Performance of LDPC Coded OTFS Systems over High Mobility Channels

ZHANG Chong, XING Wang, YUAN Jinhong, ZHOU Yiqing

Abstract： The upcoming 6G wireless networks have to provide reliable communications inhigh-mobility scenarios at high carrier frequencies. However， high-mobility or high carrierfrequencies will bring severe inter-carrier interference （ICI） to conventional orthogonal fre ?quency-division multiplexing （OFDM） modulation. Orthogonal time frequency space （OTFS）modulation is a recently developing multi-carrier transmission scheme for wireless commu ?nications in high-mobility environments. This paper evaluates the performance of coded OT?FS systems. In particular， we consider 5G low density parity check （LDPC） codes for OTFSsystems based on 5G OFDM frame structures over high mobility channels. We show the per?formance of the OTFS systems with 5G LDPC codes when sum-product detection algorithmand iterative detection and decoding are employed. We also illustrate the effect of channelestimation error on the performance of the LDPC coded OTFS systems.

Keywords： OTFS; LDPC codes; OFDM

Citation （IEEE Format）： C. Zhang， W. Xing， J. H. Yuan， et al.， “Performance of LDPC coded OTFS systems over high mobility channels ， ” ZTE Communications， vol. 19， no. 4， pp. 45 –53， Dec. 2021. doi： 10. 12142/ZTECOM.202104005.

1Introduction

Future wireless networks are expected to provide high- speed and ultra-reliable communication[ 1 –6]for a num? ber of emerging wireless applications， such as the milli? meterwave（mmWave）[7] ，low-earth-orbitsatellites （LEOSs）[8] ， high-speed trains[9] ， and unmanned aerial vehicles （UAVs）[ 10]. In practice， one of the challenges for these systems is that wireless communication for high mobility is always ac ? companied by severe Doppler spread.While orthogonal fre ? quencydivisionmultiplexing（OFDM）currentlydeployedin Long Term Evolution （LTE） and 5G systems can achieve high spectral efficiency for time-varying frequency selective chan ? nels，itsuffers fromheavyperformancedegradationinhigh Doppler conditions due to severe Doppler spread. Therefore，newmodulationandsignalprocessingtechniquesthataremore robust to time-varying channels are required to meet the challenging requirements of high mobility communications.

Recently， the orthogonal time frequency space （OTFS） mod? ulation proposed in Refs. [ 11 – 12] has drawn a lot of attention due to its advantages over OFDM in high Doppler channels where each transmitted data symbol can exploit the channels delay and Doppler variations. OTFS is a two-dimensional （2D） modulation scheme， where the information symbols are modu ? lated into the 2D delay-Doppler （DD） domain rather than into the time-frequency （TF） domain as the classic OFDM modula? tion. Thus， the time-varying channel in the TF domain can be transformed into a 2D quasi-time-invariant channel in the DD domain， where attractive properties， such as separability， sta? bility， and compactness， can be exploited[ 12]. With the applica? tion of inverse symplectic finite Fourier transform （ISFFT） in OTFS， symbols in the DD domain can be transformed to the TF domain， in which each symbol in the DD domain spans the whole bandwidth and time duration of the transmission frame. Therefore， the OTFS modulation can offer the potential of exploiting the full diversity. However， modulating information symbols into the DD domain makes conventional receiver technologies cannot be applied directly， which in turn re? quires advanced channel coding， signal estimation and detec? tion methods for OTFS to achieve the potential full diversity and reliable error performance.

Accurately estimating the channel parameters in OTFS sys ? tems is a challenging but vital requirement for reliable detec ? tion. A simple channel estimation algorithm for OTFS is to em ? bed the pilot signals in the DD domain[13 – 14] to obtain DD channel responses of each transmitting path. This method has low complexity and signal overhead. But it suffers from perfor? mance degradation for channels with fractional Doppler. In Ref. [15]， the authors proposed a channel estimation method based on pseudo-noise （PN） to estimate the Doppler frequency of each transmitting path， which has high computational com ? plexity to estimate the fractional Doppler. Using deep learning algorithms to assist in the channel estimation has recently been proposed as a potential technology to get more accurate estimation parameters. For example， in Ref. [16]， two deep learning assisted algorithms were proposed to facilitate chan ? nel estimation for massive machine-type communication， which could potentially be applied to OTFS channel estima? tion to deal with fractional Doppler.

