefly
Manuscript received 11 September 2022; accepted 21 September 2022.
Date of publication 26 September 2022; date of current version 9 January discuss several benefits of OTFS-ISAC system, i.e., unified
2023. This work is supported in part by National Natural Science Foundation signaling design framework, sensing channel exploitation, and
of China under Grant 62101232, and in part by the Guangdong Provincial communications-assisted sensing. Finally, we discuss potential
Natural Science Foundation under Grant 2022A1515011257. The work of
R. Schober and G. Caire is supported by BMBF Germany in the program of applications related to OTFS modulation.
“Souverän. Digital. Vernetzt.” Joint Project 6G-RIC (Project IDs 16KISK030 Notations:(·)H and (·)T denotes the Hermittian operation
and 16KISK023). The associate editor coordinating the review of this letter and the transpose operation, respectively; (·)N denotes the
and approving it for publication was A. Zappone. (Corresponding author:
Zhiqiang Wei.)
modulus operation w.r.t. N ; IMN denotes an identity matrix
Weijie Yuan is with the Department of Electrical and Electronic Engineer- of dimension M N ; δ (·) denotes the Dirac delta function;
ing, Southern University of Science and Technology, Shenzhen 518055, China E[·], Tr[·], and det[·] denote the expectation operation, trace
(e-mail: [email protected]). operation, and the determinant of a matrix, respectively.
Zhiqiang Wei is with the School of Mathematics and Statistics, Xi’an
Jiaotong University, Xi’an 710049, China (e-mail: [email protected]).
Shuangyang Li was with the School of Electrical Engineering and Telecom- II. OTFS-ISAC S YSTEM M ODEL
munications, University of New South Wales, Sydney, NSW 2052, Australia.
He is now with the Department of Electrical, Electronic, and Computer Let us consider a general system model with a mono-static
Engineering, University of Western Australia, Perth, WA 6009, Australia base station (BS) and P targets of interests, as shown in Fig. 1.
(e-mail: [email protected]).
Robert Schober is with the Institute for Digital Communications (IDC),
For ease of exposition, we assume that all targets of interest are
the Friedrich-Alexander University Erlangen–Nuremberg, 91054 Erlangen, communications user equipments (UEs) as well, which aligns
Germany (e-mail: [email protected]). well with the use case such as vehicular networks. The BS is
Giuseppe Caire is with the Department of Electrical Engineering and equipped with a transmit uniform linear array (ULA) of Nt
Computer Science, Technical University of Berlin, 10587 Berlin, Germany
(e-mail: [email protected]). antennas and a separate receive ULA of Nr antennas. Under
Digital Object Identifier 10.1109/LCOMM.2022.3209651 the assumption of sufficient isolation between the transmit
1558-2558 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:14:14 UTC from IEEE Xplore. Restrictions apply.
YUAN et al.: ORTHOGONAL TIME FREQUENCY SPACE MODULATION—III: ISAC AND POTENTIAL APPLICATIONS 15


with channel gain hi = c
4πfc d2i
, signal propagation speed c,
carrier frequency fc , range distance di , delay τi , and Doppler
shift νi . After passing through the channel [8], the received
signal of the i-th target can be modelled as
yi (t) = hi uH H
i aNu (θi )aNt (θi )fi si (t − τi )e
j2πνi (t−τi )
+ zi (t),
(5)
where ui ∈ CNu ×1 is the receive beamformer, zi (t) is the
time domain noise signal. We adopt a receive filter grx (t)
and after OTFS demodulation, the received OTFS signal is
expressed as1
i
YDD [l, k] = hi uH H
i aNu (θi)aNt (θi)fi XDD [(l − li )M , (k − ki )N]
i

