A Survey on Cooperative Diversity for Wireless ... - Semantic Scholar

7 downloads 147 Views 357KB Size Report
advantages of multi-antenna space diversity to single antenna networked devices ... of cooperative diversity techniques, any wireless network affected by fading ...
IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

1

A Survey on Cooperative Diversity for Wireless Networks F. Gómez-Cuba, R. Asorey-Cacheda and F.J. González-Castaño

Abstract—Diversity, i.e. transmitting multiple replicas of a signal, may mitigate fading in wireless networks. Among other diversity techniques, the space diversity of multi-antenna systems is particularly interesting since it can complement other forms of diversity. The recent cooperative diversity paradigm brings the advantages of multi-antenna space diversity to single antenna networked devices, which, through cooperation and antenna sharing, form virtual antenna arrays. However, cooperative diversity is a complex technique and research on this topic is still in its early stages. This paper aims at providing a general survey on the theoretical framework; and the physical and medium access control proposals in the literature. Index Terms—Wireless networks, cooperative diversity, MAC protocols.

H

S

D

I. I NTRODUCTION Due to time-variant fading, the attenuation in a wireless channel may vary due to multiple circumstances. Thus, wireless system designs typically include some degree of diversity so as to provide the receiver with several realizations of the signal, which increases the chances of a successful transmission [1]. Many forms of diversity are possible depending on how different available channels or subchannels replicate the signal. Time diversity consists of transmitting replicas with enough separation in time to allow signal decorrelation. Frequency diversity relies on multiple carriers, and space diversity systems have multiple antennae that are sufficiently spaced and receive the same information [1]. Wireless user devices tend to be constrained in size, complexity and power, rendering previous diversity methods unfeasible. To cope with this problem, the cooperative diversity paradigm has appeared recently. In it, single-antenna networked nodes coordinate themselves to form a virtual antenna array, seeking the advantages of MIMO spatial diversity [1]: each source associates itself with other nodes, acting as helpers that first receive the transmission of the source and then relay the information. As a result, one extra transmission is needed to send the information to the receiver and the number of hops in each route is doubled. The increase in cost is only due to the second stage, since the broadcast nature of the wireless network allows simultaneous transmission to as many helpers as needed in the first stage. Furthermore, multihop transmission [2] with cooperative diversity may favor a F. Gómez-Cuba, R. Asorey-Cacheda and F.J. González-Castaño are with the Departamento de Enxeñaría Telemática, Universidade de Vigo, ETSI Telecomunicación, Campus, 36310 Vigo, Spain, Phone: (+34) 986 814 081, Fax: (+34) 986 812 116, email: [email protected]; [email protected]; [email protected].

Fig. 1. With cooperation, more nodes are subject to interference. In this case, the helper (H) that receives the transmission of the source (S) and forwards it to the destination (D), generates extra interference that would not be present if direct transmission was used.

better reception in each receiver along the path and, thus, improvements in range, rate or autonomy. These tradeoffs are analyzed in [1] and [3], which conclude that the strategy is profitable. For this reason, future wireless network designs should consider cooperation capabilities. Cooperative diversity, as a technique to combat fading, should find its niche in the upcoming generations of mobile data networks, typically cellular architectures, granting a higher throughput by means of spectrum reutilization. Consequently, the cooperative scenario requires new analyses as the introduction of third parties tends to increase interference [4], [5] (see Figure 1). Although mobile cellular networks are the natural target of cooperative diversity techniques, any wireless network affected by fading can use them. Since the benefit increases as more potential helpers conform the network, dense sensor networks [6], [7] represent a good application scenario for low complexity cooperative techniques. A cooperative virtual backbone may also be of interest in ad-hoc networks [8], [2]. Cooperation is a versatile strategy, which can be exploited for purposes other than diversity. Cetinkaya and Orsun proposed a MAC protocol in which nodes cooperate to adapt their contention windows, improving fairness [6]. Nevertheless, cooperative diversity differs from other cooperation techniques

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

S2

Phase-I Phase-II

2

D2

S1 tx.

S2 tx.

T 2

T 2

Fig. 3. TDMA medium sharing without cooperation. The assignment period T is divided in two medium accesses.

S1

D1

Fig. 2. Wireless network example: Two sources S1 and S2 access the medium alternately to transmit to the two destinations D1 and D2 ; both nodes of each communication pair overhear the transmission of the other pair.

in the sense that the improvement opportunities rely on the physical layer (PHY). Consequently a cross-layer design is required if other layers are involved. The rest of this paper is structured in four parts. Section II reviews the background from the perspective of information theory, emphasizing the fact that a point-to-point approach cannot achieve the full capacity of a network [9]. Different theoretical transmission models are reviewed and compared. Section III analyzes recent PHY architectures. Relying on information theory, we discuss a representative group of cooperative PHY techniques. Section IV surveys cooperative MAC layers. Finally, section V concludes the paper. II. I NFORMATION T HEORETICAL A PPROACH A. Philosophy Let us consider a four-node wireless network with two transmitters S1 and S2 and two destinations D1 and D2 as shown in figure 2. The two transmitters share the medium by some means, such as, for instance, pre-assigned periodic symmetric TDMA slots (figure 3). Assuming a Rayleigh fading coefficient α that is slowly varying and flat, the link outage probability P0 is defined as the probability that the signal-to-noise ratio seen by the receiver (|α|2 SN R, where SN R is the signal-to-noise ratio at the transmitter) is lower than the minimum needed, i.e. the rate is above the mutual information of the channel [10]:    2R − 1 Po = P R > log2 (1 + |α|2 SN R) = P |α|2 < SN R (1) where R is the binary rate per Hertz, defined per slot. It is remarkable that, due to the broadcast nature of the wireless medium, a node that is idle at a certain moment can overhear the transmissions of its peers. In typical wireless systems these receptions are simply discarded. However, since different nodes experience independent realizations of the fading phenomena, the success probabilities are independent and, thus, a transmitter can rely on a helper to create a form of diversity.

S1 tx.

S2 relays S1

S2 tx.

S1 relays S

T 4

T 4

T 4

T 4

Fig. 4. TDMA medium sharing with cooperation.The assignment period T is divided in four slots, encompassing two direct transmissions and two relay transmissions. The same information is transmitted twice and the rate is halved.

It is possible -by halving the transmission rate- to allow a node to allocate half its transmission time to its own information and the other half to relaying information (figure 4). Using the above forwarding policy, the overall outage probability would be of the order of O(Po2 ), because a packet would only get lost in the case of outage in two independent routes (one composed of two links, with the helper in between). Evidently, this model is oversimplified: link outages are assumed to be independent, the PHY layer processes individual packets independently, the helper must fully (and successfully) decode the packets it retransmits, the destination decodes each replica independently, and at least one of the replicas must be successfully decoded at the destination. However, if we analyze the problem from the perspective of information theory as in [10], the system entangles much more mutual information than that defined by the hop-byhop success-requiring characterization. Mutual information is given by the combination of the direct link and the two-hop link from a source to a destination. This means that the destination can use information from both the direct transmission and the relayed transmission in order to decode the data. Laneman et al [11] considered two classical relay algorithms depending on whether or not the relay/helper decoded the received signal, (decode-and-forward (DF) and amplify-and-forward (AF), respectively). In cooperative AF, an optimum receiver accesses the information of two parallel noisy channels, one of which has a classical AF relay. Regarding DF, either the source-to-relay or the relay/source-to-destination links limit the maximum achievable rates. B. Theoretical System Models Based on classical relay techniques, Laneman et al suggest some improvements [10]: • Selection Relaying uses either AF or DF and, if the relay reception signal to noise ratio (SNR) is low, the relay stays silent, allowing the source to retransmit instead. • Incremental Relaying is based on the assumption of some kind of acknowledgement. So, regular communication takes place first and, if no acknowledgement (ACK) is

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

3

d (R nor m ) DF Direct AF Incremental AF

2

Ch1

S1 tx.

S2 relays S2

S3 relays S2

SN relays S2

Ch2

S2 tx.

S1 relays S2

S3 relays S2

SN relays S2

Ch3

S3 tx.

S1 relays S3

S2 relays S3

SN relays S3

ChN

SN tx.

S1 relays SN

S2 relays SN

SN −1 relays SN

T N

T N

T N

1

T N

Fig. 6. TDMA-based system with diversity order N and Rnorm = 1/N . In a given user channel, all other users relay information in different time slots.

