Ann. Telecommun. (2013) 68:485–501DOI 10.1007/s12243-012-0330-y

DEC-MAC: delay- and energy-aware cooperativemedium access control protocol for wirelesssensor networks

Mohammad Helal Uddin Ahmed ·Md. Abdur Razzaque · Choong Seon Hong

Received: 2 April 2012 / Accepted: 17 September 2012 / Published online: 19 October 2012© Institut Mines-Télécom and Springer-Verlag France 2012

Abstract This paper deals with two critical issuesin wireless sensor networks: reducing the end-to-endpacket delivery delay and increasing the network life-time through the use of cooperative communications.Here, we propose a delay- and energy-aware coop-erative medium access control (DEC-MAC) protocol,which trades-off between the packet delivery delay anda node’s energy consumption while selecting a cooper-ative relay node. DEC-MAC attempts to balance theenergy consumption of the sensor nodes by taking intoaccount a node’s residual energy as part of the relayselection metric, thus increasing the network’s lifetime.The relay selection algorithm exploits the process ofelimination and the complementary cumulative dis-tribution function for determining the most optimalrelay within the shortest time period. Our numericalanalysis demonstrates that the DEC-MAC protocol isable to determine the optimal relay in no more thanthree mini slots. Our simulation results show that theDEC-MAC protocol improves the end-to-end packet

M. H. U. Ahmed · C. S. Hong (B)Department of Computer Engineering,College of Electronics and Information,Kyung Hee University (Global Campus),Suwon, South Koreae-mail: cshong@khu.ac.kr

M. H. U. Ahmede-mail: helal@networking.khu.ac.kr

Md. A. RazzaqueDepartment of Computer Science and Engineering,University of Dhaka, Dhaka, 1000, Bangladeshe-mail: razzaque@cse.univdhaka.edu

delivery latency and the network lifetime significantlycompared to the state-of-the-art protocols, LC-MACand CoopMAC.

Keywords End-to-end packet delay ·Network lifetime · Cooperative communication ·Process of elimination · Complementary cumulativedistribution function

1 Introduction

With the advancement of microelectromechanical sys-tems technology, sensors are becoming increasinglysmall and economical. Since, they are typically battery-operated, thus it is very difficult to change batteryfrom a tiny node and also difficult to replace a nodefrom any hostile environment, making sensor nodesenergy-constraint. Therefore, lifetime maximization isone of the primary concerns for wireless sensor net-works (WSNs). End-to-end data packet delivery delayis another important constraint in many applicationsof wireless sensor networks [1, 2] such as tracking thetarget in a battlefield, critical patient monitoring in ahospital, and controlling temperature in industrial set-tings [3]. Hence, energy and data packet delay shouldbe jointly considered in designing protocols for criticalapplications.

Interference and signal loss due to distance andfading severely reduce the data delivery performanceof wireless networks. Multiple input multiple output(MIMO) is able to significantly improve the transmis-sion quality [4]; however, it is not possible to designsensor nodes with MIMO ability due to their small sizeand power constraints [5, 6]. However, a cooperative

486 Ann. Telecommun. (2013) 68:485–501

communication technology is able to introduce thisability in very small wireless sensor networks. Coop-erative communication exploits the wireless broadcastadvantage by using neighbor nodes as relays [4]. Aneighbor node acting as a relay node retransmits thepacket in order to facilitate better reception at thedestination end [5, 7]. There are two approaches for im-plementing cooperative communications: proactive andreactive. In the proactive approach, the relay is selectedprior to the direct transmission; in reactive approach,the relay is selected only when a direct transmission hasfailed.

Network lifetime maximization is the primary issue,and metrics such as delay are the secondary issue [8, 9]in current wireless communications research. Most ex-isting research efforts [4, 5, 8, 10] on cooperative wire-less sensor networks have focused on approaches thatattempt to maximize a network’s lifetime. Other qualitymetrics, such as the delay, throughput, and jitter, havereceived little attention. A few research efforts [9, 11]have considered delay and energy, thus developing dutycycle-based medium access control (MAC) protocols;primary results show a wide range of trade-offs betweendelay and energy consumption. Both of the publishedpapers provide solutions for lifetime maximization con-sidering primary and secondary issues that apply trafficadaptive duty-cycle MAC.

The majority of the cooperative solutions [3–5, 12–14] select the relay on the basis of channel state infor-mation. All of these protocols focus solely on energyefficiency, causing the network to partition into multi-ple parts. Since nodes with better channel conditionsfrequently participate in cooperation, they finish theirenergy earlier, thus creating energy holes [15] in thenetwork. Energy holes partition the network and ham-per the ability to acquire critical data from the targetlocation, causing an early network lifetime termination.Alternatively, the nodes with poor channel conditions,rarely participate in cooperation; energy may remainunused for those nodes resulting in unbalanced powerconsumption issues within the network. An analysis in[16] shows that energy left unused can reach up to 90 %of total initial energy of the network at the end of anetwork’s lifetime for unbalanced power consumption.

Recently, some papers [17, 18] have proposedenergy-aware cooperative MAC rather than usingchannel state information; selecting the relay with thehighest residual energy helps to balance the energyconsumption among the nodes. Therefore, their pro-tocols reduce wasted energy and increase the networklifetime. However, delays may be incurred due to thepoor channel conditions of the relay with the highestresidual energy.

In this paper, the DEC-MAC protocol is designedwith consideration for both delay and residual energy.Earlier discussions revealed that a relay having a higherresidual energy might introduce additional delay; al-ternatively, the relay with better channel conditionsforces a reduction in the network lifetime due to thenetwork partitioning problem, thus creating a criticalproblem for the network. To overcome this problem,the DEC-MAC protocol seeks to optimize the trade-off between residual energy and delay; we developeda mechanism considering both the residual energy anddelay in order to optimize the relay selection. Thismechanism selects an optimal relay that enhances tokeep balance the energy consumption among the nodesof network leading to longer network lifetime and lessend-to-end delay.

No research has been conducted on cooperativeMAC with consideration for both residual energy anddelay. Our proposed DEC-MAC protocol maximizesthe network lifetime and reduces the packet deliverydelay, thus it is applicable in many real-time applica-tions. The primary contributions of our research aresummarized as follows:

– We consider a framework that enables balancedenergy consumption among network nodes.

– We develop an algorithm to quickly select the opti-mal relay.

– We propose a cooperative MAC that increases thenetwork lifetime.

– Finally, we simulate our proposed protocol andevaluate the performance of the network. Simula-tion results demonstrate that the proposed DEC-MAC protocol is able to significantly increase anetwork’s lifetime.

The remainder of this paper is organized as follows.We first review related research in Section 2 and thendiscuss the network model and assumptions in Section 3.Our novel delay- and energy-aware medium accesscontrol protocol DEC-MAC is described in Section 4.Next, we analyze the number of required mini slots,end-to-end packet delivery delay, and network lifetimein Section 5. We present performance evaluation andsimulation results in Section 6. Finally, we conclude thepaper in Section 8 with some remarks and future scopeof works.

2 Related research

Network lifetime maximization is a great challenge forvery small energy-constrained sensor nodes. Alterna-tively, delay is another important issue in the timely

Ann. Telecommun. (2013) 68:485–501 487

delivery of data. A large number of MAC and routingprotocols for WSNs aim to increase energy efficiency[4, 5, 7, 8, 10] and to decrease the end-to-end delay[1, 2, 19, 20]; the cited references either pay attentionto energy efficiency or to delay minimization. Only acouple of studies reporting noncooperative protocols[9, 11] have included investigations of delay and energytogether in WSNs; both of these protocols have trade-offs between energy efficiency and delay in order tooptimize the lifetime and end-to-end packet delay.