+ ZDD
i
[l, k] , (6)
Fig. 1. OTFS-ISAC system model.
where the integers ki = νi N T , li = τi M Δf , is the i
ZDD [l, k]
additional white Gaussian noise sample with a power spectral
and receive arrays, the sensing echoes would not interfere density of N0 .
with the transmitted signals. For receiving the communication
information, each target is equipped with Nu antennas. B. Sensing Model
The sensing channel is given by
A. Communication Model 
K
H(t, τ ) = γi aNr (θi )aH
Nt (θi )δ(τ − ηi )e
j2πυi t
, (7)
Without loss of generality, we consider a DD domain grid i=1
of M × N , where l ∈ {0, . . . , M − 1} and k ∈ {0, . . . , N − 1} where γi , ηi , and υi are the reflected coefficient, the delay, and
denotes the delay and Doppler indices, respectively. The infor- the Doppler associated with the i-th target. Then, the acquired
mation symbols sent to the i-th target are defined as XDDi
[l, k], sensing echoes at the BS can be expressed as
which is transformed to the time-frequency (TF) domain via
inverse symplectic finite Fourier transform (ISFFT), yielding 
K
r(t) = γi aNr (θi )aH
Nt (θi )s̃(t − ηi )e
j2πυi t
+ w(t), (8)
XTFi
[m, n], where m and n denote the indices of subcarriers
i=1
and time slots. By performing multicarrier modulation, the
transmitted signal to the i-th target is given by where w(t) is the measurement noise. For massive MIMO
receive antenna arrays, steering vectors with different

N −1 M−1
 angular values are asymptotically orthogonal [10], i.e.,
si (t) = i
XTF [m, n] gtx (t − nT )ej2πmΔf (t−nT ) ,
aH
Nr (θi )aNr (θi ) ≈ 0, ∀θi = θi . Thus, the interference
n=0 m=0
between different targets in the sensing echoes can be
(1)
neglected and the BS can distinguish different targets in terms
where gtx (t) is the transmit shaping pulse, T and Δf denote of their angles-of-arrival (AoAs). As a result, the sensing echo
the symbol duration and frequency spacing, respectively, of the i-th target can be extracted from r(t) with a receive
satisfying T Δf = 1 to maintain the orthogonality. beamformer bi = aNr (θ̃i ), expressed as
Initially at the target detection stage, the BS could use the
ri (t) = γi bH H
i aNr (θi )aNt (θi )fi si (t − ηi )e
j2πυi t
+ w(t). (9)
antenna array to formulate a wide beam or an omni-directional
signal to detect all targets of interests. Afterwards, in the We note that angular parameter θi can also be inferred from the
tracking mode, the BS formulates multiple beams for informa- receive beamforming process by comparing the gains obtained
tion transmission and target tracking. Through a beamforming for different beam directions. Hence, (9) is simplified as
matrix F ∈ CNt ×P , the multi-beam signal is expressed as ri (t) = Ga si (t − ηi )ej2πυi t + wi (t), (10)
s̃(t) = Fs(t), (2) where Ga is the composite antenna array gain. By adopting
T
where s(t) = [s1 (t), . . . , sP (t)] is the transmitted signal an ideal receive filter and the OTFS demodulation, we finally
 all P targets. The i-th column of F, denoted by fi =
to arrive at the DD domain input-output relationship,
Nt aNt (θ̃i ) is used to allocate power pi and to steer the
pi −1 M−1

N 
i
RDD [l, k] = Ga i
HDD [l , k  ]
transmitted signal towards the intended direction θ̃i , where
k =0 l =0
the steering vector is given by
 T ·XDD
i
[(l − l )M , (k − k  )N ] + WDD
i
[l, k] , (11)
aNt (θi ) = 1, ejπ sin θi , . . . , ej(Nt −1)π sin θi . (3) where HDDi
[l , k  ] denotes the gain of the i-th target at
Due to the asymptotic orthogonality of the massive antenna the DD grid (bin) with indices l , k  corresponding to the
l
array, after transmit beamforming, the communication channel delay of MΔf and Doppler of NkT . By stacking all received
to the i-th target is asymptotically line-of-sight (LoS) domi- samples RDD in a vector riDD [l, k], (11) is rewritten in a
i

nated, and given by 1 For simplicity, we assume integer delay and Doppler shifts in this letter.
CiDD (τ, ν) = hi aNu (θi )aH
Nt (θi )δ(τ − τi )δ(ν − νi ), (4) For the fractional case, off-grid estimation schemes can be adopted, e.g., [9].

Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:14:14 UTC from IEEE Xplore. Restrictions apply.
16 IEEE COMMUNICATIONS LETTERS, VOL. 27, NO. 1, JANUARY 2023

matrix form as The predicted angle θ̃i can be used for steering the trans-
riDD = Ga XiDD hiDD + wDD
i
, (12) mit and receive antennas before transmitting the multi-beam
OTFS-ISAC signal and receiving the echoes. Compared to
 
where the (N k + l )-th element in hiDD is HDD
i
[l , k  ]. Based the conventional beam pairing and alignment schemes, using
on the estimate of sensing channel hDD , it is able to extract
i
the predicted angle avoids the adoption of dedicated pilots
the delay and Doppler associated to the i-th target. for estimating the angular parameters, which can improve the
resource efficiency. The predicted angle can also be contained
III. P RELIMINARY R ESULTS AND D ISCUSSIONS
in the signal to targets, which is used for designing the
A. Sensing Parameters Estimation beamformer ui [11], [12].
The classic matched filtering scheme in radar community In addition, the prediction of the range distance is obtained
can be used for obtaining the estimate of hiDD , given by as d˜i = p̃i . Consequently, the downlink channel gain hi can
r̃iDD = Xi,H i i,H i i i,H i
DD rDD = Ga XDD XDD hDD + XDD wDD . (13)
be predicted by replacing di by d˜i . Moreover, the delay and
Doppler indices in the downlink OTFS communications signal
Obviously, (13) is a linear model and it is straightforward to
model are determined as well by exploiting the predicted
find an estimator, to obtain the estimate ĥiDD . Such estimators
range and speed. Motivated by the fact that all communication
are Cramér-Rao bound (CRB) achieving if the elements of
channel parameters can be predicted at the BS side, it is
hiDD are also Gaussian distributed. However, hiDD corresponds
capable of compensating the channel effects before downlink
to the round-trip delay and Doppler shifts of the i-th target,
transmission. Provided sufficiently accurate prediction of the
indicating that hiDD is a sparse vector. The delay and Doppler
communication channel and compensation at the BS, the
indices can be determined by finding element in ĥiDD that received OTFS signal is simplified as
has the largest response, i.e., [l̂i , k̂i ], yielding the delay and
Doppler estimates η̂i and φ̂i , respectively.2 Due to the exis-
i
YDD [l, k] = Gt XDD
i
[l, k] + ZDD
i
[l, k] , (18)
tence of noise, the above estimation is biased in the low signal- with Gt being the composite array gain. It is observed
to-noise ratio (SNR) regime, which has been reported in [4]. from (18) that the communications receiver can perform data
detection while bypassing the channel estimation.
B. Sensing-Assisted OTFS Communications To verify the effectiveness of the proposed scheme, we show
From the estimates of delay η̂i , Doppler φ̂i , and angle θ̂i , the bit-error-rate (BER) performance versus the SNR based
we are now able to determine the location as well as the speed on the proposed scheme in Fig. 2. The size of OTFS frame
of target i. Assuming that the BS is located at the origin is M = 128 and N = 20. The BS is operating at a carrier
[0, 0]T , the location of the i-th target pi = [px,i , py,i ]T is frequency of 3 GHz and the subcarrier spacing is 6 kHz.
written as Therefore, the occupied bandwidth is 768 kHz. The radar
cη̂i sin θ̂i cross section (RCS) of the target is assumed to be 25, given
p̂x,i = , (14)
2 the reflection area of the target being 1m2 . Finally, the transmit
cη̂i cos θ̂i power is set to 40 dBm. Moreover, we consider the binary
p̂y,i = . (15)
2 phase shift keying (BPSK) symbol mapping. The speed of the
Similarly, the velocity of the i-th target v̂i = [v̂x,i , v̂y,i ]T can target is randomly generated from the uniform distribution
be obtained based on φ̂i and θ̂i . [10, 15] m/s. For comparison purpose, the BER performance
According to the definition of OTFS channel model, the corresponding to the classic pilot-based beam alignment and
Doppler νi is in theory half of the Doppler associated with channel estimation scheme, and the ideal case with perfect
the sensing echo φi . The slow time-varying property of knowledge of channel are illustrated. We can see that our
the OTFS communication channel motivates us to reuse the proposed algorithm can approach the performance of the ideal
sensing parameters for communications purpose. Relying on case, which implies the highly accurate channel prediction.
the estimates of target locations as well as the speeds, the Nevertheless, due to the random noise, the predicted channel
locations of the targets in the following time instant are may deviate from the actual channel, resulting in a small BER
predictable. In particular, the predicted location of the i-th performance gap. Compared to the classic pilot-based scheme,
target can be expressed as the proposed algorithm yields a significant BER performance
gain thanks to a high SNR gain. Despite the performance
p̃i = p̂i + v̂i ΔT, (16)
improvement, the proposed algorithm also reduces the com-
where ΔT is the duration of one transmission block. For the munication overhead as well as latency of the classic scheme,
speeds of the targets, it is naturally to set the predicted velocity showing the advantage of adopting OTFS-ISAC signal.
ṽi identical to the estimated velocity, since the target speed
will not have a burst change in a relatively short duration. The C. Discussions of OTFS-ISAC System
predicted locations of targets enable us to further obtain the So far, we have proposed an example of sensing-assisted
prediction of the angular parameter θi in the following time communications protocol for OTFS-ISAC systems. We show
instant, based on the geometric relationship as that benefiting from the slow-varying OTFS channel, the
p̃y,i parameters extracted from sensing echoes can be exploited for
θ̃i = arctan . (17)
p̃x,i predicting the downlink DD domain communication channel.
As a result, no pilots are required either for beam alignment
2 In addition to the adopted estimator, the sensing accuracy also depends on
or for channel estimation, leading to a reduced communi-
the resolutions of the delays and Doppler shifts, which are associated with
the bandwidth and signal duration. Therefore, the OTFS frame size, i.e., M cation overhead. Moreover, since the angle feedback and
and N will also affect the estimation performance. channel
Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on estimation
May 16,2024 areUTC
at 12:14:14 bypassed,
from IEEEthe communications
Xplore. Restrictions apply.receiver
YUAN et al.: ORTHOGONAL TIME FREQUENCY SPACE MODULATION—III: ISAC AND POTENTIAL APPLICATIONS 17