0.5

1

R nor m

Fig. 5. Diversity for direct transmission, AF, DF and incremental AF. Incremental AF achieves the upper bound of a 2-antenna system.

received, the relay sends its copy of the signal to the destination. The metric of interest is the outage probability Po (SN Rnorm ) as a function of transmitter SNR values, normalized by the minimum SNR required by each protocol (the normalization allows a fair comparison by separating the improvement due to diversity and the improvement due to the different spectral efficiencies of each protocol). The diversity order of a system is defined as the exponent that asymptotically relates SNR increase to Po decrease. This was formulated in [12] as: d≡−

lim

SN R→∞

log(Po ) log(SN R)

(2)

A system with Nt transmitter antennae and Nr receiver antennae is said to provide full diversity when d = Nt × Nr . In addition, the trade-off between diversity and normalized spectral efficiency -i.e. R normalized by its maximum sustainable value- can be analyzed: full diversity is achievable when Rnorm is zero as shown in figure 5. The normalization expression is: Rnorm :=

R 2 SN R) log2 (1 + σs,d

(3)

2 where σs,d is the channel variance. It may be identified as the improvement in R related to SNR, called multiplexing gain in [12]: R (4) r ≡ lim SN R→∞ log(SN R)

In [13], Laneman and Wornell studied systems with several helpers. Two cooperative schemas were proposed: 1) Each user channel is divided into N time slots, being N the number of transmitter (figure 6). At each time slot a different relay transmits, thus dividing R by N . 2) The user channels only have two slots (figure 7). All relays transmit simultaneously using space-time coding in the second slot. More generally, all the surrounding nodes can potentially contribute to increasing the receiver information and those

Ch1

S1 tx.

D(S1 ) relays S1

Ch2

S2 tx.

D(S2 ) relays S2

Ch3

S3 tx.

D(S3 ) relays S3

ChN

SN tx.

D(SN ) relays SN

T 2

T 2

Fig. 7. STC-based system with diversity order N and Rnorm = 1/2. In a given user channel, all other users may relay information simultaneously. The operator D(Si ) stands for the subset of relays that receive correctly from Si and actually perform the relaying.

information sources can be combined in a constructive way by some means. This diversity increase has some cost in spectral efficiency, which is known as the diversity versus multiplexing trade-off [14], which is explained in section III-A. In [15], the authors proposed dropping the orthogonality assumption on source and relay nodes, in what they called the Non-Orthogonal-AF Protocol (NAF), allowing the source to join the transmission in the second slot. For DF, they proposed the Dynamic DF Protocol (DDF) without fixed time slots, in which, assuming incremental redundancy codes, the relaying phase starts once there is enough information to decode (hence the term dynamic). Additionally, multiuser centralized schemas were outlined by analyzing Multiple Access Channel (NAF up-link) and Multi-User Broadcast Channel (DDF down-link) architectures. Contemporarily, Nabar, Bölcskei and Kneubühler designed space-time codes with three orthogonality options (table I) [16].They arrived at the same conclusion: allowing the source to transmit during the second phase (Mode 1) increases overall performance. This is compatible with AF and DF. The work in [12] extended that in [15], including the results of Nabar, Bölcskei and Kneubühler [16], whose policy-1-AF (NBK-AF) is equivalent to NAF. It achieves the upper performance bound for the AF family, surpassing the orthogonal AF of Laneman, Tse and Wornell [10] (LTW-AF). The proposed DDF was compared to LTW-DF. It outperforms LTW-DF and, TABLE I O RTHOGONALITY OPTIONS FOR SOURCE AND RELAY TRANSMISSIONS . Phase I II

Option 1 S→H,D S+H→D

Option 2 S→H,D H→D

Option 3 S→H S+H→D

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

4

d (r )

There are many open issues related to the theoretical aspects of cooperative diversity. We can classify them in three groups: generalization to non-trivial network topologies, relaxation of the ideal assumptions and design of theoretical protocols that achieve (or better approach) the theoretical performance bounds.

Ideal NAF LTW-DF NBK-DF DDF

2 1.8 1.6 1.4

III. PHY L AYER A. Theoretical Overview

1.2 1 0.8 0.6 0.4 0.2 0.2

0.4

0.6

0.8

1

r

Fig. 8. Diversity of NAF, LTW-DF, NBK-DF and DDF. Note that DDF is the best option, and NBK-DF is slightly better in a small region.

in almost all cases, NBK-DF (figure 8). In addition, DDF outperforms the bound of the AF family (NAF). In [17], the authors focused on the differences between static and dynamic architectures. They provided two new schemas, Extended Static DF protocol (ESDF), allowing a fixed cooperation in time division, and Extended Dynamic DF (EDDF), which is dynamic. ESDF is superior to NAF and other static protocols, and EDDF is a modification of DDF with a slightly better diversity d.

C. Conclusions and Open Issues Cooperative diversity brings the advantages of multiantenna transmission techniques to single antenna nodes within a network. Despite the fact that it is necessary to split resources to allocate relaying transmissions, this technique can still offer important benefits. The study of cooperative diversity is similar to that of classical MIMO, using diversity gain and multiplexing gain as metrics, and taking into account that multiplexing gain is restricted to values below 1. There are multiple options from information theory to improve cooperative diversity transmission systems. Considering the basic approach of a single helper that relays source information, the first option is to increase the number of helpers. In addition, the restriction of orthogonal transmission may be removed using coding techniques borrowed from multi-antenna transmitter design. Thus, it is possible to allow several helpers to transmit simultaneously, or the source to transmit fresh information while previous information is being relayed. Time-division planning represents another degree of freedom to enhance cooperative diversity, either by associating the activation of the second phase to the failure of the first transmission (incremental relaying), or by tuning the length of each phase so that the channel split is optimum.

The PHY layer of wireless systems with cooperative diversity is usually modeled as a MIMO system. Some designs aim at full diversity: For an N -antenna virtual array, the outage probability decreases asymptotically with SN R−N . Other designs set their performance criteria according to the well-known trade-off between diversity and multiplexing gain: for an N -antennae array, the multiplexing gain r and the diversity gain d, as defined in [12], are complementary and upper bounded by d(r) ≤ N + 1 − r [12][18]. Figures 5, 8 and 9 show examples by [10], [12] and [19]. By definition, the criteria are equivalent for a system with r = 0 and d(r) = N . Therefore, two techniques that achieve full diversity may not be equivalent in all regions. For example, Prasad and Varanasi combined non-orthogonality [16] (referred to as STC2) and source-orthogonal Space-Time Coding (STC) from [13] (referred to as STC1) [18] in their STC3 protocol, which optimally switches between STC1 and STC2 for a desired value of r. The trade-off gives hints on how to turn the increase in reliability into performance improvement: • Power saving: by transmitting with lower power, or equivalently with lower SNR, relying on cooperation to achieve the same BER with SN Rcoop < SN Rdirect . • Coverage extension: by extending the maximum distance (d) in the network while the BER requirements are still met for dcoop > ddirect . • AMC boosting: in an Adaptive Modulation and Coding (AMC) system, cooperators may allow the AMC to shift to a faster modulation schema. • Symbol rate boosting: by increasing the symbol rate Rs in a cooperative environment. • Error probability reduction: reducing the error probability that depends on fading. In practice, some authors have pursued a PHY layer design with the approaches that are commented on the following subsections. B. The Best-Relay Approach In [19], the authors proved that some helpers may be exempted from information relaying. If the best relay is selected, full diversity of order N can be achieved with only a single transmitting helper. This approach, called opportunistic relaying (OR), illustrates the fact that STC is not compulsory to design R/2 cooperative systems. Although the trade-off curves coincide (dOR (r) = dST C (r), figure 9), this does not imply that the outage behavior is exactly the same (i.e. PoOR 6= PoST C ), since d(r) is asymptotic. The interpretation is that, in an STC system, all replicas

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

5

d (r ) Ideal Repetition coding Non cooperative Space Time Coding Opportunistic Relaying

M +1

1 Conv. Code 4 1 2 CC

1 CCc 2

Source

Relay

Fig. 10. 12 / 41 distributed convolutional code. The bits in black (from the complementary 21 convolutional code) are meant to be transmitted by the helper.