Cooperative communication has the ability to pro-vide better performance in the domain of energyefficiency and end-to-end packet delay. Cooperativeprotocols [4, 12, 13, 21] work on increasing throughputor reliability by selecting their relays on the basis ofchannel state information. The cooperative protocolsin [5, 6, 17, 22] propose to maximize the lifetime onnetworks or to make more energy-efficient the pro-tocol. All of these papers concentrate on improvingthe lifetime of networks and transmission energy. Thecooperative protocols in [2, 20, 23] work with QoSin the domain of end-to-end delay or reliability. Wepropose a trades-off between energy consumption andend-to-end delay to maximize the network life time andoptimize the end-to-end delay.

A cross-layer cooperative protocol is proposed in[24] where cooperative diversity at the physical layerreflects at the networking layer. Their investigationshows that end-to-end delay and throughput increasesignificantly due to their cross-layer cooperation. In[25], the authors propose two-relay-based cooperationwhere second relay is used as back-up relay. If the firstrelay fails to transmit the packet then the back-up relayretransmits it, thus decreases packet error probability.The paper shows that two-relay-based cooperation im-proves average delay and throughput considerably.

The cooperative MAC (CoopMAC) [4] is a highlycited cooperative wireless MAC protocol, which selectsrelay based on channel state information stored inCoopTable, i.e., a node providing good channel con-dition will be selected as the relay. Therefore, a singlenode might participate frequently in cooperation, caus-ing nonuniform energy consumption among the nodesand therefore, the energy remains unused when the net-work expires. The CoopTable is updated overhearingthe ongoing transmission only. Since the channel statesvary over time, the CoopTable may not remain fullyupdated if some nodes do not transmit any packetsfor a while. Therefore, the best relay of CoopTablemight not be the best when it is required. Moreover,when the sender seeks for help it might not be ableto help due its poor performance for that moment. Asfor example, the relay might be under the interference

range of a neighbor, but the sender has no knowledgeabout the interference. The relay cannot help while theneighbor’s transmission is ongoing, although the helperis the best as per the CoopTable of sender. This effectincreases the number of collisions and retransmissionsand thus prolongs the average end-to-end packet delay.

Recently, an interesting single-phase cooperativeprotocol was proposed in link-utility-based cooperative(LC)-MAC [21] that considers both the throughputand energy consumption. In this protocol, the relay isselected based on a higher link rate and lower powerconsumption. In fact, both these criteria influence theselection of a relay with better channel conditions. Theimpact of this effect is to increase the leftover en-ergy in the network when it expires. In addition, extramessages such as group indication (GI) and memberindication (MI) as well as additional contention periodsare used in relay selection. These extra overheads serveto increase the power consumption as well as expo-nentially increase the number of network collisions,thereby increasing the end-to-end delay.

Our proposed DEC-MAC overcomes the problemswith CoopMAC and LC-MAC. The relay node is se-lected in a distributed approach, unlike CoopMAC,and does not require any storage overhead. The relaycan be selected dynamically using estimated metricvalue. In DEC-MAC, the potential candidates partic-ipate in a relay selection process, exchanging controlmessages that resolve intra-helper collisions and hiddennode problems. The relay is selected by considering aweighted metric of residual energy and delay, whereasCoopMAC considers link quality and LC-MAC consid-ers link rate and transmission power. Since the samerelay may be repeatedly selected in CoopMAC and LC-MAC, unbalanced power consumption may occur inthe network; the proposed DEC-MAC protocol pro-vides balanced energy consumption as well as reducedend-to-end delay.

3 Network model and assumptions

We assume a large wireless sensor network where thenodes are uniformly distributed. We also assume thateach source node has multi-hop routing paths to thedestination sink. At each hop, a pair of communicationnodes may exploit cooperative communication for per-formance improvement. In a single-hop environment,between a sender S and a destination D, multiplehelper nodes exist that are able to participate in thecooperative process. When one node helps anothernode in a cooperative transmission process, we call thatnode a relay node.

488 Ann. Telecommun. (2013) 68:485–501

Each node is able to compute its own weightedaverage metrics value from its own residual energy andmeasured average packet delay. The process of elimina-tion algorithm is used to select an optimal relay shownin Algorithm 1. The node with the highest metrics valueis considered to be the optimal relay. This optimal relayis selected by the receiving node D.

It is assumed that all of the nodes have an equaltransmission range with an equal initial energy, Eini. Thesink may reside anywhere within the network. Also, weassume that all of the packets are of equal size, and allof the nodes have buffers of equal size. A node is saidto be dead when its residual energy becomes less thanits threshold value Ethr; in this case, the node is unableto acquire any data from the environment. One of themost important assumptions is that the entire networkwill expire when the first node becomes nonfunctional.In this paper, we assume that residual energy remains asleftover energy in m number of nodes out of η numberof total nodes when the network expires. The residualenergy in each of m number of nodes is greater thanthe threshold energy (Eres > Ethr) while the networkremains functional.

We consider an additional queue, termed the relayqueue, at the MAC layer of each node; the relay queuehandles relayed packets at the node. In a cooperativenetwork, since a node may act as both a source anda relay, problem arises when a node act as relay. Thisis because a new packet arrives that the node mustrelay before transmitting its own packet. In coopera-tive communication, a relay node should immediatelyretransmit (after one short interframe space (SIFS))just after the transmission of sender to support themaximum ratio combining (MRC) technique [5] at thedestination . If the relay node has its own packet atthe head-of-the-line, in that case, it is unable to re-lay others packet immediately. In order to solve thisissue, we consider the fact that each node maintainstwo transmission queues: one is the data queue for itsown outgoing data and the other is the relay queue tobuffer only overheard packets requiring relay when it isrequired. A higher priority is given to the helper queuesuch that the cooperative transmission can be executedat the required time.

4 Proposed DEC-MAC protocol

4.1 Basic concept

In this section, DEC-MAC protocol is proposed,which follows the design principle of IEEE 802.11.The proposed protocol considers two scenarios of

communication. In the first scenario, a node havingdata to be transmitted operates in a two-phase cooper-ative transmission mode. In the first phase, an attemptis made to select a relay. If a relay is successfullyselected, then the protocol proceeds to the secondphase, the data transmission phase, wherein the senderS broadcasts the data packet, the relay overhears thepacket, and the relay then retransmits the packet to thereceiver D.

The second scenario depends on the first phase of thefirst scenario. If there is a failure to select a relay thenthe sender goes for the traditional direct transmissionmode. For this case, the sender S receives a feedbackmessage from the receiving node D at the end of firstphase. In response, node S transmits directly to node D.

Sender S, when it has data to transmit, sends arequest-to-send (RTS) message to the receiver D, andin response to that, D sends request-to-help (RTH)message. Here, each node knows its own weightedmetrics value, as described in Section 4.2. The nodes,hearing RTS and RTH messages, compete as candidaterelay nodes by sending an interested-to-help (ITH)message as part of the relay selection process. Aftercompletion of the relay selection process, the senderS sends the data packet to receiver D, and the relayoverhears and then retransmits the data packet in thesecond phase. The basic steps are shown in Fig. 1.

The proposed protocol employs a judicious relayselection process in cooperative communication in or-der to maximize the network lifetime and minimizethe end-to-end packet delivery delay. Relay selectionis one of the critical issues in the design of a cooper-ative communication protocol. Which node is the bestrelay? How can the best relay be selected? How canthe selection process time be decreased? These are thecommon challenges for cooperative communication.The following subsections address these issues in detail.

