Anda di halaman 1dari 4

A Simple Cooperative Extension to Wireless Relaying

Patrick Herhold, Emesto Zimmermann, Gerhard Fettweis Technische Universitat Dresden . Vodafone Chair Mobile Communications Systems . D-01062 Dresden, Germany Email: herhold@ifn.et.tu-dresden.de

Abslract-We propose and analyze a simple protocol that results, implementation issues, and conclusions are discussed simultaneously exploits two potentials offered by wireless relay in sections IV and V. systems: diversity gains and pathloss savings. An intermediate decode-and-forward relay assists bansmission from source to 11. THEPROTOCOL destination, and the destination combines the signals it receives We propose a simple adaptive decode-and-forward protocol from source and relay. The key feature nf the proposed system is that the relay node decides independently whether or not (AdDF). To our knowledge, this terminology has first been to forward information to the destination, thereby minimizing used in 117, where it is also referred to as selection relaying. the risk of errnr propagation while providing truly constructive diversity gains. In combination with pathloss savings, this leads to A. The Adaptive Decode-and-Forward Protocol significant gains over both direct transmission and conventional I ) Simple AdDF: We assume the situation as shown in relaying. The results are obtained from an analysis of the endFig. I(a), where a source sends information directly and via to-end hit e m r rate, which are confirmed by simulation.

I . INTRODUCTION

Given the constraints imposed by link budget estimates for future generations of infrastructure based networks, relaying emerges as a viable option for challenging the tradeoff between transmission range and end-to-end data rate. Essentially, relays that operate in a store-and-forward manner allow for reducing the end-to-end pathloss between an information source and its destination. Each relay in a relay chain thereby usually relies solely on the information sent by its immediate predecessor, and the destination simply listens to the last relay in this chain. We refer tu this as conventional relaying as it is known from ad hoc networking. More recently. the concepts of cooperative relaying have emerged. By allowing cooperation among the relays and by combining hll relays transmissions in the destination, the spatial diversity of relaying systems can be exploited. The information-theoretic fundamentals of two-user cooperation are discussed in [l], [2] and the references therein. Yet, most cooperative schemes suffer from one or more of the following drawbacks: they require feedback, channel state information (CSI) needs to he available at the transmitters, the total number of transmitting nodes increases compared to conventional relaying, or error propagation induced by the relays is not sufficiently tackled; see [3]. In this paper, we propose a low-complexity protocol that aims at overcoming these limitations. It exploits the spatial diversity and the broadcast nature of the wireless medium by slightly altering conventional relay protocols and adding the complexity of a combiner in the destination. We describe and analyze the protocol in sections I1 and 111, respectively. Main
This work has been supponed by Vodafone Group Research and Development.

a relay to a destination. Communication takes place in two phases as dictated by the inability of the relay to transmit and receive simultaneously at the same frequency. In phase one, the source broadcasts its information. Both relay and destination receive faded noisy versions of this signal. Based on the quality of the received signal, the relay decides whether or not to decode-and-forward. For sufficient signal quality it decodes-and-forwards, otherwise it refrains from doing so. The decision can be implemented by, e.g., a cyclic redundancy check (CRC) or by means of a simple SNR threshold as discussed later. In phase fwo, if the relay has decided to forward, then it re-sends a newly encoded version to the destination. The destination combines the received version of this signal with the stored samples it has previously received from the source. Otherwise, i.e. if the relay has decided not to decode, then it simply remains silent. The destination detects this case based on the lack of sufficient signal strength, and for demodulation it needs to rely on the samples stored in phase I . This protocol achieves spatial diversity gains, since for good channel conditions the relay frequently decodes and forwards copies of the original information over an uncorrelated channel to the destination, Simultaneously, we benefit from pathloss savings: a relay station located between source and destination will receive the messages broadcasted by the source much more reliably than the destination, and in tum it needs to use a considerably smaller transmit power to reach the destination. 2 ) Complex AdDF: This protocol, which has first been suggested in [l], differsfrom the simple AdDF protocol in the case that the relay has decided not to forward. The complex protocol prevents the occurring silence by having the source repeat its message. Clearly, the resulting gains from standard repetition coding come at the cost of increased complexity as

0-7803-8329-W04/$20.00 02004 IEEE.

36

Int. Zurich Seminar on Communications (IZS), Feb. 18-20, 2004

Source

&/

Destination

,
0
I

.
d

transmission and QPSK for the relay schemes, and (ii) QPSK for direct transmission and 16-QAM for the relay schemes.' The bit error probabilities can in both cases be evaluated using the Q-function:

(a) Symmetrical network

(h) Asymmetrical nctwork

Fig. 1.

