Abstract
Most existing approaches to coexisting communication/radar systems assume that the radar and communication systems are coordinated, i.e., they share information, such as relative position, transmitted waveforms and channel state. In this paper, we consider an uncoordinated scenario where a communication receiver is to operate in the presence of a number of radars, of which only a subset may be active, which poses the problem of estimating the active waveforms and the relevant parameters thereof, so as to cancel them prior to demodulation. Two algorithms are proposed for such a joint waveform estimation/data demodulation problem, both exploiting sparsity of a proper representation of the interference and of the vector containing the errors of the data block, so as to implement an iterative joint interference removal/data demodulation process. The former algorithm is based on classical ongrid compressed sensing (CS), while the latter forces an atomic norm (AN) constraint: in both cases the radar parameters and the communication demodulation errors can be estimated by solving a convex problem. We also propose a way to improve the efficiency of the ANbased algorithm. The performance of these algorithms are demonstrated through extensive simulations, taking into account a variety of conditions concerning both the interferers and the respective channel states.
Adaptive Interference Removal for Uncoordinated Radar/Communication Coexistence \authorLe Zheng, \emphMember, \emphIEEE, Marco Lops, \emphSenior Member and Xiaodong Wang, \emphFellow, \emphIEEE \thanksLe Zheng and Xiaodong Wang are with Electrical Engineering Department, Columbia University, New York, USA, 10027, email: le.zheng.cn@gmail.com, wangx@ee.columbia.edu.
Marco Lops is with the DIEI, Universita degli Studi di Cassino e del Lazio Meridionale, Cassino 03043, Italy (email: lops@unicas.it, e.grossi@unicas.it).
Radar/communication coexistance, atomic norm, compressed sensing, offgrid, sparsity.
\thesection Introduction
The ever increasing demand for spectrum and the consequent shortage of available bandwidths pave the way to communication/radar coexisting system architectures [chiriyath2016inner, griffiths2014t09, griffiths2015radar, hessar2016spectrum]. This inevitably produces intersystem interference that degrades the performance of both systems: in particular, the radar transmit power may be large enough to significantly degrade the performance of the communication system. Techniques such as interference mitigation [deng2013interference], precoding or spatial separation [babaei2013practical, aubry2014cognitive, khawar2014spectrum], waveform design [aubry2014radar, huang2015radar, zheng2017joint, sodagari2012projection] allow both radar and communications to share the spectrum and coexist: for example, in [babaei2013practical, sodagari2012projection, khawar2014spectrum], the radar interference is eliminated by forcing the radar waveforms to live in the null space of the interference channel between the radar transmitters and the communication receiver. Motivated by the cooperative methods in cognitive radio networks, later works [lioptimum, li2016mimo, turlapaty2014joint] exploit some prior knowledge to jointly design the radar waveform and communication codebook by minimizing a measure of the mutual interference under certain constraints. Many existing approaches assume that the radar and communication systems are aware of the existence of each other, and share information: for example, ad hoc design of radar waveforms or beamformers is proposed in [deng2013interference, aubry2014radar, babaei2013practical, aubry2014cognitive, huang2015radar] to reduce the mutual interference, while [lioptimum, li2016mimo, turlapaty2014joint, zheng2017joint, sodagari2012projection] rely on making channel information available to the communication and radar through the transmission of pilot training. Otherwise stated, these approaches rely on a centralized architecture, namely on a strict coordination between the active players in order to allow coexistence. In [khawar2014spectrum], the radar interference is eliminated by forcing the radar waveforms to live in the null space of the interference channel between the radar transmitters and the communication receiver. The algorithm can be extended to the situation with no cooperation between radar and communication by using a blind null space estimation method [manolakos2012blind]. With adaptively adjusted radar waveforms, difficulties for realtime processing arise and some