1 1 M +1

0.5

1

r

Fig. 9. The diversity curves of OR and STC are equivalent. This means that both outage probabilities tend asymptotically to SN R−N when SN R tends to infinity, but for low SN R their behavior may still differ.

contain more information as a whole than the best replica itself. The relay selection mechanism is based on medium access contention among the potential relays. The proposed algorithm is explained in detail in section IV-B4. Other MAC-PHY proposals that also select the helpers are based on medium access contention depending on Channel State Information (CSI). Nevertheless, this has the disadvantage that, unless all potential helpers lie within mutual range, some means to avoid collision are necessary. C. Code Cooperation Most classical systems include some kind of forward error correction codes at the PHY layer, and many of them are systematic, which means that each codeword begins with the original message bits followed by some redundancy. The analogy between systematic FEC and cooperative diversity is noticeable: in both cases additional information is appended to the original message. The comparison between cooperative diversity and repetition coding pointed out in [10] leads to what is formally known as code cooperation. Srefanov and Erkip [20] elaborated on the idea of a more sophisticate coding architecture shared by the source and the relay. In the proposed schema, the source message is protected by a 1/4 convolutional code that is carefully selected so that the first two bits exactly match a 1/2 convolutional code (figure 10). The source punctures its transmitted message, sending only the first two bits. The helper is expected to receive the message correctly with a 21 Viterbi decoder due to its proximity to the source, and finally the complementary code-bits of the word are computed and transmitted. If the relay failed to decode the message, in the absence of a carrier, the source would start transmitting the remaining codebits by itself. Moreover, the destination should store these, merge the two symbol streams and perform full 14 Viterbi

decoding. Diversity emerges from the fact that, the higher the s average energy per symbol E σ 2 , the better the Viterbi decoder performs; therefore providing half of the symbols through an independent channel mitigates fading losses [21]. D. Cooperation through Network Coding In this type of coding, which was originally proposed for wired networks, the nodes forward linear combinations of the messages they receive. This schema differs from channel coding in that it operates at packet level. For example, supposing that a destination receives three coded packets in three transmission slots x, y and x ⊕ y, for a given slot error rate ρ, the third packet turns Psuccess = (1 − ρ)2 into Psuccess = [1 + 2ρ](1 − ρ)2 [22]. 1) Binary Linear Combination: Xiao et al [23] criticized the cost in energy or time of cooperative relaying, proposing instead a binary XOR function to build a combined message and allocating full channel resources to its transmission. Even though the authors remark that theirs is a network coding solution, it is very similar to code cooperation: being iA L (t) (t) the information vectors and G and G the coding and iA L R R A matrices, then the XOR operation iA L (t)GL ⊕ iR (t)G R might  GL A (t)] (t)i . be expressed as a code concatenation [iA R L GR Thus, the authors proposed an iterative decoder combining the data in both streams to recover the information. 2) Non-Binary Finite Field Linear Combination: The proposal presented in [24] and extended in [25] departs from the assumption that there exist direct and orthogonal relay channels (like in TDMA cooperative diversity), and focuses on the composition of the relay messages. For example, in a four time-slot system, the binary coding schema [23] would result in the four messages [x1 |x2 ⊕x1 |x2 |x1 ⊕x2 ], which outperform the system in [10]. However, this code is still suboptimal, because, if packets 1 and 3 get lost, no information can be recovered. Moreover, the authors show that the binary field is insufficient to achieve the full potential of cooperation, and it is suggested employing Galois fields (GF(2m )) instead. As shown in figure 11, by using GF(4), the two relay components are now different and decoding is successful for any two messages. In [24], direct and relay phases alternate, while in [25] coding is extended to super-frames, consisting of multiple

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

6

Phase-I Phase-II Phase-III Phase-IV

S2

H

x 1 ⊞ 2x 2

Phase-I Phase-II

×1

×Nt

x2 x2

S1

D

x1 x1

Fig. 13. A Nt -antenna source using a cooperative single antenna relay provides the destination with Nt + 1 realizations of the message.

x1 ⊞ x2

S1

transmitter is aware of them.

Fig. 11. GF(22 ) network code for a 4-slot TDMA. Each source composes redundancy packets by combining its message and the received message (x1 and x2 ). Both original messages can be recovered by receiving any two packets

S2

D ×Nt

Frame 1

S2

x 2 [ n]

x 2 [ n] x 1 [ n]

D

Frame 2

x 2 [ n + 1] x 1 [ n + 1]

D

x 1 [ n]

S1

x 2 [ n + 2]

x 2 [ n + 2] x 1 [ n + 2]

D x 1 [ n + 2]

S1

Frame 4

S2

⊞1,2

Frame 5 ⊞2,2

D

D

⊞1,1

S1

Frame 3

x 1 [ n + 1]

S1

S2

S2

x 2 [ n + 1]

⊞2,1

S1

Fig. 12. GF-based 6/10 network code for a 5-slot super-frame. In the first three frames the sources send six messages in total. Then, four different redundancy packets are computed and transmitted within the remaining two frames.

direct and relay slots (figure 12). By establishing an analogy with the maximum distance separation in block coding, the solution is found in well-known Reed-Solomon codes. 3) Superposition Modulation: Larsson and Vojcic [26] proposed energy sharing by simultaneous transmission using superimposed p signals. The main idea is to send the discrete signal x[n] = 1 − γ 2 x(f ) [n]+γx(r) [n], in which γ 2 < 0.5 is the energy-sharing factor. Although this is not a classical finite field operation, it is definitely an algebraic linear combination and therefore it corresponds to network coding. The concept of composed constellation allows the signal to be considered as a single digital transmission. 4) Complex Field Network Coding (CFNC): Building on the principle of PHY layer network coding, Wang and Giannakis [27] proposed substituting GF(2m ) with the classical complex field C approach of symbol constellations. The sources can transmit simultaneously because their signals are separated in the complex plane, and the relay is modeled as a separate entity. Additionally, according to [28], the CFNC signal can be enhanced with dirty paper coding, which allows the effects of interference to be minimized when the

E. Cooperation through Space-Time Coding The STC transmission technique for multi-antenna systems mentioned in section III-A exploits the signals in separate antennae as diversity. In brief, a block coding model consists of mapping a L-symbol vector ~x onto an L × N matrix G(~x) whose columns correspond to the signal to be transmitted by each of the N antennae. Thus the discrete equivalent channel could be written by blocks as ~y = G(~x)~h + w ~ Many methods have been proposed in the literature by tuning different system parameters such as delay, gain, code, antenna, etc. The following is an overview to clarify the protocols in section IV (the reader may refer to [29] -Section I.A Related Work- for a deeper description, to [30] for an approach to CSI-feedback relay phase and gain adaptation based on MIMO beam-forming, and to [31] for asynchronous code design). 1) Classical Space-Time Coding: Anghel et al [32] studied space-time coding for classical multi-antenna systems and showed that it performs adequately in a distributed virtual array system. They assumed that the relays know the code, and that each relay utilizes a unique column/antenna. Laneman and Wornell [13] elaborated on deserter helpers (selected potential helpers that do not cooperate). In an Orthogonal Space Time Block Code (OSTBC), removing a helper is equivalent to shortening the code by the corresponding column, and since the columns of the matrix are orthogonal, the shortened matrix still defines an orthogonal space-time code. Nevertheless, a column subset from an N -OSTBC is not usually the optimum (N − 1)-OSTBC. Cheng et al [33] analyzed an OSTBC base station with a helper relay (figure 13). They proved a diversity order of Nt + 1 (where Nt is the number of antennae of the transmitter), thus showing that cooperative and local space diversities may benefit mutually. 2) Issues Regarding Distributed Space-Time Coding: Some authors have reviewed STC use in cooperative diversity. Some common study areas are: • STC design techniques constrained by available helpers and cooperation systems. • Asynchronous helper tolerance: in classical MIMO architectures, all signals/antennae are synchronized. • Notification of the code to the helpers. • Behavior in case of helper desertion.

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

Receiver

7

considered different random distributions [29] and together with Sharp they extended the system to asynchronous nodes in [35].

Channel Encoder

~c

Fig. 14. Architecture of a STC random combination relay. After reception (~ x) and codification (G(~ x)), the whole STC signal for all N antennae is computed and the actual antenna is fed with a linear combination of them all (~si = G(~ x)~ci ).

