Energy-Efficient RIS-Assisted Satellites for IoT Networks
11 Energy-Efficient RIS-assisted Satellites forIoT Networks
K¨urs¸at Tekbıyık,
Student Member, IEEE,
G¨unes¸ Karabulut Kurt,
Senior Member, IEEE,
Halim Yanikomeroglu,
Fellow, IEEE
Abstract —The use of satellites to provide ubiquitous coverageand connectivity for the densely deployed Internet of Things(IoT) networks are expected to be a reality in the emerging6G networks. Yet, the low battery capacity is of IoT nodesconstitute a problem for the direct connectivity to satellitesthat are located up to 2000 km altitude. As an architecturalnovelty, in this paper we propose the use of reconfigurableintelligent surface (RIS) units to help with the path loss associatedwith high transmission distances. The use of RIS units throughbroadcasting and beamforming approaches are shown to providea significant gain in terms of their signal transmission. These unitscan be placed on the reflectarrays that are already available onthe satellites. It is shown that RIS-assisted satellites can provideup to times higher downlink and uplink achievable rates forIoT networks. Index Terms —Reconfigurable intelligent surfaces, LEO satel-lites, energy-efficient IoT networks.
I. I
NTRODUCTION
Around 100 billion devices are expected to be connectedin a massive ecosystem by vendors and business advisorycompanies, [1, 2]. Internet of Things (IoT) networks areexpected to grow with approximately 20 percent compoundannual growth rates [3]. By leveraging ubiquitous IoT net-works, it is possible to enhance the efficiency in industries suchas transportation, health, and maritime. Another prominentfeature of massive IoT networks is to improve the qualityof life. Although ubiquitous and massive IoT networks havetremendous development and appealing features for human lifeand industry, it is obviously foreseen that the backhauling forthese ultra-massive networks require more attention to dealwith the connection problem due to 4 billion people withoutan Internet connection [4]. To acquire wide coverage for theInternet, satellite recently gains more attention. Moreover,supporting IoT services with satellite networks is a prominentresearch topic for both academia and industry.
A. Related Works
Recently, an architecture has been proposed for the useof geostationary orbit (GEO) satellites in narrowband-IoTapplications [5, 6]. However, it should be noted that GEOsatellites are exposed to very high path losses and very longdelay due to their relatively long distance from the ground.To eliminate the drawback of GEO satellites, low-Earth orbit
K. Tekbıyık and G.K. Kurt are with the Department of Electronics andCommunications Engineering, ˙Istanbul Technical University, ˙Istanbul, Turkey,e-mails: { tekbiyik, gkurt } @itu.edu.trH. Yanikomeroglu is with the Department of Systems and Computer Engi-neering, Carleton University, Ottawa, Canada, e-mail: [email protected] (LEO) satellite constellations for IoT is introduced in [7, 8]because LEO satellite orbits are closer to the Earth than that ofGEO satellites. Thus, LEO satellites require less transmissionpower to keep required signal-to-noise ratio (SNR) for propercommunication. But,, LEO constellations bring the drawbackswhich are steerable antenna and tracking requirements dueto motion with relatively high velocity. Obviously, theserequirements are unlikely to meet such computational andhardware complexity for IoT devices with low-complexityhardware and low battery capacity. In order to remove theeffects of these drawbacks from IoT devices, it is proposed touse reconfigurable intelligent surfaces (RISs) in LEO satellitesin this study. RISs intelligently adjust the phase shifts ofelements in order to maximize the received power [9]. Themost appealing feature of RIS is that they comprise of passivecircuit elements completely. Furthermore, complex processingor coding is not required by RISs. Moreover, its effectivenessis also demonstrated by experimental measurements in [10,11].LEO satellites do not require a serious hardware improve-ment, as RISs have a simple hardware structure consistingof passive circuit elements [12]. Moreover, the hardwarecomplexity of IoT devices can be reduced as RISs allow signalprocessing to be performed in the transmission environmentrather than on the devices [9]. Thus, battery-constrained IoTdevices can operate for a longer duration. Furthermore, RISsare able to enable energy efficient wireless communicationwhile keeping the quality of service (QoS) same [13]. AsRIS units provide the energy efficiency and low hardwarecomplexity for wireless communications, the use of RIS canbe a game-changer for satellite-IoT systems. Furthermore, ourprevious work [12] demonstrates that error probability can bereduced in LEO inter-satellite links by utilizing RIS. B. Motivations
