Power Control for Mobile Sensor Networks: AnExperimental Approach

JeongGil KoDepartment of Computer Science

Johns Hopkins UniversityBaltimore, MD

Email:jgko@cs.jhu.edu

Andreas TerzisDepartment of Computer Science

Johns Hopkins UniversityBaltimore, MD

Email:terzis@cs.jhu.edu

Abstract—Techniques for controlling the transmission powerof wireless mobile devices have been widely studied in ad-hoc andcellular networks. However, as mobile applications for wirelesssensor networks (WSNs) emerge, the unique characteristicsofthese networks, such as severe resource constraints, suggest thattransmission power control should be revisited from a WSNperspective. In this work, we take an experimental approachtoexamine the effectiveness of power control for WSN applicationsthat involve mobility at human walking speeds. Furthermore, wepropose two light-weight transmission power control schemes toimprove energy efficiency and spatial reuse. The first is an activeprobing based scheme that adjusts transmission power basedon(the lack of) packet losses and applies to all low-power radios. Onthe other hand, the second scheme requires radios that offerlinkquality indicators (LQI) to estimate the proximity between thetransmitter and receiver. We evaluate both schemes using mobilenodes in an indoor and an outdoor environment. Our resultsshow that the energy efficiency of the proposed transmissionpower control schemes can be very close to that of the optimaloffline strategy. Moreover, our schemes significantly reduce theinterference to unintended receivers and improve spatial reuse.To our knowledge, this is the first work that evaluates the effectof transmission power control in mobile WSNs.

I. I NTRODUCTION

Most existing wireless sensor network (WSN) applicationssuch as environmental [12] and structural monitoring [13],involve networks of static sensors. Nevertheless, researchershave recently began to consider applications such as residentialhealth monitoring [19], in-hospital patient monitoring [9], andsports monitoring [2] that require mobile sensing [1].

While MANETs and cellular networks already deal withnode mobility, WSNs introduce novel resource constraints.For example, while mobile devices in MANETs and cellularnetworks are considered to be always active or have looseenergy constraints –GSM devices for example have cycles of∼12% [3]– WSN nodes aggressively duty cycle their radios toconserve energy. These differences suggest the need to studyschemes that improve mobile nodes’ lifetime and networkgoodput from a WSN perspective.

Dutta and Culler proposed mechanisms to reduce the energyusage of mobile WSNs by reducing the nodes’ idle listeningtimes [5]. Instead, we are interested in reducing energy con-sumption by controlling the radio transmission power. Doingso also controls the transmitters’ interference range. Therefore,transmission power control also has the potential to increase

the spatial reuse of the wireless medium. Minimizing interfer-ence is especially important in WSNs given that many systemsoperate in low power modes using protocols such as low powerlistening [14] that use the existence of energy on the wirelesschannel to activate a node’s radio. Thereby, controlling theinterference to unintended receivers can further reduce the idlelistening times of WSN devices as well.

We begin this work by performing an empirical study thatquantifies the potential benefits of transmission power con-trol for WSNs with mobile nodes moving at human walkingspeeds in both outdoor and indoor environments. These initialexperiments show that controlling the transmission power formobile WSNs, can decrease the current draw due to packettransmissions by as much as 49.3% and also decrease thepacket interference to unintended receivers by up to 88.7%.Wealso examine the effectiveness and limitations of instantaneouslink quality indicators such as the received signal strength indi-cator (RSSI) for low-power wireless radios. We learn from ourexperiments that in mobile settings, it is difficult to estimatea mobile node’s target transmission power using RSSI valuesdue to the high variance of RSSI in mobile settings.

We leverage these findings to design a light-weight adaptivetransmission power control scheme based on active probingthat adjusts transmission power based on (the absence of)packet losses. Specifically, the proposed scheme reacts quicklyto packet losses by increasing transmission power. Conversely,after everyN consecutive packets that are successfully trans-mitted, the node reduces its transmission power to the nextlowest level. The scheme is fast enough to be responsiveto the dynamic link conditions that mobile devices observe.Moreover, we propose an enhanced scheme for radios thatindicate the level of corruption for received packets. Radiosthat implement the IEEE 802.15.4 standard [8] and reportLQI values are one such example. This enhanced schemedetermines whether the distance between the transmitter andreceiver is close enough to immediately reduce the transmis-sion power to a pre-specified power level.

We evaluate both schemes using mobile devices in the samepair of environments. The results indicate that both the basicand LQI-enhanced schemes effectively lower energy consump-tion caused by data transmissions, reducing the current drawto only 4.9% more than the offline optimal. Additionally, our

experiments show that the proposed schemes can reduce thelevel of interference at unintended receivers by 81.8%. Theseresults validate that our proposed schemes can significantlyincrease spatial reuse in the wireless medium.

The rest of the paper is organized as follows. In Section IIwe frame our contributions in the context of related work.Next, we present the results of an empirical study designed toquantify the potential benefits of transmission power controlfor mobile nodes in Section III. We describe our adaptiveschemes for transmission power control in Section IV andevaluate them in Section V. The paper concludes with asummary in Section VI.

