An Adaptive Energy-Management Framework for Sensor Nodes ...soilscape.usc.edu/~agnelors/An Adaptive...

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2013 Article ID 272849 33 pageshttpdxdoiorg1011552013272849

Research ArticleAn Adaptive Energy-Management Framework for Sensor Nodeswith Constrained Energy Scavenging Profiles

Agnelo R Silva1 Mingyan Liu2 and Mahta Moghaddam1

1 EE-Electrophysics University of Southern California Los Angeles CA USA2 EECS Department University of Michigan Ann Arbor MI USA

Correspondence should be addressed to Agnelo R Silva agnelorsgmailcom

Received 6 June 2013 Accepted 19 August 2013

Academic Editor Davide Brunelli

Copyright copy 2013 Agnelo R Silva et alThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Modern energy harvesting systems forWSNs involve power scavenging sources rechargeable batteries and supercapacitors Typicalenergy-management systems calculatepredict the remaining energy stored in a node and associated actions are dispatched involv-ing the networking protocols However long-term characteristics of the mentioned hardware components are typically neglectedpreventing the achievement of very longmaintenance-free lifetimes (eg gt5 years) for the nodes In this work a systematic analysisof this problem is provided and an open energy-management framework is proposed which promotes (a) the nontraditionalcombination of primary cells supercapacitors and harvesting systems (b) the concept of a distributed system inside a nodeand (c) the adoption of the dual duty-cycle (DDC) operation for the WSNs The DDCrsquos core component is a cross-layer protocolimplemented as an application-layer overlay which maintains the operation of the network under very high energy efficiency Itstrade-off is the reduction of the network throughputTherefore the DDC system hasmechanisms that dynamically switch theWSNoperational mode according to applicationrsquos needs Detailed guidelines are provided in order to allow the implementation of thesolution on existingWSN platformsThe energy efficiency of the low duty-cycle mode of the solution is demonstrated by simulatedand empirical results

1 Introduction

Despite the possible existence of power scavenging sourcesfor wireless sensor networks (WSN) nodes this fact does notnecessarily imply a longer lifetime or a high reliability levelfor such nodes For instance when a photovoltaic cell is usedthe energy harvesting process is typically not continuous orstable Moreover when rechargeable batteries are part of theenergy harvesting system the lifetime of the node is ulti-mately dictated by the age or by the number of charge cycles ofthose batteries among other factors Even adopting verywell-controlled charging procedures typical secondary cells forWSN nodes have a lifetime smaller than 3 years A potentialsolution to achieve a 5- to 10-year maintenance-free solutionis the adoption of a battery-free design as proposed in [1]where supercapacitors are used as temporary energy reser-voirs However the challenge of such solution is to sustaina certain level of reliability when the capability of the powersource is insufficient for the node operation

The above issue is aggravated when the duty-cycle of thenode is not solely governed by the main application This isthe case when the node must also actively collaborate inthe network Moreover the effort to incorporate energy con-sumption metrics into existing physical and higher-layernetworking protocols is still a challenge and typically suchprovision is not implemented in commercial WSN solutionsAs a result it is very difficult to achieve realistic long lifetimefor the nodes in conjunction with relatively high reliabilitylevels In this work an adaptive and flexible framework forsensor nodes with constrained energy scavenging profiles ispresented This energy-management framework has hard-ware and software components and a significant emphasison integration is given in this work in order to facilitate thepartial or integral adoption of the proposed framework onexisting WSN platforms

The motivation for this work is associated with the goalof having a reliable WSN solution with a lifetime between 5

and 10 years That is during this period of time no human

2 International Journal of Distributed Sensor Networks

intervention is expected due to power depletion of a nodeIt will be shown that to achieve this goal the complexitylevel of the solution at the design-time is relatively highAlso the initial cost of a node is expected to be at least 30higher than a traditional off-the-shelf node However thepractical functionality and the total ownership cost (TCO)of the system at long term can be very attractive Notethat the advocated vision in this work diverges from thetraditional concept that a WSN comprises hundreds of verylow-cost nodes each one individually with a relatively highprobability of failure On the contrary the focus of this workis on the achievement of a very high-quality and controlledsolution

The paper begins with the presentation of the energyeffort tripod and energy control loop concepts in Section 2It is highlighted that effective energy savings for WSN nodespotentially depends on a balanced solution in terms of hard-ware network and application demands In Section 3 thefoundations of the proposed framework are discussed (a) theoptional (but recommended) use of primary cells associatedto harvesting systems (b) the advantages of a distributedsystem inside a node and (c) the adoption of the dual duty-cycle operation (DDC) for WSN nodes The core part of theproposed framework is a cross-layer network protocol whichis presented in Section 4 This protocol is implemented as anapplication-layer overlay on top of existing WSN solutionsSuch overlay mechanism can be dynamically activated anddeactivated in order to allow the network to achieve thebest performancewhile satisfying existing energy constraintsMany of the components of the proposed framework are infact part of a long-term and ongoing project involving oneof the largest outdoors WSN deployments still in operation[2ndash4] The field results of this project in conjunction withsimulated outcomes are reported in Section 5 A discussionrelated to the integration of the framework with otherongoing WSN research efforts is provided in Section 6 andthe paper is concluded in Section 7

2 Energy Management in WSNs

In this section typical pitfalls and challenges in the design ofenergy systems for WSNs are discussed Next the importantenergy effort tripod and energy control loop concepts areintroduced

21 Design Challenges and Pitfalls Many well-designed pro-jects fail due to small details and incorrect (but generallyaccepted) assumptions Therefore before presenting the pro-posed framework it is important to highlight some aspectsassociated with the current state-of-the-art technology onenergy harvesting systems for WSNs

WSNDesign besides Long Lifetime and Reliability One criticalpitfall associated to the energy aspect of WSN designs is toperpetuate the original vision of a WSN with hundreds tothousands of very cheap nodes [5] where a high rate of nodefailures is actually expected Although such vision can stillcorrespond to the needs of some applications a quick inves-tigation at the current WSN deployments around the world

reveals a different trend for WSNs For instance few existinglong-term networks actually have more than 50 nodes Moreimpressive is the ongoing success of the infrastructure-based WSN solutions (star or tree topologies) such as theones based on IEEE 802154ZigBee [6 7] A higher nodereliability is typically crucial when WSNs move from ad hocto infrastructured architectures Accordingly to this currenttrend a significant emphasis in this work is given to thereliability of the nodes In this context the term reliability isassociated with the goal of having nodes that rarely becomeunavailable due to a noncontrolled power depletion

Energy Scavenging Does Not Imply a Perpetual LifetimeAnother pitfall associated to the design ofWSNs is to considerthat the adoption of an energy scavenging system is automat-ically associated to an endless node lifetime Besides the needto consider the life expectancy of the sensing componentssuch as a humidity sensor or a soil moisture probe typicalpower systems can rarely achieve a 5-year lifetime due toa plurality of reasons discussed in this section Thereforethe first step toward a successful low-cost WSN solution (interms of functionality + reliability + long lifetime) is theinvestigation of the expected lifetime of each of the compo-nents of an node In general the reported premature deathcause of WSN nodes is the energy subsystem in particularthe batteries Primary (nonrechargeable) batteries typicallyhave a very short lifetime in WSNs [8 9] However it isalso important to have inmind that secondary (rechargeable)batteries potentially have a lifetime smaller than 2-3 years

Possibility of Adopting a Nonrealistic Energy Model A signifi-cant number of WSN papers present three regular omissionsor inaccuracies regarding to the way the energy model isproposed or adopted First the transients such as due tothe activationdeactivation of a radio transceiver are typicallyneglected as pointed out in [10] Second many values used asinput parameters for the models are directly imported fromthe datasheets of the components without further considera-tion of the effects of integrating these components togetherFor instance based on its datasheet a radio transceiver mod-ule has a nominal sleeping current of 10 120583A However it isobserved that once it is attached to a MCU leak currents aredetectedwhich aremany folds higher than the nominal sleep-ing current Similarly a voltage regulator can be included inthe design of a WSN node in particular when energy har-vesters are also employed However many reported energymodels do not include the energy cost of a possible voltageregulator As a result an energy model that disregards theexistence of voltage conditioners can be drastically distorted(non-linearly) when it is adopted in a node The severity ofthis statement can be illustrated by the following real-caseWhen both MCU and radio modules are sleeping usually itis not possible to put the voltage regulator in sleeping mode(typically shutdownmode in the context of regulators)There-fore rather than an expected amount in 120583W as the powerconsumption for the node (as expressed inmanyWSNenergymodels) the node can potentially have an effective sleepingconsumption on the order of mW Therefore as one reducesthe application duty-cycle the adopted energy model reveals

International Journal of Distributed Sensor Networks 3

29

282726252423

21 Nov 8 Dec 24 Dec 23 Jan 3 Feb 2012

2 N

iMH

bat

terie

s

Batte

ry le

vel (

V)

Sampling time

(a)

Batte

ry le

vel (

V)

36

34

32

30

Sampling time21 Nov 2012 3 Feb 2013

1Li-S

OCl

2ba

ttery

(b)

Figure 1 Effect of subzero temperatures on secondary (a) and primary (b) cells (Ann Arbor MI USA) [9] The recharging process ofthe secondary cells is impacted by low temperatures causing node failures (lines in the figure) Primary cells are more resilient to extremetemperatures

its non-linear distortion once the node is still consuming asignificant amount of energy even in sleep mode

The third pitfall commonly observed in the WSN liter-ature is associated with the expected lifetime of batteries Ingeneral it is assumed that 100 (or a close value) of the initialnominal energy will be actually available for the operation ofthe node Some papers even justify this assumption by high-lighting the fact that bothMCUand radio support low voltagelevels such as 15V In practice it is very hard to achievevalues even close to 80 of the nominal energy capacity ofthe cell Factors that invalidate thementioned assumption arethe self-discharging current the aging of the battery tem-perature discharging regime charging regime (for secondarycells) and so forth As a rule-of-thumb when the batteryreaches its terminal state a significant amount of energy (eggt25) still remains inside the cell However only very lowdischarging currents are typically possible from that momenton Note that even if the load affords a low voltage levelthe bottom line is actually the constraint of having the loadonly draining very tiny currents This is hardly the case inparticular for radio modules in WSN nodes Therefore ifthe hardwaresoftware solution embedded on that node doeshave any provision to use this remaining energy at the cellthe adopted energy model must only consider conservativevalues for the actual initial energy stored at the battery Inmany cases such conservative value is less than half of thenominal energy of the battery

As observed realistic energy models for WSNs are inher-ently complex but they can be simplified if conservative val-ues are adoptedMoreover every time a newWSNplatform isdesigned a significant number of experiments involving thefinal hardware different kinds of batteries and realistic dis-charging regimesmust be considered before an energymodelcan be proposed for that node and also the network On theother hand it is interesting to observe that energy modelsfor batteryless solutions are typically reported as properlymatching the application needs [11] In general it is the casebecause detailed empirical investigation is realized to justifythe energy model in a critical scenario involving a smallamount of available energy at the energy reservoir (eg asupercapacitor)

Lifetime of Rechargeable Batteries Typical secondary cellsused in WSNs require special attention because besides theirinherent shelf lifetime (eg lt3 years) there are other factorsthat can drastically reduce their lifetime For instance themaximumnumber of nominal charge cycles is usually smallerthan 1000 considering the kind of cells typically reportedfor WSN nodes Therefore without a careful control of howand when the charge cycles are performed the lifetime ofsuch cells can be realistically smaller than 1 year Moreovertemperature is a critical factor in particular for secondarycells In general extreme temperature can drastically degradethe cellrsquos performance For instance in [9] it is reportedthat at sub-zero temperatures many solar-powered nodesstopped the charging process followed by long periods ofnetwork inactivity as shown in Figure 1 A solar panel com-pletely covered with snow is one of the potential reasons forthe mentioned issue However for this particular case studythe inability of the secondary cell to be charged at subzerotemperatures was the main reason behind the functionalfailures It is also observed in Figure 1 that nodes powered by aprimary cell (LithiumThionyl Chloride in this example) arenot affected by extreme temperatures

Solid-State Batteries and WSNs Solid-state batteries are arecent technological advance that can impact the design offuture WSN solutions These secondary cells are claimed tohave a lifetime between 5 and 10 years and a maximumnumber of charge cycles between 5000 and 10000 [12] Basedon these preliminary values it is possible to envision a reliableWSN solution with a very long lifetime based on such batter-ies Nonetheless these cells have three significant drawbackshigh cost low-energy density and low-power density Whilethe former aspect can be only amatter of time as a function ofthe industrial scale the remaining aspects reinforce the needof a careful designed energy-management system if such cellsare expected to be used in WSN nodes

Lifetime of Low-Cost Outdoor Energy Harvesters The lifeexpectancy of low-cost energy harvesters for outdoors is typ-ically not informed by the manufacturers For instance todate manufacturers of micro wind turbines do not provide


such information Similarly although relatively big solarpanels (eg gt30 cm times 30 cm) are typically robust and havea realistic lifetime of more than 3 years it is not the casein relation to small solar panels used in WSN nodes Weperformed outdoor tests for more than 2 years with differenttypes and models of small solar panels Unfortunately theresults were very disappointing the majority of the panelspresented a significant performance deterioration in less than1 yearThemajority of them changed their glossy surface by awhite porous surface where dust easily accumulates In one ofthe sites which experiences high temperatures (eg gt40∘C)more than 10 of the panels cracked To the date we did notfind off-the-shelf small solar panels with the typical robustencapsulation found at bigger panels Another critical aspectin relation to small solar panels left unattended outdoors isthe dirt left by birds In our outdoor deployments involvingsites in three USA states we observed the same phenomenona small solar panel mounted on top of a pole is a typical placewhere birds choose to temporarily restWithout a proper pro-tection against the birds such solar panels potentially requireperiodic cleanness The bottom line is the importance ofevaluating the robustness of the components of a harvestingsystem before assuming a perpetual lifetime for a node

22 The Concepts of Energy Effort Tripod and Energy ControlLoop In the previous section some aspects related to theenergy subsystem of a WSN node are highlighted and it wasshown that the achievement of a maintenance-free solutionfor periods of more than 5 years is not a trivial task At thatanalysis the network and application aspects are not con-sidered However the adoption of an energy-managementsolution requires the integration of energy effort in terms ofhardware the network and the application Accordingly thefocus of this section is to discuss how these three aspects areproperly integrated in an energy-management frameworkWe will conclude that the knowledge related to the energystate of the nodes is paramount

The term framework is defined as a broad outline ofinterrelated items not a detailed step-by-step set of strictguidelines The advantage of this design approach is manlythe gains in terms of flexibility one is exposed with someunderline concepts and ideas and can easily adapt them tohis problem or environment in this case a certainWSN plat-form Accordingly the basic concepts available at the WSNenergy-management literature are summarized by two pic-tures presented in this section

The first concept the energy effort tripod is illustrated inFigure 2 Assuming a certain limited energy level for a nodean efficient way to achieve functionality and reliability fora very long period of time is to balance efforts in terms ofhardware network algorithms and application demands Forinstance if advances in the hardwaresoftware of a node allowthe reduction of the sleeping energy consumption of a nodeby one order of magnitude such effort is potentially voided ifthe effective duty-cycle of the node (due to the network theapplication or both) is still very high Similarly a significantreduction of the network overhead can have little energyimpact compared to a very high and frequent applicationdemand (eg multimedia data traffic)Therefore the starting

FunctionalityReliability

NetworkApplication

Hardware

Figure 2 Energy effort tripod concept coordinated efforts involv-ing hardware network algorithms and application demands lead toan energy-balanced and efficient WSN solution

point is to limit the demand of the main application and todefine possible acceptable levels of Quality of Service (QoS)for the nodes a group of nodes and the network As expectedservice metrics are required in energy-management systemsSuch metrics involve data latency volume of data trafficfrequency of data bursts (eg scheduling in data-drivenapplications) data loss acceptancelevels and so forth

Once the application demands are clearly defined andrealistically constrained considering the energy systems andpower sources available for the nodes the next step is toevaluate how the network and the hardware of the nodes canbe improved or in other words balancedwith the applicationdemands In general the adoption of a very flexible WSNsolution (not tailored to a certain category of application) isalso associated with energy-hungry network protocols Forinstance consider a WSN application that every 5 minutesmonitors the infrastructure of a bridge One strategic ques-tion to be considered in this case is related to how the networkprotocols can be optimized considering a static topology andalso a fixed monitoring schedule

Similarly a very energy-efficient hardware module maynot be a balanced solution It is the case when the energy-saving mechanisms provided in the hardware can actuallyimpact the functionality and reliability of the network andultimately the main application For instance consider thecase of a batteryless nodewhich adopts a combination of solarpanel and supercapacitors During the daylight the hardwareof the nodes is functional and energy-efficient and the inten-sive collaboration in the network is not impacted Howeverduring the night the behavior of the nodes can drasticallychange a batteryless node may not be capable of performingregular transmissions multiple times per second even undera very low duty-cycle (eg lt1) regime Nonetheless con-sider the fact that the majority of the current WSN networkprotocols operate assuming that the node wakes up multipletimes per second Therefore it is clear that the power systemdesign and the adopted network solutions are not properlybalanced in this example


Decision

Impact

Data

Data

Energy-managementcontrol

Energy state

Node operation

Figure 3 Energy control loop concept the operation of aWSNnodeis regulated by its energy state The decisions are triggered by anenergy-management module that can be implemented internally inthe node at the network level in a centralized data server or by acombination of these options

In some cases the previous illustrated batteryless solutionfor WSN nodes imposes a certain level of data latency whichis incompatible with the application requirements and againthe balance is not achieved The main point behind thisdiscussion is not about advocating in favor or not in relationto a certain technology It is actually related to what can beadjusted in the WSN design in order to obtain a properbalanced solution in terms of energy functionality andreliability In many cases an optimized solution is complexbecause it is only achieved by combining enhancements inthe hardware in the network and also in the application (ierelaxing the required QoS metrics) For the latter aspect itis clear that control is necessary and in fact this is the roleof energy-management modules as illustrated by the nextconceptual figure

The second concept to be discussed is called energy controlloop as illustrated in Figure 3 Such control can be performedat the node level in a portion of the network by means ofa central data server or by a combination of these optionsAs shown by Figure 3 the operation of the node such asthe activation of an energy harvester the activation of theradio module or the way the node behaves in the networkis governed by the decisions of an energy-managementmodule As expected such actions impact the energy stateof the node such as the remaining energy available for thenode Therefore a proper design goal is to have an energy-management module that receives feedback related to theoperation of the node and also energy-related data Note thatthe dashed lines used in the figure are an indication thatsuch feedbacks are optional Specifically it is possible thatthe energy-management module makes inferences about theenergy state of a nodewithout receiving explicit feedback datafrom the node Next we will see how energy-managementcontrol efforts can be realized at node network and centralsystem levels

Consider a scenario where part of the energy-manage-ment processing is performed inside the node as illustratedin Figure 4(a) which is related to the activationdeactivationof an energy harvester Such control can be realized purely inthe node level without requiring any network activity In thiscase the energy state is related to themeasurements of voltage

levels and output currents of the power sourceThemaximumpower point tracking (MPPT) technique is one example ofsuch kind of energy control which allows the energy harvesterto achieve its maximum efficiency