Symeuical and asymmetrical network configurations.

the source must have information on the decoding status of the relay.
B. Parameters, Tradeoff., Performance Bounds

For the rest of the paper we will focus on the case that the forwarding decision at the relay is made on the basis of an SNR threshold: the decoding.threshold SNRd, defines the minimum SNR for which the relay decodes the signal broadcasted by the source. Note the related tradeoff while a large threshold SWRdcc lowers the probability of a decoding error (thus minimizing the risk of error propagation induced by the forwarding relay), it decreases the number of cases in which the relay decodes and forwards, thereby reducing diversity benefits. The second important parameter is related to the power allocation. In order to provide a fair comparison, it is crucial that the total consumed energy of the cooperative relay system does not exceed that of the corresponding direct system.' Hence, the second important parameter is the fraction p of the total available power which is allocated to the broadcast by the source in phase 1, which leaves a fraction of 1 - p for the second phase (0 5 p 5 1). Again there is a tradeoff choosing p very large results in a good chance that the relay will decode, but this will eventually leave no power for the relay to provide diversity gains in phase 2. The selection of the parameters SNRd,, and p is addressed in [4. Section F]. We compare the AdDF protocols with direct SlSO communication, two-branch transmit diversity using two antennas at the source for direct transmission to the destination [SI, and conventional relaying. In the latter case, the relay decodes in every case, and the destination solely receives the signal transmitted by the relay. This is also referred to as layer 3 decode-and-forward relaying (WDF). 111. PERFORMANCE ANALYSIS In this section, we aim at deriving the end-to-end bit error rate for the protocols of interest.
A. Assumptions

where a = 1, b = 2 for BPSWQPSK, and a = 1, b = 41.5 for 16-QAM for Gray-Mapping [6, eq. (5.2-80), adapted to yield BER instead of SER]. b ) Channel Model: Communication takes place over frequency $at fading channels in an uncoded symbol-by-symbol manner? The magnitudes \ h i , jof the channel coefficients follow a Rayleigh distribution; consequently, the channel powers Ihi,j 1' are exponentially distributed. Assuming the parameter of this exponential distribution to be az;: allows for capturing the effects of pathloss and distance by varying the mean powers a& of the channels [I]. c) Network Geometry: One can distinguish between symmetrical and asymmetrical network configurations; see Fig. 1. Cooperative relaying often focuses on the case of symmetrical networks to highlight the mutual benefits of cooperation among source and relay. Sacrificing symmetry allows for exploiting the pathloss savings of asymmetrical scenarios, which are frequently studied in the context of conventional relaying. To take this into account, we model the pathlosses, i.e. the mean channel powers a&, as a function of the relative relay position T without loss of generality by
U$

1,

2 - - - 0 Us,, - T

ar,d

= (1- T)--

(2)

where a is the pathloss exponent, and 0 < T < 1 as defined in Fig. 1. We focus on the case LL = 3 in this paper. d ) Energy: We assume for direct transmission SNR = Eb/No per information bit, which is shared by source and relay in the case of relaying according to the first phase power fraction p .
B. Probability of Bit Error

In the following we derive the end-to-end probability of bit error for the proposed AdDF protocols. I ) Outline of the result: The error probability of the adaptive decode-and-forward protocol can he formulated as
p;AdDF) = pd, p p

+ (1 - pd,) . p;Di=t).

(3)

Here, Pdec is the probability that the relay decodes-andforwards, which is derived in (9) in [4, Section A], and p(Div) IS . the probability that an error occurs in the combined e diversity transmission from source and relay to the destination. Furthermore, PADiRc') is the probability of error given that the
'It should be noted that the BPSWQPSK comparison is somewhat unfair: the double spectral efficiency in the relay case comer at virtually no cost as QPSK and BPSK exhibit the same performance. For this reason we study the QPSW16-QAM case, where the gains from pathloss reduction and spatial diversity compete against the SNR loss incurred by the higher order madulation scheme. 'Although the analysis is done on a per-symbol hasis, the rewlts also hold for uncoded block-based schemes under a block-fading assumption.

a ) Modulation Scheme: The two-phase nature of the relay protocols calls for doubling the spectral efficiency of the individual phases compared to that of direct transmission. Specifically, we investigate two cases: ( i ) BPSK for direct
'We note that our approach is based on a strict and conservative power constraint: allowing the relay to add additional power can only increase the attractiveness of lhe shldied protocol.