As the number of IoT applications increases, along withalmost exponentially increasing number of devices, ubiquitouscoverage requirement becomes more apparent. The emergingLEO satellite networks provide an appealing means for thisconnectivity requirement. With a high number of satellites innear Earth constellations, the likelihood of getting continuousservice from a satellite is planned to be really high. ProvidingIoT services over LEO networks constitute a synergy that hasbeen under investigation not only by the research communitybut also by the standardization organizations, such as 3GPP,and private companies, such as Satelliot. Especially withthe 5G standards, following the 3GPP TR 36.763 [14], IoT a r X i v : . [ c s . I T ] J a n Tx Antenna
RIS coated Solar Panel
IoT-rich environment
Fig. 1. It is possible to enhance the QoS for satellite-IoT systems by utilizing RIS. Therefore, the required power can be reduced for the same data rate anderror probability. deployments over satellites will be a reality in the near future,not only targeting rural areas that do not get sufficient coveragebut also improving the capacity in highly populated densedeployments in mega-cities.As the IoT devices have limited battery lives, any boost inthe link budget will be useful in satellite networks. To addressthis issue in this paper we propose the use of receive RISto improve received power levels. RIS units are composedof metasurfaces that are controlled in real time to adjust thereflection phases [9]. Their potential has already been apparentin the recent works including [13, 15]. In this paper, wepropose the use of RIS located at your satellites that canbe positioned jointly with a reflective array as illustrated inFig. 1. After formulating the problem we quantify the possibletransmit power gain for the target performance is detailedbelow.
C. Contributions
Our contributions on the way to provide ubiquitous anddense connectivity demanded by 6G and beyond are summa-rized below:C1 A novel architecture is proposed for IoT networks basedon the use of RIS units at the satellites. These are RISunits can be placed on the rectenna arrays.C2 We derive the SNR levels of LEO satellites to supportIoT networks considering broadcasting and beamformingmodes based on the transmission characteristics includingthe carrier frequency.C3 Through extensive numerical results, we quantificationof the potential reduction for the transmit power wellconsidering a realistic transmission model. Based on ouranalysis, we suggest design guidelines for future IoTnetworks that are served by LEO satellites.
D. Outline
The rest of this paper is organized as follows. Section IIintroduces the associated free space path loss models for RIS-assisted broadcasting and RIS-assisted beamforming schemes as well as the case without RIS. In Section III, the systemmodel is described for RIS-assisted LEO satellites for IoTnetworks. Section IV presents extensive numerical results andobservations for non-RIS satellites and RIS-assisted satelliteswith bot broadcasting and beamforming schemes. Open issuesare discussed in Section V. Finally, the study is concluded inSection VI. II. P
RELIMINARIES
We discuss the mathematical and physical preliminariesfor RIS-assisted satellites for IoT networks and give thefoundation information in this section. Firstly, we remind freespace path loss for conventional satellite systems. Then, pathloss modeling for RIS-assisted wireless communications isintroduced. Here, the notation is given for downlink; however,it can be generalized for uplink case in a straightforwardmanner.
