Computer Networks 224 (2023) 109597

Contents lists available at ScienceDirect

Computer Networks
journal homepage: www.elsevier.com/locate/comnet

Review article

A review on orthogonal time–frequency space modulation: State-of-art,

hotspots and challenges
Mao Li, Wei Liu ∗, Jing Lei
College of Electronic Science, National University of Defense Technology, Changsha 410073, China

ARTICLE INFO ABSTRACT

Keywords: High-speed mobility communication is one of the challenging scenarios because of the rapidly changing channel
High-speed mobility involved. In light of this, the orthogonal time–frequency space (OTFS) modulation technique is explored for
OTFS modulation meeting the high-Doppler communication requirements. This paper provides a comprehensive overview of
Time–frequency domain
the critical application technologies and state-of-art approaches to OTFS modulation. First, we summarize the
Delay–Doppler domain
input–output expression relation under the different domains and the interference analysis under different
Open issues and challenges
waveforms, which is significant for the follow-up research. Secondly, the comparison between the existing
modulation methods and OTFS modulation is discussed, which provides a reference for the improvement of
the OTFS system. Next, detailed and exhaustive reviews of current OTFS-based hotpots are provided, including
channel estimation, signal detection, etc. Further, the OTFS technologies are applied to specific application
scenarios and the corresponding advantages are analyzed. Based on the above research status, we present
our vision on the research trends and open issues, highlighting the opportunity for OTFS to be a promising
next-generation candidate in future communication.

1. Introduction The high-speed mobility scenario is one of the targets of future

communication systems, in which wireless channels behave the char-
With the continuous improvement of social science and technology, acteristics of time selective fading and frequency selective fading [7],
the mobile communication system has developed from 1G in the past as shown in Fig. 1. In the communication process, the data trans-
to beyond 5G [1] today, and then to a 6G system [2] in the coming mission way needs to show high performance in different scenarios
decades. The development of mobile communication not only pro- with different Doppler spread, time delays, and carrier frequencies.
foundly affects lifestyle but also becomes a vital engine to drive social Orthogonal frequency division multiplexing (OFDM) [8] is one of the
progress. Increasing demands are proposed in the growing volume of most widespread and available technology tools, and this technique
data and the diversity of information, making the 4G network unable achieves high-speed data while also combating interference from multi-
to meet the expected requirements of future data usage. It is manifested path channels [9]. However, OFDM performs weak performance against
in two aspects specifically: on the one hand, data traffic is proliferating the effect of Doppler shift [10] and is not suitable for employment in
at an exponential rate; on the other hand, the expanded services high mobility scenarios.
of emerging applications exceed the loading capacity of the current
A new modulation technique has been proposed, called Orthogonal
network system. In the current 5G mobile communication system, the
Time–Frequency Space (OTFS) [11,12]. OTFS technique maps the data
communication scenarios are divided into three categories according to
symbols into the Delay–Doppler (DD) domain. Each symbol can be fully
different requirements: enhanced Mobile broadband (eMBB), massive
expanded in the Time–Frequency (TF) domain, allowing the channel to
Machine Type of Communication (mMTC), ultra-Reliable and Low La-
have approximately the same effect on each data symbol, thus reducing
tency Communication (uRLLC) [3]. Although 5G has dramatically im-
the fading and interference. In other words, this technique transforms
proved communication rates and latency compared to the 4G network,
channels that exhibit fading in the TF domain into the approximately
there are still considerable challenges in communication under the
high-speed scenarios. In addition, other emerging applications, such as constant channel in the DD domain. It extracts the full diversity in
medium combat aircraft (MCA), low-earth-orbit satellites (LEOSs) [4], the time dimension, frequency dimension and angle dimension [13].
high-speed trains [5] and unmanned aerial vehicles (UAVs) [6], are also Moreover, the OFDM system needs to exploit high complexity algo-
facing the same problems. rithms to decrease inter-subcarrier interference (ICI) [14] caused by

∗ Corresponding author.
E-mail addresses: [email protected] (M. Li), [email protected] (W. Liu), [email protected] (J. Lei).

https://doi.org/10.1016/j.comnet.2023.109597
Received 22 June 2022; Received in revised form 17 November 2022; Accepted 25 January 2023
Available online 30 January 2023
1389-1286/© 2023 Published by Elsevier B.V.
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 2. Description of time–frequency and delay–Doppler domain plane grid. (𝑙𝑒𝑓 𝑡)

Fig. 1. (𝑙𝑒𝑓 𝑡) frequency selective fading; (𝑟𝑖𝑔ℎ𝑡) time selective fading; (𝑚𝑖𝑑𝑑𝑙𝑒) doubly Time–Frequency domain grid; (𝑟𝑖𝑔ℎ𝑡) Delay–Doppler domain grid.
dispersion channels.

on reviewing the various application combined with OTFS modulation.

highly dynamic channels. In contrast, the OTFS system considers prob- In Section 6, other emerging fields merged with OTFS modulation are
lems from the perspective of interference sources, so that information worth exploring and excavating are discussed further. Finally, Section 7
symbols experience minimal interference and thus improve reliability concludes the paper. The complete abbreviations of important terms are
and accuracy in highly dynamic scenarios. c According to the above given in Table 1.
description, OTFS technology shows the following advantages: (1) The
channel that exhibits fast time-variant characteristics in the TF domain 2. Model of OTFS system
can be equivalently represented in the DD domain using the impulse
response; (2) In the DD domain, the channel taps are associated with In this section, the basic concept of OTFS technology is presented,
a set of reflectors. Considering that only a few sets of reflectors have and then the input–output relation is derived both in the TF and
different delay and Doppler shift values, the parameters which are DD domains. Next, the pulse-shaping waveform design and interfer-
applied for channel estimation are also relatively few and can be ence analysis are discussed [18]. The vector representation of the
sparsely represented; (3) Compared with the channel in the TF domain, input–output relation is also shown, which can be well used in the
the channel taps in the DD domain can be considered time-invariant simulation.
for longer observation time, which facilitates channel estimation. To
sum up, it is essential to explore and study OTFS as a new modulation 2.1. Basis concepts
technology to promote the development of various applications in high-
speed dynamics [15]. In the future, OTFS technology is expected to be First of all, the time–frequency plane is discretized to a grid 𝛬 by
compatible with the existing modulation framework [16]. Also, it can sampling at interval 𝑇 (second) and 𝛥𝑓 (Hz), respectively.
be maturely applied to all kinds of fields to cope with the Doppler shift.
𝛬 = {(𝑛𝑇 , 𝑚𝛥𝑓 )}, 𝑛 = 0, … , 𝑁 − 1, 𝑚 = 0, … , 𝑀 − 1 (1)
The existing review literature [15–17] is limited to discussing part where 𝑁, 𝑀 is the integer. The duration of the whole packet is 𝑁𝑇 , the
of the related technology of OTFS-based. Different from the litera- occupied bandwidth is 𝑀𝛥𝑓 . The information representation is 𝑋[𝑛, 𝑚]
tures mentioned above, this paper provides a more detailed guide in TF domain.
for the research on the OTFS-related technologies, and the specific In DD domain, the signal representation shows the quasi-periodic
representation is as follows: form, with 1∕𝛥𝑓 as the period in delay direction and 1∕𝑇 as the period
∙ this paper summarizes the current status of OTFS technology in Doppler direction. Therefore, the delay–Doppler plane is discretized
from a more comprehensive perspective, including pulse-shaped design, in delay direction with 1∕(𝑀𝛥𝑓 ) as the interval and Doppler direction
channel coding, precoding scheme at the transmitting end, channel with 1∕(𝑁𝑇 ) as the interval to obtain the delay–Doppler plane grid 𝛤 .
estimation, signal detection, equalization at the receiving end, diver- The time–frequency domain and delay–Doppler domain plane grid as
sity analysis, PAPR reduction methods and multiple access technol- shown in Fig. 2.
ogy. Moreover, specific scenarios are considered, such as terrestrial
𝛤 = {(𝑘∕𝑁𝑇 , 𝑙∕𝑀𝛥𝑓 )}, 𝑘 = 0, … , 𝑁 − 1, 𝑙 = 0, … , 𝑀 − 1 (2)
broadcast communication, mmWave communication, joint radar and
communication, etc. Therefore, the hotspots analyzed in this paper are where 1∕(𝑀𝛥𝑓 ) and 1∕(𝑁𝑇 ) represent the quantization steps of the
exhaustive and widely. delay and Doppler frequency, respectively.
∙ the performance of different modulation methods is compared in
this paper. Under the framework of compatibility, the similarities and 2.2. Input-to-output expression in OTFS
differences between OFDM, GFDM and AFDM modulation methods and
OTFS modulation methods are described, and the simulation analysis is The OTFS system modulation diagram is shown in Fig. 3. The data
carried out. Simulation results show that OTFS modulation has better symbols 𝑥[𝑘, 𝑙] in the DD domain are mapped to the TF domain after
BER performance. experiencing the inverse symplectic Fourier transform (ISFFT) and the
The remaining sections are organized as follows: Section 2 describes addition of a transmit window function and then become TF domain
the mathematical model of OTFS, highlighting the relation between signals 𝑋[𝑛, 𝑚]. Next, the Heisenberg transform is applied to 𝑋[𝑛, 𝑚]
input–output and interference analysis. In Section 3, the comparisons to generate time domain signal 𝑠[𝑡]. The time domain signal 𝑠[𝑡] is
between current modulation methods and OTFS modulation are pre- transmitted to the receiver via the wireless channel, and the receiver
sented. The specific technical approaches are analyzed in Section 4, performs the inverse process of the transmitter. That is to say, the
including channel coding, pulse-shaping waveform design, channel es- time domain received signal 𝑟[𝑡] is transformed by Wigner transform
timation, signal detection and equalization, etc. Section 5 mainly focus and then recovered into the DD domain signal by adding a receive

2
M. Li et al. Computer Networks 224 (2023) 109597

Table 1
The summary of important acronyms.
Acronym Definition Acronym Definition
4𝐺 Fourth Generation 𝑂𝑀𝐴 Orthogonal Matching Algorithm
5𝐺 Fifth Generation 𝑀𝐴𝑃 Maximum a Posteriori
𝑂𝐹 𝐷𝑀 Orthogonal Frequency Division Multiplexing 𝐴𝐶𝐼 Adjacent Channel Interference
𝑀𝑇 𝐶 Machine Type Communication 𝐼𝐷𝐼 Inter-Doppler Interference
𝑀2𝑀 Machine-to-Machine 𝐼𝑆𝐼 Inter-Symbols Interference
𝑉 2𝑉 Vehicle-to-Vehicle 𝐼𝐶𝐼 Inter-Carrier Interference
𝑂𝑇 𝐹 𝑆 Orthogonal Time–Frequency Space 𝐼𝑆𝐹 𝐹 𝑇 Inverse Symplectic Finite Fourier Transform
𝑆𝑁𝑅 Signal-to-Noise Ratio 𝑀𝐴𝑃 Maximum a posterior probability
𝐹 𝐸𝑅 Frame Error Rate 𝑃 𝑀𝐹 Probability Mass Function
𝐵𝐸𝑅 Bit Error Rate 𝑃 𝐼𝐶 Parallel Interference Cancellation
𝑀𝐼𝑀𝑂 Multiple Input Multiple Output 𝑂𝑃 Orthogonal Precoding
𝑆𝐼𝑆𝑂 Single Input Single Output 𝐼𝑀 Index Modulation
𝑇𝐹 Time–Frequency 𝐼𝐶𝑀𝑃 Iterative Combining Message passing
𝐷𝐷 Delay–Doppler 𝑇 𝑀𝑃 Turbo Message Passing
𝐷𝐷𝐴 Delay–Doppler-Domain 𝐹 𝑆𝑆 Fractionally Spaced Sampling
𝐵𝑆 Base Station 𝐶𝑆𝐼 Channel State Information
𝑂𝑂𝐵 Out-of-Band 𝑍𝐹 Zero Force
𝐶𝑃 Circular Prefix 𝐷𝐹 𝐸 Decision Feedback Equalizer
𝐼𝐶𝐼 Inter-Subcarrier Interference 𝐿𝐷𝑃 𝐶 Low Density Parity Check
𝑁𝑂𝑀𝐴 Non-orthogonal Multiple Access 𝑀𝐿 Machine Learning
𝑀𝐶𝐴 Medium Combat Aircraft 𝐴𝐼 Artificial Intelligence
𝐿𝐸𝑂𝑆𝑠 Low-Earth-Orbit Satellites 𝑇 𝐷𝐷 Time Division Duplexing
𝑈 𝐴𝑉 𝑠 Unmanned Aerial Vehicles 𝐹 𝐷𝐷 Frequency Division Duplexing
𝑒𝑀𝐵𝐵 enhanced Mobile Broadband 𝐶𝐷𝑀𝐴 Code Division Multiple Access
𝑚𝑀𝑇 𝐶 massive Machine Type of Communication 𝑇 𝐷𝑀𝐴 Time Division Multiple Access
𝐶𝑃 𝑆𝐶 Cyclic Prefix Single Carrier 𝑆𝐶𝑀𝐴 Sparse Code Multiple Access
𝐴𝑜𝐷 Angle-of-Departure 𝐹 𝑆𝑆 Fractionally Spaced Sampling
𝐻𝑃 𝐴 High Power Amplifier 𝑊 𝐻𝑇 Walsh–Hadamard Transform
𝑆𝐹 𝐹 𝑇 Symplectic Finite Fourier Transform 𝑆𝐵𝐿 Sparse Bayesian Learning
𝑀𝑀𝑆𝐸 Minimum Mean Square Error 𝑅𝐴 Random Access
𝑃 𝐴𝑃 𝑅 Peak-to-Average Power Ratio 𝑀𝑅𝐶 Maximum Ratio Combining
𝐷𝐹 𝑇 Discrete Fourier Transform 𝑂𝐴𝑀 Orbital Angular Momentum
𝐼𝑅𝑆 Intelligent Reflecting Surface 𝑂𝑇 𝑆𝑀 Orthogonal Time Sequence Multiplexing
𝐹 𝐵𝑀𝐶 Filter Bank Multicarrier 𝐶𝑃 𝑃 Chirp-Periodic Prefix
𝐻𝐶 Hybrid Carrier 𝐼𝑄 In-phase and Quadrature
𝐷𝑁𝑁 Deep Neural Network 𝑀𝑀 Majorization Minimization

Fig. 3. The diagram block of OTFS system.

window function and applying the symplectic Fourier transform (SFFT) Eq. (3) can be transformed to Eq. (4) after adding the window
transform. The specific content is elaborated and analyzed in Ref. [13]. function.
For OTFS modulation, a set of 𝑀𝑁 information symbols 𝑥[𝑘, 𝑙] are ∑ 𝑀−1
𝑁−1 ∑
1 𝑛𝑘 𝑚𝑙
arranged on the DD domain plane grid 𝛤 , and these information sym- 𝑋[𝑛, 𝑚] = 𝑊𝑡𝑥 [𝑛, 𝑚] √ 𝑥[𝑘, 𝑙]𝑒𝑗2𝜋( 𝑁 − 𝑀 ) (4)
bols are generally QAM symbols. The data information is transformed 𝑁𝑀 𝑘=0 𝑙=0
from the TF domain to the DD domain by the ISFFT transform. The where 𝑊𝑡𝑥 [𝑛, 𝑚] denotes the transmit window function.
representation of ISFFT as follow.
Next, the TF modulator converts the samples 𝑋[𝑛, 𝑚] to continuous
1 ∑ 𝑀−1
𝑁−1 ∑ 𝑗2𝜋( 𝑛𝑘 − 𝑚𝑙 )
time waveform 𝑠[𝑡] using the transmit waveform 𝑔𝑡𝑥 (𝑡) as
𝑋[𝑛, 𝑚] = √ 𝑥[𝑘, 𝑙]𝑒 𝑁 𝑀 (3)
𝑁𝑀 𝑘=0 𝑙=0 ∑ 𝑀−1
𝑁−1 ∑
1 𝑛𝑘 𝑚𝑙
𝑠(𝑡) = 𝑊𝑡𝑥 [𝑛, 𝑚] √ 𝑥[𝑘, 𝑙]𝑔𝑡𝑥 (𝑡 − 𝑛𝑇 )𝑒𝑗2𝜋( 𝑁 − 𝑀 ) (5)
where 𝑛 = 0, … , 𝑁 − 1, 𝑚 = 0, … , 𝑀 − 1 𝑁𝑀 𝑘=0 𝑙=0

3
M. Li et al. Computer Networks 224 (2023) 109597

where 𝑔𝑡𝑥 (𝑡 − 𝑛𝑇 ) denotes the receive pulses. The mathematical method

used in Eq. (5) is Heisenberg transform.
The received signal 𝑟[𝑡] is obtained by superimposing Gaussian
white noise 𝑢[𝑡] on transmitted after passing through the channel ℎ(𝜏, 𝜈).
The detailed expression is shown below.

𝑟(𝑡) = ℎ(𝜏, 𝜈)𝑥(𝑡 − 𝜏)𝑒𝑗2𝜋(𝑡−𝜏) d𝜏d𝜈 (6)

∫𝜈 ∫𝜏
where ℎ(𝜏, 𝜈) represents channel impulse response, which characterizes
the channel response to an impulse with delay 𝜏 and Doppler 𝜈.
At the receiving end, using the received signal and received pulse
for matched filtering. First, the cross-ambiguity functions 𝐶𝑔𝑟𝑥,𝑦 (𝑡, 𝑓 ) of
𝑔𝑡𝑥 (𝑡) and 𝑟(𝑡) are needed to be calculated as shown in Eq. (7).

∗ ′ ′
𝑌 (𝑡, 𝑓 ) = 𝐶𝑔𝑟𝑥,𝑦 (𝑡, 𝑓 ) = 𝑔𝑟𝑥 (𝑡 − 𝑡)𝑟(𝑡′ )𝑒−𝑗2𝜋𝑓 (𝑡 −𝑡) d𝑡′ (7)
∫
Then, the matched filter output is obtained by sampling at intervals
𝑡 = 𝑛𝑇 , 𝑓 = 𝑚𝛥𝑓 .

𝑌 [𝑛, 𝑚] = 𝐶𝑔𝑟𝑥,𝑦 (𝑡, 𝑓 )|𝑡=𝑛𝑇 ,𝑓 =𝑚𝛥𝑓 (8)

Next, the received window function is added.

𝑌 [𝑛, 𝑚] = 𝑊𝑟𝑥 [𝑛, 𝑚]𝐶𝑔𝑟𝑥,𝑦 (𝑡, 𝑓 )|𝑡=𝑛𝑇 ,𝑓 =𝑚𝛥𝑓 (9)

where 𝑊𝑟𝑥 [𝑛, 𝑚] denotes the received window function.

The following input–output relation of OTFS in time–frequency
domain is shown below.
∑ 𝑀−1
𝑁−1 ∑
𝑦[𝑛, 𝑚] = 𝐻𝑛,𝑚 [𝑛′ , 𝑚′ ]𝑋[𝑛′ , 𝑚′ ] (10)
𝑛′ =0 𝑚′ =0

where

𝐻𝑛,𝑚 [𝑛′ , 𝑚′ ] = ℎ(𝜏, 𝜈)𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 ((𝑛 − 𝑛′ )𝑇 − 𝜏,

∬ (11)
′ ′ ′
(𝑚 − 𝑚′ )𝛥𝑓 − 𝜈)𝑒𝑗2𝜋(𝜈+𝑚 𝛥𝑓 )((𝑛−𝑛 )𝑇 −𝜏) 𝑒𝑗2𝜋𝜈𝑛 𝑇 d𝜏d𝜈
Fig. 4. The doubly-selective channel description of time–frequency domain and
Finally, the SFFT is applied on samples 𝑌 [𝑛, 𝑚] to obtain symbols delay–Doppler domain.
𝑦[𝑘, 𝑙] in the DD domain.

1 ∑ 𝑀−1
𝑁−1 ∑ 𝑛𝑘 𝑚𝑙
𝑦[𝑘, 𝑙] = √ 𝑌 [𝑛, 𝑚]𝑒−𝑗2𝜋( 𝑁 − 𝑀 ) (12) ∙ Integer multiples of delay and Doppler. When the delay and
𝑁𝑀 𝑛=0 𝑚=0 1
Doppler are integer multiples of the delay resolution 𝑀𝛥𝑓 and Doppler
1
where 𝑛 = 0, … , 𝑁 − 1, 𝑚 = 0, … , 𝑀 − 1 resolution , respectively [19]. The channel ℎ(𝜏, 𝜈) is expressed as
𝑁𝑇
( ) ( )
∑ 𝑀−1
𝑁−1 ∑ 𝑙 (𝑘)𝑁
2.2.1. Channel description ℎ(𝜏, 𝜈) = ℎ𝑖 𝛿 𝜏 − 𝛿 𝜈− (15)
The above equations describe the process of OTFS modulation and 𝑘=0 𝑙=0
𝑀𝛥𝑓 𝑁𝑇
demodulation, and the description of the channel needs to be discussed. where
The time-varying channel is described in terms of complex baseband {
channel impulse response ℎ(𝜏, 𝜈), which characterizes the channel re- 𝑘 if 𝑘 ≤ 𝑁∕2
(𝑘)𝑁 = (16)
sponse to an impulse with delay 𝜏 and Doppler 𝜈. The doubly-selective 𝑘−𝑁 otherwise
channel simulation of the time–frequency domain and delay–Doppler According to Eq. (12), the new expression of input–output is de-
domain as shown in Fig. 4 when 𝑃 = 5. Considering that there are only noted as follow.
a few reflectors with associated delays and Dopplers, a small number ( )
[𝑙−𝑙′ ]𝑀 (𝑘′ )𝑁
of parameters are needed to estimate in the DD domain. The equation ∑ ∑𝑀−1 [ ′ ′ ] 𝑗2𝜋
𝑦[𝑘, 𝑙] = 𝑁−1 ℎ 𝑘 , 𝑙 𝑒 𝑀 𝑁
𝛼𝑖 [𝑘, 𝑙] (17)
is denoted as follows. 𝑘′ =0 𝑙′ =0 [[ ] [ ] ]
𝑥 𝑘 − 𝑘 𝑁 , 𝑙 − 𝑙′ 𝑀
′

∑
𝑃
ℎ(𝜏, 𝜈) = ℎ𝑖 𝛿(𝜏 − 𝜏𝑖 )𝛿(𝜈 − 𝜈𝑖 ) (13) where
𝑖=1 {
1 𝑙′ ≤ 𝑙 < 𝑀
where 𝑃 represents the number of propagation path. ℎ𝑖 , 𝜏𝑖 , 𝜈𝑖 denote 𝛼𝑖 [𝑘, 𝑙] = 𝑘
−𝑗2𝜋 𝑁
(18)
𝑒 0 ≤ 𝑙 < 𝑙′
the path gain, delay, Doppler shift respectively. 𝛿 represents the Dirac
delta function. Among them, where [⋅]𝑀 and [⋅]𝑁 denote modulo 𝑁 and 𝑀 operations, respectively.

