low energy operation in wsns: a survey of preamble sampling

Computer Networks xxx (2011) xxx–xxx

Contents lists available at ScienceDirect

Computer Networks

journal homepage: www.elsevier .com/ locate/comnet

Low energy operation in WSNs: A survey of preamble samplingMAC protocols

C. Cano ⇑, B. Bellalta, A. Sfairopoulou, M. OliverNeTS Research Group, Department of Information and Communication Technologies, Universitat Pompeu Fabra, C/Tànger 122-140, 08018 Barcelona, Spain

a r t i c l e i n f o

Article history:Received 29 January 2011Received in revised form 21 June 2011Accepted 25 June 2011Available online xxxx

Keywords:MAC protocolsWSNsPreamble samplingLow power listeningLow energy consumption

1389-1286/$ - see front matter � 2011 Elsevier B.Vdoi:10.1016/j.comnet.2011.06.022

⇑ Corresponding author.E-mail addresses: [email protected] (C. Cano

edu (B. Bellalta), [email protected] (A. [email protected] (M. Oliver).

Please cite this article in press as: C. Cano et aNetw. (2011), doi:10.1016/j.comnet.2011.06.0

a b s t r a c t

The limited energy resources of sensor nodes are among the most important constraints inWireless Sensor Networks (WSNs). Consequently, the Medium Access Control (MAC) layerdesign is crucial, due to its influence on the transceiver, which is the most energy-consumingcomponent of a sensor node. Among the different MAC protocols designed for WSNs, pream-ble sampling techniques provide extremely low energy consumption at low loads and have anotably simple operation and a lack of synchronisation requirements, which are character-istics that are especially appealing to WSNs. In this work, a survey of the different types ofMAC protocols designed for WSNs is presented with a special focus on preamble samplingMAC protocols. The aim of this work is to give a detailed overview and classification of themost relevant preamble sampling MAC protocols, being motivated by the extremely largenumber of MAC protocols designed for WSNs in recent years. Moreover, a simple set of guide-lines for matching the most suitable MAC protocol category to a given application is providedin this work.

� 2011 Elsevier B.V. All rights reserved.

1. Introduction

Wireless Sensor Networks (WSNs) are networks formedby small and low-capability devices that are able to senseenvironmental metrics and to communicate them wire-lessly to a central unit, known as a sink [1]. There is a widerange of potential applications in various areas, includingindustrial, military, environmental, health and home auto-mation applications, which can profit from using WSNs tocollect data. However, the large number of constraints onthe sensor nodes and the special characteristics and usesof WSNs impose several challenges on the design of aWSN. Some of these characteristics are summarised next.

The most important characteristic of WSNs is their highapplication dependence [1]. Different requirements andconstraints are directly imposed by the application. This

. All rights reserved.

), [email protected]), mique

l., Low energy operation i22

issue complicates the design of a general protocol stackto be used in different deployments and applications andleads to the definition of different approaches. Whatfollows is a review of the general characteristics and con-straints of these networks; however, it is important to notethat they all depend on the application and deploymentscenario.

The deployment of dense WSNs in large, remote anddifficult-to-access areas requires keeping the size and costof the sensor nodes as low as possible. This constraintimplies that the energy, computational and memory re-sources of the sensor nodes are usually limited. Therefore,simplicity and low energy consumption are importantrequirements in the design of WSN protocols. In addition,the limited energy resources of sensor nodes, combinedwith variable environmental conditions, cause a high levelof network dynamics, which should also be taken intoaccount.

Another interesting feature is the traffic patterns thatare usually found in WSNs. In these networks, it is commonthat all of the sensor nodes periodically transmit

n WSNs: A survey of preamble sampling MAC protocols, Comput.

http://dx.doi.org/10.1016/j.comnet.2011.06.022

mailto:[email protected]

mailto:boris.bellalta@upf. edu

mailto:boris.bellalta@upf. edu

mailto:[email protected]

mailto:mique [email protected]

mailto:mique [email protected]


http://www.sciencedirect.com/science/journal/13891286

http://www.elsevier.com/locate/comnet


2 C. Cano et al. / Computer Networks xxx (2011) xxx–xxx

information (the value of the metric of interest) to the cen-tral unit, causing a communication pattern notably differ-ent from the traditional point-to-point approach. There isalso the case of event reporting (for example, when a met-ric surpasses a given threshold) to the sink. In this instance,only those nodes that detect the event will try to send anotification to the central unit. Finally, the sink can alsoquery the network about certain information. For instance,the sink can ask the network in which location a givenmetric has surpassed a certain value. In this case, the dataflows from the sink to all of the sensor nodes in a data-centric approach, in contrast to the more commoncommunication pattern based on destination addresses.

The scalability of WSNs is also a challenge becausedeployments can be formed by a large number of sensornodes. This challenge also causes extra difficulty in somescenarios in which a global identifier should be assignedto each device.

To summarise, the constraints and characteristics ofWSNs are the following:

� High application dependence� Large, remote, dense and difficult-to-access deploy

ments� Low-cost and small devices� Devices with limited memory, transmitting and com-

puting resources� Devices with limited energy resources� High network dynamics� Periodic, event-based and query-based communication

patterns� Data flowing from all or a group of sensors to the central

unit and vice versa

Among the different constraints, limitations on energyresources are the most important because this limitationdirectly affects the network lifetime. This relationship isparticularly important given that, in typical WSN deploy-ments, it is too costly or even impossible to access the sen-sor nodes to replace the batteries. In such a situation, it isobvious that the WSN protocol stack must be designed toobtain the highest possible energy savings and thus extendthe network lifetime, while maintaining the performanceof the target application. Therefore, the design of theMedium Access Control (MAC) layer is of crucial impor-tance because it controls the most energy consuming com-ponent of a sensor node: the transceiver. AsynchronousMAC protocols and, particularly, preamble sampling areespecially appealing to WSNs because of their simplicityand lack of synchronisation requirements.

In this work, a survey of the basic preamble samplingtechnique and its extensions is provided. There are othergeneral surveys of MAC protocols for WSNs presented inthe literature, such as the review performed in 2006 byDermikol et al. [2], the extensive survey performed in2007 by Kredo et al. [3], the survey performed by Lange-ndoen in 2008 [4] and the more recent review presentedby Bachir et al. [5]. The main difference between this workand those articles is in the specific focus placed here on thepreamble sampling technique and its extensions. That fo-cus allows us to provide a deeper study on this specific

Please cite this article in press as: C. Cano et al., Low energy operation iNetw. (2011), doi:10.1016/j.comnet.2011.06.022

and prolific area. Moreover, this detailed study has beenextended with a discussion of the benefits and drawbacksof each approach to provide a set of guidelines to helpresearchers and developers of WSN protocols and applica-tions to select the most suitable MAC category for a givenapplication.

The rest of this paper is organised as follows: Section2 describes the importance of the MAC layer design in aWSN. Next, in Section 3, a general overview of thedifferent categories of MAC protocols for WSNs is pro-vided. In Section 4, the basic preamble sampling andits extensions are presented. A suggested procedure toselect the MAC protocol based on the application issubsequently provided in Section 5. Finally, some conclu-sions are drawn.

2. The importance of the MAC layer in WSNs

The MAC layer is responsible for coordinating trans-missions to a shared channel by defining how and whena node will attempt transmission. The medium accessprocedure can be based on a schedule assignment (whichcan be fixed during the entire network lifetime or duringa certain amount of time) or, conversely, can be based ona random procedure in which the transmission attempttime is decided independently at each node. It is alsopossible to define hybrid approaches that combine bothtechniques.