37

Int. Zurich Seminar on Communications(IZS), Feb. 18-20, 2004

relay has decided not to decode, in which case the destination needs to rely on the direct communication from the source. Both error events are examined in the following. 2) Diversity iransmission: This occurs if the relay has decided to decode-and-forward. The destination combines the signals it receives from source and relay in the first and the second phase, respectively. The resulting error probability is p;Div) = p(s.rlpCx1 + (1 pLZ),
e e

p F 1 )

(4)
1 ; 104 D .

Here, two mutually exclusive error events may lead to a decision error at the destination - depending on whether or not the relay decodes correctly. Firstly, for a badly chosen decoding threshold there is the danger of a decision error at the relay. It occurs with probability and depends on the decoding threshold SNRd, and the effective mean SNR 7 = pu:,SNR at the relay. The probability of error at the decoding relay can be determined in closed form using the result (15) of [4, Section C ]
=

AdDF
~ i~ p ~ ~ y . , ~~ Rayleigh chl, uncoded p : s Transmit ~ diversit?
\

IO .

.-

(Alamouti)
15

10

20

25

30

Eb/No per information bit [dB]


Fig. 2. BER vs. SNR for the proposed adaptive decode-and-faward protocol in comparison to direct transmission, relaying at layer 3 (L3DF) and transmit diversity. Symmetrical network. Solid lines: simulation, squares: analysis (for AdDF).

(sRd,,,pu:r,%)

(5)

In case of such a decision error the relay transmits an erroneous signal to the destination, thus leading to error propagation. We approximate the resulting error probability with P$ 5 112. Such error propagation should he avoided by appropriately choosing SNRd, at the relay. Secondly, when the relay has decoded correctly (with probability 1 - PPI), it forwards a freshly encoded signal to the destination. This is the desired case of exploiting spatial diversity, since the destination combines this signal with the version it has received from the source. The resulting , depends on the effective SNRs of the two error rate P involved channels. These are = pu:,SNR for the channel R the from source to destination, and 72 = (1- ~ ) u , ? ~ S N for channel from relay to destination. A closed form solution of PPI is provided in (16) in [4, Section Dl for the novel case
7 1

Here, SN& = ?u,?,SNR for equal powers at both transmit antennas. 3 ) Decode-and-forward ai layer 3: In the case of conventional relaying the relay acts as a plain forwarder, and the destination only relies on the information sent by the relay. The performance of this protocol is
pAL3DF)

1- ( 1- p(s0 e ) (1 - P y ) ,

(7)

3 ) No diversify transmission: If the relay has decided not

# 72.

to decode, then the destination needs to rely on the direct channel from source to destination. The corresponding error probability is given by (17), derivedin [4, Section E]. At this point, we have achieved a closed-form solution for the error probability of the proposed protocol over Rayleigh fading channels (using (9), (4), (3,(16), and (17) in (3)).

where PF. and PFd)are the error rates for communication over the two successive Rayleigh fading channels, given by (17) in [4] with effective SNRs pu&SNR and (1- ~ ) U ; ~ S N R , respectively. The protocol benefits from pathloss savings, but not from spatial diversity. The BER is obviously minimized for P P I = PFdl,leading to an optimum power fraction of p ~ ~=~*.~ F l We now tum to comparing the AdDF protocols with these conventional schemes.

Iv. RESULTS AND DISCUSSION


A. Symmetric Nenvorks We start by discussing the performance for symmetrical networks. Fig. 2 compares the proposed simple AdDF protocol with the complex AdDF protocol and their respective performance hounds. Both AdDF protocols achieve diversity gains over direct transmission and conventional relaying. For example, at BER = lo- the cooperative protocols achieve gains of approximately 2-3 dB. Note that the complex protocol performs only slightly better than its simple counterpart, indicating that the gains from repetition coding in phase two are not substantial. Therefore we focus on the simple version in the rest of the paper. Finally, we see that analysis and simulation results match; differences are due to the approximation of Pp in (4).

C. Performance Bounds
Before examining the performance of the AdDF protocol we briefly review three different conventional altematives. 1 ) Direct transmission: The error rate is given by (17) with SN&c = o:,SNR, which is the SNR of the direct channel from source to destination. 2 ) Transmit diversify: We refer to transmit diversity as using two antennas at the source for direct communication to the destination as described in [5]. The error rate is given by [6, eq. (14.4-15)l

