You are on page 1of 6

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 61, NO.

7, APRIL 1, 2013

1657

Source Transmit Antenna Selection for MIMO Decode-and-Forward Relay Networks


Xianglan Jin, Jong-Seon No, and Dong-Joon Shin

AbstractTransmit antenna selection (TAS) is usually applied to multiple-input multiple-output (MIMO) systems because it does not require additional radio frequency (RF) chains which are quite expensive. In MIMO decode-and-forward (DF) relay networks, both source-destination and source-relay-destination paths should be simultaneously considered to nd an effective source TAS (STAS). In this paper, a new STAS is proposed based on both channel state information and transmission scheme for the MIMO DF relay networks. It is also shown that the proposed STAS which antennas among transmit antennas at the source can selects . Simulation results achieve full diversity regardless of the value of show that the proposed STAS has better average bit error probability (BEP) performance than other STASs. Also, the proposed STAS with has lower cost, complexity, overhead, and BEP than the STAS using full-rate full-diversity space-time block codes with with the same total transmit power. Index TermsDecode-and-forward (DF), diversity, multiple-input multiple-output (MIMO), relay network, transmit antenna selection (TAS).

I. INTRODUCTION When multiple antennas are used at the source, transmit diversity can be achieved by using space-time block codes (STBCs). However, STBCs require multiple antennas associated with radio frequency (RF) chains which are costly in terms of size, power, and hardware [1]. To solve this problem, low-cost and low-complexity antenna selection schemes have been studied [1][5], and transmit antenna selections (TASs) with STBCs have also been considered [6][8]. Unlike point-to-point MIMO systems, cooperative relay systems utilize two independent source-destination (SD) and source-relay-destination (SRD) paths. To select good transmit antennas at the source, we have to consider both the SD and SRD paths simultaneously. For amplify-and-forward (AF) relay networks, the optimal and suboptimal TASs at the source were investigated based on maximizing signal-tonoise ratio (SNR) at the destination [9], [10]. Unfortunately, contrary to the AF relaying case, the exact SNR for the decode-and-forward (DF) relaying is very difcult to derive. For the DF relay networks, a joint relay-and-antenna selection scheme which selects the best relay and the best antenna at both source and the selected relay was studied without considering the SD link in [11]. In [12], a suboptimal TAS of selecting two antennas at the source was proposed such that one maximizes the
Manuscript received October 31, 2011; revised June 14, 2012 and October 13, 2012; accepted December 31, 2012. Date of publication January 18, 2013; date of current version March 08, 2013. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Josep Vidal. This work was partly supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education, Science and Technology(2012-0004375) and the Korea Communications Commission (KCC), Korea, under the R&D program supervised by the Korea Communications Agency(KCA) (KCA-2012-08-911-04-003). X. Jin is with the Department of Information and Communication Engineering, Dongguk University-Seoul, Seoul 100-715, Korea (e-mail: jinxl77@gmail.com). J.-S. No is with the Department of Electrical Engineering and Computer Science, INMC, Seoul National University, Seoul 151-744, Korea (e-mail: jsno@snu.ac.kr). D.-J. Shin is with the Department of Electronic Engineering, Hanyang University, Seoul 133-791, Korea (e-mail: djshin@hanyang.ac.kr). Digital Object Identier 10.1109/TSP.2013.2241053