II. RELATED WORK

While some previous work in wireless sensor networksconsidered controlling the transmission power of resourcecon-strained motes, it mostly focuses on controlling and managingthe topology of networks of stationary nodes [7], [10], [16].The schemes proposed by Lin et al. and Son et al. rely ongathering extensive information about the channel environmentprior to deciding the transmission power [10], [16]. However,considering that channel conditions for mobile nodes changefrequently, such approaches do not apply to mobile WSNs.Specifically, Lin et al. find a one-to-one correlation betweenthe packet reception ratio (PRR) and instantaneous link qualityindicators (e.g., RSSI, LQI) to determine transmission powersfor stationary nodes [10]. We show later in Section III-B thatin dynamic channel environments, instantaneous link qualityindicators by themselves cannot provide sufficient informationto infer the appropriate transmission power level.

Hackman et al. used a fixed size window to compute theshort term PRR and infer the transmission power based onthose channel estimates [7]. While accurate channel estima-tions can be used to determine precise transmission powerlevels, the delays and unsuccessful transmissions during theestimation process can degrade the lifetime of mobile WSNsthat operate under dynamic channel environments.

Many different schemes exist to determine the transmis-sion power for mobile devices in MANETs and cellular net-works [4], [6], [11], [18], [20]. These schemes mostly usesignal strength related metrics (e.g., signal to noise ratio (SNR)or signal to interference ratio (SIR)) computed over incomingpackets and compare the resulting values to static or dynamicthresholds to determine a mobile node’s transmission power.The results in Section III-B show that the high variance of sig-nal strength measurements make the inference of transmissionpower levels from such measurements practically infeasible.

Considering the difficulties in estimating the noise andinterference levels of the receiver at the transmitter, manysignal strength-based schemes use RTS/CTS packets (802.11-based schemes) or explicit feedback packets to exchange trans-mission power information, forming a closed-loop between thereceiver and the transmitter [4], [11]. Such feedback packetsincrease the channel load and idle listening times at mobilenodes thereby increasing energy consumption [5].

Power Level Output Power (dBm) Current Draw (mA)

31 0 17.4

27 -1 16.5

23 -3 15.2

19 -5 13.9

15 -7 12.5

11 -10 11.2

7 -15 9.9

3 -25 8.5

TABLE ITRANSMISSION POWER AND CORRESPONDING CURRENT DRAW FOR THE

IEEE 802.15.4COMPLIANT CC2420RADIO USED BY THE POPULAR

TELOSB AND M ICAZ MOTES (REPLICATED FROM[17]).

Fig. 1. Pictorial overview of the outdoor and indoor testingenvironments.Yellow lines indicate the path that the mobile node follows,traveling at aconstant speed from one end of the path to the other and back. The pathlength is indicated in each figure.

Several open-loop schemes have been proposed to addressthe inefficiency of closed loop systems (see [18] and refer-ences therein). These schemes mostly use techniques such ascomplex filters to eliminate the narrowband interference andadjacent cell interference and therefore require complex radiodesigns that are not appropriate for low-cost wireless devices.Moreover, these approaches can create positive feedback loopsamong the mobile nodes [18]. Such loops arise when onenode increases its transmission power causing the interferencelevels to rise at its neighbors. In response, other nodes alsoincrease their transmission power levels. While the schemeswe propose can also create feedback loops, because theyuse packet delivery rates instead of the highly-variable signalstrength measurements we expect them to be more stable.

Last but not least, while the previously proposed schemesevaluate the energy and capacity benefits in detail, the resultsare based on mathematical analyses or simulations. To ourknowledge, our work is the first attempt to perform an em-pirical study on the effect of transmission power control forlow-power mobile nodes and evaluate them with such devices.

III. E MPIRICAL STUDY

Next, we quantify the potential benefits of transmissionpower control for mobile WSNs and test the efficacy of re-ceived signal strength measurements as a metric for controllingtransmission power for low-power radios.

0 5

10 15 20 25 30 35

0 30 60 90 120 150

Tra

nsm

issi

on P

ower

Walk Distance (m)

Outdoor

0 5

10 15 20 25 30 35

0 30 60 90 120

Tra

nsm

issi

on P

ower

Walk Distance (m)

Indoor

Fig. 2. Packet reception plots at a stationary unintended receiver for outdoor(left) and indoor (right) environments. Each dot represents a successfullyoverheard packet. The walk distance on the x-axis indicatesthe total distancethat the mobile transmitter covers as it moves away from the unintendedreceiver and back. Controlling the mobile node’s transmission power reducesthe number of packets received at theunintendedreceiver and thus can limitthe mobile source’s interference range.

A. Benefits of controlling transmission power

We investigate how transmission power control can improveenergy efficiency and reduce interference range. In terms ofenergy efficiency, Table I indicates that controlling the trans-mission power level can decrease the radio’s current draw byup to 51.7% for the popular TI CC2420 radio.

We quantify the impact of power control on reducing inter-ference range through an experiment. Specifically, we considera mobile node that moves at human walking speed (i.e.,∼5 kilometers/hour or∼1.4 meters/second) and broadcastsone packet every 256 ms. A stationary node records everysuccessfully received packet along with its reception time. Themobile node moves along a fixed path, indicated by the yellowline in Figure 1, first away from the stationary node and thentowards it. We selected this data rate to ensure that statisticallyadequate numbers of packets arrived at the stationary node.

