Simultaneous Lightwave Information and Power Transfer (SLIPT) for Indoor IoT Applications
aa r X i v : . [ c s . I T ] J un Simultaneous Lightwave Information and PowerTransfer (SLIPT) for Indoor IoT Applications
Panagiotis D. Diamantoulakis ∗ and George K. Karagiannidis ∗∗ Department of Electrical and Computer Engineering, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greecee-mails: { padiaman, geokarag } @auth.gr Abstract —We present the concept of Simultaneous LightwaveInformation and Power Transfer (SLIPT) for indoor Internet-of-Things (IoT) applications. Specifically, we propose novel and fun-damental SLIPT strategies, which can be implemented throughVisible Light or Infrared communication systems, equippedwith a simple solar panel-based receiver. These strategies areperformed at the transmitter or at the receiver, or at both sides,named
Adjusting transmission , Adjusting reception and
Coordi-nated adjustment of transmission and reception , correspondingly.Furthermore, we deal with the fundamental trade-off betweenharvested energy and quality-of-service (QoS), by maximizing theharvested energy, while achieving the required user’s QoS. To thisend, two optimization problems are formulated and optimallysolved. Computer simulations validate the optimum solutionsand reveal that the proposed strategies considerably increase theharvested energy, compared to SLIPT with fixed policies.
I. I
NTRODUCTION
The era of Internet-of-Things (IoT) opens up the opportunityfor a number of promising applications in smart buildings,health monitoring, and predictive maintenance. In the con-text of wireless access to IoT devices, radio frequency (RF)technology is the main enabler. Furthermore, the exponentialgrowth in the data traffic puts tremendous pressure on theexisting global telecommunication networks and the expec-tations from the fifth generation (5G) of wireless networks.However, it is remarkable that most of the data consump-tion/generation, which are related to IoT applications, occursin indoor environments [1]. Motivated by this, optical wirelesscommunication (OWC), such as visible light communications(VLC) or infrared (IR), have been recognized as promisingalternative/complimentary technologies to RF, in order to giveaccess to IoT devices in indoor applications [1]. The data ratesreported for indoor VLC/IR networking are much higher thanthose achieved by WiFi, especially when client and serverare closely located. Apart from the very high data rates [2],the advantages of OWC technologies include: i) increase ofavailable bandwidth, ii) easy bandwidth reuse, iii) increase ofenergy efficiency and considerable energy savings, iv) no RFcontamination, and v) free from RF interference. Moreover,nowadays, light emitting diodes (LEDs) and photodetectors(PDs) tend to be considerably cheaper than their RF counter-parts, while the cost-efficiency is further improved due to thepotential to use the existing lighting infrastructure [3]–[5].Due to the strong dependence of the IoT on wirelessaccess, their applications are constrained by the finite bat-tery capacity of the involved devices [6]. Therefore, energy harvesting (EH), which refers to harnessing energy from theenvironment or other sources and converting to electricalenergy, is a critical part of the operation and maintain of theIoT devices. Energy harvesting is regarded as a disruptivetechnological paradigm to prolong the lifetime of energy-constrained wireless networks, which apart from offering apromising solution for energy-sustainability of wireless nodes,it also reduces the operating expenses (OPEX) [6]. However,the main disadvantage of traditional EH methods is that theyrely on natural resources, such as solar and wind, which areuncontrollable. For this reason, harvesting energy from radiofrequency signals, which also transfer information, seems to bean interesting alternative. In order to enable simultaneous wire-less information and power transfer and increase efficiencyof the utilized resources, two strategies have been proposednamed power-splitting , which is based on the division of thesignals power into two streams, and time-splitting , accordingto which, during a portion of time, the received signal is usedsolely for energy harvesting, instead of decoding [7].Although RF based wireless power transfer is a