Opportunistic exploitation of arriving nodes (which were not present when the virtual array was formed). • Exploitation of very large helper sets H, for which OSTBCs are unfeasible. These problems cannot be solved separately, as there are trade-offs between them. 3) Cooperative STCs with Antenna Selection: An initial approach would be the assignation of one specific antenna of the code to each relay. Nevertheless a single helper transmitting the wrong signal may degrade the overall code, and consequently DF and CRCs are employed to ensure that only the helpers that have received a good message collaborate. This means that the actual relay set is an aleatory subset D(H) of the set of helpers H. In addition, the helper set itself may be random as long as any node with successful decoding is allowed to relay without further restrictions. To avoid notifications, random antenna selection has also been proposed. Although it is an elegant solution, the diversity order is the average of the possible results in a multinomial distribution (Na , Nh ) (where Na and Nh represent the number of antennae in the code and the number of helpers, respectively). Under-exploitation of resources is likely to occur. 4) Cooperative STCs with Virtual Antenna Linear Combination: By rewriting the equivalent channel ~y = S~h + w, ~ with S = G(~x), each column ~si corresponds to the signal transmitted by a relay. Then, the symbols transmitted by each relay can be modified by rewriting S = G(~x) × C, where C ∈ MNa ×Nh is a combination matrix that builds the transmission of the relays from the rows of G(~x). With this change the relays and the code antennae are decoupled. Furthermore, if the i-th row of C has several non-zero elements, the ith relay transmits a linear combination of signals of the code antennae (implemented as in figure 14). This model allows the gain tunning and/or antenna selection by simple adjustments in C. Summing up, any arbitrary combination of antennae at the relay is possible, yielding the same properties as any local STC of choice except for a controllable degradation of the channel ~y = G(~x)~h0 + w, ~ ~h0 = C vech. Sirkeci-Mergen and Scaglione analyzed error probability and C matrix design criteria to achieve full diversity order [29], and showed that C matrix design criteria are equivalent to classic full-rank criteria for space-time block codes applied to G(~x)C. They also formulated the maximum achievable diversity order min(Nh , Na ). Yiu et al proposed that each node should have a signature vector c~i to define its unique combination of matrix columns [34]. Random C matrices were proposed in [29] and [35], such that no signaling at all is required. Moreover, Sirkeci-Mergen and Scaglione also •

F. Conclusions and Open Issues It is possible to achieve in practice the benefits predicted by information theory at the physical layer. Some techniques are completely novel, while others are based on preexistent MIMO technologies, or they are borrowed from other areas of knowledge. Some of them focus on simplicity (best relay, binary linear combination), but most of them rely on rather complex coding strategies. The relations between the many appealing proposals for the PHY layer are not clear: as there are known bounds on the gains, the superposition of two proposals does not necessarily convey the sum of their benefits. Hybrid systems and PHY layer comparisons must be taken into account. In simultaneous transmissions such as CFNC or STC, the problem of imperfect oscillator synchronization must be addressed. In the Best-Relay approach some helpers may not need to participate. Code Cooperation shows that the relationship between the two messages may be more elaborate than mere replication. Network Coding solutions show that some extra benefit may be obtained if the relays include their own information, although the MAC layer becomes more complex. Finally, Space-Time Coding is the way classic MIMO systems transmit simultaneous replicas of the same signal. There is apparently no reason for these techniques to be mutually exclusive. It is also important to stress that some proposals depend directly on current technologies and therefore so their lifespans do. IV. MAC L AYER P ROPOSALS A. General Description and Necessity Both the telecommunications operators and the end-users would reject a wireless network with cooperative diversity if the PHY layer required manual configuration. Therefore, the role of the MAC layer is essential. In addition to cooperation control, the MAC layer must support error recovery, dynamic optimization and mobility support. As far as traditional MAC design is concerned, PHY is a lower-level service that can be requested upon will. However, in cooperative diversity MAC (CDMAC), this is an oversimplification since the behavior of the cooperative diversity PHY (CDPHY) must be finely tuned. As a result, cross-layer design is a must, in which the CDPHY serves the CDMAC with communication capabilities and the CDMAC serves the CDPHY with advanced planning capabilities. B. Examples in the Literature Next we introduce different protocol proposals from diverse authors. We refer the reader to tables III to VII to compare them. Rather than discussing the particularities of each work in depth, in section IV-C we describe the MAC design problem for cooperative diversity in general.

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

H

Phase-I Phase-II Phase-III

RRT S2 (SN RS →R )

RRT S1 RRT S1

S

8

D

Direct Path Relaying

1) CMAC: Chou and Ghosh proposed CMAC in [36]. It seeks the integration of cooperative diversity with extended IEEE 802.11g wireless local area networks. They designed a MAC layer signaling that coordinates the sources and the helpers. They also proposed an extension of the protocol for multiple relays named FCMAC, based on data separation in blocks, which are separately protected with Reed-Solomon FEC and relayed by different helpers. 2) C-MAC: The schema in [37], as others below, is based on the Distributed Coordination Function (DCF) of the IEEE 802.11 standard [7]. Performance criteria aim at energy saving, featuring CDMA transmissions. It supports routing in cooperative diversity environments, elaborating on the idea that directional information may be obtained from the received signals. 3) Relay-enabled DCF (rDCF): Zhu and Cao [38] developed a triangular handshake mechanism (Figure 15) based on DCF to coordinate communications between the source, the helper and the destination. From the result of the handshake, the receiver chooses the cooperation and the rates for each hop as in other multi-rate proposals [39]. 4) Opportunistic Relaying (OR): Noting the excessive complexity of STC, this technique is designed as a single-helper system supported by relay selection [19]. Full diversity has been analytically demonstrated. According to that analysis, the best helper is selected thanks to a DCF-based protocol in which medium contention decision is initialized with the CSI. 5) Power Aware Relay Selection (PARS): Chen et al [40] studied power-aware relay selection strategies. By modifying OR criteria, PARS selects relays using an Optimal Power Allocation (OPA) algorithm. 6) CD-MAC: Moh et al [41] seeked to improve link reliability, rather than range or rate as previously mentioned protocols. In their approach, also based on DCF, the hops in the direct route are protected by DSTC (Distributed STC) cooperative backup links, which are only activated when the signal is weak. In multi-hop transmissions, each step is independently backed by cooperation (figure 16).

H2

S

D

I

Fig. 16. Cooperation in each link of a multi-hop route: The source (S) or intermediate node (I) transmission is overheard and stored by its respective helper, which forwards the transmission to the next intermediate node or the destination (D).

RC T S Fig. 15. DCF-based triangular handshake. From the first Relay Request To Send (RRTS) packet, the helper and the destination can measure the channels that separate them from the source. The relay piggybacks the measurement on a second RRTS packet, which also allows the destination to measure the H-D channel and notify the optimal strategy to the source in the Relay Clear To Send (RCTS) packet.

H1

TABLE II D IFFERENT COOP MAC VERSIONS . Ref. [45] [45] [4] [4] [47] [48] [49]

Network WiFi WiFi WiFi WiFi WiFi WiFi WiFi

Coop. Tech. Best Relay Best Relay Best Relay Source combining Source combining RSTBC RSTBC

[52]

WiMAX

RSTBC

[50] [51]

WiFi WiFi

RSTBC RSTBC

Signaling DCF-based DCF-compatible DCF-based DCF-based DCF-based DCF-based DCF-based Modified 802.16-j OFDMA DCF-based DCF-based

7) Cooperative Triple Busy Tone Multiple Access (CTBTMA): In [42], Shan et al used carrier sensing for notification. The messages are based on DCF with the addition of busy tone medium reservation inspired by [43], and Helper Busy Tones (HBT) are used to select the best helper by contention. 8) Phoenix: This protocol integrates network coding in cooperative CSMA networks [44]. For this purpose, a reactive model called CCSMA was developed, and network coding is allowed in the relays. Coordination messages are based on DCF with a three-case based handshake process to determine what messages the destination has previously stored for reverting the network code. 9) CoopMAC family: This refers to a series of related MAC layer proposals [45], [4], [8], [46], [47], [48], [49], [50], [51] for cooperative diversity, based on IEEE 802.11 DCF and an adaptation from IEEE 802.16 [52]. All proposals are similar, changing one feature at a time. The proposals are enumerated in table II. The source performs a mapping of its surroundings, and AMC allows rate adaptation (see figure 17). The criterion for route selection is total transmission time: L Rsd L L = τc + + Rsr Rrd