The energy-management can also be realized by meansof energy-aware networking protocols as illustrated inFigure 4(b) Consider an example of a collaborative protocolthat dynamically assigns certain roles (eg cluster headrouter data aggregationfusion master node etc) to the bestqualified nodes according to their remaining energy [13]Observe in Figure 4(b) that two feedback data flows occurone related to the application data and basic network func-tionalities and the other associated specifically with energymetrics In some cases the final decision related to thetemporary role of a node at the network can be done at thenode itself without further network activity In other casesthe decision is performed by a specialized node that has amore holistic view of the network

In many cases the energy decisions performed at thenode and network levels can be insufficient for the achieve-ment of the expectedQoSmetrics At the other extreme thereare cases where the quantity of the sensing data is far beyondthe necessary level of information (eg too high samplingrate) and there is room for energy optimization For instanceconsider the case when it is enough from the viewpoint ofthe main application that only one-third of the nodes in anetwork simultaneously monitor a specific event Typicallyonly the data server can take such decision because it is closelyassociated to the historical data flow from multiple sensornodes and alsowith the data qualitymetrics stored exclusivelyat the data server Accordingly as shown in Figure 4(c)the main application at the central Data Server can issuecommands associated to the scheduled activity for each nodeat the network [2 14] Observe that the feedback line inFigure 4(c) is omitted which seems to be counter-intuitiveconsidering the fact that the current discussion is about thecontrol loop concept However in many implementations ofenergy-management systems the energy state of the nodescan be simply inferred based on the network activity of thenodes and explicit control feedback is not necessary

Considering that this work is the presentation of a frame-work not all possible forms of energy-management imple-mentations are represented by the previous illustrationsNonetheless it is secure to state based on the investigationof related work that the majority of the very energy-efficientsystems are actually a combination of efforts at the nodenetwork and central levels In general a complex design andhigher implementation costs are expectedOn the other handsuch approach is typically accompanied with the advantagesof having a balanced energy solution as highlighted by theenergy effort tripod concept In the next section the twodiscussed concepts will be translated in practical designguidelines for WSNs

3 Energy-Adaptive Framework forWSN Architectures

In this section the generic discussion in Section 2 evolves inmore practical and flexible guidelines that will compose the


Command

Measurement values

Energy harvesteronoff

MPPT control

Voltage and currentmeasurements

lowast Charges the energy reservoirlowast Powers the load

(a)

Command

App + network dataEnergy-related data

Discharges theenergy reservoir

Network protocol(cluster head)

Network protocol(node)

Radio moduleonoff

Remaining energyat the node

(b)

Command

Application data


Radio moduleonoff

Main application(data server)


(c)

Figure 4 Examples of energy-management efforts (a) At the node level optimized control of an energy harvester (b) at the network levelthe remaining energy of a node is used as criteria for its selection in network activities and (c) at a centralized level based on the sensingdata received from the nodes the data server defines what nodes will sense according to locationtime scheduling

proposed energy-adaptive framework involving WSN nodeswith constrained energy scavenging profiles It is importantto highlight that this framework is not being proposednecessarily to substitute existing ones but to optimize suchefforts in order to achieve very longmaintenance-free periodsof time for the nodes in conjunction with high levels offunctionality and reliability This section which is almostone-third of this work starts with a presentation of the threefoundations of the framework Next a discussion about thecharacteristics of the energy-efficient low duty-cycle (LDC)operational mode are provided in conjunction with potentialtarget platforms for the framework Finally high-level imple-mentation guidelines are provided

31 Foundation 1 The Strategic Use of Primary Cells Primarycells have high energy densities typically 3 times in com-parison with rechargeable batteries [9] That is for the samephysical volume primary cells provide higher energy capac-ity Another advantage of a primary cell is the possibility ofusing almost 90 of its nominal energy capacity by means ofproper techniques [9] in contrast with rechargeable batteriesas already discussed in Section 21 Moreover primary cellsare very resilient to extreme temperature outdoors Finallythe typical shelf-aging of primary cells is around 10 yearscompared to 3 years of secondary cells Nonetheless for thebest of our knowledge this is the first time that primary cellsare highlighted as an important basis for a WSN energy-management system that involves energy harvesters in par-ticular if the goal is to achieve a maintenance-free period ofmore than 5 years Also the inclusion of such cells increasesthe costs and the size of the WSN node Therefore it is veryimportant to understand the context and assumptions behindthis proposal and such analysis is divided into four topics

Optional Use of Primary Cells The use of primary cells isclosely related to the goal of long life solution while achievinghigh levels of reliability and functionality for the solutionHowever there are cases that such provision is not neces-sary and a batteryless solution fully satisfies the applica-tion requirements For instance consider sensor nodes thatharvest energy from mechanical vibrations at the enginesinstalled in an industrial plant Potentially the energy budgetof these nodes can be sufficient for a realistic zero-energy bat-teryless and WSN implementation based solely on the men-tioned harvester and supercapacitors In this case assumingthat the energy harvesters also have a long lifetime there isno need to implement almost the entire framework proposedin this work because in this case the nodes do not have acritical constrained energy profile Similarly the relative smallcosts of regular maintenance of nodes installed at indoorscan justify the avoidance of energy-managementmechanismsin particular for small networks Therefore the proposeduse of primary cells is clearly optional and depends on thecharacteristics of eachWSN application and its environment

Primary Cells Have Low-Power Density In practice the goalof using 90 or more of the energy capacity of primarycells is seldom achieved in WSNs Based on a careful studyregarding this topic [9] potentially only 10 to 30 of thenominal energy capacity of a primary cell can be realisticallyused in typical WSN nodes The main reason behind thisfact is identified in the same work these cells cannot sustaintheir nominal energy capacity if frequent peak currents (eggt15mA) occur In fact typicalWSNnodes have peak currentshigher than gt100mA although the nominal transmissionmode current is much smaller than this value Thereforethe already mentioned 3-time energy capacity advantage of


primary cells in comparison with secondary cells (assumingthe same volume) is canceled by the peak current effectTo illustrate the point assume another solution based onrechargeable batteries and the potential utilization of only50 of their nominal energy In this case this energyreservoir can still have two times the effective energy capacitycompared to the mentioned primary cell Considering thisanalysis it is not a surprise to read frequent reports aboutthe need of exchanging primary cells in WSNs in regularperiods of few months or even weeks [8] Therefore it is acommon sense in the WSN community that primary cellsmust be avoided On the other hand in this work weadvocate a balanced hybrid power system where primarycells have their strategic role defined However as expectedsuch recommendation of using primary cells in the proposedframework only holds if the peak current effect can bemitigated as proposed in [9]

Achieving a Reliable Energy Harvesting SystemThe use of pri-mary cells is introduced in the framework as a way to guaran-tee certain levels of QoS considering the occurrence of situ-ations where the energy harvesting resources are not enoughto sustain the functionality of the node part of the networkor even the overall network Ideally an energy-managementsystem would not have a battery or any other energy-relatedcomponent that requires regular maintenance But oncebatteries are used the remaining energy at these reservoirsmust be controlled Without any energy control the primarycells also become an uncertainty factor at the system whichmitigates its importance to achieve a higher reliability levelThe bottom line is that primary cells are only being proposedin this context if it is possible tomeasurepredict their currentcapacity (at node level) Moreover primary cells are high-lighted due to their high energy capacity and relative lowcost However other energy reservoirs can also be adoptedas backup modules For instance in [15] low self-dischargerechargeable batteries are used for self-powered watergasmetering nodes Also in [16] the recent fuel cell technologyis proposed with the backup role in a hybrid energy system ofa WSN node

Case Study During a period ofmore than 2 years we adopteda standard WSN solution based on ZigBee technology solarpanels and rechargeable batteries in two outdoor sites withextreme high and low temperatures [2] The reliability of thissolution was clearly impacted due to the performance of theenergy subsystem Later we designed a WSN node calledRipple-2Awith significant enhancements in terms of networkperformance However this node maintained the traditionaldesign for outdoorsWSNnodes (solar panel and rechargeablebatteries) but it was implemented by means of differenthardware components Although Ripple-2A has achieved thedesign goals in terms of efficiency with an overall networkoverhead smaller than 1 the weakest part of the solutionin terms of reliability and long-term lifetime was again theenergy subsystem As already discussed in Section 21 themain issues were related to relative small lifetime for therechargeable batteries (around 1 year) their performance

under extreme temperatures and the fragility of small solarpanels available at the market

The next step was the nontraditional design of a WSNnode which could be powered by nonrechargeable batteriesSeveral months of research were dedicated for the realiza-tion of long-term outdoor experiments that could validatethis approach The new node design is called Ripple-2Dleaving room for two additional kinds of nodes Ripple-2B(solar panel + supercapacitors) and Ripple-2C (solar panel +supercapacitors + non-rechargeable batteries) The latter oneis currently under development and it follows many of theguidelines proposed in this work The Ripple-2D solved thepulse current effect issue by slow-charging supercapacitorswhich in turn power the radio module In short the currentdrained from the battery never goes beyond 15mA and thepulse effect is voided As expected the delay introduced bythis charging step can impact the functionality of existing net-work protocols Therefore we also designed a new networksolution to address this challenge Many of the ideas behindthis design are part of the framework proposed here Basedon the results from simulated and empirical (accelerated andvery-long term experiments) results the effective capacity ofthe battery is found to be higher than 90 in relation to theirnominal value [9] The designed lifetime of the Ripple-2Dunder this project is almost 2 years (conservative value) Cur-rently the majority of the nodes are close to reach the targetlifetime and they are in continuous and reliable operation[4] proving that this solution is indeed significantly superiorcompared to original Ripple-1 and Ripple-2A architecturesbased on rechargeable batteries

These results provided the foundation for the realisticachievement of a very long lifetime for the nodes in con-junction with a high level of reliability Moreover the overallnetwork solution also provides a very deterministic way toforecast the lifetime of each individual node [4] as will bediscussed in detail in Section 4 At the data serverrsquos side it isalso possible to dynamically extend the lifetime of the nodesbymeans of spatiotemporal activation of a subset of the nodes[2] In this way the initial life expectancy of 2 years can stillbe significantly extended At node and network sides recentadvances in compressive sensing (CS) can also be added tothe solution in order to extend the lifetime of the networkwithout significant sacrifice in terms of data quality as will bediscussed in Section 6 In short this case study demonstratesthat the energy-management efforts can be realized at nodenetwork and data server levels The next generation ofWSN node for this project Ripple-2C (currently underdevelopment) combines a solar panel with supercapacitorsand non-rechargeable batteries exactly as recommended inthis section for energy-constrained scenarios The ultimategoal is to extend the lifetime for realistic values beyond 5

years in order to properly support the main project behindthis effort [2 3]

32 Foundation 2Distributed System inside a SensorNode Atthe proposed framework a node can switch between its regu-lar operation and a very low duty-cycle mode From now onwe will call a WSN node that supports such dual duty-cycle(DDC) mode of operation a DDC node The implications


Dual duty-cycle (DDC) node

Wake-up on radio(WOR)

BeaconTX

Analog sensorRadio transceiver

Digital sensorMain MCU

Real-time clock(RTC) Intelligent sensor

Energy-managementsubsystem

Power scavenging

Energy reservoir (s)

source (s)

Data linePower lineCustomizable MCU

Potential noncustom MCUWake-up interruptPower gating

Energy-management

Energy harvester

controller

Figure 5The dual duty-cycle (DDC) node hasmultiplemicrocontrollers (MCUs) and intelligent devicesThat is it encompasses a distributedsystem inside itselfThemain goal is to achieve energy savings at unprecedented levels compared to traditional nodesThe existence ofmultiplepower lines (lines without arrows in the figure) rather than a single power line is associated with the use of the power gating technique [9]and different voltage conditioning schemes for the internal modules

associated with the dual mode of operation at the design ofa node are definitely not trivial in particular when energyharvesters are employed Significant hardware and softwareadditions at the DDC node are required and the reasonswhy such changes are necessary are better understood whenthe overall framework is explained At this section a briefpresentation of the modifications in a DDC node is listedThe fundamental change is related to the evolution from anarchitecture model centralized in a single microcontroller(MCU) into a solution with multiple MCUs each one witha distinct role In other worlds a DDC node is essentially adistributed system inside a node as shown in Figure 5

The following real-word case is used to illustrate one ofthe justifications for the addition of complexity and costs in aDDC node Consider a typical scenario where a solar panel isbeing used to charge a secondary batteryThe design goal is touse the maximum amount of the energy stored in the batteryTypically a DC-DC converter is necessary considering thepotential variation of the voltage level on the battery while itis being discharged While the node is active the mentionedDC-DC converter can potentially achieve very high levels ofefficiency such as gt95 Therefore its use is clearly justifiedconsidering the mentioned goal When the node enters sleepmode the load current drastically drops as expected reach-ing values on the order of 120583A However because the DC-DC converter must be continuously active the overall powerconsumption of the node is effectively dominated by theconsumption of this converter Unfortunately its power con-sumption can be 2 or 3 orders of magnitude higher than

the sleeping consumption of the MCU and adjacent modulesbecause the DC-DC converter typically has very low effi-ciency when subjected to light loads Therefore in particularfor low duty-cycle regime the design of the node can bemodified in order to void the voltage converter while theMCU is in sleep mode Such simple goal adds significantcomplexity to the design

Because the proposed framework is founded on a mech-anism that drastically changes the operational duty-cycle ofthe DDC node the energy subsystem of the node must alsobe very efficient while operating in sleep or similar modeTherefore an increase of complexity is expected in the designof such node Accordingly the next design step is to attemptto separate the power lines and avoid the use of voltage regu-lators in modules that are constantly powered on Moreoverit is necessary to discover the power needs of the differentmodules of the DDC node For instance the dynamic voltagelevel of themainMCUmay not be the same as the radiomod-ule or as the real-time clock (RTC) Such complex scenariois better understood by means of the Figure 5 For examplenote that the concept of having a single shared power line forall the modules gives place to multiple and power-controlledlines While some aspects of this figure are discussed nextadditional details related to the software side of a DDC nodeare given in Section 4 It is important to remember that therecommended guidelines in this work can be partially fol-lowed Therefore the realization of a full DDC node may notbe the goal of aWSN designer considering the characteristicsof his application and specific energy aspects Nonetheless


some of the concepts underlying the framework and its DDCnode can be borrowed and integrated on an existing WSNplatform as discussed next

Autonomous Energy-Management Controller The control ofthe power-gating process (activationdeactivation of theinternal modules [9]) of the node is performed by this MCUMoreover the activation of voltage conditioner(s) and super-capacitors charger(s) the selection of the main energy reser-voir and the power reset of the main MCU (watchdog-timerfunction) are tasks provided by this module

Autonomous Energy Harvester While operating in very lowduty-cycles a sleeping main MCU can potentially missimportant energy harvesting events Therefore an autono-mous energy harvester module can be an ideal solution pro-vided its active power is very smallThe energy harvester andenergy-management controller roles can be integrated in thesame MCU

Main MCU this module runs the software associated withthe low duty-cycle mode in a DDC node which is mainlyimplemented as an application-layer overlay Typically thelegacy platform that is being ported into the DDC nodeis actually called Radio Transceiver module For instanceconsider a DDC node that uses a TelosB module (TinyOS 2xand IEEE 802154-based communication) In this scenarioit is important to remember that the TelosB is not the MainMCU but the radio transceiver module of the node one ofthe blocks in Figure 5

Radio Transceiver It refers to any WSN node that is beingported at the DDC node In the case of a fully customizedDDC node the radio transceiver can be any radio that pro-vides at least point-to-point communicationAs expected themajority of the existing WSN nodes fall into this category ofradio devices If a primary cell is used in order to avoid thepulse effect it is recommended that the radio module bepowered via supercapacitors [9] The main drawback of thisapproach is the significant data latency associated with thetime necessary to charge the supercapacitors In generalwhen a WSN radio transceiver is directly connected to non-rechargeable batteries the lifetime of these cells is stronglyreduced

Real-Time Clock (RTC) In low duty-cycle mode the RTCis used to wake-up the main MCU according to a certainscheduling The power consumption of the RTCmust be sig-nificantly smaller compared to the sleeping power consump-tion of the main MCU Typically the RTC device is poweredby a non-rechargeable battery in a DDC node

Wake-Up On Radio (WOR) TheWOR is an optional modulethat allows a DDC to quickly swap between low duty-cyclemode to regularmode In order to be adopted in aDDCnodethe WOR module must consume a very tiny power (eglt5 120583W) In [17] a nanopower WOR is reported making it apotential candidate for WOR in DDC nodes

Analog Sensor Typically this low-cost kind of sensor requiresa stabilized power supply Also as a passive device it is not

capable to wake up the main in the event of changes at theenvironment where it is deployed

Digital Sensor typically this kind of sensor has its own non-customizable MCU an internal voltage regulator and a serialinterface such as I2C or SPI for communication with theMCU In general it is not capable to wake up the main MCUin case of detection of events

Intelligent Sensor This is the newest generation of sensorsthat has the capability to wake-up the MCU based on theanalysis of an external event Although one can customize itsown MCU to achieve the above goal recent available tech-nology goes one step ahead and provides a way to performcontinuous monitoring of an event at the cost of few 120583W Asexpected this scheme allows the significant reduction of theduty-cycle of the main MCU in event-driven applicationsVery recently the manufacturer Atmel announced the Sleep-Walking technology [18] which is basically the integration ofthe Intelligent Sensor with the main MCU in this contextAlthough not shown in Figure 5 when aWSN node is portedin aDDCnode the threementioned kinds of sensors (analogdigital and intelligent) can be attached at the main MCU atthe radio transceiver (which is the regular WSN node) or atboth These options are associated with the function of thesesensors at the framework and also at the main application

Multiple Power Lines With this provision each individualmodule can be powered onoff independently bymeans of thepower-gating technique [9] Moreover only the devices thatrequire voltage conditioning are connected to these specialpower lines In fact there is an underlying effort to avoidif possible the use of voltage regulators [9] in any modulethat is constantly powered such as the main MCU and theRTC Nonetheless the multiple power lines effort must becompared with the simpler solution of putting a module instandby mode if this function is available The baseline forpower consumptionmust be the sleeping power consumptionof the main MCU Moreover it is important to rememberthat power gates also present a leakage current and not allelectronic switches can be used for this role As a referencefor comparisons the typical power leakage (loss) due to thepower-gating is smaller than 03120583W

Recent technological advances point into the direction ofpico- and nano-MCUs embedded in a significant number ofelectronic devices many of them considered analog devicesfor decades This is the case for power scavenging sourcessensors battery chargers and even bulb lampsTherefore thesecond foundation of the proposed framework the adoptionof a distributed system inside a node is actually well alignedwith the industry trends Nonetheless based on the energyeffort tripod concept discussed in Section 22 besides thehardware also the network and the application characteristicsmust be considered in order to achieve a truly functionalreliable and long-term maintenance-free WSN solution Sofar this work gave a significant emphasis on the energysubsystem and hardware aspects From now on the focuswill be mainly oriented toward energy-management softwaremechanisms Such software modules assume the existence of