Many existing studies focus on signal detection for OTFS modulation. In Ref. [17]， the authors gave an iterative detec ? tion based on the message passing （MP） algorithm. However， the MP detection treats the interference from other informa? tion symbols as a Gaussian variable to reduce the detection complexity， which may fail to converge and result in perfor? mance degradation. To solve this problem， the authors in Refs. [18 – 19] explored the approximate message passing algo ? rithm. They proposed a detection algorithm by covariance pro ? cessing to obtain better BER performance in Ref. [18] and a convergence guaranteed receiver based on the variational Bayes framework in Ref. [19]. An OTFS detection approach based on approximate message passing （AMP） with a unitary AMP transformation （UAMP） was developed in Ref. [20]， which enjoys the structure of the channel matrix and allows ef? ficient implementation. Inspired by the connection between the orthogonality and message passing， the authors in Ref. [21] proposed a novel cross domain iterative detection algo ? rithm for OTFS modulation， where the extrinsic information is passed between the time domain and DD domain via the corre ? sponding unitary transformations. This algorithm has low com ? putational complexity and can achieve almost the same error performance as the maximum-likelihood sequence detection even in the presence of fractional Doppler shifts.

Most papers focus on the basic principle or key algorithms for the estimation or detection of OTFS systems， while little attention has been paid to coded OTFS systems. The authors in Ref. [22] an? alyzed coded OTFS performance， but they only considered classi? cal codes， ignoring modern coding technologies like LDPC， Turbo，or Polar codes for OTFS systems， which are important to achieve high reliability and channel capacity. Therefore， we are going to evaluate modern LDPC coded OTFS performance in this work.

In this paper， we first provide a brief overview of the funda? mental concepts of OTFS. Then we consider 5G LDPC codes for OTFS systems over high mobility channels. As OTFS can be built on the top of the OFDM frame structure， we also con? sider the 5G OFDM frame structure in our evaluations. The ef? fect of channel estimation errors on the LDPC coded OTFS system performance is evaluated as well. The rest of this arti? cle is organized as follows. Section 2 introduces related knowl? edge of OTFS including the basic conceptions， system model， OTFS modulation/demodulation， and the channel input-output relationship. In Section 3， we provide an overview of the error performance analysis for OTFS systems. In Section 4， we pres? ent the details of the performance evaluation of OTFS systems with 5G LDPC codes. Section 5 concludes the paper.

2 OTFS System

In this section， we first present an OTFS system model. Then， we briefly review the input-output relations for the OT? FS systems[11 – 12， 17].

2.1 OTFS System Model

An OTFS modulation is a 2D modulation， which modulates information in the DD domain before transforming signals to TF and time domains. The time-frequency signal plane is dis? cretized to a grid by sampling time and frequency axes at inter? vals T （s） and Δf （Hz）， respectively， i.e.，

For some integers N，M> 0， where T = 1/Δf， N and M are the numbers of time and frequency grids， respectively. There? fore， we can obtain that an OTFS frame is transmitted with du ? ration Tframe = NT and occupies a bandwidth Bframe = MΔf.

Accordingly， let us define the delay-Doppler plane as，

whereandrepresent the quantization steps of the de ?lay and Doppler frequency， respectively. They are also calleddelay and Doppler resolution.

Let a set A = {a1，...，aQ} denote the constellation alphabetwithsizedQandasetof NMmodulatedsymbols x0，...，xNM - 1 ∈to be transmitted by OTFS. The OTFS modula? tion firstly maps symbols x0，...，xNM - 1to DD plane Γ. Here， we use x k，l，k = 0，...，N - 1，l = 0，...，M - 1 to represent the base? band modulated symbols to be transmitted in DD plane， where kdenotes Doppler index and l denotes delay index.

Then， the signal in DD plane Γ is transformed to TF planeΛ ， whereby the symbols in DD domain x k，l，k = 0，...，N - 1，l = 0，...，M - 1 are mapped into TF domain X n，m，n = 0，...，N - 1，m = 0，...，M - 1， where ndenotes the time index and m denotes the frequency index. Such mapping of x k，lto X n，mcan be realized by ISFFT[11].

Next， a multi-carrier modulator， such as OFDM， is used totransform the samples X n，mat each time slot to a continuous time waveform s twith a pulse shaping gtxtas the transmitted pulse. Such a transformation can be realized by discrete Heisenberg transform[11]，

The cross-ambiguity function between the transmitted pulse shaping gtxtand the received pulse shaping grxtis given by[11]

Here， we only discuss OTFS modulation/demodulation prin ? ciples without channel effects. Let r tdenote the receivedsignal. At the receiver， a match filter is applied to compute thecross-ambiguity function Agrx ，rtt，fas

BysamplingY t，fasY n，m= Y t，f|t = nT，f = mΔf， Y n，m，n = 0，...，N - 1，m = 0，...，M - 1 can be obtained.

