Group Sparse Precoding for CloudRAN with Multiple User Antennas
Abstract
Cloud radio access network (CRAN) has become a promising network architecture to support the massive data traffic in the next generation cellular networks. In a CRAN, a massive number of lowcost remote antenna ports (RAPs) are connected to a single baseband unit (BBU) pool via highspeed lowlatency fronthaul links, which enables efficient resource allocation and interference management. As the RAPs are geographically distributed, the group sparse beamforming schemes attracts extensive studies, where a subset of RAPs is assigned to be active and a high spectral efficiency can be achieved. However, most studies assumes that each user is equipped with a single antenna. How to design the group sparse precoder for the multiple antenna users remains little understood, as it requires the joint optimization of the mutual coupling transmit and receive beamformers. This paper formulates an optimal joint RAP selection and precoding design problem in a CRAN with multiple antennas at each user. Specifically, we assume a fixed transmit power constraint for each RAP, and investigate the optimal tradeoff between the sum rate and the number of active RAPs. Motivated by the compressive sensing theory, this paper formulates the group sparse precoding problem by inducing the norm as a penalty and then uses the reweighted heuristic to find a solution. By adopting the idea of block diagonalization precoding, the problem can be formulated as a convex optimization, and an efficient algorithm is proposed based on its Lagrangian dual. Simulation results verify that our proposed algorithm can achieve almost the same sum rate as that obtained from exhaustive search.
Keywords:
cloud radio access networksparse beamformingblock diagonalizationgroupsparsityantenna selection∎
1 Introduction
Cloud radio access network (CRAN)CRAN2011 () is a promising and flexible architecture to accommodate the exponential growth of mobile data traffic in the nextgeneration cellular network. In a CRAN, all the baseband signal processing is shifted to a single baseband unit (BBU) poolShi2015 (). In the meantime, the conventional basestations (BSs) are replaced by geographically distributed remote antenna ports (RAPs) with only antenna elements and power amplifiers, which are connected to the BBU pool via highspeed lowlatency fronthaul links by fiberLiu2013a (). The simple structure of RAP enables the probability to deploy ultradense RAPs in a CRAN with low cost.
With highly densed geographically distributed RAPs, significant rate gains can be expected over that with the same amount of colocated antennas in both the singleuser and multiuser casesLiu2014 (); Liu2016 (); Wang2015 (). Due to the huge differences among the distance between the user and the geographically distributed RAPs, it has been shown in Liu2014 () that the capacity in the singleuser case is crucially determined by the access distance from the user to its closest RAP. This motivates us to investigate whether it is possible to achieve a significant proportion of the sum rate by using a subset of the RAPs. In the singleuser case, it is straightforward to avoid distant RAPs transmitting, as they have little contribution to improve the capacity. In the multiuser case, the problem becomes challenging, as the beamformers of all users should be jointly designed.
To tackle this problem, a branch of sparse beamforming technologies are therefore proposed, where the beamforming vectors are designed to be sparse with respect to the total number of transmit antennasZhao2013 (); Mehanna2013 (); Dai2013 (). Motivated by the recent theoretical breakthroughs in compressive sensingZhang2015 (), the sparse beamforming problem is formulated by including the norm of the beamforming vectors as a regularization such that the problem becomes convex. By iteratively updating the weights of the norm, the sparse beamformer that minimizes the total transmit power can be obtained by iteratively solving a secondorderconicprogramming (SOCP)Zhao2013 () or a semidefiniteprogramming (SDP)Mehanna2013 (). The problem can be further simplified to a uplink beamformer design problem via uplinkdownlink dualityDai2013 (). Nevertheless, when each RAP includes multiple antennas, one RAP will be switched off only when all the coefficients in its beamformer are set to be zero. In other word, all antennas at a RAP should be selected or ignored simultaneously, which requires group sparsity instead of individual sparsity as in conventional compressive sensing. Recently, group sparse beamforming problem are proposed which can be formulated by inducing a mixed norm regularization Dai2014a (); Shi2014 (); Shi2015a (). For instance, by inducing the norm as a regularization, the weighted sum rate maximization problem is formulated as a weighted minimum mean square error (WMMSE) minimization and can be solved via a quadraticalconstrainedquadraticprogramming (QCQP)Dai2014a (). Compared to the individual sparse beamforming, the group sparse beamforming further reduces the network power consumption, and the energy efficiency can be improved as well.