where C1 indicates the acceptable sensing performance and

C2 constraints the power consumption. A possible solution of
(23) requires a specific waveform design, which can achieve
the optimal tradeoff between sensing and communication via
taking into account their respectively optimal waveforms in
the DD domain. Moreover, the objective of (23) can be
extended to include both communication and sensing metrics
for optimizing OTFS signal that can strike a good trade-off
between both functionalities. Instead of directly designing
xDD , we can also design the DD domain precoding matrix
PDD through optimization given certain constraints such that
the transmitted OTFS signal becomes PDD xDD
2) Sensing Channel Exploitation: As discussed in
Section III, the delay and Doppler parameters associated with
Fig. 2. The BER performance of different algorithms for uplink transmission. the sensing channel are reused for OTFS communication
channel prediction. In this case, we assume that all targets of
is capable of directly detecting the data symbols, which is interest are communication users as well. However, this is not
attractive for supporting ultra low latency communications in always the case in other scenarios, e.g., only part of the targets
the future networks. Relying on the same DD domain for both in the area of coverage are UEs. Therefore, a fundamental
communications and sensing, OTFS offers more promising question is how much information we can glean from the
benefits for ISAC systems, which will be briefly discussed sensing channel Hs,DD for inferring Hc,DD . As the first
in the following. step, we need to characterize the similarity between OTFS
1) Unified Signaling Design Framework: Let us consider a communication and sensing channels. A nature choice is to use
general single antenna case. When transmitting a vectorized the cross-correlation matrix Rcs = E Hc,DD HH s,DD and the
OTFS signal xDD , after OTFS demodulation, the acquired entries of Rcs will provide the correlation between sensing tar-
sensing echo and the received OTFS signal can be written gets and communication reflectors corresponding to different
in concise forms as paths. In particular, when sensing and communication channels
are highly correlated, the matrix Rcs will be diagonally
rDD = Hs,DD xDD + wDD , (19) dominant. Although OTFS modulation offers the opportunity
yDD = Hc,DD xDD + zDD , (20) to connect two channels in the same DD domain, new analysis
where Hs,DD and Hc,DD denotes the sensing and commu- tools and framework are still to be developed to further study
nication channels in DD domain, respectively. For sensing, the exploitation of sensing channel in OTFS-ISAC system.
the knowledge of xDD at the BS is exploited for inferring 3) Communications-Assisted Sensing: The OTFS channel
the sensing parameters η from channel Hs,DD . In contrast, model reveals the underlying geometry of the wireless
for communications, the transmitted data symbols xDD are propagation environments. After channel estimation, the
detected. The sensing performance is characterized by ambi- estimated delay and Doppler indices can be converted
guity function and the CRB while the communication perfor- to range distance and speed. Given a temporal sequence
mance can be evaluated using channel capacity. In particular, of estimated delays corresponding to the i-th path, i.e.,
the CRB of the sensing parameters η is derived as [13], τ i = [τ̂i,1 , . . . τ̂i,t ]T , the location of the i-th reflector can be
  −1 obtained via an ML estimator, i.e.,
∂ 2 ln p(rDD |η)
CRBη = Tr E − , (21) p̂i = arg max p(τ i |pi ), (24)
∂η 2 pi

