Reusing Wireless Power Transfer for Backscatter-assisted Cooperation in WPCN
aa r X i v : . [ c s . N I] J u l Reusing Wireless Power Transfer forBackscatter-assisted Cooperation in WPCN
Wanran Xu, Suzhi Bi, Xiaohui Lin, and Juan Wang
College of Information Engineering, Shenzhen University,Shenzhen, Guangdong, 518060, China {xuwanran2016,bsz,xhlin,juanwang}@szu.edu.cn
Abstract.
This paper studies a novel user cooperation method in a wire-less powered communication network (WPCN), where a pair of closelylocated devices first harvest wireless energy from an energy node (EN)and then use the harvested energy to transmit information to an accesspoint (AP). In particular, we consider the two energy-harvesting usersexchanging their messages and then transmitting cooperatively to theAP using space-time block codes. Interestingly, we exploit the short dis-tance between the two users and allow the information exchange to beachieved by energy-conserving backscatter technique. Meanwhile the con-sidered backscatter-assisted method can effectively reuse wireless powertransfer for simultaneous information exchange during the energy har-vesting phase. Specifically, we maximize the common throughput throughoptimizing the time allocation on energy and information transmission.Simulation results show that the proposed user cooperation scheme caneffectively improve the throughput fairness compared to some represen-tative benchmark methods.
Wireless device battery life has always been a key problem in modern wire-less communication. Frequent battery replacement/recharging may bring lotsof inconvenience and cause high probability of communication interruption. Toovercome the above difficulties, RF-enabled wireless energy transfer (WET) tech-nique has recently drawn greater attention [1–3], which can charge wireless de-vices with continuous and stable energy through the air.One useful application of WET is wireless powered communication network(WPCN) [4–10], where wireless devices (WDs) transmit information using theenergy harvested from energy node. Specifically, [4] proposed a harvest-then-transmit protocol in WPCN where one hybrid access point (HAP) with single-antenna first broadcasts energy to all users, then allows users to take turnsto perform wireless information transmission (WIT). [5] studied the placementoptimization when each pair of EN and AP is colocated and integrated as ahybrid access point. However, all the above works consider using a HAP forperforming both WET and WIT. This WPCN model will inevitable sufferedfrom a so-called “doubly near-far” problem, such that the far user can receive less wireless power than the near user, but needs to consume more energy totransmit information for achieving the same communication performance.To enhance user fairness, [11] proposed a two-user cooperation scheme wherethe near user helps relay the far user’s information to the HAP. [12] considered acluster-based user cooperation in a WPCN with a multi-antenna HAP. However,using a single HAP is the essential cause of the user unfairness problem. Tofurther enhance system performance, [13] considered using separate EN andinformation AP. Specifically, the WDs first harvest energy from the EN andthen use distributed Alamouti code to jointly transmit their information to theAP. Thanks to the achieved cooperative diversity gain, the cooperation schemecan effectively improve the throughput performance. However, the two WDsmay need to consume considerable amount of energy and time on informationexchange, which may constrain the overall communication performance of theenergy-constrained devices.Besides, a newly emerged low-cost ambient backscatter (AB) communicationtechnique provides an alternative method to reduce such cooperation