A. Free Space Path Loss
The ratio of the received and transmitted powers in a linkbetween two isotropic antennas can be given by free spacepath loss, which is defined as follows: L FS = (4 πd/λ ) α , (1)where α and λ denote the path loss exponent and wavelength,respectively. d is the distance between satellite and groundstation which can be obtained as: d = − r e sin( ϕ ) + (cid:113) r e sin ( ϕ ) + h + 2 r e h sat , (2)where r e , h sat , and ϕ are the radius of Earth, the altitude ofsatellite, and the elevation angle between the ground stationand satellite, respectively. The path loss is proportional to d α ;therefore, the elevation angle has an important role on the pathloss. IoT
NodesRIS Broadcasting RIS Beamforming(a) (b)
Fig. 2. RIS-assisted satellites can serve in two modes which are (a) RISbroadcasting and (b) RIS beamforming. In broadcasting case RIS acts asa scatterer for the incident wave while RIS performs specular reflection inbeamforming schemes.
B. Path Loss Models for RIS-assisted Communications
In this subsection, we describe the path loss models forRIS-assisted wireless communications in broadcasting andbeamforming cases.First, we consider the broadcasting case as illustrated inFig. 2(a). If the transmitter is in the near field of RIS andthe surface is at least ten times electrically larger than thewavelength, λ , the surface scatters the incident spherical wave.Scattering incident wave creates a large beam, which cancover multiple ground stations at the same time. As thetransmitter antenna is relatively near the RIS unit deployedon the satellite, the scattering paradigm should be consideredin the case of employing electrically large surfaces. The pathloss model for RIS-assisted wireless communications in thecase of broadcasting is modeled as follows [10]: P L BC ≈ (4 π ) α ( d tx + d rx ) α G t G r λ α A α , (3)where d tx and d rx are the distances from transmitter antennato RIS and the from RIS to receiver antenna, respectively. As d rx (cid:29) d tx , the distance between transmitter antenna and RIScan be omitted and d rx = d . Then, the expression can bewritten as: P L BC ≈ (4 πd ) α G t G r λ α A α . (4) G t and G r denote antenna gain for transmitter and receiverantennas, respectively. A is the amplitude of the reflectioncoefficient of meta-atoms. It should be noted that the free-space path loss for RIS broadcasting scheme is not dependentto the number of RIS elements as seen in (4). In this scheme,RIS scatters the incident wave. Thus, while a wider coverageis obtained, the amount of energy per area is slightly lower.On the other hand, the transmitted wave can be focused onthe receiver in order to improve the received signal quality as depicted in Fig. 2(b). Both the transmitter and receiver can bein the far-field of the RIS, or one of them can be found in thenear-field of the RIS. The latter is the more appropriate modelfor satellite IoT systems, which includes onboard transmitterin satellites. Therefore, we utilize the near-field beamformingscheme for RIS. In other ways, it can be said that RIS operatesas a specular reflection [16]. The path loss model for near-field beamforming scheme in RIS-assisted satellite is given asfollows [10]: P L BF = 64 π G t G r d x d y λ α A α (cid:12)(cid:12)(cid:12)(cid:12)(cid:80) Nn =1 √ F combine n d tx d rx (cid:12)(cid:12)(cid:12)(cid:12) α , (5)where N is the number of meta-atoms. d x and d y are thesize of each unit cell along the x-axis and y-axis, respectively. F combine n denotes the accounted normalized power radiationpattern on the received signal power and it is described as: F combine n = F tx (cid:0) θ txn , β txn (cid:1) F (cid:0) θ tn , β tn (cid:1) F ( θ rn , β rn ) F rx ( θ rxn , β rxn ) , (6)where F ( θ n , β n ) shows the normalized radiation pattern forthe elevation angle θ n and the azimuth angle β n between RISand transmitter antenna (or receiver antenna). The normalizedradiation pattern for the elevation