We assume that the stationary node isnot the destination ofthe mobile node’s packets, in other words the stationary nodeis anunintendedreceiver. We show experimental results with asingle unintended receiver to empirically measure the amountof interference in a controlled setting. The test is repeated forall eight power levels in Table I. Figure 2 plots the packetreception at the unintended receiver for each power level withrespect to the mobile node’s traveled distance. It is evidentthat transmission power level significantly alters the number ofpackets received and can therefore be used to increase spatialreuse of the wireless medium.

Our work targets mobile nodes that duty cycle their radios.Since idle listening consumes considerable energy in low-power radios [5], nodes that keep their radios constantly onwill benefit less from transmission power control. Even in thiscase, controlling transmission power is beneficial becauseitreduces interference range.

B. Efficacy of signal strength measurements on power levelestimation

The use of RSSI measurements to determine transmissionpower levels for WSNs has been proposed in theliterature [10], [16]. The goal of those schemes is tomaintain a constant signal strength level at the receiver by

-95-90-85-80-75-70-65-60-55

-25 -20 -15 -10 -5 0

RS

SI (

dBm

)

Output Power (dBm)

Dynamic Environment

-95-90-85-80-75-70-65-60-55

-25 -20 -15 -10 -5 0

RS

SI (

dBm

)

Output Power (dBm)

Static Environment

Fig. 3. Average and 95% confidence intervals of RSSI values ofpacketstransmitted at different power levels. The transmitter andreceiver locations arefixed in this experiment. RSSI variance increases significantly in the dynamicenvironment, complicating the use of RSSI measurements forcontrollingtransmission power.

-100-90-80-70-60-50-40-30

0 30 60 90 120 150

RS

SI (

dBm

)

Walk Distance (m)

Outdoor

-100-90-80-70-60-50-40-30

0 30 60 90 120

RS

SI (

dBm

)

Walk Distance (m)

Indoor

Fig. 4. Packet RSSI values collected in the outdoor (left) and indoor (right)environments with a mobile node sending packets at maximum power to astationary receiver. The receiver is positioned at the beginning of the path,while the mobile node travels away from the receiver and back(i.e., walkdistance 150 m in the outdoor environment corresponds to thetwo nodesbeing next to each other again). Instantaneous RSSI variation is high whenone of the devices is mobile.

informing the transmitter of the latest RSSI measurementswhich then adjusts its transmission power accordingly.

We performed an experiment to explore the applicabilityof these methods in the two environments of interest. Thetransmitter and the receiver in this experiment were stationedfive meters apart from each other in our building’s hallway.The transmitter broadcasted a packet every 256 ms using eachof the available power levels, while the receiver recorded theRSSI values of all received packets. Figure 3 plots the meanand the 95% confidence intervals of RSSI values collectedfrom packets transmitted at different power levels. We includeresults from a dynamic environment in which people consis-tently traveled in the hallway (left) and a static environmentwhere there was no movement (right). It is evident that evenwhen the nodes are static, changes in the environment inducelarge variations in RSSI values. In turn, these variations com-plicate the accurate and prompt estimation of the transmissionpower level that will achieve the receiver’s desired signalstrength levels.

While the previous experiment shows how RSSI values varyeven when both nodes are static, we now present how a node’smobility affects the RSSI trends. To do so, we performed anexperiment in which the transmitter performed a round-tripwalk on the path indicated by the yellow lines in Figure 1. Thetransmitter sent one packet every 256 ms at maximum power(i.e., 0 dBm), while a receiver was placed at the beginningof the mobile node’s path to collect packet RSSI values.One can notice from Figure 4 that RSSI measurements varysignificantly even for consecutive packets, in accordance with

Fig. 5. State diagram of the active probing scheme for selecting transmissionpower levels. Each state corresponds to one of the availabletransmissionpower levels for the CC2420 radio. Transmissions successesand failures aredetermined using acknowledgments.

what is predicted by fast fading models [15]. As before, thesevariations complicate the estimation of a mobile node’s targettransmission power using RSSI values. This is especially truesince, in the majority of the cases, the range between thehighest and lowest RSSI values is 15 dBm (-80 dBm to -95 dBm), comparable to the range of RSSI values recorded ata single location (see Fig.3).

IV. A DAPTIVE TRANSMISSION POWER CONTROL

A. Base Scheme

The goal of transmission power control is to reliably deliverpackets with minimal energy consumption and interference.Moreover, the proposed power control scheme has the follow-ing additional goals:(1) It must quickly adapt to link qualityvariations caused by mobility or changes in the environment.(2) For the same reason, knowledge of past link conditionsmust be frequently purged to avoid polluting current trans-mission power estimates.(3) Each node should independentlydetermine its transmission power levels.(4) Finally, giventhe nodes’ resource constraints, the scheme should be bothmemory and energy efficient.

While using the received signal strength is an intuitiveapproach, the results from Section III suggest that the temporaland spatial variability of signal strength measurements signifi-cantly diminish its practical use. Instead, we propose anactiveprobing scheme to determine a mobile node’s appropriatetransmission power level.

We consider applications with two different traffic patterns.In the first case, mobile nodes generate periodic streams ofdata traffic. An example of an application in this category iscontinuous vitals signs monitoring of in-hospital patients [9].The second category includes applications that generate infre-quent bursts of data traffic such as activity monitoring.