where p(rDD |η) is the likelihood function, and the capacity where p(τ i |pi ) is the likelihood function of measured delays
of the communication channel CHc can be expressed as conditioned on the location of the i-th reflector. If the
reflector is also a moving target, Bayesian estimator such
1
CHc = log2 det IMN + Hc,DD Rx HH c,DD , (22) as extended Kalman filtering (EKF) can be used to exploit
N0 the state transition information. By combining the sensing
where Rx is the covariance matrix of the transmitted signal. information gleaned from the reflected echoes at BS and from
The OTFS waveform enables us to formulate the signal- the OTFS channel estimation, we can obtain better knowledge
ing design problem in the same DD domain for achieving of the environment.
satisfactory sensing and communications performance. For
example, in a communication-centric scenario, we aim for IV. P OTENTIAL A PPLICATIONS OF OTFS
maximizing the channel capacity CHc while maintaining the In addition to unifying communications and sensing in
sensing performance. Then, it is straightforward to formulate the same DD domain, OTFS is also attractive for several
the following optimization problem other emerging applications, due to its advantages of low
complexity implementation, sparse channel representation, and
max CHc high resilience to large delay and Doppler spread.
xDD
s.t. C1 : CRBη ≤ TCRB A. Underwater Communications
1 Due to the propagation medium in underwater communica-
C2 : xDD  = PT , (23)
MN tions, mechanical waves, e.g., acoustic wave and seismic wave
Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:14:14 UTC from IEEE Xplore. Restrictions apply.
18 IEEE COMMUNICATIONS LETTERS, VOL. 27, NO. 1, JANUARY 2023

instead of the radio signal, are usually used for information will result in complex signal transmission scenarios. For exam-
transmission. The relatively low signal propagation speed ple, in high-speed train communications, the environments will
of mechanical waves results in large delay and Doppler change frequently, such as viaducts, cuttings, and tunnels,
spreads for underwater communication channels. In general, which imposes great difficulty for robust communications.
the delay could spread up to hundreds milliseconds, which Relying on the quasi-periodic property, OTFS modulation can
is much higher than typical wireless radio channels having support mobile users having a very wide range of speeds.
the delay level of μs. The Doppler spread is also severe Moreover, it is highly symmetrical to any channel distortions
due to internal waves and the movements of communication caused by delay and Doppler spreads, making OTFS a good
transceivers. Moreover, multiple reflectors from the sea-surface solution for realizing the SAGIN. Some recent contributions
and sea floor lead to complex multi-path propagation environ- have applied OTFS in LEO system and have verified its
ments. On the other hand, the inherent wideband propagation effectiveness [15]. However, how to design OTFS parameters,
in underwater communications will introduce Doppler-scale such as delay and Doppler resolutions, requires extensive
effect instead of a simple Doppler shift. To characterize channel measurements and investigations.
different delays and Doppler scales, the multiscale multilag
(MSML) channel model is usually used, expressed as V. C ONCLUSION