Regular or high duty-cycle (RDC) mode

Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH

Low duty-cycle (LDC) modeStandard WSN protocols activeBETS protocol inactive

Network segment ANetwork segment B

To data server

Standard WSN protocols inactiveBETS protocol active

Figure 6 Dual duty-cycle (DDC) operation the network switches between LDC and RDCmodes In RDCmode the network maintains itsoriginal characteristics In LDCmode the BETS protocol becomes active a planned network segmentation occurs and the network achievesits maximum energy efficiency

the hardware for DDC node with the features presented inthis section Important questions associated to the need ofhaving dual duty-cycle modes and how a DDC node actuallyoperates will be answered in the next section

33 Foundation 3DualDuty-Cycle (DDC)Operation A trulyenergy-balanced and efficient WSN solution subjected toconstrainedirregular energy resources depends on coordi-nated optimization efforts in terms of hardware networkalgorithms and flexible application demands Such coordi-nation is the main focus of this section and a discussionon how a DDC system can achieve the mentioned goals isprovided A WSN node that follows the guidelines providedin this section the so-called DDC node can be designedentirely from scratch or alternatively can be the result ofthe integration of an existing WSN with additional hardwareand software modules Details of how to implement a DDCnode by porting an existingWSN platform are provided laterin Section 35 Similarly to the hardware guidelines so farprovided one can decide to implement some of the underly-ing concepts in this section in his WSN design without fullyimplementing a DDC operational system

Intuitively the expression dual duty-cycle operation trans-mits the idea of having a system that operates in low andhigh (or regular) duty-cycle regimes Therefore a naturalquestion that raises is why not designing toward uniquely alow duty-cycle operation such as lt1 if it is more energy-efficientThe answer is related to two of the legs of the energy-effort tripod concept in Figure 2 the network and applicationcharacteristics must be also be considered for a balancedsolution Starting with the application constraints in someWSN systems a high network throughput is required suchas in a surveillance system involving video and audio trafficAs expected there are periods of time when the nodes aresubjected to a very high duty-cycle operation However it isalso possible that not all nodes are involved simultaneouslyin a high data traffic Moreover such intensive usage of thenetwork typically occurs in bursts such as when an eventof interest is detected Therefore many WSN applicationsalready present some form of dual mode of operation How-ever not allWSN solutions optimally exploit this fact in orderto achieve maximum energy savings

The above example involving a surveillance system is oneof the target scenarios for the proposed framework and theidea is relatively simple while in regular mode the existingWSN solution is fully used as it is that is without significantchanges due to the proposed framework Therefore it isexpected maximum network performance (as provided bythe current WSN solution) with potential sacrifice on theenergy performance However when the application doesnot require such high network throughput the DDC nodescan be commanded to switch from regular duty-cycle (RDC)mode to low duty-cycle (LDC) mode The overall processis illustrated in Figure 6 where three additional aspects ofthe LDC mode are also shown the 2-tier architecture thenetwork segmentation and the protocol called BETS Theseaspects will be discussed later in the next sections

It is important to highlight that nodes in LDC modeare not necessarily inactive or continuously sleeping Insteadsuch nodes are actually following a predefined low duty-cyclemaintenance scheduling A node in LDCmode does not needto wake up multiple times per second as is the usual casein traditional WSN protocols In LDC mode the nodes onlyneed to regularly wake up after long and continuous sleepingperiods (eg 15 minutes) for sending measurements or forjust updating their status in a central data server Howeverwhile in the middle of its deep sleeping period (hibernation)a node in LDC mode can be forced to return to its regularor RDC mode of operation For instance in an event-drivenapplication if an event of interest is detected by one ofmore nodes the network must quickly resume its maximumperformance operation Note that the Intelligent Sensorpreviously discussed is an important component to turn thisscenario feasible Another example is related to aWSN appli-cation for commercial buildings where the nodes employindoors photovoltaic (PV) panels In this case the networkswaps from LDC to RDCmode as soon one person starts hisactivities in his office or at the building When the humanpresence is no more detected (eg during the night) thenetwork returns to its LDC mode in order to save energy Inthis way many of the functionalities of this WSN applicationare still provided during nights weekends and holidays

For some datacollection monitoring applications theRDCLDC switching may not be necessary and the node can


permanently stay in LDCmode as in the previous mentionedcase study [2] This category of WSN applications does notrequire significant amount of data traffic and is delay-tolerant(DTN) [19 20] For instance a soil moisture monitoringsystem and many other environmental monitoring applica-tions typically have an application duty-cycle smaller than1 and only require measurements every 15 minutes ormore Therefore DDC nodes in permanent LDC mode canpotentially satisfy the needs of the application Nonethelessone can argue that a traditionalWSN solution can also satisfythis application and still be an answer for the mentionedsurveillance application Considering this argument doubtscan raise in relation to the real need of adding complexity andcosts to aWSN by implementing DDCnodesThe answer liesat the extreme (and realistic) energy savings achieved whena node operates in LDC mode In fact the simulated andempirical results in Section 5 showing the network lifetimeextension by multiple folds are related to the LDC modeTherefore the bigger is the period of time that a node staysin LDC mode and the higher is the energy efficiency of thesolution While the network is operating in RDC mode itsenergy performance is essentially the same as the originalWSN platform The term essentially is used here because theenergy overhead due to the additional hardware necessary forimplementing the DDC node is assumed to be very tiny andthis specific aspect will be discussed later in Section 35

In short the dual duty-cycle switchingmechanism is pro-vided to obtain the best of both worlds (energy and networkperformances)

(i) In regular (RDC) operation theWSNnode has essen-tially the same performance as it would have if theframework was not applied but with the cost ofpotential not optimum energy efficiency networkperformance uArr energy performance dArr

(ii) In low duty-cycle (LDC) operation the WSN is dras-tically reformulated in a process similar to virtuallyremoving all existing nodes and deploying a newnetwork but maintaining the same physical layer(radio transceiver) of the nodes The LDC mode hasexcellent energy performance achieved with somelevel of network performance sacrifice network per-formance dArr energy performance uArr

34 Characteristics of the LDC Mode In this section it isexplained why the energy savingmechanisms of a DDC nodein LDC mode hardly can be achieved by simply using off-the-shelf WSN nodes Nonetheless the achieved very highenergy efficiency has its price in terms of significant penaltieson data latency and throughput Although such drawbacksare actually expected for a node operating in LDC node (oth-erwise it would be operating in RDC mode) there are otherconstraints that can impact the adoption of a DDC systemfor all WSN scenarios Accordingly a discussion on thelimitations of the DDC system is provided such as the lack ofsupport for node mobility Also characteristics of potentialtarget applications that permanently operate in LDC mode(LDC-only applications) are also provided

341 Motivation To better understand the energy savingsassociated with the LDCmode two discussions are providedone that introduces the preliminary ideas and other that givesadditional quantitative intuition It is important to highlightthat the following discussion is crucial for a comprehensiveunderstanding of the goals behind this work

In a typical data collection application the nodes reg-ularly wake up sense their environment and transmit thesensing data to a central point In many cases the amount ofthe data is relatively small and the sampling rate is also prettysmall resulting in a very low duty-cycle operation Howeverthis conclusion is exclusively based on the viewpoint of theapplication Unfortunately the overhead of the networkingprotocols can be sufficiently high and dominate the energyconsumption of the nodes On the other hand if the applica-tionrsquos duty cycle is relatively high the overhead of the networkis typically negligible Therefore in order to effectively com-pare two solutions such as two different network protocolsit is necessary to understand the effective network overheadcaused by each one of the evaluated protocols Besides thetypical overhead added by the networking protocol in theform over additional bits or bytes at themessage payload it isalso necessary to understand the impact of that solution in thetimeline Specifically it is well known that even without anyapplication activity manyWSNs sustain some sort of networkinfrastructure traffic to maintain the nodes synchronized todetect the state of the nodes and so onTherefore even undera very low application duty-cycle the effective duty-cycle canstill be relatively high Naturally because the radio is typicallythe most power-hungry module in a WSN node the termeffective duty-cycle is related to the use of the radio transceivermodule in Figure 5 The important question to be answeredat this point of the discussion is related to the magnitude ofsuch overhead

In order to review this problemunder a quantitative view-point let us consider an antimold WSN-based solution for acommercial building where the sensor nodes are strategicallyinstalled inside thewalls Due to the fact that the nodeswill bedeployed in areas of difficult access and in many cases with-out energy scavenging opportunities non-rechargeable bat-teries are required Due to economical reasons the exchangeof such batteries must only occur after 5 years Sensingmeasurements must occur every 20min but cycles of up to60min are still acceptable The power profile of the nodestypical WSN nodes is shown in Table 2 The sensor nodecomprises a processor (MCU) a sensing module and a radiotransceiver It periodically wakes up performs some pro-cessing (1 s) takes measurements (5 s) sendsreceives datatofrom the base station (3 s) performsmore processing (1 s)and finally sleeps again Assume that the communicationperformance of the nodes and their reliability are very highand a fixed topology for the nodes is defined In this scenariodifferent stack ofWSNprotocols are tested and any additionalmeasured network overhead is assumed to be exclusively dueto the characteristics of that protocols stack Also differentsampling rates are used Some protocols require that thenodes be active for more time in comparison with othersMoreover some protocols are associated with a significantnumber of redundant data paths between two nodes All


10 5 2 10750

1

2

3

4

5

6

7

8

Network overhead added to the overall duty-cycle ()

Meas every 32min (55 app duty-cycle)Meas every 5min (33 app duty-cycle)Meas every 10min (17 app duty-cycle)Meas every 20min (08 app duty-cycle)Meas every 60min (03 app duty-cycle)

36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)

Figure 7 Lifetime of aWSNnode (see Table 2) for different networkoverheads and application duty-cycles

these differences impact the overhead of these protocolsAs a result after careful measurements it is found that theeffective duty-cycle of sensor node increases by 075 12 5 and 10according to the choice of the protocol stackAssuming that the initial energy of a node is 245KJ the goalis to run simulations that relate the lifetime of the node withthe choice of the protocol stack and application duty-cycleIt is assumed that no communication errors occur and that100 of the initial stored energy is effectively used by thenode (ideal hardware) because we want to focus only on thenetwork overhead effects on an ideal scenario

The results of these simulations are shown in Figure 7 Asexpected when the network overhead increases the lifetimeof a node decreases However such impact is particularlystronger for low duty-cycles applications For instance ifthe network overhead increases from 075 to 5 the lifeexpectancy of a node decreases around 75and 50 formea-surement cycles of 60 and 32min respectively But caution isrequired in this analysis although the difference between 75and 50 does not seem to be very drastic it is not actuallythe case These percentages are relative to different lifetimegoals and when one translates these values into years it isfound that the lifetime of the nodes decreases by around 6

years when the network overhead increases from 075 to 5and 60min schedule is used For the same increase in termsof network overhead the lifetime of the nodes decreases by 1year if the 32min schedule is used The conclusion is cleara relative slight decrease at the performance of the stackof networking protocol dramatically impacts the lifetime ofnode for low duty-cycles applications

Therefore in our ongoing example if we opt for 60minschedule in order to save energy in terms of active power

the choice of the lighter stack of protocols is the only way tomake the solution feasible Moreover if we want to improvethe data quality of the solution by increasing the samplingrate to cycles of 20min the only way to achieve the requiredlifetime gt5 years is to have a network overhead not higherthan 1 Unfortunately the typical network overhead inWSNs is much higher than this value For instance these aresome reported radio duty-cycles of MAC protocols [21] LPL10 T-MAC 25 S-MAC 23 and B-MAC 114 Notethat the overhead due to other protocols such as networkand transport are not included in these values and we stillassumed an error-free network Therefore the final answerin relation to the illustrative project is that it is potentiallyunfeasible if we simply adopt traditionalWSN technology Inthe context of our framework the goal is the development ofsolutions for this and many others scenarios while in LDCmode the DDC node must experience a very small effectivenetwork overhead (eg lt1)

As a rule-of-thumb when a WSN protocol provides ahigher level of flexibility and functionality it is also expecteda higher network overhead due to this protocol Thereforein the formulation of the operation of a node in LDC modeit is important to consider specific application needs andavoid unnecessary network features Note that in the previousexample an ad hoc deployment mobility multi-hopping andeven collaboration are not required features Considering thatthe topology is static it is possible to organize the networkinto star-based segments and at the central point of eachsegment a special node (ie cluster heads) can collect thedata from the sensors and transfer data tofrom a base stationor data serverWith such ideas inmind a 2-tier asynchronousnetwork comes naturally as a feasible architecture as shownin Figure 8 Observe that in terms of topology this is theoriginal and still dominant way to organize computers incorporative networks Moreover the majority of the IEEE802154ZigBee networks also follow such scheme some-times augmented with low-height trees topologies [6 7 22]In short in order to achieve overall network overheadssmaller than 1 (in terms of duty-cycle) the DDC systemoperating in LDC mode must adopt a very simple andefficient network topology and protocol(s) How this goal isachieved and also the drawbacks of this approach are consid-ered next

342 Network Topologies Constraints The DDC systemoperating in LDC mode is based on a cross-layer proto-col called best-effort time-slot Allocation (BETS) protocolproposed in [4] It is implemented as an application-leveloverlay because this is the simplest way to have a DDC nodeswitching between RDC and LDC modes without changingthe software layers implemented at the WSN platform thathass been ported However if one designs a DDC node fromscratch the main functionalities of the BETS protocol canbe potentially implemented at the data link layer The detailsof BETS are discussed in Section 4 For now it is enough tounderstand its general operation and requirements

The BETS design is guided by simplicity behind the con-cept called selfish node presented in [4] A regular sensor nodeacts selfishly in the sense that no message relaying is actually


DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz

DS data serverBS base stationCH cluster head (also a sensor node)ED end device (regular sensor node)BETS protocol ED-CH communication

middot middot middot middot middot middot middot middot middot middot middot middot middot middot middot middot middot middot

Figure 8 Different topologies for the nodes in LDC mode [4]

performed by the node As the name implies this idea comesin the opposite direction in relation to ad hoc networksand many current WSNs with emphasis on cooperation Thesensor node which is called end device (ED) simply wakesup takes measurements and sends the data to a specific col-lecting or cluster head (CH) node After successfully sendingits data to CH the ED receives an acknowledgment from CHwith the precise time for the next cycle and goes to sleep Inorder to avoid the need ofmessage relaying among ED nodesthe network is segmented and each segment has a star-liketopology with a predefined maximum number of EDs TheCH can communicate with all EDs in that segment and itsrole is similar to the typical sink in the WSN literature Themethod of the communication between CH and a centralpoint is a completely open aspect in this framework It can beimplemented by means of any wireless technology It is alsopossible to create a network ofCHs involving long-range linksamong them and selecting one of them as a base station (BS)whichwould be in charge of sending the data packets to a dataserver (DS) Such very flexible architecture is possible becauseBETS is an asynchronous protocol in relation to the messagedelivery between ED and DS Once the EDrsquos data reaches theCH node the transaction is concluded from the viewpoint ofthe ED nodeWhen and how CH compacts the EDmessagesif this is the case to send toDS is also an open implementationaspect

The network architecture supporting BETS (called Rip-ple-2 in [4]) is also hybrid any wireless communication tech-nology can be used provided that a simple point-to-point linkcan be implemented For instance instead of using an off-the-shelf WSN node as the radio transceiver for the DDC node(see Figure 5) one can adopt 900MHz point-to-point radio

modules without any networking layer besides the physicallayer Moreover distinct network segments can adopt differ-ent wireless link technologies Similarly the technology usedfor the ED-CH link can be different from the CH-BS CH-DS or BS-DS links Therefore the BETS protocol is a properchoice for the LDC node it is very flexible open to integra-tion and intuitively it seems to have a very tiny overhead dueto its simplicity Some of the possible topologies for the nodesin LDCmode are shown in Figure 8 Note that the maximumheight of the network tree is 2 which highlights the fact thatwith time many enhancements can be easily added to thesolution Based on the analysis of this figure some limitationsor constraints of the proposed solution can also be identified

Constraint 1 No Native Support for Mobility Node mobilityis not supported by this solution because a fixed and well-planed topology is assumed a priori for each network seg-ment When the ED node wakes up it simply takes mea-surements and transmits the data There is no provision fornetwork setup phase or any kind of search for the location ofthe CH node the ED node simply sends the data and expectsthat CH receives and acknowledges that message Thereforea mobile node could not fully implement the selfish nodebehavior which was previously described Therefore in anative DDC system with LDC and LDC modes the mobilenodes can only be part of the portion of the network that isoperating in RDC (regular) mode

Constraint 2 Not All Network Topologies Are SupportedSome physical network topologies can impede the DDC tooperate in LDC mode For instance consider a line-fashiondeployment such as in a sequence of sensor nodes in a bridge


In this case it is very hard to implement logical star topologiesunless the communication range of a node encompasses asignificant number of nodes at both directions of the linewhich is usually not the case Consequently another protocolmust replace BETS in order to include some level of lightcollaboration among EDs in LDC mode

Constraint 3 CH-DSBS Links Can Pose Challenges TypicallyallWSNnodes in a network use the samewireless technologysuch as IEEE 802154 PHY (eg TelosB-based WSN) Whenthe network (or portion of it) switches from RDC to LDCmode the new formed network obeys a predefined schemewith one or more star-based segments The assigned CHnodes now need to send the data to DS by means of the samelow-power short-range wireless links because their radiotransceiver is still the same Therefore the collaborationamong CHs is potentially necessary in this scenarioThe pro-posed framework does not offer guidelines for the CHBS orCHDS communication Nonetheless one potential solutionin this case is the adoption of two radiomodules at CHnodesIn many scenarios the second radio (CH-BSDS link) can bea Wi-Fi adapter if the place also has a WLAN infrastructureAgain in terms of energy efficiency it is necessary to investi-gate the energy costs of this approach compared to traditionalWSN solutions

343 Target Applications for LDC-Only Mode The energyefficiency of the LDC mode is achieved with the cost ofnetwork performance penalties in addition to some topologyconstraints In fact we must have in mind that the full adop-tion of the proposed framework is realistically not possible inmany cases On the other hand there are scenarios where theRDCLDC switching is not even necessary and the networkis always in LDC mode This is the case when the WSNapplication is a low duty-cycle data-collection it is delay-tolerant and it does not have to support mobile nodes

In summary three cases in relation to the applicationof the DDC system proposed by this framework can occur(a) for many regular WSN applications the DDC systemmust operate in dual mode (RDCLDC) (b) for many lowduty-cycle data-collection applications the DDC system onlyneeds to operate in LDCmode and (c) for some applicationsthe DDC system cannot be employed but some guidelines ofthis framework can still be adopted to enhance the energyperformance of the solution

Because the LDC operational mode is closely associatedwith very small network overhead the goal of maintainingthe DDC system uniquely in LDC mode is actually thebest option in terms of energy-management optimizationHowever just because the main application is low duty-cycleit does not imply that dual mode operation can be removedThere are some requirements that prevent the adoption ofa LDC-only mode and the RDCLDC switching must bepreserved One requirement is related to the characteristic ofthe application in a LDC-only network it must follow sched-ules such as of a data-collection one Also the applicationmust be delay-tolerant and the network cannot have issuesin order to follow a 2-tier and non-collaborative architecture

