IEICE TRANS. COMMUN., VOL.E98–B, NO.9 SEPTEMBER 20151785

PAPER Special Section on Emerging Technologies on Ambient Sensor Networks toward Future Generation

Rate Adaptation Based on Exposure Assessment Using RectennaOutput for WLAN Station Powered with Microwave PowerTransmission

Shota YAMASHITA†a), Koichi SAKAGUCHI†, Yong HUANG††, Student Members,Koji YAMAMOTO†, Senior Member, Takayuki NISHIO†, Member,

Masahiro MORIKURA†, Fellow, and Naoki SHINOHARA††, Member

SUMMARY This paper proposes a rate adaptation scheme (RAS) fora wireless local area network (WLAN) station powered with microwavepower transmission (MPT). A WLAN station attempting to transmit dataframes when exposed to microwave radiation for MPT, experiences a re-duction in the physical (PHY) layer data rate because frames are lost evenwhen the carrier sense mechanism is used. The key idea of the proposedscheme is to utilize the output of the rectenna used for receiving microwavepower. Using rectenna output, a WLAN station based on the proposedscheme assesses whether the station is exposed to microwave radiation forMPT. Then, using historical data corresponding to the assessment result,the station selects an appropriate PHY data rate. The historical data areobtained from previous transmission results, e.g., historical data pertainingto the data frame loss ratio. The proposed scheme was implemented andverified through an experiment. Experimental results showed that the pro-posed scheme prevents the reduction in the PHY data rate, which is causedby the use of historical data stored in a single memory. Thus, the proposedscheme leads to an improvement in the WLAN throughput.key words: wireless local area network, rate adaptation, microwave powertransmission, rectenna

1. Introduction

Microwave power transmission (MPT) [1] is a wirelesspower transmission technology in which electric power istransmitted via microwaves. MPT enables us to transmitmicrowave power over long distances, as compared withother wireless power transmission methods, such as electro-magnetic induction, magnetic resonance, and other compo-nents. Realizing the batteryless operation of a device by us-ing wireless power transmission methods, allows the elimi-nation of power cables and a reduction in battery replace-ment costs. These advantages increase as the number ofbatteryless devices increases, especially in a wireless sen-sor network.

When microwave power is transmitted to a wirelessstation, the influence of MPT on data transmission has tobe considered. In this work, we use an IEEE 802.11 wire-less local area network (WLAN), which employs the carrier

Manuscript received January 9, 2015.Manuscript revised April 17, 2015.†The authors are with the Graduate School of Informatics,

Kyoto University, Kyoto-shi, 606-8501 Japan.††The authors are with the Research Institute for Sustainable

Humanosphere, Kyoto University, Uji-shi, 611-0011 Japan.a) E-mail: info14@imc.cce.i.kyoto-u.ac.jp

DOI: 10.1587/transcom.E98.B.1785

sense mechanism. WLAN stations with a high received in-terference power level defer the transmission of data framesand are unable to receive any frames.

In [2], the authors conducted experiments in whichmicrowaves with a frequency of 2.4 GHz were transmittedto an IEEE 802.11g-based WLAN station in co-channel oradjacent channel operations. The results were as follows.Firstly, when the received interference power at the station ishigh, MPT interferes with the data transmission even whenusing an adjacent channel. Accordingly, the station is de-terred from transmitting data frames due to the carrier sensemechanism. Secondly, exposure of the station to microwaveradiation for MPT leads to the selection of a lower physical(PHY) layer data rate, the use of which is maintained evenafter exposure to the microwave radiation is discontinued.The reduction in the PHY data rate occurs because the sta-tion attempts to continue data transmission during exposureto the microwave radiation. However, the station does notreceive acknowledge (ACK) frames because of the high in-terference power. Hence, a low PHY data rate is selectedwith a rate adaptation scheme (RAS) that is implementedon the station. Note that this problem occurs even when thestation employs the carrier sense mechanism.

In most RASs, by using data related to previous trans-mission results, e.g., historical data of the data frame loss ra-tio, the PHY data rate for data transmission is selected. Thepurpose of using the historical data is to estimate currentlink quality, which depends on the distance between a datatransmitter and a data receiver and on interference power atthe data receiver. To match the link quality, the PHY datarate is selected, i.e., such that data frames transmitted at theselected PHY data rate are successfully received by the datareceiver. Related to the estimation of the link quality, manyprevious studies have attempted to propose RASs, e.g., ARF[3], RBAR [4], Onoe [5], and Sample Rate [6]. In addition,some conventional RASs employ loss differentiation mech-anisms that diagnose the cause of data frame loss as colli-sion or link quality degradation, e.g., LD-ARF [7], CARA[8], LDRA [9], and ERA [10]. However, these conventionalschemes are designed without the assumption that high in-terference power at the station causes PHY data rate reduc-tion.

Some previous studies have proposed RASs capable of

Copyright c© 2015 The Institute of Electronics, Information and Communication Engineers

1786IEICE TRANS. COMMUN., VOL.E98–B, NO.9 SEPTEMBER 2015

assessing whether a WLAN station is exposed to microwaveradiation from devices using Bluetooth, ZigBee, or from mi-crowave ovens. SGRA [11] and ARES [12] attempt to as-sess whether a WLAN station is exposed to microwave ra-diation based on both the signal-to-noise power ratio (SNR)and the data frame loss ratio. However, because of the ex-posure assessment based on the data frame loss ratio, whenthe data receiver experiences strong SNR degradation, thestation erroneously detects that it is exposed to microwaveradiation for MPT even when this is not the case. The causeof this is that the data frame loss ratio increases, not onlybecause of exposure of the station to microwave radiationfor MPT, but also because of SNR degradation at the datareceiver. The increase in the data frame loss ratio with SNRdegradation is experimentally demonstrated in [13].