Tdirect = τd + Tcoop

In these equations, Tx is the overall transmission time in mode x (the minimum of the modes is chosen), τx is the initialization time, L is the packet length and Rx is the transmission rate of each hop. 10) Multi Hop Aware Cooperative Relaying (MHA-CoopRelaying): Adam et al recall that simple substitution of direct links with pairs of cooperative links would alternate SIMO and

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

9

TABLE III TABLE OF STUDIED PROTOCOLS IN TERMS OF THE NEIGHBORHOOD MAPPING SERVICE . Protocol CMAC FCMAC rDCF C-MAC

R min R2

R... R m−1 R M a x

S R > 2R min H R min

R > 2R min

D

OR PARS CD-MAC CTBTMA Phoenix coopMAC-I coopMAC-II c-coopMAC RcoopMAC coopMAX MHA-CR fairMAC

Fig. 17. AMC rate concentric ranges and 2-hop exploitation: since the helper node is closer to the source it can receive the same packet much faster and forward it again also at a high rate; hence if the two rates are greater than 2Rmin the end-to-end rate is better than the direct one. Direct Path Relaying Second relaying

S

H1

H2

I

D

Fig. 18. MHA enhancement of all links in a multi-hop route. In addition to cooperating with a particular hop in the path, the helpers that are close to their intermediate nodes may keep enhancing the following hops along the route.

MISO steps (figure 16), instead of providing the full route with MIMO transmissions [53]. As a solution, two kinds of helpers are identified: those that are only suitable for one DSTC step and those that can cooperate in two consecutive steps. Relay selection gives a bonus to double helpers, which allows a routing schema like that in figure 18. 11) Distributed Cooperative MAC for Multi-hop Wireless Networks (DCMAC): Shan et al [5] proposed a slightly modified version of OR to support multi-rate by modifying the contention mechanism with shortened timers when AMC rate adaptation is available. 12) FairMAC: Bocherer and Mathar [54] expressed their concern about the energy cost of cooperation since there is a trade-off between energy per transmitted bit and achieved throughput. Their proposal, FairMAC, allows the selection of the desired cooperation factor α ∈ (0, 1), which represents the limit of packets to be relayed for each own packet transmitted. Packet counters and a dual ACK mechanism to notify the source are used. C. Services Required for Cooperation All proposed solutions have their benefits and drawbacks, and none of them is completely superior to the others. In this

DC-MAC

Neighborhood Mapping Not necessary Not necessary Passive listening in helpers, active distribution of willing lists Regular transmission of “hello” packets and estimation of angular position Not necessary Not necessary Passive listening at the source Not necessary Not necessary Passive listening at the source Passive listening at the source Passive listening at the source Passive listening at the source Passive listening at the source with optional pilot signals for measurement Passive listening at the source Passive listening at the source with pending ACK counter Not necessary

section, we identify the common design areas and the different approaches that the protocols follow in each area. 1) Neighborhood Mapping: Most cooperative MAC protocols require an image of their surroundings, typically implemented through a neighbor table, possibly featuring estimates of link qualities and cooperation possibilities. There are many approaches, ranging from completely passive ones (such as plain neighbor discovery when transmissions are sensed) to completely active ones (such as polling the surroundings until all neighbors are detected). There also exist hybrid approaches that seek a compromise between polling overhead and missing neighbors. Table III lists the approaches in the protocols we have reviewed. In addition, it is necessary to discover the tables of the neighbors in order to design the cooperative strategies, and again trade-offs emerge. For example, in [38], the helper creates and distributes a willing list of the source-destination pairs that it considers it may enhance, and in [46] the source merely listens. Unfortunately, even if a node is sensed as being able to cooperate, it might refuse to do so. The proposal in [46] assumes optimistically that all known neighbors are collaborative and takes their lack of relaying as packet losses. Instead, cooperation willingness may be squeezed from the environment using announcement, credit systems, game theory, etc. Finally, there is concern about mapping precision [49] because cooperative diversity is devised to combat time variant fading, and network reachability is random. Bearing this in mind, neighbor table entries must be discarded as they get old. Furthermore, even the neighbor discovery mechanism itself needs to track the network changes: if fading is too variant or the nodes are too bursty, passive hearing will not suffice, whereas if everything is slow and smooth, frequent polling would result in unnecessary overhead. 2) Helper Set Design: In a network map, the cooperation possibilities of the neighbors (i.e. potential helpers) are known

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

HS1

H D3

S

Hs

Hi

Hi

Hi

Hd

S

Ii

Ii

Ii

D

H D1 D

HS2

10

H D2

Fig. 20. An ideal example of cooperative multi-hop route with diversity order d = 2 in all steps. Each node is aided by a helper, and intermediate nodes use them both for distributed reception and transmission.

Phase I: Helper group formation

Direct Path Cooperative Enhancement

I

HS1

I

H D3

H D1

I

D

S

I I

I

S

D HS2

I

Fig. 21. Cooperative enhancement of multi-hop routes. Each intermediate node in the path (or the source itself) may be assisted by helpers (not depicted). An I node knows the full list of steps and it is allowed to replace direct forwarding with cooperative transmission towards farther I nodes (or even the final destination) in order to shorten the path.

H D2

Phase II: Nt × Nr MIMO HS1

H D3

S

H D1 D

HS2

I

H D2

Phase III: Data collection Fig. 19. A three-hop cooperative system with two helper sets. First, both the source and the destination contact their own groups of helpers to perform distributed transmission/reception. Then, a full MIMO transmission is performed, and finally the destination collects data from the distributed reception group.

individually. As there may be several options for cooperation, the information on the helpers must be processed into a unique best option for each destination. However, the structure and size of the best helper set depends heavily on the performance criteria and the PHY layer. In this regard, best-relay systems should order helpers by expected performance, while systems with any form of signal combination should simultaneously consider the combination policy and the set of signals in order to maximize performance. Table IV shows how helpers are selected and grouped in the works we have reviewed. Given a method to optimize the helper set, there is an extremely broad horizon of performance metrics. There are examples based on average information rate [55], transmission time [46], covered distance [56], error probability [51], power consumption [57][54], power saving [40], battery lifespan [40], etc. Moreover, it is possible to design hybrid techniques that allow simultaneous optimization over several of these parameter domains. In any case, there is no reason for a single helper set. For example, the 3-hop proposal with a transmitter helper group and a receiver helper group of figure 19 is used in [56], saving energy by shortening the ranges of the first and third hops (which have a lower diversity and therefore trade less range improvement per power unit).

In [58], it has been shown that the best policy for wireless network routing combines multi-hop and cooperative MIMO1 . Consequently, the longer the distance in hops towards the destination, the more set-to-set hops need to be considered (Figure 20). In general, multi-hop switching for ad-hoc networks with cooperative diversity is an open research field [2]. The example in [55] is very illustrative but it relies on an excessively regular mesh topology. In [59] and [53], classical routing is used, and the hops are enhanced a posteriori (figure 21). As initial approaches they are very good, but they do not guarantee the optimality of their solution. 3) Cooperation Analysis and Decision: It is necessary to compare the non-cooperative scenario with the cooperative options in terms of proficiency and cost. The difference with helper set design is that, instead of assuming cooperation and optimizing the helpers, the cooperation mechanism itself is selected. Optimal helper sets may be given a priori to refine the analysis criteria, or they may be designed a posteriori just for the chosen technique (except for the case of direct transmission). Table V illustrates how the protocols we have reviewed switch from direct transmission to cooperative diversity when necessary. Unfortunately the design of a fair metric to compare dissimilar transmission mechanisms is difficult. As cooperative diversity was born to fight fading, a first performance metric would likely be outage probability. Fading analysis would require a good knowledge about the geometry of the environment, which would still be very expensive in a general architecture. A centralized topology with a complex base station (BS) is more favorable for outage analysis. 1 Wireless networks are traditionally classified as extensive or dense, the main difference being that in dense networks all nodes are directly reachable (interference is the main constraint) while in extensive networks only the closest neighbors are efficiently reachable (the distance in hops is the main constraint). The authors explain that, as an intermediate point, there is a regime where both constraints apply, whose optimal routing policy consists of dividing the network into directly-reachable cells, cooperating locally and performing multi-hop routing globally.