𝜏𝑖 = 𝑙𝜏𝑖 ∕𝑀𝛥𝑓 , 𝜈𝑖 = (𝑘𝜈𝑖 + 𝜅𝜈𝑖 )∕𝑁𝑇 (14)

2.2.2. Vector forms
where 𝑙𝜏𝑖 , 𝑘𝜈𝑖 are the integers, and denote the indexes of delay tap and In order to analyze the performance of OTFS system in vector
Doppler tap associated with delay 𝜏𝑖 and Doppler 𝜈𝑖 . 𝜅𝜈𝑖 represents the forms, the formulas of subsection 2.2.2 is reorganized in vectors or
fractional Doppler. Based on above channel description, a special case matrices [20]. Assuming 𝑥 = 𝑣𝑒𝑐(𝐷) represent the signal to be trans-
needs to be noted. mitted where 𝐷 is the two-dimensional data matrix. 𝑣𝑒𝑐() represents

4
M. Li et al. Computer Networks 224 (2023) 109597

the vectorization. Hence, the ISFFT operation can be written as follow. received pulse 𝑔𝑟𝑥 (𝑡) is ideal waveform, the bi-orthogonal property is
satisfy as follow.
𝑋= 𝐹𝑀 𝐷𝐹𝑁𝐻 (19)
𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 (𝑡, 𝑓 )|𝑡=𝑛𝑇 +(−𝜏𝑚𝑎𝑥 ,𝜏𝑚𝑎𝑥 ),𝑓 =𝑚𝛥𝑓 +(−𝜈𝑚𝑎𝑥 ,𝜈𝑚𝑎𝑥 )
where 𝑋 represents the signal of TF domain. 𝐹𝑀 denotes the FFT (26)
= 𝛿[𝑛]𝛿[𝑚]𝑞𝜏𝑚𝑎𝑥 (𝑡)𝑞𝜈𝑚𝑎𝑥 (𝑓 )
transform of M points and 𝐹𝑁𝐻 denoted the IFFT of transform of N
points. Next, the Heisenberg transform is expressed as below. where 𝑞𝑎 (𝑥) = 1 for 𝑥 ∈ (−𝑎, 𝑎) and zero otherwise. That is to say, the
cross-ambiguity function 𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 (𝑡, 𝑓 ) = 0 for 𝑡 ∈ (𝑛𝑇 − 𝜏𝑚𝑎𝑥 , 𝑛𝑇 + 𝜏𝑚𝑎𝑥 )
𝐻 𝐻
𝑆 = 𝐺𝑡𝑥 𝐹𝑀 𝑋 = 𝐺𝑡𝑥 𝐹𝑀 (𝐹𝑀 𝐷𝐹𝑁𝐻 ) = 𝐺𝑡𝑥 𝐷𝐹𝑁𝐻 (20) and 𝑓 ∈ (𝑚𝛥𝑓 − 𝜈𝑚𝑎𝑥 , 𝑚𝛥 + 𝜈𝑚𝑎𝑥 ), ∀𝑛, 𝑚 except for 𝑛 = 0, 𝑚 = 0, where
where the 𝐺𝑡𝑥 represents the transmitted rectangular pulse as shown in 𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 (𝑡, 𝑓 ) = 1 for 𝑡 ∈ (−𝜏𝑚𝑎𝑥 , 𝜏𝑚𝑎𝑥 ) and 𝑓 ∈ (−𝜈𝑚𝑎𝑥 , 𝜈𝑚𝑎𝑥 ).
𝐻 denoted the IFFT of transform
Eq. (5) and is the diagonal matrix. 𝐹𝑀 ∙ Rectangular pulse waveform. The rectangular pulses are often
used in practical applications. When the transmitted pulse 𝑔𝑡𝑥 (𝑡) and
of M points.
received pulse 𝑔𝑟𝑥 (𝑡) is rectangular waveform, the cross-ambiguity func-
Then, the signal 𝑆 in TF domain is converted into the time domain
tion 𝐶𝑔rx ,𝑔tx (𝜏, 𝜈) is non-zero only for 𝑛′ = 𝑛 and 𝑛′ = 𝑛 − 1 because
signal 𝑠.
the maximum channel delay 𝜏max < 𝑇 . Hence, the approximated
𝑠 = 𝑣𝑒𝑐(𝑆) = (𝐹𝑁𝐻 ⊗ 𝐺𝑡𝑥 )𝑥 (21) cross-ambiguity function given by
𝑇
where ⊗ represents Kronecker product. 𝐶𝑔rx ,𝑔tx (𝜏, 𝜈) = ∗
𝑔tx (𝑡)𝑔rx (𝑡 − 𝜏)𝑒−𝑗2𝜋𝜈𝑡 𝑑𝑡
The received signal after the delay–Doppler channel can be written ∫0
𝑀−1−𝑙𝜏 ( ) (27)
as follow. 1 ∑ 𝑇𝑖
≈ exp 𝑗2𝜋𝜈
𝑀 𝑀
𝑟 = 𝐻𝑠 + 𝑤 (22) 𝑖=0
⌈ ⌉
𝜏
where 𝐻 represents the channel matrix. 𝑤 represents the noise vector. where 𝑙𝜏 = 𝑇 ∕𝑀
, and 𝑇 is the length of rectangular pulse.
At the receiving end, the received signal is needed to experience
the pulse-shaping, Winger transform and SFFT operation. Assuming 2.3.2. Ideal pulse in TF domain analysis
𝑟 = 𝑣𝑒𝑐(𝑅), the output signal is shown as follows. Since the transmit pulse 𝑔𝑡𝑥 (𝑡) and the receive pulse 𝑔𝑟𝑥 (𝑡) are ideal
waveforms, the bi-orthogonal property is satisfied in term of Eq. (26).
𝐻
𝑌 = 𝐹𝑀 (𝐹𝑀 𝐺𝑟𝑥 𝑅)𝐹𝑁 (23) The input–output relation can be transform as follow.
where 𝐺𝑟𝑥 represents the received rectangular pulse as shown in Eq. (7) 𝑌 [𝑛, 𝑚] = 𝐻𝑛,𝑚 [𝑛, 𝑚]𝑋[𝑛, 𝑚] (28)
and is the diagonal matrix.
where
Finally, the forms can be rewritten as
𝐻𝑛,𝑚 [𝑛, 𝑚] = ℎ(𝜏, 𝜈)𝑒𝑗2𝜋𝜈𝑛𝑇 𝑒−𝑗2𝜋(𝜈+𝑚𝛥𝑓 )𝜏 d𝜏d𝜈 (29)
𝑦 = 𝑣𝑒𝑐(𝑆) = (𝐹𝑁 ⊗ 𝐺𝑟𝑥 )𝑟 (24) ∬

Combining Eqs. (22), (23) and (24), the total expression can be 2.3.3. Ideal pulse in DD domain analysis
obtained. The SFFT as an essential transformation tool to obtain signal 𝑦[𝑘, 𝑙]
from TF domain to DD domain. The expression as below.
𝑦 = (𝐹𝑁 ⊗ 𝐼𝑀 )𝐻(𝐹𝑁𝐻 ⊗ 𝐼𝑀 )𝑥 + (𝐹𝑁 ⊗ 𝐼𝑀 )𝑤 (25)
1 ∑ ∑
𝑁−1 𝑀−1
where 𝐼𝑀 denotes the identity matrices. 𝑦[𝑘, 𝑙] = 𝑥[𝑘′ , 𝑙′ ]ℎ𝑤 [𝑘 − 𝑘′ , 𝑙 − 𝑙′ ] (30)
𝑁𝑀 𝑛=0 𝑚=0
From the vectorized perspective to view the data processing of
OTFS system, as shown in Fig. 3. The SFFT operation on 𝑀 × 𝑁 where
dimensional TF domain signal can be equivalent to the N-points IFFT
ℎ𝑤 [𝑘 − 𝑘′ , 𝑙 − 𝑙′ ] = ℎ𝑤 (𝜈, 𝜏)| ′ ′ (31)
transform firstly by column and then the M-points FFT transform by 𝜈= 𝑘−𝑘
𝑁𝑇
𝑙−𝑙
,𝜏= 𝑀𝛥𝑓
row, respectively. Correspondingly, the ISFFT operation on 𝑀 × 𝑁
in which ℎ𝑤 (𝜈, 𝜏) represents the circular convolution of channel re-
dimensional TF domain signal can be equated to the N-points FFT sponse eith SFFT of a rectangular windowing function in the TF do-
transform firstly by column and then the M-points IFFT transform by main. In addition, ℎ𝑤 (𝜈, 𝜏) denotes as shown in Eq. (32).
row, respectively. In conclusion, the entire processing flow of the OTFS
system is similar to that of the OFDM system. ℎ𝑤 (𝜈, 𝜏) = ℎ(𝜈 ′ , 𝜏 ′ )𝑤(𝜈 − 𝜈 ′ , 𝜏 − 𝜏 ′ )𝑒−𝑗𝑤𝜋𝜈𝜏 d𝜏 ′ d𝜈 ′ (32)
∬

2.3. Waveform analysis ∑ 𝑀−1

𝑁−1 ∑
𝑤(𝜈, 𝜏) = 𝑒−𝑗2𝜋(𝜈𝑛𝑇 −𝜏𝑚𝛥𝑓 ) (33)
𝑛=0 𝑚=0
The transmit pulse and receive pulse are used in data transmission,
so two waveforms design are considered in the derivation. According to 2.3.4. Rectangular pulse in TF domain analysis
two different waveforms, named ideal pulse and rectangular pulse, the Ideal pulse waveform needs to satisfy the expression of Eq. (15).
input–output relation of TF domain and DD domain can be obtained, However, the cross-ambiguity function 𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 ((𝑛−𝑛′ )𝑇 −𝜏, (𝑚−𝑚′ )𝛥𝑓 −𝜈)
respectively. is non-zero when 𝑛′ = 𝑛 and 𝑛′ = 𝑛−1 in case of using rectangular pulse.
However, the ideal pulse cannot apply to the actual operation due The time–frequency relation about rectangular pulse becomes below.
to the uncertainty principle. It is significant to study the ideal pulse ∑
𝑛 ∑
𝑀−1
waveform to explore the bound of OTFS system performance. The 𝑌 [𝑛, 𝑚] = 𝐻𝑛,𝑚 [𝑛′ , 𝑚′ ]𝑋[𝑛′ , 𝑚′ ]
purpose of examining rectangular pulse waveform is to investigate the 𝑛′ =𝑛−1 𝑚′ =0
need for application in practical engineering. The specific content is = 𝐻𝑛,𝑚 [𝑛, 𝑚]𝑋[𝑛, 𝑚]
elaborated and analyzed in literature [18]. ∑
𝑀−1
(34)
+ 𝐻𝑛,𝑚 [𝑛, 𝑚′ ]𝑋[𝑛, 𝑚′ ]
𝑚′ =0,𝑚′ ≠𝑚
2.3.1. Pulse property
∙ Idea pulse waveform. In mathematical derivation process, the ∑
𝑀−1
+ 𝐻𝑛,𝑚 [𝑛 − 1, 𝑚′ ]𝑋[𝑛 − 1, 𝑚′ 𝑐]
ideal pulse waveform is considered. If the transmitted pulse 𝑔𝑡𝑥 (𝑡) and
𝑚′ =0

5
M. Li et al. Computer Networks 224 (2023) 109597

∑
where 𝑀−1 ′ ′
𝑚′ =0,𝑚′ ≠𝑚 𝐻𝑛,𝑚 [𝑛, 𝑚 ]𝑋[𝑛, 𝑚 ] is regarded as the total interfer- 2.4.2. Inter carrier interference
ence from the samples 𝑋[𝑛, 𝑚 ] at different frequency 𝑚′ ≠ 𝑚 but in
′ In second term of Eq. (22), the expression is represented inter carrier
the same time slot 𝑛 as the current sample 𝑋[𝑛, 𝑚], which is called ICI. interference. The interference depends on the delay 𝜏 and Doppler 𝜈 and
∑𝑀−1 ′ ′
𝑚′ =0 𝐻𝑛,𝑚 [𝑛 − 1, 𝑚 ]𝑋[𝑛 − 1, 𝑚 ] accumulates the interference form the the value of cross-ambiguity function 𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 ((𝑛−𝑛′ )𝑇 −𝜏, (𝑚−𝑚′ )𝛥𝑓 −𝜈).
samples 𝑋[𝑛 − 1, 𝑚′ ] in the previous time slot 𝑛 − 1, which is called inter The expression of 𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 (−𝜏, (𝑚 − 𝑚′ )𝛥𝑓 − 𝜈) as follow.
symbol interference (ISI). ICI and ISI depend on delay 𝜏 and Doppler 𝜈
∗ ′ ′ ′
of the channel and be affected by the value of cross-ambiguity function 𝐶𝐼𝐶𝐼 ≜ 𝑔𝑟𝑥 (𝑡 + 𝜏𝑖 )𝑔𝑡𝑥 (𝑡′ )𝑒−𝑗2𝜋((𝑚−𝑚 )𝛥𝑓 −𝜈𝑖 )(𝑡 +𝜏𝑖 ) d𝑡′ (41)
∫
𝐶𝑔𝑟𝑥 ,𝑔𝑡𝑥 ((𝑛 − 𝑛′ )𝑇 − 𝜏, (𝑚 − 𝑚′ )𝛥𝑓 − 𝜈).
For the Eq. (41), the simplified form is shown in Eq. (42).
2.3.5. Rectangular pulse in DD domain analysis 𝑀−1−𝑙𝜏𝑖 𝑝
1 ∑ −𝑗2𝜋((𝑚−𝑚 )𝛥𝑓 −𝜈𝑖 )( 𝑀𝛥𝑓 +𝜏𝑖 )
′
The received signal 𝑦[𝑘, 𝑙] can be written in DD domain with rect- 𝐶𝐼𝐶𝐼 = 𝑒 (42)
angular pulse as follow. 𝑀 𝑝=0

∑
𝑃 𝑁𝑖
∑ 𝑙−𝑙𝜏
𝑖
𝑘𝜈 −𝜅𝜈
𝑖 𝑖
The amplitude of 𝐶𝐼𝐶𝐼 is
)( )
𝑦[𝑘, 𝑙] ≈ ℎ𝑖 𝑒𝑗2𝜋( 𝑀 𝑁 𝛼𝑖 (𝑘, 𝑙, 𝑞) 𝑘𝜈 +𝜅𝜈 𝑀−𝑙𝜏
𝑖=1 𝑞=−𝑁𝑖 (35) |𝑒−𝑗2𝜋(𝑚−𝑚 −
′ 𝑖
𝑁
𝑖 )( 𝑀
𝑖 )
− 1|
|𝐶𝐼𝐶𝐼 | = 𝑘𝜈 +𝜅𝜈
(43)
𝑥[[𝑘 − 𝑘𝜈𝑖 +𝑞 ]𝑁 , [𝑙 − 𝑙𝜏𝑖 ]𝑀 ] −𝑗2𝜋(𝑚−𝑚′ − 𝑖 𝑁 𝑖 1
)( 𝑀 )
|𝑀𝑒 − 𝑀|
where
According to Eq. (43), |𝐶𝐼𝐶𝐼 | decreases as 𝑚′ moves away from
⎧1 𝑚 can be observed which implies that ICI becomes less as interfering
⎪ 𝛽𝑖 (𝑞) 𝑙𝜏𝑖 ≤ 𝑙 ≤ 𝑀
𝛼𝑖 (𝑘, 𝑙, 𝑞) = ⎨ 𝑁 subcarriers are further away from the interfered subcarrier.
1
⎪ (𝛽𝑖 (𝑞) − 1) 0 ≤ 𝑙 ≤ 𝑙𝜏 .
⎩𝑁 𝑖
2.4.3. Inter symbols interference
−𝑗2𝜋(−𝑞−𝜅𝜈𝑖 ) In third term of Eq. (34), the expression is represented inter symbol
𝑒 −1
𝛽𝑖 = (36) interference.
−𝑗 2𝜋 (−𝑞−𝜅𝜈𝑖 )
𝑒 𝑁 −1 𝜈
1 ∑ −𝑗2𝜋((𝑚−𝑚 )𝛥𝑓 −𝜈𝑖 )( 𝑀𝛥𝑓 +𝜏𝑖 −𝑇 )
𝑀−1 ′
All in all, the OTFS system exists IDI when using the ideal pulse for 𝐶𝐼𝑆𝐼 = 𝑒 (44)
performance analysis in the DD domain. Similarly, there are ISI and ICI 𝑀 𝑝=𝑀−𝑙
𝜏𝑖
when using rectangular pulse.
The amplitude 𝐶𝐼𝑆𝐼 also has similar properties of 𝐶𝐼𝐶𝐼 , where it
reduces as 𝑚′ moves away from m implying that ISI is smaller for
2.4. Interference analysis
interfering symbols further away from the interfered symbol.
From Eqs. (42) and (44), it can be obtained that ISI and ICI are not
From the input–output relationship as described in Section 2.3, we
compatible with each other. When 𝑝 from 0 to 𝑀 − 1 − 𝑙𝜏𝑖 , the ICI exists;
can know that there are interferences such as ICI, IDI and ISI in the
when 𝑝 from 𝑀 − 𝑙𝜏𝑖 to 𝑀 − 1, the ISI exists. This situation is beneficial
OTFS system. This subsection focuses on the analysis of three different
to distinguish the ICI and ISI effects in DD domain.
kinds of interference: IDI, ICI and ISI.
In summary, the basic description of the OTFS system is presented
from the input–output relationship and interference analysis, which can
2.4.1. Inter Doppler interference
provide theoretical support for subsequent research on related tech-
Taking the sparse channel manifestation (Eq. (13)) into the input–
nologies. The following section focuses on the comparison of existing
output relation (Eq. (29)) in the DD domain when using the ideal pulse,
modulation methods and OTFS modulation.
the expression can be obtained:
∑
𝑃
3. Comparison of existing modulation methods
ℎ𝑤 (𝜈, 𝜏) = ℎ𝑖 𝑒−𝑗2𝜋𝜈𝑖 𝜏𝑖 𝑤(𝜈 − 𝜈𝑖 , 𝜏 − 𝜏𝑖 )
𝑖=1
(37) Section 2 is mainly a review of the input–output relations of the
∑
𝑃
= ℎ𝑖 𝑒 −𝑗2𝜋𝜈𝑖 𝜏𝑖
𝜑(𝜈, 𝜈𝑖 )𝜓(𝜏, 𝜏𝑖 ) OTFS system (mathematical expressions and vector expressions). This
𝑖=1 chapter mainly discusses the commonalities and differences between
where OTFS and existing modulation techniques, such as OFDM, Generalized
frequency division multiplexing (GFDM) and Affine Frequency Division
∑
𝑀−1
′
𝜑(𝜈, 𝜈𝑖 ) ≜ 𝑒𝑗2𝜋(𝜏−𝜏𝑖 )𝑚 𝛥𝑓 (38) Multiplexing (AFDM).
𝑚′ =0
∑
𝑁−1 3.1. Comparison between OFDM and OTFS
′
𝜓(𝜏, 𝜏𝑖 ) ≜ 𝑒−𝑗2𝜋(𝜈−𝜈𝑖 )𝑛 𝑇 (39)
𝑛′ =0 OFDM technology is widely adopted in the 4G mobile communi-
According to a series of mathematical operations, the finally expres- cation system due to its high data rates, reliable coverage for mobile
sion is shown in Eq. (28). users, and superiority against multipath interference. However, OFDM
𝑘′ =[𝑘−𝑘𝜈𝑖 +𝑁𝑖 ]|𝑁 can only solve the challenges posed by the application scenarios of the
∑
𝑃 ∑ 𝑒−𝑗2𝜋(𝑘−𝑘𝜈𝑖 −𝑘 −𝜅𝜈𝑖 ) − 1
′
5G network in a limited way, which does not reflect the advantages of
𝑦[𝑘, 𝑙] ≈ ( 2𝜋
)
𝑖=1 𝑘′ =[𝑘−𝑘𝜈 −𝑁𝑖 ]|𝑁 𝑁𝑒−𝑗 𝑁 (𝑘−𝑘𝜈𝑖 −𝑘 −𝜅𝜈𝑖 ) − 𝑁
′ (40) OFDM. For example, MTC and M2M [21] need a strict synchronization
𝑖
−𝑗2𝜋𝜈𝑖 𝜏𝑖 ′
process under the circumstance of orthogonality between subcarriers;
ℎ𝑖 𝑒 𝑥[𝑘 , [𝑙 − 𝑙𝜏𝑖 |𝑀 ]]
the short burst of data in V2V applications [22] brings the problems
From the Eq. (40), it is known that the received signal 𝑦[𝑘, 𝑙] is of low spectral efficiency; In addition, the high OOB emission of
∑𝑃
a linear combination of 𝑆 = 𝑖=1 2𝑁𝑖 + 1 transmitted signals. The
OFDM [23] will result in the challenges for opportunistic and dynamic
transmitter signals that are neighboring 𝑥[[𝑘 − 𝑘𝜈𝑖 |𝑁 ], [𝑙 − 𝑙𝜏𝑖 ]|𝑀 ] are spectrum access. The above problems demonstrate the limitations of
called IDI in the DD domain. The number of transmitted signals 𝑆 OFDM modulation in high-speed scenarios. In literature [24–26], a
affecting received signal 𝑦[𝑘, 𝑙] is smaller than 𝑁𝑀 in Eq. (18). Hence, fair comparison between OFDM and OTFS is provided, and [25,26]
the linear system is sparsely-connected. mainly discuss the achievable rate of OTFS. From the perspective of

6
M. Li et al. Computer Networks 224 (2023) 109597

3.2. Comparison between GFDM and OTFS

GFDM modulation has been proposed for the air interface of the
5G network, which is a promising multicarrier modulation method for
the physical layer [28]. The modulation objects of the GFDM scheme
are dependent blocks, where each block comprises subcarriers and
subsymbols. The subcarriers are filtered with a prototype filter that is
circularly shifted in the time and frequency domain. The block diagram
Fig. 5. The scheme diagram of OFDM-based OTFS system. of GFDM is depicted in Fig. 7. The data source generates the binary
data vector, which experiences the process of encoding, mapper, GFDM
modulator and cyclic prefix, and then the signal is received by the
receiver antenna. The initial signal can be restored by reversing the
MIMO technology, literature [27] compares the performance between
operation at the receiver side [29].
the MIMO–OFDM system and the MIMO–OTFS system, and finds that
Based on the above description of GFDM, Nimr et al. [30] present
MIMO–OTFS outperforms the MIMO–OFDM modulation scheme. The
above issues illustrate OFDM is not the most promising waveform an extended representation of GFDM, which shows that OTFS can be
for the next generation networks, especially in dealing with commu- processed as a GFDM signal with simple permutation. That is to say,
nication data processing under doubly-selective channels. Therefore, the OTFS block results from a permutation of the corresponding con-
alternative multicarrier schemes (OTFS) are currently being evaluated ventional GFDM block and the received OTFS block can be represented
as candidates for the physical layer of the next generation of mobile as a GFDM block in addition noise channel. Furthermore, GFDM is also
communication systems. a reasonable choice for moderate and low latency circumstances due to
OTFS as one of the research hotspots, the current situation is re- its advantage over frame error rate (FER).
graded OFDM modulation as a template. The OFDM-based OTFS system GFDM and OTFS exist similar processing steps. As shown in Fig. 8,
diagram is shown in Fig. 5. The difference from Fig. 3 is that the the specific steps of GFDM modulation are divided into spreading,
OFDM-based OTFS system utilizes a precoder instead of ISFFT in the windowing, transformation and allocation. Compared with formula
transmitting end, and a decoder replaces SFFT in the receiving end. In (19) (20) (21), the difference between OTFS and GFDM is that there
other words, the OFDM system can realize the functions of the OTFS is no allocation step for OTFS modulation when 𝐾 = 𝑁 and 𝑄 = 𝑀.
system by adding pre-processing and post-processing modules to a cer- In fact, the OTFS block is a permutation of the GFDM block with the
tain degree, and the new framework can further achieve compatibility square commutation matrix 𝑷 𝑁,𝑀 ∈ ℜ𝑁𝑀×𝑁𝑀 .
with the existing OFDM system. Based on the framework proposed The comparison of BER between GFDM modulation and OTFS mod-
in Fig. 5, we simulate the BER performance of OFDM and OFDM at ulation is provided in Fig. 9 under different Doppler. When Doppler
different speeds, respectively, in Fig. 6. is equal to 300 Hz, as shown in Fig. 9(a), the performance of OTFS
As shown in Fig. 6, the comparison of BER performance between is better than GFDM, and the BER of difference between GFDM and
OFDM and OTFS is provided at speeds 𝑣 = 30 km/h, 80 km/h, and OTFS is not significant. When Doppler is set to 3000 Hz, as shown in
130 km/h, respectively, where the blue line represents the BER of
Fig. 9(b), the BER value of OTFS is lower than that of GFDM. The above
OFDM modulation at different Eb/No, the red line represents the BER of
phenomena indicate that OTFS modulation is more suitable for high
OTFS modulation at different Eb/No. It can be seen from Figs. 6(a), 6(b)
Doppler scenarios compared with GFDM modulation. All in all, OTFS
and 6(c) that the BER of OTFS is lower than that of OFDM under the
modulation and GFDM modulation have certain commonalities. Future
same Eb/No. The above simulation result proves that the anti-Doppler
research on OTFS technology can learn from the research on GFDM
characteristics of OTFS will be better than OFDM, and it is suitable for
future high-speed mobile communication scenarios. In addition, Fig. 6 technology.
shows the Eb/No values required to achieve the same BER at different
speeds, which indicates that OTFS requires higher overhead at higher 3.3. Comparison between AFDM and OTFS
speeds.
Compared with the commonly used OFDM modulation, the novelty Affine Frequency Division Multiplexing (AFDM) is a novelty chirp-
OTFS modulation will show excellent quality in terms of anti-Doppler based multicarrier transceiver scheme for high-speed mobility commu-
characteristics and is expected to become the modulation method in nication for a future wireless system, which is able to reduce the ICI and
the next generation of mobile communication or high-speed scenarios. perform low-complexity detection. The AFDM system can be efficiently
The future direction can be to study the hybrid modulation mode of implemented by adding a simple phase-correction block to the OFDM
OFDM and OTFS to achieve reasonable optimization and allocation of modulator/demodulator. The diagram of AFDM modulation is shown
resources. in Fig. 10. The inverse DAFT is used to map data symbols into the time

Fig. 6. Comparison of BER between OFDM modulation and OTFS modulation under different speed. (For interpretation of the references to color in this figure legend, the reader
is referred to the web version of this article.)