system’soverhead [14–16]. Specifically, AB allows WDs to transmit information by pas-sively backscatter environment RF signals, e.g., WiFi and cellular signals, inthe neighbouring area. Several recent works have focused on improving the datarates of AB, such as applying new signal detection methods and more advancedbackscattering circuit designs [15, 16]. However, due to the dependency on time-varying environment RF signals, backscatter technology suffers many problems,e.g., sensitive to transmission distance and uncontrollable transmissions.In this paper, we present a novel user cooperation method assisted bybackscatter in WPCN. As shown in Fig. 1, we consider a similar setup in [13],where two devices first harvest WET from the HAP and then transmit jointly tothe AP by forming a virtual antenna array. However, unlike the conventional in-formation exchange in [13], we allow the two closely-located WDs to use backscat-ter communication to exchange their messages in a passive manner during theWET stage. The key contributions of this passage are summarized as follows:1. We propose a novel user cooperation method assisted by backscatter commu-nication during the information exchange stage. By reusing the WET signal,the proposed method can achieve simultaneous energy harvesting and infor-mation exchange, and thus potentially improves the throughput performanceof the energy-constrained system.2. We derive the individual throughput of the two WDs for the proposedbackscatter-assisted cooperation method, and formulate an optimizationproblem that maximizes the minimum data rates (common throughput) be-tween the two users. By optimizing the time allocation on WET and WIT,we can effectively enhance the throughput fairness of the system.3. We show that the throughput maximization problem can be cast as a convexoptimization, such that its optimum can be efficiently obtained. By compar-ing with some representative benchmark methods, we show that the proposedbackscatter-assisted cooperation can effectively improve the throughput per-formance under various practical network setups. (cid:1)(cid:2) (cid:3) (cid:1)(cid:2) (cid:4) (cid:5)(cid:6) (cid:7)(cid:8) (cid:5)(cid:9)(cid:10)(cid:11)(cid:12)(cid:13)(cid:14)(cid:15)(cid:11)(cid:16)(cid:9)(cid:17)(cid:18)(cid:10)(cid:11) (cid:19)(cid:16)(cid:20)(cid:21)(cid:17)(cid:20)(cid:16)(cid:15)(cid:15)(cid:10)(cid:11)(cid:14)(cid:20)(cid:22)(cid:23)(cid:23)(cid:24)(cid:9)(cid:25)(cid:20)(cid:16)(cid:15)(cid:25)(cid:22)(cid:9) (cid:26)(cid:9)(cid:18)(cid:22)(cid:11)(cid:23)(cid:16)(cid:15)(cid:25)(cid:22)(cid:9)(cid:14)(cid:15)(cid:11)(cid:16)(cid:9)(cid:17)(cid:23)(cid:25)(cid:17)(cid:17)(cid:25)(cid:22)(cid:9) (cid:5)(cid:6) (cid:1) (cid:14)(cid:1)(cid:2) (cid:3) (cid:27)(cid:1)(cid:2) (cid:4) (cid:14)(cid:14)(cid:14)(cid:14)(cid:14)(cid:1)(cid:5)(cid:28) (cid:1)(cid:2) (cid:3) (cid:27)(cid:1)(cid:2) (cid:4) (cid:1) (cid:7)(cid:8)(cid:1)(cid:26)(cid:28) (cid:1)(cid:2) (cid:4) (cid:1) (cid:14)(cid:1)(cid:2) (cid:3) (cid:19)(cid:16)(cid:20)(cid:21)(cid:17)(cid:20)(cid:16)(cid:15)(cid:15)(cid:10)(cid:11) (cid:29)(cid:5) (cid:1) (cid:2) (cid:1) (cid:3) (cid:1) (cid:4)(cid:3) (cid:1) (cid:4)(cid:4) (cid:1) (cid:5)(cid:3) (cid:1) (cid:5)(cid:4) (cid:1)(cid:2) (cid:3) (cid:27)(cid:1)(cid:2) (cid:4) (cid:1) (cid:7)(cid:8)(cid:1)(cid:26)(cid:28) (cid:1)(cid:2) (cid:3) (cid:1) (cid:14)(cid:1)(cid:2) (cid:4) (cid:19)(cid:16)(cid:20)(cid:21)(cid:17)(cid:20)(cid:16)(cid:15)(cid:15)(cid:10)(cid:11) (cid:5)(cid:6) (cid:1) (cid:1)(cid:2) (cid:3) (cid:27)(cid:1)(cid:2) (cid:4) (cid:1)(cid:5)(cid:28) (cid:30) (cid:3)(cid:4) (cid:30) (cid:4)(cid:3) (cid:30) (cid:5)(cid:3) (cid:30) (cid:5)(cid:4) (cid:30) (cid:3)(cid:7) (cid:30) (cid:4)(cid:7)
Fig. 1.
The proposed user cooperation method and operating protocol.