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

TABLE IV TABLE OF STUDIED PROTOCOLS IN TERMS OF HELPER SET DESIGN . Protocol CMAC FCMAC rDCF C-MAC OR PARS CD-MAC CTBTMA Phoenix coopMAC-I coopMAC-II c-coopMAC RcoopMAC coopMAX MHA-CR fairMAC DC-MAC

Helper Set Design One random helper selected by contention Random helper set selected by contention One optimal helper selected by the source Optimal CDMA helper set selected by the source One optimal helper selected by contention One optimal helper selected by contention One optimal helper selected by the source One optimal helper selected by contention One optimal helper selected by contention One optimal helper selected by the source One optimal helper selected by the source One optimal helper selected by the source Optimal helper set selected by the source Optimal helper set selected by the source One optimal helper selected by the source considering multi-hop double relaying. One optimal helper selected by the source One optimal helper selected by contention

11

TABLE VI TABLE OF STUDIED PROTOCOLS IN TERMS OF N OTIFICATION AND AGREEMENT. Protocol CMAC FCMAC rDCF C-MAC OR PARS CD-MAC CTBTMA Phoenix coopMAC-I coopMAC-II c-coopMAC RcoopMAC coopMAX MHA-CR fairMAC DC-MAC

Helper Notification No ACK No ACK / Reception of NACK DCF based triangular handshake Multiple explicit messages based on DCF DCF with SNR-proportional contention DCF with energy-proportional contention Helper ID in data packets. C-DCF DCF with BT MAC and helper contention NACK retransmission, DCF handshakes DCF with ACK, helper indication on RTS DCF without ACK, helper ID in data packets DCF with ACK, helper indication on RTS DCF without ACK, opportunistic selection Helper announcement/allocation to/by BS Helper ID in data packets. C-DCF DCF with preACK when packets are accepted and jointACK when delivered DCF with contention

TABLE V TABLE OF STUDIED PROTOCOLS IN TERMS OF COOPERATION DECISION . Protocol CMAC FCMAC rDCF C-MAC OR PARS CD-MAC CTBTMA Phoenix coopMAC-I coopMAC-II c-coopMAC RcoopMAC coopMAX MHA-CR fairMAC DC-MAC

Cooperation Decision Reactive rtx after direct link failure Reactive rtx after direct link failure Proactive selection at the source with heuristic credit system Iterative increment of cooperation gain Cooperation choice assumed Proactive, source contention against relays Reactive backup upon direct link failure Proactive relay announcement and contention Reactive in case of NACK Min tx time, proactive source Min tx time, proactive source Min tx time, proactive source Min tx time, proactive source Min tx time, proactive source Reactive backup upon direct link failure Min tx time, proactive source with pending ACK counter Proactive offer of helpers (hi packet)

A design with limited resources cannot afford a complete outage analysis of a general scenario with aggressive channels. With limited knowledge, the MAC layer should take its chances and decide. For this purpose, performance metrics should be as generalist as possible to accept any available data. There are examples that employ expected packet loss [52], number of routes [55], achievable rate [47][41], elapsed time [46][8], and so on. In addition, the distinction in [5] between proactive and reactive cooperation, depending on whether the decision is made before or after transmission in the case of primary link failure, should also be taken into account. 4) Cooperator Notification and Agreement: The helpers need to be notified of the cooperative aspects of the transmission and the receiver should be aware of them as well. The notification is even more important when the cooperative mechanism requires the exchange of initialization values. Table VI shows the notification mechanisms of the protocols in our review. There are difficulties when the helper set is chosen without information on helper willingness. Then, either

helper acknowledgement is required [42] or cooperation is optimistically assumed [49]. If no ACK returns, some policy is necessary. The receiver may contribute to the agreement, but it is necessary to decide if the transmitter, the helpers, or both send notifications to the receiver and which of them must receive the feedback from the receiver (if any). Additionally, it is necessary to decide which node is responsible of the final decisions in case of exception. There are examples of decisions taken at the source [46], the receiver [39], the helper [38], and some approaches provide different choices to the most informed peer [47]. 5) Cooperative Transmission Design: The most versatile PHY layers can be modified regarding number of relays, transmission time, code, modulation schema, etc. The MAC layer should be capable of tuning these parameters to maximize the chances of meeting the system requirements. Of course, a one-size-fits-all solution, if this exists, would avoid the complexity of dynamic set-up switching. Nevertheless, for strong adaptiveness and exhaustive capacity exploitation, some form of on-the-fly parameter design or selection is indispensable. Table VII summarizes the different approaches to cooperative transmission in the protocols we have reviewed. D. Comments on Security and Fairness Zhu and Cao [38] anticipated security issues. Provided that helper nodes are in fact other users, malicious relays might steal, modify or forge messages from fair users. Such attacks are possible in all communication systems; hence the use of cryptography, authentication and integrity checking is extended in them. The fact that cooperative diversity is a new area where these attacks can be performed should not invalidate the effectiveness of well-known protection techniques already in use. On the other hand, new forms of malice might arise in cooperative diversity services if greedy users distort the coordination mechanism to their own benefit. Regarding fairness, adequate incentives are necessary for the network users to share their resources willingly. For rapid

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

TABLE VII TABLE OF STUDIED PROTOCOLS IN TERMS OF TRANSMISSION METHODS . Protocol CMAC FCMAC rDCF C-MAC OR PARS CD-MAC CTBTMA Phoenix coopMAC-I coopMAC-II c-coopMAC RcoopMAC coopMAX MHA-CR fairMAC DC-MAC

Helper Notification Plain relaying Plain relaying Plain relaying Simultaneous CDMA relaying Not specified Not specified OSTBC of source and relay Plain relaying with AMC boosting CFNC and plain relaying Plain relaying with AMC boosting Plain relaying with AMC boosting Source combination and AMC boosting RDSTC relaying with AMC boosting RDSTC relaying with AMC boosting DSTC with multi-hop double relaying Plain relaying during fraction of time Plain relaying with AMC boosting

deployment wireless networks cooperation is natural since all nodes in the network are deployed under the same command. The same is expected for wireless sensor networks since all sensor nodes are typically deployed by the same agent. In cellular data networks, mobile devices are independent. However, they are controlled by few service providers, which improve the efficiency of their networks according to cost incentives. Those providers could offer cooperation incentives to end users by means of a well designed policy of requirements and rewards. Finally, in ad-hoc wireless data networks, the problem of cooperation incentives is hard to solve, since the users are not grouped by providers or subject to service contracts. As a consequence, cooperation should be imposed by communication standards with fairness-forcing mechanisms. E. On Interference in Cellular Networks The simulation results in [4] and [46] for the coopMAC protocols show that interfering signals are lower in cooperative scenarios, which is apparently contradictory to the interference area extension in figure 1. This is due to the influence of the time dimension, which must be taken into account. Although it is true that helper support does extend the interference area, it also reduces the time that interference power is sustained. Also, the relay and the source do not transmit simultaneously, so each influence area is active half the time. Therefore, there is not an increase of interference in the interfered areas, but a redistribution. This reasoning could be extended to any other protocol with few modifications. It has to be considered that, if the nodes were allowed to transmit simultaneously, their interfered areas would become concatenated and, as a result, interfering power would effectively increase. Nevertheless, two transmissions have twice as much energy and consequently the comparison would be unfair. If we perform an interference normalization on transmission power, using SINR metrics instead of plain Watt power units, the effect of interference extension would be canceled by the power increase and the basic results of coopMAC would be extensible to any other MAC protocol.