A summary of the characteristics of LDC-only applications islisted in Table 3 A detailed discussion is provided next

Data latency in LDC-only mode can be a problem forsome applications The complete message data latency of anED999450999456DSmessage can be on the order of seconds or evenmoremainly due to the asynchronous behavior of the network(LDCmode)The latency for the other way (DS999450999456ED) is evenworse and can be in the order of minutes Fortunately typi-cally for a data-collection application the ED999450999456DS messagedirection is the one of interest Even in this case the reasonsfor a significant data delay are as follows

(i) Once a measurement-cycle starts the ED node mustwait for its time to transmit the data

(ii) Once the data from this ED node arrives at CHit is necessary to wait until CH concludes the datacollection from all nodes at that cycle in order to havethe transmission to DS

(iii) At the end of each cycle CH must activate its CHBSor CHDS link to perform the data transfer Such acti-vation delay can be significant For instance to acti-vate a SMS based (text) the SMS modem can takealmost 1min to conclude the link activation andmes-sage transmission

(iv) If multiple CHs send data to a shared device (BS) thedelay due to the coordination among CHs and alsodue to the activation of the BSDS link can also besignificant

A second aspect to be evaluated before deciding by theLDC-only network operation is related to the energy require-ments for the CH nodes The energy consumption of CHfollows at the best theoretical case a linear relation with thenumber of EDs in that network segment For instance if a CHis in charge of 20 nodes its lifetime is expected to be around120 the lifetime of a regular (ED) node assuming that allnodes start with the same initial energy capacity Thereforethe lifetime of CHs can be strongly impacted if the WSNdesign does not carefully consider the energy challenges inCHnodesThe increase of the number of segments (thus alsothe number of CH nodes) can be a technique that alleviatesthe energyworkload on each assignedCHnodeThedynamicsharing of the CH role among nodes in the segment such asin a round-robin fashion can also be a solution Finally theadoption of a hybrid power source system for CHs based onenergy scavenging and non-rechargeable batteries can extendthe lifetime of CHs The overall characteristics of a DDCsystem operating in LDC mode are summarized in Figure 9Note that two of the highlighted challenges in this designapproach are the need of some form of power hibernationfor both ED and CH nodes and also an efficient time syn-chronization technique for the EDCH communication Bothaspects are considered in our previous works the former oneinvolving hardware techniques is considered in detail in [9]and the latter one is presented in a short-format in [4] Thedetailed discussion of the BETS protocol the core componentof a DDC system is discussed in detail in Section 4 whereits algorithms are presented The performance evaluation ofBETS is provided in Section 5


Design goal

Scalability

Energy efficiency(regular node)

Energy efficiency (cluster head)

Design approach

lowast Network segmentationlowast 2-tier asynchronous network lowast Higher data latency

lowast Need of cluster heads (CHs)

lowast Selfish node conceptlowast Hibernation mode (deep sleep)

lowast Hibernation mode (deep sleep)

Trade-offschallenges

lowast Static topology

lowast Sense-send-sleep application

lowast Need of smart time synchronization lowast Higher data latency

lowast Need of time synchronizationlowast Very low application duty-cycle

lowast Less frequent communication wbase station

Figure 9The characteristics of the network while in LDCmode the BETS protocol is designed to provide high energy efficiency for both EDand CH nodes

35 DDC System Implementation Guidelines The main goalof this section is to clarify what are the steps necessary forthe partial or full implementation of the proposed energy-management framework Also it is a good place to summa-rize the exposed concepts and acronyms specifically intro-duced in this work as shown in Table 1 In relation to thefull implementation of the framework there are at least 3important scenarios to be analyzed

Case 1 Node in LDC-Only Mode and Using Ordinary RadiosA low duty-cycle data collection application is being usedThis application is delay-tolerant and does not have mobilitysupport In this case the DDC system never switches toRDC mode and this is the simplest scenario to implementThe ordinary radios provide a simple point-to-point com-munication link and in general there is no need of furtherconfigurations for these radio modules

Case 2 Node in LDC-Only Mode and Using Legacy WSNNodes A low duty-cycle data collection application is beingused and the system never switches to RDC mode It is nec-essary to configure the legacy WSN node to provide a simplepoint-to-point communication with the minimum possiblenetworking functionalities Any feature or service beyond thephysical and medium access control (MAC) layers is unnec-essary and even worse can potentially increase the networkoverhead of the solution Therefore such features must bepermanently deactivated For instance a target legacy WSNnode can have hidden dynamic topology control and timesynchronization features that are not necessary for this sce-nario and must be deactivated Good ways to discover if thisis the case are (a) to observe how quickly two WSN modulesexchangemessages just after they are initialized and (b) to usea radio frequency (RF) sniffer device to monitor the contentof the packets For the latter case low-cost monitoring toolsfor the 24GHz ISM band are available such as the device in[23] which is used in our ongoing project If it is not possibleto completely deactivate unnecessary features in legacyWSNnodes maybe the better option for this specific scenario is toexchange the legacy nodes to ordinary radio modules

Case 3 Node inDualMode andUsing LegacyWSNNodesThisis the general case where the WSN application is more strictin relation to network QoS metrics While in LDC mode itis necessary to configure the legacy WSN node to providea simple point-to-point communication with the minimumnetworking functionalities as discussed above Howeverwhen the node returns to RDC mode such networkingfeatures at the legacy WSN nodes must be activated againTherefore the integration effort in this scenario involvesthe addition of an application-layer module at the legacyWSN module to dynamically activatedeactivate high-levelnetworking functionalities of the device If the unnecessaryfeatures in legacy WSN nodes cannot be deactivated whilethe node is operating in LDC mode the optimum energyefficiency provided by the BETS protocol cannot be achieved

A typical example of a device with the role of DDCrsquos radiotransceiver (Figure 5) under Case 1 is XBee-Pro 802154Series 1 (Digi Inc) For Cases 2 and 3 the open-sourceTelosB mote and the commercial XBee-PRO ZB ZigBee(Digi Inc) are good examples considering their popularitydocumentation market availability and ease of integrationNote that all these devices operate at 24GHz band and areIEEE 802154 PHY-compliant which is the general trend forcurrent WSNs

The following are implementation guidelines labeled asGx which can be applied in all 3 mentioned cases unlessexplicitly indicated(G1) To implement a DDC node as illustrated in Figure 5

it is necessary to have the main MCU separated fromthe radio transceiver in all 3 mentioned cases TheBETS protocol implementation software resides at themain MCU

(G2) When an energy scavenger is employed the EnergyManagement Controller is separated from the mainMCU In this case it runs the softwaremodule associ-ated with local energy-management decisions One ofthese decisions is the selection of the energy reservoirused to power the Radio Transceiver or any otherpower-hungry module Moreover the charging of


Table 1 Main terms and acronyms introduced in the context of this work

Term Explanation

Energy effort tripod Fundamental frameworkrsquos concept an energy-balanced WSN solution involves hardwarenetwork and application

Energy control loop Fundamental frameworkrsquos concept the operation of a WSN node must be regulated by its energystate

Selfish node concept Ultralight energy consumption of a node wake up sense send and sleep The expected networkoverhead is lt1

Dual duty-cycle (DDC) 2modes of WSN operation (a) current WSN solution (RDC) and (b) constrained but highlyenergy-efficient mode (LDC)

Regular duty-cycle (RDC) Provides the same network performance as the existing solution with potential energy penalties

Low duty-cycle (LDC) Maintenance mode with excellent energy performance It can also be used for low duty-cycledata-collection applications

DDC switching Framework provision which allows that the applicationrsquos needs (or an event) automatically switchthe DDC mode

DDC system Part of the proposed energy-management framework which involves hardware and softwareadditions to the WSN

DDC node Distributed system (multiple MCUs) inside a single device The main MCU runs the BETSprotocol

Power gating Technique extensively used in a DDC node which activatesdeactivates the hardware modules of anode (not sleep mode)

BETS protocol Best-effort time-slot allocation protocol is the core DDC module that controls the behavior of thenetwork in LDC mode

Energy reservoir In a DDC system the energy reservoirs are typically supercapacitors primary batteries or acombination of them

Wake-up on radio (WOR) Micro- or nanopower technology that allows the DDC node to switch to RDC mode when anevent of interest occurs

Beacon TX Radio transmitter used to trigger remote WORs Typically it is separated from the regular radiotransceiver module

Intelligent sensor It is not the WSN node but refers to a sensor probe which has the capability to wake up the mainMCU via interrupt

End device (ED) The regular sensor node in the BETS architecture while in operating in LDC mode

Cluster head (CH) The data-collector node in the BETS architecture while in LDC mode ED nodes onlycommunicate with a CH node

Table 2 Power profile used in the simulations

MCU Sensors RadioActive 5mW 30mW (5 s) 70mW (3 s)Sleeping 001mW 01mW 01mW

Table 3 Target applications for LDC-only mode

Design space OptionGoal Sense onlyTime Periodic data collectionDeployment Planned a prioriTopology StaticData Rate LowDelay tolerance Must support

supercapacitors is directly controlled by this moduleas well as the power gating ofmanymodules such as asensing probeWhen an energy-harvester is not used

the functions of the energy-management controllercan be absorbed by the Main MCU

(G3) For Case 3 the mainMCU is sleeping while the nodeis in RDC mode and the radio transceiver (ie alegacy WSN node) is active However sometimes itis necessary to wake up the main MCU It occursbecause in order to power onoff any module suchas a digital sensor connected to the legacy WSNnode there is an energy-management hierarchy to befollowed radio transceiverrArrmain MCUrArr energy-management controller rArr power-gating device Inthis case the legacyWSN node can wake up the mainMCU by means of an interruption line as shown inFigure 5

(G4) In general for Cases 1 and 2 the sensor devices arephysically connected to the Main MCU For Case 3they can be either connected to the Main MCU(recommended) or to the Radio Transceiver

(G5) The power hibernation refers to a continuous and longperiod of sleeping time defined in terms of minutes


or hours [9] While hibernating it is possible to shutdown the majority of the modules of the node whichleads to significant energy savings The drawback ofthis approach is the resulting delay to have the nodeready to resume its tasks

(G6) The operational schedule of the node in LDCmode isdefined by a centralized energy-management modulerunning at the data server and this information is sentto the nodes via BETSprotocol ForCases 1 and 2 thatis LDC-only mode such scheduling information istypically defined in terms ofminutes such as cycles of15 or 20minutes for some environmental applications[3] Not all nodes at the segment need to follow thesame schedule In fact heterogeneous scheduling isone of the provisions of a central data server basedenergy-management to evenly balance the energyresources in the network For Case 3 the schedule ofthe nodes in LDC mode is potentially defined by theradio transceiver which is the legacy WSN node Inthis case the value of the LDR cycle is mainly associ-ated with the expected network QoS metrics and thecharacteristics of the application

(G7) The wake-up on radio (WOR) module is connectedto the main MCU Its function is to wake up the nodewhen it is hibernating in LDC mode and an events ofinterest is detected Without such provision a nodein LDC mode could not quickly resume to the RDCmode in case of a critical event For instance considera surveillance system under Case 3 and assume thatthe nodes are hibernating while in LDC mode Oneintelligent sensor attached to a specific sensor nodedetects the presence of an intruder and it promptlyawakes the main MCU of the node where it isinstalled However the major challenge is to wake upall EDs of the same segment and eventually all the net-work in order to switch back to RDC mode and pro-vide the expected functionalities of the surveillanceapplicationTherefore the key-answer to address thischallenge is the WOR module once the first node isawaken by its intelligent sensor it transmits a specialbeacon by means of the beacon TX module shown inFigure 5This beacon triggers theWORmodule of theCH node of that segment In turn the CH wakes upall EDs of the segment using the same technique Anultra-low-power WOR is a recent and sophisticatedtechnology [16] and its use is recommended forevent-driven scenarios

(G8) For Cases 2 and 3 one preliminary and critical testto be performed is related to the feasibility of usingthe legacyWSN node as the radio transceiver moduleof the DDC node It is necessary to evaluate a point-to-point communication with two nodes by meansof their serial ports because this is the typical waythat the main MCU communicates with the radiotransceiver For instance for TelosB there are the pinsUART0RX and UART0TX at the TelosBrsquos expansionconnector that allows such test Similarly the XBeemodules provide the serial ports TXD and RXD for

the same purpose For the integration of any WSNnode that supports pure ad hoc communication (iewithout involving any formof network hierarchy) theprocess is relatively straightforward However legacyWSNnodes that are natively based on infrastructuredtopologies such as XBee-PRO ZB ZigBee requirespecial attention For this specific example a nodewith the CH role in LDC mode can have the ZigBeeCoordinator (or Router) profile defined for itself andthe nodes with the ED role can have the ZigBee enddevice profile

(G9) TheBETS protocol assumes that the network segmen-tation is already in place when it is in operation Itmeans that for Case 3 an application-layer softwarerunning at the nodes in RDC mode must logicallyconfigure the network segment(s) One ED node canonly be logically attached to a single segment andeach ED node has assigned a logical address which isused by BETS to identify each ED In this case twonodes can have the same logical address providedthey belong to distinct network segments However ifa node of segment A is also in the communicationrange of the CH at the segment B potential errorscan occurThe simplest solution is to assign a distinctRF channel for each segment in particular if they arephysically close to each other However if the fre-quency-hopping spread spectrum (FHSS) technologyis being employed by the nodes further investigationis necessary Moreover for Case 3 it is possible thatthe network segment of the node in RDC mode doesnot correspond to the BETS segment (LDC mode)as in the case shown in Figure 6 That is an existingnetwork hierarchy in RDCmodemay not be the sameCH-ED hierarchy in LDC mode It is not the casefor legacy WSN nodes that have ad hoc operationHowever this problem can potentially occur whenZigBee-based nodes are used In this case the switch-ing between operational duty-cycle modes must bepreceded by a dynamic setup of the WSN modulesFor the mentioned ZigBee example potentially theBETS segmentation will be achieved by configuringchannel frequency node profile PAN id or a mix ofthem

4 Cross-Layer Protocol for Very LowDuty-Cycle (LDC) Mode

From now on this work assumes that the DDC system isoperating in LDC mode Therefore the network has alreadybeen logically segmented in multiple sets each one with aCH node and its associated ED nodes In this section wefocus our attention to one of these network segments anddiscuss how exceptional energy-performance is possible witha DDC node running the best-effort time slot Allocation(BETS) protocol in LDCmode BETS is an example of a cross-layer protocol that is compliant with the goals of the proposedframework However one can design a similar cross-layerprotocol that can also be efficiently used by the nodes in LDC


Delay tolerance

Periodic sensing

2-tier architectureSegmentationStatic topology

Appl

icat

ion

Net

wor

k

Implements

Selfish node concept

Time synchronization

Very highenergy savings

BETSprotocol

To achieve Requirements

Power-gating (onoff)

Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle

controlPower-saving mode

control

Assumptions

Figure 10 The cross-layer nature of the BETS protocol a candidate of choice for the LDC mode

mode The section starts with an overview of related work inthis context followed by an overview of BETS and its ownterminology Next the design goals of BETS are presentedand the functional details are discussed by means of itsalgorithms

41 Related Work The virtual elimination of collaborationamong wireless nodes is not a novelty The IEEE 802154standard [6] was introduced as a low-rate short-range com-munication solution for Wireless Personal Area Networks(WPANs) One of the network topologies defined in thisstandard is a star topology where a PAN coordinator is incharge of the communication with the remaining devicesSimilarly the Bluetooth technology is based on a star topol-ogy with a master node as the central point [24] Note thatsuch arrangements are similar to the relation CH-EDs in theBETS solution In fact the design of the network architectureassociated with BETS Ripple-2 was influenced by thesestandards and their outstanding success

IEEE 802154 only defines the specifications in relationto the physical and MAC layers However upper ISOOSIlayers can be optionally used to allow ad hoc deploymentsmultihopping trees and mesh networks One example ofsuch augmentation is the ZigBee standard [7] which definesthe network application and security layers Initially adoptedas a WPAN solution the functionalities of ZigBee becomepretty similar to the ones in traditional WSNs As expectedmany WSN deployments based on ZigBee devices have beenreported [2 22 24 25]

In this context BETS can be seen as an effort to addscalability and extreme energy efficiency to IEEE 802154(star topology mode) solutions without incurring in thehigher overhead and complexity of the ZigBee standard Alsothe overhead similar to the one associated with the PANassociation procedure [6] in the 802154 standard does notexist in the BETS solution Moreover BETS is also designedto support any underlying point-to-point physical layer notonly 802154 PHY In fact the ED-CH link implementation isnot even limited to a radio operation Finally different from aPAN coordinator (or ZigBee router or ZigBee coordinator)the CH node in the BETS solution can sleep and even betterit can hibernate This fact drastically reduces the energyrequirements of the data-collector device which still is achallenge for typical 802154-based solutions

A simplified version of WSN based on multiple stars ispresented in [26] for forest monitoring but the details related

to the underlying networking protocols are not providedA hierarchical architecture for delay-tolerant networks ispresented in [27] and a customized MAC protocol calledLiteTDMA is employed Hardware specialization of WSNnodes in particularwith the introduction of the power-gatingtechnique is proposed in [28] and it is extended in [9] Slot-based MAC implementations have been proposed such asTRAMA [29] PMAC [30] Z-MAC [31] and H-MAC [32]AlthoughBETS is not aMACprotocol its core functionalitiesin relation to the time-synchronization among nodes ofthe same network segment have some similarities with thementioned MAC protocols

To the best of our knowledge this work (as an extensionto [4]) is the first to propose a non-collaborative model forWSNs by means of the implementation of the selfish nodeconcept presented in Section 34 As already discussed inthe previous sections such low-energy model is exclusivelyadopted while the DDC node is in LDCmode Even for someapplications the nodes can permanently operate in such LDCmode (where BETS lies)

42 Protocol Overview BETS is a novel cross-layer protocolimplemented as an application-level overlay It is designedfor low data rate low duty-cycle and sense-and-send WSNsWhen the DDC node operates in LDC mode BETS is theprotocol of choice BETS operates at the MAC and upper-level networking layers If an existing MAC protocol alreadyexists in the target sensor platform the MAC functionalitiesof that platform can be disabled or simply ignored if notcausing significant overhead as discussed in Section 35 Inother words the ultimate actions necessary to achieve afair contention-free and reliable communication channel aretaken by BETS As shown in Figure 10 BETS assumes thata periodic sensing application is in place which matches theway the network behaves while in LDC mode The protocolhas a provision to capture the scheduling data sent by themain application (data server) to the nodes and defines theproper allocation of the wireless channel in the time domainIn this sense the term schedule refers to the same object forboth application and network discussions

Also shown in Figure 10 the energy efficiency of theprotocol is mainly achieved by sacrificing the network per-formance in terms of data latency Therefore the main appli-cation must afford such higher delay which can vary fromseconds to hours according to the final implementation Norouting-related functions are actually provided by BETS and


ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next

middot middot middot

ED MEAS CH CTRL ED CTRL


ED1 rarr CH CH rarr ED1 ED1 rarr CH


Figure 11 BETS functionality (ED side) an implementation of the selfish node concept

it is assumed that the network is divided into multiple star-like segments following an asynchronous 2-tier architectureapproach Another assumption shown in Figure 10 statictopology calls our attention to the fact that BETS protocolcannot be easily modified to support mobile nodesThere areoptional assumptions to be considered in the BETS contextAs shown at the right side of Figure 10 if the ultimate designgoal is to achieve very high energy efficiency (ie more than1 order of magnitude compared to state-of-the-art solutions)the combination of very low application duty-cycles (ielt1)and power hibernation techniques [9 28] is a required stepAs already discussed a DDC node has such capabilities

The adoption of star-like segments potentially reduces thecomplexity of a network design However it is important toverify if the center point (access-point controller or cluster-head) does not become the real bottleneck of the solutionFor instance the energy issues related to the CH node inWSNs have been studied for a long time and a CH role-rotation scheme has been proposed [13] In the architecturalcontext where BETS is implemented such role rotation is noteasy to be implementedTherefore in order to achieve energyefficiency for both ED and CH nodes a nontrivial solution isrequired and this is themain challenge of the BETS design Infact besides the role rotation we did not find in the literaturean approach to efficiently reduce energy consumption of CH-like nodes in a static topology In BETS we have the goal ofhaving the energy profile of the CH node linearly followingthe average energy profiles of the ED nodes in the segmentNote that besides a higher energy capacity (if CH role rota-tion is not adopted) and strategic communication coverage(which depends on the location of the node) theCHrole doesnot require special hardwareprocessing capabilities and anynode can potentially have this role assigned to itself

From the EDrsquos viewpoint the fundamentals goals of BETSare the implementation of the selfish node concept and anefficient way to avoid that two or more EDs try to use thecommunication channel simultaneously Because the currentMAC protocols do not fully implement the selfish nodeconcept BETS must have MAC-related functions in order tofulfill the mentioned goals The messages exchanged by ED

and CH nodes are shown in Figure 11 From the viewpoint ofthe selfish node the process occurs as follows

(1) ED node wakes up and takes measurements(2) Without any channel negotiation ED sends the

ED MEAS message to the CH node (unicast) Thismessage basically contains themeasurements and fewcontrol data such as its energy state and communica-tion metrics In some scenarios instead of individualsensing measurements the ED MEAS message cancontain aspects related to the status of the nodecompressed data derived from a historical sequenceof measurements and so forth

(3) Without any channel negotiation CH sends back aCH CTRL message to the ED node (channel broad-cast logical unicast) This message contains theschedule for the next cycle related to that node

(4) Without any channel negotiation ED sends aED CTRL message to the CH node (unicast) Thismessage acknowledges the reception of the schedule

(5) ED node configures its wake-up circuitry accordinglyto the received schedule and sleeps

As expected no collaboration among nodes exists andthe implementation of the protocol becomes significantlysimple at the ED side In fact such simplicity clearly indicatesthe proper realization of the selfish node concept One canargue that even the CH CTRL and ED CTRL messages canbe eliminated for a full implementation of a selfish nodeconcept However the reliability of BETS requires that somesort of minimum communication-quality control exists aswill be explained later in this section In fact the smalloverhead associated with these messages becomes irrelevantin low duty-cycle applications

From the CHrsquos viewpoint the implementation of theBETS protocol is not so straightforward as in the ED caseBesides the proper support for EDs it is important that CHsleeps in optimum cycles Different schedules for the nodes ofthe same segment can potentially cause energy inefficienciesfor the CH node Even a global schedule may not provide


ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS

ED CTRLCH CTRLED MEASED rarr CH ED rarr CHCHrarr ED

Figure 12 An example of regular (no errors) operation of BETS from the CHrsquos perspective At inactive MTSs all nodes (EDs and CH) aresleeping However the CH node can still use an inactive MTS for the CH-BS (or CH-data server) communication

the best energy performance for the CH node For instanceassuming only 5min application cycles oneEDcan follow thesequence 0ndash5ndash10ndash15 in the line of time while another EDthat was initialized later follows the 3ndash8ndash13ndash18 sequenceIn this case both EDs have the same schedule but the CHnode cannot take a longer sleep although such goal is clearlypossible to be achieved in this case Therefore BETS mustprovide a way to accommodate schedules in the best efficientway possible both for ED and for CH nodes In fact the firstreason for the term Best-Effort in the BETS acronym is relatedto this aspect To solve the mentioned issue which hereafteris called dispersion BETS adjusts the first cycle of the secondnode which was turned on at moment 119905 = 3 Therefore thisnode will have this wake-up sequence under BETS 3ndash5ndash10ndash15 Note that the first programmed cycle and only this oneis adjusted in order to group all EDs with the same scheduleIt is worth to highlight that such adjustment is fundamentalfor saving energy at the CH side while such mechanism doesnot affect the EDs In other words BETS is a dispersion-freeprotocol

The second reason for theBest-Effort term is related to thereliability of the solution In order to achieve very high end-to-end reliability metrics (in this case for the ED-CH link)multiple handshakemessagesmay be necessary However forevery active node BETS provides a time-slot with a smalland fixed length Bigger andor dynamic slot lengths can beadopted to increase the communication reliability but thereare energy penalties to be considered In our simulations to beshown in Section 5 different communication channel errorrates are analyzed under BETS in order to evaluate energy andreliability metrics Also in one of our current BETS imple-mentations the average ED-CHdistance is around 210mandfor that case the total data loss was smaller than 18 duringmultiple weeks The slot length was parametrized to be largeenough to allow just a single additional round of ED MEASminusCH CTRL ndash ED CTRL messages if necessary Doing so thesolution became more reliable but less energy-efficient andclearly the data latency of the solution increases The bottomline in this discussion is that the provision of large time-slots (multiple messages in sequence) in order to increase the

reliabilitymay not be really necessary Nonetheless due to theBETS flexibility the slot length can be modified

43 Definitions and Terminology Before proceeding with adetailed explanation of BETS some terms shown in Figure 12must be properly introduced or better defined In additionsome contextual aspects are discussed in order to ease theadoption of BETS for the LDC mode of DDC systems

Logical Network Segment (or Simply Segment) A fundamen-tal assumption for BETS is that the network is divided intological clusters or segments This division is realized consid-ering the physical topology of the network Accordingly it isexpected that the CHnode be located at the center of a virtualcircle where all nodes inside that circle are able to communi-cate with that CH (unit disk graph approach) Realisticallythe communication range of the ED nodes will significantlyvary due to many reasons Moreover the location of thenodes must be primarily governed by the application needsTherefore it is very hard to achieve an ideal division of thenetwork into circles that do not overlap Due to this factadditional communication techniques must be employed toenforce that a node solely communicates with a single CHnode even if more than one CH can be reached For instanceby using different channelsfrequencies or even by usingsoftware filters it is possible to deal with the overlappingcircles issue This enforced concept of segmentation dividesthe overall network into logical network segments ED nodesof the one (logical) network segment can only communicatewith the CH node of that segment and vice versa

CHrsquos Children All ED nodes of the same network segmentassociated with a certain CH node are children of that node

Registered Children When CH communicates with DS thelatter can potentially send explicit information about thenumber of children EDs and their respective schedules Inthis case the ED nodes are considered registered childrenand the CH can properly calculate how much time to spendwaiting for the contact of an ED node based on the number


of registered children Such information is easily availablein planned deployments with a static topology and can alsobe modified to reflect possible node failures The number ofregistered children is not required for the functionality ofBETS but it increases the energy efficiency of the CH nodebecause it can hibernate as soon as possible

Major Time Slot (MTS) OnceCH is initialized (boot) the lineof time is divided into fixed periods of time calledMTSs eachone with the lengthmtsLength a software variable expressedin units of seconds In our implementation and also in thesimulations the value formtsLength is 300 s (5min) In orderto achieve a collision-free solution and maximum energyefficiency it is assumed that the application schedules are alsogiven in mtsLength units

Active and Inactive MTS The CH node does not have to benecessarily active all the time that is for every MTS Asexpected during some MTSs the CH node is sleepingbecause all its children are also sleeping and such MTSsare called inactive MTSs When CH is ready to hear an EDnode the respective MTS is called an active MTS The activeMTS (AM) is divided into three sequential parts with variablelengths ETS BTS and STS as explained next

ED Time Slot (ETS) It refers to the initial part of an activeMTS (AM)which is used specifically for communicationwithchildren EDsThe dynamic length of ETS corresponds at leastto the sum of the assigned time-slots for the active childrenat that MTS One of the BETS algorithms uses the registeredchildren parameter in order to determine the optimum lengthof ETS for each AM Its ultimate goal is to allow CH to sleepas soon as possible

BS Time Slot (BTS) It refers to the second part of an activeMTS which is used for the CH-BS (also CH-DS) communi-cationThe length of a BTS varies as a function of the amountof data to be sent to the BS node CH-BS link throughputand possible errors at this link To save energy ED data fromdistinct active MTSs can be aggregated Doing so the BTSlength is 0 for the majority of MTSs and it is maximum forfew MTSs Although the CH-BS data transfer can be dividedinto multiple MTSs a BTS transaction cannot conflict withan ETS of an active MTS However by also using inactiveMTS for CH-BSDS communication the BTS transactioncan last longer as shown in Figure 12 Finally it is possibleto change the CH-BS transfer scheduling according to theenergy performance metrics at the CH node Due to theasynchronous nature of the second network layer (CH-BS)the communication with the BS must not impact the BETSperformance for the ED nodes

Sleeping Time Slot (STS) It refers to the third part of an activeMTS which is actually not being used During this period oftime the CH is inactive and potentially sleeping

Time Slot (TS) This the time slot allocated to each individualED The TS has a fixed and unique length tsLength for eachsegment a software variable expressed in units of secondsSuch parameter basically corresponds to the time necessary

for a single ED MEAS minus CH CTRL minus ED CTRL transactionas illustrated in Figure 12 However in practice tsLength isa little bigger and it is influenced by many factors such asthe characteristics of the EDrsquos radio transceiver and wirelesschannel the use of a power-gating technique and numberof possible retransmissions In our implementation and alsoin the simulations tsLength is 8 s (1 retransmission is sup-ported) When the retransmission is not necessary which isusually the case the corresponding reserved time period atthe end of a TS window provides a gap between TSs andpotentially mitigates drift clock and channel contentionissues Moreover multipath effect is significantly reducedwhen very large gaps are employed As a result it is possibleto extend the communication range of the nodes to valuesvery close to their maximum ones mentioned in their radiomodule datasheets [4]

Homogeneous Scheduling It refers to the scenario where allED nodes of the same network segment have exactly the samesleeping schedule That is they wake upsleep at the sameAMs

Heterogeneous Scheduling It refers to the scenario where atleast two ED nodes of the same network segment have dif-ferent schedules A network with multiple segments can havehomogeneous and heterogeneous scheduling schemes at thesame time

Dispersion In heterogeneous scheduling the dispersion isdefined as an anomaly characterized by having some EDnodes using improper MTSs causing a negative impact onthe energy efficiency of the CH node In other words thegoal of having the maximum number of inactive MTSs is notachieved although it is possible As already discussed BETSis designed to be dispersion-free

EmergencyMode (EM) In normal BETS operation all EDs ina segment are properly synchronized with the CH nodeThatis they correctly follow their assigned time slots Howeverwhen a node is (a) deployed for the first time (b) restartedor (c) does not receive CH CTRL after sending ED CTRLmessage it does not have any time-slot assignment In thiscase the node follows a different algorithm (EM) in order tocommunicate with the CH node

Convergence When all ED nodes in a segment are in regularoperation (not in EM) that network segment is said to beconvergent In a nonconvergent segment one or more nodecan try to contact CH while it is sleeping Also the node inEM can interfere with the current assigned TSs and that seg-ment is nomore contention-free (temporarily)While in non-convergent state the network has significant energy penaltiesTherefore a design goal for BETS is to have segments thatquickly converge

Node Qualification Because BETS is a loose protocol in termsof wireless channel error control mechanisms it is importantto establish ways to preventmitigate errors at the ED-CHlink The node qualification is a procedure used to addressthis challenge During more than 2 years we have been


deploying nodes in different outdoors sites and the followingnode qualification guidelines are based on our field workexperience Before the final deployment of an ED nodeit is recommended to verify the conditions of the wirelesschannel at the location where an ED node is expected to bedeployed in particular for the ones in critical areas (longdistance ED-CH line-of-sight issues trees topology etc)First the average noise floor level is measured (NF) Secondit is measured the received signal strength (RSS) as providedby the ED node in relation to messages sent by CH TheRSS value must be significantly higher than NF (eg 5 dBor more) For instance if the RSS = minus87 dBm and NF =minus93 dBm this location is potentially close to the boundaries ofthe network segment but it is acceptable In this limit-case itis also recommended to run communication tests to evaluatethe channel conditions for a final decision If the node cannotpass the qualification test a new network segment must beused the CH must change its position or new antennaschemes must be adopted and so forth The bottom line isthat the energy performance of BETS is strongly affected bythe wireless channel errors as will be discussed in Section 5

44 Design Goals The main goals of the BETS protocol aresummarized as follows

(1) To be functional in relation to

(a) providing a way to send fixed or dynamic sched-ules for EDnodes that are originated at themainapplication running at the BS node or above(data server)

(b) collecting sensing and basic control data (net-work-related errors and energy-related metrics)from ED nodes and send such data to the mainapplication

(2) To implement the selfish node concept and maintainfairness for the wireless channel access

(3) To be energy efficient (ED side) by imposing anetwork overhead not higher than 075 even if theprobability of errors at the communication channel(ED-CH link) is as high as 5 This overhead limitis motivated by the analysis of the scenario illustratedin Figure 7 specifically related to very low duty-cyclesense-and-send applications

(4) To be energy efficient (CH side) by allowing CH tohave optimum sleeping cycles even in case of hetero-geneous scheduling

(5) To isolate network problems between segments(6) To isolate problems related to ED-CH and CH-BSDS

communication links(7) To support network management tasks as follows (a)

verify the reliability of individual ED nodes and theED-CH communication links and (b) isolate erraticED nodes

(8) To mitigate the wireless channel contention amongthe ED nodes of the same segment BETS must be

insensitive to the existence or not of a MAC protocolrunning above the physical layer

(9) To provide support for optional use of power hiber-nation schemes in order to achieve very high energysavings

Observe that no specific attention is given to networkperformance metrics such as maximum transmit delay orthroughputThis fact anticipates themost important trade-offof BETS the network performance is expected to be sacrificedin order to obtain impressive energy achievements in con-junction with excellent scalability and reasonable reliabilityThe main reason for this trade-off is the isolation betweenED-CH and CH-BSDS data flows (asynchronous approach)Therefore the possibility to switch from LDC (BETS) to RDCmode is a necessary provision for critical WSN applications

45 BETS Normal Operation In order to realize the designgoals number 2 (selfish node concept) and number 8 (con-tention-free) a TDMA approach is used to avoid contentionamongnodes and allow a fair usage of thewireless channel Byreserving a time slot (TS) for each ED node that will use thesame future active MTS (AM) the wireless channel becomespotentially contention-free When an ED wakes up in itsassigned TS it immediately starts sensing and without anydelay sends the data to CHObserve that this sense-and-sendapproach provides the best energy efficiency possible becauseno care is given by the ED node related to the possibility ofa busy channel or the availability of CH Clearly it corre-sponds to the selfish node concept In fact this procedure isautonomously repeated by the ED node from time to timeeven if the CH node is not operating In normal operationonce ED MEAS is received it is followed by the CH CTRLmessage sent by CH The CH CTRL message has two pur-poses First this message acknowledges the proper receptionof ED MEAS Second CH CTRL contains configuration datato ED such as the next time that the node must be activeOnce ED receives CH CTRL message it sends back theED CTRL message This message also has two goals Besidesserving as an acknowledgment for CH CTRL the ED nodeuses this message to send its control (log) data battery statuspower shortages communication errors and so forth Atthe BS side (or above) the main application can recognizeerratic patterns associated with a specific node Therefore byassigning a very long schedule specifically for that node (eghours or days) this one can be virtually isolated from thenetwork

The ED MEAS + CH CTRL + ED CTRL transactionforms the core of a handshake-based procedure in BETSHowever in contrast with the traditional usage of acknowl-edgmentsmissing of one of themessages does not necessarilytrigger a retransmission In our implementation of BETS weprovide a second transaction round at the same TS if the firstone fails This explanation helps to clarify why the defaulttsLength used in our simulations (and real implementations)is 8 s while the associated messages only sum up to 3 saccording to the Table 2 Besides the retry timing tsLengthalso encompasses the potential timeouts associated withcollisions and other communication errors Observe that the


mentioned message redundancy provision is not a formalspecification of BETS but tsLength can be increased evenmore to support multiple retries In this way a critical net-work segment in terms of channel communication errors canhave a higher tsLength to increase the likelihood of successfultransaction

In our real-world BETS implementation all the 3 men-tioned messages are sent twice with a small delay betweenthe messages Empirically we figured out that such provisionhighlymitigates the possibility of amissingmessage in partic-ular for outdoorsThis second form of redundancy provides away to increase the reliability of the communication withouthaving to introduce timeouts or additional complexity atthe protocol Again such effort is a design possibility butnot a specification of BETS Nonetheless this discussion isimportant to highlight the strength of BETS in terms of itsadaptability for different network scenarios

The acknowledgments provided by CH CTRL andED CTRL messages are primarily used as a networkmanagement tool In other words it is possible to identifyenergy and communication problems related to a certainnode or a group of nodes Also the communication quality isevaluated in both directions (ie ED-CH and CH-ED) Thislater aspect is very important because CH and ED nodes mayhave different antennas For instance an omni-directionalone for CH and a directional one for EDs Alternatively ahigher antenna gain (typically a bigger antenna) for CH anda regular one for EDs The bottom line is that in all casesthe goal number 7 (network management) is fully satisfiedin BETS That is it is possible to detect and correct reliabilityissues at the network already deployed On the other handthe 3 mentioned messages only satisfy the goals number2 and number 8 when the segment is operating in normalconditions that is it is convergent and all nodes are wellsynchronized Erratic scenarios under BETS are considerednext

46 BETS Dealing with Erratic Scenarios So far the energy-efficiency fairness and collision-free channel characteristicsof BETS are achieved provided that all nodes properlyhave and follow their TS assignments (convergence) In thissection the erratic scenarios are considered and the BETSconvergence process is discussed

The first erratic scenario is related to the well-knownclock drift issue Even when all EDs are properly synchro-nized the minimal differences among their internal clockswill eventually cause overlapping between TSs BETS solvesthis problem by continuously providing a schedule adjust-ment for each ED node Such adjustment occurs every timean ED node receives a CH CTRL message Therefore ingeneral the clock drift hardly impacts the solution Howeverthe combination of very long schedules (eg gt5 h) and low-quality clock modules must be avoided in order to mitigatethe risk of clock drift issues If such schedules are reallyexpected two guidelines can be employed Firstly higherquality clock systems with temperature compensation canbe used at CH and ED nodes Alternatively a higher valuefor tsLength can be used in order to enhance the effectivegap between consecutive time slots Note that this parameter

is a key one to properly adapt BETS to the existing topol-ogynetwork scenario

The remaining erratic scenarios are basically associatedwith the same final result and they can be analyzed in a singlescenario under BETS Specifically no matter if the ED-CHtransaction fails (collision or other communication error)CHED node restarts or an ED node is recently deployed inall cases the network temporarily is non-convergent Whenan ED node timeouts the reception of an expected CH CTRLmessage it automatically enters in emergency mode (EM)A related solution typically used in MAC protocols is thecombination of channel overhearing with random back-offsIn BETS only the second part of this technique (randomback-offs) is used The reason for this approach is related tothe support of hibernation mode at EDs as stated in the goalnumber 9 and explained next