well in-vestigated topic in the last five years, optical wireless powertransfer (OWPT) is a new topic and only a few works havebeen reported so far in the open literature. In the pioneeringwork of Fakidis et. al. [8], the visible and infra-red partsof the electromagnetic (EM) spectrum was used for OWPT,through laser or LEDs at the transmitter and solar cells atthe receiver side. Also, in [9] and [10] energy harvestingwas performed by using the existing lighting fixtures forindoor IoT applications. Regarding the simultaneous opticalwireless information and power transfer, in [11] the sum ratemaximization problem has been optimized in a downlink VLCsystem with simultaneous wireless information and powertransfer. However, in this paper the utilized energy harvestingmodel does not correspond to that of the solar panel, whereonly the direct current DC component of the modulated lightcan be used for energy harvesting, in contrast to the alternatingcurrent AC component, which carries the information. Theseparation of the DC and AC components was efficientlyachieved by the self-powered solar panel receiver proposedin [12], [13], where it was proved that the use of the solarpanel for communication purposes does not limit its energyharvesting capabilities. Thus, the utilization of the power-splitting in the useful recent work [14], where the receivedphotocurrent is splitted in two parts with each of them toinclude both a DC and a AC part, reduces the EH efficiency.oreover, in [14] an oversimplified energy harvesting modelwas used, assuming that the harvested energy is linearly pro-portional to the received optical power, while an optimizationof the splitting technique was not presented. Furthermore, inthe significant research works [15], [16], a dual-hop hybridVLC/RF communication system is considered, in order toextend the coverage. In these papers, besides detecting theinformation over the VLC link, the relay is also able to harvestenergy from the first-hop VLC link, by extracting the DCcomponent of the received optical signal. This energy canbe used to re-transmit the data to a mobile terminal over thesecond-hop RF link. Also, in [15] the proposed hybrid systemwas optimized, in terms of data rate maximization, while in[16] the packet loss probability was evaluated.In this paper, we present for first time a framework forsimultaneous optical wireless information and power transfer,called from now on as
Simultaneous Lightwave Informationand Power Transfer (SLIPT) , which can be efficiently usedfor indoor IoT applications through VLC or IR systems. Morespecifically, we propose novel and fundamental strategies inorder to increase the feasibility and efficiency of SLIPT,when a solar panel-based receiver is used. These strategiesare performed at the transmitter or at the receiver, or atboth sides, named
Adjusting transmission , Adjusting reception ,and
Coordinated adjustment of transmission and reception .Regarding adjusting transmission two policies are proposed:i)
Time-splitting (TS) , according to which the time frame isseparated in two distinct phases, where in each of them themain focus is either on communication or energy transferand, ii)
Time-splitting with DC bias optimization , which isa generalization of TS. In contrast to RF-based wirelesspowered networks, where the TS strategy and adjustment ofthe related parameters takes place at the receiver’s side, TS inSLIPT refers to the adaptation of specific parameters of thetransmitted signal. Regarding adjusting reception, the
Field-of-view (FoV) adjustment policy is proposed, while accordingto the coordinated adjustment of transmission and receptionstrategy, we propose the simultaneous optimization of theformer policies at both transmitter and receiver, in order tomaximize the harvested energy, while achieving the requiredQuality-of-Service (QoS) (e.g. data rate and signal-to-noiseplus interference ratio (SINR)). Finally, the resulting twooptimization problems are formulated and optimally solved.II. S
YSTEM AND C HANNEL MODEL
We consider the downlink transmission of an OWC system,consisting of one LED and a single user. We also assumethat the user is equipped with the functionality of energyharvesting. The VLC/IR transmitter/receiver design is shownin Fig. 1, while the VLC/IR downlink communication isdepicted in Fig. 2.