7
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 10. The specific steps of GFDM modulation process.

Fig. 7. The scheme diagram of GFDM modulator.

Fig. 11. Comparison of BER between AFDM modulation and OTFS modulation.

Fig. 8. The specific steps of GFDM modulation process. domain, and DAFT is performed at the receiver. Similarity to OFDM,
the OFDM (circular prefix) CP is replaced by the chirp-periodic prefix
(CPP) [31].
AFDM is a promising new waveform for future mobility communi-
cation, which has attracted many attention of scholars at home and
abroad. In literature [32,33], the comparison between (Orthogonal
Code division Multiplexing) OCDM, AFDM and OTFS is provided and
shows that AFDM can achieve the full diversity of time-varying chan-
nels. We simulate the BER performance of OTFS and AFDM at different
Eb/No values. The simulation result is shown in Fig. 11 when 𝑁 = 8
and 𝑀 = 8. It can be seen from Fig. 11 that the OTFS modulation
possesses lower BER performance compared with AFDM modulation.
At higher Eb/No, the performance of AFDM and OTFS is almost the
same. Hence, AFDM modulation also can take advantage of a reference
to further promote OTFS technology development.

3.4. Comparison with other modulation methods

In addition to the above modulation methods [26,28,31], the system

structure of OTFS is equivalent to asymmetric OFDM in the static
multipath channel, which bridges between cyclic prefix single carrier
(CPSC) and traditional OFDM in literature [34]. Moreover, orthogonal
time sequence multiplexing (OTSM) is also proposed in literature [35].
OTSM offers similar performance to OTFS, and a low complexity detec-
tion method by taking advantage of zero-padding and a low complexity
channel estimation method are presented. Besides, filter bank multicar-
rier (FBMC)-based OTFS [36] certificate superiority in error probability
and time spectral efficiency terms.
From the perspective of compatibility, OTFS is the generalization
of code division multiple access (CDMA) and OFDM. OTFS modulates
each information symbol to a set of two-dimensional (2D) orthogonal
basis functions. When these basis functions are used as one-dimensional
(1D) spread spectrum codes, OTFS degrades to CDMA; when basis
functions are regarded as sub-channel carriers, OTFS becomes OFDM.
Time division multiple access (TDMA) and OFDM are the exceptional
cases of OTFS when the delay period tends to infinity and Doppler
tends to infinity, respectively [13]. Hence, OTFS has the advantages
Fig. 9. Comparison of BER between GFDM modulation and OTFS modulation under of TDMA, CDMA and OFDM, and it is also motived for investigating
different Doppler. the OTFS technology.

8
M. Li et al. Computer Networks 224 (2023) 109597

4. System related technology discussion Table 2

The relationship between parameters and waveforms.

Section 4 concentrates on the design of transmitter and receiver. Parameter 1 Parameter 2 Waveforms

From the point of view of system structure, the framework diagram of 0<𝛼<1 𝐿=𝑄 Hybrid Carrier (HC)
OTFS system is divided into transmitting end and receiving end. The 0<𝛼<1 1<𝐿<𝑄 WFRFT–OTFS/ asymmetric-HC (A-HC)
𝛼=0 1<𝐿<𝑄 OTFS/ A-OFDM
transmitter design includes channel coding, waveform design scheme, 𝛼=0 𝐿=𝑄 OFDM
precoding technology, etc. The receiver design is involved in synchro- 𝛼=1 𝐿=1 Single Carrier-FDE (SC-FDE)
nization, channel estimation, signal detection and equalization.

4.1. Transmitting end and its technology

To sum up, it is known from the existing literatures that the ideal
In the transceiver design, in-phase and quadrature (IQ) impairment pulse-shaping waveform is suitable as a tool for exploring OTFS per-
arises from inevitable imperfections between IQ branches, and litera- formance boundaries, and the most commonly used waveform is the
ture [37] establishes the mathematical model to illustrate the physical rectangular pulse-shaping waveform. Next, it is hoped to find a general
effect of the IQ mismatched transmitter, which results in a new type of waveform that can reach the rational boundary that is achieved by the
interference named mirror Doppler interference (MDI). The simulation ideal pulse-shaping waveform and realize further improvement of OTFS
shows that IQ imbalanced transmitter causes a significant reduction in system performance.
the diversity gain. In addition to the above problems, the design of the
receiver can be considered from the perspectives of waveform pulse 4.1.2. Channel coding
shaping, channel coding and precoding. To further improve transmission reliability, channel coding tech-
nology is applied to the OTFS modulation system. The architecture of
4.1.1. Pulse-shaping waveform design channel coding in OTFS system as shown in Fig. 13. At present, the
From the Eq. (26) to the Eq. (29) can be obtained, ideal pulse- most used channel coding techniques are LDPC codes and Turbo codes.
shaping waveforms are used that satisfy orthogonal conditions in both
time and frequency, which is not practically feasible due to Heisen- In literature [48], Liu et al. analyze the performance of uncoded
berg’s uncertainty principle [38]. Hence, rectangular pulse shaping and coded OTFS systems and OFDM systems with different iterations,
exploits instead of ideal pulse-shaping, but the main challenge of the and the results show that coded scheme is better than uncoded scheme.
rectangular pulse-shaping waveform is high OOB power. Excessive OOB Starting from the study of conditional pairwise-error probability (PEP),
power emissions result in adjacent channel interference (ACI) [39] the performance upper bound of the coded OTFS system is derived in
and degrade communication quality. In order to reduce OOB power literature [49], and the impact of channel coding parameters on the
radiation [40], a new pulse-shaping waveform is needed to develop. performance of the coded OTFS system is unveiled. In literature [50],
Mitigating the high OOB power emission is a crucial issue. For this 5G LDPC code is considered and the effect of channel estimation
purpose, Raviteja et al. [41] derive a simple form of input–output re- errors on LDPC coded OTFS system performance is evaluated as well.
lation from rectangular pulse shaping waveform to arbitrary waveform Similarly, Tian et al. [51] is dedicated to investigating nonbinary
in the DD domain, where the waveform used is a prolate spheroidal LDPC–OTFS system and propose a layered min–max decoding algo-
waveform (PSW) [42]. Biglieri et al. [43] use a general waveform to rithm based on the early termination criterion to reduce the decoding
compute error probability and the amount of diversity that can be complexity. Take LDPC code as an example, the simulation results are
achieved on fading channels. The general or arbitrary waveforms are shown in Fig. 12.
studied in the two waveform design, but the perspective considered The simulation parameters are set as speed = 30 km/h, 80 km/h,
differs. The waveform design proposed in literature [41] implements 130 km/h, respectively. When the value of BER is the same, it can be
a tradeoff between OOB and BER; The other waveform design pro- known that the performance of the LDPC–OTFS modulation scheme is
posed in literature [43] considers the number of diversities that can better than that of the LDPC–OFDM modulation scheme from Fig. 12;
be achieved. In addition, Bomfin et al. in literature [44] propose a similarly, When the value of Eb/No is the same, the LDPC–OTFS modu-
low-complex waveform based on sparse Walsh–Hadamard transform lation scheme is also better than the OTFS modulation scheme. The first
(WHT) to realize the same frame error rate performance as OTFS. A phenomenon proves the effectiveness of the LDPC–OTFS system. The
windowing and restructuring OTFS system to reduce OOB power is second phenomenon shows that channel coding can effectively improve
proposed by Hossain et al. [45]. What is more, Tiwari et al. [46] present the system’s performance and reduce the BER.
a circular pulse shaping framework named circular Dirichlet pulse To conclude, an effective coding method that can freely switch
shaped OTFS (CDPS) and decreases the PAPR of OTFS without any between low-speed and high-speed can be designed for future commu-
loss of BER, which using the block-circulant matrix decomposition. In nication scenarios.
literature [47], Wang et al. propose an integrated waveform framework
based on weighted-type fractional Fourier transform (WFRFT), and the 4.1.3. Precoding design
proposed framework has the advantage of two-dimensional adjustable
In the MU–MIMO system, the transmitter precoding can effectively
parameters and robust scalability.
eliminate multi-users interferences. For the traditional precoding algo-
Most newly waveforms are still designed on the basis of OFDM
rithm, there is high computational complexity to be realized in practice,
architecture. At present, it is necessary to derive a more generalized
especially since the BS need to communicate with multiple high-speed
waveform framework. Combined with the above literature, the relation-
users simultaneously under high mobility scenarios.
ship between parameters and waveforms is shown in Table 2. It can be
In literature [52], Cao et al. focus on studying the low complexity
seen from Table 2 that the WFRFT–OTFS framework is equivalent to
transmitter precoding for MU MIMO–OTFS, and the proposed precod-
OTFS when 𝛼 = 0. From the perspective of FFT, WFRFT–OTFS is able
ing algorithm significantly reduces the computational complexity with
to simplify as a Hybrid Carrier (HC) when 𝐿 = 𝑄. Moreover, OFDM
a slight performance loss compared with the block diagonalization (BD)
and SC also can be derived from OTFS or HC.
precoding method. To simplify the receiver design, literature [53] stud-
where 𝑄 is the length of data stream, and the 𝐾 × 𝐿 denotes the
ies the precoding methods for the CP–OTFS system, and the provided
size of delay–Doppler grid. 𝛼 is the order of 𝐿-point precoding matrix,
algorithm can pre-cancel the inter-slot-interference and dramatically
in which
improves the BER performance. From the perspective of security, Sun
𝐖𝛼,𝑁 = 𝑤0 (𝛼)𝐈𝑁 + 𝑤1 (𝛼)𝐅𝑁 + 𝑤2 (𝛼)𝐅2𝑁 + 𝑤3 (𝛼)𝐅3𝑁 (45) et al. [54] propose a rotate OTFS (R-OTFS) waveform to achieve both

9
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 12. Comparison between coded-OTFS modulation and uncoded-OTFS modulation.

become one of the significant problems and has not been solved in the
OFDM system. ICI not only degrades the system performance but also
hinders channel estimation and signal detection. The classical solution
is to design a shorter OFDM symbol in the time domain to overcome
the time-varying characteristics of the channel. Although some scholars
have studied ICI elimination or suppression scheme in the OFDM sys-
tem, the method proposed is complex and cannot be adapted to higher
dynamic communication scenarios. Recently, OTFS modulation was
proposed to apply in high mobility, which shows significant advantages
over OFDM. The difficulty of the channel estimation technique in OTFS
Fig. 13. The architecture of channel coding in OTFS system.
systems is how to estimate the channel state information in the DD
domain accurately.
The existing channel estimation schemes mainly distinguish the
secure and reliable transmissions while degrading the signal inter- antennas of different base stations by transmitting multiple pulses
cepted by the eavesdropper. In literature [55], Pandy et al. propose between two adjacent pulses and setting a reasonable protection in-
a novel multi-user OTFS-based low-complexity precoder to achieve terval. In massive MIMO systems [60], it is necessary to make use of
high spectral and energy efficiency and show more robust in channel transmitting quadrature pilots to distinguish antennas, which causes
induced Doppler spread. Moreover, Zhang et al. [56] propose an adap- relatively high pilot overhead. Therefore it is eager to investigate 3D
tive transmission scheme with a frequency-domain precoding matrix channel architecture to cope with massive MIMO systems based on
composed of the eigenvectors of the channel matrix, and the lower and OTFS technology.
upper bounds for diversity performance are derived under two extreme ∙ Delay–Doppler–Angle Domain Architecture. At present, most of
channel conditions. the research on OTFS–MIMO channels is focused on the DD plane. The
From the perspectives of hardware overhead and antenna, precod- study of massive MIMO channels is not sufficient alonely in the DD
ing designs can be combined with hybrid beamforming techniques [57] plane, so it is essential to add the angle dimension to spread the DD
to achieve higher diversity gain. domain to the DDA domain. The structure of the DDA domain channel
is shown in Fig. 15. In literature [20], Shen et al. demonstrate that
4.2. Receiving end and its technology OTFS massive MIMO channel exhibits a DDA structured sparsity. Since
the number of dominant propagation paths is limited, the 3D channel
Designing highly effective receivers is also one of the vital objectives is sparse along the delay dimension; As the Doppler frequency of a path
in the OTFS system, which has engaged crucial research attention. is usually much smaller than the system bandwidth, the 3D channel is
When the signals pass through the time-varying channel, it is sus- block-sparse along the Doppler dimension; Since the angle-of-departure
ceptible to interference such as IDI and ISI. Therefore, it is expected (AoD) spread of a path at the BS is usually small, the 3D channel is
for a receiver scheme that is simple in structure and can suppress burst-sparse along the angle dimension [61]. Based on the above 3D
interference in function. The framework of the OTFS-based receiver structure of the massive OTFS–MIMO channel, the channel estimation
scheme is illustrated in Fig. 14. Similar to transmitter design, Neelam technique is detailed and discussed. In addition, the Ref. [62] analyzes
et al. study the effects of receive IQ imbalance on the performance the multi-access method under this framework. Currently, the extent
of the OTFS system in literature [58], and the error flooring effect is of research on the 3D framework is limited that only investigates the
removed by pilot aided channel estimation. From the perspective of IQ channel estimation method and multi-access technology.
imbalance (IQI) compensation, Naikoti et al. [59] present a deep neural Currently, channel estimation schemes can be divided into two cat-
network (DNN) based transceiver architecture for channel training and egories. The first is to exploit the superposition property of multipath
detection. signals; the second is to study the sparsity of DD domain channels. In
the first category, better system performance is realized by employing
4.2.1. Channel estimation the pilot design in channel estimation. Kollengode et al. [63] propose
As the need for communication becomes more and more of a a simple and efficient channel estimation scheme that uses impulses in
concern in high-speed scenarios, the inevitable requirement for modu- the DD domain as pilots in the MIMO–OTFS system. The effect of syn-
lation technology for next-generation communication systems is able to chronization errors can be seen as part of the channel, literature [64]
support the tolerance of fast time-varying channels, where the channel leverages the appropriate channel coefficients to compensate for the
parameters are no longer constant in each symbol period. The time- effect of time-varying multipath and residual synchronization errors.
varying characteristics of wireless channel results in the ICI, which has In order to deal with fractional Doppler, Hashimoto et al. [65] give

10
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 14. The diagram of OTFS-based receiver scheme.

Fig. 17. (𝑙𝑒𝑓 𝑡) Tx symbols arrangement (×: data symbols; ◦: guard symbols; □: pilot).
(𝑟𝑖𝑔ℎ𝑡) Rx symbols arrangement (△: data detection; ⋄: channel estimation) under the
circumstances of fractional Doppler.

tensor-based orthogonal matching pursuit (OMP) channel estimation

algorithm to decrease the enormous computational complexity by ten-
Fig. 15. The delay–Doppler-angle domain channel architecture. sor decomposition and parallel support detection, which verify the
superiority and robustness. Zhao et al. [72] explores sparse Bayesian
learning (SBL) to highlight the superiority of pilot overhead, pilot
power consumption and anti-noise interference. In literature [73], a
novel method is proposed based on a joint sparse basis expansion
model to resist ICI. As same as literature [20], Shen et al. analyze
the 3D structured OMP algorithm in a massive MIMO–OTFS system.
Literature [74] also implements the 3D channel estimation over DDA
domain as shown in Figs. 16 and 17. Several works have explored the
off-grid methods to handle the fractional delay and Doppler shift. In
literature [75], Wei et al. study an off-grid channel estimation scheme
by adopting the SBL framework and regard the channel estimation
Fig. 16. (𝑙𝑒𝑓 𝑡) Tx symbols arrangement (×: data symbols; ◦: guard symbols; □: pilot). problem as a sparse signal recovery problem based on a virtual sam-
(𝑟𝑖𝑔ℎ𝑡) Rx symbols arrangement (△: data detection; ⋄: channel estimation) under the
pling grid. Similarly, the literature [76] cast the channel estimation
circumstances of integer Doppler.
problem as a block sparse signal recovery problem, which exceeds to
traditional BSBL method. The literature [77] presents a deterministic
pilot design and puts forward a modified sensing matrix based chan-
the corresponding solution. For the downlink channel estimation, Liu nel estimation (MSMCE) algorithm to acquire the downlink CSI. In
et al. [62] propose an uplink-aided method and adopt the expectation literature [78], Zhang et al. propose a novel frequency-domain pilot-
maximization based variational Bayesian (EM-VB) framework in the aided channel estimation scheme to obtain the CSI, which is lower than
3D downlink. In literature [66], a superimposed pilot (SP) based so- those of the current popular schemes. Additionally, the literature [79]
lution to improve spectral efficiency by the iteration. For both uplink investigates an iterative channel estimation and data detection design
and downlink, an embedded pilot-aided scheme is provided to avoid for the superimposed pilot (SP)-based system, which is able to improve
interference between pilot and data symbols in MIMO system [67]. its spectral efficiency. Moreover, the literature [80] proposes a low-
The method in literature [67] is similar to literature [68] proposed, complexity detection and low-overhead pilot pattern design based on
it mainly uses several guard zero-symbols to prove significant error a large-scale antenna array. Mohammed et al. in [81] numerically
performance advantages. In addition, Fish et al. [69] introduce a new investigate the receiver scheme, which can directly transform the time
flag method by using group representation techniques to compute the domain signal to the DD domain. Additionally, literature [82] applies
channel parameters. In literature [70], Zhang et al. proposed a 2D struc- the low complexity ZF and MMSE methods to receiver design with
tured Turbo compressed sensing algorithm and the MMSE estimator to imperfect receive CSI and then shows excellent performance compared
validate the convergence and MSE performance. The two pilot design with other design schemes. Furthermore, the marriage of OTFS and
schemes are shown in Figs. 16 and 17. large scale antenna array in receiver design can result in the low
The second categories mainly transform the channel estimation overhead pilot [83]. What is more, the low complexity MMSE receiver
problem into the sparse recovery problem. Wu et al. [71] present a is proposed in Ref. [84] without any performance degradation of BER.

11
M. Li et al. Computer Networks 224 (2023) 109597

For the fractional Doppler channel, Hashimoto et al. [85] propose a

solution to channel estimation with low complexity. For the MIMO-STC
system, the literature [86] presents a unique word (UW)-based channel
estimation approach based on frame optimization, the FER results
reveal that OTFS becomes more advantageous than OFDM for MIMO-
STC systems with higher order modulation and code rate. Similar to
literature [86], literature [87] introduces an Orthogonal Matching Pur-
suit with Binary-division Refinement (OMPBR) estimation algorithm
based on compressed sensing, because integer approximations cannot
accurately model non-integer delay and Doppler shifts in the channel.
In literature [88], Liu et al. parameterize the DD domain channel by Fig. 18. (𝑙𝑒𝑓 𝑡) message passing from observation nodes to variable nodes; (𝑟𝑖𝑔ℎ𝑡)
estimating the channel gains and fractional Doppler shift and also message passing from variable node to observation nodes.
develop an efficient message passing algorithm. To mitigate the tandem
off-grid distortion, a 2D off-grid decomposition and SBL combination
method is proposed in the literature [89], and the simulation results of synchronization. Cramer–Rao low bound (CRLB) provides a bench-
show the effectiveness of the proposed methods. Ge et al. in [90] mark for different channel estimation algorithms. In literature [104], a
present a fractionally spaced sampling (FSS) receiver in which the discrete-time system model for non-ideal pulse-shaping OTFS is derived
sampling rate is an integer multiple of symbol rate and demonstrates and the CRLB of channel estimation for SCP-OTFS, MCP-OTFS, and
the more robust performance and robustness of CSI. The Ref. [91] multiple cyclic prefix approximation (MCPA) OTFS is provided.
designs a novel receiver based on multi-antennas under the V2X system. Channel estimation is a fundamental task in wireless communica-
In addition, a data-pilot channel estimation to improve the spectral tions, which refers to calculating the channel parameters experienced
efficiency and the proposed method can approach the baseline scheme by the signal as it travels from the transmitter to the receiver. In high-
with perfect channel estimation in literature [92]. Long et al. [93] speed mobile scenarios, higher pilot overhead and pilot power are
present a joint channel estimation and equalization approach based on required to estimate the channel state information due to the Doppler
the expectation propagation (JCEE-EP), and the algorithm has advan- shift. In the above proposed method, the perspective can be considered
tages in BER performance. The literature [94] constructed an MMSE in terms of spectral efficiency, pilot overhead, computational complex-
DD domain channel estimator, and the proposed algorithm performs ity, the MSE performance, etc., which can provide better adaptability
better than the conventional algorithm in the literature both in terms and higher reliability in dynamic scenarios. Future work can be focused
of the channel estimation accuracy and the bit error rate performance on the 3D channel estimation of massive MIMO–OTFS in the DDA
at the cost of fewer guard symbols. The literature [95] investigates a domain.
method that exploits smoothness regularization in TF to suppress the
leakage observed in the DD domain, which can improve the channel 4.2.2. Signal detection
estimation accuracy while reducing signaling overhead. Li et al. study At the receiving end, the received signals are required to detect.
a 3D dynamic support detection (DSD) algorithm to solve the problem One of the bases for judging the advantages and disadvantages of the
that the pilot overhead grows linearly with the number of antennas. detection algorithm is the complexity. Hence, it is urgent to design a
Compared with the traditional OMP and 3D-structured orthogonal low-complexity detection to decrease the difficulty.
matching pursuit algorithms, the proposed dynamic support detect From the Eq. (18), the input–output relationship can be denoted
algorithm has higher channel estimation accuracy and lower pilot over- as the linear system. In addition, the introduction of Eq. (13) makes
head. In literature [96], Qu et al. propose a corner-inserted pilot pattern it become a sparse linear system where each received signal can be
that targets the low pilot overhead and satisfactory CE performance. expressed as a linear combination of only a few transmitted signals.
The literature [97] proposes a prior channel statistics-based scheme Therefore, the low complexity detection algorithm is needed to explore
to maximize the ergodic system capacity. The results demonstrate the when the channel is sparsity. Moreover, OTFS modulation is able to
superiority of the proposed scheme. In literature [98], an improved offer the potential to exploit the full channel diversity compared with
Basis Expansion Model (BEM) combined with an Extended Kalman conventional OFDM modulation. In order to achieve the potential full
Filter (EFK) channel estimation method to reduce the complexity while channel diversity, advanced detection methods are also required for
tracking the dynamic changes of the channel. Similar to literature [98], OTFS detection. The most commonly used detection method is iterative
Qin et al. [99] adopt BEM combined with LMMSE channel estimation detection based on message passing (MP).
method to estimate channel impulse response. Compared with the Firstly, the symbol-by-symbol maximum a posterior probability
OFDM system, the OTFS applied to the presented method has a more (MAP) detection [105] rule is used as follow.
considerable SNR gain and improved symbol error rate performance. ̂ = 𝑎𝑟𝑔𝑚𝑎𝑥𝑃𝑟 (𝑥|𝑦, 𝐻)
𝑥[𝑐]
In literature [100], a low-order generalized complex exponential BEM
= 𝑎𝑟𝑔𝑚𝑎𝑥𝑃𝑟 (𝑥[𝑐] = 𝑎𝑗 |𝑦, 𝐻)
(GCEBEM) aided rough channel estimation is proposed by deriving an
1 (46)
analytical BEM OTFS system model. In literature [101], Li et al. present = 𝑎𝑟𝑔𝑚𝑎𝑥 𝑃𝑟 (𝑦|𝑥[𝑐] = 𝑎𝑗 , 𝐻)
𝑄
a variational autoencoder (VAE) based receiver for OTFS modulation ∏
that achieves a joint estimation and detection without a pilot in the ≈ 𝑎𝑟𝑔𝑚𝑎𝑥 𝑃𝑟 (𝑦[𝑑]|𝑥[𝑐] = 𝑎𝑗 , 𝐻)
DD domain. where 𝑄 denotes the number of QAM symbols. 𝑥[𝑐] denotes variable
In addition to the high-complexity issue, the existing pilot pattern to nodes and 𝑦[𝑑] denotes observation nodes. In order to further solve
estimate the unknown channel accurately requires significant overhead approximate symbol-by-symbol MAP detection, the low complexity
to avoid the pilot being contaminated, which is spectrally inefficient. message passing algorithm is proposed [106] as illustrated in Fig. 18
From the perspective of improving the spectrum efficiency without and is mainly divided into following five steps: (1) Message passing
degrading the channel estimation and symbol detection performance, from observation nodes 𝑦[𝑑] to variable nodes 𝑥[𝑐] where the mean
literature [102] focuses on the pilot design and optimization and an and variance is needed to computed; (2) Message passing from variable
OTFS pilot sequence optimization method based on particle swarm to nodes 𝑥[𝑐] to observation nodes 𝑦[𝑑]. The probability mass function
address the pilot contamination. In literature [103], a low complexity (pmf) is also needed to update; (3) The computation of convergence
channel estimation scheme is obtained to accommodate for the coef- indicator; (4) the update decision of transmitted symbol; (5) Stopping
fects of TVMC as well as the errors that occur during the procedure criteria which can choose according to different conditions.