12

F. Conclussions and Open Issues The MAC layer is of the utmost importance for cooperative diversity feasibility, as this technique relies on identifying alternative ways of transmission within a networked context. In other words, the advantages of cooperative transmissions are only possible if the MAC layer is able to efficiently trace, classify and coordinate helpers at reasonable cost. For this purpose, most protocols rely on small control messages, such as the DCF of 802.11, or on a central controller, such as that of 802.16. There is no clear winner among the studied protocols, because their features are better or worse depending on the application domain. In this regard, there are five noteworthy services that must be taken into account when selecting or designing a cooperative diversity MAC: neighborhood mapping, helper set design, cooperation analysis and decision, cooperator notification and agreement, and cooperative transmission design. Some side effects that should be managed by the MAC layer may arise from cooperative diversity: High energy expenditure of nodes in “good” positions (since many peers use them as relays), new security concerns, and interference redistribution across cells. Most MAC protocols achieve their goals by relying on previously existing protocols and tailoring them to cooperative diversity. Thus, it may be interesting to design new protocols aiming primarily at the accomplishment of the theoretical bounds rather than at retro-compatibility. Energy expenditure deserves more analysis as the results of [54] show a trade-off between energy per bit and throughput increase, whereas the results in [46], taking into account idle energy consumption, show potential power savings due to a reduction of idle waiting time. Integration with other network aspects, like routing, must be extended, as for example most models require sourcedestination signaling, preventing transmission towards nodes out of direct range that are cooperatively reachable. In general, routing and forwarding are heavily affected because node reachability depends on the cooperation environment. Another interesting field is the effect of network saturation as most models assume that the incoming flow of packets is unlimited and therefore any acceleration of the delivery process would increase throughput. Therefore, analysis of relaxed networks with plenty of resources should be investigated to determine how cooperation affects jitter, delay, etc. V. C ONCLUSIONS Cooperative diversity reaches the performance of multiantennae systems at a lower cost, with smaller, single-antenna networked nodes. Based on recent analyses, the relation between transmission improvement and cooperation is evident even for conveniently placed nodes. In their case, the gain comes from a faster medium release from inefficient neighbors. The overall increase of throughput has a cost in energy that the peers need to consider when they decide to cooperate. In this review we have discussed the information theoretical models that support cooperative diversity. However, these

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

models are far from complete, and further work on them is expected. In addition, research in protocols must continue from new problem perspectives. Different PHY layer technologies have been reviewed, showing a wealth of possible implementations, some of them even mutually exclusive. Their development must take into account off-the-shelf non-cooperative technologies to absorb future enhancements. Simultaneous utilization of compatible PHY solutions in hybrid systems should be evaluated to investigate how far their gains hold when deployed cumulatively. The advent of completely new PHY technologies for cooperative diversity should not be discarded. The review of the MAC layer has also revealed the many proposals in the literature, especially cooperative add-ons for preexistent networks. These represent excellent proofs of concept and offer effective short-term implementation methods, and therefore work in this direction is of great practical interest. However, more challenging research should be pursued, comparing multiple PHY support or even protocols that would switch PHY techniques if an optimal solution is not found. Integration with other fields of wireless technologies must also be considered. Finally, cooperative diversity is also affected by the existence of too many architectures. This is a direct consequence of cross-layer design: once the layer frontiers are removed almost any preexistent layer-specific technique is suitable for hybrid designs. A large space of possibilities is never a drawback by itself, but, in the close future, a consensus on a subset of good solutions for industry transfer might be desirable as research goes on. ACKNOWLEDGEMENTS This research has been supported by the following grants: CALM (TEC2010-21405-C02-01), funded by MICINN, Spain, and MEFISTO (10TIC006CT), funded by Xunta de Galicia, Spain. R EFERENCES [1] A. Sendonaris, E. Erkip, and B. Aazhang, “User cooperation diversity. Part I. System description,” IEEE Transactions on Communications, vol. 51, no. 11, pp. 1927–1938, 2003. [2] P. Pathak and R. Dutta, “A survey of network design problems and joint design approaches in wireless mesh networks,” IEEE Communications Surveys & Tutorials, 2010. [3] A. Sendonaris, E. Erkip, and B. Aazhang, “User cooperation diversity. Part II. Implementation aspects and performance analysis,” IEEE Transactions on Communications, vol. 51, no. 11, pp. 1939–1948, 2003. [4] P. Liu, Z. Tao, Z. Lin, E. Erkip, and S. Panwar, “Cooperative wireless communications: A cross-layer approach,” IEEE Wireless Communications, vol. 13, no. 4, pp. 84–92, 2006. [5] H. Shan, W. Zhuang, and Z. Wang, “Distributed cooperative MAC for multihop wireless networks,” IEEE Communications Magazine, vol. 47, no. 2, pp. 126–133, 2009. [6] C. Cetinkaya and F. Orsun, “Cooperative medium access protocol for dense wireless networks,” in Proc. Third Annual Mediterranean Ad Hoc Networking Workshop - Med Hoc Net 2004, pp. 197–207. [7] A. Bachir, M. Dohler, T. Watteyne, and K. Leung, “Mac essentials for wireless sensor networks,” IEEE Communications Surveys & Tutorials, vol. 12, no. 2, pp. 222–248, 2010. [8] T. Korakis, Z. Tao, Y. Slutskiy, and S. Panwar, “A cooperative MAC protocol for ad hoc wireless networks,” Mitsubishi Electric Research Laboratories, Tech. Rep., Apr. 2007.

13

[9] P. Kumar and P. Gupta, “The capacity of wireless networks,” IEEE Transactions on Information Theory, vol. 49, no. 11, pp. 3117–3117, 2003. [10] J. Laneman, D. Tse, and G. Wornell, “Cooperative diversity in wireless networks: Efficient protocols and outage behavior,” IEEE Transactions on Information Theory, vol. 50, no. 12, pp. 3062–3080, 2004. [11] J. Laneman, G. Wornell, and D. Tse, “An efficient protocol for realizing cooperative diversity in wireless networks,” in Proc. IEEE Intl. Symp. on Information Theory 2001, p. 294. [12] K. Azarian, H. E. Gamal, and P. Schniter, “On the achievable diversitymultiplexing tradeoff in half-duplex cooperative channels,” IEEE Transactions on Information Theory, vol. 51, no. 12, pp. 4152–4172, 2005. [13] J. Laneman and G. Wornell, “Distributed space-time-coded protocols for exploiting cooperative diversity in wireless networks,” IEEE Transactions on Information Theory, vol. 49, no. 10, pp. 2415–2425, 2003. [14] L. Zheng and D. Tse, “Diversity and multiplexing: A fundamental tradeoff in multiple-antenna channels,” IEEE Transactions on Information Theory, vol. 49, no. 5, pp. 1073–1096, 2003. [15] K. Azarian, H. El Gamal, and P. Schniter, “Achievable diversity-vsmultiplexing tradeoffs in half-duplex cooperative channels,” in Proc. Information Theory Workshop, 2004. IEEE, pp. 292–297. [16] R. U. Nabar, H. Bölcskei, and F. W. Kneubuhler, “Fading relay channels: performance limits and space-time signal design,” IEEE Journal on Selected Areas in Communications, vol. 22, no. 6, pp. 1099–1109, 2004. [17] N. Prasad and M. K. Varanasi, “High performance static and dynamic cooperative communication protocols for the half duplex fading relay channel,” in Proc. IEEE Global Telecommunications Conference 2005, GLOBECOM’05. [18] N. Prasad and M. Varanasi, “Diversity and multiplexing tradeoff bounds for cooperative diversity protocols,” in Proc. Intl. Symp. on Information Theory, ISIT 2004, p. 268. [19] A. Bletsas, A. Khisti, D. Reed, and A. Lippman, “A simple cooperative diversity method based on network path selection,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 3, pp. 659–672, 2006. [20] A. Stefanov and E. Erkip, “Cooperative coding for wireless networks,” IEEE Transactions on Communications, vol. 52, no. 9, pp. 1470–1476, 2004. [21] T. Hunter and A. Nosratinia, “Diversity through coded cooperation,” IEEE Transactions on Wireless Communications, vol. 5, no. 2, pp. 283– 289, 2006. [22] Y. Chen, S. Kishore, and J. Li, “Wireless diversity through network coding,” in Proc. IEEE Wireless Communications and Networking Conference, vol. 3, pp. 1681–1686. [23] L. Xiao, T. E. Fuja, J. Kliewer, and D. J. C. Jr, “A network coding approach to cooperative diversity,” IEEE Transactions on Information Theory, vol. 53, no. 10, pp. 3714–3722, 2007. [24] M. Xiao and M. Skoglund, “M-user cooperative wireless communications based on nonbinary network codes,” in Proc. IEEE Information Theory Workshop on Networking and Information Theory, ITW 2009, pp. 316–320. [25] J. L. Rebelatto, B. F. U. Filho, Y. Li, and B. Vucetic, “Multi-user cooperative diversity through network coding based on classical coding theory,” CoRR, vol. abs/1004.2757, 2010, informal publication. [26] E. Larsson and B. Vojcic, “Cooperative transmit diversity based on superposition modulation,” IEEE Communications Letters, vol. 9, no. 9, pp. 778–780, 2005. [27] T. Wang and G. B. Giannakis, “Complex field network coding for multiuser cooperative communications,” IEEE Journal on Selected Areas in Communications, vol. 26, no. 3, pp. 561–571, 2008. [28] N. Fawaz, D. Gesbert, and M. Debbah, “When network coding and dirty paper coding meet in a cooperative ad hoc network,” IEEE Transactions on Wireless Communications, vol. 7, no. 5-2, pp. 1862–1867, 2008. [29] B. Sirkeci-Mergen and A. Scaglione, “Randomized space-time coding for distributed cooperative communication,” IEEE Transactions on Signal Processing, vol. 55, no. 10, pp. 5003–5017, 2007. [30] E. Koyuncu, Y. Jing, and H. Jafarkhani, “Distributed beamforming in wireless relay networks with quantized feedback,” IEEE Journal on Selected Areas in Communications, vol. 26, no. 8, pp. 1429–1439, 2008. [31] N. Wu and H. Gharavi, “Asynchronous Cooperative MIMO Systems Using a Linear Dispersion Structure,” IEEE Transactions on Vehicular Technology, vol. 59, no. 2, pp. 779–787, 2010. [32] P. Anghel, G. Leus, and M. Kavehl, “Multi-user space-time coding in cooperative networks,” in Proc. IEEE Intl. Conf. on Acoustics, Speech, and Signal Processing, (ICASSP’03), vol. 4. [33] S. Chen, W. Wang, X. Zhang, and Z. Sun, “Performance analysis of OSTBC transmission in amplify-and-forward cooperative relay networks,”