A. Optical Wireless Transmission
Let m ( t ) denote the modulated electrical signal that corre-sponds to the bit stream from the information source. A DCbias B is added to m ( t ) to ensure that the resulting signal Fig. 1. SLIPT transceiver designFig. 2. VLC/IR downlink communication is non-negative, before being used to modulate the opticalintensity of the LED and regulate the LED in the properoperation mode. The transmitted optical signal from the LEDis [15] P t ( t ) = P LED [ B + m ( t )] , (1)where P LED is the LED power. The electrical signal variesaround the DC bias B ∈ [ I L , I H ] with peak amplitude A ,where I L is the minimum and I H is the maximum input biascurrents, correspondigly. In order to avoid clipping distortionby the nonlinearity of the LED, by restraining the inputelectrical signal to the LED within the linear region of theLED operation, the following limitation is induced A ≤ min( B − I L , I H − B ) , (2)where min( z, y ) denotes the minimum between z and y . B. Channel Model
The channel power gain is given by [17]–[19] h = L r d R ( ϕ ) T s ( ψ ) g ( ψ ) cos( ψ ) , (3)here L r is the physical area of the photo-detector, d is thetransmission distance from the LED to the illuminated surfaceof the photo-detector, T s ( ψ ) is the gain of the optical filterand g ( ψ ) represents the gain of the optical concentrator, givenby [17], [19] g ( ψ ) = ( ρ sin (Ψ fov ) , ≤ ψ ≤ Ψ fov , , ψ > Ψ fov . (4)with ρ and Ψ fov being the refractive index and FOV, respec-tively. Also in (3), R ( ϕ ) is the Lambertian radiant intensityof the LED, given by R ( ϕ ) = ξ + 12 π cos ξ ϕ, (5)where ϕ is the irradiance angle, ψ is the incidence angle, and ξ = − cos(Φ / ) , (6)with Φ / being the semi-angle at half luminance. C. Received Electrical SINR
The electrical current i r ( t ) at the output of the PD can bewritten as i r = η ( hP t ( t ) + P o ) + n ( t ) = I DC ( t ) + i ( t ) + n ( t ) , (7)where η is the photo-detector responsivity in A / W , P o isthe received optical signal from other sources, e.g. otherneighboring LEDs, I DC is the DC component, i ( t ) is theAC component, and n ( t ) is the additive white Gaussian noise(AWGN), which is created from background shot noise andthermal noise.The AC component i ( t ) is composed of two terms, i.e. i ( t ) = i ( t ) + i ( t ) , where i ( t ) = ηhP LED m ( t ) (8)is due to the dedicated LED, and i ( t ) is due to otherinterfering sources. Thus, the received SINR can be writtenas γ = ( ηhP LED A ) P I + σ , (9)where σ is the noise power and P I is the electrical power ofthe received interference. D. Energy Harvesting Model
As it has already been mentioned the photocurrent consistesof both the DC and AC signals. In order to perform energyharvesting, the DC component is blocked by a capacitorand passes through the energy harvesting branch [12]. Theharvested energy is given by [20] E = f I DC V oc , (10)with f being the fill factor [20] and I DC = I + I being theDC component of the output current, where I = ηh n P LED B (11) is due to the dedicated LED, while I is due to different lightsources, e.g. neighboring LEDs. Also, V oc is V oc = V t ln(1 + I DC I ) , (12)where V t is the thermal voltage and I is the dark saturationcurrent of the PD. Moreover, f is the fill factor, defined as theratio of the maximum power from the solar cell to the productof the open-circuit voltage V oc and I sc ,III. SLIPT S TRATEGIES
In this section we propose fundamental SLIPT strategiesfor use in VLC/IR communication systems. These strategiesare performed either at the transmitter or at the receiver, or atboth sides:
Adjusting transmission , Adjusting reception , and
Coordinated adjustment of transmission and reception . A. Adjusting Transmission
Next, we introduced two policies for the adjusting trans-mission strategy, named
Time-splitting and
Time-splitting withDC Bias Optimization .
1) Time-splitting:
According to the Time-splitting policythe received optical signal is used for a portion of time solelyfor energy harvesting, instead of decoding. During this periodof time the LED transmits by using the maximum DC bias,in order to maximize the harvested energy by the receiver.Thus, assuming time frames of unitary duration, there are thefollowing two distinct phases during a time frame:Phase : The AC component of the received signal isused for information decoding and the DC component forenergy harvesting. Let A and B ∈ [ I L , I H ] denote thepeak amplitude of m ( t ) and DC bias, respectively. DuringPhase 1, the aim is to maximize the received SINR. SinceSINR is an increasing function with respect to A , then A takes its maximum value, which, considering (2) is given by A = I H − I L and similarly, B = I H + I L . The duration ofthis phase is denoted by ≤ T ≤ , which can be optimizedaccording to the QoS requirements. For a specific value of T ,the amount of harvested energy is given by E [1] T S = f T ( ηhP LED I H + I L I ) V t × ln(1 + ηhP LED I H + I L + I I ) . (13)Phase : In the time period − T , the aim is to maximize theharvested energy, which is an increasing function with respectto B . Thus, during Phase 2 the transmitter eliminates the ACpart and maximizes the DC bias, i.e., A = 0 and B = I H ,where A and B ∈ [ I L , I H ] denote the values of A and B ,respectively. Thus, the amount of harvested energy during thisphase, is given by E [2] T S = f (1 − T )( ηhP LED I H + I ) V t × ln(1 + ηhP LED I H + I I ) . (14)Considering both phases, the total harvested energy is givenby E T S = E [1] T S + E [2] T S . (15) ) Time-Splitting with DC Bias Optimization: This policy isa generalization of Time-splitting. During Phase , the DC biasis optimized in order to increase the harvested energy, whileit simultaneously enables information transfer, i.e., A > .In this case, the total harvested energy is given by E TSBO = f T ( ηhP LED B + I ) V t × ln(1 + ηhP LED B + I I ) + E [2] T S , (16)where B is the DC bias during Phase . B. Adjusting Reception
We propose the
Adjustment of the field of view (FOV) policyfor the adjusting reception strategy, in order to balance thetrade-off between harvested energy and SINR. ControllingFOV is particularly important especially when, except forthe used VLC/IR LED, there are extra light sources in theserving area [21], e.g. neighboring LEDs that serve otherusers. For the practical and efficient implementation of thispolicy, electrically controllable liquid crystal (LC) lenses is apromising technology [22].When the aim is to maximize the SINR, the FOV is tunedup to receive the beam of the dedicated LED only (if possible),in order to reduce the beam overlapping. This is achieved bytuning the FOV to the narrowest setting, that allows receptiononly from that LED. On the other hand, when the aim is toachieve a balance between SINR and harvested energy, a widerFOV setting could be selected.For the sake of practicality, we assume that the VLC/IR re-ceiver has discrete FOV settings, i.e. Ψ fov ∈ { Ψ [1]fov , .., Ψ [ M ]fov } .Also, note that except for h , both P I and I are also discretefunctions of Ψ fov , i.e., P I = P I (Ψ fov ) and I = I (Ψ fov ) . C. Coordinated Transmission and Reception Adjustment
Considering (9), (15), and (16), it is revealed that bothSINR and harvested energy -apart from A , B and T -also depend on the selection of Ψ fov , despite the utilizedadjusting transmission technique. This dependence motivatesthe coordinated transmission and reception adjustment, i.e. thecoordination between the strategy III-A1 or III-A2 and III-B,which results in the following two policies, i.e. • Policy 1: Time-splitting with tunable FOV (III-A1 andIII-B) • Policy 2: Time-splitting with DC bias optimization andtunable FOV (III-A2 and III-B)Note that in both policies, during Phase 2, where the aimis to maximize the harvested energy, the FOV setting thatmaximizes E [2] T S should be used. This is not necessarily thewidest setting, because although it increases the receivedbeams (if there are neighboring LEDs), it reduces g ( ψ ) . Onthe other hand, the preferable FOV setting during phase 1,denoted by Ψ fov , , cannot be straightforwardly determined,since it also depends on the required QoS. IV. SLIPT O PTIMIZATION