In WSNs, the MAC protocol design faces several limita-tions, such as the low computational and synchronisationcapabilities of sensor nodes, as well as their low memorycapacities [5]. Moreover, in this type of network in whichthe energy consumption is a major constraint, the MAClayer design becomes an important issue, as it directly con-trols the transceiver operation, which is the most energy-consuming component in a sensor node [1]. Therefore,the MAC layer must focus on the reduction of energy con-sumption of the sensor nodes. The different sources of en-ergy waste can be classified into the following [3]:

1. Idle Listening: Idle listening occurs when a node listensto the channel for a possible reception, but nothing isreceived. Idle listening has been identified as the majorsource of energy waste in a sensor node, mainly due tothe low traffic loads commonly found in WSNs.

2. Collisions: When two (or more) nodes simultaneouslytransmit, the recipient may be unable to decode anyof the packets involved. This problem implies that thesenders waste energy through transmitting and thatthe receiver expends energy receiving without obtain-ing any benefit, as senders may eventually retry trans-mission. In the presence of hidden terminals, thiseffect is even more important because transmitters can-not sense each other.

3. Overhearing: This source of energy dissipation occurswhen a sensor node wastes energy receiving a packetthat is intended for a different destination.

4. Overhead: Data packets in WSNs are usually small;therefore, headers and other types of overhead (suchas control messages) imply a high level of energy wastefor WSNs.



C. Cano et al. / Computer Networks xxx (2011) xxx–xxx 3

MAC protocols designed for Wireless Local AreaNetworks (WLANs) or mesh networks, such as the well-known Distributed Coordination Function (DCF) definedin the IEEE 802.11 standard [6], are not suitable for thespecial requirements of WSNs for several reasons. One ofthe reasons is that they do not consider the different con-straints of WSNs. Usually, these protocols assume thatnodes are full-capability devices with high memory foot-prints and precise synchronisation hardware and, mostimportantly, unlimited energy resources. Consequently,the protocols do not address the sources of energy wastepreviously described. The common behaviour of these pro-tocols is to continuously listen to the channel for possiblereception. However, in WSNs, this design will rapidly con-sume the energy resources of a sensor node, due to idle lis-tening. Moreover, as nodes remain constantly listening tothe channel, overhearing is common in that type of proto-col. The overhead is also considerable in traditional MACprotocols because the length of the data packets of theapplications that they are addressed to are significantlyhigher than the lengths found in WSNs. The only sourceof energy waste that these MAC protocols address is colli-sions because they (apart from wasting energy) consider-ably reduce the performance of the network. Anotherreason for their unsuitability is that traditional MAC proto-cols are commonly focused on throughput or latency, while,in WSNs, these metrics are traded off for a decrease inenergy consumption. Although sporadic increases in thenetwork load can happen because of, for example, eventdetections, WSNs usually work at extremely low loads(with delay-tolerant traffic). Therefore, it is not always nec-essary to provide a high throughput or reduced delay to theapplication.

3. A general overview of MAC protocols for WSNs

In previous years, a large number of MAC protocols thatare specially designed for WSNs have been defined [3]. Tosave energy, especially the energy wasted because of idlelistening, the most common approach is to put the trans-ceiver into sleep mode for as much time as possible becausesleep mode consumes substantially less energy than theother available modes (idle, transmitting or receiving). Insleep mode, a sensor node is not able to receive/transmitpackets from/to the medium. This solution greatly reducesthe idle listening energy waste. However, specific mecha-nisms should be defined to ensure that a sensor node willbe awake if a node sends something to it. There are threegeneral categories to address this issue [5], as follows:

� Scheduled (Time Division Multiple Access (TDMA)-like)protocols [7,8]: in this category, sensor nodes areassigned to specific slots in a frame to transmit and/orreceive; then, the nodes only wake up in those slotsand sleep in the rest. A node must know the time slotin which a neighbour will be awake before it can senddata to it. A schematic representation is shown inFig. 1(1); observe that nodes transmit on their assignedslots and wake up to receive on the slots of their neigh-bours. This type of protocol provides good performance


(controlled delays, high throughput and reduced or zerocollision probabilities) but requires good synchronisa-tion among sensor nodes. Moreover, there is alwaysan overhead cost associated with creating and main-taining the slot assignment and keeping the synchroni-sation among the nodes.� Protocols with Common Active Periods [9,10]: these pro-

tocols organise sensor nodes to wake up and go to sleepat the same time. A sensor node must wait for the nextwake-up period to send data. Observe Fig. 1(2), whichrepresents the functionality of one of the most well-known protocols with common active periods: theS-MAC protocol. It can be seen that sensor nodes readyto transmit wait until the active period. Moreover, as anextension to the previously described behaviour ofS-MAC presented in [11], the authors of S-MAC definein [9] that a first handshake (using RTS/CTS) is per-formed in the active period, but data communicationis done in the sleep period, allowing the overhearingnodes to go to sleep. Although a certain degree of syn-chronisation is also needed in these protocols, thisrequirement is not as stringent as in scheduled TDMA-like protocols. Additionally, protocols with commonactive periods have a cost associated with the creationand maintenance of the schedule (when to sleep andwhen to wake up).� Asynchronous MAC protocols [12,13]: in this type of pro-

tocol, no synchronisation is required, because each nodegoes to sleep and wakes up independently of the others.These protocols implement different mechanisms toensure that a receiver will be awake to receive the data.The protocols can be divided into preamble samplingand receiver-initiated approaches. In preamble sam-pling, sensor nodes wake up only for checking whetherthe channel is busy, in which case they remain awake.Next, a sensor node wanting to transmit first sends along preamble to overlap with the channel samplingtime of the receiver and subsequently sends the data.This behaviour is represented in Fig. 1(3); it can beobserved that the long preamble makes both the recei-ver and an overhearing node stay awake to finallyreceive the message. Conversely, in receiver-initiatedprotocols, sensor nodes send an advertisement whenthey wake up. Then, a sensor node with a packet totransmit waits until it receives the advertisement sentby the receiver.

Among the different MAC protocols designed for WSNs,asynchronous approaches and preamble sampling providesome interesting features that make them especiallyappealing to WSNs. Observe that, compared to protocolswith common active periods, preamble sampling providesless complexity and cost because no common scheduleneeds to be created, communicated and maintained. Inaddition, preamble sampling consumes less energy com-pared to protocols with common active periods when thetraffic load is low. This difference is caused by the extre-mely short channel sampling time, which allows a sensornode to return immediately to sleep if the medium is foundidle. On the contrary, in protocols with common activeperiods, the active period of the duty cycle is commonly



Data

Rx Data

Rx

Rx Data

Rx

Data

Rx

Rx

Node 1

Node 2

Node 3

Node 4

Slot 1 Slot 2 Slot 3 Slot nt

t

t

t

1 2 3 4

Transmitting Receiving Idle Listening Sleeping

Sleep Period

t

t

SYNC

Data

RxACK

Tx

Rx

CTS

Active Period

RTS

Overhearingt

3. Asynchronous MAC Protocols (Preamble Sampling)

2. Protocols with Common Active Periods

1. TDMA−like MAC Protocols

t

Long PreambleTx

Rx

Data

Rx

t

Rx

t

Overhearing

Fig. 1. Schematic representation of the different MAC categories.

Table 1Summary of the characteristics of each MAC category. The ‘++’, ‘+’, ‘�’ and ‘� �’ signs denote how well each category is able to provide a specific characteristiccompared with the others.