angle θ and azimuth angle β is defined as follows [17]: F ( θ, β ) = cos ( θ ) , θ ∈ (cid:104) , π (cid:105) , β ∈ [0 , π ]0 , θ ∈ (cid:16) π , π (cid:105) , β ∈ [0 , π ] (7)It is worth to noting that in order to maximize the receivedpower, the transmitter antenna has to be deployed to satellitesuch that its normal line is orthogonal to the surface. Namely, θ txn = θ tn = 0 for all unit cells. Without loss of generality,it is assumed that θ rxn = θ rn = π − ϕ and d rx (cid:29) d tx . Themost important issue in the RIS beamforming scheme is thatthe loss gradually decreases with the increasing number ofelements. C. Rain Attenuation
Rain attenuation is one of major propagation impairmentfor satellite systems. Rain can cause scattering and absorptionof the wave propagating through the atmosphere. The rainattenuation is described by ITU-R P.618-13 as [18]:
P L rain = ξ R L E (dB) , (8)where ξ R and L E are is frequency-dependent specific at-tenuation coefficient described by ITU-R P.838 [19] and theeffective path length. Firstly, we introduce the steps to find thevalue of ξ R . ξ R = k ( R . ) ν (dB / km) , (9)where R . is the rainfall rate and can be obtained from ITU-R P.837 [20] for the location of a ground station. k and ν denote the frequency-dependent coefficients given in ITU-RP.838 [19] as follows: k = (cid:2) k H + k V + ( k H − k V ) cos ( ϕ ) cos(2 τ ) (cid:3) / ν = (cid:2) k H ν H + k V ν V + ( k H ν H − k V ν V ) cos ( ϕ ) cos(2 τ ) (cid:3) / , (10) where τ = π for circular polarization and all coefficients arelisted in Table I for . GHz and . GHz.In order to find the effective path length L E , which is L E = L R v . (km) , (11)where v . is the vertical adjustment factor modeled asfollows: v . = 11 + (cid:112) sin( ϕ ) (cid:16) (cid:0) − e − ( ϕ/ (1+ χ )) (cid:1) √ L R ξ R f − . (cid:17) . (12) f is the frequency in GHz and χ is defined as [21]: χ = (cid:26) − | latitude | , | latitude | < ◦ , otherwise. (13)Also, L R is described as: L R = (cid:40) L G r . cos( ϕ ) , tan − (cid:16) h R − h S L G r . (cid:17) > ϕ ( h R − h S )sin( ϕ ) , otherwise. (14)where h S is the altitude of the ground station in km. Also, L G is defined as: L G = (cid:40) L S cos( ϕ ) , h R − h S > , h R − h S ≤ (15)where L S is slant-path length calculated as follows: L S = h R − h S ) (cid:113) sin ( ϕ )+ hR − hS ) re +sin( ϕ ) , ϕ ≤ ◦ h R − h S sin( ϕ ) , ϕ > ◦ (16)where h R is the effective height of the rain which is describedby ITU-R P.839 [22] as: h R = h + 0 .
36 ( km ) , (17)where h is the mean ◦ isotherm height above mean sea leveland it is site-specific value.III. LEO S ATELLITES - ENABLED I O T N
ETWORKS
In this section, we describe the system model for RIS-assisted satellite communications with a transmitter antenna in the near-field of RIS deployed on satellite. In order to maxi-mize the received power, the transmitter antenna is required tobe aligned with the normal line of the surface, which makesthe angle between the normal line and antenna beam zero.Thus, the normalized radiation pattern takes maximum valueas observed in (7). Furthermore, the distance d tx betweentransmit antenna and the surface is very short compared tothe distance to ground station. Hence, the wireless channelbetween transmitter and RIS can be omitted through theanalysis. The received signal y can be introduced as y = (cid:114) P t P L g T Φh x + w, (18)where x and w stand for the transmitted signal with power P t and additive white Gaussian noise (AWGN) at receiver, re-spectively. The noise term, w , can be assumed to be distributedwith CN (0 , N ) . h is the channel coefficient vector for the It should be noted that it is a receiver antenna for uplink communications. TABLE IT
HE FREQUENCY - DEPENDENT COEFFICIENTS WHICH USED IN RAINATTENUATION CALCULATION FOR CIRCULAR POLARIZATION (ITU-R R EC . P.838). Coefficients 4.25 GHz 10.5 GHz k H . − . − k V . − . − ν H . . ν V . . Polarization Circular Circular k . − . − ν .