Given that mobile nodes move at relatively low speedscompared to their data generation rate, the transmission powerlevel used for the previous successful transmission is a rea-sonable estimate for the transmission power of the currentpacket. However, if more thannlower consecutive packetssucceed at the current power level, the proposed algorithmattempts to lower the transmission power by one level. Thevalue ofnlower depends on the application’s data rate and themobile node’s speed. In practice,nlower is set to be a smallvalue, e.g., four in our experiments. On the other hand, whena transmission is not successful (i.e., not acknowledged) atpower levelp, the source retransmits the packet using powerlevel p + 1, until the maximum power level is reached. If thenode has not transmitted any packets during the lasttcancel

seconds, it transmits the next packet with maximum power.Table I indicates the reason behind the decision to restarttransmissions at full power. Specifically, it is more energyefficient to transmit a packet once at the maximum power andsucceed than to retransmit the packet multiple times at lowerpower levels (i.e., power levels 7 to 27 for the CC2420 radio).

Figure 5 presents the basic transmission power controlscheme as a state diagram. While this diagram is based onthe characteristics of the CC2420 radio, it can be adapted toother radios. We also note that since acknowledgment framesplay an essential role in our scheme, we transmit them withmaximum power levels in all cases to assure reliable deliverywith a single transmission attempt.

B. Using the Link Quality Indicator (LQI)

While the previous scheme is flexible enough to adapt todynamic channel environments it introduces two inefficiencies.First, it may inadvertently increase the transmission powerlevels quickly due to fluctuations in instantaneous link quality.This inefficiency represents a trade-off between having multi-ple failed transmissions at lower power levels and temporarilyincreasing the transmission power above the minimum levelnecessary. The second inefficiency is related to the strategyused to decrease transmission power. Specifically, if a nodeincreases its transmission power due to transient link qualitychanges, thennlower × increased number of levels packetswill be transmitted at higher power levels before reducing thepower level back to its previous state.

We leverage the link quality indicator (LQI) that IEEE802.15.4-compliant radios report for each successfullyreceived packet [8] to address the second inefficiency. Whilethe IEEE 802.15.4 standards do not specify a method forcomputing LQI, the CC2420 radio returns a value that isinversely proportional to the packet’s chip error rate [17].The data source collects LQI values with no extra overheadusing the receiver’s acknowledgment frames.

Figure 6 presents the LQI values of the packets that astationary node receives as a mobile transmitter travels alongthe linear path shown in Figure 1. One notices in both casesthe small variance in LQI values when the distance betweenthe transmitter and receiver is small and LQI levels are high.

60

70

80

90

100

110

0 30 60 90 120 150

LQI

Walk Distance (m)

Outdoor

60

70

80

90

100

110

0 30 60 90 120

LQI

Walk Distance (m)

Indoor

Fig. 6. LQI values for received packets in the outdoor (left)and indoor (right)environments when a mobile transmitter sends packets at maximum power toa stationary receiver. The receiver is positioned at the beginning of the pathwhile the sender moves at constant speed over the path shown in Figure 1.Consecutive high LQI values can be observed when the transmitter-receiverdistances are small in the beginning and the end of the round-trip path.

0

0.2

0.4

0.6

0.8

1

35 40 45 50 55 60 65 70 75 80 85 90 95 100 105

CD

F

LQI Value

Power region 3Power region 7

Power region 11Power region 15Power region 19Power region 23Power region 27Power region 31

0

0.2

0.4

0.6

0.8

1

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

CD

F

Continuous LQI values > 100

Power region 3Power region 7

Power region 11Power region 15Power region 19Power region 23Power region 27Power region 31

Fig. 7. CDF of acknowledgment frames’ LQI values (top) and number ofconsecutive LQI values higher than 100 (bottom) for different optimal powerlevel regions. Each line represents different positions with different minimumpower levels that assure 90% reliability. Results suggest that consecutive LQIvalues larger than 100 are an effective indicator that the power level shouldbe less or equal to 15.

Despite showing a positive correlation between high LQIvalues and good link quality, the previous graph does notprovide a method that the transmitter can use to map theacknowledgments’ LQI values to transmit power levels. Inorder to derive this rule, we set up a transmitter that senta sequence of packets to a static receiver placed at the endof the 75 m path in the outdoors testing environment shownin Figure 1. The transmitter sent 500 packets for each powerlevel and recorded the LQI values of all the acknowledgmentsit received. We manually moved the transmitter to the nextposition along the path at the end of each measurement cycle.Based on these data we computed the minimum power levelp

necessary to achieve packet reception ratio (PRR) above 90%.Then, for each power levelp we define the set of positionswherep is the minimum power level that achieves a PRR of90% asp’s optimal transmission power region.

In Figure 7 we plot the cumulative density function (CDF)of the LQI values from the acknowledgments collected at thetransmitter (top) and the consecutive number of acknowledg-ment frames with LQI higher than 100 (bottom) for one posi-tion in each transmission power region. Within each optimal

transmission power region, the plots looked similar and weplot the edge cases for each region. These figures suggest thatLQI = 100 can be a reasonable cutoff for “high” LQI valuesand that consistently observing LQI> 100 from acknowledg-ment frames indicate that a mobile node’s transmission powershould not be higher than 15.

Therefore, when more thannlqi continuous acknowledg-ment frames show LQI values higher thanlqihigh (in our caselqihigh = 100), and the current transmission powerpT is higherthanplqi (in our environmentplqi = 15), we can immediatelysetpT = plqi to assure fast convergence to the lowest possibletransmission power. The results from Figure 7 also imply thatnlqi can be set as low as two. We note that we observed similartrends (plqi = 15, lqihigh = 100) in the indoor environment.