performance loss may happen to the radar. The situation we refer to in this contribution is one wherein a communication system should share its spectrum with an ensemble of potential interferers, i.e., a set of radar/sensing systems. We assume full bandwidth overlapping of the active systems, but that not necessarily all of the potentially active radars are transmitting: thus, when observed on a conveniently long time interval, this situation is akin to a highly nonstationary environment, wherein the sources of interference may vary over time, and so do the corresponding waveforms, timing and channels. While the interference produced by the (unique) communication system on the active radars may be neglected, due to both the order of magnitude of the powers in play and some specific countermeasure that can be taken (for example, the active radars may use suitable beamforming techniques to get rid of interference from a known location [liu2014joint]), the interference produced onto the communication system may be highly detrimental, and must be dealt with. Unlike the antijamming in uncoordinated wireless networks [xiao2012jamming, xiao2012mac], our communication system can be a single TX/RX pair, so it cannot rely on the collaborative broadcasting scheme. In this uncoordinated scenario, the only information the communication system can rely upon is that the interfering waveforms live in the subspace of a known dictionary, and that they impinge on the communication receiver (RX) with unknown, possibly timevarying delays and coupling coefficients. As a consequence, the communication RX must be made adaptive, in order to accomplish jointly the two tasks of interference estimation/removal and data demodulation. The approach we propose here focuses on guaranteeing the performance of the communication system and relies on the size of the dictionary from which the radar waveforms are picked up: we first show that, adopting a suitable representation domain, the interfering signals hitting the communication RX are sparse. On a parallel track, if iterative demodulation/remodulation algorithms are implemented, the vector containing the demodulation errors of a data block should be itself sparse (and become sparser and sparser as the iterations go), whereby a joint interference removal/data demodulation process can take great advantage of existing algorithms forcing sparsity constraints. Unfortunately, however, the scenario we consider in this paper is not ensured to lend itself to direct application of compressed sensing (CS) theory [foucart2013mathematical], which relies on the fact that signals can be sparsely represented by a finite discrete dictionary [stankovic2013compressive, jokanovic2015reduced, studer2012recovery, yang2014exact]: the presence of relevant continuous parameters, such as delays, could indeed lead to remarkable degradations from model mismatch, should a simple discretization of the parameter space be implemented [chi2011sensitivity]. We thus also explore the applicability of the recently developed mathematical theory of continuous sparse recovery for superresolution [candes2013super, candes2014towards, tang2013compressed] and in particular of AtomicNorm (AN) minimization techniques successfully employed for continuous frequency recovery from incomplete data [tang2013compressed, bhaskar2013atomic], directionofarrival estimation [tan2014direction], interference mitigation [al2016atomic] and line spectral estimation [tang2014robust]. Given the above framework, we thus propose two algorithms for joint waveform estimation and data demodulation in the overlaid radar/communication architecture, the former based on classical ongrid CS, the latter forcing an AN constraint: in both cases the radar parameters and the communication demodulation errors can be estimated by solving a convex problem. We also propose a way to improve the efficiency of the ANbased algorithm. The merits of these algorithms are demonstrated through extensive simulations, taking into account a variety of conditions concerning both the interferers and the respective channel states. The remainder of the paper is organized as follows. In Section II, we present the signal model of the coexisted radar and communication system. In Section III, we develop the proposed CSbased algorithms using both the atomic norm and the norm. In Section IV, an accelerated algorithm for solving the atomic normbased algorithm is proposed. Simulation results are presented in Section V. Section VI, finally, contains concluding remarks.