P
In the last part of this tutorial on OTFS modulation, we aim
h(t, τ ) = hi δ(τ − τi − αi t)e−jαi fc t , (25) for unveiling its great potentials of supporting future networks.
i=1 We first studied the OTFS-based ISAC system, which shows
where P is the number of paths, αp is the time scaling the superiority of sensing-assisted OTFS communications.
parameter. In fact, in underwater environments, the received Then several possible benefits of adopting OTFS as the
signal is a time-delayed, time-scaled, and phase-rotated ver- ISAC signaling waveform were discussed. As a further step,
sion of the transmitted signal caused by moving scatterings. we summarized several OTFS modulation-related applications.
Although OFDM has been applied in underwater communica-
tions because of its robustness against long delay spread, the R EFERENCES
inter-carrier interference (ICI) induced by multiple Dopplers [1] Z. Wei et al., “Orthogonal time-frequency space modulation: A promis-
will deeply degrade the performance. Compared with OFDM ing next-generation waveform,” IEEE Wireless Commun., vol. 28, no. 4,
modulation, OTFS modulation could benefit from the possible pp. 136–144, Aug. 2021.
[2] F. Liu et al., “Joint radar and communication design: Applications, state-
sparsity and stability of the DD domain channel response, of-the-art, and the road ahead,” IEEE Trans. Commun., vol. 68, no. 6,
enabling efficient channel estimation and equalization. pp. 3834–3862, Jun. 2020.
[3] P. Raviteja et al., “Orthogonal time frequency space (OTFS) modulation
B. Optical Transmission based radar system,” in Proc. IEEE Radar Conf. (RadarConf), Apr. 2019,
pp. 1–6.
The optical spectrum offers huge and unregulated band- [4] L. Gaudio et al., “On the effectiveness of OTFS for joint radar parameter
width without electromagnetic interference. Information sym- estimation and communication,” IEEE Trans. Wireless Commun., vol. 19,
no. 9, pp. 5951–5965, Sep. 2020.
bols are transmitted via modulating the luminous intensity [5] W. Yuan et al., “Integrated sensing and communication-assisted orthog-
of light-emitting diodes (LED) while the received signal is onal time frequency space transmission for vehicular networks,”
converted to electrical signal using photodiodes. In optical IEEE J. Sel. Topics Signal Process., vol. 15, no. 6, pp. 1515–1528,
Nov. 2021.
transmission, the channel is generally modeled as static [6] S. Li et al., “A novel ISAC transmission framework based on spatially-
multi-path one. While OTFS is recognized for its superi- spread orthogonal time frequency space modulation,” IEEE J. Sel. Areas
ority in high Doppler scenarios, its capability of exploiting Commun., vol. 40, no. 6, pp. 1854–1872, Jun. 2022.
full time-frequency diversity will guarantee its performance [7] K. Wu et al., “OTFS-based joint communication and sensing for future
industrial IoT,” IEEE Internet Things J., early access, Dec. 31, 2021,
in static channels as well. Recent results have shown that doi: 10.1109/JIOT.2021.3139683.
OTFS-based optical transmission system achieves a better [8] W. Yuan et al., “A simple variational Bayes detector for orthogonal
performance than the conventional OFDM-based counterpart time frequency space (OTFS) modulation,” IEEE Trans. Veh. Technol.,
vol. 69, no. 7, pp. 7976–7980, Jul. 2020.
with a single LED [14]. It is worth to extend the research [9] Z. Wei et al., “Off-grid channel estimation with sparse Bayesian learning
to multi-LED and multi-photodiodes systems based on for OTFS systems,” IEEE Trans. Wireless Commun., vol. 21, no. 9,
OTFS modulation. pp. 7407–7426, Sep. 2022.
[10] T. L. Marzetta, Fundamentals of Massive MIMO. Cambridge, U.K.:
Cambridge Univ. Press, 2016.
C. Space-Air-Ground Integrated Networks (SAGIN) [11] F. Liu et al., “Radar-assisted predictive beamforming for vehicular links:
To provide ubiquitous connectivity and global coverage, the Communication served by sensing,” IEEE Trans. Wireless Commun.,
vol. 19, no. 11, pp. 7704–7719, Nov. 2020.
future communications will be SAGIN. The low-earth-orbit [12] W. Yuan et al., “Bayesian predictive beamforming for vehicular net-
(LEO) satellite communications, the aerial communications, works: A low-overhead joint radar-communication approach,” IEEE
and the vehicular communications will extend the coverage Trans. Wireless Commun., vol. 20, no. 3, pp. 1442–1456, Mar. 2021.
[13] S. M. Kay, Fundamentals of Statistical Signal Processing: Estimation
range of the current deployed terrestrial networks. On the one Theory. Upper Saddle River, NJ, USA: Prentice-Hall, 1993.
hand, extremely high-mobility is the principal characteristic [14] A. Sharma et al., “Performance analysis of OTFS over mobile multipath
for the satellite and space communications. Such a high speed channels for visible light communication,” in Proc. IEEE Region 10
will inevitably introduce a very high Doppler shift, which is Conf. (TENCON), Nov. 2020, pp. 490–495.
[15] J. Shi et al., “Outage probability for OTFS based downlink LEO
challenging for reliable information transmission. On the other satellite communication,” IEEE Trans. Veh. Technol., vol. 71, no. 3,
hand, the networking with different communication devices pp. 3355–3360, Mar. 2022.

Authorized licensed use limited to: UNIVERSITE DE GABES. Downloaded on May 16,2024 at 12:14:14 UTC from IEEE Xplore. Restrictions apply.