Category Energy awareness Lack of sync. requirements Simplicity Traffic load adaptability

TDMA-like ++ � � � � �CAP + � � �Asynchronous ++ ++ ++ �


large because it should give time for transmitting synchro-nism messages, as well as control and/or data packets. Ifcompared to TDMA-like MAC protocols, preamble sam-pling is much less complex, because no schedule shouldbe computed and distributed [3]. Table 1 shows a sum-mary of the characteristics that each category of the MACprotocols provide. This table shows how asynchronousMAC protocols are able to reduce energy consumptionwhile taking into consideration the lack of synchronisationand simplicity requirements. It is also important to notethat none of the basic MAC protocols in each category isable to provide traffic load adaptability by itself.


The special suitability of preamble sampling for WSNshas motivated this work, which provides a deep survey ofthe basic preamble sampling technique and its extensions.

4. Preamble sampling MAC protocols

In the basic preamble sampling technique, sensor nodessleep and periodically wake up only to sample the channel.If the channel is determined to be busy, the nodes remainawake; otherwise, they return to sleep. The time betweenchannel samples, called the check interval (Tci), is fixedand is known by all of the nodes in the network. To ensure




t

Long PreambleTx

Rx

Data

Rx

Overhearing

Data Tx Start Data

t

t

t

t

t

t

Tx

Rx

Tx

Rx

Data

Rx

Rx

Overhearing

Overhearing

Early ACK

t

t

ACKTx

Rx

Data

Rx

Overhearing

Rx

Dual Sampling

t

t

ACKTx

Rx

Data

Rx

Overhearingt

t

t

t

1.Basic LPL

2.Short PreamblesDest Info

3.Short Preambles

4.Short Preambles

5.Short Preambles

Fig. 2. Schematic representation of preamble sampling with short packet burst extensions.


the correct reception of packets, each message is precededby a long preamble transmission that must overlap withthe listening time of the receiver; thus, its duration mustbe at least equal to Tci (see Fig. 2(1)).

El-Hoiydi [14] and Hill et al. [15] presented the pream-ble sampling technique at approximately the same time in2002. The former combined this technique with Aloha andcalled it preamble sampling, while the latter combined itwith Carrier Sense Multiple Access (CSMA) and called itLow Power Listening (LPL).1

Later, Polastre et al. defined Berkeley MAC (B-MAC) [12], apreamble sampling MAC protocol with an improved Clear

1 In this work, preamble sampling and LPL will be used interchangeably torefer to the same technique: sending a long preamble before each datatransmission.


Channel Assessment (CCA). Using CCA, the noise floor is esti-mated by taking samples when the channel is supposed to befree (for instance, immediately after transmitting a packet).Moreover, the decision of whether the channel is clear ismade based on the detection of outliers (channel energy sig-nificantly below the noise floor) because a valid packet couldnever have an outlier. This technique reduces the number offalse negatives compared with taking only one sample andcomparing that sample with the noise floor. B-MAC also de-fines a set of interfaces to control the protocol parameters:back-off (BO), enable/disable CCA, set the value of the dutycycle and enable/disable the acknowledgement (ACK).

Waking up to sample the channel allows preamble sam-pling MAC protocols to consume small amounts of energywhen the traffic load is small (which is the usual case inWSNs). The main reason for this energy waste reduction




is the decrease in idle listening. However, basic preamblesampling protocols have several disadvantages, due tothe transmission of the long preamble, as follows:

(1) High overhead: The long preamble transmissiondepends on the sleep period of the duty cycle. There-fore, if a long sleep period is set, a high overhead isinserted at each transmission due to the requiredlength of the preamble. This construct results in alow saturation throughput and in a high transmit-ting and receiving cost.

(2) Costly collisions: If there is a collision, then the entirelong preamble is involved, causing a high collisionduration (the long preamble can be in the order ofa thousand milliseconds). Moreover, if Aloha is usedor if hidden terminals exist, then the vulnerabilitytime is equal to the duration of the long preambleand the packet transmissions. This effect furtherreduces the achieved saturation throughput.

(3) High overhearing cost: Using the basic preambletechnique, overhearing nodes receive the entire pre-amble and the data message because they are notable to identify the destination of the transmissionuntil the data packet is received. This effect resultsin a considerable waste of energy consumption.

(4) Incompatibility with newer radios: Sending a longpreamble with a variable size is not compatible withmodern radios, such as the CC2420 [16], in whichonly fixed short preambles can be transmitted.

Moreover, the basic preamble sampling technique lackstraffic load adaptability because the duty cycle is fixed forall of the sensor nodes independent of their traffic loadrequirements. It has to be considered that variations inthe duty cycle of a sensor node must be known by itsneighbours for them to adapt the duty cycle accordingly.

Several approaches exist that address some of the previ-ously described disadvantages of the basic preamble sam-pling technique. These approaches can be divided intothree different categories: i) mechanisms that divide thelong preamble into a burst of short packets, ii) approachesthat take advantage of synchronisation information and iii)protocols that adapt the duty cycle.

4.1. Division of the long preamble into a burst of short packets

This category of preamble sampling MAC protocols di-vides the long preamble into a series of short packets withthe main goal of including some useful information inthem. Moreover, this approach is fully compatible withmodern packetised radios.

The most common approach is to include the destina-tion identifier (ID) in the short packets, thus allowing over-hearing nodes to return to sleep for the purpose ofreducing the overhearing cost (see Enhanced B-MAC (ENB-MAC) [17]). This type of protocol is represented in Fig. 2(2).Observe that when the destination information is included,the overhearing node can immediately go to sleep.

Other proposals also include information about whenthe data transmission will start; by doing this, the destina-tion can go to sleep after receiving the short packet and can


wake up to receive the data (see also Fig. 2(3), where thisbehaviour is represented). Examples of this behaviour arethe BMAC+ [18], SpeckMAC-B [19], Divided Preamble Sam-pling MAC (DPS-MAC) [20], Synchronous Wake-Up Frame(SyncWUF) [21] and Signalling-Embedded Short PreambleMAC (SESP-MAC) [22] protocols.

4.1.1. Transmission of an early acknowledgementSome approaches also permit the transmission of an

early ACK that stops the transmission of the burst. To allowthe transmission of the early ACK, a gap should be allowedbetween short packets. This approach implies that an extramechanism should be defined to avoid a receiver’s goingback to sleep in case it wakes up in a gap of the burst. Mostof the existing approaches define that sensor nodes mustlisten during at least the gap period each time they wakeup to sample the medium. These protocols are representedin Fig. 2(4) in which the increased wake-up period can beobserved. Some examples of this behaviour are the MACprotocols Transmitter Initiated CyclEd Receiver (TICER)[23], Minimum Preamble Sampling MAC (CSMA-MPS)[24], X-MAC [25], Adaptive Schedule MAC (AS-MAC) [26],Patterned Preamble MAC [27], Asynchronous Real-timeEnergy-efficient and Adaptive MAC (AREA-MAC) [28], Pre-amble Sampling with State Information [29] and 1-hopMAC [30]. Other approaches, in contrast, define a doublecheck mechanism in which the channel is sampled twiceeach time that a sensor node wakes up. The interval be-tween channel samples is such that, even if the first sampleis taken in a gap, the second sample detects the short pack-et burst. Refer also to Fig. 2(5) in which this mechanism isrepresented and in which it can be seen that the first wake-up does not detect the short preamble, while the seconddoes. The first definition of this technique appeared in DualPreamble Sampling MAC (DPS-MAC’) [31] and is also usedin Convergent MAC (CMAC) [32].