\thesection Problem Formulation
We consider a situation with one communication system and active radars. Suppose the th radar transmits the coded waveform
(\theequation) 
where is the th code, is the code length, and satisfies the Nyquist criterion with respect to . From now on we assume that such a basic pulse is a Square Root Rased Cosine (SRRC) with excess bandwidth , so that the transmit bandwidth is . We also set with when is odd and when is even. Let be the Discrete Fourier Transform (DFT) of , i.e., with denoting the DFT matrix. In practice, the radar waveforms usually have some characteristics to achieve certain performance, so we assume lives in a lowdimensional subspace of , spanned by the columns of a known matrix with and , i.e., for some unknown such that for . An example of such a situation is waveform diversity, wherein radar systems may vary the transmit waveforms by selecting them in a set  a dictionary  so as to cope with interference, clutter, coexistence with competing systems [hassanien2016dual]. Obviously, if , the number of unknown variables exceeds , and the problem is for sure infeasible: in general, since the radar waveforms may contain a number of unknown parameters, the condition is mathematically necessary, on top of being per se reasonable. \colorblack For simplicity, we assume that there is one path between the radar TX and the communication RX. This assumption is true for narrowband radar systems [li2016mimo] or as the interference is dominated by the direct path between the radar TX and the communication RX. The interference produced by active radars  with possibly unknown  onto the communication RX can be expressed as
(\theequation) 
where and denote the delay and complex coupling coefficient of the th radar, respectively. We assume the communication TX is not moving and its position is known by the radars, so its effect on the radar can be dealt with via beamforming. The communication TX transmits data symbols with . Let where denotes the set of possible values. Defining the data received at the RX side as , we have
(\theequation) 
where , and is the channel matrix. This model subsumes a number of communication systems. For example, in a CodeDivision Multiple Access (CDMA) system [guo2005randomly], the elements in are the symbols transmitted by active users: The columns of are the signatures of the users, and is a diagonal matrix representing the channel gains. Another example is an OFDM system [zhang2015maximum], in which is the IDFT matrix with , is the channel matrix and is the received data in time domain. In practice, is known and the channel matrix can be obtained through the transmission of pilot signals. Let be the DFT of , i.e., . The received communication signal is given by
(\theequation) 
where denotes the overall delay of the communication transmission. As the communication TX and RX are synchronized, we let for the convenience of derivation. Notice also that the full bandwidth overlap scenario considered in this paper makes it reasonable to assume that the communication system occupies the full available bandwidth and uses the same SRRC basic pulse as the radar systems. At the communication RX, the signal contains both the communication signal and the radar interference, i.e.,
(\theequation)  
where is the measurement noise. Projecting onto results in
(\theequation)  
where is the autocorrelation function of , i.e., with denoting inner product, and . The autocorrelation function is considered substantially timelimited in a finite interval, say: This is a common assumption which is justified by the fact that is always vanishingly small for large , and in particular goes to zero as in the considered scenario (see Appendix A). Letting and be the minimum and the maximum delays, respectively, tied to the corresponding minimal and maximum distances of all of the potential radar systems from the receiver, we define in \eqrefeq:truebarr
(\theequation)  
for , where . For simplicity, we define . It is assumed that is a complex Gaussian variable with zero mean and variance , i.e., . In communication receivers, is typically available and can be estimated offline. Suppose that satisfies for . The above condition is rigorously true for small excess bandwith (in particular, ), which includes the relevant case that the communication system employs an OFDM format, corresponding to , while being only approximately true for larger values of [glover2010digital]. Since in an efficient spectrum exploitation context it is mandatory to choose small excess bandwiths, we henceforth assume
(\theequation) 
Plugging \eqrefeq:integ into \eqrefeq:barr, we have
(\theequation)  
where and are the th element of and , respectively. We define . As outlined in the introduction, the communication RX has to remove the radar interference from the measurement with no knowledge of the delays and the waveforms of the active radars, i.e., under uncertainty concerning and . In principle, demodulation may be undertaken simply ignoring the presence of interference, i.e., through the operation , with the decoding function operating on the received signal, which would obviously lead to an uncontrolled symbolerrorrate (SER). The approach we take here instead relies on a joint interferenceestimation symboldemodulation process. In particular, defining as the “estimated communication signal”, the presence of errors in the decision process results into a nonzero difference vector where . Obviously, the th element of is given by
(\theequation)  
where we have defined , with , and the th column of the identity matrix . Once estimates of and are available, the radar interference can be obtained and canceled from the measurements and the symbols redemodulated. Hence, the main problem is to estimate and from the noisy measurements . Notice that contains both the radar interference and the residual of communication signal caused by the misdemodulations. The mixing of both signals causes great difficulties for the estimation, which inspires us to exploit some structural information about the desired solution, and in particular sparsity, as detailed in the next section.