In this paper, we propose an RAS based on exposureassessment using rectenna output for a WLAN station ex-posed to microwave radiation for MPT. Then, we carry outexperiments to evaluate the performance of the system onwhich the proposed scheme is implemented. The rectennawas installed in the device powered with MPT, and convertsthe microwave power it receives into direct current power.The rectenna output enables a station based on the proposedscheme to assess whether it is exposed to microwave radi-ation for MPT. The historical data corresponding to the as-sessment result are used by the station to select an appropri-ate PHY data rate, where two independent memories, eachof which contains a different set of historical data for rateadaptation, are prepared in advance.

The proposed scheme was implemented on a WLANstation by modifying a mac80211 device driver [14] that iswidely used as a WLAN driver in Linux systems. This de-vice driver uses two conventional RASs by default, i.e., PID[15] and Minstrel [16]. By modifying either PID or Min-strel, we implemented PID-based or Minstrel-based pro-posed schemes on a WLAN station.

This paper is organized as follows. In Sect. 2, we ex-plain conventional RASs (PID and Minstrel) used in themac80211 device drivers. Section 3 contains our pro-posal for an RAS with rectenna output. The experimentsconducted to evaluate the performance are described inSect. 4. In Sect. 5, we discuss the advantages of the pro-posed scheme. Finally, we conclude this paper in Sect. 6.

2. Conventional Rate Adaptation Schemes of mac80211Device Driver

This section presents a description of the two conventionaldefault RASs of the mac80211 device driver, i.e., PID andMinstrel. In Sects. 2.1 and 2.2, we explain the operationof the station based on each of these RASs. Note that theperformances of these algorithms have been investigated in[17] and [18].

2.1 PID

A station based on PID selects an appropriate PHY data rate

based on the general proportional integral derivative con-troller such that the data frame loss ratio converges to apredetermined target value, RL,target. The station selects thePHY data rate every sampling period TPID.

The kth rate selection operates as follows. First, the sta-tion calculates the data frame loss ratio between the (k−1)thand kth rate selections, rL[k]. Second, the station calculatesthe kth error between RL,target and rL[k] (denoted by e[k]) as

e[k] = RL,target − rL[k]. (1)

Third, the station calculates the difference between e[k] ande[k − 1] (denoted by Δe[k]) as

Δe[k] = e[k] − e[k − 1]. (2)

Moreover, the station computes an exponential moving av-erage of e[k] (denoted by e[k]) by using e[k] and e[k − 1]computed at the (k − 1)th rate selection as

e[k] =(αsmooth − 1)e[k − 1] + e[k]

αsmooth, (3)

where αsmooth is a smoothing factor (αsmooth > 1). Fourth,the station computes an adjustment value, vadj[k], as

vadj[k] = CPe[k] +CIe[k] +CDΔe[k](1 + αsharp), (4)

where αsharp is a sharpening factor (if αsharp � 0, a fast re-sponse is achieved), and CP, CI, and CD are proportional,integral, and derivative coefficients, respectively. Finally, ifvadj[k] ≤ −1, the station decreases the PHY data rate. Ifvadj[k] ≥ 1, the station increases the PHY data rate, and if−1 < vadj[k] < 1, the station maintains the current PHY datarate.

2.2 Minstrel

A station based on Minstrel selects an appropriate PHYdata rate by using the previous data frame transmission per-formance such that the highest throughput performance isachieved. The station updates a rate table every samplingperiod TMinstrel. The rate table indicates the PHY data ratebased on the retransmission count.

The station updates the kth rate table as follows. First,N is a set of rate indexes that indicate the supported rates,and i is an element of N , i.e., i ∈ N . Let the PHY datarate corresponding to i be denoted by R(i). Second, the sta-tion calculates ps(i)[k] for every i, which represents the esti-mated success probability of data frame transmission at R(i)as follows. Note that the term “success” represents the sit-uation in which the station transmits a data frame and thenreceives an ACK frame. Let Δntransmit(i)[k] denote the in-creased number of data frames transmitted at R(i) betweenthe (k − 1)th and kth updates. Similarly, let Δnsuccess(i)[k]denote the increased number of data frames transmitted suc-cessfully at R(i) between the (k−1)th and kth updates. UsingΔntransmit(i)[k] and Δnsuccess(i)[k], ps(i)[k] is calculated as

ps(i)[k] =(1 − αscale)Δnsuccess(i)[k]Δntransmit(i)[k]

YAMASHITA et al.: RATE ADAPTATION BASED ON EXPOSURE ASSESSMENT USING RECTENNA OUTPUT FOR WLAN STATION1787

Table 1 Minstrel rate table created at the kth update.