So far, most algorithms focus on the situation where each user has a single antennaZhao2013 (); Mehanna2013 (); Dai2013 (); Dai2014a (); Shi2014 (); Shi2015a (). As suggested by the multipleinputmultipleoutput (MIMO) theory, the capacity increases linearly with the minimum number of transmit and receive antennasTelatar1999 (). In a CRAN, it is desirable to employ multiple antennas at each user to exploit the potential multiplexing gains, which, however, further complicates the sparse precoder design. The difficulty originates from the fact that the problem is typically nonconvex, and the transmit and receive beamformers, which are mutual coupling, should be jointly designed. This paper focuses on designing group sparse precoder based on block diagonalization (BD) Spencer2004 (), which has gained widespread popularity thanks to its low complexity and nearcapacity performance when the number of transmit antennas is largeShen2006 (); Shen2007 (); Shim2008 (); Ravindran2008 ().
In this paper, we address the joint problem of RAP selection and joint precoder design in a CRAN with multiple antennas at each user and each RAP. Whereas the problem is typically NPhard, we show that the problem becomes convex by inducing the reweighted norm of a vector that indicates the transmit power at each RAP as a regularization. Based on its Lagrangian dual problem, we propose an iterative algorithm by iteratively updating the weights of the norm to generate sparse solution. Simulation results verify that the proposed algorithm can achieve almost the same sum rate as that achieved from exhaustive search.
The rest of this paper is organized as follows: Section II introduces the system model and formulates the problem. Section III proposes an iterative algorithm to solve the group sparse precoding problem. Simulation results are presented and discussed in Section IV. Section V concludes this paper.
Throughout this paper, italic letters denote scalars, and boldface uppercase and lowercase letters denote matrices and vectors, respectively. The superscripts and denote transpose and conjugate transpose, respectively. denotes the expectation operator. denote the ceiling operator. denotes the norm of vector . and denote the trace and determinant of matrix , respectively. denotes an diagonal matrix with diagonal entries . denotes an identity matrix. and denote matrices with all entries zero and one, respectively. denotes the cardinality of set .
2 System Model and Problem Formulation
Consider a CRAN with a set of remote antenna ports (RAPs), denoted as , and a set of users, denoted as , with and , as shown in Fig. 1. Suppose that each RAP is equipped with antennas, and each user is equipped with antennas. The baseband units (BBUs) are moved to a single BBU pool which are connected to the RAPs via highspeed fronthaul links, such that the BBU pool has access to the perfect channel state information (CSI) between the RAPs and the users, and the signals of all RAPs can be jointly processed. With a high density of geographically distributed RAPs, the access distances from each user to the RAPs varies significantly, and the distant RAPs have little contribution to improve the capacity. This motivates us to find a subset of RAPs that can provide near optimal sum rate performance.
In particular, let denotes the set of active RAPs, with . To utilize the multiplexing gains from the use of multiple user antennas, we assume that . The received signal at user can be then modeled as
(1) 
where and denote the transmit and receive signal vectors, respectively. is the channel gain matrix between the active RAPs and user . denotes the additive noise, which is modeled as a Gaussian random vector with zero mean and covariance . With linear precoding, the transmit signal vector can be expressed as
(2) 
where is the precoding matrix. is the information bearing symbols. It is assumed that Gaussian codebook is used for each user at the transmitter, and therefore . The transmit covariance matrix for user can be then written as . It is easy to verify that . The sum rate can be written from (1) as
(3) 
Note that it is difficult to find the optimal linear precoder that maximizes the sum rate due to the nonconvexity of (3), and the mutual coupling of the transmit and receive beamformers makes it difficult to jointly optimize the beamformersCai2011 (). In this paper, we assume that block diagonalization (BD) Spencer2004 () is adopted, where an interferencefree block channel is obtained by projecting the desired signal to the null space of the channel gain matrices of the other users, such that , or equivalently for all . With BD, the sum rate can be obtained as
(4) 
As the signals come from more than one RAPs, they need to satisfy a set of perRAP power constraint, i.e.,
(5) 
where denotes the perRAP power constraint. is a diagonal matrix, whose diagonal entries is defined as
(6) 
This paper focuses on the tradeoff between the sum rate and the group sparsity. In particular, the problem can be formulated as the following optimization problem:
(7a)  
s.t.  (7b)  
(7c)  
(7d) 
where (7b) is the zeroforcing (ZF) constraint, which ensures that the interuser interference can be completely eliminated at the optimum. is the tradeoff constant, which controls the sparsity of the solution, and thus the number of active RAPs. With , the problem reduces to a BD precoder optimization problem with perRAP power constraint. The group sparsity can be improved by assigning a larger .