Then， by transforming TF plane Λ signal Y n，mback to DD plane Γ signal y [ k，l ] with symplectic finite Fourier trans ? form （SFFT）， symbols in DD domain can be recovered as[11]

2.2 Input-Output Relations of OTFS Signal

In this subsection， we present the channel model and the in? put-output relations between the transmitter and the receiv ? er[17]. We assume that only a few reflectors are moving within one OTFS frame duration in practice， and only a small number of channel taps are associated with Doppler shift. Therefore， the channel response of the DD domain is sparse comparedwith the whole DD plane. Considering a channel with P inde? pendent distinguishable paths， the channel response can be represented as

where hi， τiand ν idenote the channel coefficient， delay and Doppler shift associated with the i-th path， respectively. With? out considering additive white Gaussian noise （AWGN）， the relation of s tand r tis given by[17]

The relation ofXn，mand Y n，min TF domain is given by[17]

in which

The second term in Eq. （10） is the samples X n，m′at different frequencies m ′≠ m， which can be seen as the interference to the current sample X n，min the same time slot n.On the other hand， the third term in Eq. （10） accumulates theinterference from the samples X n - 1，m′in the previoustime slot n - 1. Hence， we call the second and third terms asthe inter-carrier interference （ICI） and inter-symbol interference （ISI）， respectively. It is clear that since the delay spreadand Doppler spread， there are severe ICI and ISI in the TFdomain.

The relation of x k，land y [ k，l ] in DD domain is given by[17]

and βi-q - κνi- 1.

In Eq. （12）， hidenotes the channel coefficient and Niis the number of neighboring transmitted signals in the Doppler do ? main， where 0 < Ni? N. The lτiand k νiare the delay and Doppler indices corresponding to the i-th path， respectively， and we have

Note that the term -<κνi≤denotes the fractional Doppler shifts which correspond to the fractional shifts from the nearest Doppler indices[17]. For wideband systems， the typical value of the sampling time in the delay domain is usually suffi ? ciently small. Therefore， the impact of fractional delays in typi? cal wide-band systems can be neglected[24]. Here， we havekmax≤ ki+ κ i≤kmax， where kmax is the maximum Doppler in? dex satisfying kmax = NTνmax[17].

From Eq. （12）， we can see that the received signal y [ k，l ] isa linear combination of S =2Ni+ 1 transmitted signals. The signal corresponding to q = 0， x[ k - k νi ] N ， [ l - lτi ] M contributes the most， and all the other 2Nisignals can be seen as interference. Such interference is due to the transmitted signals neighboring x[ k - k νi ] N ， [ l - lτi ] Min the Doppler domain caused by fractional Doppler and we refer to this interfer? ence as inter-Doppler interference （IDI）. In the TF domain， the delay spread and Doppler spread cause severe ICI and ISI as the second and the third terms in Eq. （10）， which are hardto distinguish in the received signals. However， in the DD do? main， they mainly affect the phase shifts as e j2π（ ） l - lτi M （ ） kνi + κνi N .

Now， let us represent the DD domain input-output relation inavectorform.Letx ?vecX∈MN and y ?vecY∈MNdenote the vector forms of the transmitted symbols X and the received symbols Y in DD domain， respec? tively. According to Eq. （9）， we have

where w is the corresponding noise vector and Heff of size MN × MN is the effective channel matrix in the DD domain. Assuming that both gtx （t） and grx （t） are rectangular pulses， with a reduced CP frame format， the effective channel matrix Heff is given by[23]

where Π is the permutation matrix （forward cyclic shift）， i.e.，

and Δ is a diagonal matrix，

3 Error Performance Analysis

In this section， we discuss the error performance of the un ? coded and coded OTFS systems， respectively. We assume that ideal channel state information （CSI） is available at the receiver.

3.1 Uncoded OTFS System Performance

According to Ref. [26]， Eq. （14） can be rewritten as

where Φ τ，νxis referred to as the equivalent code-word ma? trix and it is a concatenated matrix of size MN × P construct? ed by the column vector Ξi x， i.e.，

and Ξi is given by

In Eq. （18）， h is the channel coefficient vector of size P × 1，i.e.， h = h 1 ，h2 ，...，hPT， where the elements in h are assumed tobe independently and identically distributed complex Gaussian random variables. Besides， we assume a uniform power delay and Doppler profile of the channel so that the channel coeffi ? cient hihas mean μ and variance 1/2 P per real dimension for 1 ≤i≤ P and is independent from the delay and Doppler indi? ces[25]. In particular， we note that if μ = 0， hifollows the Ray? leigh distribution， which will be considered as a special case in the error performance analysis and code design.