RetransmissionNormal rate

Lookaround rate (generating irand[k] ∈ N)count R(irand[k]) ≤ R(ibest tp[k]) R(irand[k]) > R(ibest tp[k])

0 R(ibest tp[k]) R(ibest tp[k]) R(irand[k])1 R(inext best tp[k]) R(irand[k]) R(ibest tp[k])2 R(ibest prob[k]) R(ibest prob[k]) R(ibest prob[k])3 R(ibaserate[k]) R(ibaserate[k]) R(ibaserate[k])

+ αscale ps(i)[k − 1], (5)

where αscale is a scaling value (0 < αscale < 1). Third, thestation computes the throughput at R(i) (denoted by tp(i)[k])using the estimated maximum number of data frames trans-mitted successfully at R(i) per unit time. Note that the sym-bol “tp” represents “throughput.” Fourth, the station deter-mines the following three rate indexes.

• ibest tp[k]: the rate index of the PHY data rate that willachieve the highest throughput performance• inext best tp[k]: the rate index of the PHY data rate

that will achieve the second highest throughput perfor-mance• ibest prob[k]: the rate index of the PHY data rate that will

achieve the highest success probability of data frametransmission

These rate indexes are defined as

ibest tp[k] = arg maxi∈N

tp(i)[k], (6)

inext best tp[k] = arg maxi∈N ,i�ibest tp[k]

tp(i)[k], (7)

ibest prob[k] = arg maxi∈N

ps(i)[k], (8)

respectively. Finally, the station updates the rate table inTable 1, where ibaserate represents the rate index of the lowestsupported PHY data rate.

Using Table 1, the station selects an appropriate PHYdata rate as follows. The “Normal rate” is used to transmit90% of data frames and the remaining 10% are transmittedat the “Lookaround rate,” for which the station generates arandom rate index, irand[k] ∈ N . The station then determinesthe PHY data rate for the current retransmission count asshown in Table 1.

3. Rate Adaptation Scheme Based on Exposure Assess-ment Using Rectenna Output

In this section, we propose an RAS based on exposure as-sessment using rectenna output. The design of this proposedscheme focuses on the use of historical data in both of theconventional RASs, i.e., PID and Minstrel. Recall that his-torical data are obtained from previous transmissions, e.g.,historical data of the data frame loss ratio. This proposedscheme has the following two features: (i) the station as-sesses whether it is exposed to microwave radiation for MPTby using rectenna output power; (ii) the station selects an ap-propriate PHY data rate using historical data corresponding

Fig. 1 Flowchart of kth rate adaptation in the proposed scheme, whereα[k] is the historical data obtained during [tk−1, tk].

to the assessment result. The reason for using rectenna out-put power is twofold. First, a rectenna has been installedin a station powered with MPT and thus there is no needto install other devices to perform the exposure assessment.Second, the use of rectenna output power enables a stationequipped with the rectenna to directly assess whether it isexposed to microwave radiation for MPT.

Figure 1 illustrates the flowchart of the proposedscheme. We explain kth rate adaptation at the time tk asfollows. Immediately before rate adaptation, the stationmeasures the rectenna output power, po[k]. Then, by us-ing po[k], the station assesses whether it is exposed to mi-crowave radiation for MPT. Let the power threshold be de-noted by Pth. Pth should be determined from the power thestation receives from microwaves for MPT. The reasons areas follows. The output power of the rectenna that receivesmicrowaves for communication is less than 1 μW. The rea-son is that the rectenna output power is not more than themaximum transmission power that limited to 10 mW/MHzin Japan, and that at 2.4 GHz, the free space propagationloss at 1 m distance is equal to 40 dB in theory. On theother hand, the output power of the rectenna that receivesmicrowaves for MPT is more than that of the rectenna thatreceives microwaves for communication, where, in this pa-per, we assume that the supplied power to the station is ofthe order of milliwatt.

Then, by using the historical data in the memory cor-responding to the assessment result, the station selects anappropriate PHY data rate. The two independent memo-ries in which the historical data for rate adaptation purposesare stored are denoted by ME and MNE, where the subscripts“E” and “NE” represent “exposure” and “non-exposure,” re-spectively. When po[k] > Pth, the station determines that

1788IEICE TRANS. COMMUN., VOL.E98–B, NO.9 SEPTEMBER 2015

it is exposed to microwave radiation for MPT, after whichit selects an appropriate PHY data rate using the historicaldata stored in ME, which contains the corresponding his-torical exposure data. At the same time, the historical dataobtained during [tk−1, tk] is stored in ME. On the other hand,when po[k] ≤ Pth, the station determines that it is not ex-posed to microwave radiation for MPT and then selects anappropriate PHY data rate using the historical data storedin MNE, which contains the historical data corresponding tonon-exposure to microwave radiation for MPT. At the sametime, historical data obtained during [tk−1, tk] are stored inMNE.

The scheme proposed in this paper does not define aconcrete RAS; instead, it uses historical data correspondingto the state of exposure to microwave radiation for MPT.Hence, a station based on the proposed scheme is able touse RASs corresponding to the exposure assessment result.In the experiments described in Sect. 4, we implemented theproposed scheme on a station such that the PHY data rateis selected on the basis of a single conventional RAS re-gardless of the magnitude of po[k]. This has the purpose ofenabling a comparison with the conventional RAS. For ex-ample, in the “PID-based proposed scheme,” po[k] is mea-sured immediately before the kth rate selection is measured,following which the PHY data rate is selected much in thesame way as described in Sect. 2.1. Unlike in conventionalPID, in the PID-based proposed scheme, historical data cor-responding to the magnitude of po[k] is used to select thePHY data rate. When po[k] > Pth, the PHY data rate isselected by using both e[klast,E] and e[klast,E] that satisfy

klast,E = maxk′∈NE

k′, (9)

where NE = { n | po[n] > Pth, n = 1, . . . , k − 1 }. Whenpo[k] ≤ Pth, the PHY data rate is selected by using bothe[klast,NE] and e[klast,NE] that satisfy

klast,NE = maxk′∈NNE

k′, (10)

where NNE = { n | po[n] ≤ Pth, n = 1, . . . , k − 1 }.