Note that the optimization problem (7) needs to jointly determine the subset and design the transmit covariance matrices for users, which is a combinatorial optimization problem and is NPhard. A bruteforce solution to a combinatorial optimization problem like (7) is by exhaustive search. Specifically, we must check all possible combinations of the active RAPs. For each combination, we must search for the optimal that satisfies the constraints (7b7d). In the end, we pick out the combination that maximizes the sum rate. However, the complexity grows exponentially with , which can not be applied to realworld application. Instead, we use the concept of norm to reformulate problem (7). In particular, define as
(8) 
where , with denoting the transmit signal vector from all RAPs in to user . The th to the th entries of are zero if RAP is inactive. It is clear that the th entry is the transmit power of RAP , which is nonzero if and only if RAP . It is easy to verify that . We then have the following lemma:
Theorem 2.1
Problem (7) is equivalent to the following optimization problem:
(9)  
s.t.  
where is the channel gain matrix from all RAPs in to user , .
Proof
Please refer to Appendix A for detailed proof.
Lemma 2.1 indicates that instead of searching over the possible combinations of and then optimizing according to the corresponding channel gain matrices , (7) can be solved based on the channel between the users and all RAP antennas, i.e., . However, the problem (9) is nonconvex due to the existence of the norm, making it difficult to find the global optimal solution.
In compressive sensing theory, the norm is usually replaced by a norm, and sparse solution can be achieved. However, simply substituting by in (9) will not necessarily produce sparse solution in general, as equals the sum power consumption instead of the number of nonzero entries. By replacing by , the transmit power at all RAPs still tend to satisfy the power constraints with equality at the optimum, leading to a nonsparse solution. In this paper, we propose to solve (9) heuristically by iteratively relaxing the norm as a weighted norm. In particular, at the th iteration, the norm is approximated by
(10) 
where , with denoting a small positive constant. (9) can be then reformulated as
maximize  (11)  
s.t. 
where
(12) 
It is clear that problem (11) is convex, and can be solved by standard convex optimization techniques, e.g. interior point method Boyd2004 (), which, however, is typically slow. In fact, by utilizing the structure of BD precoding, the problem can be efficiently solved by its Lagrangian dual.
3 Reweighted Based Algorithm
In this section, the algorithm to solve the group sparse linear precoding problem will be presented. To solve (11), it is desirable to remove the set of ZF constraints in the first place. It has been proved in Zhang2010 () that the optimal solution for BD precoding with perRAP constraint is given by
(13) 
where . is given from the singular value decomposition (SVD) of as
(14) 
where is the last columns of the right singular matrix of . It is easy to verify that for all .
Therefore, by substituting (13) into (11), the problem reduces to
maximize  (15)  
s.t.  
Note that (15) is a convex problem, its Lagrangian dual can be written as
(16) 
where , with denoting the Lagrangian dual variables. The Lagrangian dual function can be given as
(17) 
We can then obtain the Lagrangian dual problem of (15) as
(18) 
Since the problem (15) is convex and satisfies the Slater’s condition, strong duality holds. The respective primal and dual objective values in (15) and (18) must be equal at the global optimum, and the complementary slackness must hold at the optimumBoyd2004 (), i.e.,
(19) 
where and are the optimal primal and dual variables, respectively. As BD requires that the total number of transmit antennas should be equal or greater than the total number of user antennas, the minimum number of active RAPs should be . We can then conclude from (19) that the maximum number of positive ’s is
For fixed , the Lagrangian dual function can be obtained by solving
(20) 
where . By noting that , we have
(21) 
Let . (20) can be written as
(22) 
It is clear that for given , the problem (22) can be solved from uncoupled subproblems:
(23) 
Let us then solve the subproblems (23). By introducing the (reduced) SVD:
(24) 
where and are unitary matrices. , with denoting the th singular value of . The optimal can be then obtained from the standard waterfilling algorithmCover2006 ():
(25) 
where , with
(26) 
where . The optimal for given can be then obtained as
(27) 
With the optimal is achieved for given , we can then find the Lagrangian dual variables by the projected subgradient method. Projected subgradient methods following, e.g., the square summable but not summable step size rules, have been proved to converge to the optimal valuesBertsekas2003 (). In particular, a subgradient of with respect to is . With a step size , the dual variables can be updated as
(28) 
From the complementary slackness in (19), a stopping criterion for updating (28) can be
(29) 
Once the optimal is obtained, the optimal precoding matrices can be achieved by using the fact that as
(30) 
The optimal set and the corresponding precoding matrices can be obtained from . The algorithm is summarized as Algorithm 1.