SNR of source-relay (SR) link, and the other maximizes the SNR of the SD link. Also, the maximum diversity was achieved by using Alamouti code [13]. However, this scheme can be used only for two-antenna selection and has not been extended to general multiple-relay networks. In this paper, we consider DF relay networks of one source, one des, , and antennas, respectively. tination, and relays with We assume that the relay-destination (RD) channels are orthogonal, which decreases the data transmission rate. The reason for this assumption is that if the relays transmit signals via the same channel, the potential maximum diversity may be difcult to achieve. To achieve such maximum diversity, joint coding for multiple relays should be investigated and it is far from the scope of this paper. In this paper, we propose a criterion of source TAS (STAS) of seantennas among transmit antennas at the source based lecting on the upper bound on the pairwise error probability (PEP) derived in [14]. The proposed STAS can be performed at the destination, and then the information on the selected transmit antennas is fed back to the source. In the rst phase, the source transmits an uncoded single ) or a codeword of a full-diversity STBC with symbol ( transmit antennas. During the second phase, anthe relays decode, re-encode, and re-transmit signals from tennas, and so the relays may transmit erroneous signals. Finally, the destination decodes the received signals from the source and the relays by using the near-maximum-likelihood (near-ML) decoding scheme [15]. antennas We prove that the proposed STAS which selects transmit antennas achieves the maximum diversity among in the DF relay networks. We also compare the average bit error probability (BEP) of the proposed STAS with those of other STASs through Monte Carlo simulation. The simulation results show that the proposed STAS has better average BEP than other STASs [12]. Moreover, with the same total transmit has lower cost, complexity, power, the proposed STAS with using full-rate overhead, and BEP than the STAS with full-diversity STBCs. The following notations are used in this paper: the capital letter dedenotes the identity matrix; denotes a notes a matrix; complex matrices; represents the Frobenius norm of set of denotes the denotes the expectation; the superscript a matrix; , complex conjugate transpose. For denotes that the elements of are independent and identically distributed (i.i.d.) circularly symmetric Gaussian random variables with zero mean and variance . II. SYSTEM MODEL AND SOURCE TRANSMIT ANTENNA SELECTION A. System Model A cooperative DF relay network with one source, one destination, relays with , , and antennas, respectively, is and considered as shown in Fig. 1. The half-duplex transmission and frequency-at quasi-static fading channels are assumed. It is also assumed relay knows the channel state information (CSI) of the corthat the SR link and the destination knows the CSIs of all SR, responding and be the numbers of transmitted symSD, and RD links. Let bols at the source and the relay during the rst and the second phases, respectively, and be a set of message symbols from the -ary signal and constellation. Let be the channel coefcient matrices SR link and the SD link, respectively, where and are of the and channel vectors from the transmit antenna

1053-587X/$31.00 2013 IEEE

1658

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 61, NO. 7, APRIL 1, 2013

B. Source Transmit Antenna Selection To select good source transmit antennas, we should consider both the SD and SRD paths simultaneously. Unlike AF relay networks, it is difcult to nd the optimal solution for the STAS in the DF relay networks due to the difculty in deriving their error probabilities. Instead, the union bound on BEP can be used as a criterion of selecting good source antennas by deriving PEPs. However, it is still difcult to derive the exact PEP. Therefore, for the MIMO relay networks with the source antenna subset, we can use the following upper bound on PEP , for ,

Fig. 1. A DF relay network with multiple relays. The solid line denotes the rst phase transmission and the dashed line denotes the second phase transmission.

(4)

at the source to antennas in the relay and antennas in the destination, respectively. In the rst phase, the source broadcasts a codeword encoded from a full-diversity code with -tuple mesto the relays and the sage vector selected andestination by using the possible source antenna subsets can tennas. Thus, be selected. We dene them as , . We assume that the antennas in the source are selected. Then, the column vectors and compose

means by adopting the result of Theorem 2 in [14], where . As proved in [14], the expectation of the upper bound in (4) taken over the random variables denoting CSIs is , where is the diversity of MIMO DF relay proportional to networks. Therefore, the union bound on BEP derived from the upper bounds for all pairs of and in (4) can be used as a performance criterion. Let

antenna subset

the channels matrix and , respectively. Hence, the received relay and the destination can be written as signal at the Then, selecting the antennas in the subset which satises

(5)

(1) (2) is the average transmit power at the source, is the noise matrix at the relay with dis, and tribution represents the noise matrix at the destination with distribution . relays transmit the codewords reencoded In the second phase, orthogonal RD channels. from their decoded symbols through orthogonal Thus, the received signal at the destination through the channel is given as respectively, where (3) is the codeword constructed from -tuple where decoded by the message vector relay in the rst phase. A relay may decode correctly or incorrectly, may be different from . is the average transmit and therefore is the channel coefcient matrix power of each relay, of the RD channel distributed as , and is the noise matrix at the destination for the orthogonal channel with .

(6) the union bound on BEP derived from the upper bounds on PEPs in (4) can be minimized. The performance of this proposed STAS is analyzed in the following sections. III. DIVERSITY ANALYSIS OF THE PROPOSED STAS In this section, we show that the proposed STAS achieves full diversity. Using (4), the upper bound on the average PEP of the proposed STAS based on (6) can be written as

(7) Let be the minimum rank among the ranks of for all , and and be the unitary matrices whose columns are the eigenvectors of and for any , respectively. We dene an matrix with , where

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 61, NO. 7, APRIL 1, 2013

1659

means the column of a matrix. Since the multiplication of the unitary matrix does not change the statistical distribution of the matrix with , circularly symmetric complex Gaussian entries, the entries of , and have the same distribution as the entries of , and be the minimum values and , respectively. Let among nonzero eigenvalues of and for all , respectively. is a full-diversity code, we have Since

, and . Then, by using the upper bound in Lemma 1, the expectation in (9) can be upper bounded as Proof: Let

By from the Appendix D in [14], the expectation can also be upper bounded as

Finally, an upper bound on the average PEP for the DF relay networks with the proposed STAS is derived as