4. Experiments

This section presents an explanation of the experiments inthis paper. All measurements were performed in a radio fre-quency (RF) anechoic chamber to ensure the absence of anysources of wireless transmissions except for the system de-scribed in Sect. 4.1.

4.1 Experimental Setup

Figure 2 illustrates the experimental setup for the perfor-mance evaluation of the proposed scheme. The experimen-tal system consists of the following three devices, an en-ergy source (ES), a data transmitter (DT), and a data receiver(DR).

The details of each of these devices are as follows. The

ES intermittently transmits continuous microwaves for MPTto the DT and consists of a transmission antenna, an ampli-fier, and an RF signal generator. The microwaves generatedby the RF signal generator are amplified via the amplifier,and are then transmitted from the transmission antenna tothe DT.

The DT continuously transmits data frames to the DR.It consists of a commercial WLAN adapter (Logitec LAN-W150NU2AB, rt2800usb device driver), a Linux machine(Raspberry Pi Model B+ [19], Linux kernel 3.12.23+), alaptop (Apple MacBook Air, OS X 10.9.5), a microcon-troller board (Arduino Uno board [20]), and a rectenna. Themicrocontroller board measures the rectenna output powerand assesses whether the DT is exposed to microwave ra-diation for MPT, and then shares the information of the ex-posure assessment with the Linux machine. The Linux ma-chine, which is remotely operated by the laptop, generatesthe data that are transmitted from the WLAN adapter.

The DR manages the network, receives the data framestransmitted from the DT, and captures the frames transmit-ted by the network as a packet sniffer. The DR consists ofan access point (AP, Allied Telesis AT-TQ2403) and a lap-top (Apple MacBook Pro, OS X 10.8.4). The AP generatesand manages an IEEE 802.11g-based WLAN. The laptopreceives the data frames transmitted from the DT via the APand captures the frames transmitted from both the DT andthe DR using its internal WLAN device. Note that the framecapture at the DR laptop is for the purpose of evaluating thethroughput performance of the proposed scheme and is in-dependent of the rate selection at the DT.

In this paper, the experiments were conducted only inthe environment consisting of one DT and one DR, whereasthe performance evaluation of the proposed scheme in anenvironment consisting of multiple DTs and DRs is beyondthe scope of this paper. The reason is that the performancein the environment consisting of multiple nodes depends notonly on the RAS but also on other factors, e.g., contentionbetween the multiple nodes and the method of transmittingmicrowave power to the multiple nodes. In addition, we canverify the performancfe of the proposed scheme either in theenvironment consisting of multiple DTs and DRs or in theenvironment consisting of one DT and one DR. There aretwo reasons for this. One is that, in RASs, the PHY data ratefor each pair of DT and DR is selected independently. Theother one is that the two features of the proposed scheme,i.e., the exposure assessment using rectenna output and theuse of historical data corresponding to the assessment result,are also performed independently on each DT.

4.2 Parameters

Figure 2 shows the position of each of the devices, and Fig. 3shows the entire DT and the power transmission antenna ofthe ES. Figures 2 and 3 also show that the DT is placed infront of the transmission antenna, whereas the DR is placedbehind the transmission antenna such that the microwavepower transmitted from the ES only affects the DT. The DR

YAMASHITA et al.: RATE ADAPTATION BASED ON EXPOSURE ASSESSMENT USING RECTENNA OUTPUT FOR WLAN STATION1789

Fig. 2 Experimental setup.

Fig. 3 DT and power transmission antenna.

AP, transmission antenna, and DT are placed in a straightline. The distances from the transmission antenna to the DRAP, DR laptop, and DT are 3.06 m, 1.63 m, and 2.96 m, re-spectively.

The ES repeatedly transmits continuous microwavesfor MPT to the DT for 1.4 s and then pauses for 5.0 s. Notethat the duration of the microwave transmission, 1.4 s, sat-isfies the following two conditions. It ensures that the DTdoes not become disassociated from the DR AP because ofthe consecutive non-reception of beacon frames, and, addi-tionally, it ensures that the PHY data rate is reduced duringthis period. The central frequency of the microwaves is setto be 2.457 GHz, and the bandwidth is less than 2 kHz. Thegain of the power transmission antenna is 16.4 dBi. Whenthe ES transmits the microwaves, the transmission power ofthe ES is set to be 17.9 W. In this case, the power density ateach antenna position has the value listed in Table 2. More-over, the DT rectenna output power when the ES transmitedmicrowaves for MPT is 31 mW, whereas the DT rectennaoutput power when the ES did not transmit microwaves forMPT is 0 mW.