Initilization: Set iteration counter , Lagrangian dual variable , , and initial ellipsoid .
Repeat:

Update according to (28).

Update the iteration counter and
Stop if , where is a predefined tolerance threshold.
4 Simulation Results
In this section, simulation results are presented to illustrate the results in this paper. We assume that the channel gain matrices from RAP to user are independent over and for all and , and all entries of are independent and identically distributed (iid) complex Gaussian random variables with zero mean and variance . The pathloss model from RAP to the user is
(31) 
and , where is in the unit of kilometer. The largescale fading coefficient from RAP to user can be then obtained as
(32) 
The transmit power constraint at each RAP are assumed to be identical, which is set to be dBm/Hz, , and the noise variance is set to be dBm/Hz.
We consider the case that RAPs with antennas each and users with antennas each. The positions of users and RAPs and the corresponding smallscale fading coefficients are randomly generated. Fig. 2 shows how the number of active RAPs^{1}^{1}1A RAP is said to be active if its transmit power . varies with iterations, where the tradeoff constant is set to be , and . The number of active RAPs corresponds to . Fig. 2 shows that the number of active RAPs decreases with . Specifically, with , the problem reduces to a BD precoder design with perRAP constraint, and all RAPs are active. With , the sparsest solution can be achieved, i.e., . We should mention that the value of that corresponds to the sparsest solution varies with the system configuration.
As Fig. 2 shows, the first several iterations lead to biggest improvement. As iterations go on, there is no further improvement after the 15th iteration. Compared to the full cooperation case, i.e., , as RAPs are switched off, 70% of the power consumption can be saved with, however, limited rate performance loss, which will be illustrated later in this section. Fig. 3 plots the transmit power distribution over all 10 RAPs. From Fig. 3, we can see that as the iterations progress, RAP , and form a serving cluster and transmit with almost the maximum power, while all the other RAPs eventually drop their transmit power to zero after 15 iterations.
Fig. 4 plots how the average sum rate varies with the number of active RAPs. The average sum rate is obtained by averaging over 20 realizations of smallscale fading and 30 realizations of the positions of RAPs and users. The results of exhaustive search is also presented for comparison, which is obtained by searching over all possible combinations of active RAPs, and compute its achievable sum rate by the algorithm given in Zhang2010 (). For the proposed algorithm, we simulate a series of different ’s to get different points along the curve. As we can see from Fig. 4, our proposed algorithm achieves almost the same average sum rate as the exhaustive search, which verifies the optimality of our proposed algorithm. Fig. 4 further plots the average sum rate with a fixed number of instead of uniformly distributed antennas are deployed for comparison, which is denoted as “Fixed ” in Fig. 4. We can clearly see that the proposed algorithm can achieve much better rate performance over that with the same amount of transmit RAP. With 6 selected antennas, for instance, the group sparse precoder reduces the average sum rate for only 3 bit/s/Hz, whereas if only antennas were installed instead of , an additional of 9 bit/s/Hz rate loss can be observed compared to the proposed algorithm. Moreover, the rate gap between the proposed algorithm and that with a fixed number of full cooperative RAPs further increases as the number of selected RAPs decreases. It highlights the importance of group sparse precoder design in CRAN with a large number of distributed RAPs.
5 Conclusion
In this paper, we study the group sparse precoder design that maximizes the sum rate in a CRAN under perRAP power constraint. We show that the joint antenna selection and precoder design problem can be formulated into an norm problem, which is, however, combinatorial and NPhard. Inspired by the theory of compressive sensing, we propose an approach that solves the problem via reweighted norm. Simulation results verify the optimality of our proposed algorithm in that it achieves almost the same performance as that obtained form the exhaustive search. Compared to full cooperation, the group sparse precoding can achieve a significant proportion of the maximum sum rate that achieved from full cooperation with, however, much fewer active RAPs, which highlights the importance to employ sparse precoding in CRAN with ultradense RAPs.
Appendix
Appendix A Proof of Lemma 2.1
Define the channel gain matrix and the precoding matrix between user and the antennas of RAP as and , respectively, and . Let us reorder the channel gain matrices and the precoding matrices as
(33) 
and
(34) 
respectively. We have
(35) 
for all , where (a) follows from the fact that if RAP is not active, i.e., .