(8) where Let is due to Fact 1 in Appendix A, and and . is from

(10) Therefore, the DF relay networks with the proposed STAS can achieve , and when the diversity , the maximum diversity is achieved. IV. SIMULATION RESULTS AND DISCUSSION In this section, we compare the average BEPs of MIMO DF relay networks with the proposed STAS and other STASs, and also compare . For other the performance of the proposed STAS for various antennas among transmit antennas at the STAS of selecting source, the following schemes are considered. The rst one selects antennas with largest SNRs of the SR link, called MAX-SR, and another one selects antennas with largest SNRs of the SD link, called MAX-SD as shown in [12]. Also, the random selection which selects antennas , the STAS which randomly is considered. For the case of selects one antenna with the maximum SNR of the SR link and the other antenna with the maximum SNR of the SD link [12] is also considered, which will be called MAX-SR-SD. For the simulation, quadrature phase shift keying (QPSK) is used under the channel condition of , and total transmit powers at the source , respectively. Furthermore, the ML and at each relay are 1 and decoder is used at each relay and the near-ML decoder [15] is used at the destination. , To begin with, we consider the case of , , and . Fig. 2 compares the average BEPs of various STASs in the DF relay network. It is easy to nd that the proposed STAS has better average BEP performance than and much better performance MAX-SD by about 1.5 dB at BEP= and than random selection and MAX-SR for both cases of . In Fig. 3, we compare the average BEP of the proposed STAS with , , , and those of other STASs for in the Alamouti-coded DF relay network with and , where

and

The PDF of is very difcult to derive for general multiple-antenna cases. However, by doing integration by parts, we can rewrite the last part of (8) as

(9) Then, an upper bound on the average PEP can be derived by calculating an upper bound on cumulative density function (CDF) of , which is derived in the following lemma. can be upper bounded as Lemma 1: The CDF of

, , and . Proof: See the Appendix B. Using Lemma 1, the following theorem for the achievable diversity can be established. antennas Theorem 1: The proposed STAS which selects transmit antennas at the source can achieve the maximum among diversity in the MIMO DF relay , , networks of one source, one destination, and relays with antennas, respectively. and

where

The proposed STAS also shows better average BEP than MAX-SD and MAX-SR-SD, and much better average BEP than random selection and MAX-SR.

1660

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 61, NO. 7, APRIL 1, 2013

Fig. 2. Comparison of average BEPs of various STASs with , and in DF relay networks with one relay.

Fig. 4. Comparison of average BEPs of the proposed STAS with various in MIMO DF relay networks with and .

Fig. 3. Comparison of average BEPs of various STASs with , , and in Alamouti-coded DF relay networks with one relay.

Fig. 5. Comparison of average BEPs of the proposed STAS with various in MIMO DF relay networks with and .

Next, we discuss and compare the proposed STAS for various . requires one RF chain, The proposed STAS with feedback bits, and calculations of the metric in (5). Therefore, without considering the error correction performance, the proposed is the most benecial case. To fairly compare the STAS with error probabilities on similar decoding complexity level, we consider by using single-symbol-decodthe proposed STAS with various able full-rate full-diversity STBCs at the source. and with First, we consider the case of . While the uncoded single symbol is , Alamouti code used for the case of is used for the case of . Fig. 4 shows that the BEP curves of and with have the same the proposed STAS with and even though the case of has slope for the same with the better average BEP performance than the case of same total transmit power.

Additionally, we consider the case of and with in Fig. 5. For the same symbol is used for rate and diversity, an uncoded single symbol , the Alamouti code is used for the case of , and the coordinate interleaved STBC (CISTBC) [16],

is used for the case of , where and with the optimal rotation angle . The also show the same BEP curves of the proposed STAS for various shows the better average BEP perforslope, and the case of with the same total transmit power. mance than the cases of can be a good STAS Therefore, the proposed STAS with scheme at the source.

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 61, NO. 7, APRIL 1, 2013

1661

V. CONCLUSION In this paper, a new STAS for MIMO DF relay networks was proposed by considering the SD and SRD paths. It was also proved that the proposed STAS can achieve the maximum diversity regardless of the number of the selected antennas. The simulation results showed that the proposed STAS has better average BEP performance than other existing STASs ( and ). Surprisingly, the proposed STAS with has lower cost, lower complexity, lower overhead, and better BEP perforwith full-rate full-diversity STBCs. mance than the cases of can be a good STAS Therefore, the proposed STAS with scheme. APPENDIX A FACT 1 matrix , there exists a unitary maFact 1: [15]: For an and a real diagonal matrix such that , where , are the eigenvalues of , and are the corresponding eigenvectors. Suppose that the columns of are nonzero and the remaining eigenvalues are all zero, and is the minimum nonzero eigenvalue. Then, for any matrix , the inequality always holds where is an matrix constructed by using the column of as its column, . trix APPENDIX B PROOF OF LEMMA 1 The CDF of can be written as