4.2 Metric design

In this section, an average weighted metrics (W) isformulated, consisting of the residual energy (Eres) andthe average delay D for successful packet transmissionof a node in order to select the most optimal relay.Each node is able to calculate its own residual energyby subtracting the energy required for the immediate

RTS RTHRelay

Selection

SenderTx

ACKTime

ReceiverTx

Data Data

RelayTx

Dest. Tx

SenderTx

Fig. 1 Basic cooperative transmission process

Ann. Telecommun. (2013) 68:485–501 489

transmission or reception based on the previous valueof Eres. The average packet transmission delay of a nodecan also be measured by using the link data rate, packetsize, and link error rate.

In DEC-MAC, a source node S sends data to Dthrough any relay rn from n number of candidate re-lays. Different links exist from S to rn and rn to D.The transmission quality of a wireless link typicallydepends on shadowing, fading, interference, and noise;in particular, the error probability for a specific nodesignificantly increases due to these parameters. There-fore, cooperative transmission delay D varies and isdependent on the links from S to rn and rn to D. Relayselection thus has a significant impact on single-hopdelays.

D is calculated by summing the expected transmis-sion time (ETT) from S to rn, ETTSrn , and that from rn

to D, ETTrn D, as given by Eq. 1:

D = ETTSrn + ETTrn D (1)

The definition of ETT is given by

ETT = ETX × S

B(2)

where S is the packet size and B is the link data rate.The expected retransmission count ETX depends onpacket error rate. The packet loss includes the packetdrops for link errors. Thus, the ETX incorporates theeffects of wireless link loss [26], and the increased linkloss forces to increase the expected transmission time(ETT). This local delay has a significant impact onthe end-to-end packet delay. Therefore, we incorporateETT as a metric parameter that attempts to select arelay with lower link losses.

In the proposed DEC-MAC, a candidate relay nodern, having the highest metric value Wn among the avail-able candidate nodes n, is considered the optimal relay.The average weight Wn of an optimal relay is calculatedsuch that

Wn = w1Eres

Eini+ w2

Dmax − D

Dmax(3)

where w1 and w2 are smoothing factors for the sub-metric residual energy and delay, respectively, andw1 + w2 = 1, 0 < Wi ≤ 1. Determining the weightingvalues w1 and w2 is critical. By performing extensivesimulation experiments, we were able to determine thatbetter performance was achieved by setting w1 = 0.3and w2 = 0.7.

In Eq. 3, the value of Wn will be at the maximumof Eres, i.e., Eres = Eini and D will be the minimum,i.e., Dmin = 0; 0 < D ≤ Dmax. In practice, the packetdelay at any node rn can never be equal to zero. The

maximum packet delay Dmax will be when both of theexpected transmission times for ETTSrn and ETTrn D

are at their respective maximum values. In fact, themaximum value for ETT is when the channel conditionis very poor i.e., the link loss is very high, and thenumber of retransmissions (ETX) rises to a high value(maximum retry limit), which in turn decreases themetrics value Wn for the node. The proposed DEC-MAC protocol searches for the candidate having thehighest value of Wn.

4.3 Relay selection

Relay selection is a vital issue for cooperative commu-nication. It has been proven that a good helper is ableto significantly improve network performance [14], incontrast to an inferior helper, which might considerablydegrade the performance. In the following subsections,we present the DEC-MAC method of selecting theoptimal relay within the shortest time period.

4.3.1 Relay selection algorithm

In the proposed DEC-MAC protocol, a proactive relayselection approach is used, where the relay is selectedprior to the start of data transmissions. Proactive relayselection has been proven to be more energy-efficientthan its reactive counterpart [12] since, in the formerapproach, only the selected candidate relay node needsto expend energy to overhear the transmission of thesource.

We consider the fact that there are n candidaterelay nodes in between sender S and receiver D. Whensender S has data to transmit, it sends a RTS; in re-sponse to the RTS, the receiving node D sends a RTHafter which the relay selection process starts. Activenodes, listening to the RTS and RTH messages, cancompete as a relay. Each node n, based on its weightedmetrics value Wn, participates in the relay selectionprocess. In this process, a time slot T is assigned totransmit a single data burst, as shown in Fig. 2. Eachtime slot is divided into two parts: Tr and Td. The first

Minislot

Data

TTr

Td

T

Fig. 2 Time slots for two data bursts

490 Ann. Telecommun. (2013) 68:485–501

part Tr is used for relay selection; the second part Td isused to transmit data packets. Tr is again divided intoseveral mini slots (time slots). A candidate relay node nsends an ITH message at the beginning of a mini slot ifits weighted metrics value Wn satisfies condition WL ≤Wn ≤ WH, where WH and WL are the higher and lowerthresholds of the metrics value defined by the receivingnode D, respectively. In response, a feedback messageFITH (feedback of ITH) is sent by the receiving nodeD in the same mini slot.

If there exists more than one candidate node whoseweighted metrics value that remains within the rangeWL and WH in a slot, then an ITH message collisionoccurs in that slot; in this case, the receiving nodeincreases the lower threshold WL, as shown in Fig. 3,using function WL = upper(WL, WH). This strategy re-duces the collision probability and also increases theprobability to find the optimal relay in the next minislot. In order to make the relay selection process faster,the process of elimination (PE) is employed. In PE, ifa node i hears an ITH message from any other nodej and Wi < W j, then the node i will not participatein the relay selection process for the subsequent minislots, thus ensuring a further reduction in the collisionprobability.

If there is no candidate relay between WL and WH,meaning an empty slot, then the feedback messageis msg = empty. Both of the values for WL and WH

are lowered and new values are calculated such thatWH = WL and WL = lower(Wlower, WL), as shown inFig. 3. Here, Wlower is the lower threshold above whichcertainly the optimal relay exists. If an empty slot isfound after a single-node slot, it means that the optimalrelay has been found, and the relay search process isterminated. If only one candidate node transmits in amini slot, then it is called a single-node slot. In this case,

WlowerWL WH

Wlower WL WH

WL WH

Wlower WL WH

Wlower

Lower bound shift to right after collision or single node transmission

Lower and higer bound shift to left after having empty slot

If slot is non-idle

If slot is idle

One or more relay’s metrics remain in this range

No relay’s metric remains in this range

Fig. 3 Upper and lower threshold update procedure

a greedy strategy is followed in order to find an optimalrelay. Here, we consider the possibility that a bettercandidate node might exist above the upper threshold.In this case, we assume that WL = WH and WH = 1 forthe next mini slot. If again it is found to be a singlenode-slot, then the relay of this mini slot is consideredas an optimal relay and stop the relay searching process.

The upper and lower threshold values can be deter-mined by the following functions [27]:

upper(WL, WH) = F−1c

(Fc(WL) + Fc(WH)

2

)

and

lower(Wlower, WL) = F−1c

(Fc(Wlower) + Fc(WL)

2

)

The proposed optimal relay selection algorithm,DEC-MAC, is presented in Algorithm 1. If more thanone candidate node transmits an ITH message, then afeedback message multi is sent back by D, indicatinga collision. An empty message is sent when no nodetransmits any ITH message in a slot. The receiver Dsends msg_coop when a relay is finally selected. If noneof the relays is interested to help, then the feedback ismsg_dir for direct transmission.

4.4 Terminology and challenges

Some terminology and determination process of fewparameters are detailed in this section.