12
M. Li et al. Computer Networks 224 (2023) 109597

Table 3
The overview of detection algorithm and its complexity.
Paper Scenario MP optimization Other technologies Complexity
Raviteja et al. [106] 𝑆𝐼𝑆𝑂 Message passing 𝑂(𝑛𝑖𝑡𝑒𝑟 𝑆|𝐴|)
Murali et al. [107] 𝑆𝐼𝑆𝑂 Markov chain Carlo sampling 𝑂(𝑀 3 𝑁 3 )
Pandey et al. [108] 𝑀𝐼𝑀𝑂 Multi-users precoder 𝑂(𝑀𝑁log(𝑀𝑁))
Li et al. [109] 𝑆𝐼𝑆𝑂 Improved message passing 𝑂(𝑀𝑁𝑆) + 𝑂(𝑀𝑁|𝐴|)
Yuan et al. [110] 𝑆𝐼𝑆𝑂 Approximate message passing 𝑂(𝑀𝑁log(𝑀𝑁) + 𝑂(𝑀𝑁|𝐴|))
Zemen et al. [111] 𝑆𝐼𝑆𝑂 Orthogonal precoding
Li et al. [112] 𝑆𝐼𝑆𝑂 Hybrid MAP and PIC
Liang et al. [113] 𝑆𝐼𝑆𝑂 Index Modulation 𝑂(𝐶𝑛𝑘 )
Yuan et al. [114] 𝑆𝐼𝑆𝑂 Variational Bayes approach 𝑂(|𝐴|𝑁𝑀𝑃 2 )
Kollengode et al. [63] 𝑀𝐼𝑀𝑂 Iterative detection 𝑂(|𝐴|𝑁𝑀𝑃 2 )
Ge et al. [90] 𝑆𝐼𝑆𝑂 ICMP and TMP algorithm 𝑛𝑖𝑡𝑒𝑟 𝑀𝑁𝐺𝑅(11𝑄 + 1)
Cheng et al. [91] 𝑀𝐼𝑀𝑂 Maximum ratio combining 𝑂(|𝑃 𝑀𝑁𝑆|𝐴|)
Qu et al. [115] 𝑆𝐼𝑆𝑂 B-PIC-DSC 𝑂(𝑀 3 𝑁 3 + 𝑀 2 𝑁 2 𝑛𝑖𝑡𝑒𝑟 )
Gong et al. [116] 𝑀𝐼𝑀𝑂 Transform-domain maximal ratio combining 𝑂(|𝑄|𝑀𝑁log𝑀𝑛𝑖𝑡𝑒𝑟 )
Enku et al. [117] 𝑀𝐼𝑀𝑂 2D-CNN 𝑂(𝑀𝑁)
Xu et al. [118] 𝑀𝐼𝑀𝑂 Damped GAMP 𝑂(𝑀𝑁𝑆𝑇 )

According to the above proposed low complexity MP algorithm, detection is proposed [114]. What is more, Cheng et al. [91] propose a
the better detection algorithms have received profound attention from joint MP-maximum ration combining (MRC) iterative detection method
academia [17]. In literature [118], a damped generalized approximate to prove the convergence performance of iteration.
message passing (GAMP) algorithm is proposed to reduce the receiver In literature [123], Liu et al. propose a novel detection method
complexity, which can outperform the classical GAMP algorithm and based on a sum–product algorithm (SPA) with low complexity to
the MP algorithm. The literature [119] proposes a low complexity achieve near-optimal MAP detection. Wu et al. [124] demonstrate that
matched filtering based message passing detector (MF-MPD), which the OTFS-BFDM system is sparsely-connected, and the low-complexity
combined with a novel probability clipping (PC) solution, and the MP decoding algorithm can be adopted. The literature [115] provides
complexity and storage requirement of this solution can be reduced, a Bayesian based parallel interference and decision statistics combining
and better BER performance in low SNR regimes also can be achieved. (B-PIC-DSC) OTFS detector, which can achieve a high BER perfor-
In the literature [120], Xiang et al. conceive a novel low-complexity mance under high ICI environments. In literature [125], Gao et al.
Gaussian approximate MP (GA-MP) detector, which models the a pos- investigate an efficient clipping noise recovery method using reliability-
teriori probabilities propagated through the proposed joint factor graph based compressed sensing (CS) is proposed to recover the clipping
by Gaussian distributions. The literature [121] presents a unitary AMP noise efficiently. The literature [126] proposes a novel cross domain
(UAMP) which provides the robustness of UAMP relative to AMP. In iterative detection algorithm to enhance the error performance of OTFS
literature [117], EnKu et al. propose a two-dimensional convolutional modulation, which is also applied to fractional Doppler channels. In
neural network (2D-CNN) based detector and also use a MP algo- literature [116], Gong et al. design a signal detector for OTFS modula-
rithm to improve the learning ability of the proposed method. The tion based on transform-domain maximal ratio combining (TD-MRC),
literature [122] introduces the average approximation to circumvent and its reliability performance improves as the maximal relative ve-
the matrix inversion operation and then develops a low-complexity locity increases. In literature [127], an enhanced data detection (EDD)
VAMP-based algorithm, which outperforms the existing MMSE algo- scheme to iteratively demodulate the superposed multi-antenna signal
rithm and the MP algorithm in terms of BER performance. In addition, is developed, which can realize the linear-complexity interference can-
Ge et al. [90] investigate an iterative combining message passing cellation. In literature [128], a low complexity BP detection algorithm
(ICMP) algorithm and turbo message passing (TMP) algorithm. In the is provided and also adds a damping factor to improve the convergence
MIMO–OTFS system, an iterative detective algorithm is also proposed of the proposed algorithm. The proposed detection scheme effectively
in [63] based on the MP algorithm. The above research is based on the improves the bit error rate (BER) performance for OTFS systems. From
traditional MP algorithm. the deep learning perspective, Naikoti et al. investigate the problem of
A report by Murali et al. [107] reveals that the Markov chain Monte low complexity detection by using deep neural networks [129], which
Carlo Sampling (MCMC) is also can be applied to the detection method have less complexity. The Ref. [130] presents a sample-based method to
to decrease the complexity further. Pandey et al. [108] propose a novel estimate spatial correlation and find the optimized combining weights
multi users-based precoder and a corresponding detector, and it can for MRC from the estimated correlation matrix. The proposed detec-
achieve high spectral and energy efficiency in the downlink. Compared tor performance and complexity improve over the LMMSE and MP
with the low complexity MP detection proposed in literature [106], detectors. In literature [131], the transmission codeword is designed
Li et al. [109] and Yuan et al. [110] both explore the approximate and a low complexity linear detection is proposed, the bit error rate
message passing algorithm from different views. Li et al. exhibit better performance of the proposed scheme has superiority over the existing
BER performance by covariance processing, while Yuan et al. use works. From the perspective of the upper and lower bounds, the
the unitary transformation to allow efficient implementation. Parallel literature [132,133] derives closed-form BER expression for OTFS-ZF
interference cancellation (PIC) is also an excellent way to deal with receiver with perfect and imperfect receive channel information. From
detection. The literature [111] and literature [112] both study the the perspective of the MIMO system, Singh et al. [134] designs and
application of PIC, but literature [111] investigate the scalar signal analyzes low-complexity ZF and MMSE receivers with perfect and
model for orthogonal precoding (OP), and the literature [112] propose imperfect receive channel state information (CSI). The results show the
a hybrid MAP and PIC detection algorithm and a partition rule to sep- lower BER and lower complexity of the proposed designs.
arate symbols which can approach the near-optimal symbol-wise MAP At present, a part of the method is used to improve detection
algorithm. Liang et al. [113] study an integrated MMSE-ML detection performance by combining with other techniques, and another part
strategy under index modulation (IM) based on the OTFS system (OTFS- of the method is optimized the traditional MP algorithm by iterative
IM) to achieve better BER and maintain good tolerability against the operation. The majority of methods can reduce the complexity of the
Doppler effect. In order to improve the convergence rate under the detection algorithm to some degree. The intricate complexity is shown
traditional MAP algorithm, the variational Bayes approach based on in Table 3.

13
M. Li et al. Computer Networks 224 (2023) 109597

Note: 𝑀 and 𝑁 are the numbers of sub-divisions of the delay and

Doppler domain, respectively; 𝑛𝑖𝑡𝑒𝑟 is the number of iterations; 𝐴 is
the modulation alphabet; 𝑛 = 𝑀𝑁∕𝑔, 𝑔 represents groups of incoming
data bits; 𝑃 indicate the number of propagation paths; 𝑆 represents the
sparsity level of channel matrix; 𝑄 represents the numbers of Doppler
taps; 𝑇 𝐴 is the number of cascaded layers.

4.2.3. Equalization
The equalization scheme is essential to the performance of a prac-
tical OTFS system. Compared with the linear equalizations, non-linear
equalizations are more compatible with the OTFS system. In non-linear
equalizations, the complexity is key that should be highlighted, and
Fig. 19. Date multiplexing in DD domain.
the complexity of non-linear equalizations is subjected to the sparsity
of channels in the DD domain. Hence, it is essential to investigate the
low complexity equalization schemes.
5. Other technologies and applications
MMSE [135] and ZF [136] are the traditional and classical equaliza-
tion design approaches. Nevertheless, the conventional MMSE and ZF Section 4 describes the pulse-shaped waveform design, channel cod-
equalizers involve the inversion of matrices for computing, and their ing, precoding design in the transmitting end, and channel estimation,
computational complexity is high, especially when the dimensions of signal detection, and equalization in receiving end. Section 5 mainly
matrices are large. Therefore, a large number of studies have emerged analyzes the performance of OTFS from other technologies. The specific
recently. Pfadler et al. in [137] propose improved MMSE equalizer aspects include diversity gain, PAPR, multiple access technology, other
to instantaneously estimate the self-interference power. Ref. [138] field applications, etc. The detailed presentation is as follows.
provides mobility modes with distinct grids and pulse matching for
different doubly dispersive channels. The reports by Cheng et al. [139] 5.1. Diversity analysis
and Xu et al. [140] both investigate the optimal ZF and MMSE equaliz-
ers to reduce the computational complexity. Literature [141] proposes Diversity is one parameter that needs to be studied in OTFS sys-
tems. Full diversity refers to the number of clustered reflectors in
the linear equalizer for MIMO–OTFS system. Thaj et al. [142] study the
the channel. That is to say, full diversity is the number of multipath
iterative decision feedback equalizer (DFE) based on MRC to improve
components separable in either the delay or the Doppler dimension. In
the performance further. Linear minimum mean square error based PIC
literature [155], the full diversity can be achieved in the DD domain,
(LMMSE-PIC) equalizer is adopted in [143] which can achieve great but no detailed proof process is given. When channel fading is fast
gain. Li et al. [144] propose a simple two-stage equalizer to remove with a more significant Doppler frequency shift, OTFS exploits both
the multipath effect and eliminate the effect of residual interference. time and frequency diversity, providing a significant improvement by
In addition, [145] realizes the LMMSE equalizer in the presence of combating the Doppler effect. Hence, many scholars have researched
hardware impairment. Furthermore, Zemen et al. [146] also study diversity techniques.
the MMSE equalizer by using soft-symbol feedback for interference Diversity technology is analyzed from two aspects. The first aspect
cancellation. In literature [147], the MMSE equalizer in the TF domain is the proof process of the diversity technology; another is the imple-
is considered for improved TF linear domain receiver, which has lower mentation process of the diversity technology. As shown in Fig. 19,
complexity. The literature [148] investigate a tuned one-tap minimum The N-point FFT extends each data symbol in 𝑥 into each segment
mean square error equalizer (MMSEEQ) with low complexity over of 𝑠 so that each data symbol experiences a different time shift as it
doubly dispersive channels. In literature [149], a MMSE based turbo passes through the channel due to the Doppler shift; after the received
equalization scheme is provided as a low complexity version requiring signal 𝑟 is converted to 𝑦, the received signal achieves the time diversity
provided by the fast fading channel. OTFS shows similar characteristics
only a single matrix inversion. The results verify that OTFS is superior
to SC-FDMA. When 𝑀 ⩾ 𝐿, it can achieve full frequency diversity.
to OFDM in the performance of block error rate (BLER).
When 𝐾 ⩽ 𝑁, full-time diversity can be exploited in the Doppler
In literature [150], Zou et al. propose two low complexity linear domain. Therefore, OTFS can achieve full diversity in the DD domain
equalizers, where FFT and lower–upper (LU) factorization are effi- with diversity order 𝐾 × 𝐿.
ciently leveraged to reduce the complexity, and the proposed equalizers In addition to the proof of full diversity, scholars at home and
improve the performance without additional complexity. In Ref. [151] abroad have researched how to implement full diversity techniques.
an efficient block equalizer for OFDM-based OTFS via expectation Surabhi et al. [156] present a formal analysis of the diversity achieved
propagation (EP) while reducing the overall equalizer complexity or- by OTFS modulation along with supporting simulations and propose a
der from cubic to log-linear. The literature [152] considers a neural phase rotation scheme. As shown in Fig. 20, the utilization of space–
network-based supervised learning framework for OTFS equalization time coding in literature [157], and firstly investigates the use of
with lower complexity. Additionally, a novel block decision feedback space–time coding along with phase rotation to achieve full diversity
equalization (BDFE) algorithm is constructed using the band feature of under the finite SNR condition in MIMO–OTFS, which can be applied to
the channels to combat both interferences jointly in [153]. Moreover, low-latency applications. According to [157], the asymptotic diversity
in literature [154], an efficient improved approximate message passing order of 2 × 1 phase rotated STC–OTFS is defined as
{ ( ) }
(AMP) equalizer with better BER performance and lower complexity. 𝜙
𝜌STC−OTFS = min min rank 𝜟̃ 𝑖𝑗 , 2𝑃 (47)
For the investigation of equalization, the linear equalizations and 𝑖,𝑗

the non-linear equalizations schemes are considered the basis for divi- ̃ 𝑖𝑗 ≜ 𝐗̃ 𝑖 − 𝐗̃ 𝑗 , 𝐗̃ 𝑖
where 𝑃 represents the number of channel taps, ∆
sion. The complexity is a crucial judgment in evaluating the perfor- and 𝐗̃ 𝑗 are two distinct STC–OTFS codeword matrices, respectively,
mance of proposed equalizers. Hence, the low-complexity equalizers and 𝜙𝑖 = 𝑒𝑗𝑎𝑖 . In literature [156], the asymptotic diversity order of
have become an integral part of the OTFS system. SISO-OTFS is one. Based on [156], the asymptotic diversity order of

14
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 20. The space–time coding in OTFS system.

2 × 1 STC–OTFS is two when 𝑃 ≥ 1. Additionally, when 𝑃 ≤ 𝑀𝑁, where 𝑅𝑛 (0) and 𝑅′′ 𝑛 (0) are connected to pulse shape 𝑔𝑡𝑥 (𝑡). When the
the asymptotic diversity order of 2 × 1 phase rotated STC–OTFS is 2𝑃 . 𝑔𝑡𝑥 (𝑡) is the rectangular pulse, due to 𝑅𝑛 (0) = 1 independent to 𝑡 and
When the system model is extended to 2 × 𝑛𝑟 MIMO case, only 2𝑃 𝑛𝑟 𝑅′′
𝑛 (0) = 0, 𝑠(𝑡) is a stationary Gaussian random process. Hence, the
diversity is attained in MIMO–OTFS with ML detection. Eq. (43) can be rewrite as follow.
From the spectral efficiency perspective, Raviteja et al. [155] study ( √𝜋 )
the diversity of OTFS assuming rectangular waveforms and a delay– −𝑒−𝛾 𝑁 3(
𝑀 2 −1) log(𝑀𝑁) (51)
𝑝rect { PAPR > 𝛾} ≈ 1 − 𝑒
Doppler channel with two paths. Similar to [156], Zhang et al. [158]
examine the diversity order through the received signal to noise ratio In literature [162], Surabhi et al. firstly derive an upper bound on
analysis under the 5G tapped-delayline channel model. All of the above the PAPR of OTFS signals and prove the relation between maximum
methods of implementing diversity make use of the phase rotation PAPR and the number of subcarriers. The PAPR of the OTFS transmit
scheme in the OTFS system. signal can be upper bounded as
Currently, there is relatively little research on the diversity tech-
𝑀 2 𝑁 2 max𝑘,𝑙 |𝑥[𝑘, 𝑙]|2 𝐵1 𝑁 max𝑘,𝑙 |𝑥[𝑘, 𝑙]|2 𝐵1
nology of OTFS systems. The timing/frequency offset [159], synchro- PAPR ≤ = (52)
nization effects [160] and link adaptation in OTFS can be considered 𝑀 2 𝑁𝜎𝑎2 𝐵2 𝜎𝑎2 𝐵2
for future work. More robust systems targeting ultra-reliable and low- ∑ ∑𝑁−1 ∑𝑀−1
where 𝐵1 ≜ 𝑚𝑎𝑥 𝑁−1 2
𝑛=0 |𝑔𝑡𝑥 ([𝑟 + 𝑞𝑀 − 𝑛𝑀]𝑀 𝑁)| , 𝐵2 ≜
latency communication can be considered with proper configuration ∑𝑁−1 2 2
𝑛=0 𝑟=0
𝑞=0 |𝑔𝑡𝑥 ([𝑟 + 𝑞𝑀 − 𝑛𝑀]𝑀 𝑁)| , and 𝜎𝑎 = E|𝑥[𝑘, 𝑙]| . It can be known
2
for M, N, coding, and considering the natural presence of residual syn-
chronization effects in combination with the proposed rotation scheme. that a bound on the PAPR of transmit signal is also related to pulse
From another perspective, the new methods to achieve full diversity shape 𝑔𝑡𝑥 (𝑡). When the rectangular pulse shape is considered, the value
also need to be explored in the OTFS system. of 𝐵1 ∕𝐵2 is equal to 1, and the formula (18) can be rewrite as

𝑁 max𝑐∈A |𝑐|2
5.2. Peak-to-average power ratio PAPRmax = (53)
𝜎𝑎2
Since the information symbol is superimposed by multiple indepen- Note that the upper bound on the PAPR of the OTFS signal grows
dently modulated subcarrier signals, when the subcarrier phase is the linearly with the number of Doppler bins 𝑁 and not with the number
same or close, the superimposed signal will be modulated by the same of subcarriers 𝑀. This consequence is different from other multicarrier
initial phase signal, resulting in a more prominent instantaneous power modulations where PAPR depends on the number of subcarriers.
peak. The above problem is likely to cause the signal to enter the non-
The above analysis mainly gives the mathematical expression of
linear region of the power amplifier quickly, leading to a spectrum
PAPR from the perspective of performance boundary as shown in (43)
spread interference and in-band signal distortion that generate the
severe degradation of system performance. Hence, the new method is and (45). From the perspective of technical ways to reduce PAPR, Gao
developed urgently to reduce the effect of PAPR. et al. [163] propose an efficient PAPR reduction method based on the
Generally, the PAPR of the OTFS signal is defined as iterative clipping and filtering (ICF) framework by permitting the guard
[ ] symbols to be padded with signals of small magnitudes, which can
max |𝑠(𝑡)|2
PAPR = [ ] (48) achieve the tradeoff between PAPR and BER. Similarly, literature [164]
E |𝑠(𝑡)|2 provides an autoencoder structure and designs loss function to mini-
where E[⋅] denotes the expectation operator. Since the PAPR is a mize the BER and PAPR at the same time. Naveen et al. [165] also study
random variable, the complementary cumulative distribution function a conventional MuLaw companding technique under the condition of a
(CCDF) can be utilized to characterize the statistical properties of PAPR, slight degradation in BRE performance. Moreover, Surabhi et al. [162]
which can be expressed as present the simulated CCDF for different pulse shapes compared with
GFDM. In addition, Hossain et al. [166] detailed analyze the PAPR
𝐶𝐶𝐷𝐹 = 𝑃 (𝑃 𝐴𝑃 𝑅 > 𝛾) (49)
characteristics and propose a concept of a DFT-Spread OTFS system
where 𝛾 denotes a constant, and 𝑃 (⋅) presents the probability function. to reduce PAPR efficiently. Furthermore, indexing technology is also a
To further detailed analyze the impact of PAPR, Wei et al. [161] reveal feasible solution to solve the high PAPR. Francis et al. [167] enhances
that the complex-valued OTFS signal weakly converges to a Gaussian the asymptotic diversity order to two by eliminating all the rank one
distribution when the number of data points near infinity, and an difference matrices. Meanwhile, the above method also offers improved
accurate expression of the PAPR distribution is derived. When 𝑀, 𝑁
PAPR performance of OTFS.
approach infinity, the CCDF for the PAPR of 𝑠(𝑡) can be approximated
PAPR Reduction of the OTFS system is inevitable to improve the
as statistical properties of PAPR, which can be expressed as
efficiency of a high power amplifier and consequently improve the com-
𝑝{ PAPR > 𝛾} munication quality in nonlinear situations. By referring to decreasing
( √( ) )
𝜋 ∑𝑁−1 2 ∑𝑁−1 ′′ log(𝑀𝑁) (50) the PAPR in the OFDM system further investigate the PAPR problem in
−𝑒−𝛾 𝑁 3 (𝑀 2 −1) 𝑛=0 𝑅𝑛 (0)− 𝑀𝜋 𝑛=0 𝑅𝑛 (0) 𝑁
.
≈1−𝑒 the OTFS system.

15
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 21. The schematic diagram of SCMA model based on OTFS. (a) describes the block diagram of OTFS–SCMA in downlink. (b) depicts the block diagram of OTFS–SCMA in
uplink.