SLIPT induces an interesting trade-off between harvestedenergy and QoS. In this section, we aim to balance this trade-off by maximizing the harvested energy, while achieving therequired user QoS. In the present work, we focus on thecoordinated adjustment of transmission and reception strategy,which can be considered as a generalization of the otherSLIPT strategies. The following optimization problems canbe formulated, based on the two techniques presented insubsection III-C.Regarding the QoS, two different criteria are taken intoaccount, namely SINR and information rate. Note that thesetwo criteria are not equivalent to each other, when eitherof the two techniques is used, due to the time-splitting.More specifically, since only Phase is used for informationtransmission (the duration of which is T ), the lower bound ofthe capacity is given by [23] R = T log (cid:16) e π γ (cid:17) . (17) A. Time-Splitting with Tunable FOV
The corresponding optimization problem can be expressedas max T, Ψ fov , E TS h s.t. C : R ≥ R th , C : γ ≥ γ th , C : 0 ≤ T ≤ , C : Ψ fov , ∈ { Ψ [1]fov , .., Ψ [ M ]fov } , (18)where R th and γ th denote the information rate SINR andthreshold, respectively. Theorem 1:
The optimal value of T in (18) is given by T ∗ = R th log (cid:16) e ( ηhP LED ( I H − I L )) π ( P I (Ψ ∗ fov , )+ σ ) (cid:17) , (19)where ( · ) ∗ denotes optimality. Proof:
The optimization problem (18) is a combinatorialone. In order to find the optimal solution, all possible valuesof Ψ fov , have to be checked before selecting the value thatmaximizes the harvested energy, E TS h , while satisfying theconstraints C , C , and C . For a specific specific value of Ψ fov , , if ( ηhP LED I H − I L ) P I (Ψ fov , ) + σ < γ th , (20)then the optimization problem is infeasible, since C is notsatisfied. Also, due to constraint C , the following limitationis induced for T , T ≥ R th log (cid:18) e ( ηhP LED IH − IL ) π ( P I (Ψ fov , )+ σ ) (cid:19) . (21)Moreover, the harvested energy is decreasing with respect to T . Thus, the optimal value of T is given by (19) and the proofis completed.Note that if T ∗ > , the optimization problem in (18) isinfeasible, due to C . . Time-Splitting with DC Bias Optimization and tunable FOV The corresponding optimization problem can be formulatedas max B ,A ,T, Ψ fov , E TSBO h s.t. C : R ≥ R th , C : γ ≥ γ th , C : A ≤ min( B − I L , I h − B ) , C : 0 ≤ T ≤ , C : A ≥ , C : I L ≤ B ≤ I H , C : Ψ fov , ∈ { Ψ [1]fov , .., Ψ [ M ]fov } . (22) Proposition 1:
The optimal value of B in (22) belongs inthe range (cid:2) I H + I L , I H (cid:3) . Proof:
The constraint C can be rewritten as C : A ≤ B − I L , C : A ≤ I H − B . (23)For a specific value of B , only one of the constraints C and C is activated. Now, let assume that the optimal solution is B ∗ < I H + I L , for which all the constraints are satisfied. Inthis case, C is activated. However, by setting B = I H + I L the objective function is increased, while the constraints arestill satisfied. Thus, we can infer that that B ∗ is not optimal.Consequently, Proposition 1 has been proved by contradiction.The optimal value Ψ fov , is calculated similarly to thesolution of (18). Regarding the rest optimization variables of(22) they are optimized according to the following theorem: Theorem 2:
For a specific value of Ψ fov , , the optimal valueof T is given by T ∗ = argmax K ≤ T ≤ K ˜ E TSBO h (24)with ˜ E TSBO h being solely a function of T and given by (16),by replacing A and B by A = 1 nhP LED s π ( P I (Ψ fov , ) + σ )(2 R th T − e , (25)and B = I H − A , (26)respectively. Also, K = R th log (cid:18) e ( ηh P LED ( I H − I L )) π ( P I (Ψ ∗ fov , )+ σ ) (cid:19) (27) K = min R th log (cid:0) eγ th π (cid:1) , ! . (28)Finally, the optimal values of A and B are given by (25) and(26), by replacing Ψ fov , and T by Ψ ∗ fov , and T ∗ , respectively. Proof:
Considering Proposition 1 and for a specificvalue of Ψ fov , the optimization problem in (22) can be re-formulated as max B ,A ,T E TSBO h s.t. C : R ≥ R th , C : γ ≥ γ th , C : A + B ≤ I H , C : 0 ≤ T ≤ , C : A ≥ , C : B ≥ I H + I L . (29)The optimization problem in (29) still cannot be easilysolved in its current form, since the objective function as wellas the constraints C and C are not concave. However, itcan be solved with low complexity by using the followingreformulation.First, the inequalities in C and C are replaced by equali-ties. Then, A and B are given by (25) and (26), respectively.By substituting T and B by (25) and (26), C , C , and C of (29) vanish, and the optimization problem is rewritten as max B ,A,T ∀ n ˜ E TSBO h s.t. C : T ≤ R th log ( eγ th2 π ) , C : 0 ≤ T ≤ , C : T ≥ R th log (cid:18) e ( ηh P LED( IH − IL ) ) π ( PI + σ ) (cid:19) , (30)which is equivalent to (24), and, thus, the proof is completed.V. S IMULATIONS AND DISCUSSION
We assume the downlink VLC/IR system of Fig. 2, wherethe user is located in a distance d = 1 . m from the LED, ψ = 0 , and the transmitter plane is parallel to the receiver one,i.e., ϕ = ψ . In the same room there are N other LEDs, whichsimultaneously use the same frequency band. The distancebetween each of them and from the dedicated LED is D = 1 . m. We also assume f = 0 . , P LED = 20
W/A, Φ / = 60 deg, σ = 10 − A , L r = 0 . m , η = 0 . A/W, I = 10 − A, I L = 0 A, I H = 12 mA [15], T s = 1 , ρ = 1 . , γ th = 10 dB, and two settings for the FOV, i.e., Ψ fov ∈ { , } deg,are considered.Regarding the neighboring LEDs, we assume that the DCbias and the peak amplitude are given by A ′ n = B ′ n = 6 mA, ∀ n ∈ { , ..., N } , while the rest parameters are equal to thoseof the dedicated LED. Furthermore, the channel between themand the user’s receiver, denoted by h n is modeled accordingto (3), using the corresponding parameters. Thus, when thewidest FOV setting is selected, P I and I are given by P I = N X n =1 ( ηh n P LED A ′ n ) (31)and I = N X n =1 ηh n P LED B ′ n , (32)otherwise their values are zero.The performance of both optimized policies of Section III-Care compared for N = 1 , while they are also presented against H a r v e s t e d E n er gy ( m W ) R th (bps/Hz) Policy 2 Policy 1 A =B =6 mA, T =1, fov,1 =30 deg A =B =6 mA, T =1, fov,1 =50 degInfeasible for these values of R th Fig. 3. Harvested energy vs R th for N = 1 . H a r v e s t e d E n er gy ( m W ) N Policy 2 Policy 1 A =B =6 mA, T =1, fov,1 =30 deg Fig. 4. Harvested energy vs N for R th = 7 bps/Hz. the case of fixed A , B , T , and Ψ fov , , which is consideredas the baseline policy. More specifically, in Fig. 3 the harvestedenergy is plotted against the rate threshold. As it is observed,both policies significantly outperform the baseline for bothvalues of Ψ fov , . Regarding the baseline, the value Ψ fov , = 50 deg reduces the harvested energy compared to Ψ fov , = 30 deg, because g ( ψ ) decreases and thus, cancels the benefitof receiving the beam of the neighboring LED. Also, thebaseline policy with Ψ fov , = 50 deg is infeasible for mediumand high values of R th , because the rate threshold cannot bereached, due to the received interference. Interestingly, Policy2 outperforms Policy 1, especially for the high region of R th , which is due to the extra degrees of freedom. Similarconclusions can be obtained by Fig. 4, where the harvestedenergy is plotted against the number of neighboring LEDS.For this specific setup, the baseline with Ψ fov , = 50 deg is notfeasible, and, thus, it is omitted. We notice here that for a smallnumber of neighboring LEDs, the harvested energy remainsconstant, since the receiver prefers the smallest FOV setting.However, as the number of neighboring LEDs increases, thereceiver prefers the widest FOV setting and the harvested energy increases with the increase of LEDs.R EFERENCES[1] M. Ayyash, H. Elgala, A. Khreishah, V. Jungnickel, T. Little, S. Shao,M. Rahaim, D. Schulz, J. Hilt, and R. Freund, “Coexistence of WiFiand LiFi toward 5G: Concepts, Opportunities, and Challenges,”