Based on Eq. （18）， for a given channel realization， we de? fine the conditional Euclidean distance d h（2），τ，νx，x′between a pair of code-words x and x′x ≠ x′as

where e = x - x′ is the corresponding code-word difference（error） sequence and Ωτ，νe= Φ τ，νeH Φ τ，νeis referred to as the code-word difference matrix. Here we have

Note that the code-word difference matrix Ω （e） is positivesemidefinite Hermitian with a rank r， where r ≤ P. Let us denote by {v1，v2，...，vP} the eigenvectors of Ω （ e） and {λ1，λ2，...，λP} the corresponding nonnegative real eigenvalues sorted in the descending order， where λi> 0 for 1 ≤ i ≤ r and λi = 0 for r + 1 ≤ i ≤ P. Thus， the conditional pairwise-error probability （PEP） [27–28] is upper-bounded by

where Es is the average symbol energy and h ? i = hi ? vi ， for 1 ≤ i ≤ r. It can be shown that {h ? 1，h ? 2，...，h ? r} are independent com? plex Gaussian random variables with mean μh ? i = E[ h ] ? vi and 1/2 P variance per real dimension. It has been defined in the previous work[26， 29–30] that the rank of Ω （ e） is the diversity gain of the uncoded OTFS system. Specifically， it has been shown in Ref. [30] that the diversity gain of uncoded OTFS modulation systems can be one but the full diversity can be obtained by suitable precoding schemes. Furthermore， Ref. [26] has shown that the full diversity can be achieved almost surely for the case of P = 2 when the frame size is sufficiently large， even for un? coded OTFS modulation systems.

3.2 Coded OTFS System Performance

Based on the previous analysis， Ref. [22] gave the error per? formance of the coded OTFS systems. With the assumptions of the wide-sense stationary-uncorrelated scattering （WSSUS） chan? nel and Rayleigh fading， Eq. （23） can be further simplified as

and it is approximated by

Based on Eq. （25）， we note that the unconditional PEP forOTFS modulation only depends on d E（2） e， the rank of Ω e，and the number of independent resolvable paths P， and is independent of the specific distribution of delay and Doppler indi ? ces. The power of the signal-to noise ratio （SNR） is referred to as the diversity gain， and the term d E（2） e/P is referred to as the coding gain， which characterizes the approximate improvement of coded OTFS systems over the uncoded counterpart with the same diversity gain， i.e.， the same exponent -r[27]. Considering the total diversity， there exists a fundamental trade-off between the diversity gain and the coding gain. Based on the previous work[26， 29 –30]， we can notice that the diversity gain depends on the number of independent resolvable paths P. When P （the rank of Ω e） is small， the diversity gain is small. It is crucial for OTFS systems that using an optimized channel code can greatly improve the error performance. For a large value of P and a reasonably high SNR， the unconditional PEP of OTFS systems can be approximately upper-bounded[22] by

Form Eq. （26）， we can notice that the channel with a large number of diversity paths approaches an AWGN model[35]， which indicates that it is reasonable to use coding gain approx ? imation for AWGN channels for evaluating the coding gain of OTFS systems with a large P. A preliminary guideline for the code design of the OTFS systems is to maximize the minimum value of d E（2） eamong all pairs of code-words of the code. Note that， the error performance of coded OTFS systems still de ? pends on the channel parameters， which is widely observed in the system design for fading channels[36].

4 Evaluation of Coded OTFS with 5G LD? PC Codes

In this section， we provide the performance evaluation of the coded OTFS with 5G LDPC codes. Without loss of general? ity， we consider the sum-product algorithm （SPA）[31] or mes? sage passing algorithm for detection， where the details can be found in Refs. [17， 32]. In specific， we consider 5 MHz chan? nelbandwidth1 and 15 kHz sub-carrier for a time slot. Fol? lowed by a 5G frame structure， 5 MHz channel bandwidth con? tains 4.5 MHz efficient bandwidth with 300 sub-carriers and a time slot contains 14 OFDM symbols[33]， where N = 14 and M = 300. In all simulations， we consider the Rayleigh fading case. If not otherwise specified， we only consider the integer delay and Doppler case and set the maximum delay index as lmax = 4 and the maximum Doppler index as kmax = 4. For each channel realization， we randomly select the delay and Doppler indices according to the uniform distribution， so thatwe have -kmax≤ k νi≤kmax and 0 ≤lτi≤lmax. We first present the performance for uncoded OTFS systems with SPA detection. Then， we discuss the performance of codedOTFS sys ? tems with 5G LDPC codes.