5.3. Multiple access while guaranteeing the high mobility users’ targeted data rate. In order
to harvest both multipath and multiuser diversity, Ding et al. [176]
In the radio wave coverage of the wireless communication environ- also present a new NOMA-OTFS transmission protocol which can be
ment, the problem of establishing the connection of wireless channels applied to both downlink and uplink, it can remove inter-symbol
between users is multiple access (MA) technology. MA technology can interference in a way. Consider the optimization from uplink, the
be divided into orthogonal multiple access (OMA) and non-orthogonal literature [168] explore a novel method by allocating non-overlapping
multiple access (NOMA). In literature [168,169], OMA methods are portions and have significantly higher spectral efficiency. The three dif-
proposed, which are shown to achieve multiuser interference (MUI) ac- ferent delay–Doppler resource blocks allocation scheme is also provided
cessible communication with the ideal pulses. However, the problem of in literature [169] to solve uplink. In addition, Agustine et al. [177]
MUI is not studied under practical rectangular pulses. Literature [170] propose an interleaved time–frequency multiple access (ITFMA) to
derives the expression for the achievable sum SE, and the simulations reduce the detection complexity and improve bit error performance.
show that the sum SE with rectangular pulses is improved compared From the code domain perspective, OTFS–SCMA is a worth studying
with [168,169]. technology to apply for massive MIMO communication. A downlink
As a complement to OMA technology, NOMA is a popular multiple scenario with 𝐾 receiving users and one transmitting BS is described
access framework particularly suited for applications involving mas- in Fig. 21(a) and an uplink scenario with K transmitting users and one
sive connectivity [171]. In a conventional NOMA network, spectrum receiving BS is depicted in Fig. 21(b).
sharing among multiple users is encouraged because users can be
distinguished based on channel conditions. According to the current 5.3.1. Downlink communication
study contents, the NOMA schemes can be divided into two domains: As shown in Fig. 21(a), the superimposed input signal is given by
power domain [172] and code domain [173]. In power domain NOMA,
different users are identified using different assigned power levels. ∑
𝐾
𝐗sum = 𝐗𝑘 (54)
In code domain NOMA, the multiple users are distinguished using 𝑘=1
different codewords. A typical representation of code domain NOMA
where 𝐗𝑘 is the delay–Doppler symbol frame for 𝑘th user. The signal
is sparse code multiple access (SCMA) [174]. Therefore, these two
𝐗sum goes through the OTFS modulator to produce the vector 𝐒sum .
domains can be used as the basis for discussion.
Hence, the input–output relationship between the BS and the 𝑘th user
From the power domain perspective to review the combination of
becomes as follow.
OTFS and NOMA technologies, Ding et al. [175] transform a beam-
forming design to maximize the low mobility NOMA users’ data rate 𝐲𝑘, vec = 𝐇𝑘 𝐬sum,vec + 𝐳𝑘 (55)

16
M. Li et al. Computer Networks 224 (2023) 109597

( ) ( )
where 𝐬sum,vec = vec 𝐗sum , 𝐲𝑘,vec = vec 𝐘𝑘 , 𝐇𝑘 is the 𝑀𝑁 × 𝑀𝑁 5.4. More application field
complex coefficient matrix, and 𝐙𝑘 is the complex AWGN. The output
signal 𝐲𝑘,vec goes through OTFS detector and SCMA detector to generate Apart from above application of OTFS, the OTFS technology also
the estimate value 𝐱̂ sum, vec . can be adapted in other fields, such radar [185], vehicular commu-
nication [186], millimeter wave (mmWave) communication [187],
5.3.2. Uplink communication terrestrial broadcast [188] and so on. The detailed analysis below and
As shown in Fig. 21(b), the input–output expression of 𝐲𝑘, vec is the application of OTFS technology in other fields are summarized in
given as Table 4.
∑
𝐽
𝐲vec = 𝐇𝑘 𝐱𝑘, vec +𝐳 (56) 5.4.1. Terrestrial broadcast communication
𝑘=1 The fast development of mobile communication makes the strong
( ) demand for mobility video, which is provided the higher require-
where 𝐱𝑘,vec = vec 𝐗𝑘 , and 𝐳 is the complex AWGN. We rewritten the
Eq. (49), and the new expression is given as ment on aspects of data transmission rate and spectrum efficiency.
For example, Advanced Television Systems Committee (ATSC), (Digital
𝐲vec = 𝐇all 𝐱all + 𝐳 (57) Video Broadcasting Terrestrial) DVB-T, or Integrated Services Digital
where Broadcasting Terrestrial (ISDB-T). These standards can overcome the
[ ] shortcoming of analog wireless TV, which is susceptible to interference,
𝐇all = 𝐇1 , 𝐇2 , … , 𝐇𝐾 (58) and significantly improve the utilization of the wireless spectrum.
[ ]𝑇 However, the currently broadcasting standards cannot support the case
𝑇
𝐱all = 𝐱1, 𝑇
, 𝐱2, 𝑇
, … , 𝐱𝐾, (59) of extensive coverage with the high-mobile speed in the same mode
vec vec vec
due to the limitation of OFDM modulation in the contradiction between
From formula (52), it can be seen that 𝐇all and 𝐱all is not the longer CP and larger subcarrier spacing. Therefore, the new modulation
superposition of SCMA data from all users, but a vertical concatena- technology is proposed to settle the above problems, which is called
tion of them. Therefore, the delay–Doppler interference of the OTFS OTFS technology.
modulator and the multiuser interference of SCMA cannot be separated. In literature [189], a robust broadcast system under time-varying
Based on above Figs. 21(a) and 21(b), the signal detectors designed channels based on OTFS technology is proposed, and the simulation
are different for uplink and downlink communication. For example, a results show that the proposed broadcast system outperforms the con-
simple LMMSE-based detector is used in literature [178] for downlink, ventional ATSC 3.0 system under high Doppler shift environments.
and the 𝐱̂ sum, vec can be obtained as follow. The literature [190] builds the bridge between OTFS and LTE-based
[ ]−1 terrestrial broadcast in high Doppler scenarios. In addition, Blazek
𝐱̂ sum,vec = 𝐇† 𝐇𝑘 𝐇† + 𝜎𝑛2 𝐈𝑀𝑁
𝑘 𝑘
𝐲𝑘, vec (60)
et al. [191] present an analysis when it is applied to a realistic vehicular
The basic system framework of SCMA based on OTFS is given in channel situation under two different conditions. Combined with the
Fig. 21. For the design of the signal detectors, scholars at home and above research status, OTFS will be a promising technology to further
abroad have made many contributions. Deka et al. [178] propose a study in the broadcast when high mobility is required.
multi-user OTFS based on SCMA both in downlink and uplink, which
can yield better BER performances than a traditional OTFS–OMA sys- 5.4.2. Vehicular-to-everything communication
tem. Kang et al. [179] also propose a two-dimension OTFS–NOMA (2D– Vehicular-to-Everything communication has presented the strict re-
OTFS–NOMA) transmission scheme based on heterogeneous mobility quirements for wireless channels, especially in high-speed mobile sce-
users grouping. That is to say, power-domain NOMA is fully utilized narios. Wireless channels are displayed the doubly selective character-
between the users under high mobility, and the SCMA technique is istics where multipath effects result in ISI and Doppler shifts due to
applied to the users under low mobility. Similarly, Ge et al. [180] the time selectivity. The currently used modulation scheme performs
propose an approximate orthogonal message passing with linear mini- weakly in a high-mobility environment, such as OFDM shows excellent
mum mean squared error (OAMP–LMMSE) and a Gaussian approximate performance against multipath effects. However, it is vulnerable to the
message passing with expectation propagation (GAMP–EP) based on ICI caused by high Doppler shifts. The two available ways to solve
successive interference cancellation (SIC) to overcome the co-channel high Doppler shift have been studied: the first way is pulse-shaping
interference (CCI). From a system level perspective, the literature [181] waveform design, but it is ineffective in dealing with time selectivity
presents an MMSE SIC-based receiver and designs a realizable code- in high-mobility channels; the other method is the decomposition of
word level SIC receiver for evaluating NOMA–OTFS’s performance. It the received signal into delay and Doppler shifted. Hence, it is urgent
also improves the spectral efficiency from link level perspective. In to explore a new multiplexing scheme to respond to the high-mobility
a 3D structured OTFS system, Shao et al. [182] investigate a DDA environments. The OTFS modulation method is investigated, which is
domain NOMA scheme which schedules multiusers with overlapped different from the second method proposed in the OFDM system. The
angle signatures to employ the same DD domain resources. OTFS is designed in the DD domain rather than the TF domain.
In addition to the above multiple access technologies, there are In literature [192], a comparative performance study of OTFS with
others novel MA methods proposed in recent research. Li et al. [74] multicarrier code division multiple access (MC-CDMA) in doubly dis-
focus on a new path division multiple access (PDMA) method under persive channels is presented from the precoder perspective, and it
the similar 3D network. Random access (RA) technology [183] is also proves that OTFS is able to apply for high-speed V2X communication.
combined with OTFS, and in literature [183] proposes a novel RA From the 5G application perspective, literature [193] proposes a less
preamble waveform to apply in the OTFS system. Moreover, joint complex NR-based IoT (NR-IoT) system with OTFS that performs better
tandem spreading multiple access (TSMA) and OTFS technology is than NR-IoT with LDPC. Ramachandran et al. [194] illustrate the slow
proposed for high-speed railway [184]. variability and sparse nature of the delay–Doppler channel under an
Furthermore, the combination of emerging spectrally efficient 5G urban multi-lane scenario. Additionally, Wiffen et al. [195] compare
solutions, such as 5G New Radio Bandwidth Part (5G-NR-BWP) and the performances of OTFS with OFDM for line-of-sight mobility auto-
software-controlled metasurfaces, with OTFS–NOMA is also a promis- motive channel and simulate the channel using the Kirchhoff model. In
ing topic. In the foreseeable future, the emerging spectrum-efficient a railway communication scenario, Lin et al. [196] develop a channel
5G solutions, combined with OTFS–NOMA, are the subject of future and phase shift estimator that initially adopts the linear minimum mean
research. squared error to obtain the end-to-end channel and iteratively optimize

17
M. Li et al. Computer Networks 224 (2023) 109597

the phase shift using the majorization-minimization (MM) method, and Table 4
the simulation results show that better performance from the perspec- Comparison results of SC, OFDM and OTFS in mmWave Communication.

tive of convergence. The future work will include designing the training Property SC OFDM OTFS

signal, equalizer, and detector specific to the V-IoT applications. PAPR 𝐿𝑜𝑤 𝐻𝑖𝑔ℎ 𝐻𝑖𝑔ℎ
Latency 𝐿𝑜𝑤 𝐻𝑖𝑔ℎ 𝐻𝑖𝑔ℎ
Implementation Degree 𝐿𝑜𝑤 𝐻𝑖𝑔ℎ 𝐻𝑖𝑔ℎ
5.4.3. Joint radar and communication Spectral Efficiency 𝐻𝑖𝑔ℎ 𝐿𝑜𝑤 𝐻𝑖𝑔ℎ
As the emerging application such as joint radar and communication Overhead 𝐿𝑜𝑤 𝐻𝑖𝑔ℎ 𝐻𝑖𝑔ℎ
system (RadCom) [197] has been aroused the interest of many scholars BER depend on SNR 𝐺𝑜𝑜𝑑 𝐺𝑜𝑜𝑑
at home and abroad. The RadCom system not only can provide com- Robustness (synchronization) 𝐿𝑜𝑤 𝐿𝑜𝑤 𝐿𝑜𝑤
Compatibility (MIMO) 𝐻𝑖𝑔ℎ 𝐻𝑖𝑔ℎ 𝐻𝑖𝑔ℎ
munication links with other vehicles but also enables to offer active
perimeter awareness. Of course, this system is also applied to other
fields, like the military field. However, there are limitations in the
RadCom system due to the high requirements. Under the high mobility In literature [200], Gaudio et al. propose an efficient ML-based ap-
and dense traffic environments, it is indispensable for providing the proach to detect targets and estimate the corresponding delay, Doppler,
high data rate, ultra reliability and ultra low latency communication. and angle of arrival parameters. Compared with literature [198], the
In this case, exploiting an appropriate waveform to satisfy the demands proposed algorithm analyzes the angle of arrival parameters and esti-
for both radar and communication is difficult. With the development of mates the super-resolution of radar parameters. From the beamforming
OTFS technology, the combination of OTFS and the RadCom system is perspective, literature [201] design hybrid digital-analog beamform-
a critical issue in future research. ing at radar transmitter adapted to the different operating phases. Li
As a novel waveform, OTFS can be applied to parameter estimation et al. [202] present a novel integrated sensing and communication
in radar. First, according to the description of the first part, the vector (ISAC) transmission framework based on spatially spread OTFS (SS-
input–output relation is obtained as OTFS), the key novelty is the angular domain discretization enabled
𝑃∑
−1 by the spatial spreading. Yuan et al. [203] develop a novel downlink
𝒚= ℎ′𝑝 𝜳 𝑝 𝒙 + 𝒘 = 𝐇𝐱 + 𝐰 (61) transmission scheme for vehicular networks based on ISAC signals and
𝑝=0 OTFS modulation. The above literature [203] is only for downlink
where 𝜳 𝑝 is the 𝑁𝑀 × 𝑁𝑀 matrix, 𝒘 denotes the AWGN with zero parameters estimation, the roadside unit (RSU) can infer the delays
mean and identity covariance. Based on above equation, three classes and Dopplers associated with different channel paths based on the
of parameter estimators are considered: Maximum Likelihood (ML) dynamic topology of the vehicular network for uplink transmission
Estimator, Matched Filter (MF) Estimator and Generalized Likelihood in literature [204]. Liu et al. [205] investigate an efficient Bayesian
Ratio Test (GLRT) Estimator. learning scheme that leverages the structural sparsity of radar channels
(1) ML estimator. The log-likelihood function [198] to be min- in the DD domain, the provided method realizes the low-complexity
imized is given as follow parameter estimation for radar targets. Zhang et al. [206] study a
( when) ML estimator is used to estimate
unknown parameters 𝜽 = ℎ′ , 𝜏, 𝜈 and 𝑃 = 1. high-peak-to sidelobe ratio matrix-inversion based method for symbol
cancellation and also provide an approximate spectrum-division based
𝑙(𝒚 ∣ 𝜽, 𝒙) = ||𝒚 − ℎ′ 𝜳 (𝜏, 𝜈)|| (62) method to achieve low complexity.
where symbols in 𝒙 are known at the radar receiver. When parameter To sum up, the estimation algorithm of parameters and the complex-
(𝜏, 𝜈) is known, the channel gain ℎ′ is shown as ity analysis are the main hotpots of the current research on the joint
communication and radar. For a massive MIMO system, beamforming
𝒙𝐻 𝜳 𝐻 𝒚 based on OTFS is needed to be studied in the spatial domain, such as
ℎ̂ ′ = (63)
𝒙𝐻 𝜳 𝐻 𝜳 𝒙 hybrid beamforming.
where 𝐻 indicates the transpose complex conjugate. Based on the
expression of formula (57), the value of (𝜏, ̂ is shown as follow.
̂ 𝜈) 5.4.4. mmWave communication
|𝒙𝐻 𝜳 (𝜏, 𝜈)𝐻 𝒚 |2 With the massive popularity of 5G construction, 5G has been di-
(𝜏, ̂ = arg max
̂ 𝜈) | | (64) vided into two scenarios according to different spectrums: sub-6 GHz
(𝜏,𝜈)∈𝛤 𝒙𝐻 𝜳 (𝜏, 𝜈)𝐻 𝜳 (𝜏, 𝜈)𝒙 and mmWave. The requirement for high data rates and the availabil-
(2) MF estimator. Similar, a matched filter (MF) based ity of a large spectrum in mmWave frequency bands have prompted
method [19] is proposed to estimate ℎ(𝜏, 𝜈), which is equivalent to the design of high throughput wireless systems. However, it is sus-
detect range and velocity of targets. In order to obtain the estimate ceptible to mmWave communication when the high-carrier frequency
of ℎ, the following MF detection is expressed as follow. is utilized regardless of low-mobility or high-mobility scenarios. Yan
̃H 𝐲 = 𝐗
̂𝐡 = 𝐗 ̃ H 𝐗𝐡 ̃ H 𝐰 = 𝐆𝐡 + 𝐰
̃ +𝐗 et al. [207] develop a low-complexity OTFS-based hybrid precoding
̃ (65)
algorithm with the rectangular waveform for mmWave A2G communi-
where the gain matrix 𝐆 = 𝐗 ̃ H 𝐗̃ ∈ C𝑀𝑁×𝑀𝑁 , and 𝐰 ̃ H 𝐰.
̃=𝐗 cation. The literature [208] investigates the performance of OTFS over
(3) GLRT estimator. In literature [199], Keskin et al. design a the vehicular channels, which are characterized by different channel
low-complexity detection and estimation scheme, the GLRT can be characteristics under 60 GHz mmWave channels. Ji et al. [209] propose
denoted as a multi-dimensional sparse vector compression (M-SVC) method for the
( ) mmWave spectrum, and the M-SVC method shows robustness against
maxℎ′ ,𝜏,𝜈 𝑝 𝐲 ∣ 1 ; ℎ′ , 𝜏, 𝜈 1
𝛬(𝐲) = ( ) ≷ 𝜂 (66) phase noise and the Doppler effect. Similarly, Surabhi et al. [210]
𝑝 𝐲 ∣ 0 0
investigate the effect of phase noise on the performances of OTFS in the
where 𝜂 is the threshold. mmWave communication. In literature [211] provides the performance
What is more, CRLB provides a standard for comparing the per- evaluation of OTFS, which has lower BER than OFDM.
formance of unbiased estimators, where the Fisher information matrix Based on the above research status at home and abroad [212], the
defined as comparison results of SC, OFDM and OTFS in mmWave communication
{ [ ] [ ]}
∑ 𝜕𝑠𝑛,𝑚 ∗ 𝜕𝑠𝑛,𝑚 are shown in Table 4. As can be seen from Table 4, the multicarrier
𝐈(𝜽, 𝑨)𝑡𝑖,𝑗 = 2𝑃avg ℜ (67) technology is better than the single-carrier technology on the whole.
𝑛,𝑚
𝜕𝜃𝑖 𝜕𝜃𝑗
When multicarrier technology adopts OTFS, the spectral efficiency is
The CRLB expressions follow by applying the Fisher information matrix. higher than OFDM technology. As a result, OTFS is a novel multicarrier

18
M. Li et al. Computer Networks 224 (2023) 109597

Table 5
An overview of application about OTFS technology.
Application scenarios Reference Channel environment Performance description Compatibility with existing system
[190] doubly-selective channel lower BER with high-speed LTE-based 5G csystem
Terrestrial Broadcast [191] realistic vehicular channel good overall performance 60GHZ V2V system
[189] doubly-selective channel robust reception performance ATSC 3.0 system
[194] doubly-selective channel detection and parameter estimation 4 GHz and 28 GHz frequency bands
[195] doubly-selective channel Line-of-Sight (LOS) model sub-6 GHz and mmWave system
[198] vehicular channel obtain detection CRLB mono-static radar
V2X Communication
[192] doubly-selective channel precoder based on WH and ZC MC-CDMA system
[193] low latency channel achieve higher SNR gain NR-IoT system
[196] doubly-selective channel better convergence performance railway communication system
[19] integral Doppler channel MF-based parameter estimation OFDM-based radar
[200] fractional Doppler channel ML-based parameter estimation OFDM-based radar
[201] space sparse channel hybrid digital-analog beamforming mono-static MIMO radar
[202] space sparse channel beam tracking and angle estimation SS-OTFS framework
Joint Radar and Communication [203] doubly-selective channel predict downlink channel parameter ISAC-assisted OTFS system
[204] doubly-selective channel downlink and uplink parameter ISAC-assisted OTFS system
[205] doubly-selective channel low-complexity parameter estimation Automotive radar
[206] doubly-selective channel modulation symbol cancellation mono-static radar
[199] doubly-selective channel GLRT-based parameter estimation OFDM and OTFS radar
[210] mmWave channel more robust to phase noise 28 GHz OFDM-based system
[211] mmWave channel lower BER than OFDM 5G mmWave frequencies
[212] mmWave channel overview 5G mmWave technology 5G IoT system
mmWave Communication
[207] mmWave channel design hybrid precoding MIMO–OTFS system
[208] mmWave channel measure vehicular channel 60 GHz mmWave system
[209] mmWave channel propose M-SVC scheme mmWave and terahertz spectrum
[213] doubly-selective channel propose descent data detection DFT-S-OTFS system
Terahertz Communication
[214] doubly-selective channel channel estimation data detection DFT-S-OTFS system
[215] two-ray channel comprehensive performance in BER LTE-based 5G system
Airborne Communication [216] doubly-selective channel rescue of positioning operation B5G system
[217] doubly-selective channel OTFS-modulated PRACH scheme OFDM system
[218] doubly-selective channel two-stage successive ATI scheme MU–MIMO system
[219] doubly-selective channel NOMA-OTFS for LEO satellite LEO satellite mMTC system
LEO Satellite Communication
[220] doubly-selective channel derive outage probability 5G OFDM system
[221] doubly-selective channel develop TDSBLFM for estimation MIMO–OTFS system
Underwater Communication [222] UWA channel design PTR structure receiver OTFS-UWA communication system
[223] doubly-selective channel trade-off in rate and overhead cell-free massive MIMO system
Cell-Free Network
[224] doubly-selective channel EP-CHE and SP-CHE schemes cell-free massive MIMO system
Throughput [225] doubly-selective channel improve communication capacity MIMO–OTFS system
Spectral Efficiency [226] AWGN or multipath reduce OOB power or ISI SE-OTFS system
[227] doubly-selective channel synchronization and CID estimation OTFS-based cellular system
Synchronization
[103] doubly-selective channel synchronization error OTFS system
[228] doubly-selective channel a modified LLR detector OTFS-based system
Index Modulation
[229] doubly-selective channel lower complexity detector OTFS-IM system
[230] doubly-selective channel physical layer abstraction model DFT-spread-OFDM system
Others Aspects
[231] doubly-selective channel better BER and Lower PAPR DFrFT-OFDMc system

modulation method to settle the mmWave communication compared 6.1. Artificial intelligence in OTFS
with the OFDM technology, which is vulnerable to Doppler shift and
phase noise effects. Machine learning (ML) [232] and artificial intelligence (AI) [233]
have achieved great success in a variety of fields, especially in im-
5.4.5. Other aspects age recognition, natural language processing, etc. In recent year, the
The contents of part 5.4.1 to part 5.4.4 analyze the research status of application of ML and AI to the communication field have attracted
terrestrial broadcast communication, V2X communication, joint radar curious concerns, such as burst signal identification, modulation recog-
and communication and mmWave communication from the perspective nition [234], channel modeling, network security and edge AI.
of technological innovation or improvement. Apart from the above re- As a kind of artificial intelligence method, deep learning (DL) [235]
search, OTFS also has many application combinations in other aspects, can be widely used to promote the development of communication in-
such as UAV communication, underwater acoustic communication, telligence further. In literature [236,237], a data-driven OTFS receiver
synchronization, etc. The specific description and details are shown in based on Deep Neural Network (DNN) is presented, and the proposed
Table 5. algorithm can effectively counter channel distortion, which achieves
good robustness for the variation of channel parameters. Based on the
6. Open issues and challenges original technology of the wireless communication system, the model
structure is not changed, and the deep learning network is used to
OTFS is an emerging research technology instead of OFDM. The pre- replace a particular module or a specific training-related parameter to
vious Sections 4 and 5 mainly discusses the critical technologies, such improve the performance of a certain module, as shown in Fig. 22.
as pulse-shaping waveform design, signal detection, precoding design, It can be seen from Fig. 22 that the red part can be regarded as a
channel estimation, equalization and application for many fields. In this black box, and the deep learning model is directly used to perform
section, the potential of OTFS technology will be explored. its function, which is more advantageous than traditional methods to