V. EVALUATION

We use three metrics to evaluate the proposed transmis-sion power control schemes. Thenumber of transmissionattempts is the first metric. This metric represents the re-source efficiency of a scheme and also measures the wastedbandwidth. Moreover, this metric is directly related to ournextmetric,energy consumption. Both power level selection andthe number of transmission attempts will affect this metric.Finally, we use thenumber of packet receptions at thereceiver as the third metric. While high packet reception atthe destination is desirable, unintended receivers overhearingpacket transmissions can lead to decrease in spatial reuse.These metrics combined represent the efficiency of a schemeboth in terms of energy usage and bandwidth consumption.

We compare both schemes with the optimal transmissionpower levels and also with a naıve scheme in which pack-ets are always transmitted at maximum power. The optimaltransmission power levels represent the minimum transmissionpower necessary to achieve 90% PRR at a given position andare computed offline. For example, if a transmitter’s position isfar away from the receiver and packets are transmitted at lowpower levels, not all packets will be successfully received. Ifthis PRR is below 90%, we try transmitting packets at higherpower levels on the same link and select the lowest power levelthat achieves at least 90% PRR as the optimal transmissionpower for the transmitter’s position. We select 90% as thePRR threshold given that this is an application requirementfor one of our potential applications [9]. Using higher PRRthresholds did not generate different optimal power levels.

All experiments use TMote Sky motes equipped with IEEE802.15.4 compliant TI CC2420 radios [17]. In all cases,nlower

= 4 andtcancel = 10 seconds. We selected these values aftermultiple rounds of experiments in our target environments.Finally, based on the observations from Section IV-B, we setnlqi = 3. These parameters can be customized for differentenvironments by performing an initial round of channel mea-surements. Nevertheless, the values used across two very dif-ferent environments are similar, suggesting that the parametersare not very sensitive to the deployment conditions.

5 10 15 20 25 30 35

0 30 60 90 120 150

Tra

nsm

issi

on P

ower

Walk Distance (m)

Outdoor - Adaptive

5 10 15 20 25 30 35

0 30 60 90 120

Tra

nsm

issi

on P

ower

Walk Distance (m)

Indoor - Adaptive

5 10 15 20 25 30 35

0 30 60 90 120 150

Tra

nsm

issi

on P

ower

Walk Distance (m)

Outdoor - LQI-enhanced

5 10 15 20 25 30 35

0 30 60 90 120

Tra

nsm

issi

on P

ower

Walk Distance (m)

Indoor - LQI-enhanced

Fig. 8. Selected power levels for the basic adaptive scheme (top) and LQI-enhanced scheme (bottom) using periodic traffic. The left column correspondsto the outdoor testing environment while the figures on the right are from theindoor environment. Offline optimal transmission power values are shown ingray. Compared to the basic adaptive scheme, the LQI-enhanced scheme triesaggressively to lower its power levels when link conditionsallow.

Periodic/outdoor Optimal Adaptive Enhanced Naıve

Current Draw (mA) 13.65 13.76 13.03 17.40

TX Attempts 1 1.10 1.11 1

Energy Savings 21.55% 12.93% 16.95% 0%

PRR 100% 98.31% 98.06% 98.25%

Periodic/indoor Optimal Adaptive Enhanced Naıve

Current Draw (mA) 12.79 14.23 13.36 17.40

TX Attempts 1 1.08 1.10 1

Energy Savings 26.46% 11.81% 15.46% 0%

PRR 100% 98.18% 98.19% 98.00%

TABLE IICOMPARISON OF AVERAGE CURRENT DRAW AND TRANSMISSION

ATTEMPTS FOR SUCCESSFUL RECEPTION WITH PERIODIC TRAFFIC. ALSOSHOWN, THE PRRTHAT EACH SCHEME ACHIEVES.

A. Energy efficiency

We evaluate the energy efficiency of the proposed schemesusing two different traffic patterns: periodic traffic and a pat-tern with multiple packet bursts. All experiments took place inthe outdoor and indoor environments described in Section III.A mobile node departs from a stationary receiver and movesalong the yellow path shown in Figure 1, then returns to thereceiver’s location at a constant speed of∼1.4 m/s. Notice thatthe path that mobile devices traverse can affect the amount ofenergy consumption savings. We use fixed tracks with constantspeed to equally cover areas close and far from the receiver.

1) Periodic Traffic: The transmitter generates one unicastpacket every 128 ms, requesting acknowledgments from thereceiver. Figure 8 presents the power levels that the twoproposed schemes select as the mobile node travels away fromthe receiver and back. Also shown, are the optimal powerlevels for the two environments. The naıve scheme alwaystransmits at power level 31. It is clear that more packets aresent using power levels around 15 in the LQI-enhanced scheme

5 10 15 20 25 30 35

0 30 60 90 120 150

Tra

nsm

issi

on P

ower

Walk Distance (m)

Outdoor - Adaptive

5 10 15 20 25 30 35

0 30 60 90 120

Tra

nsm

issi

on P

ower

Walk Distance (m)

Indoor - Adaptive

5 10 15 20 25 30 35

0 30 60 90 120 150

Tra

nsm

issi

on P

ower

Walk Distance (m)

Outdoor - LQI-enhanced

5 10 15 20 25 30 35

0 30 60 90 120

Tra

nsm

issi

on P

ower

Walk Distance (m)

Indoor - LQI-enhanced

Fig. 9. Selected power levels for the basic adaptive scheme (top) andLQI-enhanced scheme (bottom) for a bursty traffic source. The left columncorresponds to the outdoor testing environment while the figures on the rightare from the indoor environment. Offline optimal transmission power valuesare shown in gray. When the inter-node distance is small, theLQI-enhancedscheme quickly decreases the power levels, compared to the basic scheme.