4.1 Uncoded OTFS System via SPA Detection

In Fig. 1， it shows the receiver of the uncoded OTFS system via the iterative SPA detection. The SPA detector needs to it? erate the soft information between the detector and the demod ? ulator.

The block-error-rate （BLER） performance of the uncoded OT? FS systems with P = 4 is shown in Fig. 2. Here， we compare theBLERperformanceunderdifferentiterations.InFig.2， “ite × n ”means n iterations in the iterative detections. It can be observed that SPA detection can achieve better performance with iterations. For BPSK modulation， increasing the iteration numberovertwodoesnotprovidemuchimprovement.The same phenomenon for QPSK modulation is observed for over four iterations. We notice that there is a trade-off between the numbers of iterations and BLER performance. For higher order modulation， we need more iterations for better performance.

4.2 Coded OTFS System via SPA Detection

Here，weprovideperformanceevaluationof codedOTFS based on 5G LDPC codes[34]. The decoder uses the sum-prod? uctdecodingalgorithm witha floating value，and themaxi? mum decoding iteration number is 50. We show the coded OT? FSperformanceforvariousmodulationanddifferentcode lengths.

Fig. 3 shows the receiver of the coded OTFS system via iter? ative SPA detection. We firstly only consider iterative detec ? tion in the coded OTFS system. The received symbols after the OTFSdemodulation willbeputintotheSPAdetector. The SPA detector gives soft bit messages corresponding to the re ? ceived symbols to the demodulator. Soft bit messages after de ? modulation will back to the SPA detector via the detection iter? ation loop. After several iterations， soft bit messages will be sent to the decoder for decoding. Compared with Fig. 1， Fig. 3 adds the function of coding and decoding.

Fig. 4 shows the BLER performance of the OTFS systems with P = 4. We list the parameters of the codes we use in Ta? ble 1. Here， K is the information bit length and R is the code rate. We can observe that for the BPSK modulation， increasing the iteration number does not provide much improvement. For the QPSK modulation， two iterations can obtain better perfor? mance.For16QAMmodulation，fouriterationscanobtain larger performance gain. We notice that there exists an opti ? mal iteration number for different modulations.

Next， we only consider the decoding iteration shown in Fig. 5， where we can notice that the detection iteration loop is replaced by decoding iteration loop. Here， the soft bit messages after de ? modulation will be straightly sent to the decoder. After decod? ing， the soft bit messages will be passed to the SPA detector via the decoding iteration loop.

We compare the BLER performance with iterative detectiononly and with iterative detection-decoding in Figs. 6 and 7. For the BPSK modulation in Fig. 6， the performances are similar.

For the QPSK modulation in Fig. 7， we notice that the per? formance of the iterative detection is better than that of the it? erative detection-decoding.

Here， conducting two and four iterative detections outper? forms the iterative decoding at BLER 10- 2 ， with about 0.8 dB and 0.6 dB more gain， respectively. We can notice here that conducting iterative detections can improve the performance more than only iterating between decoding and detection.

Now， we consider the hybrid iterative decoding and detec ? tion. In Fig. 8， we have two iteration links. For detection iter? ation loop （we call it the inner iterative layer）， we do SPA de? tection and demodulation as shown in Fig. 3. For decoding it? eration loop （we call it the outer iterative layer）， we do decod? ing and SPA detection as shown in Fig. 5. After several inner iterations， the demodulator sends the soft bit messages to the decoder for decoding， and after decoding， the soft bit messages will be sent back to the SPA detector via the outer itera? tive layer. Here， “ite 2 × 2”means doing two decoding itera? tions and in each decoding iteration there are two detection iterations. In Fig. 9， we can observe that for the BPSK modu ? lation， doing hybrid iterative decoding and detection cannot obtainmuchgain.FortheQPSKmodulation，theperfor? mance of “ite 2 × 2”hybrid iteration can approach the per? formance of doing 8 iterative detections. For 16QAM modula? tion， the performance of“ite 2 × 2”hybrid iterative decoding and detection is similar to the performance of doing 8 itera? tive detections.

Consequently， we have some observations for coded OTFS systems as follows. For the iterative SPA detection， the opti? mal iterative numbers depend on the modulation order. Higher order modulation needs a higher number of iterations. Only it? erating between decoding and detection does not improve the performancemuch.Hybriditerativedecodinganddetection canachievebetterperformance，especiallyforhigherordermodulations.