Mini slot A mini slot is a duration assigned to sendITH message by a candidate relay and to receive afeedback message from receiver. The duration of eachmini slot β is equal to the round-trip time required for anode to transmit an ITH and receive a FITH message.If K mini slots are required to find a relay, then the totalrelay selection time is Tr = Kβ.

Number of candidate relays Each node is able to com-pute its distance from a neighboring node using theEuclidian method [28]. The expected number of can-didate relays can be expressed as n = ξπ

(R2

)2, where ξ

is the node density and R is the distance between thesender S and the receiver D.

Initialization We apply the complementary cumula-tive distribution function (CCDF) to determine theinitial lower threshold WL [27]. At first, we assume thatthe weight of only one node out of n nodes between Sand D is above WL and transmits with probability 1/nin the first mini slot. The weight of all of the other nodes

Ann. Telecommun. (2013) 68:485–501 491

Algorithm 1 DEC-MAC algorithm

1: Initialization: Initialize f lag = 0, WL = F−1c ( 1

n ),WH = 1, Wlower = 0;

2: while WH − WL > constant do3: if msg = multi then4: Wlower = WL; WL = upper(WL, WH);5: if msg = single then6: f lag = 1;7: WL = WH ; WH = 1;8: if msg = single then9: exit loop

10: end if11: end if12: else13: if msg = empty then14: if f lag = 1 then15: exit loop16: else17: WH = WL; WL = lower(Wlower, WL);18: end if19: end if20: end if21: end while22: if f lag = 1 then23: Msg = msg_coop24: else25: Msg = msg_dir26: end if

are below the initial lower threshold WL. The CCDFof the average weighted metrics of the nodes can beexpressed as Fc = P(W > WL) = 1

n . The value of WL

is determined by using inverse CCDF such that

WL = F−1c

(1

n

). (4)

Upper threshold WH Setting the value of the upperthreshold WH for each data burst is another challengingissue for the proposed DEC-MAC protocol. Initially,we assumed that if the weight of each candidate relaysatisfies the condition 0 ≤ Wn ≤ 1 then it immediatelytransmits an ITH message. According to Eq. 3, theinitial upper threshold should be 1, if Eres = Eini andD = 0. However, this is impractical since Eres decreasesover time, and the transmission delay can never be zero.Therefore, the value of WH is much less than 1 as timegoes on and hence, if we always set the upper thresholdto 1, then the probability of an empty slot will increaseover time. As a result, the required number of minislots might be increased, thereby increasing the relayselection time (Tr) . To solve this problem, the initial

upper threshold is updated by each node according tothe following equation:

Wini_h = w1max

(maxE source

res , maxE destres

)Eini

+ w2(Dmax − Dmin)

Dmax(5)

The term maxE sourceres denotes the node having the

highest residual energy in the transmission range of thesource node. Similarly, the term maxE dest

res denotes thenode having the highest residual energy in the trans-mission range of the destination node. The entire termmax(maxE source

res , maxEdestres ) represents the node contain-

ing the highest residual energy within the transmissionrange of both the source and destination. Dmin is theminimum transmission delay TSD = T min

SD .

Feedback message In each mini slot, potential can-didate relay nodes receive a FITH message from thereceiver D. The feedback message contains the direc-tion of action and the values of the upper and lowerthresholds for the next mini slot.

Algorithm termination condition The loop will repeatuntil the condition WH − WL > constant is satisfied.An appropriate constant is carefully selected in orderto avoid having an excess number of mini slots or aninfinity loop. It is evident that the lower the value ofthe constant is, the better the relay; however, the relayselection requires additional time. An optimal valueof the constant is investigated based on simulation,i.e., constant = 0.01 but is not the main focus of ourresearch.

4.5 Protocol operation

Figure 4 explains our proposed IEEE 802.11 distrib-ution coordination function (DCF)-based DEC-MACprotocol operation details. A sender S initiates its trans-mission by sending a RTS packet after completing,its distributed inter-frame space (DIFS) and back-offperiods . The receiver D responds by sending a RTHmessage in order to initiate cooperative communica-tion; the relay selection process is started after the SIFS.The RTH message contains the initial lower (WL) andupper thresholds (WH) of the mini slot. A neighbornode n, hearing both RTS and RTH messages andsatisfying WL < Wn ≤ WH sends an ITH message. TheITH message contains the node’s ID and metrics value.After the ITH message, the candidate relays receive afeedback message from the receiver D.

At the end of the relay selection phase, the receivernode sends back either msg_coop or msg_dir. This final

492 Ann. Telecommun. (2013) 68:485–501

Fig. 4 DEC-MAC protocoloperation diagram

Source

Destination

Relay-1

Relay-2

Relay-i

Data

Data

SIF

S

SIF

S

Mini slots to select relay

NA

V

DIF

S

RT

S

Bac

k-of

f

ACK

NA

VN

AV

NA

VN

AV

NAV

NAV

Time

Time

Time

Time

Time

NAVRTS

SIF

S

SIF

S

SIF

S

msg_coop/msg_dir

RT

H

NAVRTS

NAVRTS

feedback message is required by the sender to decidewhether to transmit the data packet in cooperativemode or in direct transmission mode. The feedbackmessage msg_coop from receiver D indicates that theoptimal relay has been selected. This message is lis-tened to by the sender S and all candidate relays. Thus,the other candidate relays stop their participation inthe relay selection process. After receiving msg_coop, Ssends data to receiver D after the SIFS interval. The re-lay node overhears and decodes this data and forwardsit to D after another SIFS interval. Then, the node Dsends an ACK message upon receiving data after anSIFS interval. The relay selection for a single packetmight not be cost-effective because of the introductionof some overhead by the selection process. Therefore,the selected relay can continue its services as a helperfor the consecutive continuous packets between S andD until its weight is reduced by a predefined thresholdvalue, for instance 0.01. However, if the relay fails tohold the transmission performance as calculated at thetime of relay selection process, it stops services and theprotocol searches for a new relay.

The feedback message msg_dir means that a suitablerelay is not available, and thus the sender S transmits itsdata directly to the receiver D after an SIFS interval.After another SIFS, D sends an ACK message uponsuccessful data reception. The nature of this protocolis such that if a suitable relay is unavailable, then thesender uses a noncooperative transmission mode, alsoknown as a traditional direct transmission mode.

The proposed DEC-MAC protocol has the abilityto quickly switch from cooperative to noncooperativemode. When both the relay and the receiver fail todecode the sent data packet, then the receiver does notsend an ACK message, and, hence, the sender retrans-mits the packet after a PCF interframe space (PIFS)(i.e., SIFS + slot time) time period. Similarly, whenboth the sender and relay transmit but the receiver isunable to decode, the packet is retransmitted by thesender after a PIFS time period.

5 Analysis

In this section, we derive analytical expressions for theexpected number of mini slots required for relay selec-tion. In addition, analyses for single-hop packet delay,power cost and network lifetime are accomplished.