4.1.2. Repetitions of the data packetOther approaches, instead of sending a burst of short

packets, including some useful information, send repeti-tions of the data message. By sending repetitions, the des-tination can immediately receive the packet when it wakesup. These approaches are usually combined with the earlyACK technique to reduce the overhearing cost. Examples ofthis approach are the MX-MAC [33] and the Preamble Sam-pling with State Information [29] protocols. On the con-trary, SpeckMAC-D [19] sends repetitions of the datamessage until it is ensured that the destination has wokenup and received the message. In this case, the sender doesnot request any type of ACK from the recipient.

4.1.3. Packet-dependent behaviourOther protocols, such as Multi-mode Hybrid MAC (MH-

MAC) [34], DPS-MAC’ [31], MIX-MAC [35] and TrafficAware MAC (TrawMAC) [36], perform different operations,depending on the type of packet to be sent. In the MH-MACand DPS-MAC’ protocols, the early ACK technique is al-lowed for unicast messages, while the time until the datatransmission start is included in broadcast messages,allowing receivers to go to sleep until the beginning ofthe transmission. MIX-MAC selects from a pool of MAC




protocols (compatible among them) the best to be used foreach packet transmission. For instance, for small packetsand a low sending rate versus receiving rate, the X-MACprotocol (based on short preambles) is used, while forthe rest of the cases, except broadcast messages, the MX-MAC protocol (based on repetitions of the data message)is used. In both cases, the early ACK is allowed. For broad-cast messages, MIX-MAC uses SpeckMAC-D. In contrast,TrawMAC uses short packet preambles for long data mes-sages and repetitions of the data packet for smallmessages.

4.1.4. Other approachesIt is worth mentioning the Micro-Frame Preamble MAC

(MFP-MAC) [37]. This protocol includes a hash of the datain the short packets to allow nodes to identify whetherthey have already received the message, which is usefulfor reducing the load of broadcast communications.

More elaborate approaches are the protocols CMAC[32], 1hopMAC [30] and RA-MAC [38]. In CMAC and 1hop-MAC, anycast transmission is used. Then, those nodes witha better routing cost to the sink reply before using an earlyACK. In CMAC, the early ACK stops the preamble transmis-sion, while, in 1hopMAC, the sender selects the destinationbased on the early ACK responses of the neighbourhood. Adifferent approach is defined in RA-MAC, which addressesthe problem of concurrent data transmissions and definesthat sensor nodes sending a short packet burst as a pream-ble are capable of aggregating bursts from other nodes ifthey overhear their burst transmissions. The destinationthus creates a schedule for data transmission, alleviatingthe collision probability.

A completely different technique is the SEESAW [39]protocol. Using SEESAW, nodes willing to transmit send anumber of advertisements during their active time thateventually overlap with the listening time of the receiver.Moreover, with the goal of increasing the network lifetime(defined by the authors as the time until the first node runsout of battery), the overhead cost is balanced among sensornodes. Each node adapts its own parameters (listeningtime and number of advertisements sent per active time)based on the remaining energy, the listening time andthe number of advertisements of its neighbours.

4.2. Taking advantage of synchronisation information

This category of MAC protocols combines preamblesampling techniques with a synchronisation mechanismand has the main goal of reducing the length of the longpreamble. These protocols can be divided into two maingroups: protocols that make sensor nodes learn thewake-up time of the receivers and protocols that synchro-nise the wake-up time of all of the neighbouring nodes.

4.2.1. Remembering the wake-up time of the receiversThe first group of MAC protocols make sensor nodes

remember the wake-up time of the receivers with the goalof reducing the long preamble duration specifically to ac-count for the possible synchronisation error. This techniquefirst appeared in the definition of Wireless Sensor MAC(WiseMAC) [13], which was presented by El-Hoiydi et al.


in 2004. In this protocol, sensor nodes include the informa-tion of the next channel sampling time in ACK messages. Inthis way, transmitters can store this information and use itthe next time that new data is available for that node. Know-ing the channel sampling time allows the sender to becomesynchronised with the receiver and to transmit a preambleof length Lp = min (4htsync, Tci) where h is the clock drift, tsync

is the time since the last synchronisation happened and Tci isthe sampling period. The computed value of the preambleaccounts for the possible clock drifts of the sender and thedestination with the maximum value of the usual long pre-amble duration set to the listen plus sleep time. Observe inFig. 3, where a representation of the WiseMAC protocol isprovided, how these techniques are able to considerably re-duce the long preamble transmissions. However, note thatone disadvantage of these approaches is that short pream-bles cannot be used to send broadcast messages. In this case,long preambles should be used instead.

This approach is easily combined with the division of thelong preamble into a burst of short packets. Immediatelybefore the sampling time of the receiver, the sender startstransmitting the burst of short packets that can be stoppedby an early ACK or that can include information of when thedata transmission will start. Examples of this behaviour arethe protocols CSMA-MPS [24], SyncWUF [21] and TrawMAC[36], which were already seen in the previous section.CSMA-MPS, SyncWUF and TrawMAC adapt the length ofthe short packet burst according to the time since the lastsynchronisation information from the receiver is known.Moreover, they allow the reception of an early ACK thatmakes the transmitter stop the burst, further reducing thetransmitting and receiving costs compared with WiseMAC.

One disadvantage of this category of MAC protocols isthat they can suffer from continuous collisions when sen-sor nodes share the same wake-up time. The MAC protocolAsynchronous Scheduled MAC (AS-MAC’) [40] tries to ad-dress this problem by forcing nodes in a neighbourhoodto select a different wake-up time. It allows for a reductionin contention, but there is a cost associated with the adver-tisement of the schedule because each node broadcasts anadvertisement if it wakes up and nothing is received.Moreover, broadcast communication is poorly supported(repetitions of the message should be performed or a longpreamble should be transmitted). The authors of WiseMACalso define a technique to address this issue in [13]. Beforetransmitting the short preamble, a medium access reserva-tion preamble of randomised length is sent. Then, colli-sions to destinations that share the same wake-up timeby nodes using nearly equal short preambles are reduced.

4.2.2. Synchronising wake-up schedulesConversely, the second group of MAC protocols that take

advantage of synchronisation information schedule sensornodes to wake up at the same instant and use preamble sam-pling techniques to decrease the impact of the synchronisa-tion error. This type of solution is compatible with broadcastcommunication because all nodes wake up at nearly thesame moment; however, they normally suffer from an in-creased channel contention. In these protocols, if comparedwith the approaches that remember the wake-up time of thereceivers, there is a reduction of the information that is re-




Basic LPL

t

Long PreambleTx

Rx

Data

Rx

Overhearingt

Rx

t

Rx Wake upTime Storage

t

t

Rx

Tx

Rx

Data

Overhearingt

Fig. 3. Schematic representation of preamble sampling with remember the wake-up time of the receiver extension.


quired to be stored (because all nodes share the same wake-up time), but, in contrast, there is an increased cost of syn-chronisation to maintain the schedule.