The DT transmits data frames at eight different PHYdata rates: 6, 9, 12, 18, 24, 36, 48, and 54 Mbit/s. The cen-tral frequency for data transmission is set to be 2.457 GHz.Recall that MPT interferes with data transmission evenwhen an adjacent channel is used for the purpose. The DTuses Iperf 2.0.5 [21] and generates user datagram protocol(UDP) data traffic with an offered load of 54 Mbit/s, i.e., the

Table 2 Power density at each antenna when the ES transmits mi-crowaves for MPT.

Antenna position Power densityDT rectenna 0.86 mW/cm2

DT WLAN adapter 0.92 mW/cm2

DR laptop 1.33 μW/cm2

DR AP 0.97 μW/cm2

network is saturated by traffic. The UDP datagram size isset to be 1,470 B and this size has no effect on the advantageof the proposed scheme because the throughput degradationdiscussed in this paper is caused by the ACK frame loss dis-cussed in Sect. 1.

The proposed scheme was implemented on the DTLinux machine by modifying the mac80211 device driver.Now, Pth was set to be 6 mW taking both that the DTrectenna output power of 0 mW when the ES did not trans-mit microwaves for MPT and that the DT rectenna outputpower of 31 mW when the ES transmited microwaves forMPT into account. We emphasize that this value is for as-sessing that the rectenna output power is either 0 mW or31 mW and does not sensitively affect the performance ofthe proposed scheme. In addition, to compare the role ofthe exposure assessment by using the rectenna output in theproposed scheme, we investigated the RAS without rectennaoutput for comparison as described in Sect. 4.3. As de-scribed in Sect. 3, regardless of the use of the rectenna out-put, the proposed scheme is implemented such that the PHYdata rate is selected on the basis of either PID or Minstrel.The proposed scheme was modified as follows when imple-mented. As shown in Fig. 4, when the station determinedthat it was not exposed to microwave radiation for MPT inthe (k− 1)th rate adaptation, but instead determines that it isexposed to microwave radiation for MPT in kth rate adap-tation, the station copies over the contents of MNE to ME

immediately before the rate selection using ME. Even whenwe add these modifications to the proposed scheme, we areable to confirm its effectiveness because the content of MNE

shows the same behavior regardless of these modifications.Based on the above discussion, we investigated the per-

formance of each of the three PID-based and each of thethree Minstrel-based RASs, i.e., the PID, PID-based pro-

1790IEICE TRANS. COMMUN., VOL.E98–B, NO.9 SEPTEMBER 2015

Fig. 4 Flowchart of kth rate adaptation in the proposed scheme imple-mented for the experiments, where α[k] is the historical data obtained dur-ing [tk−1, tk].

Table 3 PID default parameters.

TPID RL,target αsmooth αsharp CP CI CD

125 ms 14% 8 0 15 9 15

Table 4 Minstrel default parameters.

TMinstrel αscale

100 ms 0.25

posed scheme with and PID-based scheme without rectennaoutput for comparison, Minstrel, and Minstrel-based pro-posed scheme with and Minstrel-based scheme withoutrectenna output for comparison. The parameters used inboth the PID-based and Minstrel-based RASs were initial-ized with the values in Tables 3 and 4, respectively, and wereconstant.

The experiments enabled us to evaluate the UDPthroughput, PHY data rate used in the data frame trans-mission, and data frame loss ratio for each RAS. Firstly,the UDP throughput was estimated based on the number ofACK frames the DR laptop captured with Wireshark [22].Recall that the frame capture at the DR laptop is for the pur-pose of evaluating the throughput performance of the pro-posed scheme and is independent of the rate selection at theDT. Even though the use of Iperf would enable us to calcu-late the UDP throughput, we estimated the UDP throughputwith the captured frames. The reason is that the period of theUDP throughput estimation with the captured frames can beset shorter than that of the UDP throughput calculation withIperf. Secondly, the PHY data rate in the PID-based RASswas measured on the DT Linux machine by using the de-bug function. On the other hand, the PHY data rate in theMinstrel-based RASs was estimated on the DR laptop by us-ing captured data frames. In the Minstrel-based RASs, thereason for using captured data frames is that the PHY datarate used in each data transmission is unknown even whenthe debug function is used [16]. Lastly, the data frame lossratio was measured on the DT Linux machine by using the

debug function. We emphasize that the DT Linux machinewas not able to completely obtain the data frame loss ratiobecause it did not have the required performance.

4.3 Rate Adaptation Scheme Based on Exposure Assess-ment Not Using Rectenna Output for Comparison

The performance of the proposed scheme was evaluated de-pending on whether the station uses rectenna output. Forthis reason we now explain the operation of an RAS with-out rectenna output for comparison. The station based onthe scheme without rectenna output uses the results of pre-vious data frame transmissions to assess whether the stationis exposed to microwave radiation for MPT.