127 1 . link between RIS and receiver such that h = [ h , h , . . . , h N ] .On the other hand, g stands for the channel coefficient forthe link between transmitter and RIS. As transmit antenna isclose to RIS, we can ignore the channel effects in betweentransmitter and RIS. Therefore, g can be selected as the vectorof all ones such that g = N . In this study, it is assumed thatchannel coefficients follow Rician distribution with the shapeparameter of K = 10 . Φ denotes the RIS element responsesand it can be shown as Φ = diag (cid:8) A e jφ , . . . , A N e jφ N (cid:9) , (19)where A i and φ i are the amplitude and phase response of i -th RIS element. Throughout this study, RIS is assumed as alossless device; therefore, A i = A = 1 , ∀ i .Next, regarding the received signal, the instantaneous SNR γ can be given as: γ = (cid:12)(cid:12) g T Φ h (cid:12)(cid:12) P t N P L , (20)where
P L is the total loss including free-space path loss and rain attenuation. For free-space path loss calculation, (1)is utilized for satellite communications without RIS. In RIS-assisted case, there two different modes, which are relatedto the dimension of the RIS element unit. For RIS-assistedbroadcasting scheme, (4) gives the free-space path loss, while(5) is the free-space path loss expression for RIS-assistedbeamforming. Considering the SNR, the achievable data ratecan be expressed as R = log (1 + γ ) (bits/s/Hz) . (21)IV. N UMERICAL R ESULTS AND D ISCUSSIONS
In this section, we evaluate comprehensive simulation anddiscuss simulation results. Firstly, we focus on downlinkcapacity for satellite IoT systems. Then, we investigate up-link capacity of satellite-supported IoT communications. Weconsider the simulation results for the two prominent bands forsatellite IoT systems, namely C- and X-band. In simulations,the path loss model given in (1) is used for the case withoutRIS, while the models given with (4) for RIS broadcasting and(5) are used for RIS beamforming. It can be either RIS-assisted wireless communications or not. For RIS-assisted communications,
P L BC and P L BF are employed for RIS broadcast-ing case and RIS beamforming case, respectively. On the other hand, it isequal to P L FS in case of wireless communications without RIS. TABLE IIT
HE LOCATION - SPECIFIC COEFFICIENTS WHICH USED IN RAINATTENUATION CALCULATION FOR CIRCULAR POLARIZATION (ITU-R R EC . P.838 & ITU-R R EC . P.839). Parameters Values
Location Istanbul, TurkeyLatitude ◦ NLongitude ◦ E h . km R . . h S m h sat km r e km As the rain attenuation seriously changes with respect tothe operating frequency, it is important to properly acquirethe frequency-dependent coefficients such as k and ν . Forexample, k is found for C- and X-band as . − and . − , respectively. Besides, ν is equal to . and . for C- and X-band, respectively. It is worth to saying thatthese values are valid for circular polarization, i.e. τ = π . Allfrequency-dependent coefficients are given in Table I.Also, in simulations, we assume that the ground stations, i.e.IoT devices, are located in Istanbul, Tukey, which is locatedat ◦ North latitude and ◦ East longitude. Site-specificcoefficients such as the mean ◦ isotherm height above meansea level h and the rainfall rate R . is found as . km and . for Istanbul, Turkey. These parameters are summarizedin Table II. As we investigate the LEO satellites, the satellitealtitude h sat is chosen as km.Another crucial point is that the characteristics of transmitterand receiver antennas as well as RIS characteristics. Weevaluate the simulations by employing the antennas and RISgiven in [10]. These antennas have the normalized radiationpattern F ( θ, β ) defined as F tx ( θ, β ) = (cid:40) cos ( θ ) , f = 4 . GHz (C-band) cos ( θ ) , f = 10 . GHz (X-band) . (22)Furthermore, antenna gains are given as . dB and dB forC-band and X-band antennas, respectively. As aforementionedabove, the distance between transmit antenna and the surfaceis small. Hence, d tx is selected as m to keep near-fieldcondition for