5.1 Expected number of mini slots

Determining the number of mini slots required forfinding the optimal relay is one of the key challenges forthe proposed protocol. The overhead increases as thenumber of mini slots increases. Assuming that a slot isnon-idle when one or more candidate helpers transmitin that slot, we can express the probability that thefirst j th minislot is non-idle as

(nk

)pk(1 − jp)n−k, where

k ≤ n is the number of candidate nodes transmittingan ITH message in a slot, and p = 1

n is the probability

Ann. Telecommun. (2013) 68:485–501 493

that a node will transmit an ITH message in any slot.Therefore, the average number of slots prior to the firstnon-idle slot is given by

E1(X) =n∑

j=1

n∑k=1

(nk

)pk(1 − jp)n−k j. (6)

Also, the average number of mini slots required toresolve the collision among k nodes is given by

E2(X) =n∑

j=1

n∑k=1

(nk

)pk(1 − jp)n−ke(Xk). (7)

Here, the term e[Xk] denotes the number of minislots that may be required to resolve the collisionamong k number of nodes. e(Xk) is a recursive functionand the value of e(Xk) is given by [27]

e[Xk] = 0.5k ∑k−1l=2

(kl

)e(Xk) + 1

1 − (0.5)k−1, ∀ k ≥ 1, (8)

where e[Xk] ≤ log2(k) + 1 for all k. For simplicity, wecan write the Eq. 8 as follows:

e[Xk] = log2(k) + 1, ∀ k ≥ 1. (9)

Thus, the total number of mini slots required to findthe optimal relay is given by

K = E1(X) + E2(X). (10)

From Eqs. 6, 7, and 10, we can write

K =n∑

j=1

n∑k=1

(nk

)pk(1 − jp)n−k j

+n∑

j=1

n∑k=1

(nk

)pk(1 − jp)n−ke(X). (11)

Putting the value for e[Xk] in Eq. 11, we have

K =n∑

j=1

n∑k=1

(nk

)pk(1 − jp)n−k j

+n∑

j=1

n∑k=1

(nk

)pk(1 − jp)n−k(log2(k) + 1). (12)

We plot a graph of average number of required minislots K vs. the total number of candidate relays n, asshown in Fig. 5. The graph reveals that the number ofmini slots increases from 2 to 3.07 with the candidaterelays ranging from 1 to 100. The number of minislots remain close to 3.07 even though n → ∞. Themost noticeable finding is that the number of mini slotsremains less than 3 (K < 3) until n ≤ 10. Typically, formost of the cases, less than ten potential candidatesremain in a transmission environment [12]. Therefore,we conjecture that three slots are required for findingthe optimal relay.

5.2 Delay analysis

The local single-hop delay has a great impact on theend-to-end packet delay. This local delay is defined asindividual hop delay of a multi-hop network. We con-sider the following three parameters-expected trans-mission time, relay selection time (To), and protocoloperation overhead (Tr) for calculating the total single-hop delay DSD from node S to node D and can beexpressed as follows:

DSD = ETT + To + Tr. (13)

Here, we have neglected propagation delay since itcarries very small value. The measurement method ofETT has been discussed in Section 4.2. We assumethat our cooperative communication method uses the

Fig. 5 Number of candidaterelays vs. average number ofmini slots

0 10 20 30 40 50 60 70 80 90 1002

2.2

2.4

2.6

2.8

3

3.2

3.4

Number of candidate relays

Exp

ecte

d nu

mbe

r of

min

i−sl

ots

494 Ann. Telecommun. (2013) 68:485–501

MRC technique at the receiving node D. Therefore, thetotal signal-to-noise ratio (SNR) value for decode andforward at the receiving node D can be written as

ϒcoop = ϒSD + ϒrn D (14)

where ϒSD is the SNR value at node D for the directlink SD and ϒrn D is the SNR value at D for the linkrn D. The probability of packet error for cooperativecommunication is given by [29]:

δcoop = δSDδSrn + (1 − δSrn)δSrn D (15)

where δSD and δSrn are the probabilities of packet errorsfor the links sender–receiver (SD) and sender-relay(Srn), respectively, and δcoop is the probability of thepacket error for the cooperative link Srn D. The firstterm of the right-hand side of Eq. 15 corresponds tothe probabilities of packet error at the links sender–receiver and sender-relay. The second term corre-sponds to the probabilities of a successful reception atthe relay for the sender-relay link and packet error atthe receiver for the cooperative link Srn D. In terms ofthe SNR value, the probability of a packet error can beexpressed as

δ = 1 −(

1 − 1

2e− ϒ

2

)S

where S is the packet length. Therefore, the values forδSD, δSrn , and δSrn D of Eq. 15 can be calculated using thevalues ϒSD, ϒrn D, and ϒcoop, respectively.

Based on 802.11 DCF, the sender retransmits apacket if the original transmission is not successful,with a standard retry limit. Initially, we assume that theprobability of the successful transmission of a packetfrom S to D, after ϕ attempts, can be written as

δsuccess = (δcoop)ϕ−1(1 − δcoop). (16)

Finally, the expected number of retransmissions re-quired to transmit a packet successfully from S to D canbe written as

ETX =∞∑

ϕ=1

ϕδsuccess (17)

Therefore, using the value of ETX, we are able todetermine the value of ETT such that

ETT = (ETX)S

B= S

B

∞∑ϕ=1

ϕδsuccess (18)

If ϕ is the required number of retransmissions for thesuccessful transmission of a packet, then we can find the

delay D successSD for the hop SD using Eqs. 13 and 18 such

that

D successSD =

∞∑ϕ=1

(S

Bϕ ∗ δsuccess + To,ϕ + Tr,ϕ

)(19)

Equation 14 demonstrates that SNRcoop ≥ SNRdir.Thus, the probability of a packet error with cooperativecommunication is less than with direct communication,δcoop ≤ δdir. Therefore, the required number of retrans-missions ϕ for cooperative communication is less thanthe number required for direct communication. Theresult of Eq. 19 shows that the hop-by-hop delay isreduced. Our proposed relay selection algorithm alsoreduces the value of the relay selection overhead Tr.

5.3 Power cost

The power cost in cooperative communication includestransmission cost by sender S is Ttx,s and the receivingcost both by relay rn and receiver D are 2Prx. A relayretransmits a packet at the cost (Ptx,r + Prx)(1 − δsr),where, (1 − δsr) is the probability that the relay decodedthe packet correctly and then retransmitted. Thus,the power cost for cooperative transmission [29] is asfollows:

Pcoop = (Ptx,s + 2Prx) + (Ptx,r + Prx)(1 − δsr). (20)

The power cost for direct transmission can be writ-ten as

Pdir = Ptx,s + Prx (21)

The required transmission energy for cooperativetransmission is Ecoop = PcoopL

Cand for direct transmis-

sion is Edir = PdirLC . Here, L is the packet length and

C is the transmission rate.The expected number of retransmissions [22, 30]

for one hop is given by ETX = 11−δ

, where δ is theprobability of a packet error. Thus, the required energyexpected for successful transmission can be writtenas Eexpected

coop = Ecoop

1−δcoopfor cooperative transmission and

Eexpecteddir = Edir

1−δdirfor direct transmission. The packet

error rate is inversely proportional to the SNR, δ ∝1ϒ

. From Eq. 14, we find that ϒcoop ≥ ϒdir, whereϒdir = ϒSD. Therefore, the expected number of retrans-missions for cooperative communication is ETXcoop ≤ET Xdir; ET Xdir is the expected number of retransmis-sions for direct transmissions. Thus, cooperative com-munication requires less transmission energy.