(bottom) than the basic scheme (top). As link conditionsimprove (i.e., as the mobile node approaches the receiver),the LQI-enhanced scheme attempts to reduce its transmissionpower to 15 (plqi), leading to multiple packet transmissions inthat transmission power range.

Thus, as Table II shows, the LQI-enhanced scheme sacri-fices a small number of transmission attempts to aggressivelytry lower power levels resulting in higher energy savings.The current draw is computed by recording both the numberof transmission attempts and the power levels used for eachattempt. We calculate the average current draw by dividing thetotal current draw by the total number of transmissions.

While the basic adaptive and LQI-enhanced schemes suc-cessfully conserve a noticeable amount of energy compared tothe naıve scheme, we can notice that inefficiencies exist whencompared to the offline optimal values. This inefficiency ismostly due to the rapid increases and slow decreases of thetransmission power levels. The LQI-enhanced scheme attemptsto minimize this inefficiency by aggressively trying lowerpower levels as link quality improves. As a result, in theoutdoors test case, the per packet current draw is lower for theLQI-enhanced scheme than the offline optimal. This initiallycounter-intuitive result can be explain by the observationthatthe optimal values represent the minimum power level thatassures 90% PRR while the proposed scheme sometimes suc-ceeds in transmitting packets with lower power levels. How-ever, as Table II shows, this lower current draw also leads tomore transmission attempts. The net result is that the proposedscheme is still less efficient than the optimal. Last, we notethat the energy savings came without sacrificing PRR. Of theeight packets transmitted each second, the basic adaptive,LQI-enhanced and naıve schemes delivered an average of 7.86,7.85, 7.85 packets, respectively.

2) Bursty Traffic: To test the performance of our schemesfor bursty traffic sources, we use a source node that transmits

Bursty/outdoor Optimal Adaptive Enhanced Naıve

Current Draw (mA) 13.52 13.88 13.25 17.40

TX Attempts 1 1.04 1.07 1

Energy Savings 22.31% 17.10% 18.51% 0%

PRR 100% 98.61% 97.10% 97.78%

Bursty/indoor Optimal Adaptive Enhanced Naıve

Current Draw (mA) 12.63 14.51 13.61 17.40

TX Attempts 1 1.05 1.09 1

Energy Savings 27.41% 12.59% 14.77% 0%

PRR 100% 97.50% 99.10% 98.36%

TABLE IIICOMPARISON OF AVERAGE CURRENT DRAW AND TRANSMISSION

ATTEMPTS FOR SUCCESSFUL TRANSMISSION WITH BURSTY TRAFFIC

ALONG WITH THE PRRFOR EACH SCHEME.

a 9-second burst of packets followed by 11 seconds of idling.Given thattcancel = 10 s, this idle time is enough to invalidatethe previous transmission power level and force the mobilenode to re-learn its power levels. Within each burst, packetsare transmitted with an interval of 128 ms.

Figure 9 plots the power level selections of the proposedschemes in both environments. For the outdoor experiments(left), a total of six packet bursts were transmitted, wherethelast packet burst did not finish its transmissions and for theindoor experiments (right) five packet bursts were transmittedin full. In order to minimize the number of unsuccessfultransmissions the algorithm transmits the first packet of eachburst using maximum power and attempts to reduce the trans-mission power of subsequent packets since idle time is longerthan tcancel. We also notice that the LQI-enhanced schemesuccessfully reduces the transmission power quickly with onlya few transmission attempts at high power levels when thedistance is close (i.e., second burst in both environments).

Overall, as shown in Table III, the LQI-enhanced schemeoutperforms the basic adaptive scheme in terms of energy effi-ciency in both environments. Both schemes achieve noticeableenergy savings compared to the naıve scheme with only asmall number of additional transmission attempts. Moreover,out of the 72 packets per burst, the basic adaptive, LQI-enhanced, and naıve schemes each received an average of70.60, 70.63, 70.61 packets, respectively. Thus, the proposedschemes save energy with out sacrificing the PRR.

We also point out that since the beginning of each packetburst in Figure 9 represents a new “learning period” of thetransmission power, this test can be used to show how ourschemes would perform when handoffs happen for mobiledevices. If a mobile node decides to select a new next hopit should discard its existing transmission power informationand restart the learning process. The results from Figure 9indicate that this learning process is fast.

B. Spatial Reuse Benefits

We demonstrate the benefits of transmission power controlin terms of decreasing interference and increasing spatialreusethrough a one hour experiment in an indoor environment

Fig. 10. Pictorial overview of experimental setup for spatial reuse exper-iments. Three receivers were placed in an indoor environment. R1 is theintended receiver for the packets thatT1 andT2 transmit.