19
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 22. The OTFS system based on deep learning. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

some extent. In addition, modulation recognition [238] and channel such as the comparison of LDPC code and polar code [242]. At present,
decoding [239] based on DL have a wide range of applications in there is no literature on the theoretical analysis of the OTFS system
signal monitoring, signal interception, interference recognition and based on polar code. Moreover, how to choose the coding parameters
electronic countermeasures in the civil and military fields to improve for near-capacity joint detection and decoding remains an interesting
the transmission performance of physical layer. However, it should be open issue. Furthermore, it is also possible to design an adaptive
noted that the intelligent identification of modulation methods is based rate coding scheme based on OTFS and OFDM in a mixed Doppler
on the OFDM system architecture. It is a considerable problem to be environment.
compatible with the above proposed identification method in the OTFS It is worth noting that the diversity gain and the coding gain are
system, although OTFS can be regarded as a general form of OFDM. incompatible. The diversity gain of the OTFS system improves with the
What is more, whether the eigenvalues extracted in the OTFS system number of independent resolvable channel paths, while the coding gain
will change compared to the OFDM system, and how to extract the declines. Hence, it is a considerable problem to balance the trade-off
eigenvalues in the OTFS system, these issues will be the key to solving between the diversity gain and coding gain in the future.
the combination of OTFS and AI.
With the explosion of smart mobile devices and IoT devices, most 6.3. Security and privacy issues in OTFS system
intelligent applications will be deployed at the edge of the wireless
network. The conventional ML or AI algorithms need to handle the
Security and privacy issues [243] have been a critical problem from
vast data in service so that its burden is so large, but the edge comput-
ancient times to the present. With the advent of the Big Data era, per-
ing [240] can solve this problem. In the face of high-mobility scenarios,
sonal privacy and information security suffer enormous challenges. The
OTFS technology as a tool can be combined with edge computing,
corresponding solution is to investigate encryption technology or covert
which provides the possibility for the combination of ML or AI and
communication in the physical layer. Currently, there is no research on
OTFS. In addition, resource allocation is the key issue to allowing
the safety aspects of the OTFS system. Therefore, it is possible to borrow
multi-users access, and the ML is able to become the criteria to define
encryption techniques from the OFDM system and then combine the
‘‘more or less important’’ data. It is promising to view the prospect of
proposed encryption techniques with the OTFS system.
combining OTFS with ML or AI.
From the perspective of the physical layer, the wireless channel
Intelligence is the core content of the future information society,
has the characteristics of time-varying, randomness and reciprocity,
and it is indispensable to combine with machine learning and artificial
which makes it possible to use these characteristics to solve the com-
intelligence technologies. However, it should be noted that OTFS and
munication security problem [244]. For example, the OFDM system
OFDM can share a similar framework, there are still differences when
migrating the OFDM method to the OTFS system, it is mainly a problem can provide extra randomness because of the use of multiple sub-
that the intelligent OTFS framework needs to solve in high-speed channels compared to a single carrier system, and how to make full
situations. use of the subchannels to increase the available keys is still an open
issue [245]. However, as OTFS is a more representative multicarrier
6.2. Coded OTFS system performance analysis system, whether it can make full use of the delay–Doppler channel to
extract better key features is also a topic worthy of discussion. Simi-
Effective channel coding and decoding techniques play an essential larly, whether it is possible to borrow the subcarrier power allocation
role in reliable channel transmission, which is exceptionally indispens- scheme of the OFDM system to generate the key in the OTFS system.
able in high-mobility data transmission scenarios. The channel codes Different from traditional encryption methods, the purpose of covert
have to be carefully designed for OTFS modulation due to channel communication is to realize information transmission without causing
impairments. From the OTFS channels analyzed above, it is known that third-party perception, it is mainly to investigate the content security
a few groups of reflectors have different delay and Doppler shift values of information and the behavioral security of communication. Kumar
where the channel taps are related to the number of reflectors, so the et al. [246] introduce the covert communication integrated into the
channel is sparsity in the DD domain. According to this characteristic, wavelet packet transform OFDM (WPT–OFDM) system and evaluate the
Low Density Parity Check (LDPC) code [241] is suitable for the OTFS BER and quality assessment of convert communication under OFDM.
system to improve the reliability, because LDPC code is a linear block OTFS is a general form of OFDM, it is expected to put forward an
code of sparse check matrix. effective covert communication method to solve the security problem
Due to its good performance close to the Shannon limit and low of the OTFS system in future work.
decoding complexity, LDPC code is able to combine with SCMA to The development of wireless communication means that informa-
realize joint detection under the OTFS system. In addition, considering tion interaction eliminates the shackles of location, time and objects
the performance bounds that can be achieved in an OTFS system by and accelerates the process of social informatization. However, the
comparing different coding schemes in terms of reliability and BER, privacy and security issues of wireless communication have not been

20
M. Li et al. Computer Networks 224 (2023) 109597

Fig. 23. The brief classification and description about open issues.

Fig. 24. The brief classification and description about challenges.

21
M. Li et al. Computer Networks 224 (2023) 109597

guaranteed, especially the data transmission under high dynamic con- 6.6. Practical application of OTFS system
ditions in the future. The solution can not only be considered from the
traditional network layer cryptography but also can use the information
At present, the majority of current literature only analyses the OTFS
theory related knowledge of the physical layer. The process of massive
system’s performance from the theoretical point of view. For example,
data information transmission requires a safe and secure environment.
pulse-shaping waveform design, channel estimation methods, signal
6.4. New physical dimensions in OTFS system detection and equalization algorithm, etc. The amount of references
on the implementation of OTFS from the hardware perspective is
With the explosive growth of multimedia services and the vigorous negligible. In future work, the system-level simulation should be paid
development of mobile Internet applications, the contradiction between more attention, and then the practical engineering applications or
capacity demand and the shortage of spectrum resources has become prototypes of OTFS technology are realized by using the embedded
increasingly prominent, especially in wireless communications, where chips or FPGA. Therefore, OTFS technology has the opportunity to
the shortage of spectrum resources has become an important bottleneck replace OFDM technology in engineering.
restricting information transmission. The OTFS system mainly provides
In addition, the hardware object studied in the limited literature is
a solution for the Doppler shift caused by high speed, but it does not
the implementation of OTFS based on the OFDM framework, and its
solve the lack of spectrum resources or spectrum utilization problems.
Therefore, settling the trouble of spectrum efficiency or increasing the specific manifestation is the addition of a pre-processing module at the
transmission capacity from a new perspective is indispensable. transmitting end and a post-processing module at the receiving end.
In the following research content, it is one of the essential tasks to
6.4.1. Orbital angular momentum focus on the development of an OTFS architecture suitable for future
Orbital angular momentum (OAM) is a natural attribute of electro- communications. It is able to integrate and encapsulate this architecture
magnetic waves, which is similar to linear momenta such as amplitude, to achieve commercialization.
phase, and frequency, and is also a new dimension of information From the modulator’s point of view, Thaj et al. in Ref. [250] discuss
transmission in the communication field [247]. Compared with tradi-
the influence of receiver impairment in OTFS modem by using the same
tional electromagnetic waves, vortex electromagnetic waves (carrying
hardware setup and environment for natural indoor doubly selective
OAM) have more freedom in information modulation. The infinity
and orthogonality between different modes can significantly improve channels. Farhang et al. in literature [251] study the channel impact
channel capacity and spectrum utilization. in discrete time and derive the simplified modulator and modulator
structures. The above work is beneficial for addressing the challenges
6.4.2. Intelligent reflecting surface of OTFS practical application in a real environment.
The intelligent reflecting surface (IRS) is a new programmable The theoretical research of OTFS is to solve the Doppler frequency
metamaterial with the characteristics of low power consumption, low shift phenomenon, and the practical application of the OTFS system
complexity and low cost [248]. It mainly controls many passive electro- is to make life more convenient or accelerate development. In the fu-
magnetic reflective elements on the plane through digital coding and ture, the existing theoretical basis and practical applications should be
other means. The intelligent reflective surface provides a new degree of
combined to design a communication framework for future high-speed
space for the system by intelligently controlling the reflected phase and
mobility scenarios.
amplitude of the incident signal and achieves the purpose of enhancing
the performance of the communication link. Furthermore, IRS technol-
ogy can actively control the wireless environment to reduce Doppler 6.7. Visible light communication for OTFS system
frequency shift during data transmission in high-speed dynamic scenes.
Combined with the Doppler frequency shift characteristics of the OTFS
system, it can achieve double resistance to Doppler. Visible light communication (VLC) is rapidly emerging as a com-
As current new technologies in the communications field, the QAM pelling technology for supplementing traditional radio frequency com-
technology and IRS technology are hopeful of becoming part of the munication and enabling new wireless device use cases that are
future communications framework. OTFS may be used as one of the uniquely achievable with this technology [252,253]. The essential
future communication architectures, and more consideration should to visible light communication design is to achieve faster data rates
be given to collisions with hotspot technologies to produce a new without compromising illumination. In literature [254], a direct current
combination mode. biased optical orthogonal time–frequency space (DCOOTFS) based full-
duplex relay-assisted visible light communication system is proposed
6.5. TDD or FDD OTFS system
to enhance the spectral efficiency of visible light communications,
and the proposed method can achieve higher spectrum efficiency
The two standard modes of duplexing are considered in this subsec-
tion. Time division duplexing (TDD) mode has been the most widely and robustness. In an indoor VLC setting, Sinha et al. [255] present
used duplexing technique due to its simplicity. Due to the reciprocity of an NDC-OTFS scheme that uses a 2D Hermitian symmetry operation
channels in TDD systems, the channel estimation process is much more to convert complex signals to positive, real-valued signals suitable
straightforward than in frequency division duplexing (FDD) systems. for transmission in the optical domain. Based on standardized IEEE
Hence, TDD mode is the first choice for massive MIMO systems. In 802.15.7 VLC channel, literature [256] investigates the performance
the massive MIMO system based on FDD, acquiring downlink CSI may of OTFS over static multipath channels. For multi-LED indoor visible
require a vast training overhead, which reduces the spectrum efficiency light wireless communications, Sinha et al. [257] investigate two
of the network. In the future OTFS system, the TDD mode in the OFDM quad-LED OTFS schemes named QCM–OTFS and SM–DCM–OTFS. The
system can be used as the reference for massive MIMO–OTFS, and the proposed schemes achieve significantly better BER performance. From
research is conducted on adopting FDD for massive MIMO by reducing
the perspective of security, a 2D encryption based on the Arnold trans-
training and feedback overhead.
formation is proposed for implementing a secure DC-biased optical or-
FDD mode mainly separates the receive and transmit channels by
thogonal time–frequency multiplexing (DCO–OTFM) in optical-wireless
using the protected band. In the OTFS system, FDD is the tool to
distinguish between uplink and downlink. For TDD mode, which can communications (OWCs) [258].
separate channels by times, it works on inferring the downlink channel At present, there are relatively few studies on the application of
from the uplink channel [249]. Moreover, it is also essential to conduct OTFS modulation technology to visible light communication scenarios.
a systematic comparative study of TDD or FDD-aided OTFS systems. Therefore, its related research is one of the future research hotspots.

22
M. Li et al. Computer Networks 224 (2023) 109597

6.8. Other challenges [6] L. Bai, R. Han, J.W. Liu, Q. Yu, J. Choi, W. Zhang, Air-to-ground wireless links
for high-speed UAVs, IEEE J. Sel. Areas Commun. 38 (12) (2020) 2918–2930.
In addition to the issues discussed in part 6.1 ∼ 6.7 above, there [7] K. Anand, Y.L. Guan, X.B. Liu, Z.L. Liu, Y. Yang, Z.C. Zhou, P.Z. Fan, E.
Gunawan, Pilot design for BEM-based channel estimation in doubly selective
are still many research hotspots that are worth pondering or not
channel, IEEE Trans. Veh. Technol. 69 (2) (2020) 1679–1694.
being involved, as shown in Fig. 23, such as fundamental information [8] S.Z. Chen, S.H. Sun, G.X. Xu, X. Su, Y.M. Cai, Beam-space multiplexing: Practice,
theoretical limits, uRLLC [259] and so on. The information theory theory, and trends, from 4G TD-LTE, 5G, to 6G and beyond, IEEE Wirel.
as a fundamental discipline of theoretical knowledge has been tried Commun. 27 (2) (2020) 162–172.
and tested in various areas of social sciences. The application of this [9] X.Q. Liu, H.H. Chen, W.X. Meng, B.Y. Lyu, Successive multipath interference
subject to the mathematical derivation of OTFS is no exception. As a cancellation for CP-Free OFDM systems, IEEE Syst. J. 13 (2) (2019) 1125–1134.
[10] L. Ma, H.B. Jia, S.Z. Liu, I.U. Khan, Low-complexity Doppler compensation
result, the capacity scaling [260] of the OTFS system for the multiple
algorithm for underwater acoustic OFDM systems with nonuniform Doppler
antennas and multi-users system can be analyzed. The fundamental shifts, IEEE Commun. Lett. 24 (9) (2020) 2051–2054.
information theoretical limitations are also an alternative perspective. [11] R. Hadani, A. Monk, OTFS: A new generation of modulation addressing the
In addition, the transmission protocols of OTFS, coexisting of 5G and chanllenges of 5G, 2018, arXiv preprint arXiv:1802.02623.
OTFS signaling, filter design and the application of the OTFS system [12] A. Monk, R. Hadani, M. Tsatsanis, S. Rakib, OTFS-orthogonal time frequency
in Tera Hertz [261–263] are all the significant hot issues to further space, 2016, arXiv preprint arXiv:1608.02993.
[13] Z.Q. Wei, W.J. Yuan, S.Y. Li, J.H. Yuan, G. Bharatula, Orthogonal time–
investigate in the future work.
frequency space modulation: A promising next-generation waveform, IEEE
Finally, this paper gives a brief description of the topics that need Wirel. Commun. 28 (4) (2021) 136–144.
to be solved and the challenges that need to be faced in the future as [14] J.X. Lu, X.H. Chen, S.Y. Liu, P.Y. Fan, Location-aware ICI reduction in MIMO–
shown in Fig. 24. Not only pay attention to current common problems, OFDM downlinks for high-speed railway communication systems, IEEE Trans.
but also focus on unresearched areas to give unique insights. Veh. Technol. 67 (4) (2018) 2958–2972.
[15] Z.Q. Wei, W.J. Yuan, S.Y. Li, Orthogonal time–frequency space modulation:
7. Conclusion A promising next-generation waveform, IEEE Wirel. Commun. 28 (4) (2021)
136–144.
[16] S.Y. Li, W.J. Yuan, Z.Q. Wei, A tutorial to orthogonal time frequency space
High-speed mobility scenarios are extremely challenging in wire- modulation for future wireless communications, in: 2021 IEEE/CIC International
less communication, in which the most representative product is the Conference on Communications in China, ICCC Workshops, 2021, pp. 439–443.
doubly selective channel. This paper provides a more comprehensive [17] Z.Q. Zhang, H. Liu, Q. Wang, A survey on low complexity detectors for OTFS
summary and analysis of OTFS techniques compared with the existing systems, ZTE Commun. 19 (4) (2021) 03–15.
review literature [15–17], and also illustrates the advantages of OTFS [18] P. Raviteja, K.T. Phan, Y. Hong, E. Viterbo, Interference cancellation and
iterative detection for orthogonal time frequency space modulation, IEEE Trans.
modulation from the level of multicarrier modulation comparison. We
Wirel. Commun. 17 (10) (2018) 6501–6515.
commenced with a system model of OTFS, including input–output [19] P. Raviteja, K.T. Phan, Y. Hong, E. Viterbo, Orthogonal time frequency space
relationship and interference analysis. The related technologies of OTFS (OTFS) modulation based radar system, in: 2019 IEEE Radar Conference,
modulation, such as channel coding, channel estimation, signal detec- RadarConf, 2019, pp. 1–6.
tion, etc. are discussed and highlighted. Furthermore, the potential [20] W.Q. Shen, L.L. Dai, J.P. An, P.Z. Fan, R.W. Heath, Channel estimation for
applications and the research direction are proposed in this paper. orthogonal time frequency space (OTFS) massive MIMO, IEEE Trans. Signal
Process. 67 (16) (2019) 4204–4217.
Finally, the seamless transition from the OFDM system to the OTFS
[21] Y. Cao, T. Jiang, Z. Han, A survey of emerging M2M systems: Context, task,
system is the necessary demand for future air interface design, it is and objective, IEEE Internet of Things J. 3 (6) (2016) 1246–1258.
expected that this paper will be helpful for follow-up research on OTFS [22] J. Thota, N.F. Abdullah, A. Doufexi, S. Armour, V2V for vehicular safety
modulation technology. applications, IEEE Trans. Intell. Transp. Syst. 21 (6) (2020) 2571–2585.
[23] M. Sharique, A.K. Chaturvedi, Transmitter pulse shaping to reduce OOB power
Declaration of competing interest and ICI in OFDM system, Wirel. Pers. Commun. 83 (2) (2015) 1567–1578.
[24] V. Rangamgari, S. Tiwari, S.S. Das, S.C. Mondal, OTFS: Interleaved OFDM with
Block CP, in: 2020 National Conference on Communications, NCC, 2020, pp.
The authors declare that they have no known competing finan-
1–6.
cial interests or personal relationships that could have appeared to [25] L. Gaudio, G. Colavolpe, G. Gaire, OTFS vs. OFDM in the presence of sparsity:
influence the work reported in this paper. A fair comparison, IEEE Trans. Wirel. Commun. 21 (6) (2021) 4410–4423.
[26] R.X. Chong, S.Y. Li, J.H. Yuan, Achievable rate upper-bounds of uplink
Data availability multiuser OTFS transmissions, IEEE Wirel. Commun. Lett. 11 (2022) 4.
[27] M. Joshi, G. Punjabi, B. Sainath, Comparative performance investigation of
No data was used for the research described in the article. MIMO–OTFS and MIMO–OFDM using deep neural network modeling, in: 2021
IEEE 18th India Council International Conference, INDICON, 2021, pp. 1–6.
[28] A. Farhang, N. Marchetti, L.E. Doyle, Low-complexity modem design for GFDM,
Acknowledgment IEEE Trans. Signal Process. 64 (6) (2016) 1507–1518.
[29] N. Michailow, M. Matthé, I.S. Gaspar, A.N. Caldevilla, L.L. Mendes, A. Festag, G.
All the authors listed have approved the manuscript that is enclosed. Fettweis, Generalized frequency division multiplexing for 5th generation cellular
networks, IEEE Trans. Commun. 62 (9) (2014) 3045–3061.
References [30] A. Nimr, M. Chafii, M. Matthe, G. Fettweis, Extended GFDM framework: OTFS
and GFDM comparison, in: 2018 IEEE Global Communications Conference,
[1] Y.H. Huang, X. Xu, N. Li, H. Ding, X. Tang, Prospect of 5G intelligent networks, GLOBECOM, 2018, pp. 1–6.
IEEE Wirel. Commun. 27 (4) (2020) 4–5. [31] T. Erseghe, N. Laurenti, V. Cellini, A multicarrier architecture based upon the
[2] S.P. Dang, O. Amin, B. Shihada, M.S. Alouini, What should 6G be? Nature affine Fourier transform, IEEE Trans. Commun. 53 (5) (2005) 853–862.
Electron. 3 (1) (2020) 20–29. [32] A. Bemani, N. Ksairi, M. Kountouris, AFDM: A full diversity next generation
[3] L. Wan, Z.H. Guo, Y. Wu, W.P. Bi, J.H. Yuan, M. Elkashlan, L. Hanzo, 4G/5G waveform for high mobility communications, in: 2021 IEEE International
spectrum sharing: Efficient 5G deployment to serve enhanced mobile broadband Conference on Communications Workshops, ICC Workshops, 2021, pp. 1–6.
and Internet of Things applications, IEEE Veh. Technol. Mag. 13 (4) (2018) [33] A. Bemani, G. Cuozzo, N. Ksairi, Affine frequency division multiplexing for
28–39. next-generation wireless networks, in: 2021 17th International Symposium on
[4] L. You, K.X. Li, J.H. Wang, X.Q. Gao, X.G. Xia, B. Ottersten, Massive MIMO Wireless Communication Systems, ISWCS, 2021, pp. 1–6.
transmission for LEO satellite communications, IEEE J. Sel. Areas Commun. 38 [34] P. Raviteja, E. Viterbo, Y. Hong, OTFS performance on static multipath
(8) (2020) 1851–1865. channels, IEEE Wirel. Commun. Lett. 8 (3) (2019) 745–748.
[5] G. Noh, B. Hui, I. Kim, High speed train communications in 5G: Design elements [35] T. Thaj, E. Viterbo, Y. Hong, Orthogonal time sequency multiplexing modula-
to mitigate the impact of very high mobility, IEEE Wirel. Commun. 27 (6) tion: Analysis and low-complexity receiver design, IEEE Trans. Wirel. Commun.
(2020) 98–106. 20 (12) (2021) 7842–7855.