Ann. Telecommun. (2013) 68:485–501 495

5.4 Lifetime analysis

The network lifetime is defined as the duration of thenetwork operating time up until the first node failsdue to battery depletion [10, 18]. This is a meaningfulmeasurement in the sense that failure of the first node isan indication of the remaining limited network lifetime;in some applications, the first node failure can cause anetwork to become partitioned and additional servicescan be interrupted. In general, the lifetime (LT) of anynode i with initial energy Eini can be expressed as

LTi = Eini

ei, ∀i = 1, 2, ...η (22)

where ei is the average energy consumed per unit time.First, we assume that all of the network energy is

used in transmitting, receiving, processing, idle, and lis-tening modes. Therefore, the virtual lifetime of the net-work is defined as LTvir = (ηEini)/ei where no energyremains leftover in the network. In reality, however,some energy remains leftover in the network when thenetwork expires. Let each of a m number of nodeshave residual energy Eres,i, where the total number ofnodes is η. This leftover residual energy is treated aswasted energy. The average amount of wasted energyin a network can be expressed as

Ew =m∑

i=1

Eres,i, t ≥ lifetime (23)

The network lifetime can be written in terms ofwasted energy Ew such that

LT = ηEini − Ew

ei(24)

In Eq. 24, η and Eini are constants for a network; thus,the network lifetime depends on Ew. In [8], it is reportedthat the network lifetime depends not only on the aver-age consumed energy but also on the residual energyleft after the network lifetime expires. Therefore, inorder to maximize the network lifetime, the protocolshould reduce the wasted energy Ew to a minimumvalue minEw. Thus, the network lifetime is increased toits maximum as max LT and given by

maxLT = ηEini − minEw

ei(25)

Since our protocol searches for an optimal relayhaving a higher residual energy and shorter delay, itinfluences the balance of energy consumption amongall of the nodes, which decreases the leftover wastedenergy, thereby increasing the network lifetime.

6 Performance evaluation

6.1 Simulation environment

We evaluated the performance of the proposed DEC-MAC using simulation experiments conducted on NS-2[31] and compared the results with those of CoopMAC[4] and LC-MAC [21]. The configuration of the simu-lation environment parameters is listed in Table 1. Thedata traffic bursts from three randomly chosen events,listed in Table 2, are considered in the performancestudies. Each data point shown in the graphs are theresults of the average of ten simulation runs.

6.2 Performance metrics

6.2.1 Average end-to-end packet delay

The end-to-end delay of a single packet is measuredas the time difference between when the packet is

Table 1 Configuration of parameters

Deployment Area size 2,000 × 2,000 mDeployment type Uniform randomNetwork architecture Homogeneous flatNumber of nodes 2,000Sink (1,000, 1,000)Initial node energy 100 JBuffer size 50Sources in one event 15 nodesRadio range 100 mLink layer trans. rate 512 kbpsTransmit power 7.214e−3 WRcv. signal threshold 3.65209e−10 WPHY error model Uniform randomLink failure rate 0.30

Task Application type Event-drivenPacket size 64 bytesTraffic type CBR, 3 pkts/s

DEC-MAC MAC header 272 bitsPHY header 192 bitsRTS 128 bitsRTH 128 bitsACK 112 bitsITH 112 bitsFITH 112 bitsSIFS 10 µsDIFS 32 µsaCWMin 31 slotsaCWMax 1,023 slotsRetry limit 5

Simulation Time 200 s

496 Ann. Telecommun. (2013) 68:485–501

Table 2 Events and bursts description

Event 1 (s) Event 2 (s) Event 3 (s)

Burst 1 10∼40 20∼50 30∼60Burst 2 90∼120 100∼130 110∼140

received at the sink from and its the time it is generationgenerated time at the source node. Delays experiencedby individual data packets are averaged over the totalnumber of individual packets received by the sink. Thelower the value is, the better the performance is.

6.2.2 Network lifetime

The network lifetime is defined as the length of timefrom the network’s deployment until the point in timewhen the first of its nodes is fully drained of energy andbecomes nonfunctional.

6.2.3 Per-packet energy consumption

Energy consumption per packet is measured as the ra-tio of the total amount of energy dissipated by all sourceand forwarding nodes during the simulation period tothe number of packets received by the sink, e.g., theaverage amount of energy expended for each successfulpacket reception.

6.2.4 Standard deviation of residual energy levels

The standard deviation of residual energy levels atsensor nodes defines the average variance between theresidual energy levels at all nodes and is measured byEq. 26,

σ =√√√√ 1

N

N∑i=1

(Eres,i − μres)2, (26)

where Eres,i and μres are the residual energy of nodei and the mean residual energy for all nodes, respec-tively. The value of σ indicates how well the energyconsumption is distributed among the sensor nodes; thesmaller the value, the better the capability of DEC-MAC to balance the energy consumption.

6.2.5 Protocol operation overhead

In order to compare the protocol operation overheadincurred by the control packets, the following arecounted during the entire simulation period: the num-ber of bytes in RTS, RTH, ITH, FITH, and ACK con-trol packets transmitted by DEC-MAC; the number of

bytes in RTS, HTS, CTS, and ACK packets transmittedby CoopMAC; and the number of bytes in RTS, CTS,GI and MI messages, RTH and RC (recontention)messages, and ACK packets transmitted by LC-MAC.The lower the values are for the number of bytes, thehigher the protocol’s performance.

6.3 Simulation results

6.3.1 Impacts of node density

In this section, we show the impacts of node densityon the performance of the protocols. The graph ofFig. 6 shows that average end-to-end delay of DEC-MAC has significantly improved over CoopMAC andLC-MAC. Although initially end-to-end delay of Coop-MAC is lower than DEC-MAC, it increases with thenode densities. Since, the helpers in CoopMAC donot exchange any message before transmission, so theneighbors might start to transmit data while the trans-mission of helper is ongoing. Therefore, collision in-creases in CoopMAC as node density increases andhence average end-to-end delay also increases.

Measuring the lifetime of a sensor network in a sim-ulation environment is a tedious task, typically rangingfrom several months up to 2 years. It depends uponthe initial energy of the nodes, the amount of traffic,and the MAC and routing protocols employed in thenetwork. We executed simulation runs for varying nodedensities, with an initial energy of 20 J for each node,in order to compare the performances of the protocols.Figure 7 shows that the network lifetime of DEC-MACis significantly increased compared to that of LC-MACand CoopMAC. The reason behind this interesting

Fig. 6 Average end-to-end delay vs. number of deployed nodes

Ann. Telecommun. (2013) 68:485–501 497

Fig. 7 Network lifetime vs. number of deployed nodes

result is that residual energy is considered in the relayselection in DEC-MAC, which implements balancedenergy consumption among the nodes. Therefore, lessleftover energy remains when a network dies out. LC-MAC and CoopMAC do not consider the residualenergy when selecting a relay node; thus, the betterhelper nodes frequently participate in cooperation anddie earlier. As a result, a significant amount of energyremains unused as leftover energy in the network.

Figure 8 shows that DEC-MAC is more energyefficient than LC-MAC and CoopMAC. This is basedon the fact that the number of successful transmissionof data packets in the simulation period is higher withDEC-MAC than with the other protocols. Therefore,

Fig. 8 Per-packet energy consumption vs. number of deployednodes

per data packet energy consumption is lower in DEC-MAC. The per data packet energy consumption ishigher in CoopMAC and LC-MAC due to the lowernumber of successful packet transmissions resultingfrom the higher number of collisions during the simula-tion period. Energy efficiency increases as node densityincreases; along with the node increase is an increasein cooperative diversity, which leads to an improvedchance of having a better helper node.

Figure 9 shows the standard deviation of the residualenergy in different protocols with varying node densi-ties. The figure reveals that the proposed DEC-MACperforms quite well in terms of energy load balancingamong the network nodes. Even at very low node den-sities, DEC-MAC outperforms LC-MAC and Coop-MAC. This is due to the use of an energy-aware relayselection mechanism in DEC-MAC, which is absent inLC-MAC and CoopMAC.