We explain the operation in the kth rate adaptation atthe time tk using the value of rL[k] described in Sect. 2.1.Recall that rL[k] denotes the data frame loss ratio during[tk−1, tk]. Note that, even in the Minstrel-based scheme with-out rectenna output for comparison, rL[k] is calculated. Letthe number of data frames transmitted during [tk−1, tk] bedenoted by Δntransmit[k]. In the same way as the proposedscheme, two memories ME and MNE are prepared. Usingthese memories, the kth rate adaptation is performed as fol-lows. When rL[k − 1] < 50%, if rL[k] ≥ 50%, the sta-tion determines that it is exposed to microwave radiationfor MPT, and then selects an appropriate PHY data rateusing the historical data stored in ME and then stores inME the historical data obtained during [tk−1, tk]. Then, thestation continues to use historical data stored in ME untilΔntransmit[m] ≥ 3Δntransmit[m − 1] in the mth rate adaptationfor any m > k. On the other hand, when the station usesthe historical data stored in ME in the (k − 1)th rate adap-tation, if Δntransmit[k] ≥ 3Δntransmit[k − 1], it determines thatit is not exposed to microwave radiation for MPT, and thenselects an appropriate PHY data rate using historical datastored in MNE and stores in MNE the historical data obtainedduring [tk−1, tk]. Then, the station continues to use the his-torical data stored in MNE until both rL[m − 1] < 50% andrL[m] ≥ 50% in the mth rate adaptation for any m > k.

4.4 Experimental Results

Figure 5 shows both the UDP throughput and PHY data ratefor each of the PID-based RASs. In addition, Fig. 6 showsboth the data frame loss ratio and PHY data rate for eachof the PID-based RASs. The red lines in Fig. 5 confirmthat the PHY data rate is reduced when the DT is exposedto microwave radiation for MPT. Comparing both the UDPthroughput and the PHY data rate in Fig. 5(a) with those inboth Figs. 5(b) and 5(c), it is clear that there has been an im-provement. Note that, in Figs. 5(b) and 5(c), there is a periodduring which the DT uses the lowest PHY data rate imme-diately after the MPT is discontinued. This is because therate selection interval in the PID scheme is set to be 125 ms.

Figure 7 shows both the UDP throughput and PHY datarate for each of the Minstrel-based RASs. In addition, Fig. 8shows both the data frame loss ratio and PHY data rate for

YAMASHITA et al.: RATE ADAPTATION BASED ON EXPOSURE ASSESSMENT USING RECTENNA OUTPUT FOR WLAN STATION1791

Fig. 5 UDP throughput and PHY data rate for each of the three PID-based RASs, where the grey area indicates the time during which the DT isexposed to microwave radiation for MPT.

each of the Minstrel-based RASs. Except for the changein the link quality, the UDP throughput in Fig. 7(a) can beconsidered the same as that in Figs. 7(b) and 7(c). On theother hand, after the MPT is discontinued, a PHY data rateof less than 48 Mbit/s is sometimes used in Fig. 7(a), whilethat of 54 Mbit/s is always used in Figs. 7(b) and 7(c).

5. Discussion

Taking the observations in Sect. 4 into account, we discussthe advantages of the proposed scheme.

Exposure assessment based on the rectenna output al-lows a device equipped with the rectenna to precisely rec-ognize in real-time that the device is exposed to microwaveradiation for MPT. Here it should be noted that if a device ispowered with MPT, the device has a rectenna installed. Ac-cording to either Figs. 5(b) and 5(c) or Figs. 7(b) and 7(c),the UDP throughput of the proposed scheme can be regardedas the same as that of the RAS without rectenna output forcomparison, except for the change in the quality of the link.However, the use of the scheme without rectenna output forcomparison has the following two disadvantages. Firstly,as described in Sect. 1, because of the exposure assessment

Fig. 6 Data frame loss ratio and PHY data rate for each of the three PID-based RASs, where the grey area indicates the time during which the DT isexposed to microwave radiation for MPT.

based on the data frame loss ratio, when severe SNR degra-dation occurs at the receiver, the station based on the schemeerroneously determines that it is exposed to microwave ra-diation for MPT even when it is not. Secondly, because ofthe exposure assessment based on the number of transmitteddata frames, when the station transmits data frames underunsaturated traffic conditions or when there are other nodesthat operate in the same band as the station, the station canerroneously determine that it is exposed to microwave radi-ation for MPT even when this is not the case. The causeof this is that the number of transmitted data frames fluc-tuates depending both on the traffic conditions and on thecontention level.

Using historical data corresponding to the state of ex-posure to microwave radiation for MPT has the followingadvantages in comparison with using continuously histori-cal data stored in a single memory.

• After discontinuation of the MPT, the high PHY datarate is maintained, and thus throughput degradationdoes not occur. Compared with Fig. 5(b), Fig. 5(a)shows that the PHY data rate was reduced over theperiod from 24.8 to 26.2 s temporarily, after which aPHY data rate below 48 Mbit/s was used over the pe-

1792IEICE TRANS. COMMUN., VOL.E98–B, NO.9 SEPTEMBER 2015

Fig. 7 UDP throughput and PHY data rate for each of the three Minstrel-based RASs, where the grey area indicates the time during which the DT isexposed to microwave radiation for MPT.

riod from 26.2 to 28.6 s. On the other hand, Fig. 5(b)shows that a PHY data rate below 48 Mbit/s was notused over the period from 24.2 to 26.6 s.• Using historical data corresponding to the exposure

assessment result prevents the content of historicaldata from being corrupted. Compared with Fig. 7(b),Fig. 7(a) shows that a PHY data rate below 48 Mbit/swas sometimes used over the period from 18.7 to19.8 s. On the other hand, Fig. 7(b) shows that a PHYdata rate below 48 Mbit/s was not used over the pe-riod from 34.9 to 35.8 s. The reason for this is thatin Fig. 7(a), the contents stored in MNE were corrupteddue to a data frame loss when the DT is exposed tomicrowave radiation for MPT.