\thesection Proposed Algorithms
Equation \eqrefeq:z highlights that data demodulation and interference mitigation are coupled, in the sense that they should be accomplished jointly and that poor performance in estimating either one has detrimental effects on the estimate of the other. In fact, in order to remove the radar interference, we need to estimate the matrix from the observations \eqrefeq:z, but this would obviously require also estimating the error vector , which boils down to correctly demodulating the data block. To this end, we design iterative algorithms, exploiting structural information on the desired solution. In the th iteration, the demodulated symbols are denoted as , and where is the estimate of . \colorblack The following two types of sparsity are exploited in the problem:

The signal is a combination of complex exponentials with unknown modulation , and the number of complex exponentials is much smaller than the dimension of , i.e., .

Ideally, the vector should be an allzero vector. As a consequence, denoting the result of the th iteration, we want to be as small as possible, and in any case we want to force the condition .
\thesubsection Joint Waveform Estimation and Demodulation Based on Ongrid CS Algorithm
In the first iteration, so . According to \eqrefeq:z, jointly estimating , and is a nonlinear problem. However, it can be linerized by using an overcomplete dictionary matrix
(\theequation) 
with denoting sets of uniformly spaced points of the radar delays. Define as the number of columns of where . For sufficiently large , the delay is densely sampled. Let be the sparse vector whose nonzero elements correspond to in \eqrefeq:z. As usual, forcing a constraint onto the norm is impractical, since it results in an NPhard nonconvex optimization problem, and norm regularization is used instead, i.e., and . Define as the set of all possible differences when the two vectors both belong to . We notice that the constraint results in a nonconvex problem, and would cause much difficulty in solving the optimization problem. This constraint is relaxed and the nonlinear joint estimation problem is thus reduced to a linear parameter estimation problem, i.e., the estimation of the linear amplitude vectors and , under a sparsity constraint:
(\theequation)  
where
(\theequation) 
and is given by \eqrefeq:Upsilon.
(\theequation) 
The parameters and are weights determining the sparsity of the reconstruction. In practice, we set . As \eqrefeq:cs is convex, it can be solved with standard convex solvers. The computational complexity of problem solving depends on the dimension of and . Specifically, if \eqrefeq:cs is solved by using the interior point method [boyd2004convex], the complexity in each iteration is . By solving \eqrefeq:cs, we obtain the estimates and , whereby the symbols can be corrected and redemodulated as:
(\theequation) 
Notice that the redemodulation process makes use of the structural information of the communication symbols, i.e., . After the redemodulation, the demodulated symbols belong to the constellation alphabet, whereby . When the estimates of the symbols in the first iteration is accurate and the measurement noise is small, i.e., and are both small, then all the mistakenly demodulated symbols can be corrected by applying \eqrefeq:estb. In many cases, however, the interference from the radars is strong, and contains many demodulation errors. Thus, we need to iterate the joint interference removal/data demodulation process. Specifically, in the th iteration (), and are estimated as
(\theequation)  
Then redemodulation is undertaken by \eqrefeq:estb with replaced by and replaced by . As the iteration goes, some wrong symbols are corrected, and the demodulation error becomes sparser as increases. The proposed algorithm iterates until or the maximum number of iterations is reached. Let be the estimate of when the algorithm terminates. The radar delays can be identified by locating the nonzero entries of . If the solution is either nonzero or has elements larger than a preset threshold, i.e., , then a radar interference exists at delay . Notice that one cannot resolve the inherent scaling ambiguity between each and the corresponding , which is in any case not essential since it is the product the measure of interest for radar interference removal purposes. The estimated time domain radar waveform is then given by
(\theequation) 
The CS algorithm based on minimization (CSL1) is capable of superresolving the spectrum of the sparse signal under certain conditions of the matrices and [eldar2009robust]. For comparison purposes, and to quantify, at the performance assessment stage, the loss induced by the uncertainty on the active radars positions, it is worth exploring also the case that the delays of such active radar systems are known at the communication receiver. In such a simplified scenario, which assumes that training pilot signals are periodically transmitted for channel sensing purposes, a modified version of the proposed algorithm can be easily derived, by using arguments similar to those employed above. Indeed, since the radar delays are known precisely, we let , and \eqrefeq:z can be rewritten as
(\theequation) 
where . The algorithm with known timing and radar delay has the same procedure as that of CSL1. Notice that is not sparse. In the th iteration, and the demodulation error can be estimated by solving \eqrefeq:cs2 with , removed and replaced by .