Figure 10 compares the operational overhead of thedifferent protocols. The graphs indicate that the oper-ational overhead of DEC-MAC is significantly lowerthan that of LC-MAC and CoopMAC. Initially, theoperational overhead is higher than that of CoopMACsince the number of control packets in CoopMAC islower than that in DEC-MAC. However, with an in-crease in node density, the number of collisions andretransmissions increases in CoopMAC and thus theprotocol’s operational overhead also increases. The useof extra GI and MI messages, along with the commoncontrol messages in LC-MAC, also serves to increasethe protocol’s operational overhead; LC-MAC’s oper-ational overhead increases exponentially as collisionsincrease with increases in node density.

Fig. 9 Standard deviation of residual energy vs. number of de-ployed nodes

498 Ann. Telecommun. (2013) 68:485–501

Fig. 10 Protocol operation overhead vs. number of deployed ofnodes

6.3.2 Impacts of link failure rate

This section discusses the impacts of different channelconditions on the performance of the protocols. Themore link failures in a unit of time, the more unstable awireless network will be [32]. In this paper, we expressthe channel condition in terms of link failure rate f andwe vary the value of f from 0.05 to 0.50 in incrementsof 0.05; all other parameters in Tables 1 and 2 are fixed.

The simulation results in Fig. 11 show that DEC-MAC has a better average end-to-end packet delaythan the other protocols at higher link failure rates.Since the number of control messages for CoopMAC islower than that for DEC-MAC, so initially CoopMAChas a lower end-to-end packet delay. However, the

Fig. 11 Average end-to-end delay vs. link failure rate

end-to-end packet delay of CoopMAC is higher thanthat of DEC-MAC at higher link failure rate. This is be-cause, in CoopMAC and LC-MAC, relays are selectedbased on the transmission rate; however, the node witha better transmission rate always might not be a betterhelper due to its heavy traffic load. Therefore, DEC-MAC has less end-to-end packet delay than CoopMACand LC-MAC at higher link failure rates.

The graphs in Fig. 12 reveal that DEC-MAC pro-vides a more than 60 % higher lifetime as comparedto CoopMAC and LC-MAC, even with a 50 % linkfailure rate. This is due to the uniform distribution ofthe energy consumption among the nodes, even with ahigher link failure rate. This strategy reduces the left-over energy in a network when it dies out. As a result,the network lifetime of DEC-MAC is much higher thanthat of the other protocols, even with a higher linkfailure rate.

As the link failure rate increases, the throughputdecreases, and the end-to-end packet delay increases,thereby increasing the per-packet energy consumption.Figure 13 shows that the per-packet energy consump-tion of all protocols increases as the link failure rateincreases. However, the per-packet energy consump-tion for DEC-MAC is much lower than for CoopMAC.This is because CoopMAC select a relay based onCoopTable which updates overhearing the neighbor’stransmission, so it is less dynamic. If the link failurerate increases, then the probability to select a relay withpoor-linked path increases. In turn, this fact increasesthe retransmissions and thus increases transmission en-ergy. On the other hand, in LC-MAC, it has higher pro-tocol overhead and higher susceptible to collision thanDEC-MAC. Thus, the per-packet energy consumption

Fig. 12 Network lifetime vs. link failure rate

Ann. Telecommun. (2013) 68:485–501 499

Fig. 13 Energy efficiency vs. link failure rate

in DEC-MAC is much lower than LC-MAC. Thisproves that the proposed DEC-MAC protocol, with itshigher link failure rate, is more energy-efficient thanLC-MAC and CoopMAC.

A desirable result is found for the standard deviationof residual energy against link failure rate, as shownin Fig. 14. The graphs of Fig. 14 show that there islittle change in the standard deviation of the residualenergy in DEC-MAC even with an increasing link fail-ure rate. This is because DEC-MAC selects an optimalrelay having a lower delay and higher residual energy,causing the power consumption among the nodes toremain balanced. Alternatively, the standard deviationof LC-MAC and CoopMAC changes considerably with

Fig. 14 Standard deviation of residual energy vs. link failure rate

Fig. 15 Protocol operation overhead vs. link failure rate

the rate of link failures, thus proving that DEC-MACprovides a better distribution of energy consumptionamong the sensor nodes.

Figure 15 shows the changes in a protocol’s oper-ation overhead as a function of the link failure rate.Initially, the operation overhead of DEC-MAC ishigher than that of CoopMAC. The graphs in Fig. 15reveal that the increasing rate of operation overheadfor DEC-MAC is much lower than for CoopMAC andLC-MAC. Therefore, DEC-MAC, with a higher linkfailure rate, achieves lower operation overhead thaneither LC-MAC or CoopMAC. The reason for this isthat the increasing rate of collision between the relayand neighbor in DEC-MAC is lower than that of Coop-MAC and LC-MAC.

7 Discussion

In this section, we would like to direct the readers’ at-tention to the discussions on the limitations and weak-nesses of our paper.

In the relay selection mechanism, we use PE andthe CCDF to select the most optimal relay. However,if two or more potential relays have equal averageweighted metric (W) value, then the required numberof mini slots for relay selection goes to infinity. Thus,the relay selection time also goes to infinity. One of thepossible solutions for such situation is as follows. Therelays that send ITH message after fourth mini slot, allare selected as optimal relays to break the tie. This isbecause average number of required mini slots is 3.07,thus if more than one relays transmit ITH messages atfifth mini slot, then all the respondents of fifth mini

500 Ann. Telecommun. (2013) 68:485–501

slot will be selected as optimal relays. After selectionof the relays, sender transmits to relays and then relaystransmit together to receiver. Since, the relays transmitsame data at the same rate to the destination node, thusit does not result any collision there [25].

Relay selection is an important part for cooperativetransmissions to improve the network performance.In our scheme, we select a relay which has higherresidual energy and lower data delivery delay. If weselect a relay with highest residual energy but with poorchannel condition, it increases packet error rate as wellas transmission energy and thus increases per-packetenergy consumption. In contrast, if we select a relaywhich has lowest data delivery delay, then particular re-lay may participate in cooperation frequently. Thus, itfinishes its energy early and may partition the networkwhich reduces the network lifetime. Therefore, we dotrade-off between delay and energy consumption tooptimize end-to-end delay and network lifetime. Due tothese facts, end-to-end delay does not increase sharplylike other protocols such as CoopMAC and LC-MACshown in Fig. 6. On the other hand, per-packet en-ergy consumption does not decrease sharply like otherprotocols such as CoopMAC and LC-MAC shown inFig. 8. However, network lifetime of DEC-MAC im-proves significantly over the protocols CoopMAC andLC-MAC as shown in Fig. 7. This is because our relayselection mechanism balances the energy consumptionamong the nodes and reduces the leftover energy whenthe network dies out.

8 Conclusion

The proposed DEC-MAC protocol provides significantservice differentiation in relay selection and networklifetime domains. Our relay selection algorithm con-tributes to find the most optimal relay in reduced num-ber of time slots. For the lifetime domain, our method-ologies help the network to maintain balanced energyconsumption among the nodes, thereby increasing thenetwork lifetime significantly.

Like other cooperative protocols, our DEC-MACalso has relay selection and additional transmissionoverheads. Furthermore, relayed transmission mightinterfere transmissions from other hidden or exposednodes. In the future, we work on methodologies thatcould diminish the relay selection overheads and theinterferences.