6. Conclusion

This paper proposed an RAS based on exposure assessmentusing rectenna output for a WLAN station powered withMPT. The objective of the proposed scheme was to pre-vent the throughput degradation that occurs when the sta-tion is exposed to microwave radiation for MPT. This objec-tive was achieved by utilizing the rectenna output to assess

Fig. 8 Data frame loss ratio and PHY data rate for each of the threeMinstrel-based RASs, where the grey area indicates the time during whichthe DT is exposed to microwave radiation for MPT.

whether the station is exposed to microwave radiation forMPT. Note that a device powered with MPT has a rectenna.The proposed scheme is based on the use of two indepen-dent memories in which historical data for rate adaptationwere stored with the aim of mitigating the PHY data ratedegradation when the station is exposed to microwave radi-ation for MPT. The historical data stored in the two memo-ries were used to determine when the station was either ex-posed or not exposed to microwave radiation for MPT. Weimplemented the proposed scheme on a station using a com-mercial WLAN device and a widely used device driver. Theresults confirmed the effectiveness of the proposed scheme.

Acknowledgment

This work was supported in part by JSPS KAKENHI GrantNumber 24360149. This experiment was carried out us-ing the Microwave Energy Transmission Laboratory (MET-LAB) system of the Research Institute for Sustainable Hu-manosphere at Kyoto University.

References

[1] N. Shinohara, “Power without wires,” IEEE Microw. Mag., vol.12,

YAMASHITA et al.: RATE ADAPTATION BASED ON EXPOSURE ASSESSMENT USING RECTENNA OUTPUT FOR WLAN STATION1793

no.7, pp.S64–S73, Dec. 2011.[2] N. Imoto, S. Yamashita, T. Ichihara, K. Yamamoto, T. Nishio,

M. Morikura, and N. Shinohara, “Experimental investigation ofco-channel and adjacent channel operations of microwave powerand IEEE 802.11g data transmissions,” IEICE Trans. Commun.,vol.E97-B, no.9, pp.1835–1842, Sept. 2014.

[3] A. Kamerman and L. Monteban, “WaveLAN-II: A high-perfor-mance wireless LAN for the unlicensed band,” Bell Labs Tech. J.,vol.2, no.3, pp.118–133, June 1997.

[4] G. Holland, N. Vaidya, and P. Bahl, “A rate-adaptive MACprotocol for multi-Hop wireless networks,” Proc. 7th AnnualInternational Conference on Mobile Computing and Network-ing — MobiCom’01, pp.236–251, 2001.

[5] “Madwifi Onoe specification,” http://sourceforge.net/projects/madwifi/

[6] J.C. Bicket, “Bit-rate selection in wireless networks,” MIT Master’sThesis, Feb. 2005.

[7] Q. Pang, V.C.M. Leung, and S.C. Liew, “A rate adaptation algorithmfor IEEE 802.11 WLANs based on MAC-layer loss differentiation,”Proc. 2nd International Conference on Broadband Networks, 2005,vol.1, pp.659–667, 2005.

[8] J. Kim, S. Kim, S. Choi, and D. Qiao, “CARA: Collision-awarerate adaptation for IEEE 802.11 WLANs,” Proc. IEEE INFOCOM2006. 25TH IEEE International Conference on Computer Commu-nications, pp.1–11, 2006.

[9] S. Biaz and S. Wu, “Loss differentiated rate adaptation in wirelessnetworks,” Proc. 2008 IEEE Wireless Communications and Net-working Conference, pp.1639–1644, 2008.

[10] S. Wu, S. Biaz, and H. Wang, “Rate adaptation with loss diagno-sis on IEEE 802.11 networks,” Int. J. Commun. Syst., vol.25, no.4,pp.515–528, April 2012.

[11] J. Zhang, K. Tan, J. Zhao, H. Wu, and Y. Zhang, “A practicalSNR-guided rate adaptation,” Proc. IEEE INFOCOM 2008 — The27th Conference on Computer Communications, pp.2083–2091,2008.

[12] K. Pelechrinis, I. Broustis, S.V. Krishnamurthy, and C. Gkantsidis,“A measurement-driven anti-jamming system for 802.11 Networks,”IEEE/ACM Trans. Netw., vol.19, no.4, pp.1208–1222, Aug. 2011.

[13] D. Aguayo, J. Bicket, S. Biswas, G. Judd, and R. Morris, “Link-levelmeasurements from an 802.11b mesh network,” ACM SIGCOMMComputer Communication Review, vol.34, no.4, pp.121–132, 2004.

[14] “mac80211,” http://wireless.kernel.org/en/developers/Documentation/mac80211/

[15] “PID,” http://wireless.kernel.org/en/developers/Documentation/mac80211/RateControl/PID/

[16] “Minstrel specification,” http://madwifi-project.org/svn/madwifi/trunk/ath rate/minstrel/minstrel.txt/

[17] W. Yin, P. Hu, J. Indulska, and K. Bialkowski, “Performance ofmac80211 rate control mechanisms,” Proc. 14th ACM InternationalConference on Modeling, Analysis and Simulation of Wireless andMobile Systems — MSWiM’11, pp.427–436, 2011.