One of the most well-known approaches is the Sched-uled Channel Polling MAC (SCP-MAC) [41] protocol. InSCP-MAC, the maintenance of the common schedule isachieved by periodically sending synchronisation mes-sages and/or piggybacking synchronisation information indata packets, depending on the rate at which packets aresent (i.e., if the rate is low, then explicit synchronisationmessages are needed). It is important to note that piggy-backing synchronisation information allows for a reduc-tion in the cost of maintaining the schedule. Moreover,the authors of SCP-MAC address the increased contentionby defining two contention phases per packet, one for thepreamble and the other for the data message. In thisway, only winners of the first contention (those nodes thatwere able to transmit the preamble, i.e., no busy channeldetected before transmission) try to send the data mes-sage; therefore, the collision probability is reduced. A sim-ilar approach is followed in Crankshaft [42] where apreamble is used to reduce the contention regarding thedata message. This protocol belongs to the TDMA-like cat-egory because sensor nodes have slots assigned to listenfor data from their neighbouring nodes. However, the sameslot can be owned by different sensor nodes. This reasonmotivates the use of long preambles to reduce contention.

Another approach of this group is the MIX-MAC [35]protocol, which takes advantage of the information of thewake-up times of the receivers to synchronise all nodesin a path and to stagger their transmissions with the goalof reducing the end-to-end delay of a burst of packets froma sensor node to the sink.

4.3. Duty cycle adaptation

There are different approaches in the literature toadapting the duty cycle and, also, the preamble durationto achieve a specific goal, usually to handle increases inthe traffic load.


4.3.1. Request-basedSome protocols adapt the duty cycle based on requests

from the neighbourhood (normally embedded in packettransmissions). For instance, the WiseMAC more bit [13]is used by senders to indicate to the receiver to stay awake(not follow its duty cycle) when they have more packets tosend. An extension to this approach is the Stay AwakePromise introduced in [43]. This extension dictates thatwhen a sensor node receives a message with the more bitset to 1, it replies with an ACK with the stay awake promisebit also set. As a result, other nodes willing to transmit tothe same destination can contend for the channel immedi-ately after the current data transmission. This approach isuseful for handling bursts of packets from different nodes;however, it is not effective for different sensor nodes thatwant to send only one message. A similar approach is de-fined in the Burst-Aware Energy-Efficient Adaptive MAC(BEAM) [44]: the sender marks a bit in the header of apacket if it has more packets to send. If the bit is marked,the destination doubles its duty cycle, allowing the senderto reduce the long preamble and thus increase thethroughput. Another example is the AS-MAC [26] protocol.In AS-MAC, transmitters include the expected next packettransmission time in the messages, so that the receivercan set its duty cycle accordingly. If the sender indicatesthat there are no more packets to transmit, the receiversets the sleep time to the maximum. On the contrary, ifthe transmitter has another packet pending, the receiversets the sleep time to the minimum, which is defined asthe time needed to forward the current received packet.Also, in AREA-MAC [28], sensor nodes adapt their duty cy-cle based on specific requests of their neighbours, whichare performed to handle real-time data to minimise the de-lay from source to sink. Finally, the last example is theAsynchronous Sensor MAC (AS-MAC’’) [45], in which thetransmitter indicates the duty cycle to follow by the recei-ver when there is a burst of packets to be sent. Receiverscan have requests from different nodes in that case.AS-MAC’’ defines that the request with the smallest sleepinterval should be considered.




4.3.2. Based on the traffic loadAnother technique is to base the duty-cycle adaptation

on the observed traffic load. An example is the work pre-sented in [46], which proposes an extension of B-MAC+and defines that, after detecting an increase in the numberof incoming packets, a receiver must increase its dutycycle. After that, the sensor node informs its neighboursabout that change. The authors specify that the notificationis included in the short packets of the burst sent beforetransmitting a packet. A set of predefined listening modesassociated with an incoming rate must be defined andknown by all of the sensor nodes. A similar approach isthe BoostMAC [47] protocol. In this approach, sensor nodesincrease their sleep times in additive steps (until a certainvalue is reached) if they find the medium idle when theywake up. Otherwise, if a high level of traffic is observed,they decrease the sleep time in a multiplicative manner.Instead of announcing the sleep times that each node isusing, senders decide which preamble length is used basedon their own traffic and the cost of sending at different pre-amble lengths, given that the probability that the packetmust be retransmitted because the receiver is sleeping(this probability is estimated and fine tuned throughoutthe lifetime of the network). Another example is theMaximally Traffic-Adaptive MAC (MaxMAC) [48] protocol,which provides an extension for WiseMAC-like protocols.Using MaxMAC, nodes double their duty cycle if thenumber of received packets exceeds a threshold. They dou-ble the duty cycle once again if a second threshold issurpassed. Finally, after exceeding a third threshold, thesensor nodes start working in a CSMA fashion withoutgoing to sleep. ACK messages are used to inform aboutthe change in the duty cycle and the duration of the newconfiguration; therefore, neighbouring nodes can adaptthe length of the short packet burst accordingly. A similarapproach is the Low Power Listening with Wake-Up afterTransmissions MAC (LWT-MAC) protocol [49] in whichall nodes that overhear a packet transmission wake up atthe end of it and are allowed to send packets (to any des-tination) without using the long preamble first. In this pro-tocol, sensor nodes behave in a CSMA operation as soon asthe network load increases, which improves the networkperformance. In a similar way, SCP-MAC [41] adds extrachannel samples immediately after the reception of a mes-sage. The sensor node returns to its regular duty cycle ifnone of these extra channel samples is useful, i.e., no moremessages are received.

Two more approaches that fit in this category are thecontrol algorithms used for setting the sleep schedulepresented in [50]: the Asymmetric Additive Duty CycleControl (AADCC) and the Dynamic Duty Cycle Control(DDCC). In AADCC, each node increments its sleep time in100 ms increments when it successfully sends 5 packetsto the destination. On the contrary, the sleep time is de-creased 250 ms if there is a single packet failure. The goalof this algorithm is to react to an increased channel conten-tion: if 5 consecutive packets are correctly transmitted,then the channel contention is low and nodes can sleepfor longer periods to save energy. In contrast, a packet fail-ure can indicate a high contention for the channel; there-fore, nodes should sleep for shorter periods. This


algorithm can work better than a static sleep time; how-ever, the algorithm does not take into account the energyconsumption or the required data rate. To this aim, theauthors define DDCC. The goal of DDCC is to regulate thesleep interval to accommodate the number of packets tobe sent and to minimise the energy consumption. The algo-rithm is well-suited for single-hop networks; however, inmulti-hop sensor networks, there are several key aspectsthat complicate its design: the variability of the one-hopdelay of a path should be reduced and sensor nodes mustknow information about the entire path to the destination.Both in AADCC and DDCC, nodes must exchange informa-tion about their current sleep time. Finally, another ap-proach is the Zero Configuration Algorithm (ZeroCal) [51],in which sensor nodes select the sleep interval that pro-vides the required bandwidth and minimises their energyconsumption and the energy consumption of their chil-dren. To compute that sleep interval, an energy consump-tion model is used. It is assumed that sensor nodes will runthe energy consumption model to compute the sleep inter-val to use. The computation is made at fixed intervals orwhen the number of sent packets of a child exceeds a giventhreshold. The sleep interval that each node is using is pig-gybacked in data packets. A similar approach is the trafficadaptation mechanism of X-MAC [25], which allows sensornodes to select the sleep interval to use based on their esti-mation of the traffic load. However, in this case, sensornodes do not need to compute an analytical model; theyjust look up values and interpolate them from a table ofpre-computed values.

A completely different approach appears in the workpresented in [52]. Sensor nodes compute their nextwake-up time based on the expectation of receiving anew packet with the goal of minimising the delay or theenergy consumption. To compute the expectation ofreceiving a packet, the distribution of the packet inter-arri-val times should be known, although they present a learn-ing algorithm to approximate it.