23
M. Li et al. Computer Networks 224 (2023) 109597

[36] S.M. Pishvaei, B.M. Tazehkand, J. Pourrostam, Design and performance eval- [64] S. Tiwari, V. Rangamgari, Estimation and equalization of residual synchroniza-
uation of FBMC-based orthogonal time–frequency space (OTFS) system, Phys. tion errors in CP-OTFS, 2020.
Commun. 53 (2022) 101723. [65] N. Hashimoto, N. Osawa, K. Yamazaki, Channel estimation and equalization
[37] A. Tusha, S. Dogan-Tusha, F. Yilmaz, Performance analysis of OTFS under in- for CP–OFDM-based OTFS in fractional Doppler channels, in: 2021 IEEE
phase and quadrature imbalance at transmitter, IEEE Trans. Veh. Technol. 70 International Conference on Communications Workshops, ICC Workshops, 2021,
(11) (2021) 11761–11771. pp. 1–7.
[38] P. Busch, T. Heinonen, P. Lahti, Heisenberg’s uncertain principle, Phys. Rep. [66] H.B. Mishra, P. Singh, A.K. Prasad, OTFS channel estimation and data detection
452 (6) (2007) 155–176. design with superimposed pilots, IEEE Trans. Wirel. Commun. 21 (4) (2021)
[39] A. Hisham, D. Yuan, E.G. Ström, F. Brännström, Adjacent channel interference 2258–2274.
aware joint scheduling and power control for V2V broadcast communication, [67] P. Raviteja, K.T. Phan, Y. Hong, Embedded pilot-aided channel estimation for
IEEE Trans. Intell. Transp. Syst. 22 (1) (2021) 443–456. OTFS in delay–Doppler channels, IEEE Trans. Veh. Technol. 68 (5) (2019)
[40] M. Ailasgari, A. Abbasfar, S.M. Fakhraie, Out-of-band power reduction by 4906–4917.
using computationally efficient cancellation pulses, Phys. Commun. 13 (2014) [68] P. Raviteja, K.T. Phan, Y. Hong, E. Viterbo, Embedded delay–Doppler channel
142–155. estimation for orthogonal time frequency space modulation, in: 2018 IEEE 88th
[41] P. Raviteja, Y. Hong, E. Viterbo, E. Biglieri, Practical pulse-shaping waveforms Vehicular Technology Conference, VTC-Fall, 2018, pp. 1–5.
for reduced-cyclic-prefix OTFS, IEEE Trans. Veh. Technol. 68 (1) (2019) [69] A. Fish, A. Sayeed, S. Gurevich, Delay–Doppler channel estimation with almost
957–961. linear complexity, IEEE Trans. Inform. Theory 59 (11) (2013) 7632–7644.
[42] J.M. Li, H. Lin, Z. Teng, C. Liang, Digital prolate spheroidal window-based [70] M.C. Zhang, F.G. Wang, X.J. Yuan, L. Chen, 2D structured turbo compressed
S-transform for time-varying harmonic analysis, Electr. Power Syst. Res. 187 sensing for channel estimation in OTFS systems, in: 2018 IEEE International
(2020) 106512. Conference on Communication Systems, ICCS, 2018, pp. 45–49.
[43] E. Biglieri, P. Raviteja, Y. Hong, Error performance of orthogonal time fre- [71] X. Wu, S. Ma, X. Yang, Tensor-based low-complexity channel estimation for
quency space (OTFS) modulation, in: 2019 IEEE International Conference on mmWave massive MIMO–OTFS systems, J. Commun. Inf. Netw. 5 (3) (2020)
Communications Workshops, ICC Workshops, 2019, pp. 1–6. 324–334.
[44] R. Bomfin, A. Nimr, M. Chafii, G. Fettweis, A robust and low-complexity walsh- [72] L. Zhao, W.J. Gao, W.B. Guo, Sparse Bayesian learning of delay–Doppler
Hadamard modulation for doubly-dispersive channels, IEEE Commun. Lett. 25 channel for OTFS system, IEEE Commun. Lett. 24 (12) (2020) 2766–2769.
(3) (2021) 897–901. [73] W.J. Gao, S. Li, L. Zhao, Research on OTFS performance based on joint-
[45] M.N. Hossain, Y. Sugiura, T. Shimamura, H. Ryu, Waveform design of low sparse fast time-varying channel estimation, in: International Conference on
complexity WR–OTFS system for the OOB power reduction, in: 2020 IEEE Communications and Networking in China, 2019, pp. 707–719.
Wireless Communications and Networking Conference Workshops, WCNCW, [74] M.Y. Li, S. Zhang, F.F. Gao, P.Z. Fan, O.A. Dobre, A new path division multiple
2020, pp. 1–5. access for the massive MIMO–OTFS networks, IEEE J. Sel. Areas Commun. 39
[46] S. Tiwari, S.S. Das, Circularly pulse-shaped orthogonal time frequency space (4) (2021) 903–918.
modulation, Electron. Lett. 56 (3) (2019) 157–160. [75] Z.Q. Wei, W.J. Yuan, S.Y. Li, Off-grid channel estimation with sparse Bayesian
[47] Z.D. Wang, X.Y. Chen, X.Y. Ning, BER analysis of integrated WFRFT–OTFS learning for OTFS systems, IEEE Trans. Wirel. Commun. 21 (9) (2022)
waveform framework over static multipath channels, IEEE Commun. Lett. 25 7407–7426.
(3) (2021) 754–758. [76] L. Zhao, J. Yang, Y.L. Liu, Block sparse Bayesian learning-based channel
[48] M.M. Liu, S.Y. Li, C.Q. Zhang, Coded orthogonal time frequency space estimation for MIMO–OTFS systems, IEEE Commun. Lett. 26 (4) (2022)
modulation, ZTE Commun. 19 (4) (2021) 54–62. 892–896.
[49] S.Y. Li, J.H. Yuan, W.J. Yuan, Performance analysis of coded OTFS systems over [77] D. Shi, W.J. Wang, L. You, Deterministic pilot design and channel estimation for
high-mobility channels, IEEE Trans. Wirel. Commun. 20 (9) (2021) 6033–6048. downlink massive MIMO–OTFS systems in presence of the fractional Doppler,
[50] C. Zhang, W. Xing, J.H. Yuan, Performance of LDPC coded OTFS systems over IEEE Trans. Wirel. Commun. 20 (11) (2021) 7151–7165.
high mobility channels, ZTE Commun. 19 (4) (2022) 45–53. [78] H.Y. Zhang, X.J. Huang, J.A. Zhang, Frequency domain pilot-aided channel
[51] H. Tian, D.F. Zhao, H.X. Jia, Research on nonbinary LDPC–OTFS scheme in estimation for OTFS over fast fading channels, in: 2021 IEEE 94th Vehicular
high mobile communication scenarios, in: 2021 International Conference on Technology Conference, VTC2021-Fall, 2021, pp. 1–5.
Engineering and Emerging Technologies, ICEET, 2021, pp. 1–6. [79] H.B. Mishra, P. Singh, A.K. Prasad, Iterative channel estimation and data
[52] B.H. Cao, Z. Xiang, P. Ren, Low complexity transmitter precoding for MU detection in OTFS using superimposed pilots, in: 2021 IEEE International
MIMO-otfs, Digit. Signal Process. 115 (2021) 103083. Conference on Communications Workshops, ICC Workshops, 2021, pp. 1–6.
[53] Y. Zhang, P. Du, J. Wang, Partial and statistical CSI based precoding for [80] Y.R. Shan, F.G. Wang, Low-complexity and low-overhead receiver for OTFS via
reduced-CP OTFS with inter-slot-interference pre-cancellation, IET Commun. 16 large-scale antenna array, IEEE Trans. Veh. Technol. 70 (6) (2021) 5703–5718.
(8) (2022) 878–887. [81] S.K. Mohammed, Time-domain to delay–Doppler domain conversion of OTFS
[54] J.J. Sun, Z.L. Wang, Q. Huang, Secure precoded orthogonal time fre- signals in very high mobility scenarios, IEEE Trans. Veh. Technol. 70 (6) (2021)
quency space modulation, in: 2021 13th International Conference on Wireless 6178–6183.
Communications and Signal Processing, WCSP, 2021, pp. 1–5. [82] P. Singh, A. Gupta, H.B. Mishra, Low-complexity ZF/MMSE receivers for
[55] B.C. Pandey, S.K. Mohammed, P. Raviteja, Low complexity precoding and MIMO–OTFS system with imperfect CSI, 2020, arXiv preprint arXiv:2010.04057.
detection in multi-user massive MIMO OTFS downlink, IEEE Trans. Veh. [83] Y.R. Shan, F.F. Wang, Low-complexity and low-overhead receiver for OTFS via
Technol. 70 (5) (2021) 4389–4405. large-scale antenna array, IEEE Trans. Veh. Technol. 70 (6) (2021) 5703–5718.
[56] H.Y. Zhang, X.J. Huang, J.A. Zhang, Adaptive transmission with frequency– [84] S. Tiwari, S.S. Das, V. Rangamgari, Low complexity LMMSE receiver for OTFS,
domain precoding and linear equalization over fast fading channels, IEEE Trans. IEEE Commun. Lett. 23 (12) (2019) 2205–2209.
Wirel. Commun. 20 (11) (2021) 7420–7430. [85] N. Hashimoto, N. Osawa, K. Yamazaki, Channel estimation and equalization
[57] I. Ahmed, H. Khammari, A. Shahid, A survey on hybrid beamforming techniques for CP–OFDM-based OTFS in fractional Doppler channels, in: 2021 IEEE
in 5G: Architecture and system model perspectives, IEEE Commun. Surv. Tutor. International Conference on Communications Workshops, ICC Workshops, 2021,
20 (4) (2018) 3060–3097. pp. 1–7.
[58] S.G. Neelam, P.R. Sahu, Channel estimation and data detection of OTFS system [86] R. Bomfin, M. Chafii, A. Nimr, Channel estimation for MIMO space time coded
in the presence of receiver IQ imbalance, in: 2021 National Conference on OTFS under doubly selective channels, in: 2021 IEEE International Conference
Communications, NCC, 2021, pp. 1–6. on Communications Workshops, ICC Workshops, 2021, pp. 1–6.
[59] A. Naikoti, A. Chockalingam, A DNN-based OTFS transceiver with delay– [87] F. Gomez-Cuba, Compressed sensing channel estimation for OTFS modulation
Doppler channel training and IQI compensation, in: 2021 IEEE 32nd Annual In- in non-integer delay–Doppler domain, in: 2021 IEEE Global Communications
ternational Symposium on Personal, Indoor and Mobile Radio Communications, Conference, GLOBECOM, 2021, pp. 1–6.
PIMRC, 2021, pp. 628–634. [88] F. Liu, Z.D. Yuan, Q.H. Guo, Message passing-based structured sparse signal
[60] P. Händel, D. Rönnow, MIMO and massive MIMO transmitter crosstalk, IEEE recovery for estimation of OTFS channels with fractional Doppler shifts, IEEE
Trans. Wirel. Commun. 19 (3) (2020) 1882–1893. Trans. Wirel. Commun. 20 (12) (2021) 7773–7785.
[61] A. Liu, V.K.N. Lau, W. Dai, Exploiting burst-sparsity in massive MIMO with [89] Q.L. Wang, B.W. Jia, Z.Q. Zhang, OTFS channel estimation via 2D off-
partial channel support information, IEEE Trans. Wirel. Commun. 15 (11) grid decomposition and SBL combination, in: 2021 IEEE/CIC International
(2016) 7820–7830. Conference on Communications in China, ICCC Workshops, 2021, pp. 416–421.
[62] Y.S. Liu, S. Zhang, F.F. Gao, J.P. Ma, X.B. Wang, Uplink-aided high mobility [90] Y. Ge, Q.W. Deng, P.C. Ching, Z. Ding, Receiver design for OTFS with
downlink channel estimation over massive MIMO–OTFS system, IEEE J. Sel. fractionally spaced sampling approach, IEEE Trans. Wirel. Commun. 20 (7)
Areas Commun. 38 (9) (2020) 1994–2009. (2021) 4072–4086.
[63] M. Kollengode Ramachandran, A. Chockalingam, MIMO–OTFS in high-Doppler [91] J.Q. Cheng, C.L. Jia, H. Gao, W.J. Xu, Z.S. Bie, OTFS based receiver scheme
fading channels: Signal detection and channel estimation, in: 2018 IEEE Global with multi-antennas in high-mobility V2X systems, in: 2020 IEEE International
Communications Conference, GLOBECOM, 2018, pp. 206–212. Conference on Communications Workshops, ICC Workshops, 2020, pp. 1–6.

24
M. Li et al. Computer Networks 224 (2023) 109597

[92] W.J. Yuan, S.Y. Li, Z.Q. Wei, Data-aided channel estimation for OTFS systems [118] X.K. Xu, M.M. Zhao, M. Lei, A damped GAMP detection algorithm for OTFS
with a superimposed pilot and data transmission scheme, IEEE Wirel. Commun. system based on deep learning, in: 2020 IEEE 92nd Vehicular Technology
Lett. 10 (9) (2021) 1954–1958. Conference, VTC2020-Fall, 2020, pp. 1–5.
[93] F. Long, K. Niu, J.R. Lin, Joint channel estimation and equalization for OTFS [119] H.J. Zhang, T.T. Zhang, A low-complexity message passing detector for OTFS
based on EP, in: 2021 IEEE Global Communications Conference, GLOBECOM, modulation with probability clipping, IEEE Wirel. Commun. Lett. 10 (6) (2021)
2021, pp. 1–6. 1271–1275.
[94] G. Guo, Z.Y. Jin, X.Y. Zhang, Joint iterative channel estimation and symbol [120] L.P. Xiang, Y.S. Liu, L.L. Yang, Gaussian approximate message passing detection
detection for orthogonal time frequency space modulation, in: 2021 IEEE 94th of orthogonal time frequency space modulation, IEEE Trans. Veh. Technol. 70
Vehicular Technology Conference, VTC2021-Fall, 2021, pp. 1–6. (10) (2021) 10999–11004.
[95] A. Pfadler, T. Szollmann, P. Jung, Leakage suppression in pulse-shaped OTFS [121] Z.D. Yuan, F. Liu, W.J. Yuan, Iterative detection for orthogonal time frequency
delay–Doppler-pilot channel estimation, IEEE Wirel. Commun. Lett. 11 (6) space modulation with unitary approximate message passing, IEEE Trans. Wirel.
(2022) 1181–1185. Commun. 21 (2) (2022) 714–725.
[96] H.Y. Qu, G.H. Liu, L. Zhang, Low-dimensional subspace estimation of [122] N. Wu, Y.K. Zhang, Y.S. Ma, Vector approximate message passing based
continuous-Doppler-spread channel in OTFS systems, IEEE Trans. Commun. 69 iterative receiver for otfs system, in: 2021 IEEE/CIC International Conference
(7) (2021) 4717–4731. on Communications in China, ICCC Workshops, 2021, pp. 422–426.
[97] R.N. Liu, Y.H. Huang, D.Z. He, Optimizing channel estimation overhead for [123] H.Y. Liu, Y.M. Liu, M. Yang, A low complexity MAP detector for OTFS
OTFS with prior channel statistics, in: 2021 IEEE Wireless Communications and modulation in logarithmic domain, 2021, arXiv preprint arXiv:2109.10232.
Networking Conference, WCNC, 2021, pp. 1–6. [124] T. Wu, Z. Bie, C. Wu, A sparsely-connected OTFS–BFDM system using message-
[98] Z.Y. Huang, S. Cao, H. Wei, Research on channel estimation algorithm based passing decoding, IEICE Trans. Fundam. Electron. Commun. Comput. Sci. 105
on OTFS system, in: 2022 IEEE 6th Information Technology and Mechatronics (3) (2022) 576–583.
Engineering Conference, vol. 6, ITOEC, 2022, pp. 352–356. [125] S. Gao, J.P. Zheng, Clipping noise recovery of OTFS signals using reliability-
[99] F. Qin, Y.F. Ji, X.Y. Sun, Research on frequency domain channel estimation based compressed sensing, IEEE Trans. Veh. Technol. 70 (7) (2021)
method based on otfs system, in: 2022 IEEE 6th Information Technology and 7187–7192.
Mechatronics Engineering Conference, vol. 6, ITOEC, 2022, pp. 973–978. [126] S.Y. Li, W.J. Yuan, Z.Q. Wei, Cross domain iterative detection for orthogonal
[100] Y.J. Liu, Y.L. Guan, D.G. Gonzalez, Near-optimal BEM OTFS receiver with low time frequency space modulation, IEEE Trans. Wirel. Commun. 21 (4) (2022)
pilot overhead for high-mobility communications, IEEE Trans. Commun. 70 (5) 2227–2242.
(2022) 3392–3406. [127] H.Y. Qu, G.H. Liu, M.A. Imran, Efficient channel equalization and symbol
[101] Q. Li, Z. Xiang, P. Ren, Variational autoencoder based receiver for orthogonal detection for MIMO OTFS systems, IEEE Trans. Wirel. Commun. (2022) 1.
time frequency space modulation, Digit. Signal Process. 117 (2021) 103170. [128] Y. Zhang, Q.F. Zhang, C.B. He, Low-complexity belief propagation detector for
[102] S.Q. Wang, J. Guo, X.Y. Wang, Pilot design and optimization for OTFS OTFS system, in: 2021 IEEE International Conference on Signal Processing,
modulation, IEEE Wirel. Commun. Lett. 10 (8) (2021) 1742–1746. Communications and Computing, ICSPCC, 2021, pp. 1–5.
[103] H. Yan, M. Wang, A low complexity channel estimation scheme for orthogonal [129] A. Naikoti, A. Chockalingam, Low-complexity delay–Doppler symbol DNN for
time frequency space (OTFS) system with synchronization errors, in: 2021 IEEE OTFS signal detection, in: 2021 IEEE 93rd Vehicular Technology Conference,
6th International Conference on Computer and Communication Systems, ICCCS, VTC2021-Spring, 2021, pp. 1–6.
2021, pp. 576–581. [130] T. Thaj, E. Viterbo, Low-complexity linear diversity-combining detector for
[104] J. Pan, Cramer–Rao low bound of channel estimation for orthogonal time MIMO–OTFS, IEEE Wirel. Commun. Lett. 11 (2) (2022) 288–292.
frequency space modulation system, IEEE Trans. Veh. Technol. 70 (10) (2021) [131] D. Wang, B.L. Sun, F.G. Wang, Transmit diversity scheme design for rectangular
9646–9658. pulse shaping based OTFS, China Commun. 19 (3) (2022) 116–128.
[105] N. Lee, MAP support detection for greedy sparse signal recovery algorithms in [132] P. Singh, K. Yadav, H.B. Mishra, BER analysis for OTFS zero forcing receiver,
compressive sensing, IEEE Trans. Signal Process. 64 (19) (2016) 4987–4999. IEEE Trans. Commun. 70 (4) (2022) 2281–2297.
[106] P. Raviteja, K.T. Phan, Q. Jin, Y. Hong, E. Viterbo, Low-complexity iterative [133] K. Yadav, P. Singh, H.B. Mishra, Closed form BER for ZF OTFS receivers,
detection for orthogonal time frequency space modulation, in: 2018 IEEE in: 2021 IEEE 22nd International Workshop on Signal Processing Advances in
Wireless Communications and Networking Conference, WCNC, 2018, pp. 1–6. Wireless Communications, SPAWC, 2021, pp. 136–140.
[107] K.R. Murali, A. Chockalingam, On OTFS modulation for high-Doppler fading [134] P. Singh, A. Gupta, H.B. Mishra, Low-complexity ZF/MMSE MIMO–OTFS
channels, in: 2018 Information Theory and Applications Workshop, ITA, 2018, receivers for high-speed vehicular communication, IEEE Open J. Commun. Soc.
pp. 1–10. 3 (2022) 209–227.
[108] B.C. Pandey, S.K. Mohammed, P. Raviteja, Y. Hong, E. Viterbo, Low complexity [135] O. Castañeda, S. Jacobsson, G. Durisi, T. Goldstein, C. Studer, Finite-alphabet
precoding and detection in multi-user massive MIMO OTFS downlink, IEEE MMSE equalization for all-digital massive MU–MIMO mmWave communication,
Trans. Veh. Technol. (2021) 1. IEEE J. Sel. Areas Commun. 38 (9) (2020) 2128–2141.
[109] L.J. Li, Y. Liang, P.Z. Fan, Y.L. Guan, Low complexity detection algorithms [136] Y. Wang, J. Gui, Y. Yin, J. Wang, J.L. Sun, G. Gui, H. Gacanin, H. Sari,
for OTFS under rapidly time-varying channel, in: 2019 IEEE 89th Vehicular F. Adachi, Automatic modulation classification for MIMO systems via deep
Technology Conference, VTC2019-Spring, 2019, pp. 1–5. learning and zero-forcing equalization, IEEE Trans. Veh. Technol. 69 (5) (2020)
[110] Z.D. Yuan, F. Liu, W.J. Yuan, Iterative detection for orthogonal time fre- 5688–5692.
quency space modulation using approximate message passing with unitary [137] A. Pfadler, P. Jung, S. Stanczak, Pulse-shaped OTFS for V2X short-frame com-
transformation, IEEE Trans. Wirel. Commun. 21 (2) (2021) 714–725. munication with tuned one-tap equalization, in: WSA 2020; 24th International
[111] T. Zemen, M. Hofer, D. Löschenbrand, C. Pacher, Iterative detection for ITG Workshop on Smart Antennas, 2020, pp. 1–6.
orthogonal precoding in doubly selective channels, in: 2018 IEEE 29th [138] A. Pfadler, P. Jung, S. Stanczak, Mobility modes for pulse-shaped OTFS with
Annual International Symposium on Personal, Indoor and Mobile Radio linear equalizer, in: GLOBECOM 2020 - 2020 IEEE Global Communications
Communications, PIMRC, 2018, pp. 1–7. Conference, 2020, pp. 1–6.
[112] S.Y. Li, W.J. Yuan, Z.Q. Wei, Hyhrid MAP and PIC detection for OTFS [139] J.Q. Cheng, H. Gao, W. Xu, Low-complexity linear equalizers for OTFS ex-
modulation, IEEE Trans. Veh. Technol. 70 (7) (2021) 7193–7198. ploiting two-dimensional fast fourier transform, 2019, arXiv preprint arXiv:
[113] Y. Liang, L.J. Li, P.Z. Fan, Y.L. Guan, Doppler resilient orthogonal time– 1909.00524.
frequency space (OTFS) systems based on index modulation, in: 2020 IEEE [140] T.T. Zou, W.J. Xu, H. Gao, Low-complexity linear equalizers for OTFS sys-
91st Vehicular Technology Conference, VTC2020-Spring, 2020, pp. 1–5. tem with rectangular waveforms, in: 2021 IEEE International Conference on
[114] W.J. Yuan, Z.Q. Wei, J.H. Yuan, D.W.K. Ng, A simple variational Bayes detector Communications Workshops, ICC Workshops, 2021, pp. 1–6.
for orthogonal time frequency space (OTFS) modulation, IEEE Trans. Veh. [141] G.D. Surabhi, A. Chockalingam, Low-complexity linear equalization for 2 × 2
Technol. 69 (7) (2020) 7976–7980. MIMO–OTFS signals, in: 2020 IEEE 21st International Workshop on Signal
[115] X.W. Qu, A. Kosasih, W. Hardjawana, Bayesian-based symbol detector for Processing Advances in Wireless Communications, SPAWC, 2020, pp. 1–5.
orthogonal time frequency space modulation systems, in: 2021 IEEE 32nd [142] T. Thaj, E. Viterbo, Low complexity iterative rake decision feedback equalizer
Annual International Symposium on Personal, Indoor and Mobile Radio for zero-padded OTFS systems, IEEE Trans. Veh. Technol. 69 (12) (2020)
Communications, PIMRC, 2021, pp. 1154–1159. 15606–15622.
[116] Z.H. Gong, S.H. Liu, Y.M. Huang, Doppler diversity reception for OTFS [143] F. Long, K. Niu, C. Dong, J.R. Lin, Low complexity iterative LMMSE–PIC
modulation, in: Conference: 2022 IEEE 95th Vehicular Technology Conference, equalizer for OTFS, in: ICC 2019 - 2019 IEEE International Conference on
VTC2022-Spring, 2022, pp. 1–5. Communications, ICC, 2019, pp. 1–6.
[117] Y.K. Enku, B. Bai, F. Wan, Two-dimensional convolutional neural network-based [144] L. Li, H. Wei, Y. Huang, A simple two-stage equalizer with simplified orthogonal
signal detection for OTFS systems, IEEE Wirel. Commun. Lett. 10 (11) (2021) time frequency space modulation over rapidly time-varying channels, 2017,
2514–2518. arXiv preprint arXiv:1709.02505.