[18] D. Xia, J. Hart, and Q. Fu, “Evaluation of the minstrel rate adapta-tion algorithm in IEEE 802.11g WLANs,” Proc. 2013 IEEE Interna-tional Conference on Communications (ICC), pp.2223–2228, 2013.

[19] “Raspberry Pi model B+,” http://www.raspberrypi.org/products/model-b-plus/

[20] “ArduinoBoardUno,” http://arduino.cc/en/Main/arduinoBoardUno/[21] “Iperf,” http://www.iperf.fr/[22] “Wireshark,” https://www.wireshark.org/

Shota Yamashita received the B.E. degreein electrical and electronic engineering fromKyoto University in 2013. He is currently study-ing toward the M.E. degree at the GraduateSchool of Informatics, Kyoto University. His re-search interests include microwave power trans-mission to WLAN devices. He is a studentmember of the IEEE.

Koichi Sakaguchi received the B.E. de-gree in electrical and electronic engineeringfrom Kyoto University in 2014. He is cur-rently studying toward the M.E. degree at theGraduate School of Informatics, Kyoto Univer-sity. His current research interests include full-duplex wireless communication systems.

Yong Huang received the B.S. degree inintegrated mechanical and electrical engineer-ing from Jiangnan University, Jiangsu, China, in2008, the M.S. degree in electric and electronicengineering from the University of Toyama, To-yama, Japan, in 2011, and is currently workingtoward the Ph.D. degree in electrical engineer-ing at Kyoto University, Kyoto, Japan. His cur-rent research interests include rectifying circuitintegration and power electronics of low-powerrectenna design.

Koji Yamamoto received the B.E. degree inelectrical and electronic engineering from KyotoUniversity in 2002, and the M.E. and Ph.D. de-grees in informatics from Kyoto University in2004 and 2005, respectively. From 2004 to2005, he was a research fellow of the Japan So-ciety for the Promotion of Science (JSPS). Since2005, he has been with the Graduate Schoolof Informatics, Kyoto University, where he iscurrently an associate professor. From 2008to 2009, he was a visiting researcher at Wire-

less@KTH, Royal Institute of Technology (KTH) in Sweden. His researchinterests include the application of game theory, spectrum sharing, andM2M networks. He received the PIMRC 2004 Best Student Paper Awardin 2004, the Ericsson Young Scientist Award in 2006, and the Young Re-searcher’s Award from the IEICE of Japan in 2008. He is a member of theIEEE.

1794IEICE TRANS. COMMUN., VOL.E98–B, NO.9 SEPTEMBER 2015

Takayuki Nishio received the B.E. de-gree in Electrical and Electronic Engineeringfrom Kyoto University in 2010. He received theM.I. and Ph.D. degrees in Communications andComputer Engineering, Graduate School of In-formatics from Kyoto University, Kyoto, Japan,in 2012 and 2013, respectively. He is currentlyan Assistant Professor in Communications andComputer Engineering, Graduate School of In-formatics, Kyoto University. His current re-search interests include communication system

design, protocol design particularly MAC and TCP, and QoS and resourcemanagement in wireless networks. He is a member of the IEEE.

Masahiro Morikura received his B.E.,M.E., and Ph.D. degrees in electronics engi-neering from Kyoto University, Kyoto, Japan in1979, 1981 and 1991, respectively. He joinedNTT in 1981, where he was engaged in the re-search and development of TDMA equipmentfor satellite communications. From 1988 to1989, he was with the Communications Re-search Centre, Canada, as a guest scientist.From 1997 to 2002, he was active in the stan-dardization of the IEEE 802.11a based wireless

LAN. His current research interests include WLANs and M2M wirelesssystems. He received the Paper Award and the Achievement Award fromIEICE in 2000 and 2006, respectively. He also received the Education, Cul-ture, Sports, Science and Technology Minister Award in 2007 and MaejimaAward in 2008. Dr. Morikura is now a professor in the Graduate School ofInformatics, Kyoto University. He is a member of the IEEE.

Naoki Shinohara received the B.E. de-gree in electronic engineering, the M.E. andPh.D. (Eng.) degrees in electrical engineeringfrom Kyoto University, Japan, in 1991, 1993and 1996, respectively. He was a research as-sociate in the Radio Atmospheric Science Cen-ter, Kyoto University from 1998. He was aresearch associate of the Radio Science Cen-ter for Space and Atmosphere, Kyoto Univer-sity by recognizing the Radio Atmospheric Sci-ence Center from 2000, and there he was an as-

sociate professor since 2001. He was an associate professor in ResearchInstitute for Sustainable Humanosphere, Kyoto University by recognizingthe Radio Science Center for Space and Atmosphere since 2004. From2010, he has been a professor in Research Institute for Sustainable Hu-manosphere, Kyoto University. He has been engaged in research on SolarPower Station/Satellite and Microwave Power Transmission system. Heis IEEE MTT-S Technical Committee 26 (Wireless Power Transfer andConversion) member, IEEE MTT-S Kansai Chapter TPC member, advisorycommittee member of IEEE Wireless Power Transfer Conference, execu-tive editor of international journal of Wireless Power Transfer (CambridgePress), Japanese committee C member of Radio Science for URSI, chair oftechnical committee on Wireless Power Transfer, communications society,IEICE, chair of Wireless Power Transfer Consortium for Practical Appli-cations (WiPoT), and chair of Wireless Power Management Consortium(WPMc).