4.3.3. Based on topology informationThere are other approaches that adapt the duty cycle

based on the position of the node in the topology. For in-stance, Energy Aware Adaptive LPL (EA-ALPL) [53] assumesa tree topology and allows each sensor node to set its ownduty cycle according to its offered load and its number ofdescendants in the routing tree. Nodes learn the numberof children by counting the number of messages they haveforwarded in a given interval. The duty cycle information isexchanged among neighbours and stored. Similarly, inPreamble Sampling with State Information [29], each nodeselects its duty cycle depending on its position on the tree.Nodes that are closer to the sink set a small sleep time(decreasing the energy of sending preambles), while nodesin the leaves configure a longer sleep time. However, theauthors leave open how to set the values of the sleep inter-vals depending on the position in the tree.

4.4. Summary

As a summary of the previous section, the classificationof the different MAC protocols is shown in Table 2.



Table 2List of preamble sampling MAC protocols.

Category List of MAC protocols

Basic Preamble sampling [14], LPL [15], B-MAC [12]

Short preambleburst

ENBMAC [17], BMAC+ [18], SpeckMAC-B [19],DPS-MAC [20], SyncWUF [21], SESP-MAC[22], TICER [23], CSMA-MPS [24], X-MAC [25],AS-MAC [26], Patterned Preamble MAC [27],AREA-MAC [28], 1-hopMAC [30], DPS-MAC’[31], CMAC [32], MX-MAC [33], Preamblesampling with state information [29],SpeckMAC-D [19], MH-MAC [34], MIX-MAC[35], TrawMAC [36], MFP-MAC [37], RA-MAC[38], SEESAW [39]

Taking advantageof sync. info.

WiseMAC [13], CSMA-MPS [24], SyncWUF[21], TrawMAC [36], AS-MAC’ [40], SCP-MAC[41], MIX-MAC [35]

Adaptative dutycycle

WiseMAC more bit [13], Stay Awake Promise[43], BEAM [44], AS-MAC [26], AREA-MAC[28], AS-MAC’’ [45], extension to B-MAC+[46], BoostMAC [47], MaxMAC [48], LWT-MAC [49], SCP-MAC [41], AADCC and DDCC[50], ZeroCal [51], X-MAC [25], EA-ALPL [53],Preamble Sampling with State Information[29]


In Table 3, the characteristics of the different extensionsof preamble sampling are reviewed. In the case of shortpreamble burst solutions, the only modification that needsto be introduced to achieve a notable reduction of the over-hearing energy consumption is the division of the long pre-amble into short bursts and the insertion of extrainformation into them; as a result, the low level ofcomplexity is maintained. If the early ACK is enabled, thecomplexity is increased but not in a notable manner. Theextensions that take advantage of synchronisation infor-mation are able to reduce substantially the energy wasteby slightly increasing the synchronisation requirementsand the complexity. Finally, the approaches that adaptthe duty cycle are able to adapt to the traffic load at thecost of an increased complexity. Observe that both the traf-fic load adaptation and the complexity are higher in thoseprotocols that adapt the duty cycle based on the estimationof the traffic load.

5. Selection guideline

As previously pointed out, the selection of the MAC pro-tocol to use strictly depends on the application to be pro-vided. The usual recommendation is to use TDMA-likeMAC protocols for high traffic loads, protocols with com-mon active periods for applications with periodic traffic

Table 3Summary of the characteristics of each preamble sampling extension. The ‘++’, ‘+’, ‘�characteristic compared with the others.

Category Energy awareness Lack of sy

Short preamble burst + ++Taking advantage of sync. info. ++ +Adaptive (requests) + ++Adaptive (load) + ++


and asynchronous protocols for low traffic loads [5]. In thissection, a more elaborated guideline to select the best MACcategory is provided. The guideline takes into account thepreamble sampling extensions reviewed in the previoussection.

In Fig. 4, the recommended guideline is presented in aflowchart format. The most suitable MAC protocol is sug-gested, given the most important requirement of anapplication:

� It is recommended to use TDMA-like protocols forapplications with high traffic load requirements andprotocols with common active periods in cases of lowtraffic load and periodic traffic profiles only, as sug-gested in [5].� Also, for applications that require the network to sup-

port increases in the traffic load (due to event-basedtraffic) with high QoS requirements, TDMA-likeapproaches are also recommended. Having slotsassigned to sensor nodes enables the network to easilyhandle increases in the network load and, therefore,maintain the required QoS for the application.� Current preamble sampling with adaptive duty cycles

cannot provide strict QoS to the application because,although there is a reduction of the time to send a packetwhen the network load is high, contention can stilldegrade the performance of the network. For this reason,these protocols are recommended when increases in thenetwork load caused by events are expected, but nostrict QoS requirements are needed by these messages.� The last part of the flowchart provides the suggested

protocols when the traffic load is low, when it is aperi-odic and when increases in the network load are notexpected. In these cases, the basic preamble samplingor any of the extensions that divide the long preambleinto short bursts or that take advantage of synchronisa-tion can be used. However, a fine election is providedhere based on some trade-offs and network configura-tions, as follows:– First, it is recommended to use preamble sampling

with some type of technique that exploits synchro-nisation information in which the complexity ofthe solution can be traded off for a reduction ofenergy consumption. As previously discussed, thesesolutions can substantially reduce the energy con-sumption but at the cost of a slightly increased com-plexity. Taking advantage of synchronisationinformation requires sensor nodes to either obtainand store the wake-up time of the receivers or tocreate and maintain a scheduled wake-up time(depending on the subcategory used). This construct

’ and ‘� �’ signs denote how well each category is able to provide a specific

nc. requirements Simplicity Traffic load adaptability

++ �+ �� +� � ++



be traded−off for a reductionin energy consumption?

Can complexity

Appl. & NetworkRequirements

no

yes

application have a high loadrequirement?MAC Protocols

Use TDMA−like

PeriodsCommon Active

Use protocols with yes

no

network need to supportperiodic traffic only?

no

yes

Does the

Does the

no

yes

with adaptive duty cycleUse preamble sampling

Is the network density high?

no

yes

yes

samplingUse basic preamble

no

with short packet burstUse preamble sampling

Use preamble samplingtaking advantage of synchronisation info

due to events?increases of traffic

need to support sporadicDoes the networkDoes the

messages?for event−based

strict QoS requirementsapplication impose

Fig. 4. Guideline to select the MAC protocol for a given application.


PleNe

implies an increased complexity if compared to theprotocols that divide the long preamble into burstsof short packets and basic preamble sampling.

– Next, it has been suggested that one of the majoradvantages of the division of the long preamble intoa short packet burst is the reduction of the overhear-ing energy consumption. For this reason, this cate-gory has been recommended for networks withhigh density (a high number of neighbouring sensornodes).

ase cite this article in press as: C. Cano et al., Low energy operation itw. (2011), doi:10.1016/j.comnet.2011.06.022

– Finally, the use of basic preamble sampling is sug-gested in cases in which simplicity is more impor-tant than a reduction in energy consumption andthe density is not high.

6. Concluding remarks

Given the high application dependence of WSNs, thereis not one MAC protocol able to satisfactorily work in everydeployment. However, preamble sampling is able to pro-




vide some interesting capabilities, which are especiallyappealing to WSNs that usually work in low traffic loadconditions and that are formed by low-capability and en-ergy-constrained devices.

The benefits and disadvantages of basic preamble sam-pling have been outlined and used as the basis for the ex-tended review of the different preamble samplingprotocols that are presented here.