IEEE WIRELESS COMMUNICATIONS VOL. X, NO. X, FEBRUARY 2011

[34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53]

[54]

[55]

IEEE Transactions on Vehicular Technology, vol. 59, no. 1, pp. 105–113, 2010. S. Yiu, R. Schober, and L. Lampe, “Distributed space-time block coding,” IEEE Transactions on Communications, vol. 54, no. 7, pp. 1195–1206, 2006. M. Sharp, A. Scaglione, and B. Sirkeci-Mergen, “Randomized cooperation in asynchronous dispersive links,” IEEE Transactions on Communications, vol. 57, no. 1, pp. 64–68, 2009. C. Chou and M. Ghosh, “Cooperative communication MAC (CMAC)-a new MAC protocol for next generation wireless LANs,” in Proc. Intl. Conf. on Wireless Networks, Communications and Mobile Computing, 2005, vol. 1. A. Azgin, Y. Altunbasak, and G. AlRegib, “Cooperative MAC and routing protocols for wireless ad hoc networks,” in Proc. IEEE Global Telecommunications Conference 2005, GLOBECOM’05, vol. 5. H. Zhu and G. Cao, “rdcf: A relay-enabled medium access control protocol for wireless ad hoc networks,” IEEE Trans. Mob. Comput., vol. 5, no. 9, pp. 1201–1214, 2006. G. Holland, N. Vaidya, and P. Bahl, “A rate-adaptive MAC protocol for multi-hop wireless networks,” in Proc. 7th Annual International Conference on Mobile Computing and Networking, ACM, 2001, pp. 236–251. Y. Chen, G. Yu, P. Qiu, and Z. Zhang, “Power-aware cooperative relay selection strategies in wireless ad hoc networks,” in Proc. IEEE 17th Intl. Symp. on Personal, Indoor and Mobile Radio Communications, 2006. S. Moh, C. Yu, S. Park, H. Kim, and J. Park, “CD-MAC: Cooperative diversity MAC for robust communication in wireless ad hoc networks,” in Proc. IEEE Intl. Conf. on Communications, 2007, ICC’07, pp. 3636– 3641. H. Shan, P. Wang, W. Zhuang, and Z. Wang, “Cross-layer cooperative triple busy tone multiple access for wireless networks,” in Proc. IEEE Global Telecommunications Conference 2008, IEEE GLOBECOM 2008, pp. 4992–4996. Z. Haas and J. Deng, “Dual busy tone multiple access (DBTMA)-a multiple access control scheme for ad hoc networks,” IEEE Transactions on Communications, vol. 50, no. 6, pp. 975–985, 2002. E. Fasolo, A. Munari, F. Rossetto, and M. Zorzi, “Phoenix: a hybrid cooperative-network coding protocol for fast failure recovery in ad hoc networks,” in Proc. 5th Ann. IEEE Communications Society Conf. on Sensor, Mesh and Ad Hoc Communications and Networks 2008, SECON’08, pp. 404–412. P. Liu, Z. Tao, and S. Panwar, “A cooperative MAC protocol for wireless local area networks,” in Proc. IEEE Intl. Conf. on Communications, 2005, ICC 2005, vol. 5, pp. 2962–2968. P. Liu, Z. Tao, S. Narayanan, T. Korakis, and S. S. Panwar, “CoopMAC: A cooperative MAC for wireless LANs,” IEEE Journal on Selected Areas in Communications, vol. 25, no. 2, pp. 340–354, 2007. F. Liu, T. Korakis, Z. Tao, and S. S. Panwar, “A MAC-PHY cross-layer protocol for ad hoc wireless networks,” in Proc. WCNC, IEEE, 2008, pp. 1792–1797. F. Verde, T. Korakis, E. Erkip, and A. Scaglione, “On avoiding collisions and promoting cooperation: Catching two birds with one stone,” in Proc. IEEE 9th Workshop on Signal Processing Advances in Wireless Communications 2008, SPAWC 2008, pp. 431–435. P. Liu, Y. Liu, T. Korakis, A. Scaglione, E. Erkip, and S. S. Panwar, “Cooperative MAC for rate adaptive randomized distributed space-time coding,” in Proc. IEEE Global Telecommunications Conference 2008, IEEE GLOBECOM 2008, pp. 4997–5002. P. Liu, C. Nie, E. Erkip, and S. Panwar, “Robust cooperative relaying in a wireless LAN: cross-layer design and performance analysis,” in Proc. IEEE Global Telecommunications Conference 2009, GLOBECOM 2009. F. Verde, T. Korakis, E. Erkip, and A. Scaglione, “A Simple Recruitment Scheme of Multiple Nodes for Cooperative MAC,” IEEE Transactions on Communications, vol. 58, no. 9, pp. 2667–2682, 2010. C. Nie, P. Liu, T. Korakis, E. Erkip, and S. S. Panwar, “CoopMAX: A cooperative MAC with randomized distributed space-time coding for an IEEE 802.16 network,” in Proc. IEEE ICC 2009. H. Adam, C. Bettstetter, and S. Senouci, “Multi-hop-aware cooperative relaying,” in Proc. IEEE 69th Vehicular Technology Conf. 2009, VTC Spring 2009. G. Bocherer and R. Mathar, “On the throughput/bit-cost tradeoff in CSMA based cooperative networks,” in Proc. 2010 Intl. ITG Conf. on Source and Channel Coding (SCC), IEEE. H. Gharavi, B. Hu, and N. Wu, “A Design Framework for High-Density Wireless Ad-Hoc Networks Achieving Cooperative Diversity,” in Proc. 2010 IEEE Intl. Conf. on Communications (ICC).

14

[56] H. Shen, H. Yang, B. Sikdar, and S. Kalyanaraman, “A distributed system for cooperative MIMO transmissions,” in Proc. IEEE Global Telecommunications Conference 2008, IEEE GLOBECOM 2008. [57] X. Bai, D. Liu, G. Yue, and H. Wu, “Joint Relay Selection and Power Allocation in Cooperative-Diversity System,” in Proc. 2010 IEEE Intl. Conf. on Communications and Mobile Computing (CMC), vol. 2, pp. 361–365. [58] A. Ozgur, R. Johari, D. Tse, and O. Lévêque, “Information-theoretic operating regimes of large wireless networks,” IEEE Transactions on Information Theory, vol. 56, no. 1, pp. 427–437, 2009. [59] G. Jakllari, S. Krishnamurthy, M. Faloutsos, P. Krishnamurthy, and O. Ercetin, “A framework for distributed spatio-temporal communications in mobile ad hoc networks,” in Proc. IEEE INFOCOM 2006.