25
M. Li et al. Computer Networks 224 (2023) 109597

[145] R. Marsalek, J. Blumenstein, D. Schützenhöfer, M. Pospisil, OTFS modulation [173] F. Wei, W. Chen, Y.P. Wu, J. Li, Y. Luo, Toward 5G wireless interface
and influence of wideband RF impairments measured on a 60 GHz testbed, technology: Enabling nonorthogonal multiple access in the sparse code domain,
in: 2020 IEEE 21st International Workshop on Signal Processing Advances in IEEE Veh. Technol. Mag. 13 (4) (2018) 18–27.
Wireless Communications, SPAWC, 2020, pp. 1–5. [174] S. Lim, N. Kim, H. Park, Uplink SCMA system with multiple antennas, IEEE
[146] T. Zemen, M. Hofer, D. Loeschenbrand, Low-complexity equalization for equal- Trans. Veh. Technol. 66 (8) (2017) 6982–6992.
ization for orthogonal time and frequecy signaling (OTFS), 2017, arXiv preprint [175] Z.G. Ding, R. Schober, P. Fan, H. Vincent Poor, OTFS–NOMA: An efficient ap-
arXiv:1710.09916. proach for exploiting heterogenous user mobility profiles, IEEE Trans. Commun.
[147] L.F. Xu, N. Li, W. Xiang, Improved time–frequency domain linear re- 67 (11) (2019) 7950–7965.
ceiver for OFDM-based OTFS, in: 2021 IEEE/CIC International Conference on [176] Z.G. Ding, Robust beamforming design for OTFS–NOMA, IEEE Open J. Commun.
Communications in China, ICCC, 2021, pp. 410–414. Soc. 1 (2020) 33–40.
[148] A. Pfadler, P. Jung, T. Szollmann, Pulse-shaped OTFS over doubly-dispersive [177] R.M. Augustine, A. Chockalingam, Interleaved time–frequency multiple access
channels: One-tap vs. Full LMMSE equalizers, in: 2021 IEEE International using OTFS modulation, in: 2019 IEEE 90th Vehicular Technology Conference,
Conference on Communications Workshops, ICC Workshops, 2021, pp. 1–6. VTC2019-Fall, 2019, pp. 1–5.
[149] W. Shin, K. Kim, K. Chang, Coded multicarrier systems for high mobility [178] K. Deka, A. Thomas, S. Sharma, OTFS–NOMA based on SCMA, 2020, arXiv
environments in 30 GHz band, in: 2021 International Conference on Information preprint arXiv:2005.03216.
and Communication Technology Convergence, ICTC, 2021, pp. 628–633. [179] Z.Q. Kang, H. Zhao, H. Wang, An efficient two-dimension OTFS–NOMA scheme
[150] T.T. Zou, W.J. Xu, H. Gao, Low-complexity linear equalization for OTFS based on heterogenous mobility users grouping, in: 2021 IEEE 21st International
systems with rectangular waveforms, in: 2021 IEEE International Conference Conference on Communication Technology, ICCT, 2021, pp. 726–730.
on Communications Workshops, ICC Workshops, 2021, pp. 1–6. [180] Y. Ge, Q.W. Deng, P.C. Ching, OTFS signaling for uplink NOMA of
[151] F. Long, K. Niu, J.R. Lin, Low complexity block equalizer for OTFS based on heterogeneous mobility users, IEEE Trans. Commun. 69 (5) (2021) 3147–3161.
expectation propagation, IEEE Wirel. Commun. Lett. 11 (2) (2022) 376–380. [181] A. Chatterjee, V. Rangamgari, S. Tiwari, Non orthogonal multiple access with
[152] Z. Zhou, L.J. Liu, J.R. Xu, Learning to equalize OTFS, IEEE Trans. Wirel. orthogonal time frequency space signal transmission, 2020, arXiv preprint
Commun. (2022) 1. arXiv:2003.06387.
[153] Y. Zhang, Q. Zhang, C. He, Delay–Doppler domain decision feedback turbo [182] W.D. Shao, S. Zhang, C.J. Zhong, X.F. Lei, P.Z. Fan, Angle-delay–Doppler
equalization for OTFS modulation, Phys. Commun. 52 (2022) 101699. domain NOMA over massive MIMO–OTFS networks, in: 2020 IEEE/CIC
[154] Y. Zhang, Q.F. Zhang, L.L. Zhang, A low-complexity approximate message International Conference on Communications in China, ICCC, 2020, pp. 74–79.
passing equalizer for otfs system, in: 2021 IEEE/CIC International Conference [183] A.K. Sinha, S.K. Mohammed, P. Raviteja, Y. Hong, E. Viterbo, OTFS based
on Communications in China, ICCC Workshops, 2021, pp. 449–454. random access preamble transmission for high mobility scenarios, IEEE Trans.
[155] P. Raviteja, Y. Hong, E. Viterbo, E. Biglieri, Effective diversity of OTFS Veh. Technol. 69 (12) (2020) 15078–15094.
modulation, IEEE Wirel. Commun. Lett. 9 (2) (2020) 249–253. [184] Y.Y. Ma, G.Y. Ma, N. Wang, OTFS–TSMA for massive Internet of Things in
[156] G.D. Surabhi, R.M. Augustine, A. Chockalingam, On the diversity of uncoded high-speed railway, IEEE Trans. Wirel. Commun. 21 (1) (2022) 519–531.
OTFS modulation in doubly-dispersive channels, IEEE Trans. Wirel. Commun. [185] O.B. Akan, M. Arik, Internet of radars: Sensing versus sending with joint
18 (6) (2019) 3049–3063. radar-communications, IEEE Commun. Mag. 58 (9) (2020) 13–19.
[157] R.M. Augustine, G.D. Surabhi, A. Chockalingam, Space-time coded OTFS mod- [186] L. Liang, H.X. Peng, G.Y. Li, X.M. Shen, Vehicular communications: A physical
ulation in high-Doppler channels, in: 2019 IEEE 89th Vehicular Technology layer perspective, IEEE Trans. Veh. Technol. 66 (12) (2017) 10647–10659.
Conference, VTC2019-Spring, 2019, pp. 1–6. [187] L.M. Li, D. Wang, X. Niu, L. Chen, L. He, mmWave communication for 5G:
[158] H.Y. Zhang, X.J. Huang, J.A. Zhang, Comparison of OTFS diversity performance Implementation challenges and advances, Sci. China Inf. Sci. 61 (2) (2018).
over slow and fast fading channels, in: 2019 IEEE/CIC International Conference [188] M. Säily, C.B. Estevan, J.J. Gimenez, F. Tesema, W. Guo, D. Gomez-Barquero,
on Communications in China, ICCC, 2019, pp. 828–833. D. Mi, 5G radio access network architecture for terrestrial broadcast services,
[159] X.L. Feng, M.J. Wang, L. Chen, W.X. Zhu, A new timing and frequency IEEE Trans. Broadcast. 66 (2) (2020) 404–415.
offset estimation algorithm for filtered OFDM system., EURASIP J. Adv. Signal [189] H. Kim, E. Cha, J. Kim, A robust broadcast system under time- varying
Process. 2020 (38) (2020). channels based on OTFS modulation, in: 2021 IEEE International Symposium
[160] Y.B. Zou, H.P. Liu, Semidefinite programming methods for alleviating clock on Broadband Multimedia Systems and Broadcasting, BMSB, 2021, pp. 1–4.
synchronization bias and sensor position errors in TDOA localization, IEEE [190] B.Y. Xu, Z. Xia, R. Liu, Y. Zhang, J. Hu, W. Xie, Research on OTFS modulation
Signal Process. Lett. 27 (2020) 241–245. applied in LTE-based 5G terrestrial broadcast, in: 2020 International Wireless
[161] P. Wei, Y. Xiao, W. Feng, N. Ge, M. Xiao, Charactering the peak-to-average Communications and Mobile Computing, IWCMC, 2020, pp. 514–519.
power ratio of OTFS signals: A large system analysis, IEEE Trans. Wirel. [191] T. Blazek, D. Radovic, Performance evaluation of OTFS over measured V2V
Commun. 21 (6) (2022) 3705–3720. channels at 60 GHz, in: 2020 IEEE MTT-S International Conference on
[162] G.D. Surabhi, R.M. Augustine, A. Chockalingam, Peak-to-average power ratio Microwaves for Intelligent Mobility, ICMIM, 2020, pp. 1–4.
of OTFS modulation, IEEE Commun. Lett. 23 (6) (2019) 999–1002. [192] D. Basak, S. Sruti, K. Giridhar, Comparing precoded MC–CDMA and OTFS for
[163] S. Gao, J.P. Zheng, Peak-to-average power ratio reduction in pilot-embedded high-speed V2X communications, in: 2021 IEEE 94th Vehicular Technology
OTFS modulation through iterative clipping and filtering, IEEE Commun. Lett. Conference, VTC2021-Fall, 2021, pp. 1–5.
24 (9) (2020) 2055–2059. [193] C.S. Reddy, D. Sen, C. Singhal, Performance analysis of NR based vehicular
[164] M.X. Liu, M.M. Zhao, M. Lei, Autoencoder based PAPR reduction for OTFS mod- IoT system with OTFS modulation, in: 2021 IEEE 94th Vehicular Technology
ulation, in: 2021 IEEE 94th Vehicular Technology Conference, VTC2021-Fall, Conference, VTC2021-Fall, 2021, pp. 1–6.
2021, pp. 1–5. [194] M.K. Ramachandran, G.D. Surabhi, A. Chockalingam, OTFS: A new modulation
[165] C. Naveen, V. Sudha, Peak-to-average power ratio reduction in OTFS modu- scheme for high-mobility use case, J. Indian Inst. Sci. 100 (2) (2020) 315–336.
lation using companding technique, in: 2020 5th International Conference on [195] F. Wiffen, L. Sayer, M.Z. Bocus, A. Doufexi, A. Nix, Comparison of OTFS and
Devices, Circuits and Systems, ICDCS, 2020, pp. 140–143. OFDM in ray launched sub-6 GHz and mmWave line-of-sight mobility channels,
[166] M.N. Hossain, Y. Sugiura, T. Shimamura, DFT-Spread OTFS communication in: 2018 IEEE 29th Annual International Symposium on Personal, Indoor and
system with the reductions of PAPR and nonlinear degradation, Wirel. Pers. Mobile Radio Communications, PIMRC, 2018, pp. 73–79.
Commun. 115 (3) (2020) 2211–2228. [196] J.L. Lin, G.P. Wang, R.T. Xu, Channel and phase shift estimation for TM-aided
[167] J.K. Francis, R.M. Augustine, A. Chockalingam, Diversity and PAPR enhance- OTFS railway communications, in: 2021 IEEE/CIC International Conference on
ment in OTFS using indexing, in: 2021 IEEE 93rd Vehicular Technology Communications in China, ICCC Workshops, 2021, pp. 444–448.
Conference, VTC2021-Spring, 2021, pp. 1–6. [197] S.A. Hassani, V. Lampu, K. Parashar, L. Anttila, A. Bourdoux, B.v. Liempd, M.
[168] V. Khammammetti, S.K. Mohammed, OTFS-Based multiple-access in high Valkama, F. Horlin, S. Pollin, In-band full-duplex radar-communication system,
Doppler and delay spread wireless channels, IEEE Wirel. Commun. Lett. 8 (2) IEEE Syst. J. 15 (1) (2021) 1086–1097.
(2019) 528–531. [198] L. Gaudio, M. Kobayashi, B. Bissinger, G. Caire, Performance analysis of joint
[169] G.D. Surabhi, R.M. Augustine, A. Chockalingam, Multiple access in the delay- radar and communication using OFDM and OTFS, in: 2019 IEEE International
Doppler domain using OTFS modulation, 2019, arXiv preprint arXiv:1902. Conference on Communications Workshops, ICC Workshops, 2019, pp. 1–6.
03415. [199] M.F. Keskin, H. Wymeersch, A. Alvarado, Radar sensing with OTFS: Embracing
[170] V. Khammametti, S.K. Mohammed, Spectral efficiency of OTFS based orthogonal ISI and ICI to surpass the ambiguity barrier, in: 2021 IEEE International
multiple access with rectangular pulses, IEEE Trans. Veh. Technol. (2022) 1–16. Conference on Communications Workshops, ICC Workshops, 2021, pp. 1–6.
[171] Y. Zhang, H.M. Wang, T.X. Zheng, Q. Yang, Energy-efficient transmission design [200] L. Gaudio, M. Kobayashi, G. Caire, G. Colavolpe, Joint radar target detection
in non-orthogonal multiple access, IEEE Trans. Veh. Technol. 66 (3) (2017) and parameter estimation with MIMO OTFS, in: 2020 IEEE Radar Conference,
2852–2857. RadarConf20, 2020, pp. 1–6.
[172] O. Maraqa, A.S. Rajasekaran, S. Al-Ahmadi, H. Yanikomeroglu, S.M. Sait, A [201] L. Gaudio, M. Kobayashi, G. Caire, G. Colavolpe, Hybrid digital-analog beam-
survey of rate-optimal power domain NOMA with enabling technologies of forming and MIMO radar with OTFS modulation, 2020, arXiv preprint arXiv:
future wireless networks, IEEE Commun. Surv. Tutor. 22 (4) (2020) 2192–2235. 2009.08785.

26
M. Li et al. Computer Networks 224 (2023) 109597

[202] S.Y. Li, W.J. Yuan, C. Liu, A novel ISAC transmission framework based on [229] H.R. Ren, W.K. Xu, L. Wang, Orthogonal time–frequency space with improved
spatially-spread orthogonal time frequency space modulation, IEEE J. Sel. Areas index modulation, in: 2021 15th International Conference on Signal Processing
Commun. 40 (6) (2022) 1854–1872. and Communication Systems, ICSPCS, 2021, pp. 1–6.
[203] W.J. Yuan, S.Y. Li, Z.Q. Wei, Bypassing channel estimation for OTFS transmis- [230] W. Anwar, A. Kumar, N. Franchi, G. Fettweis, Performance analysis using
sion: An integrated sensing and communication solution, in: 2021 IEEE Wireless physical layer abstraction modeling for 5G and beyond waveforms, in: 2019
Communications and Networking Conference Workshops, WCNCW, 2021, pp. IEEE Global Communications Conference, GLOBECOM, 2019, pp. 1–6.
1–5. [231] R. Mallaiah, V.V. Mani, A novel OTFS system based on DFrFT–OFDM, IEEE
[204] W.J. Yuan, Z.Q. Wei, S.Y. Li, Integrated sensing and communication-assisted Wirel. Commun. Lett. (2022).
orthogonal time frequency space transmission for vehicular networks, IEEE J. [232] F.O. Olowononi, D.B. Rawat, C. Liu, Resilient machine learning for networked
Sel. Top. Signal Process. 15 (6) (2021) 1515–1528. cyber physical systems: A survey for machine learning security to securing
[205] C.W. Liu, S.H. Liu, Z.H. Mao, Low-complexity parameter learning for OTFS machine learning for CPS, IEEE Commun. Surv. Tutor. 23 (1) (2021) 524–552.
modulation based automotive radar, in: ICASSP 2021 - 2021 IEEE International [233] J. Wang, R. Li, Y. Ge, Q. Zhang, W. Shi, Artificial intelligence and wireless
Conference on Acoustics, Speech and Signal Processing, ICASSP, 2021, pp. communication, Front. Inf. Technol. Electron. Eng. 21 (10) (2020) 1413–1426.
8208–8212. [234] Y.B. Liu, Y. Liu, C. Yang, Modulation recognition with graph convolutional
[206] H.J. Zhang, T.T. Zhang, Y. Shen, Modulation symbol cancellation for OTFS- network, IEEE Wirel. Commun. Lett. 9 (5) (2020) 624–627.
Based joint radar and communication, in: 2021 IEEE International Conference [235] L.L. Dai, R.C. Jiao, F.Y. Adachi, H.V. Poor, Deep learning for wireless communi-
on Communications Workshops, ICC Workshops, 2021, pp. 1–6. cations: An emerging interdisciplinary paradigm, IEEE Wirel. Commun. 27 (4)
[207] Y.P. Yan, J.Y. Zhang, C.Z. Shan, Low-complexity OTFS-based hybrid precoding (2020) 133–139.
in mmWave high mobility systems, Digit. Signal Process. 126 (2022) 103496. [236] Q. Li, Y. Gong, Y. Liu, Explore the performance of receiver algorithm in OTFS
[208] D. Radovic, C.F. Mecklenbrauker, T. Blazek, OTFS performance over different based on DNN, 2022.
measured vehicular 60 GHz millimeter-wave channels, in: 2021 Interna- [237] Q.Y. Li, Y. Gong, F. Meng, Data-driven receiver for OTFS system with deep
tional Conference on Software, Telecommunications and Computer Networks, learning, in: 2021 7th IEEE International Conference on Network Intelligence
SoftCOM, 2021, pp. 1–5. and Digital Content, IC-NIDC, 2021, pp. 172–176.
[209] H. Ji, Y. Kim, T. Yim, Compressed OTFS transmission over millimeter-wave and [238] K. Yogesh, S. Manu, J. Gaurav, Y.S. Kumar, Automatic modulation classification
terahertz spectrum, in: 2021 IEEE International Conference on Communications based on constellation density using deep learning, IEEE Commun. Lett. 24 (6)
Workshops, ICC Workshops, 2021, pp. 1–6. (2020) 1275–1278.
[210] G.D. Surabhi, M.K. Ramachandran, A. Chockalingam, OTFS modulation with [239] F. Liang, C. Shen, F. Wu, An iterative BP–CNN architecture for channel
phase noise in mmWave communications, in: 2019 IEEE 89th Vehicular decoding, IEEE J. Sel. Top. Signal Process. 12 (1) (2018) 144–159.
Technology Conference, VTC2019-Spring, 2019, pp. 1–5. [240] W.Z. Khan, E. Ahmed, S. Hakak, I. Yaqoob, A. Ahmed, Edge computing: A
[211] R. Hadani, S. Rakib, A.F. Molisch, C. Ibars, A. Monk, M. Tsatsanis, J. survey, Future Gen. Comput. Syst. 97 (2019) 219–235.
Delfeld, A. Goldsmith, R. Calderbank, Orthogonal time frequency space (OTFS) [241] C.T. Healy, A.A. Rawi, C.C. Tsimenidis, LDPC-coded modulation performance
modulation for millimeter-wave communications systems, in: 2017 IEEE MTT-S analysis and system design, IEEE Access 8 (2020) 212166–212176.
International Microwave Symposium, IMS, 2017, pp. 681–683. [242] K. Niu, Y. Li, Polar codes for fast fading channel: Design based on polar
[212] V.D.T. Hai, Benefits of 5G Millimeter-wave communication in IoT applications. spectrum, IEEE Trans. Veh. Technol. 69 (9) (2020) 10103–10114.
[243] Y.C. Yang, L.F. Wu, G.S. Yin, L.J. Li, H.B. Zhao, A survey on security and
[213] Y.Z. Wu, C. Han, T. Yang, DFT-Spread orthogonal time frequency space
privacy issues in Internet-of-Things, IEEE Internet of Things J. 4 (5) (2017)
modulation design for terahertz communications, in: 2021 IEEE Global
1250–1258.
Communications Conference, GLOBECOM, 2021, pp. 1–6.
[244] O.A. Topal, M.O. Demir, Z. Liang, A physical layer security framework for
[214] Y.Z. Wu, C. Han, Z. Chen, DFT-Spread orthogonal time frequency space system
cognitive cyber-physical systems, IEEE Wirel. Commun. 27 (4) (2020) 32–39.
with superimposed pilots for terahertz integrated sensing and communication,
[245] J.Q. Zhang, A. Marshall, R. Woods, Design of an OFDM physical layer
2022, arXiv preprint arXiv:2202.10035.
encryption scheme, IEEE Trans. Veh. Technol. 66 (3) (2017) 2114–2127.
[215] T.M.C. Chu, H.J. Zepernick, A. Westerhagen, Performance assessment of OTFS
[246] K. Sanjeev, S. Amarpal, K. Manoj, Covert communication integrates into wavelet
modulation in high Doppler airborne communication networks, Mobile Netw.
packet transform OFDM system over Rayleigh fading channel, Wirel. Netw. 26
Appl. (2022) 1–11.
(1) (2020) 81–89.
[216] A. Albanese, V. Sciancalepore, X. Costa-Perez, First responders got wings: UAVs
[247] D. Lee, H. Sasaki, H. Fukumoto, Orbital angular momentum (QAM) multiplex-
to the rescue of localization operations in beyond 5G systems, IEEE Commun.
ing: An enabler of a new era of wireless communications, IEICE Trans. Commun.
Mag. 59 (11) (2021) 28–34.
E100B (7) (2017) 1044–1063.
[217] F. Linsalata, A. Albanese, V. Sciancalepore, OTFS-superimposed PRACH-aided
[248] Q. Tao, J.W. Wang, C.J. Zhong, Performance analysis of intelligent reflecting
localization for UAV safety applications, in: 2021 IEEE Global Communications
surface aided communication systems, IEEE Commun. Lett. 24 (11) (2020)
Conference, GLOBECOM, 2021, pp. 1–6.
2464–2468.
[218] X.Y. Zhou, K. Ying, Z. Gao, Active terminal identification, channel esti-
[249] H.X. Xie, F.F. Gao, S. Zhang, S. Jin, A unified transmission strategy for
mation, and signal detection for grant-free NOMA–OTFS in LEO satellite
TDD/FDD massive MIMO systems with spatial basis expansion model, IEEE
internet-of-things, 2022, arXiv preprint arXiv:2201.02084.
Trans. Veh. Technol. 66 (4) (2017) 3170–3184.
[219] Y.Y. Ma, G.Y. Ma, N. Wang, OTFS enabled NOMA for MMTC systems over LEO
[250] T. Thaj, E. Viterbo, OTFS modem SDR implementation and experimental study
satellite, ZTE Commun. 19 (4) (2022) 63–70.
of receiver impairment effects, in: 2019 IEEE International Conference on
[220] J. Shi, J.F. Hu, Y. Yue, Outage probability for OTFS based downlink LEO Communications Workshops, ICC Workshops, 2019, pp. 1–6.
satellite communication, IEEE Trans. Veh. Technol. 71 (3) (2022) 3355–3360. [251] A. Farhang, A. RezazadehReyhani, L.E. Doyle, B. Farhang-Boroujeny, Low
[221] B.X. Shen, Y. Wu, J. An, Random access with massive MIMO–OTFS in LEO complexity modem structure for OFDM-based orthogonal time frequency space
satellite communications, 2022, arXiv preprint arXiv:2202.13058. modulation, IEEE Wirel. Commun. Lett. 7 (3) (2018) 344–347.
[222] L.Y. Jing, N. Zhang, C. He, OTFS underwater acoustic communications based [252] A. Jovicic, J. Li, T. Richardson, Visible light communication: Opportunities,
on passive time reversal, Appl. Acoust. 185 (2022). challenges and the path to market, IEEE Commun. Mag. 51 (12) (2013) 26–32.
[223] M. Mohammadi, H.Q. Ngo, M. Matthaiou, Cell-free massive MIMO meets OTFS [253] M. Li, F. Jiang, C. Pei, Research on visible light indoor positioning algorithm
modulation, IEEE Trans. Commun. (2022). based on fire safety, Opt. Appl. 50 (2) (2020).
[224] M. Mohammadi, H.Q. Ngo, M. Matthaiou, Cell-free massive MIMO with OTFS [254] D. Zheng, H.M. Zhang, J. Song, DCO–OTFS-based full-duplex relay-assisted
modulation: Power control and resource allocation, 2022, arXiv preprint arXiv: visible light communications, Optics Express. 29 (25) (2021) 41323–41332.
2203.07549. [255] S. Sinha, A. Chockalingam, OTFS modulation in dual-LED indoor visible light
[225] C.Y. An, H.G. Ryu, High throughout mobile communication based on OTFS communication systems, in: 2021 IEEE 94th Vehicular Technology Conference,
system with the delay–Doppler compensation, Wirel. Pers. Commun. 106 (2) VTC2021-Fall, 2021, pp. 1–7.
(2019) 473–486. [256] A. Sharma, S. Jain, R. Mitra, Performance analysis of OTFS over multipath
[226] C.Y. An, H.G. Ryu, Design and performance evaluation of spectral efficient channels for visible light communication, in: 2020 International Conference on
orthogonal time frequency space system, in: 2019 IEEE 2nd 5G World Forum, Communications, Signal Processing, and their Applications, ICCSPA, 2021, pp.
5GWF, 2019, pp. 249–252. 1–6.
[227] M.S. Khan, Y.J. Kim, Q. Sultan, Downlink synchronization for OTFS-based [257] S. Sinha, A. Chockalingam, Quad-LED OTFS modulation in indoor visible light
cellular systems in high Doppler environments, IEEE Access 9 (2021) communication systems, in: 2021 IEEE Global Communications Conference,
73575–73589. GLOBECOM, 2021, pp. 1–6.
[228] H. Zhao, D. He, Z. Kang, Orthogonal time frequency space (OTFS) with [258] J. Zhong, J. Zhou, S. Gao, Secure orthogonal time–frequency multiplexing with
dual-mode index modulation, IEEE Wirel. Commun. Lett. 10 (5) (2021) two-dimensional encryption for optical-wireless communications, Chin. Opt.
991–995. Lett. 19 (5) (2021) 050603.

27
M. Li et al. Computer Networks 224 (2023) 109597

[259] G. Gourab, Cooperative relaying for URLLC in V2X networks, IEEE Wirel. Wei Liu received the B.Sc., M.Sc., and Ph.D. degrees from
Commun. Lett. 10 (1) (2021) 97–101. the National University of Defense Technology (NUDT),
[260] J. Nam, G. Caire, M. Debbah, H.V. Poor, Capacity scaling of massive MIMO in Changsha, China, in 2002, 2004, and 2010, respectively. He
strong spatial correlation regimes, IEEE Trans. Inform. Theory 66 (5) (2020) is currently associate professor of Department of Communi-
3040–3064. cations Engineering, College of Electronic Science, National
[261] S. Mumtaz, J.M. Jornet, J. Aulin, Terahertz communication for vehicular University of Defense Technology. He has published many
networks, IEEE Trans. Veh. Technol. 66 (7) (2017) 5617–5625. papers in various journals and conference proceedings.
[262] S. Ghafoor, N. Boujnah, M.H. Rehmani, A. Davy, MAC protocols for terahertz His research interests include information theory, space–
communication: A comprehensive survey, IEEE Commun. Surv. Tutor. 22 (4) time coding, physical layer security, wireless communication
(2020) 2236–2282. technology and 5G technology, etc.
[263] A. Alexiou, S. Andreev, G. Fodor, T. Nagatsuma, THz communications: A
catalyst for the wireless future, IEEE Commun. Mag. 58 (11) (2020) 12–13.
Jing Lei received the B.Sc., M.Sc., and Ph.D. degrees from
the National University of Defense Technology (NUDT),
Changsha, China, in 1990, 1994, and 2009, respectively.
Mao Li is currently working toward the Ph.D. degree She is currently a distinguished professor of Department of
in the Department of Communication Engineering, School Communications Engineering, College of Electronic Science,
of Electronic Science, NUDT. His research interests in- National University of Defense Technology, the leader of
clude orthogonal time–frequency space modulation, radar communication coding group. She was a visiting scholar in
beamforming, 5G technology. the School of Electronics and Computer Science, University
of Southampton, UK. She has published many papers in vari-
ous journals and conference proceedings and five books. Her
research interests include information theory, LDPC, space–
time coding, advanced multiple access technology, physical
layer security, and wireless communication technology, etc.

28