both bands. d rx is equal to the distance givenby (2). In addition, some parameters of RISs are determinedas follows: It is reported that RIS elements are designed insquares and their edge lengths are . m and . m for . GHz and . GHz, respectively.The path loss exponent α is chosen as , which is agenerally accepted value. Additionally, the small-scale fadingfor the channel between RIS and receiver is modeled by Ricedistribution with the shape parameter of K = 10 in order toallow few non-line-of-sight (NLOS) paths. Last, the effectivenoise power for the overall system is chosen as − dB.Table III summarizes the parameters which are employed inthe calculation process of free-space path loss. Here, the normalized radiation pattern is given for only transmit antenna.It should be noted that this expression can also be utilized for receive antenna. TABLE IIIT
HE PARAMETERS FOR FREE - SPACE PATH LOSS CALCULATION FOR
RIS
SAND ANTENNAS OPERATING AT C- BAND AND X- BAND . Parameters 4.25 GHz 10.5 GHz d tx m m d rx d dα d x [10] . m . m d y [10] . m . m G t [10] . dB dB G r [10] . dB dB F tx ( θ, β ) [10] cos ( θ ) cos ( θ ) F rx ( θ, β ) [10] cos ( θ ) cos ( θ ) θ tx θ t θ rx π − ϕ π − ϕθ r π − ϕ π − ϕK
10 10 N − dB − dB A. Downlink Performance Analysis
In this section, we investigate the downlink performance ofsatellite IoT networks in terms of achievable data rate in threecases: Without RIS, RIS broadcasting, and RIS beamforming.First, we compare the achievable data rate for conventionalsatellite systems with RIS-assisted satellites in Fig. 3(a) for C-band. The simulation results show that RIS-assisted satellitesprovide much higher capacity than conventional satellites,regardless of which scheme they operate. In fact, the RISbroadcasting scheme can provide up to times higherachievable rate than the non-RIS case, while RIS beamformingcan provide up to times higher rate. In other words, thedesired capacity can be obtained with a lower transmit powerby using RIS structures in satellites. Next, we investigate theimpact of the number of RIS elements in Fig. 3(b).As the number of RIS elements increases, the achievablecapacity for the RIS beamforming case increases, as expected.However, as given in (4), in the case of RIS broadcasting,the number of elements does not affect the performance. Dueto specular reflection, RIS beamforming even with a singleelement can achieve more data rate than the broadcasting case.The main reason behind that is the scattering of energy in awide area. It can be said that the broadcasting scheme cansupport more IoTs than the beamforming scheme because of alarger coverage area provided by the broadcasting. Last, as rainattenuation coefficients and the normalized radiation patternare seriously dependent on the operating frequency the sameanalysis in C-band is evaluated for X-band in order to observeimpacts. The simulation results are depicted in Fig. 3(c). Forlarge N values, the achievable rate decreases to less than half,while the ratio of performance loss increases as the numberof elements decreases. For example, in the case of a singleelement, the achievable rate decreases to almost one-fourth ofthat in the C-band.Since LEO satellites are mobile with respect to the earth,the distance between them and the ground stations variesdepending on the elevation angle. Distance is one of themajor contributors to free space path loss. Therefore, we haveperformed the simulations by considering varying elevation A c h i e v ab l e R a t e ( b i t s / s / H z ) No RISRIS BroadcastingRIS Beamforming (N = 4) -5 (a) A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting (b) A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting17 18 190.010.020.030.040.05 (c)Fig. 3. Downlink achievable rate versus transmit power P t for (a) comparison between non-RIS case and RIS-assisted satellites, (b) the various number ofRIS elements in C-band, and (c) the various number of RIS elements in X-band. It should be noted that the elevation angle is π . Elevation Angle (rad) A c h i e v ab l e R a t e ( b i t s / s / H z ) No RISRIS BroadcastingRIS Beamforming (N = 4) -5 (a) Elevation Angle (rad) A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting (b)