The fixed-length nature of a TS (deterministic approach)highly promotes the adoption of supercapacitors at the EDnodes as part of the hibernation solution However becausethe power used to overhear a wireless channel is very highit would be necessary to charge supercapacitors with anamount of energy multiple times the expected one for asingle TS transaction thus resulting in drastic energy inef-ficiencies Because this design aspect is closely related to thehardware characteristics we empirically evaluated differentpower management solutions for typical WSN nodes and weconcluded that under the context of BETS instead of theoverhearing scheme a more energy-efficient solution wouldbe using the back-off technique alone The EM procedurewith an optional support for supercapacitors is provided bytwo algorithms (Algorithms 1 and 2) one for the EDnode andthe other for the CH node

In order to provide support for different hardware plat-forms some of the parameters in Algorithms 1 and 2 areclearly hardware-dependent Algorithm 1 is executed onceED node enters in EM and it is reexecuted for each unsuc-cessful tryout while in that mode To generate randomTime1and randomTime2 a discrete uniform distribution functionis used Assuming that the Node Qualification proceduredescribed in Section 43 was considered all ED nodes areexpected to have similar behavior in relation to the EH-CH link performance Therefore the mentioned uniformfunction can potentially be adopted in our model

Algorithm 2 is executed once CH node starts a new activeMTS and it is reexecuted each time a newED MEASmessageis received Note that the basic idea behind the Algorithm 2 isto make the CH node extending the current ETS slot in orderto wait more time for the missingnon-convergent childrennodes At the upper part of the algorithm it is considered thecase where no information about the number of the nodesare provided and a higher time extension is provided (notthe best energy-efficient approach) At the bottom part ofthe algorithm we consider the expected case where the CHknows howmany nodes are expected to belong to its segmentIn this case two extra ETS slots (in terms of tsLength time)are included in the time extension for ETS considering thefact that when one node tries to communicate with CH it canalso collide with another one properly assigned ED and bothare can go EM state Once one of the non-convergent nodes


RequiremaxContinuousEM max time in EMmode before rebootingRequire currTime time elapsed since the beginning of the active MTSRequire randomTime1 (random value between 5 to 11) lowast tsLengthRequire randomTime2 (random value between 300 to 720) lowast tsLengthRequire SCchargingTime time to charge SCs in EMmode (default 0)Require tryouts incremented for each unsuccessful CH transactionEnsure A segment with up119873 = 08 lowast (119898119905119904119871119890119899119892119905ℎ119905119904119871119890119899119892119905ℎ) EDs converges

119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890

elseif supercapacitor(s) is(are) used then

119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0

backoffTimelarr 1199031198861198991198891199001198981198791198941198981198902

sleep 119889119906119903119894119899119892 backoffTime 119901119890119903119894119900119889elsebackoffTimelarr 1199031198861198991198891199001198981198791198941198981198901

idle (radio off) 119889119906119903119894119899119892 backoffTime 119901119890119903119894119900119889end if

end if

Algorithm 1 Emergency mode (implemented at ED node)

Require expectedNumChildren registered children (119873)

Require numChildrenCurrentMTS pending EDs with TSs in this MTSRequire numFutureChildren EDs with assigned TSs in future MTSsEnsure A segment with up119873 = 08 lowast (119898119905119904119871119890119899119892119905ℎ119905119904119871119890119899119892119905ℎ) EDs converges

return waitEDTime how much time CH must wait hearing childrenif CH just booted then

119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ

elseif 119890119909119901119890119888119905119890119889119873119906119898119862ℎ119894119897119889119903119890119899 = 0 (CH never contacted BS node) thenif 119899119906119898119862ℎ119894119897119889119903119890119899119862119906119903119903119890119899119905119872119879119878 = 0 (no assigned TS in this MTS)then

119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ

else119908119886119894119905119864119863119879119894119898119890 larr 119905119904119871119890119899119892119905ℎ lowast (119899119906119898119862ℎ119894119897119889119903119890119899119862119906119903119903119890119899119905119872119879119878 + 5)

end ifelse (BS had already sent the number of EDs in this segment)119886119906119909 larr 119899119906119898119862ℎ119894119897119889119903119890119899119862119906119903119903119890119899119905119872119879119878 + 119899119906119898119865119906119905119906119903119890119862ℎ119894119897119889119903119890119899

if 119886119906119909 = 119890119909119901119890119888119905119890119889119873119906119898119862ℎ119894119897119889119903119890119899 then[(convergence achieved)]119908119886119894119905119864119863119879119894119898119890 larr 119905119904119871119890119899119892119905ℎ lowast (119899119906119898119862ℎ119894119897119889119903119890119899119862119906119903119903119890119899119905119872119879119878 + 2)

else (potentially one or more EDs missing in this AM)119908119886119894119905119864119863119879119894119898119890 larr 119905119904119871119890119899119892119905ℎ lowast (119890119909119901119890119888119905119890119889119873119906119898119862ℎ119894119897119889119903119890119899 minus 119899119906119898119865119906119905119906119903119890119862ℎ119894119897119889119903119890119899 + 2)

end ifend if

end if

Algorithm 2 Emergency mode (implemented at CH node)


finally succeeds the ETS is extended again considering thenew scenario of how many nodes already synchronized andhow many are missing

In this work we will limit the verification of the correct-ness of these algorithms by analyzing the simulated results inSection 5 In our simulations and context the goal is to havethe network convergence achieved in less than the time of(4lowast119904119888ℎ119890119889min) where 119904119888ℎ119890119889min is the smallest of the schedulesamong the ED nodes that are already synchronized with CHIf all nodes are in EM 119904119888ℎ119890119889min is the default schedule usedby CH temporarily without children Because such values areon order of many minutes it is clear that when a dual system(RDCLDC modes) is used the frequency of the switchingprocess is clearly impacted by the long time (eg gt15min)necessary to converge the network under BETS If this modeswitching occurs a few times per day this is not an issueHowever the need of more frequent switching requires akind of support not natively provided by BETS One potentialand relatively easy way to mitigate this issue in dual systemsis to sequentially add EDs to the segment (rather than allnodes at the same time) when the DDC system switchesfrom RDC to LDC Again although such provision is not aBETS mechanism it is clear that this procedure just requiresa small development effort at the software side of the radiotransceiver module

Themost important aspect behind these algorithms is thefact that the convergence can be achieved with any numberof ED nodes provided that the CH is active for enoughtime because the maximum continuous time the CH canwait for children is limited by mtsLength Therefore thisparameter essentially governs the convergence feasibilityHowever tsLength is ultimately the parameter that influencesmtsLength For instance with a tsLength = 8 s and 100 EDnodes the minimum mtsLength is 1000 s (166min) As aresult the application schedulemust be an integermultiple ofthis value such as 20 40 and 60min In our implementationwe opted by a maximum number of 30 nodes leading tomtsLength of 5min which is more useful from the viewpointof many existing environmental monitoring applications

With smaller values of tsLengthmuchmore nodes can besupported in a segment while practical values for mtsLengthare still maintainedTherefore one natural question is associ-ated with the adoption of a high value for tsLength (eg 8 s)when it is well known that typical RF transceivers can realizethe full transaction in less than half a second Besides theadditional time for a possible retransmission and for a securegap between TSs the answer lies in the choice of the hardwareplatform In our implementation a single transaction wasactually achieved with less than 2 s and finally 8 s were deter-mined as a secure value based on experiments and takinginto account our transceiver choice power-gating latency [9]reliability aspects and critical outdoor environments Thelatter aspect is important for our ongoing project becausewe would like to achieve ED-CH distances very close to theinformed by the manufacturers of the radio modules Whenthe messages in the network are separated by significantgaps the multi-path effects in hilly regions are significantlyreduced as already highlighted and this fact is observed

empirically as discussed in Section 5 Nonetheless a highertsLength typically will not affect the energy efficiency ofEDs provided that network errors do not occur frequentlyHowever a higher tsLength can definitely impact the CHnode because it must be active more time during each cycleThe next section discusses the energy efficiency of the CHnode

47 Energy Efficiency of the CH Node The maximum energyefficiency at the CH side is achieved with a homogeneousscheduling In this case the nodes of one network segmentalways use the same AMs whenever they are expected tosense-and-send and the goal of achieving the maximumnumber of inactive MTSs is trivially achieved Althoughthis fact does not represent any additional energy-relatedadvantage for the ED nodes it definitely extends the sleepingtime of theCHwhich is the nodewith themost critical energyconstraint in the network segment Moreover continuousand long sleeping periods maximize the efficiency of thepower-gating technique because there are energy penaltiesassociated to the switching transients

However considering that an energy-aware system run-ning is on top of the BSDS (which is a guideline of this frame-work) such system can potentially select different schedulesfor EDs In this case the already mentioned dispersionproblem can occur andmust be addressed byBETSAlthoughthe schedules are sent by themain application running on theBS side due to its cross-layer nature the CH is effectively thenode that controls the distribution of the schedules to its chil-dren Based on this observation it is possible to implement analgorithm at the CH side that can properly adjust (advancingor delaying) the next active-time information (via CH CTRLmessage) sent to each ED in order to avoid dispersion Suchapproach is used by Algorithm 3 which is a dispersion-freeprocedure for the CH node providing maximum energyefficiency for this node as stated in the goal 4 There aresome interesting aspects that need to be highlighted in theAlgorithm 3

(i) The time of the entire network segment is referenceduniquely by the CH clockWhen the CH is initializedthemoment 119905

0is defined and theMTS 0 has its begin-

ning No real-time information is exchanged throughBETS when an ED node transmits ED MEAS CHtimestamps it with a real-time value based on itslocal clock This scheme helps to maintain BETS verylight but as expectedmeasurements with time-stamperrors on the order of seconds can occur Fortunatelyit is rarely an issue for non-real-time applicationsrunning in the LDC mode of the network

(ii) The least common multiple (LCM) principle is usedto avoid dispersion For instance if currMTS is 13 andan ED node (just initialized) has an assigned scheduleof 5 we can initially expect that it wakes up again atMTS 18 However this is not the case In Algorithm 3the variable 119886119906119909 has the value 119891119897119900119900119903(135) = 2 andbecause 119899119890119909119905119872119879119878 = (2 + 1) lowast 5 = 15 this nodewill actually wake up again 2MTSs ahead (MTS 15 =

13 + 2) not at MTS 18 This adjustment only occurs


Require schedED Table ED schedules assigned by main app (BS side)fields (ED id x) where x is an integer multiple ofmtsLength

RequireMTS Table MTS allocation table (old current and future AMs)fields (ED idMTS seq) whereMTS seq is the next AM for ED id

Require fixedDelta readiness time for CH (from wake-up to hear EDs)Require currMTS current value of MTS (numbered sequentially)Require currED id of the ED on the current transactionEnsure TS overlapping does not occurEnsure Dispersion-free TS allocation maximum energy efficiency

return adjSch next ED activation time (to be sent via CH CTRL) dispersion-free119909 larr 119904119888ℎ119890119889119864119863 119879119886119887119897119890[119864119863 119894119889]

119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)

no TS overlapping continuous TSs assignment119886119889119895119878119888ℎ larr 119899119890119909119905119872119879119878 lowast 119898119905119904119871119890119899119892119905ℎ

119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)

119890119905119904119860119889119895 larr (119899119906119898119864119863119904119878119886119898119890119872119879119878 minus 1) lowast 119905119904119871119890119899119892119905ℎ

119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895

Algorithm 3 Dispersion-free TS allocation (CH node)

at the first assigned cycle for that node From thatmoment on the node continues following the expect-ed scheduling In short the dispersion is avoidedbecause this node has this wake-up sequence in theline of time (13 15 20 25 )

(iii) After calculating the new AM for the node anadjustment is performed by CH in order to avoid TSoverlapping In other words if 3 nodes have the samecycles (schedules) to operate we do not want all ofthem to send their messages at the same time thatis at the beginning of that AM To avoid this issuethe number of current EDs that share the same futureAM is recorded Every time CH allocates a new EDfor that AM it adds a specific delay in order to haveall the nodes accessing CH in an efficient sequentialand contention-free manner The maximum energyefficiency is achieved for the CH node

(iv) The algorithm is implemented without any historicalcontrol or complex proceduresdata structures Suchsimplicity provides the path to implement the CHside of BETS in hardware platforms already used byregular WSN nodes This aspect is extremely impor-tant in a dual mode system (with RDCLDC modes)because the CH nodes are assigned among regularWSN nodes If the CH requires a very powerful mainMCU (which is not the case) the assignment of CHswould be impacted

5 Simulated and Empirical Results(LDC Mode)

In order to verify the correctness of the BETS algorithms andalso to determine the constraints of the solution experimentsare performed In addition to simulations preliminary results

from our field deployments are provided The following aresome questions of the particular interest in this context

(i) Are the algorithms correct and is goal number 3 (EDenergy-efficiency) feasible

(ii) Assuming that CHwas just installed howmuch addi-tional time is necessary to achieve convergence

(iii) Given a certain probability of errors for the commu-nication channel between ED and CH nodes what isthe associated energy penalty

(iv) When will the network not converge

51 Experimental Setup For this work we developed a spe-cific network simulator for BETS (MATLAB environment)A single segment is simulated but any number of EDs is sup-ported Due to its asynchronous nature we omit the CH-BScommunication (BTS length = 0) Each individual simulatedscenario involves aminimumof 1000 iterations In relation tocommunication errors a uniform distribution is consideredas explained in Section 46 Also such probability is indepen-dent among nodes increasing the likelihood of errors in thenetworkThe probability of communication channel issues isindependent of the probability of collisions better represent-ing a real scenario Also for the simulations involving 1minschedule CH is assumed to not sleep for practical reasonsSpecifically there is a significant energy cost associated withthe power activation of a node If the activationdeactivationcycles associated with a hibernating node are very frequent itis possible that the energy cost due to the transients is higherthan the hibernating savings

When not specified the parameters used in the simula-tions are the same ones used in our real-world implementa-tion of BETS also discussed in this section Such parameters


Table 4 Default parameters for simulations

ED MEAS time 1 s CH CTRL time 1 sED CTRL time 1 s ED MEAS timeout 3 sED CTRL timeout 2 s RandomTime1 5ndash11 sRandomTime2 5ndash12 min Measurements time 1 sMaxTimeStartLastED 5min SCchargingTime 1140 sExpectedNumChildren value of119873 tsLength 8 sMaxContinuousEM 1 h

are listed in Table 4 The parameter maxTimeStartLastEDrequires an explanationWhen convergence tests are realizedthe EDs are randomly turned on during a certain amountof time and maxTimeStartLastED represents a limit for thistime For energy-related calculations Table 2 is used and a36V 19Ah non-rechargeable battery is assumed as the singlesource of power Accordingly many results are given in termsof life expectancy for the node and one can simply convertback to energy values in Joules assuming that 100 of theinitial energy is actually used by the nodeThis assumption isrealistic if the power matching technique presented in [9] andsupported in this framework is used With such techniquethe current pulse effect discussed in Section 3 is avoided

Note that by assuming the usage of non-rechargeablebattery as the single power source it becomes straightforwardto verify if the solution would reach a very long lifetime inde-pendently of the existence of an energy scavenging systemfor the nodes Similarly such estimated lifetime considersthe LDC-only mode Therefore the energy costs associatedwith a more critical WSN application (RDC mode) are notcomputed In short both user-defined energy unknownsfrom the viewpoint of the framework the possible amount ofharvested energy and the energy spent in a demandingWSNapplication are removed in this analysis However in a real-world implementation such information naturally must beincluded according to the available energy resources andapplication demands In our current project the nodes arebeing deployed with non-rechargeable batteries and thesystem is currently LDC-only due to the low duty-cycle char-acteristics of the application In this case the estimated valuesfor lifetime provided in this section are exactly the ones usedin the energy analysis and decisions under this project

Related to the empirical investigation the BETS solutionhas been implemented in eight distinct outdoor networkssince August 2011 Six adjacent sites have a total of 150 nodes(110 already deployed) covering a 36 km times 36 km continuousarea The results presented here are from the two oldest net-works in terms of operation time The first network features1 BS 1 CH and 26 ED nodes The CH and EDs are attachedto 3 soil moisture sensors The maximum ED-CH distance isaround 150m The irregular topography and the existence ofobstacles (trees and plants) are the main challenges for thissite The second site a cow farm is composed of 21 EDnodes and the maximum ED-CH distance is around 350mPreviously we had faced problems in this site related to (a)solar panels versus cows (b) rechargeable-batteries versusextreme temperatures (c) complex support for unattended802154-based routers and (d) scalabilityoverhead issues

associated with the ZigBee protocol and sparse networks[2 4] Accordingly the BETS design was strongly influencedby the lessons of that work

52 Performance Evaluation In this section the experimen-tal results are discussed in relation to three aspects (a)convergence time (b) impact of communication errors and(c) impact of heterogeneous scheduling

Convergence Time In Figure 13 the convergence time is givenas a function of the number of nodes for different nodersquosschedules and a homogeneous scheduling is assumed Alsoonly communication errors due to the contention collisionsare considered The convergence value here represents theadditional time besides the schedule value for the lastED node to converge Because all the nodes are randomlyturned on such scenario represents the worst-case which isassociated with a stronger channel contention for the initialmoments For dual mode systems (RDCLDC) such conver-gence impact can be eliminated by providing a progressiveactivation of the nodes as already discussed

For clarity only the variance of the 20min schedule caseis presented As expected higher duty-cycles aggravate theconvergence For instance a segment with 30 nodes takesaround 15min to achieve convergence if 20min schedule isused However if a 5min schedule is defined for the nodesthe converge can take around 35min Similarly a highernumber of nodes impact the convergence time A segmentwith 20 nodes typically converges in less than 15min even if a5min schedule is in place On the other hand for the sameschedule it is possible that a network with more than 30

nodes only achieves convergence after hours It is importantto highlight that if CH randomly restarts a scenario simi-lar to this simulation will occur Therefore such convergenceanalysis provides important insights for the parametrizationof BETS in a given segment In our real-world deployments(20min-sched) we do not observe convergence time higherthan 1 h and the average value for convergence is found to bearound 40min Such value is close to the upper bound of thevariance line for 20min cycles in the figure Because this sim-ulation does not include any communication channel errorbesides the collisions the theoretical model has a fair agree-ment with our empirical evaluation Finally it is importantto highlight that a nonconvergent network does not implysignificant lack of functionality Even for thementioned casesinvolving 40min for convergence time typically the majorityof the nodes synchronize at the first MTS However it takesmore time to have all the nodes synchronized


10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)

1min schedule5min schedule10min schedule

20min schedule2h schedule

Empirical(20min schedule)