PRR Enhanced Adaptive Naıve

Receiver 1 97.78% 98.43%

Receiver 2 20.26% 96.15%

Receiver 3 13.52% 73.90%

TABLE IVAVERAGE PACKET RECEPTION RATIO AT EACH RECEIVER FOR BOTH

SCHEMES. THE PRRAT UNINTENDED RECEIVERS(RECEIVERS2 AND 3)ARE SIGNIFICANTLY LOWER WHEN USING THE ADAPTIVE SCHEME.

(see Figure 10). Specifically, we placed three receivers indistinct positions. Receiver 1 was the destination for twomobile transmitters which periodically transmitted two packetsper second. The other two receivers recorded all overheardmessages. The mobile nodes mostly moved within the roomwith frequent walks to the hallway shown in Figure 10. Con-sidering that packet reception at unintended receivers translatesto interference, we try observing how the LQI-enhanced powercontrol scheme reduces interference compared to the naıvescheme. The two systems used separate 802.15.4 channels (25and 26) and each of the two volunteers carried one transmitterfor each system simultaneously.

We plot the reception of packets at each receiver over time inFigure 11 for both schemes. One can see thatR1 successfullyreceived packets sent fromT 1 and T 2 in both schemes.Moving our attention toR2 (positioned closer to the room inthe hallway), one can see thatR2 overhears a larger number ofT 1’s packets compared toT 2. This difference can be explainedby the observation thatT 1 spent the majority of its time fartheraway fromR1 compared toT 2. This causedT 1’s packets to betransmitted at a higher power thanT 2 to reach the destination(R1), thus, increasing its interference range to be large enoughto reachR2. However, despite the larger interference range,the reception atR3 (positioned∼18 m down the hallway fromthe room) implies that the selected power levels were moreefficient than the naıve scheme. WhileR3 received most ofT 1 andT 2’s packets when using the naıve scheme, it received81.7% fewer packets with the adaptive scheme. Most of thepackets thatR2 and R3 overheard with the adaptive schemeoccurred when the volunteers traveled across the hallway.

We organize the average PRR with respect to the totalnumber of packets transmitted for each scheme in Table IV.

Naive-T2

Naive-T1

Adaptive-T2

Adaptive-T1

0 10 20 30 40 50 60

Time (Minutes)

Receiver 1 (Destination)

Naive-T2

Naive-T1

Adaptive-T2

Adaptive-T1

0 10 20 30 40 50 60

Time (Minutes)

Receiver 2 (Unintended Receiver)

Naive-T2

Naive-T1

Adaptive-T2

Adaptive-T1

0 10 20 30 40 50 60

Time (Minutes)

Receiver 3 (Unintended Receiver)

Fig. 11. Sequence of packets received by each of the three receivers in Figure 10. Compared to the naıve scheme, the adaptive scheme effectively reducespacket interference to the unintended receivers (2 and 3) while PRR to the actual destination (receiver 1) is not affected.

Additional Memory Usage RAM ROM

Basic Adaptive Scheme 16 Bytes 108 Bytes

LQI-Enhanced Scheme 20 Bytes 172 Bytes

TABLE VADDITIONAL RAM AND ROM USAGE OF THE PROPOSED SCHEMES ON

THE TMOTE SKY MOTE PLATFORM. BOTH THE BASIC ANDLQI-ENHANCED SCHEMES REQUIRE ONLY A SMALL AMOUNT OF

ADDITIONAL MEMORY SPACE.

One can observe that with the proposed adaptive scheme, thePRR at R1 stayed high while reducing the interference atthe other receivers significantly. These results indicate thatthe proposed adaptive transmission power control scheme suc-cessfully decreases interference at unintended receivers, whilecontinuing to transmit packets to its destination.

In terms of energy efficiency, the current draw of the adap-tive scheme was 22.41% lower than the naıve scheme atT 1 (13.50 mA) and 44.31% lower forT 2 (9.69 mA). Thedifferences in the two values are also due to the differentpositions that the two transmitters took during the test.

C. Discussion

Both proposed schemes are implemented in the PacketLinkcomponent of the CC2420 radio stack in TinyOS 2.x. Consid-ering that Tmote Sky motes have 10KB of RAM and 48KB ofROM, keeping memory usage low is critical. Fortunately, asTable V shows, the additional amount of memory necessaryis minimal.

A worry that might arise from the use of transmission powercontrol is the increasing occurrence of the hidden terminalproblem (HTP). Once a mobile node decreases its transmissionpower to the minimum level necessary to reach the intendedreceiver, carrier sensing from other nodes is more likely tofail. Nevertheless, the proposed adaptive scheme addresses theHTP problem by promptly increasing transmission power inresponse to unacknowledged packets. In other words, whenpacket collisions occur due to transmissions from hidden ter-minals, the original source reacts by increasing its transmissionpower and thus re-enables carrier sensing for the other sources.

Finally, while the results in Section V-B show that our adap-tive scheme can significantly reduce the interference caused byunintended packet reception, interference can also be caused

from packets that arenot received (e.g., a packet can getcorrupted due to low SNR but can still affect the receptionquality of other incoming packets). Quantifying the ability ofthe proposed power control schemes to reduce this form ofinterference is part of our future work.