38

Int. Zurich Seminar on Communications (IZS), Feb. 16-20, 2004 relies on the knowledge of the average SNR and the longterm average pathlosses between the corresponding nodes. Explicit knowledge of channel state information is therefore not needed at the transmitters, but as usual estimates of the channel coefficients are needed for combining and demodulation at the receivers. With respect to synchronization, the same requirements hold as for conventional relaying. V. CONCLUSIONS A simple adaptive decode-and-forward protocol has been proposed that is designed to combine the pathloss savings o f conventional relaying with the spatial diversity gains and broadcast advantages of cooperative relaying. Due to its fixed two-phase nature, the proposed scheme can be understood as an extension of conventional relaying. Two parameters, namely a decoding threshold and a power allocation allow to adapt the protocol to certain SNR regimes and pathloss conditions. The derived closed-form expressions for the end-to-end error probability help finding rules for adjusting the parameters. These are based on knowledge of the average SNR and mean pathlosses; channel state information is not necessary at the transmitters. The simple AdDF scheme exhibits SNR gains of up to seven decibels over direct transmission, while conventional relaying was shown to achieve three decibels. These results were obtained under a strict normalization of power, bandwidth, and delay for symbol-by-symbol communication over Rayleigh fading channels. Finally, we note that the cooperative protocol has the same number of transmitting nodes as conventional relaying, which might be a first promising indication of the system-level performance in interference-limited scenarios. In consideration of these outcomes, we believe the proposed adaptive decode-and-forward protocol to be an interesting candidate for future wireless networks. ACKNOWLEDGEMENTS The authors gratefully acknowledge the funding received from Vodafone Group Research and Development. Funhermore, we would like to thank W. Rave for the valuable discussions.

,
-5

0.

: relaying (L3DF)
0.3
0.4

1
0.7

..
A

0.1

0.2

0.5

0.6

0.8

Relative position of relay r


Fig. 3 . S N R gains of simple AdDF and conventional relaying (L3DFj over direct transmisSion. While L3DF exploits pathloss savings, the cooperative AdDF pmtocol additionally benefits from diversity gains. Parameten are RER=10W2 and pathloss exponent a = 3.0

B. Asymmetric Networks
The proposed AdDF protocol is designed to benefit from diversiiy gains as well as fmm pathloss savings. The latter advantage becomes apparent in asymmetric networks, where the relay is located between source and destination. To examine this, Fig. 3 depicts the SNR gain over direct transmission at BER = 10W2 as a descriptive performance measure for uncoded settings for both studied modulation schemes (BPSWQPSK and QPSW16-QAM). We note that substantial improvements can he obtained using the simple AdDF protocol. By exploiting diversity, the cooperative protocol outpelforms the conventional relay protocol by at least 3 dB for both spectral efficiencies and for all studied relay positions. Next, we observe that the SNR loss [I] incurred by using a higher spectral efficiency leads to a correspondingly worse performance of the relay protocols compared to direct transmission. For example, having the relay located halfway between source and destination (T = 0.5) and using QPSK for cooperative relay transmission leads to an SNR gain of 6 . 6 dB over direct transmission using BPSK. This gain reduces to 3.6 dB if 16-QAM relay transmission is used instead of direct QPSK transmission. Eventually, the negative SNR gain suggests that the conventional L3DF protocol with 16-QAM performs worse than direct QPSK transmission. It is worth repeating that these results have been obtained under a strict nomlization of power; delay, and bandwidth.

REFERENCES
[ I ] J. N. Laneman, D. N. C. Tse, and G . W.Womell, "Cooperative Diversity in Wireless Networks: Efficient Prolecols and Outage Behavior," IEEE Trona Inform. Theory, Apr. 2W3, (accepted for publication). [2] M. Dohler, A. Gkelias, and H. Aghvami, "2-Hop Distributed MIMO Communication System:' IEEE Elecrmnics Lerrers, June 2003. 1 3 1 E. Zimmermann, P. Herhold, and G . Fettweis, "On the Performance of Cooperative Diversity Pmtocols in Practical Wireless Systems,'' in Proc. 58th IEEE Vehic. Techno/. Con$, Orlando, FL,Oct. 2003. [ 4 lP . Herhold, "A Simple Cooperative Extension to Wireless Relaying - Appendix," Feb. 2004, Available at www.ifn.et.tudresden.de/-herhol~~/publicalions. [5] S. M. Alamouti, "A Simple Transmit Diversity Technique for Wireless Communications," IEEE J. Select. Areas Comm., vol. 16, no. 8, pp. 1451-1458. Oct. 1998. 1 6 1 John G . Proakis, Digital Communicorions. McGraw-Hill International t h edition. 200?. Editions, 4 [7l I.S. Gmdshteyn and I.M. Ryzhik, Table ojlntegro0. Series, and Pmducts, Academic Press, San Diego, CA, 5 t h edition, 1994.

C. Implementation Aspects
The proposed simple AdDF protocol is similar to conventional relaying protocols. Both are static two-phase protocols that impose the same end-to-end delay. As the relay decides independently whether or not to decode and forward, there is no need for feedback information. In fact, the protocol only

Anda mungkin juga menyukai