Figure 13 Convergence time assuming that all EDs are turned on randomly during a 5min period

Impact of Communication Errors Once the convergence at thenetwork segment is achieved BETS becomes highly deter-ministic and its theoretical network overhead is negligible(eg≪075) assuming low duty-cycles applications and nocommunication errors However for the next simulation wewant to verify if this design goal number number 3 also holdswhen the probability of communication errors is high (eg5) It is assumed that the network has already convergedand 30 EDs follow 20min cycles In Figure 14 the additionalnetwork overhead is given as a function of the channel errorprobability This overhead represents the additional numberofmessages due to the retransmissions and also possible addi-tional collisions These secondary collisions occur because anon-convergent node can transmit at the TS assigned to othernode As shown in the figure even a small error rate such as1 causes an overhead of 163 A higher error rate such as10 doubles the network trafficHowever whenwe calculatethe effective duty-cycle of BETS for this worst error case itturns out to be around 053 and the design goal number 3is shown to be satisfied

Nonetheless if the mentioned error rates are not tempo-rary a strong lifetime reduction for the nodes is expectedThis result highlights that the lack of extensive error controlin BETS has the clear trade-off of making the protocolvery sensible to communication channel errors The lifetimeimpact is slightly smaller than the additional traffic rate but itis still very high For our empirical investigation we analyzedthe data that arrived at the BS side Because it is possibleto easily detect duplicates missing data and the number ofretries that each node had experienced an accurate energyprofile for the nodes is feasible The results of such analysisare also shown in Figure 15 but now including the options of20 and 40 nodes in the segment It is clear that the mentionedtrade-off is aggravated with a higher number of nodes Forinstance consider the 4 error rate case the network with40 nodes has its life expectancy shortened by more than 1

01 05 075 1 2 3 4 100

25

50

75

100

Probability of errors at the communication channel (ED-CH link)

Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction

Lifetimereduction 85

Empirical (site no 2)lifetime reduction 464Empirical (site no 1)

lifetime reduction 34

lowast Simulation based on 20

scheduling with 30 ED nodeslowast Default hardware parameters (table 1)lowast Starting point network already converged

388 (64 to 39 yr)

min homogeneous

Figure 14 Impact of the communication channel error (ED-CHlink) on the BETS network performance

year compared with the case where the network only has20 nodes Therefore the effect of the channel error on theBETS performance is strongly aggravated when more nodesare involvedThis fact is explained by the best-effort approachused by BETS to achieve convergence the process is relativelyslow when the number of EDs is relatively high In otherwords it can take many minutes for a node that experiencedcommunication error to converge again and while in EMstate trying to contact CH it is wasting energy

Returning to the analysis of Figure 14 it also has empiricalresults to be discussed For sites number 1 and number 2 theaverage error rates of around 06 and 11 respectively arecalculated based on measured network metricsThese resultsare in agreement with the fact that site number 2 has veryhigh ED-CH distances Now using Figure 15 it is possible to


01 05 075 1 2 3 4 103

35

4

45

5

55

6

637

Probability of errors at the communication channel (ED-CH) link

Expe

cted

nod

e life

time (

year

s)

40 ED nodes30 ED nodes20 ED nodes

Baseline ideal network (no errors)

485reduction

Figure 15 Impact of the communication channel error (ED-CHlink) on the energy performance Similar scenario of Figure 14 fordifferent number of nodes

infer about the expected lifetime of the nodes For instancefor site number 2 (21 nodes 11 average error rate) itis possible to forecast a lifetime of around 4 years For anexisting WSN solution to be superior in terms of energyefficiency it needs to have a total network overhead smallerthan 18 (see Figure 7) To date site number 2 has been incontinuous operation for 24 months and so far the solutionis comparable to any existing solution that has an effectivenetwork overhead of less than 4 These results are prettysignificant considering the coverage area and the numberof nodes involved it is a sparse network and many WSNprotocols potentially would fail in such site [4]Moreover justa single CH node is being used and the energy consumptionamongEDnodes is guaranteed to be homogeneous assumingthat they have the same application schedule and the sameaverage communication error rate Finally although theenergy costs are significant with high communication errorrates our implementation of BETS proved to also have goodcommunication reliability the data losses for sites 1 and 2

are smaller than 14 and 18 respectivelyThese numbers arerelatively superior than the average for outdoors deploymentsreported in the WSN literature

Impact of Heterogeneous Scheduling So far the simulationsconsidered homogeneous scheduling However in somescenarios an adaptive sensing scheduling is desired [2 14] Inthe next simulation the energy performance of BETS underhomogeneous and heterogeneous scheduling is evaluated InFigure 16 the relative energy consumption of the CH nodefor a 24-hour period is shown as a function of differentscheduling schemes The reference case is a 20min scheduleinvolving 30 nodes The goal is to estimate how much energy(additional or reduction) is associated if the schedulingscheme changes The first 3 left most cases involve the same

Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds

lowastlowastlowastlowast

8 ED

s

lowast

lowastlowast

lowastlowastlowast


Homogeneous 20 min scheduleHomogeneous 10 min scheduleHomogeneous 10min and 20min scheduleHomogeneous 5 min schedule

Figure 16 CH energy consumption for homogeneous and hetero-geneous schedulings

20min schedule and 10 20 and 30 nodes Next 15 nodes ina 10min scheduleThe following case is special 20 nodes in aheterogeneous scheduling half following a 10min scheduleand half a 20min scheduleThe last case (rightmost) involvesonly 8 nodes but with a more frequent 5min schedule It isassumed a convergent and error-free segment

The number of ED MEASmessages received per hour bythe CH node is already calculated Note that in Figure 16 thesame number of ED MEASmessages received by CH in threedistinct scheduling schemes 90meash does not imply thesame energy consumption of the CH node The underlyingfactor that mainly governs the energy performance of theCH node is the number of AMs for a certain period oftime such as 24 hours If homogeneous scheduling is inplace the optimum scenario in relation to the number ofAMs is achieved However for heterogeneous scheduling thenumber of AMs can potentially increase and the calculationof the energy spent by the CH node is more complex asexplained next

(i) In eachMTS there is an extra time (dynamically valuedetermined by software) allocated at the ETS as aprovision for possible communications failures Thehigher the number of AMs is the higher the sumof these time provisions spent by CH and worse itsenergy efficiency are

(ii) Before the ETS beginning theCHnodemust be readyfor the radio reception and a certain amount of suchreadiness time is provided In our implementationsuch CPU time has an average of 30 s (it also performs


sensor measurements) and the corresponding con-sumed energy is relatively high In short the higherthe number of AMs the higher the energy spent withthis processing phase

(iii) Hardware state transition the higher is the numberof AMs the higher is also the energy cost due to thetransients while turning-on the modules [10] In par-ticular such cost is critical when power-gating tech-niques are used

The basic scenarios associated with an increase of thenumber of AMs are illustrated in Figure 16 An interestingaspect to highlight is related to the three leftmost cases usingthe same 20min schedule A linear relation exists in relationto the number of nodes and energy consumption Howeverthere is a hidden fixed energy cost in all cases and it is relatedto the establishment of AMs In this case the number of AMsis exactly the same (72 per day) and the mentioned fixed costis the sameThis analysis indicates that the higher the numberof EDs sharing the same AM the higher the CH energyefficiencyThenext two cases involve 15nodes (homogeneousscheduling) and 10 + 10 nodes (heterogeneous scheduling)The number of AMs involved in both cases is the same (144per day) The heterogeneous scheduling is not adding moreAMs because 20 is a multiple of 10 and all AMs related tothe nodes that follow the 20min schedule are also sharedwith the nodes that follow the 10min scheduleTherefore theheterogeneous scheduling is not imposing an additional costcomparing these two cases as proved by this simulation

Finally if we compare one of these two just mentionedcases with the one with 30 nodes the number of ED MEASmessages received by CH per hour is exactly the same (90)However the number of AMs for the case with 30 nodes issmaller (72 per day) and this fact explains why the CH isconsuming less energy compared to the other case In shortto save energy at the CH side both the number of EDs andthe number of AMs cannot be very high and the latter aspectis directly related to the employed scheduling scheme

The analysis of the results in this section strongly suggeststhat the LDCmode can be better exploited in many scenarioswhere the network is not continuously under high demandNonetheless the proposed framework based on a switchingRDCLDC scheme or simply on the LDC-only mode can stillbe enhanced by integrating it with state-of-the-art approachesinvolving energy efficiency for WSNs as will be discussednext

6 Future Directions

As already discussed in Section 2 the energy-managementefforts can be applied at different levels and the componentsof the proposed framework are implemented in all theselevels node network and at a centralized server Howeverthe framework can also be extended by beingmerged to otherstate-of-the-art efforts and this discussion is presented in thissection

In this work in particular for scenarios where the powerdepletion is a real risk the importance of using some sort

of backup energy reservoir in conjunction with an energy-harvesting system has been highlighted So far the focus hasbeen on the nontraditional use of primary cellsThemain rea-son for this choice is the combination of a high energy densityrobustness in relation to extreme temperatures lifetime andrelative low cost However besides this option other formsof backup energy storage have been reported For instancein [33] a survey on multisource energy harvesting systemsis provided and different combinations of energy storagecomponents are discussed including the recent fuel celltechnology In this context of hybrid energy system the samework highlights the advantages associated with the adoptionof the Smart Power Unit (SPU) [16] or System A architectureSuch energy system architecture has its own MCU and itcorresponds to the energy-management subsystem discussedin Section 32 shown in Figure 5

Still at the node level besides the mentioned effortsinvolving the hardware of the node it is also possible toincrease its energy efficiency by means of software actionsIn this track techniques such as data aggregation and datacompression can be investigated [34] In this case the maingoal is to exploit the data correlation among nodes or datasets in order to reduce the network traffic and ultimately theenergy spent by the nodes Inmany cases such techniques areassociated with penalties in terms of data quality that must beproperly evaluated as acceptable or not For instance basedon the compressive sensing (CS) theory it can be feasible torepresent sparse signals with fewer samples than required bytheNyquist sampling theorem [35] Accordingly the problemofmonitoring soilmoisture is studied following this approachwith excellent results [36] Similarly manyWSN applicationscan achieve better energy efficiency by adopting a proper CStechnique The bottom line in this discussion is that in orderto achieve a higher energy efficiency typically it is necessaryto sacrifice one or more QoS metrics for the WSN nodeIn our framework the focus is on acceptablemanageablepenalties in terms of network performance Similarly CS andrelated techniques are associated with different penalty levelsin terms of data quality Such conclusion is in accordancewiththe previous discussion about the energy effort tripod con-cept (Figure 2) where the importance of investigating howflexible the application requirements are also highlighted

Returning to the analysis of the proposed frameworksome future directions and open research aspects are thefollowing

Temporal Segmentation One can think that the RDCLDCswitching is an action that affects all the network as awhole However it is not necessarily the case and the DDCoperation can actually be applied only to a small part of thenetwork In this way a temporal and very efficient networksegmentation can be adopted assuming that a certain level ofnode redundancy is also in place For instance assume that itis detected that 15 of the nodes in a network are currentlyapproaching critical remaining energy levels By assigning theLDCmode only to this group of nodes it is possible to removethem from the regular node operation without losing controlof them In this example they can continue to regularly reporttheir state to a central application (eg every 6 hours) Note


Energy-management framework

Dual duty-cycle (DDC) system

LDC-only mode Dual mode operation

LDC

BETS protocol

Network segmentation2-tier CH and ED nodes

Scheduled sense-send-sleep

RDC

Long-term energy repository

Energy-management(control data server)

Primary cell

Distributed system inside a node

Intelligent sensorWake-up on radio (WOR)

Energy-management subsystem

Main MCU separated from radioWSN module

Figure 17 Proposed Energy-Framework

that this solution is significantly superior compared to themajority of existingWSN solutions because while the existingWSN application is preserved the lifetime of the nodes canbe realistically extended Moreover subsets of nodes can beperiodically selected to go to LDC mode

AutosegmentationWhile the previous idea ismainly based ona central server as the trigger agent to perform the RDCLDCswitching actions it is also possible to let such decision becommanded by the node itself For instance assume that anode detects that it is approaching a critical remaining energylevel While in RDC mode it can send a message to a centralserver warning its decision to switch to LDCmode and it canpotentially receive information about the existingCHnode tobe contacted and additional details related to the LDC setup(RF channel ID etc) In this way a virtual parallel network isautobuilt where the associated nodes continue to report theirsensing data or simply their health state in a relative low paceto a central server

Two Radio Transceivers One interesting way to implement aDDC node is the adoption of two radio transceivers one forthe LDCmode and the other for the RDCmode Such out-of-band solution has the advantage of easy implementation andit is highly recommended if the application requires frequentRDCLDC switching

7 Conclusions

An open energy-management framework for WSNs is pro-posed in this work with a strong emphasis at the realisticachievement of a functional and reliable solution with a verylong maintenance-free lifetime (eg gt5 years) for the nodesThe components of the proposed framework are shown inFigure 17 By means of a detailed and systematic preliminaryanalysis it is shown that energy scavenging systems aretypically necessary to achieve this goal However in order

to also increase the reliability of the solution a long-termenergy repository is recommendedThe traditional candidatefor this role is a rechargeable battery but considering thementioned very long-term lifetime a primary cell is a betterchoice Besides the selection of the proper energy resourcesfor the node it is also very important to control the energyused by the nodes and in the network in general An excellentstrategy to achieve an energy-efficient solution is to balancethe efforts at the node network and application levels Thelatter kind of effort is realized by means of an energy-management systemhosted in a central data server By energymeasurements received from the nodes by estimation tech-niques or both the centralized energy-management systemcan evaluate the current and future remaining energy atthe nodes and depending on the application QoS metrics itcan typically activate and deactivate sensing nodes based onlocation and time In this work the emphasis is given on theenergy-management efforts at node and network levels

It is proposed that multiple MCUs compose the sensornode that is a distributed system inside a node With thetechnological advances such proposal does not imply ahigher energy consumption but a better way to control indi-vidual modules of a node although adding complexity andcosts to theWSN project Inside this node architecture calleddual duty-cycle (DDC) node the traditionalWSN node is nomore considered the main MCU of the node but simply aradio transceiver module Moreover besides the mentionedmain MCU and radio transceiver modules the energy-management subsystem also has its own MCU In this con-text recent availability of intelligent sensor probes is alsodiscussed and this component can represent another MCUin such DDC node

Besides the focus on energy-related hardware aspects thiswork also provides guidelines related to the energy efforts atthe network level A dual duty-cycle (DDC) system is pro-posed as part of the framework From the viewpoint of aDDCsystem low duty-cycles (LDC) data-collection applications


are classified as belonging to the LDC-only class Howevermany WSN applications are not LDC-only and they areclassified as regular or high duty-cycle (RDC) applicationsIn order to achieve maximum energy efficiency the networkmust operate in LDC mode the majority of the time ifpossible LDC-only class of applications permanently satisfiesthis goal However for RDC applications it is still possible tohave the network operating temporarily in LDC mode andreturning back to RDC mode This is the case for instanceof indoors WSNs deployed in office buildings and operatingduring nonoffice hours In order to save energy during thisperiod of time the network can operate in a form of stand-bymode and this is actually what the LDC mode provides forthe network As expected the network eventually will returnto the RDC mode and the mechanisms of the DDC systemprovide such operational mode switching

While operating in LDC mode the nodes are actuallyrunning the code at themainMCU not the code at the legacyWSN nodes Such code includes the novel cross-layer proto-col called best-effort time-slot allocation (BETS) Under theBETS protocol the nodes assume a different logical topologyaccording to the BETS logical network segmentation andits 2-tier architecture By simulations and empirical resultsBETS is proved to be very energy efficient and typically hasan effective network overhead smaller than 1 Its maindrawback is a low data throughput which is the reason whythis solution is not applied for the RDC or regular mode ofoperation of many WSN applications Nonetheless becauseit is possible to switch between LDC and RDC modesmany current WSN platforms can still be enhanced with theproposed energy framework For instance when it is allowedto have the network temporarily operating with low datathroughput the LDCmode can be activated and the networkis then controlled by the BETS protocol In this way thenetwork can experience its best energy performance for theduration of the LDC period

Finally some ways to take advantage of the LDCmode innetworks that mainly operate in RDC mode are highlightedFor instance when a node approaches critical energy levelsit can autoswitch its operating mode to LDC thus reportingmeasurements (or health status) in a very low pace fashionwhile recovering from its energy depletion Nonetheless asidentified in this work the potential main drawback associ-ated with the RDCLDCmode switching is the need of someWSN applications of quickly resuming the hibernation modewhen a nonregular event occurs The key-component is anultra-low-power wake-up on radio (WOR) module for DDCnodes Although recent advances in theWORarea promise toaddress such need additional research effort is still necessaryin order to integrate WOR modules into DDC nodes

Acknowledgments

This work was performed at the University of SouthernCalifornia and at the University of Michigan under supportfrom the National Aeronautics and Space AdministrationEarth Science Technology Office and Advanced InformationSystems Technology Program Agnelo R Silva is also sup-ported by the Brazilian National Research Council (CNPq)

scholarship under the Brazilian Programme Science withoutborders

References

[1] X Jiang J Polastre and D Culler ldquoPerpetual environmentallypowered sensor networksrdquo in Proceedings of the 4th Interna-tional Symposium on Information Processing in Sensor Networks(IPSN rsquo05) pp 463ndash468 Los Angeles Calif USA April 2005

[2] M Moghaddam D Entekhabi Y Goykhman et al ldquoA wirelesssoil moisture smart sensor web using physics-based optimalcontrol concept and initial demonstrationsrdquo IEEE Journal ofSelected Topics in Applied Earth Observations and Remote Sens-ing vol 3 no 4 pp 522ndash535 2010

[3] M Moghaddam X Wu M Burgin et al ldquoGround networkdesign and dynamic operation for validation of spaceborne soilmoisture measurements Initial developments and resultsrdquo inProceedings of the Earth Science Technology Forum conference(ESTF rsquo10) June 2010

[4] A Silva M Liu andMMoghaddam ldquoRipple-2 a non-collabo-rative asynchronous and open architecture for highly-scalableand lowduty-cycleWSNsrdquo inProceedings of the 1st ACMAnnualInternational Workshop on Mission-Oriented Wireless SensorNetworking (MiSeNet rsquo12) pp 39ndash44 ACM New York NYUSA 2012

[5] J M Rabaey M J Ammer J L da Silva Jr D Patel and SRoundy ldquoPicoRadio supports ad hoc ultra-low power wirelessnetworkingrdquo IEEE Computer vol 33 no 7 pp 42ndash48 2000

[6] IEEE 802 15 4 Standard ldquoPart 15 4 wireless medium accesscontrol (MAC) and physical layer (PHY) specifications forlow-rate wireless personal area networks (LR-WPANs)rdquo IEEEPiscataway NJ USA 2006

[7] ZigBee Alliance ZigBee Specifications ZigBee Standard Orga-nization San Ramon Calif USA 2008

[8] A S Weddell N R Harris N M White et al ldquoAlternativeenergy sources for sensor nodes rationalized design for long-term deploymentrdquo in Proceedings of the IEEE InternationalInstrumentation and Measurement Technology Conference(IMTC rsquo08) pp 1370ndash1375 Victoria Canada May 2008