As shown in Fig. 1, we consider a WPCN consisting of an EN, two WDs, and aninformation AP, where all the devices have single antenna each. The EN is as-sumed to have stable energy supply and able to broadcast RF energy at constantpower P . Besides, it has a time-division-duplexing (TDD) circuit structure toswitch between energy transfer and communication, e.g, for performing channelestimation. The two WDs have no other embedded energy source thus need toharvest RF energy for performing information transmission to the AP.The circuit block diagram of a WD is shown in Fig. 2. With the two switches S and S , a WD can switch flexibly among three operating modes as follows.1. RF (active) communication mode ( S = 0): the antenna is connected to theRF communication circuit and WD is able to transmit or receive informa-tion using conventional RF wireless communication techniques, e.g., QAMmodulation and coherent detection.2. energy harvesting mode (S = 1 and S is open): the antenna is connectedto the energy harvesting circuit, which can convert the received RF signal toDC energy and store in a rechargeable battery. The energy is used to powerthe operations of all the other circuits.3. backscatter (passive) communication mode (S =1 and S is closed): the an-tenna is connected to backscatter communication and energy harvesting cir-cuits. In this case, the WD transmits information passively by backscat-tering the received signal. Specifically, by setting the switch S = 0, theimpedance-matching circuit absorbs most of the received signal such that a“0” is transmitted; otherwise when S = 1, due to the imbalance of trans-mission line impedance, the received signal is reflected and broadcasted by Energy harvesting circuit
Backscatter data transmitterBackscatter data receiver
RF communication circuit
Data flow
Energy flow S S Backscatter circuit
Fig. 2.
Circuit block diagram of the RF-powered backscatter wireless device. the antenna such that a “1” is transmitted. Meanwhile, non-coherent detec-tion techniques, e.g., energy detector [17], can be used to decode backscattertransmissions from other devices.Notice that a WD can harvest RF energy simultaneously when the backscat-ter circuit transmits or receives information. Specifically, as shown in Fig. 3, apower splitter is used to split the received RF signal into two parts. We denotethe portion of signal power for backscatter communication by (1 − β ), where β ∈ [0 , β for energy harvesting (EH). The received signal iscorrupted by an additive noise N ∼ CN (0 , σ ) at the receiver antenna. Besides,the power splitting circuit and the information decoding circuit are also intro-duced by an additional noise N s ∼ CN (0 , σ s ), which is assumed independent ofthe antenna noise N . As a result, the equivalent noise power for informationdecoding is (1 − β ) σ + σ s . The value of β can be adjusted according to differentreceive signal power and is assumed constant for the time being. EN Energy transfer
Backscatter communication WD WD Power
Splitter
Information
Decoding (cid:163) (cid:163) + (cid:49) (cid:19) + (cid:49) (cid:54) Energy
Harvest
Fig. 3.
The power splitter structure used during backscatter communication
Evidently, backscatter communication does not need to generate RF carriersignals locally and insteads using simple energy encoding/decoding circuits, thusis more energy-efficient than conventional active wireless communication. How-ever, its application is often limited by the strength of the ambient RF signaland its short communication range (within a couple of meters) due to the weak signal strength after reflection. In the following, we propose a method to reuseWET to achieve controllable backscatter communication in WPCN.
We consider a block fading channel model where all the channels are reciprocaland the channel gains remain constant during each transmission block of du-ration T . At the beginning of a transmission block channel estimation (CE) isperformed within a fixed duration t . Then, a three-stage operating protocol isused in the remainder of a tagged transmission block, as shown in Fig. 1. Specif-ically, in the second stage, the EN continuous to broadcast energy for t amountof time, during which the WD and WD take turns to backscatter their localinformation for t and t amount of time, respectively, where t = t + t .With the power splitter structure in Fig. 3, each of the two WDs can decodeinformation from the other’s transmission and harvest RF energy simultaneously.Notice that we neglect the backscatter signal received by the AP due to the muchlarger distance separation between a WD and the AP in practice. In the thirdstage of length t , the two users transmit jointly their information to the AP.Specially, t amount of time is allocated to transmit user WD ’s information,and the rest of t is for transmitting WD ’s information. Accordingly, we havea total time constraint t + t + t + t + t + t = T. (1)For convenience, we normalize T = 1 in the sequel without loss of generality.We denote the complex channel coefficient between WD and WD as α .Similarly, the other channel coefficients are denoted as α E , α E , α A , α A , α .In the CE stage, user WD and WD broadcast their pilot signals, so that ENhas the knowledge of α E and α E , the AP knows α A and α A , and user WD (WD ) knows α ( α ) respectively. Then, each node feeds back their known CSIto a control point, which calculates and broadcasts the optimal time allocation( t ∗ , t ∗ , t ∗ , t ∗ , t ∗ ) to all the nodes in the network. During the WET phase, we denote the baseband equivalent pseudo-random en-ergy signal transmitted by the EN as x ( t ) with E [ | x ( t ) | ] = 1. The receivedsignal at WD i , i = 1 ,