The energy consumption of the basic preamble samplingtechnique can be further reduced in protocols that dividethe long preamble into short burst packets and in protocolsthat exploit synchronisation information. The adaptiveapproaches, in contrast, apart from attempting to keep theenergy consumption low, aim to adapt to the traffic load,improving the system throughput and reducing the delayof the generated packets. These last approaches also implya greater complexity compared with the basic preamblesampling technique, especially the adaptive techniques thatare based on estimating the traffic load.

Acknowledgment

This work has been partially supported by the SpanishGovernment under projects TEC2008-06055 (Plan NacionalI+D) and CSD2008-00010 (Consolider-Ingenio Program),and by the Catalan Government (SGR2009n#00617).

References

[1] I. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, Wireless sensornetworks: a survey, Computer Networks 38 (4) (2002) 393–422.

[2] I. Demirkol, C. Ersoy, F. Alagoz, MAC protocols for wireless sensornetworks: a survey, Communications Magazine IEEE 44 (4) (2006)115–121.

[3] K. Kredo II, P. Mohapatra, Medium access control in wireless sensornetworks, Computer Networks 51 (4) (2007) 961–994.

[4] K. Langendoen, Medium access control in wireless sensor networks,Medium Access Control in Wireless Networks, Volume II: Practiceand Standards.

[5] A. Bachir, M. Dohler, T. Watteyne, K.K. Leung, MAC essentials forwireless sensor networks, IEEE Communications Surveys andTutorials 12 (2) (2010) 222–248.

[6] IEEE Standard 802.11, Wireless LAN Medium Access Control (MAC)and Physical Layer (PHY) Specifications, ANSI/IEEE Std 802.11.

[7] S. Ergen, P. Varaiya, PEDAMACS: power efficient and delay awaremedium access protocol for sensor networks, IEEE Transactions onMobile Computing 5 (7) (2006) 920–930.

[8] L. Van Hoesel, P. Havinga, A Lightweight medium access protocol(LMAC) for wireless sensor networks, in: Proceedings of the 1stInternational Workshop on Networked Sensing Systems (INSS), 2004.

[9] W. Ye, J. Heidemann, D. Estrin, Medium access control withcoordinated adaptive sleeping for wireless sensor networks, IEEE/ACM Transactions on Networking 12 (3) (2004) 493–506.

[10] T. van Dam, K. Langendoen, An adaptive energy-efficient MACprotocol for wireless sensor networks, in: Proceedings of the FirstInternational Conference on Embedded Networked Sensor Systems(SenSys ’03), 2003, p. 171.

[11] W. Ye, J. Heidemann, D. Estrin, An energy-efficient MAC protocol forwireless sensor networks, in: Proceedings of the Twenty-FirstAnnual Joint Conference of the IEEE Computer andCommunications Societies, vol. 3, 2002, pp. 1567–1576.

[12] J. Polastre, J. Hill, D. Culler, Versatile low power media access for wirelesssensor networks, in: Proceedings of the 2nd International Conference onEmbedded Networked Sensor Systems (SenSys ’04), 2004.

[13] A. El-Hoiydi, J. Decotignie, WiseMAC: an ultra low power MACprotocol for multi-hop wireless sensor networks, AlgorithmicAspects of Wireless Sensor Networks (2004) 18–31.

[14] A. El-Hoiydi, Aloha with preamble sampling for sporadic traffic in Adhoc wireless sensor networks, in: Proceedings of the IEEEInternational Conference on Communications. (ICC 02), vol. 5,2002, pp. 3418–3423.


[15] J. Hill, D. Culler, Mica: a wireless platform for deeply embeddednetworks, IEEE Micro (2002) 12–24.

[16] Chipcon, CC2420. 2.4 GHz IEEE 802.15.4/ ZigBee-ready RFTransceiver. <http://focus.ti.com/lit/ds/symlink/cc2420.pdf>.

[17] S. Lim, S. Kim, J. Cho, S. An, Medium access control with anenergy-efficient algorithm for wireless sensor networks, in:Proceedings of the Personal Wireless Communications, 2006,pp. 334–343.

[18] M. Avvenuti, P. Corsini, P. Masci, A. Vecchio, Increasing the efficiencyof preamble sampling protocols for wireless sensor networks, in:Proceedings of the First Mobile Computing and WirelessCommunication International Conference (MCWC 06), 2006, pp.117–122.

[19] K. Wong, D. Arvind, SpeckMAC: Low-power decentralised MACprotocols for low data rate transmissions in specknets, in:Proceedings of the 2nd International Workshop on Multi-hop Adhoc Networks: from Theory to Reality, 2006, pp. 71–78.

[20] S. Lim, Y. Ji, J. Cho, S. An, An ultra low power medium access controlprotocol with the divided preamble sampling, in: Proceedings of theThird International Symposium on Ubiquitous Computing Systems,2006, pp. 210–224.

[21] X. Shi, G. Stromberg, SyncWUF: an ultra low-power MAC protocol forwireless sensor networks, IEEE Transactions on Mobile Computing 6(1) (2007) 115–125.

[22] K. Han, S. Lim, S. Lee, J. Lee, S. An, Signaling-embedded shortpreamble MAC for multihop wireless sensor networks, in:Proceedings of the International Conference on InformationNetworking. Towards Ubiquitous Networking and Services, LectureNotes in Computer Science, 2008, pp. 1–10.

[23] E. Lin, J. Rabaey, A. Wolisz, Power-efficient Rendez-vous schemes fordense wireless sensor networks, in: Proceedings of the IEEEInternational Conference on Communications (ICC 04), 2004, pp.3769–3776.

[24] S. Mahlknecht, M. Bock, CSMA-MPS: A minimum preamble samplingMAC protocol for low power wireless sensor networks, in:Proceedings of the IEEE International Workshop on FactoryCommunication Systems, 2004, pp. 73–80.

[25] M. Buettner, G. Yee, E. Anderson, R. Han, X-MAC: A shortpreamble MAC protocol for duty-cycled wireless sensor networks,in: Proceedings of the fourth International Conference onEmbedded Networked Sensor Systems (Sensys ’06), 2006, pp.307–320.

[26] L. Bing, Z. Lin, Z. Huimin, An adaptive schedule medium accesscontrol for wireless sensor networks, in: Proceedings of theSixth International Conference on Networking (ICN’07), 2007, pp.12–12.

[27] I. Joe, H. Ryu, A patterned preamble MAC protocol for wirelesssensor networks, in: Proceedings of the 16th InternationalConference on Computer Communications and Networks, 2007,pp. 1285–1290.

[28] P. Kumar, M. Gunes, Q. Mushtaq, B. Blywis, A real-time and energy-efficient MAC protocol for wireless sensor networks, in: Proceedingsof the IFIP International Conference on Wireless and OpticalCommunications Networks, WOCN ’09., 2009.

[29] E. Sanchez, C. Chaudet, M. Rebaudengo, Improving preamblesampling performance in wireless sensor networks withstate information, in: Proceedings of the Sixth InternationalConference on Wireless On-Demand Network Systems andServices, 2009.

[30] T. Walteyne, A. Bachir, M. Dohler, D. Barthe, I. Augé-Blum, 1-hopmac: An energy-efficient mac protocol for avoiding 1-hopneighborhood knowledge, in: Proceedings of the Third Annual IEEECommunications Society on Sensor and Ad Hoc Communicationsand Networks, 2006, pp. 639–644.

[31] H. Wang, X. Zhang, F. Naït-Abdesselam, A. Khokhar, DPS-MAC: Anasynchronous MAC protocol for wireless sensor networks, in:Proceedings of the 14th International Conference on HighPerformance Computing, 2007, pp. 393–404.