VI. SUMMARY

We experimentally investigate the effectiveness of trans-mission power control for WSNs that include mobile nodesmoving at human walking speeds. Moreover, we propose twoschemes for controlling transmission power that improve en-ergy efficiency and increase spatial reuse. The first schemeuses active probing and can be implemented on all packet-based radios while the second is an enhancement for radiosthat provide link quality indicator (LQI) values. Measurementswith mobile devices in two realistic environments show thatthe proposed schemes can achieve power levels close to theoffline optimal and also significantly decrease the interferencerange compared to the naıve approach.

ACKNOWLEDGMENTS

We would like to thank the anonymous reviewers for theircomments that helped us improve the quality of this paper.JeongGil Ko and Andreas Terzis are partially supported bythe National Science Foundation under grant #085591.

REFERENCES

[1] Tarek Abdelzaher, Yaw Anokwa, Peter Boda, Jeff Burke, DeborahEstrin, Leonidas Guibas, Aman Kansal, Samuel Madden, and Jim Reich.Mobiscopes for human spaces.IEEE Pervasive Computing, 2007.

[2] S. Armstrong. Wireless connectivity for health and sports monitoring:a review. British Journal of Sports Medicine, 41:285–289, 2007.

[3] C.F.Chiasserini and R.R.Rao. Pulsed battery dischargein communica-tion devices. InInternational Conference on Mobile Computing andNetworking (MobiCom), 1999.

[4] Bor-Sen Chen, Bore-Keun Lee, and Sheng-Kai Chen. Adaptive powercontrol of cellular cdma systems via the optimal predictivemodel. IEEETransactions on Wireless Communications, 4(4), 2005.

[5] Prabal Dutta and David Culler. Practical asynchronous neighbor discov-ery and rendezvous for mobile sensing applications. InProceedings ofthe 6th ACM conference on Embedded network sensor systems, 2008.

[6] J. Gomez and A.T. Campbell. Variable-range transmission power controlin wireless ad hoc networks.Mobile Computing, IEEE Transactions on,6(1):87–99, Jan. 2007.

[7] Gregory Hackman, Otav Chipara, and Chenyang Lu. Robust TopologyControl for Wireless Sensor Networks. InProceedings of the ACMSensys, November 2008.

[8] IEEE Standard for Information technology – Telecommunications andinformation exchange between systems – Local and metropolitan areanetworks. Specific requirements – Part 15.4: Wireless Medium AccessControl (MAC) and Physical Layer (PHY) Specifications for Low-RateWireless Personal Area Networks (LR-WPANs). Available at http://www.ieee802.org/15/pub/TG4.html, May 2003.

[9] JeongGil Ko, JongHyun Lim, Yin Chen, Razvan Musaloiu-E., AndreasTerzis, Gerald Masson, Tia Gao, Walt Destler, Leo Selavo, and RichardDutton. MEDiSN: Medical Emergency Detection in Sensor Networks.ACM Transactions on Embedded Computing Systems (TECS), SpecialIssue on Wireless Health Systems, 2010.

[10] Shan Lin, Jingbin Zhang, Gang Zhou, Lin Gu, John A. Stankovic, andTian He. ATPC: Adaptive Transmission Power Control for WirelessSensor Networks. InProceedings of the 4th ACM Sensys, 2006.

[11] Alaa Muqattash and Marwan M. Krunz. A distributed transmissionpower control protocol for mobile ad hoc networks.IEEE Transactionson Mobile Computing, 3(2):113–128, 2004.

[12] Razvan Musaloiu-E., Andreas Terzis, Katalin Szlavecz, Alex Szalay,Joshua Cogan, and Jim Gray. Life Under your Feet: A Wireless SensorNetwork for Soil Ecology. InProceedings of EmNets, May 2006.

[13] Jeongyeup Paek, Krishna Chintalapudi, John Cafferey,Ramesh Govin-dan, and Sami Masri. A wireless sensor network for structural healthmonitoring: Performance and experience. InProceedins of the SecondIEEE Workshop on Embedded Networked Sensors (EmNetS-II), May2005.

[14] Joseph Polastre, Jason Hill, and David Culler. Versatile Low PowerMedia Access for Wireless Sensor Networks. InProceedings of the2nd ACM Sensys Confence, 2004.

[15] Theodore S. Rappaport.Wireless Communications: Principles & Prac-tices. Prentice Hall, 1996.

[16] Dongjin Son, B. Krishnamachari, and J. Heidemann. Experimental studyof the effects of transmission power control and blacklisting in wirelesssensor networks. InIEEE SECON, 2004.

[17] Texas Instruments. 2.4 GHz IEEE 802.15.4 / ZigBee-ready RFTransceiver, 2006.

[18] Jiangzhou Wang. Open-loop power control in cdma overlay. BroadbandWireless Communications 3G, 4G and Wireless LAN, 620, 2006.

[19] A. Wood, J. Stankovic, G. Virone, L. Selavo, Z. He, Q. Cao, T. Doan,Y. Wu, L. Fang, and R. Stoleru. Context-Aware Wireless Sensor Net-works for Assisted Living and Residential Monitoring.IEEE Network,2008.

[20] Murtaza Zafer, Bongjun Ko, and Ivan Wang-Hei Ho. Cooperativetransmit-power estimation under wireless fading. InMobiHoc ’08:Proceedings of the 9th ACM international symposium on Mobile adhoc networking and computing, 2008.