Since

where we have

means the element of

row and

column of a matrix,

(13) where is due to

and the fact that and are i.i.d., and is due to the exponential distribution of random variable with the rate parameter the last term in (12), we have . For

Since the random variables are not statistically independent, the CDF of derive. However, we have

, is very difcult to Since

(14)

and thus, the CDF can be upper bounded as

by the similar derivation in (13), we have (11) (15) where the row vectors and are statically independent and where . Also, from the fact

the row vectors are i.i.d. Therefore, we can rewrite the upper bound on the CDF of in (11) as we have (12) (16)

1662

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 61, NO. 7, APRIL 1, 2013

where . By plugging (15) and (16) into (14), can be upper and plugging (13) and (14) into (12), the CDF of bounded as the equation at the top of the page.

REFERENCES
[1] S. Sanayei and A. Nosratinia, Antenna selection in MIMO systems, IEEE Commum. Mag., vol. 42, no. 10, pp. 6873, Oct. 2004. [2] A. F. Molisch and M. Z. Win, MIMO systems with antenna selection, IEEE Microw. Mag., vol. 5, no. 1, pp. 4656, Mar. 2004. [3] I. Berenguer, X. Wang, and V. Krishnamurthy, Adaptive MIMO antenna selection via discrete stochastic optimization, IEEE Trans. Signal Process., vol. 53, no. 10, pp. 43154329, Nov. 2005. [4] Z. Chen, J. Yuan, and B. Vucetic, Analysis of transmit antenna selection/maximal-ratio combining in Rayleigh fading channels, IEEE Trans. Veh. Technol., vol. 54, no. 4, pp. 13121321, Jul. 2005. [5] H. Zhang and H. Dai, Fast MIMO transmit antenna selection algorithms: a geometric approach, IEEE Commun. Lett., vol. 10, no. 11, pp. 754756, Nov. 2006. [6] D. A. Gore and A. Paulraj, MIMO antenna subset selection with space-time coding, IEEE Trans. Signal Process., vol. 50, no. 10, pp. 25802588, Oct. 2002. [7] Z. Chen, J. Yuan, B. Vucetic, and Z. Zhou, Performance of Alamouti scheme with transmit antenna selection, Electron. Lett., vol. 39, no. 23, pp. 16661668, Nov. 2003.

[8] D. J. Love, On the probability of error of antenna-subset selection with space-time block codes, IEEE Trans. Commun., vol. 53, no. 11, pp. 17991803, Nov. 2005. [9] S. Peters and R. W. Heath, Nonregenerative MIMO relaying with optimal transmit antenna selection, IEEE Signal Process. Lett., vol. 15, pp. 421424, 2008. [10] H. A. Suraweera, P. J. Smith, A. Nallanathan, and J. S. Thompson, Amplify-and-forward relaying with optimal and suboptimal transmit antenna selection, IEEE Trans. Wireless Commum., vol. 10, no. 6, pp. 18741885, Jun. 2011. [11] M. Ju, H. K. Song, and I. M. Kim, Joint relay-and-antenna selection in multi-antenna relay networks, IEEE Trans. Commum., vol. 58, no. 12, pp. 34173421, Dec. 2010. [12] G. Zhang, W. Zhan, and J. Qin, Transmit antenna selection in the Alamouti-coded MIMO relay systems, Wireless Pers. Commun., vol. 64, no. 4, pp. 879891, Feb. 2012. [13] S. Alamouti, A simple transmit diversity technique for wireless communications, IEEE J. Sel. Areas Commum., vol. 16, no. 8, pp. 14511458, Oct. 1998. [14] X. Jin, J.-S. No, and D.-J. Shin, Relay selection for decode-and-forward cooperative network with multiple antennas, IEEE Trans. Wireless Commun., vol. 10, pp. 40684079, Dec. 2011. [15] X. Jin, D.-S. Jin, J.-S. No, and D.-J. Shin, Diversity analysis of MIMO decode-and-forward relay network by using near-ML decoder, IEICE Trans. Commun., vol. E94-B, no. 10, pp. 28282836, Oct. 2011. [16] M. Z. A. Khan and B. S. Rajan, Single-symbol maximum-likelihood decodable linear STBCs, IEEE Trans. Inf. Theory, vol. 52, no. 5, pp. 20622091, May 2006.

You might also like