\thesubsection Joint Waveform Estimation and Demodulation Based on Offgrid CS Algorithm
As anticipated, \eqrefeq:z reveals that is a linear combination of modulated complex exponentials with arbitrary phases, which do not in general correspond to the point of a discrete grid: the offgrid radar position can lead to mismatches in the model and deteriorate the performance. In this subsection, we use the atomic norm to build a sparse representation which does not suffer from the offgrid problem. We define the atomic norm [yang2016super] associated to as
where denotes the convex hull of the input atom set, and the set of atoms is defined as
(\theequation) 
For future developments, we introduce the following equivalent form of the atomic norm for the atom set [yang2016super]:
(\theequation) 
where is a complex vector whose first entry is real, denotes the Hermitian Toeplitz matrix whose first column is , and is a Hermitian matrix. Based on \eqrefeq:z, and paralleling the arguments outlined in the previous subsection, the th iteration achieves estimates, and say, of and by processing and solving the optimization problem:
where and are the weight factors. In practice, we set and . In the light of \eqrefeq:atomic, the above can be transformed into the following SemiDefinite Programming (SDP):
(\theequation)  
where denotes the Toeplitz matrix whose first column is the input vector. The above problem is convex, and can be solved by using a convex solver. The corresponding computational load is per iteration if the interior point method is applied. We name the algorithm based on solving \eqrefeq:atomicnorm as the CS AtomicNorm (CSAN) based algorithm. Similar to the CSL1 algorithm, CSAN iterates \eqrefeq:SDP and \eqrefeq:estb until . From now on, denotes the estimate of when the algorithm terminates. We hasten to underline here that the atomic norm can be exploited to enforce sparsity in the continuous domain without any discretization [tang2013compressed, yang2016super], thus leading to better estimation performance compared to ongrid CS techniques. Atomicnormbased optimization has also been shown to achieve better denoising performance than subspace methods such as MUSIC [bhaskar2013atomic]. Notice that if has low rank, it is amenable to nuclearnormbased estimation. However, the nuclear normbased approach cannot fully take advantage of the signal structure (e.g., in our scenario is a combination of multiple complex exponentials with unknown modulation ), whereby its performance has been shown to be inferior to that of atomic normbased optimization [fernandez2016demixing, chen2014robust]. Solving \eqrefeq:SDP does not directly provide estimates of the delays of the active radars, . Notice however that each row of is a linear combination of several complex exponentials, in that, denoting the th row of , we have . Hence, and can be obtained by MUSIC [naha2015determining] or prony’s method [stoica1997introduction] with as input. Denoting the operation of the MUSIC algorithm, it outputs components with different amplitudes and delays^{1}^{1}1Here is not necessarily the same as because can be zero while for some . For example, the radar has candidate waveforms, and select one for transmission. In such case, is a vector having zero elements and one element with magnitude 1. As a result, the number of output delays should satisfy ., i.e.,
(\theequation) 
for . An association procedure is thus needed to combine the different s and come up with the estimates, say, of the radar signals delays. In practice, the calculation of may not be accurate due to computational errors. Hence, if the estimated delays for different are closer than a small threshold , then they are regarded as generated by the same radar, and the corresponding s are combined. For clarity, we summarize the association process in Algorithm 1, where is the set of the estimated radar delays and waveform parameters. In the algorithm, for , if there exists such that and , then both components belongs to the same radar and the radar delay is updated by the weighted summation of and . Otherwise, an additional radar with the estimated delay and amplitude is added to the set . Notice that, once again, the inherent scaling ambiguity between each and the corresponding . Hence, we only estimate via \eqrefeq:hatg with replaced by . {algorithm}[htbp] Input for , . 1, , . For For If there exists such that and 2,