[32] S. Liu, K. Fan, P. Sinha, CMAC: an energy-efficient MAC layer protocolusing convergent packet forwarding for wireless sensor networks,ACM Transactions on Sensor Networks 5 (4) (2009) 1–34.

[33] C. Merlin, W. Heinzelman, Network-aware adaptation of MACscheduling for wireless sensor networks, in: Proceedings of theConference on Distributed Computing in Sensor Systems (PosterSession), 2007, pp. 24–28.

[34] L. Bernardo, R. Oliveira, M. Pereira, M. Macedo, P. Pinto, A wirelesssensor MAC protocol for bursty data traffic, in: Proceedings of theIEEE 18th International Symposium on Personal, Indoor and MobileRadio Communications, 2007, pp. 1–5.


http://focus.ti.com/lit/ds/symlink/cc2420.pdf



[35] C. Merlin, W. Heinzelman, Schedule adaptation of low-power-listening protocols for wireless sensor networks, IEEE Transactionson Mobile Computing 9 (5) (2010) 672–685.

[36] X. Zhang, J. Ansari, P. Mähönen, Traffic aware medium access controlprotocol for wireless sensor networks, in: Proceedings of the SeventhACM International Symposium on Mobility Management andWireless Access – MobiWAC ’09, 2009, p. 140.

[37] A. Bachir, D. Barthel, M. Heusse, A. Duda, Micro-frame preambleMAC for multihop wireless sensor networks, in: Proceedings of theIEEE International Conference on Communications (ICC 06), 2006,pp. 3365–3370.

[38] D. Lee, K. Chung, RA-MAC: an energy efficient and low latency MACprotocol using RTS aggregation for wireless sensor networks, in:Proceedings of the International Conference on AdvancedTechnologies for Communications, 2008, pp. 150–153.

[39] R. Braynard, A. Silberstein, C. Ellis, Extending network lifetime usingan automatically tuned energy-aware MAC Protocol, in: Proceedingsof the Third European Wireless Sensor Networks Workshop (EWSN2006), 2006, pp. 244–259.

[40] B. Jang, J. Lim, M. Sichitiu, AS-MAC: An asynchronous scheduled MACprotocol for wireless sensor networks, in: Proceedings of the FifthIEEE International Conference on Mobile Ad Hoc and SensorSystems. (MASS 2008), 2008, pp. 434–441.

[41] W. Ye, F. Silva, J. Heidemann, Ultra-low duty cycle MAC withscheduled channel polling, in: Proceedings of the FourthInternational Conference on Embedded Networked Sensor Systems,2006, p. 334.

[42] G. Halkes, K. Langendoen, Crankshaft: an energy-efficientMAC-protocol for dense wireless sensor networks, WirelessSensor Networks. Lecture Notes in Computer Science (2007)228–244.

[43] P. Hurni, T. Braun, Increasing throughput for WiseMAC, in:Proceedings of the Fifth Annual Conference on Wireless onDemand Network Systems and Services, 2008.

[44] M. Anwander, G. Wagenknecht, T. Braun, K. Dolfus, BEAM: A burst-aware energy-efficient adaptive MAC protocol for wireless sensornetworks, in: Proceedings of the Seventh International Conferenceon Networked Sensing Systems, 2010.

[45] J. On, J. Kim, J. Lee, Y. Kim, H. Chong, A MAC protocol with adaptivepreloads considering low duty-cycle in WSNs, in: Proceedings of theThird International Conference on Mobile Ad-hoc and SensorNetworks, 2007, pp. 269–280.

[46] M. Avvenuti, A. Vecchio, Adaptability in the B-MAC+ protocol, in:Proceedings of the IEEE International Symposium on Parallel andDistributed Processing with Applications, 2008, pp. 946–951.

[47] K. Stone, M. Colagrosso, Efficient duty cycling through predictionand sampling in wireless sensor networks, WirelessCommunications and Mobile Computing. Special Issue: CognitiveRadio, Software Defined Radio and Adaptive Wireless Systems(2007) 1087–1102.

[48] P. Hurni, T. Braun, Maxmac: A maximally traffic-adaptive MACprotocol for wireless sensor networks, in: Proceedings of the SeventhEuropean Conference, EWSN, 2010, pp. 289–305.

[49] C. Cano, B. Bellalta, A. Sfairopoulou, J. Barceló, A low power listeningMAC with scheduled wake up after transmissions for WSNs, IEEECommunication Letters 13 (4) (2009) 221–223.

[50] C. Merlin, W. Heinzelman, Duty cycle control for low-power-listening MAC protocols, IEEE Transactions on Mobile Computing(2010) 497–502.

[51] A. Meier, M. Woehrle, M. Zimmerling, L. Thiele, ZeroCal: automaticMAC protocol calibration, Distributed Computing in Sensor Systems1 (2010) 1–14.

[52] X. Ning, C. Cassandras, Dynamic sleep time control in wirelesssensor networks, ACM Transactions on Sensor Networks 6 (3)(2010) 1–37.

[53] R. Jurdak, P. Baldi, C. Videira Lopes, energy-aware adaptive lowpower listening for sensor networks, in: Proceedings of theInternational Workshop for Networked Sensing Systems (INSS),2005.


Cristina Cano obtained the Telecommunica-tions Engineering Degree at the UniversitatPolitecnica de Catalunya (UPC) in February2006. Next, she received an M.Sc. (2007) and aPh.D. (2011) on Information, Communicationand Audiovisual Media Technologies from theUniversitat Pompeu Fabra (UPF). Currently,she is working at UPF on topics related toWireless Networks, Wireless Sensor Net-works, Quality of Service, MAC layer designand Multi-Channel operation.

Boris Bellalta received a B.Sc. degree inTelecommunications from the UniversitatPolitecnica de Catalunya (UPC) in 2002 and aPh.D. from the Universitat Pompeu Fabra(UPF) in 2007, where he combined Ph.D.studies with a fulltime assistant professorposition. Since 2007, he is a post-docresearcher and fulltime lecturer at UPF. Hismain research interests are in the area ofwireless communications, MAC protocols andqueuing models.

Anna Sfairopoulou graduated in ComputerScience from the University of Ioannina,Greece on 2000. She received her M.Sc. (2004)and Ph.D. (2008) degree on Computer Scienceand Digital Communications from the Uni-versitat Pompeu Fabra (UPF). Since 2002, shehas been working at the UPF as a fulltimeresearcher and teaching assistant and from2008 as a Visiting Professor in the Departmentof Information and Communication Technol-ogies of the UPF. Her research interests arefocused in the area of performance analysis

and QoS optimisation of VoIP and streaming video over 802.11 networks,MAC protocols for 802.11 and sensor networks, routing protocols and P2Poverlay architectures for real-time applications.

Miquel Oliver received a degree in Telecom-munications Engineering (1994, UPC), adegree in Business Administration (2009,UOC) and a Ph.D. in Electrical Engineering(1999, UPC). He received a postdoctoralaward, joining the Wireless Information Net-works Laboratory (WINLAB) at Rutgers Uni-versity (NJ, USA). He has been involved inresearch projects regarding wireless accessnetworks following neutrality according to hisongoing research. His research topics includeradio resource management for cellular sys-

tems, quality of service in wireless data networks, IP mobility and radioaccess protocols, including wireless sensor networks and distributedpeer-to-peer protocols. He has also recently researched the ICT impact
upon society, as well as business models for incoming technologies andnetworks.


low energy operation in wsns: a survey of preamble sampling

Documents