4.3 Effect of Channel Estimation on Coded OTFS System

Furthermore， we consider the OTFS performance with the channel estimation， where the DD domain channel estimation is employed[ 14]. In Fig. 10， we compare the performances of un? coded OTFS and coded OTFS system with the channel estima? tion. Here， we only consider the QPSK modulation with SPAdetection，wheretheiterationnumberis2.“Pilot30dB” means the SNR of the pilot is 30 dB. We can see from the fig?ure that the channel estimation has more impact on the perfor?mance of the uncoded system than that of the coded OTFS sys ?tem. For example， the performance gap between “Pilot 30 dB” and “Pilot 40 dB”of the uncoded OTFS system is about 2 dB.For the coded OTFS system， the performance gap between“pi?lot 30 dB”and “pilot 40 dB”is about1 dB. In other words，the channel codes reduce the effect of channel estimation onthe system. However， the performance of the coded OTFS sys ?tem shows an error floor below BLER at10- 2 ， particularly athigh SNR due to the channel estimation error.

5 Conclusions

OTFS has great potential in providing reliable communica? tions for next generation wireless communication systems. In this paper， we gave a brief overview of the fundamental con ? ceptofOTFSandpresentedthecodedOTFSperformance based on 5G LDPC codes. We showed the BLER performance of uncoded OTFS system and coded OTFS system over high mobility channels， and discussed three different coded OTFS schemes with iterative detection and decoding. We also evalu? ated the impact of channel estimation on the performance of coded OTFS systems.

References

[ 1] ZHOU Y Q， LIU L， WANG L， et al. Service-aware 6G： an intelligent and open network based on the convergence of communication， computing and caching [J]. Digital communications and networks， 2020， 6（3）： 253 – 260. DOI： 10. 1016/ j.dcan.2020.05.003

[2] ZHOU Y Q， TIAN L， LIU L， et al. Fog computing enabled future mobile commu ? nication networks： a convergence of communication and computing [J]. IEEE com ? munications magazine， 2019， 57（5）： 20 –27. DOI： 10. 1109/MCOM.2019. 1800235

[3] LIU L， ZHOU Y Q， YUAN J H， et al. Economically optimal MS association for multimedia content delivery in cache-enabled heterogeneous cloud radio access networks [J]. IEEE journal on selected areas in communications， 2019， 37（7）： 1584 – 1593. DOI： 10. 1109/JSAC.2019.2916280

[4] LIU L， ZHOU Y Q， ZHUANG W H， et al. Tractable coverage analysis for hexag? onalmacrocell-basedheterogeneousUDNswithadaptiveinterference-aware CoMP [J]. IEEE transactions on wireless communications， 2019，18（ 1）： 503 – 517. DOI： 10. 1109/TWC.2018.2882434

[5] LIU L， ZHOU Y Q， GARCIA V， et al. Load aware joint CoMP clustering and in ? ter-cell resource scheduling in heterogeneous ultra-dense cellular networks [J]. IEEEtransactionsonvehicular technology，2018，67（3）：2741– 2755.DOI：10. 1109/TVT.2017.2773640

[6] GARCIA V， ZHOU Y Q， SHI J L. Coordinated multipoint transmission in dense cellular networks with user-centric adaptive clustering [J]. IEEE transactions on wirelesscommunications，2014，13（8）：4297–4308.DOI：10. 1109/TWC.2014.2316500

[7] JAMEEL F， WYNE S， NAWAZ S J， et al. Propagation channels for mmWaveve ? hicularcommunications：state-of-the-artandfutureresearchdirections[J]. IEEEwirelesscommunications，2019，26（ 1）：144–150.DOI：10. 1109/MWC.2018. 1800174

[8] SU Y T， LIU Y Q， ZHOU Y Q， et al. Broadband LEO satellite communications： architectures and key technologies [J]. IEEE wireless communications， 2019， 26（2）： 55 –61. DOI： 10. 1109/MWC.2019. 1800299

[9] NOH G， HUI B， KIM I. High speed train communications in 5G： Design ele? ments to mitigate the impact of very high mobility [J]. IEEE wireless communi? cations， 2020， 27（6）： 98 – 106. DOI： 10. 1109/MWC.001.2000034

[ 10] BAI L， HAN R， LIU J W， et al. Air-to-ground wireless links for high-speed UAVs [J]. IEEE journal on selected areas in communications， 2020， 38（ 12）： 2918 –2930. DOI： 10. 1109/JSAC.2020.3005471

[ 11] HADANI R， RAKIB S， TSATSANIS M， et al. Orthogonal time frequency space modulation[C]//2017 IEEE Wireless Communications and Networking Confer? ence（WCNC）.SanFrancisco，USA：IEEE，2017：1– 6.DOI：10. 1109/WCNC.2017.7925924