References
 (1) CRAN: The Road Towards Green RAN. White paper, ver. 2.5.
 (2) Shi, Y., Zhang, J., Letaief, K.B., Bai, B., Chen, W. (2015): Largescale convex optimization for ultradense cloudran. IEEE Wireless Commun., 22(3), 84–91.
 (3) Liu, C., Zhang, L., Zhu, M., Wang, J., Cheng, L., Chang, G.K. (2013): A novel multiservice smallcell cloud radio access network for mobile backhaul and computing based on radiooverfiber technologies. J. Lightwave Technol., 31(17), 2869–2875.
 (4) Liu, Z., Dai, L. (2014): A comparative study of downlink mimo cellular networks with colocated and distributed basestation antennas. IEEE Trans. Wireless Commun., 13(11), 6259–6274.
 (5) Liu, Z.: On the scaling behavior of the average rate performance of largescale distributed MIMO systems. IEEE Trans. Veh. Technol. To appear.
 (6) Wang, J., Dai, L. (2015): Asymptotic rate analysis of downlink multiuser systems with colocated and distributed antennas. IEEE Trans. Wireless Commun., 14(6), 3046–3058.
 (7) Zhao, J., Quek, T.Q.S., Lei, Z. (2013): Coordinated multipoint transmission with limited backhaul data transfer. IEEE Trans. Wireless Commun., 12(6), 2762–2775.
 (8) Mehanna, O., Sidiropoulos, N.D., Giannakis, G.B. (2013): Joint multicast beamforming and antenna selection. IEEE Trans. Signal Process., 61(10), 2660–2674.
 (9) (2013): Sparse beamforming for limitedbackhaul network mimo system via reweighted power minimization. In: Proc. IEEE Globecom,
 (10) Zhang, Z., Xu, Y., Yang, J., Li, X., Zhang, D. (2015): A survey of sparse representation: Algorithms and applications. IEEE Access, 3, 490–530.
 (11) Dai, B., Yu, W. (2014): Sparse beamforming and usercentric clustering for downlink cloud radio access network. IEEE Access, 2, 1326–1339.
 (12) Shi, Y., Zhang, J., Letaief, K.B. (2014): Group sparse beamforming for green cloudran. IEEE Trans. Wireless Commun., 13(5), 2809–2823.
 (13) Shi, Y., Zhang, J., Letaief, K.B. (2015): Robust group sparse beamforming for multicast green cloudran with imperfect csi. IEEE Trans. Signal Process., 63(17), 4647–4659.
 (14) Telatar, E. (1999): Capacity of multiantenna Gaussian channels. Euro. Trans. Telecommun., 10(6), 585–595.
 (15) Spencer, Q., Swindlehurst, A., Haardt, M. (2004): Zeroforcing methods for downlink spatial multiplexing in multiuser MIMO channels. IEEE Trans. Signal Process., 52(2), 461–471.
 (16) Shen, Z., Chen, R., Andrews, J., Heath, R., Evans, B. (2006): Low complexity user selection algorithms for multiuser MIMO systems with block diagonalization. IEEE Trans. Signal Process., 54(9), 3658–3663.
 (17) Shen, Z., Chen, R., Andrews, J., Heath, R., Evans, B. (2007): Sum capacity of multiuser MIMO broadcast channels with block diagonalization. IEEE Trans. Wireless Commun., 6(6), 2040–2045.
 (18) Shim, S., Kwak, J.S., Heath, R., Andrews, J. (2008): Block diagonalization for multiuser MIMO with othercell interference. IEEE Trans. Wireless Commun., 7(7), 2671–2681.
 (19) Ravindran, N., Jindal, N. (2008): Limited feedbackbased block diagonalization for the MIMO broadcast channel. IEEE J. Sel. Areas Commun., 26(8), 1473–1482.
 (20) Cai, D.W.H., Quek, T.Q.S., Tan, C.W. (2011): A unified analysis of maxmin weighted sinr for MIMO downlink system. IEEE Trans. Signal Process., 59(8), 3850–3862.
 (21) Boyd, S., Vandenberghe, L. (2004): Convex Optimization. Cambridge University Press.
 (22) Zhang, R. (2010): Cooperative multicell block diagonalization with perbasestation power constraints. IEEE J. Sel. Areas Commun., 28(9), 1435–1445.
 (23) Cover, T.M., Thomas, J.A. (2006): Elements of Information Theory. WileyInterscience.
 (24) Bertsekas, D.P. (2003): Convex Analysis and Optimization. Athena Scientific.