[9] A Silva M Liu and M Moghaddam ldquoPower-managementtechniques for wireless sensor networks and similar low-powercommunication devices based on nonrechargeable batteriesrdquoJournal of Computer Networks and Communications vol 2012Article ID 757291 10 pages 2012

[10] A G Ruzzelli P Cotan G M P OrsquoHare R Tynan and PJ M Havinga ldquoProtocol assessment issues in low duty cyclesensor networks the switching energyrdquo in Proceedings of theIEEE International Conference on Sensor Networks Ubiquitousand Trustworthy Computing (SUTC rsquo06) pp 136ndash143 TaichungTaiwan June 2006

[11] H Danneels V D Smedt C D Roover et al ldquoAn ultra-low-power batteryless microsystem for wireless sensor networksrdquoin Proceedings of the 26th European Conference on Solid-StateTransducers (EUROSENSOR rsquo12) Krakow Poland September2009

[12] EnerChip Smart Solid State Batteries httpwwwcymbetcomproductsenerchip-solid-state-batteriesphp

[13] RMisra andCMandal ldquoClusterHead rotation via domatic par-tition in self-organizing sensor networksrdquo in Proceedings of the2nd International Conference on Communication System Soft-ware and Middleware and Workshops (COMSWARE rsquo07) Ban-galore India January 2007


[14] J C Lim andC Bleakley ldquoAdaptiveWSNscheduling for lifetimeextension in environmental monitoring applicationsrdquo Interna-tional Journal of Distributed Sensor Networks vol 2012 ArticleID 286981 17 pages 2012

[15] D Zenobio K Steenhaut M Celidonio E Sergio and Y Ver-belen ldquoA self-powered wireless sensor for watergas meteringsystemsrdquo in Proceedings of the IEEE International Workshopon Energy Harvesting for Communication pp 5772ndash5776 June2012

[16] M Magno S Marinkovic D Brunelli E Popovici B OrsquoFlynnand L Benini ldquoSmart power unit with ultra low power radiotrigger capabilities for wireless sensor networksrdquo in Proceedingsof the Design Automation and Test in Europe Conference (DATErsquo12) Dresden Germany 2012

[17] S J Marinkovic and E M Popovici ldquoNano-power wirelesswake-up receiver with serial peripheral interfacerdquo IEEE Journalon Selected Areas in Communications vol 29 no 8 pp 1641ndash1647 2011

[18] Atmel Corp ldquoSleepwalking helps conserve energyrdquo httpatmelcorporationwordpresscom20130416sleepwalking-helps-conserve-energy

[19] P Dutta D Culler and S Shenker ldquoProcrastination might leadto a longer and more useful liferdquo the 6th Workshop on HotTopics in Networks (HotNets-VI rsquo07) 2007

[20] K Lu Y Qian D Rodriguez W Rivera and M RodriguezldquoWireless sensor networks for environmental monitoring appli-cations a design frameworkrdquo in Proceedings of the 50th AnnualIEEEGlobal Telecommunications Conference (GLOBECOM rsquo07)pp 1108ndash1112 Washington DC USA November 2007

[21] G Halkes MAC Protocols for Wireless Sensor Networks andTheir Evaluation 2009

[22] S FarahaniZigBeeWireless Networks and Transceivers ElsevierOxford UK 2008

[23] ldquoCc2531 usb evaluation module kitrdquo httpwwwticomtoolcc2531emk

[24] C Buratti A Conti D Dardari and R Verdone ldquoAn overviewonwireless sensor networks technology and evolutionrdquo Sensorsvol 9 no 9 pp 6869ndash6896 2009

[25] H R Bogena J A Huismana H Meierb U Rosenbauma andA Weuthena ldquoHybrid wireless underground sensor networksquantification of signal attenuation in soilrdquo Vadose Zone Jour-nal vol 8 no 3 pp 755ndash761 2009

[26] Z G Kovacs G E Marosy and G Horvath ldquoCase study of asimple low powerWSN implementation for forestmonitoringrdquoin Proceedings of the 12th IEEE Biennial Baltic Electronics Con-ference (BEC rsquo10) pp 161ndash164 Tallinn Estonia October 2010

[27] L Selavo A Wood Q Cao et al ldquoLUSTER wireless sensornetwork for environmental researchrdquo in Proceedings of the 5thACM International Conference on Embedded Networked SensorSystems (SenSys rsquo07) pp 103ndash116 November 2007

[28] M A Pasha S Derrien and O Sentieys ldquoToward ultra low-power hardware specialization of a wireless sensor networknoderdquo in Proceedings of the IEEE 13th International MultitopicConference (INMIC rsquo09) pp 1ndash6 Islamabad Pakistan Decem-ber 2009

[29] V Rajendran K Obraczka and J J Garcia-Luna-AcevesldquoEnergy-efficient collision-free medium access control forwireless sensor networksrdquo in Proceedings of the 1st InternationalConference onEmbeddedNetworked Sensor Systems (SenSys rsquo03)pp 181ndash192 ACM Press Los Angeles Calif USA November2003

[30] T Zheng S Radhakrishnan andV Sarangan ldquoPMAC an adap-tive energy-efficient MAC protocol for wireless sensor net-worksrdquo in Proceedings of the 19th IEEE International Paralleland Distributed Processing Symposium (IPDPS rsquo05) pp 65ndash72Denver Colo USA April 2005

[31] I Rhee AWarrier M Aia J Min andM L Sichitiu ldquoZ-MACa hybrid MAC for wireless sensor networksrdquo IEEEACM Trans-actions on Networking vol 16 no 3 pp 511ndash524 2008

[32] S Mehta and K Kwak ldquoH-MAC a hybrid MAC protocol forwireless sensor networksrdquo International Journal of ComputerNetworks amp Communications vol 2 no 2 Article ID 10871172010

[33] A Weddell M Magno G Merrett D Brunelli B Al-Hashimiand L Benini ldquoA survey of multi-source energy harvestingsystemsrdquo in Proceedings of the Design Automation and Test inEurope Conference (DATE rsquo13) pp 905ndash910 2013

[34] I F Akyildiz W Su Y Sankarasubramaniam and E CayircildquoWireless sensor networks a surveyrdquo Computer Networks vol38 no 4 pp 393ndash422 2002

[35] W Bajwa J Haupt A Sayeed and R Nowak ldquoCompressivewireless sensingrdquo in Proceedings of the 5th International Confer-ence on Information Processing in Sensor Networks (IPSN rsquo06)pp 134ndash142 Nashville Tenn USA April 2006

[36] X Wu andM Liu ldquoIn-situ soil moisture sensing measurementscheduling and estimation using compressive sensingrdquo in Pro-ceedings of the 11th ACMIEEE Conference on Information Pro-cessing in Sensing Networks (IPSN rsquo12) pp 1ndash11 Beijing ChinaApril 2012

Submit your manuscripts athttpwwwhindawicom

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2013Part I


DistributedSensor Networks


ISRN Signal Processing



Mechanical Engineering

Advances in

Modelling amp Simulation in EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2013

Advances inOptoElectronics


Volume 2013

ISRN Sensor Networks


VLSI Design


Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawi Publishing Corporation httpwwwhindawicom Volume 2013

The Scientific World Journal

ISRN Robotics



Antennas andPropagation


ISRN Electronics



thinspJournalthinspofthinsp

Sensors


Active and Passive Electronic Components

Chemical EngineeringInternational Journal of



Electrical and Computer Engineering

Journal of

ISRN Civil Engineering


Advances inAcoustics ampVibration



intervention is expected due to power depletion of a nodeIt will be shown that to achieve this goal the complexitylevel of the solution at the design-time is relatively highAlso the initial cost of a node is expected to be at least 30higher than a traditional off-the-shelf node However thepractical functionality and the total ownership cost (TCO)of the system at long term can be very attractive Notethat the advocated vision in this work diverges from thetraditional concept that a WSN comprises hundreds of verylow-cost nodes each one individually with a relatively highprobability of failure On the contrary the focus of this workis on the achievement of a very high-quality and controlledsolution

The paper begins with the presentation of the energyeffort tripod and energy control loop concepts in Section 2It is highlighted that effective energy savings for WSN nodespotentially depends on a balanced solution in terms of hard-ware network and application demands In Section 3 thefoundations of the proposed framework are discussed (a) theoptional (but recommended) use of primary cells associatedto harvesting systems (b) the advantages of a distributedsystem inside a node and (c) the adoption of the dual duty-cycle operation (DDC) for WSN nodes The core part of theproposed framework is a cross-layer network protocol whichis presented in Section 4 This protocol is implemented as anapplication-layer overlay on top of existing WSN solutionsSuch overlay mechanism can be dynamically activated anddeactivated in order to allow the network to achieve thebest performancewhile satisfying existing energy constraintsMany of the components of the proposed framework are infact part of a long-term and ongoing project involving oneof the largest outdoors WSN deployments still in operation[2ndash4] The field results of this project in conjunction withsimulated outcomes are reported in Section 5 A discussionrelated to the integration of the framework with otherongoing WSN research efforts is provided in Section 6 andthe paper is concluded in Section 7

2 Energy Management in WSNs

In this section typical pitfalls and challenges in the design ofenergy systems for WSNs are discussed Next the importantenergy effort tripod and energy control loop concepts areintroduced

21 Design Challenges and Pitfalls Many well-designed pro-jects fail due to small details and incorrect (but generallyaccepted) assumptions Therefore before presenting the pro-posed framework it is important to highlight some aspectsassociated with the current state-of-the-art technology onenergy harvesting systems for WSNs

WSNDesign besides Long Lifetime and Reliability One criticalpitfall associated to the energy aspect of WSN designs is toperpetuate the original vision of a WSN with hundreds tothousands of very cheap nodes [5] where a high rate of nodefailures is actually expected Although such vision can stillcorrespond to the needs of some applications a quick inves-tigation at the current WSN deployments around the world

reveals a different trend for WSNs For instance few existinglong-term networks actually have more than 50 nodes Moreimpressive is the ongoing success of the infrastructure-based WSN solutions (star or tree topologies) such as theones based on IEEE 802154ZigBee [6 7] A higher nodereliability is typically crucial when WSNs move from ad hocto infrastructured architectures Accordingly to this currenttrend a significant emphasis in this work is given to thereliability of the nodes In this context the term reliability isassociated with the goal of having nodes that rarely becomeunavailable due to a noncontrolled power depletion

Energy Scavenging Does Not Imply a Perpetual LifetimeAnother pitfall associated to the design ofWSNs is to considerthat the adoption of an energy scavenging system is automat-ically associated to an endless node lifetime Besides the needto consider the life expectancy of the sensing componentssuch as a humidity sensor or a soil moisture probe typicalpower systems can rarely achieve a 5-year lifetime due toa plurality of reasons discussed in this section Thereforethe first step toward a successful low-cost WSN solution (interms of functionality + reliability + long lifetime) is theinvestigation of the expected lifetime of each of the compo-nents of an node In general the reported premature deathcause of WSN nodes is the energy subsystem in particularthe batteries Primary (nonrechargeable) batteries typicallyhave a very short lifetime in WSNs [8 9] However it isalso important to have inmind that secondary (rechargeable)batteries potentially have a lifetime smaller than 2-3 years

Possibility of Adopting a Nonrealistic Energy Model A signifi-cant number of WSN papers present three regular omissionsor inaccuracies regarding to the way the energy model isproposed or adopted First the transients such as due tothe activationdeactivation of a radio transceiver are typicallyneglected as pointed out in [10] Second many values used asinput parameters for the models are directly imported fromthe datasheets of the components without further considera-tion of the effects of integrating these components togetherFor instance based on its datasheet a radio transceiver mod-ule has a nominal sleeping current of 10 120583A However it isobserved that once it is attached to a MCU leak currents aredetectedwhich aremany folds higher than the nominal sleep-ing current Similarly a voltage regulator can be included inthe design of a WSN node in particular when energy har-vesters are also employed However many reported energymodels do not include the energy cost of a possible voltageregulator As a result an energy model that disregards theexistence of voltage conditioners can be drastically distorted(non-linearly) when it is adopted in a node The severity ofthis statement can be illustrated by the following real-caseWhen both MCU and radio modules are sleeping usually itis not possible to put the voltage regulator in sleeping mode(typically shutdownmode in the context of regulators)There-fore rather than an expected amount in 120583W as the powerconsumption for the node (as expressed inmanyWSNenergymodels) the node can potentially have an effective sleepingconsumption on the order of mW Therefore as one reducesthe application duty-cycle the adopted energy model reveals


29

282726252423


2 N

iMH

bat

terie

s

Batte

ry le

vel (

V)

Sampling time

(a)

Batte

ry le

vel (

V)

36

34

32

30


1Li-S

OCl

2ba

ttery

(b)














NetworkApplication

Hardware






Decision

Impact

Data

Data


Energy state

Node operation












Command

Measurement values


MPPT control



(a)

Command





Radio moduleonoff


(b)

Command

Application data


Radio moduleonoff



(c)


















BeaconTX





Power scavenging


source (s)



Energy-management

Energy harvester

controller






















Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH



To data server



















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of






29

282726252423


2 N

iMH

bat

terie

s

Batte

ry le

vel (

V)

Sampling time

(a)

Batte

ry le

vel (

V)

36

34

32

30


1Li-S

OCl

2ba

ttery

(b)














NetworkApplication

Hardware






Decision

Impact

Data

Data


Energy state

Node operation












Command

Measurement values


MPPT control



(a)

Command





Radio moduleonoff


(b)

Command

Application data


Radio moduleonoff



(c)


















BeaconTX





Power scavenging


source (s)



Energy-management

Energy harvester

controller






















Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH



To data server



















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of











NetworkApplication

Hardware






Decision

Impact

Data

Data


Energy state

Node operation












Command

Measurement values


MPPT control



(a)

Command





Radio moduleonoff


(b)

Command

Application data


Radio moduleonoff



(c)


















BeaconTX





Power scavenging


source (s)



Energy-management

Energy harvester

controller






















Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH



To data server



















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of






Decision

Impact

Data

Data


Energy state

Node operation












Command

Measurement values


MPPT control



(a)

Command





Radio moduleonoff


(b)

Command

Application data


Radio moduleonoff



(c)


















BeaconTX





Power scavenging


source (s)



Energy-management

Energy harvester

controller






















Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH



To data server



















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of






Command

Measurement values


MPPT control



(a)

Command





Radio moduleonoff


(b)

Command

Application data


Radio moduleonoff



(c)


















BeaconTX





Power scavenging


source (s)



Energy-management

Energy harvester

controller






















Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH



To data server



















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of

















BeaconTX





Power scavenging


source (s)



Energy-management

Energy harvester

controller






















Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH



To data server



















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of





















Sink

To data server

Dual modeswitching

ED

ED

ED ED ED ED

EDED

ED ED ED

EDCH CH



To data server



















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of















10 5 2 10750

1

2

3

4

5

6

7

8



36

V 1

9Ah

batte

ry li

fetim

e (ye

ars)











DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of






DS

DS

Wi-Fi

BSDS

3G

DSBS

BSBS

CH

CH

CHCH

ED

CH CH

CH

CH

ED ED

ED

ED

SMS

Tier

-1Ti

er-2

900MHz 900MHz

24GHz

24GHz























Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of



















Design goal

Scalability



Design approach






















Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of







Term Explanation





































Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of















Delay tolerance

Periodic sensing


Appl

icat

ion

Net

wor

k

Implements




BETSprotocol



Appl

icat

ion

Har

dwar

ePower hibernation

Very low duty-cycle


control

Assumptions












ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of






ED1 node

Measurements

Measurements

Timeline

ED2 nodewakes up

wakes upED1 node

sleeps

ED2 nodesleeps

CH nodewaits next



















ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of






ETS ETS

ActiveMTS

ActiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

InactiveMTS

BTS BTS BTS BTSSTS

TS TS TS TS

Measurements

TS TS

ETS ETSSTS STS

ActiveMTS

ActiveMTS
























































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of



















































119905119903119910119900119906119905119904 larr 0

if 119888119906119903119903119879119894119898119890 gt 119898119886119909119862119900119899119905119894119899119906119900119906119904119864119872 then119877119890119887119900119900119905 119899119900119889119890


119888ℎ119886119903119892119890 119904119906119901119890119903119888119886119901119886119888119894119905119900119903(119904) for 119878119862119888ℎ119886119903119892119894119899119892119879119894119898119890

end ifif 119905119903119910119900119906119905119904 gt 5 then

119905119903119910119900119906119905119904 larr 0




end if





119908119886119894119905119864119863119879119894119898119890 larr 20 lowast 119898119905119904119871119890119899119892119905ℎ


119908119886119894119905119864119863119879119894119898119890 larr 31 lowast 119905119904119871119890119899119892119905ℎ





end ifend if

end if




















119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of























119886119906119909 larr 119891119897119900119900119903(119888119906119903119903119872119879119878119909)

119899119890119909119905119872119879119878 larr (119886119906119909 + 1) lowast 119909

119906119901119889119886119905119890119872119879119878 119879119886119887119897119890(119864119863 119894119889 119899119890119909119905119872119879119878)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119891119894119909119890119889119863119890119897119905119886

119899119906119898119864119863119904119878119886119898119890119872119879119878 larr 119892119890119905119873119906119898119864119899119905119903119894119890119904119878119886119898119890119872119879119878(119872119879119878 119879119886119887119897119890 119864119863 119894119889)


119886119889119895119878119888ℎ larr 119886119889119895119878119888ℎ + 119890119905119904119860119889119895


























10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of

















10 20 30 40 500

15

30

45

60

75

Number of ED nodes

Tim

e to

achi

eve f

ull c

onve

rgen

ce (m

in)







01 05 075 1 2 3 4 100

25

50

75

100


Addi

tiona

l net

wor

k tr

affic (

)

Tota

l app

licat

ion

+ BE

TS d

uty-

cycle

05

3

2883

163116

262

411

523

1118

Lifetime reduction






388 (64 to 39 yr)

min homogeneous





01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of






01 05 075 1 2 3 4 103

35

4

45

5

55

6

637


Expe

cted

nod

e life

time (

year

s)



485reduction





Relat

ive e

nerg

y co

nsum

ptio

n (

)

30

mea

sh

60

mea

sh

90

mea

sh

90

mea

sh

90

mea

sh

96

mea

sh

minus548

minus274

REF

+178

+588

lowast10

ED

s

lowast20

ED

s

lowast30

ED

s

lowastlowast

15 E

Ds

lowastlowastlowast

10+

10 E

Ds


8 ED

s

lowast

lowastlowast

lowastlowastlowast














6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of











6 Future Directions











LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of









LDC

BETS protocol



RDC



Primary cell









7 Conclusions









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of









Acknowledgments



References








































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of































Journal of



RotatingMachinery


Volume 2013Part I








Advances in




Volume 2013



VLSI Design




ISRN Robotics





ISRN Electronics




Sensors







Journal of





An Adaptive Energy-Management Framework for Sensor Nodes ...soilscape.usc.edu/~agnelors/An Adaptive...

Documents

Transcript of An Adaptive Energy-Management Framework for Sensor Nodes ...soilscape.usc.edu/~agnelors/An Adaptive...