2, is then expressed as y (1) i ( t ) = p P α Ei x ( t ) + n i ( t ) , (2)where n i ( t ) denotes the receiver noise at WD i with n i ( t ) ∼ CN (0 , N ). It isassumed that P is sufficiently large such that the energy harvested due to the receiver noise is negligible. Hence, the amount of energy harvested by WD andWD can be expressed as E (1)1 = P ηh E t , E (1)2 = P ηh E t , (3)where 0 < η < first transmits its information to WD for t amount of time. We assume a fixed data transmission rate R b bit/s, thus theduration of transmitting a bit is 1 /R b second. In particular, when WD transmitsa bit 0, the switch 3 is open, and WD receives only the energy signal from theHAP and WD . Otherwise, when WD transmits a bit 1, the received signal atWD is a combination of both the HAP’s energy and the reflected signal fromWD . Those signals can be jointly expressed as y (2)2 ( t ) = α E p P x ( t ) + Bµ α E α p P x ( t ) + n (2)2 ( t ) , (4)where µ denotes the backscatter reflection coefficient of WD , and B denotesthe information bit transmitted by WD through backscattering. Due to the useof the power splitter at each user, the energy and information signals receivedby WD can be respectively expressed as y (2)2 ,E ( t ) = p βy (2)2 ( t ) , y (2)2 ,I ( t ) = p (1 − β ) y (2)2 ( t ) . (5)It is assumed that the probabilities of transmitting 0 and 1 are equal. Thereforethe harvested energy by WD can be expressed as E (2)2 = 12 ηβt ( E [ | y (2)2 , ( t ) | ] + E [ | y (2)2 , ( t ) | ])= ηβt P [ h E + µ α E α E α + 12 µ h E h ] . (6)Notice in (4), we assume that the signals received directly from the HAPand that reflected from WD are uncorrelated due to the random phase changeduring backscatter. We denote the sampling rate of WD ’s backscatter receiveras S , such that it sampled N = SR b samples during the transmission of a bitinformation, where R b denotes the fixed backscatter rate in bits per second. Inthe following lemma, we derive the bit error rate (BER) of a backscatter receiverusing an optimal energy detector. Lemma 3.1 : The BER of WD for the considered backscatter communicationwith an optimal energy detector is P e = 12 erf c " (1 − β ) P √ N − β ) σ + 4 σ s ( µ h E h ) . (7) Proof : The proof is omitted here due to the space limitation.As the backscatter communication can be modeled as a binary symmetricchannel, the channel capacity (in bit per channel use) of the transmission fromWD to WD can be expressed as C = 1 + (1 − P e ) log (1 − P e ) + P e log ( P e ) . (8)By symmetry, we can get the BER and channel capacity from WD to WD as P e and C , where P e = 12 erf c " (1 − β ) P √ N − β ) σ + 4 σ s ( µ h E h ) , (9) C = 1 + (1 − P e ) log (1 − P e ) + P e log ( P e ) . (10)As a result, the communication rates of WD and WD in this stage can beexpressed as function of time allocation t = [ t , t , t , t , t , t ] R (2)1 ( t ) = R b t C , R (2)2 ( t ) = R b t C . (11)In the last WIT stage, we assume that both user WD and WD exhaust theharvested energy, and each transmits with a constant power. Then, the transmitpowers of WD and WD is P = E (1)1 + E (2)1 t , P = E (1)2 + E (2)2 t , (12)where t = t + t . In this stage, the two users use Alamouti STBC transmitdiversity scheme [13] for joint information transmission with t = t , wherethe achievable data rates from user WD to AP is R (3)1 ( t ) = t (1 + P h A σ + P h A σ ) . (13)Likewise, we have R (3)1 ( t ) = R (3)2 ( t ) for user WD . With the considered cooperation scheme, the overall achievable date rates of userWD and WD are R ( t ) = min n R (2)1 ( t ) , R (3)1 ( t ) o , R ( t ) = min n R (2)2 ( t ) , R (3)2 ( t ) o . (14)In this paper, we focus on maximizing the common throughput (max-minthroughput) of two users by jointly optimizing the time allocated to the HAP,WD and WD .( P
1) : max t min ( R ( t ) , R ( t ))s. t. t + t + t + t + t + t = 1 ,t , t , t , t , t ≥ . (15)By introducing an auxiliary variable Z , (P1) can be equivalently written as ( P
2) : max t ,Z Z s. t. t + t + t + t + t + t = 1 ,t , t , t , t , t ≥ ,Z ≤ R (2)1 ( t ) , Z ≤ R (2)2 ( t ) ,Z ≤ R (3)1 ( t ) , Z ≤ R (3)2 ( t ) . (16)Notice that R (3)1 ( t ) , R (3)2 ( t ) are both concave functions, therefore (P2) is a convexproblem whose optimum can be efficiently solved using off-the-shelf algorithms,e.g., interior point method. In this section, we evaluate the performance of the proposed backscatter-assistedcooperation with that without backscatter in [11] (the No B.S. scheme). Unlessotherwise stated, it is assumed that users are separated by 4 meters. The noisepower N is set 10 − W for all receivers, and the additional noise power forID circuit is N s = 10 − W. The transmit power of EN is P = 1W, and thewireless channel gain h ij = G A ( ∗ πdf d ) λ , where ij ⊂ { E E
2; 1 A ; 2 A ; 12; 21 } , f d denotes 915 MHz carrier frequency, λ = 2 . GA = 2, the signal bandwidth is 10 Hz, andthe sampling rate S = 6 × . Without loss of generality, we assume the powersplitting factor β = 0 .