[ 12] WEI Z Q， YUAN W J， LI S Y， et al. Orthogonal time-frequency space modula? tion：apromisingnext-generationwaveform[J].IEEEwirelesscommunica? tions， 2021， 28（4）： 136 – 144. DOI： 10. 1109/MWC.001.2000408

[ 13] HEBRON Y， RAKIB S， HADANI R， et al. Channel acquisition using orthogonal time frequency space modulated pilot signals： PCT/US20 17/025 166 [P]. 2016

[ 14] RAVITEJA P， PHAN K T， HONG Y. Embedded pilot-aided channel estima? tion for OTFS in delay – Doppler channels [J]. IEEE transactions on vehicular technology， 2019， 68（5）： 4906 –4917. DOI： 10. 1109/TVT.2019.2906357

[15] MURALI K R， CHOCKALINGAM A. On OTFS modulation for high-DopplerThe Performance of LDPC Coded OTFS Systems over High Mobility ChannelsSpecial Topic ZHANG Chong， XING Wang， YUAN Jinhong， ZHOU Yiqingfading channels [J]. 2018 information theory and applications workshop （ITA）， 2018： 1 – 10. DOI： 10. 1109/ITA.2018.8503182

[ 16] LIU B， WEI Z Q， YUAN W J， et al. Channel estimation and user identification with deep learning for massive machine-type communications [J]. IEEE trans? actions on vehicular technology， 2021， 70（ 10）： 10709 – 10722. DOI： 10. 1109/TVT.2021.3111081

[ 17] RAVITEJA P， PHAN K T， HONG Y， et al. Interference cancellation and itera? tive detection for orthogonal time frequency space modulation [J]. IEEE trans? actionsonwirelesscommunications，2018，17（ 10）：6501– 6515.DOI：10. 1109/TWC.2018.2860011

[ 18] LI L J， LIANG Y， FAN P Z， et al. Low complexity detection algorithms for OTFS under rapidly time-varying channel [C]//2019 IEEE 89th Vehicular Technology Conference（VTC2019-Spring）.KualaLumpur，Malaysia： IEEE， 2019：1– 5. DOI： 10. 1109/VTCSpring.2019.8746420

[ 19] YUAN W J， WEI Z Q， YUAN J H， et al. A simple variational Bayes detector for orthogonal time frequency space （OTFS） modulation [J]. IEEE transactions onvehiculartechnology，2020，69（7）：7976–7980.DOI：10. 1109/TVT.2020.2991443

[20] YUAN Z D， LIU F， YUAN W J， et al. Iterative detection for orthogonal time frequencyspacemodulationwithunitaryapproximatemessagepassing[J]. IEEE transactions on wireless communications， 7173， （99）：1. DOI：10. 1109/TWC.2021.3097173

[21] LI S Y， YUAN W J， WEI Z Q， et al. Cross domain iterative detection for or? thogonal time frequency space modulation [J]. IEEE transactions on wireless communications， 0125， （99）： 1. DOI： 10. 1109/TWC.2021.3110125

[22] LI S Y， YUAN J H， YUAN W J， et al. Performance analysis of coded OTFS systems over high-mobility channels [J]. IEEE transactions on wireless commu ? nications， 2021， 20（9）： 6033 –6048. DOI： 10. 1109/TWC.2021.3071493

[23] RAVITEJA P，HONG Y， VITERBO E， et al. Practical pulse-shaping wave? forms for reduced-cyclic-prefix OTFS [J]. IEEE transactions on vehicular tech ? nology， 2019， 68（ 1）： 957 –961. DOI： 10. 1109/TVT.2018.2878891

[24] TSE D， VISWANATH P. Fundamentals of wireless communication [M]. Cam? bridge： Cambridge University Press， 2005. DOI： 10. 1017/cbo9780511807213

[25] MOLISCH A F. Wireless communications [M]. Hoboken， USA： John Wiley & Sons， 2012

[26] RAVITEJA P， HONG Y， VITERBO E， et al. Effective diversity of OTFS modu ? lation [J]. IEEE wireless communications letters， 2020， 9（2）： 249 – 253. DOI：10. 1109/LWC.2019.2951758

[27] TAROKH V， SESHADRI N， CALDERBANK A R. Space-time codes for high data rate wireless communication： Performance criterion and code construction [J].IEEE transactionsoninformation theory，1998， 44（2）： 744– 765.DOI： 10. 1109/ 18.661517

[28] VUCETIC B， YUAN J H. Space-time coding [M]. Chichester， UK： John Wiley & Sons， Ltd， 2003. DOI： 10. 1002/047001413x

[29] BIGLIERI E， RAVITEJA P， HONG Y. Error performance of orthogonal time frequencyspace（OTFS）modulation[C]//IEEEInternationalConferenceon CommunicationsWorkshops.Shanghai，China：IEEE，2019：1– 6.DOI：10. 1109/ICCW.2019.8756831