IEEECommun. Mag. , vol. 54, no. 2, pp. 64–71, Feb. 2016.[2] M. S. A. Mossaad, S. Hranilovic, and L. Lampe, “Visible Light Com-munications Using OFDM and Multiple LEDs,”
IEEE Trans. Commun. ,vol. 63, no. 11, pp. 4304–4313, Nov. 2015.[3] M. Kavehrad, “Sustainable Energy-Efficient Wireless Applications Us-ing Light,”
IEEE Commun. Mag. , vol. 48, no. 12, pp. 66–73, Dec. 2010.[4] S. Arnon, Ed.,
Visible Light Communication . Cambridge UniversityPress, 2015.[5] D. Bykhovsky and S. Arnon, “Multiple Access Resource Allocation inVisible Light Communication Systems,”
J. Lightw. Technol. , vol. 32,no. 8, pp. 1594–1600, Apr. 2014.[6] S. Sudevalayam and P. Kulkarni, “Energy Harvesting Sensor Nodes:Survey and Implications,”
IEEE Commun. Surveys Tuts. , vol. 13, no. 3,pp. 443–461, Jul. 2010.[7] S. Nikoletseas, Y. Yang, and A. Georgiadis, Eds.,
Wireless PowerTransfer Algorithms, Technologies and Applications in Ad Hoc Com-munication Networks . Springer, 2016.[8] J. Fakidis, S. Videv, S. Kucera, H. Claussen, and H. Haas, “IndoorOptical Wireless Power Transfer to Small Cells at Nighttime,”
J. Lightw.Technol. , vol. 34, no. 13, pp. 3236–3258, Jul. 2016.[9] C. Carvalho and N. Paulino, “On the Feasibility of Indoor LightEnergy Harvesting for Wireless Sensor Networks,”
Procedia Technology ,vol. 17, pp. 343–350, 2014.[10] A. Nasiri, S. A. Zabalawi, and G. Mandic, “Indoor Power Harvestingusing Photovoltaic Cells for Low-Power Applications,”
IEEE Trans. Ind.Electron. , vol. 56, no. 11, pp. 4502–4509, Nov. 2009.[11] Y. Li, N. Huang, J. Wang, Z. Yang, and W. Xu, “Sum Rate Maximizationfor VLC Systems with Simultaneous Wireless Information and PowerTransfer,”
IEEE Photon. Technol. Lett. , vol. 29, no. 6, pp. 531–534,Mar. 2017.[12] Z. Wang, D. Tsonev, S. Videv, and H. Haas, “Towards Self-PoweredSolar Panel Receiver for Optical Wireless Communication,” in
Proc.IEEE International Conference on Communications (ICC) , Jun. 2014,pp. 3348–3353.[13] ——, “On the Design of a Solar-Panel Receiver for Optical WirelessCommunications with Simultaneous Energy Harvesting,”
IEEE J. Sel.Areas Commun. , vol. 33, no. 8, pp. 1612–1623, Aug. 2015.[14] H. Sandalidis, A. Vavoulas, T. Tsiftsis, and N. Vaiopoulos, “Illumination,Data Transmission and Energy Harvesting: The Threefold Advantage ofVLC,” 2017.[15] T. Rakia, H. C. Yang, F. Gebali, and M. S. Alouini, “Optimal Design ofDual-Hop VLC/RF Communication System with Energy Harvesting,”
IEEE Comm. Lett. , vol. 20, no. 10, pp. 1979–1982, Oct. 2016.[16] ——, “Dual-Hop VLC/RF Transmission System with Energy HarvestingRelay under Delay Constraint,” in
Proc. IEEE Globecom Workshops ,Dec. 2016, pp. 1–6.[17] T. Komine and M. Nakagawa, “Fundamental Analysis for Visible-LightCommunication System Using LED Lights,”
IEEE Trans. Consum.Electron. , vol. 50, no. 1, pp. 100–107, Feb. 2004.[18] H. Ma, L. Lampe, and S. Hranilovic, “Coordinated Broadcasting forMultiuser Indoor Visible Light Communication Systems,”
IEEE Trans.Commun. , vol. 63, no. 9, pp. 3313–3324, Sep. 2015.[19] J. M. Kahn and J. R. Barry, “Wireless Infrared Communications,”
Proc.IEEE , vol. 85, no. 2, pp. 265–298, Feb. 1997.[20] C. Li, W. Jia, Q. Tao, and M. Sun, “Solar Cell Phone Charger Perfor-mance in Indoor Environment,” in
Proc. IEEE 37th Annual NortheastBioengineering Conference (NEBEC) , 2011, pp. 1–2.[21] A. Al-Ghamdi and J. Elmirghani, “Performance and Field of View Opti-mization of an Optical Wireless Pyramidal Fly-Eye Diversity Receiver,”
Journal of optical communications , vol. 23, no. 6, pp. 215–222, 2002.[22] Y. S. Tsou, Y. H. Lin, and A. C. Wei, “Concentrating PhotovoltaicSystem using a Liquid Crystal Lens,”
IEEE Photon. Technol. Lett. ,vol. 24, no. 24, pp. 2239–2242, Dec. 2012.[23] J. B. Wang, Q. S. Hu, J. Wang, M. Chen, and J. Y. Wang, “Tight Boundson Channel Capacity for Dimmable Visible Light Communications,”