7, energy harvesting efficiency η = 0 .
8, and backscatterreflection coefficient µ = µ = 0 . h /h M a x - m i n T h r oughpu t ( bp s / H z ) R b =100kR b =50k R b =100kR b =50kNo-B.S Fig. 4.
The impact of user-to-AP channel disparity to the common throughput per-formance
Fig. 4 shows the impact of user-to-AP channel disparity to the optimalcommon throughput performance. Here we set h E = h E = 8 . × − , fix h A = 8 . × − as a constant and show the performance when h A becomessmaller. Notice that when h A /h A changes from 1 to 10, all the schemes show adecreasing trend in system performance, which is due to the weaker user-to-APchannel. In particular, the backscatter communication rate R b has a significantlyeffect on system performance. As we can see in Fig. 4, when R b = 50 kbps theperformance of backscatter system is similar to the case without backscattering.However, the former decrease faster than the latter, which is because the worsechannel h A will affect both two user’s communication rate. When R b increasesto 100 kbps, the system performance increases and outperforms the one withoutbackscattering in all cases. This is because the higher backscatter communica-tion rate can effectively reduce the time spent on cooperation, thus leaving moretime on energy harvesting and information transmission to the AP. D M a x - m i n T h r oughpu t ( bp s / H z ) R b =100kR b =50k R b =100kR b =50kNo-B.S Fig. 5.
The impact of inter-user channel to the common throughput performance
Fig. 5 further studies the impact of inter-user channel strength to the through-put performance. Here we set h E = h E = 8 . × − , and h A = h A =8 . × − . We consider the distance between WD and WD varies form 1 to 5m.It is observed that the max-min throughput of all schemes decreases with D ,due to the worse inter-user channel h . We can see that cooperation withoutbackscatter performs relatively well when d is small. However, as the distancebetween users increases, its performance quickly degrades due to the larger timeand energy consumed on information exchange, and in general is worse than theproposed backscatter-assisted method, e.g., when 2 < d < . R b = 100 kbps.When the inter-user distance becomes very large, e.g., larger than 4 . assisted cooperation has advantage over that without backscattering when theinter-user channel is relatively weak. This paper studied a novel user cooperation method in a two-user WPCN as-sisted by backscatter communication. In particular, the considered backscatter-assisted method reuses wireless power transfer for simultaneous information ex-change during the energy harvesting phase, which can effectively save the energyand time consumed by conventional active transmission schemes. We derivedthe maximum common throughput of the proposed method through optimiz-ing the time allocation on WET and WIT. By comparing with existing bench-mark method, we showed that the proposed method can effectively improve thethroughput fairness performance under various practical network setups.
Acknowledgement
The work of S. Bi was supported in part by the National Natural ScienceFoundation of China under Project 61501303, the Foundation of ShenzhenCity under Project JCYJ20160307153818306 and JCYJ20170818101824392, theScience and Technology Innovation Commission of Shenzhen under Project827/000212, and the Department of Education of Guangdong Province underProject 2017KTSCX163. The work of X. H. Lin was supported by researchGrant from Guangdong Natural Science Foundation under the Project number2015A030313552. X. H. Lin is the corresponding author of this paper.