[30] SURABHI G D， AUGUSTINE R M， CHOCKALINGAM A. On the diversity of uncodedOTFS modulation in doubly-dispersive channels[J]. IEEE transac? tions on wireless communications， 2019，18（6）： 3049 – 3063. DOI：10. 1109/TWC.2019.2909205

[31] KSCHISCHANG F R， FREY B J， LOELIGER H A. Factor graphs and the sum- product algorithm[J]. IEEE transactions on information theory， 2001， 47（2）： 498 –519. DOI： 10. 1109/ 18.910572

[32] LI S Y， YUAN W J， WEI Z Q， et al. Hybrid MAP and PIC detection for OTFS modulation [J]. IEEE transactions on vehicular technology， 2021， 70（7）： 7193 – 7198. DOI： 10. 1109/tvt.2021.3083181

[33] 3GPP. Technical specification group radio access network： 3GPP TS 38.214 V15.0.0 [S]. 2017

[34] 3GPP. Technical specification group radio access network： 3GPP TS 38.212 V15.0.0 [S]. 2017

[35] VENTURA-TRAVESET J， CAIRE G， BIGLIERI E， et al. Impact of diversity reception on fading channels with coded modulation. Part I： coherent detection [J].IEEEtransactionsoncommunications，1997，45（5）： 563– 572.DOI： 10. 1109/26.592556

[36] BIGLIERI E， PROAKIS J， SHAMAI S. Fading channels： Informationtheoretic andcommunicationsaspects[J].IEEEtransactionsoninformationtheory， 1998， 44（6）：2619 –2692

Biographies

ZHANG Chong （zhangchong@ict.ac.cn） received the B.S. degree in electron? ic&informationengineering fromHuaqiaoUniversity ，Chinain2017，and the M. S. degree in electronic and communication engineering from University of Chinese Academy of Science， China in 2020. He is currently pursuing the Ph. D. degree in the Institute of Computing Technology ， Chinese Academy of Science，China.Hisresearchinterestsincludechannelcoding，modulation， and wireless communications.

XING Wang received the B.S. degree from University of Science and Technology Beijing， China in 2019. He is currently pursuing the Ph. D. degree at the Wireless Communication Research Center， Institute of Computing Technology， Chinese Academy of Sciences. His current research interests include orthogonal time frequency space modulation and the convergence of communication ， computation， and caching.

YUAN Jinhong received the B. E. and Ph. D. degrees in electronics engineer ing from the Beijing Institute of Technology， China， in 1991 and 1997， respectively. In 2000， he joined the School of Electrical Engineering and Telecommunications， University of New South Wales， Australia， where he is currently a professor and Head of Telecommunication Group with the School. He has pub lished two books， five book chapters， over 300 papers in telecommunications journals and conference proceedings， and 50 industrial reports. He is a co-in ventor of one patent on MIMO systems and two patents on low-density-parity check codes. He has co-authored four Best Paper Awards and one Best Poster Award. He is an IEEE Fellow and currently serving as an Associate Editor for the IEEETransactions onWireless Communications.HeservedastheIEEE NSW Chapter Chair of Joint Communications/Signal Processions/Ocean Engineering Chapter during 2011-2014 and served as an Associate Editor for the IEEE Transactions on Communications during 2012-2017. His current research interestsincludeerror controlcodingandinformation theory，communication theory， and wireless communications.

ZHOU Yiqing received the B.S. degree in communication and information en ? gineering and the M.S. degree in signal and information processing from South ? east University， China in 1997 and 2000， respectively， and the Ph. D. degree in electrical and electronic engineering from The University of Hong Kong ， China in2004.Sheiscurrentlyaprofessor withtheWirelessCommunicationRe ? search Center， Chinese Academy of Sciences. She has published over 150 arti? cles and four books/bookchapters.She received the best paper awards from WCSP2019，IEEEICC2018，ISCIT2016，PIMRC2015，ICCS2014，and WCNC2013. She also received the 2014 Top 15 Editor Award from IEEE TVT and the 2016 – 2017 Top Editors of ETT. She is also the TPC Co-Chair of Chi? naCom2012， an Executive Co-Chair of IEEE ICC2019， a Symposia Co-Chair of ICC2015， a Symposium Co-Chair of GLOBECOM2016 and ICC2014， a Tutorial Co-Chair of ICCC2014 and WCNC2013， and the Workshop Co-Chair of Smart? GridComm2012 and GlobeCom2011. She is also the Associate/Guest Editor of some renowned journals.