Acknowledgements This research was supported the MKE(TheMinistry of Knowledge Economy), Korea, under the ITRC(Infor-mation Technology Research Center) support program super-

vised by the NIPA(National IT Industry Promotion Agency)”(NIPA-2012-(H0301-12-2001). Dr. CS Hong is the correspondingauthor.

References

1. Gama O, Carvalho P, Afonso JA, Mendes PM (2008) Qualityof service support in wireless sensor networks for emergencyhealthcare services. In: 30th annual international conferenceof the IEEE Engineering in Medicine and Biology Society.EMBS 2008, pp 1296–1299

2. Felemban E, Member S, gun Lee C, Ekici E (2006) Mmspeed:multipath multi-speed protocol for QoS guarantee of reliabil-ity and timeliness in wireless sensor networks. IEEE TransMobile Comput 5:738–754

3. Liang X, Balasingham I, Leung V (2009) Cooperative com-munications with relay selection for QoS provisioning inwireless sensor networks. In: IEEE global telecommuni-cations conference. GLOBECOM, Honululu, HI, Nov. 302009–Dec. 4 2009

4. Liu P, Tao Z, Narayanan S, Korakis T, Panwar SS (2007)CoopMAC: a cooperative MAC for wireless LANS. IEEEJ Sel Areas Commun 25(2):340–354

5. Zhai C, Liu J, Zheng L, Xu H (2009) Lifetime maximiza-tion via a new cooperative MAC protocol in wireless sensornetworks. In: IEEE global telecommunications conference.GLOBECOM, 30, Honululu, HI, Nov. 30 2009–Dec. 4 2009

6. Sadek AK, Yu W, Liu KJR (2010) On the energy efficiencyof cooperative communications in wireless sensor net-works. ACM Trans Sen Netw 6:5:1–5:21 [Online]. Available:http://doi.acm.org/10.1145/1653760.1653765

7. Nacef AB, Senouci S-M, Ghamri-Doudane Y, Beylot A(2009) Enhanced relay selection decision for cooperativecommunication in energy constrained networks. In: 2009 2ndIFIP Wireless Days (WD), pp 1–5

8. Chen Y, Zhao Q (2005) Maximizing the lifetime of sensornetwork using local information on channel state and resid-ual energy. In: 2005 conference on information sciences andsystems. The Johns Hopkins University

9. Durresi A, Paruchuri V, Barolli L (2005) Delay-energy awarerouting protocol for sensor and actor networks. In: Proceed-ings of the 11th international conference on parallel and dis-tributed systems, vol 1. ser. ICPADS ’05. IEEE ComputerSociety, Washington, DC, pp. 292–298. [Online]. Available:http://dx.doi.org/10.1109/ICPADS.2005.121

10. Kang I, Poovendran R (2003) Maximizing static networklifetime of wireless broadcast ad hoc networks. In: IEEEinternational conference on communications. ICC ’03, vol 3,pp 2256–2261

11. Kim J, Lin X, Shroff NB, Sinha P (2010) Minimizing delayand maximizing lifetime for wireless sensor networks withanycast. IEEE/ACM Trans Network 18:515–528

12. Adam H, Elmenreich W, Bettstetter C, Senouci S-M (2009)CoRe-MAC: a MAC-protocol for cooperative relaying inwireless networks. In: GLOBECOM, pp 1–6

13. Shan H, Zhuang W, Wang Z (2009) Distributed cooperativeMAC for multihop wireless networks. Commun Mag 47:126–133. [Online]. Available: http://dx.doi.org/10.1109/MCOM.2009.4785390

14. Shah V, Mehta N, Yim R (2010) Splitting algorithms for fastrelay selection: generalizations, analysis, and a unified view.IEEE Trans Wirel Commun 9(4):1525–1535

Ann. Telecommun. (2013) 68:485–501 501

15. Wu X, Chen G, Das SK (2006) On the energy hole problemof nonuniform node distribution in wireless sensor networks.In: 2006 IEEE international conference on mobile adhoc andsensor systems (MASS), pp 180–187

16. Lian J, Naik K, Agnew GB (2005) Data capacity improve-ment of wireless sensor networks using non-uniform sensordistribution. Int J Distrib Sens Netw 2:121–145

17. Liu E, Zhang Q, Leung K (2010) Residual energy-awarecooperative transmission (REACT) in wireless networks. In:2010 19th annual wireless and optical communications con-ference (WOCC), pp 1–6

18. Banerjee A, Foh CH, Yeo C, Lee BS (2010) A networklifetime aware cooperative MAC scheme for 802.11b wirelessnetworks. In: 2010 7th IEEE consumer communications andnetworking conference (CCNC), pp 1–5

19. Razzaque MA, Mamun-Or-Rashid M, Alam MM, Hong CS(2009) Aggregated traffic flow weight controlled hierarchicalMAC protocol for wireless sensor networks. Ann Télécom-mun 64(11–12):705–721

20. Sarkar S, Yen H-H, Dixit S, Mukherjee B (2008) Anovel delay-aware routing algorithm (DARA) for a hy-brid wireless-optical broadband access network (WOBAN).IEEE Network 22(3):20–28

21. Zhou Y, Liu J, Zheng L, Zhai C, Chen H (2011) Link-utility-based cooperative MAC protocol for wireless multi-hop net-works. IEEE Trans Wirel Commun 10(3):995–1005

22. Wireless RM-H, Banerjee S (2002) Adapting transmissionpower for optimal energy. In: Multi-hop wireless communica-tion, wireless optimization workshop (WiOpt03). Tech. Rep.,Sophia-Antipolis

23. Liang X, Chen M, Xiao Y, Balasingham I, Leung VCM(2010) A novel cooperative communication protocol for QoSprovisioning in wireless sensor networks. Int J Sen Netw

8(2):98–108. [Online]. Available: http://dx.doi.org/10.1504/IJSNET.2010.034619

24. Sheng Z, Leung K, Ding Z (2011) Cooperative wireless net-works: from radio to network protocol designs. IEEE Com-mun Mag 49(5):64–69

25. Khalid M, Wang Y, ho Ra I, Sankar R (2011) Two-relay-based cooperative MAC protocol for wireless ad hoc net-works. IEEE Trans Veh Technol 60(7):3361–3373

26. Park JC, Kasera S (2005) Expected data rate: an accuratehigh-throughput path metric for multi-hop wireless routing.In: 2005 second annual IEEE communications society con-ference on sensor and ad hoc communications and networks.IEEE SECON 2005, pp 218–228

27. Qin X, Berry R (2004) Opportunistic splitting algorithms forwireless networks. In: INFOCOM 2004. Twenty-third annualjoint conference of the IEEE Computer and Communica-tions Societies, vol 3, pp 1662–1672

28. Langendoen K, Reijers N (2003) Distributed localization inwireless sensor networks: a quantitative comparison. ComputNetw 43:499–518. [Online]. Available: http://portal.acm.org/citation.cfm?id=957368.957374

29. Shi L, Fapojuwo A (2010) Energy efficiency and packet errorrate in wireless sensor networks with cooperative relay. In:2010 IEEE sensors, pp 1823–1826

30. Draves R, Padhye J, Zill B (2004) Routing in multi-radio,multi-hop wireless mesh networks. In: ACM MobiCom.ACM Press, pp 114–128

31. The Network Simulator NS-2. http://www.isi.edu/nsnam/ns/.Accessed 10 July 2011

32. Shu X, Li X (2007) Link failure rate and speed of nodesin wireless network. In: International conference on wirelesscommunications, networking and mobile computing. WiCom2007, pp 1441–1444