Elevation Angle (rad) A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting3 /8 /20.020.04 (c)Fig. 4. Downlink achievable rate versus elevation angle ϕ for (a) comparison between non-RIS case and RIS-assisted satellites, (b) the various number ofRIS elements in C-band, and (c) the various number of RIS elements in X-band. It should be noted that the transmit power is W. angles between zero and π rad. Fig. 4 denotes that the RISbeamforming overperforms than broadcasting and non-RIScases. Since θ rx and θ r are defined as θ rx = θ r = π − ϕ ,the elevation angle change the normalized received radiationpattern. Therefore, for the lower number of RIS elements atlower elevation angles, the RIS broadcasting scheme providesslightly better achievable rate than RIS broadcasting. However,increasing the number of RIS elements compensates for thepath loss increasing owing to the increased distance anddecreased received radiation pattern as seen in Fig. 4(b)and Fig. 4(c). B. Uplink Performance Analysis
In this section, we analyze the uplink capacity of satelliteIoT networks in three cases which are without RIS, broadcast-ing RIS, and beamforming RIS. Similar to downlink analysesabove, we firstly investigate the advantage of utilizing RISsin satellites for IoT networks. Fig. 5(a) shows that RIS beam-forming can provide up to times higher uplink rate thannon-RIS case. This result indicates that the battery and lifetimeof battery-limited IoTs can be significantly increased with RIS-assisted satellites. Moreover, Fig. 5(b) shows that it is possibleto obtain 3 dB more gain from transmit power by doublingthe number of RIS elements (from 512 to 1024) for the sameachievable data rate value. It is seen in Fig. 5(c) that the uplinkdata rate decreases significantly if the operating frequencyis increased. As the number of RIS elements increases, the rate of decrease in the achievable rate due to the increasein frequency degrades. Although increasing in the operatingfrequency reduces the achievable data rate performance, thehigher frequency can be employed for small IoTs whichcannot accommodate larger antennas. Fig. 6(a) shows that theRIS-beamforming scheme can serve with higher performanceover a wider range of elevation angles than RIS broadcastingand non-RIS satellites. As seen in Fig. 6(b), increasing thenumber of RIS elements marginally improves the operationrange in terms of elevation angle. Higher frequency limits thecommunications in a narrow elevation angle range as depictedin Fig. 5(c), which means that the communication durationin one period of a satellite is very short. It may cause manyhandovers between satellites in a short while.Since the computational capacities and batteries of IoTs aresmall, it is possible to say that there are two main factors in ex-tending battery life. These two basic factors are the operationof computationally complex processes such as transmissionpower and equalization. RIS-assisted satellite systems havetwo-fold advantages for IoT networks. First, RISs can providethe desired capacity at lower transmission power. The secondis that it enables complex operations to be performed in anexternal environment (i.e. propagation medium) rather than onthe device, thanks to the use of RIS units. -20 -15 -10 -5 000.050.10.150.20.25 A c h i e v ab l e R a t e ( b i t s / s / H z ) No RISRIS BroadcastingRIS Beamforming (N = 4)-3 -2 -1 046 10 -7 (a) -20 -15 -10 -5 00123456 A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting-4 -2 000.05 (b) -20 -15 -10 -5 000.050.10.150.20.250.30.350.40.45 A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting-5 0024 10 -4 (c)Fig. 5. Uplink achievable rate versus transmit power P t for (a) comparison between non-RIS case and RIS-assisted satellites, (b) the various number of RISelements in C-band, and (c) the various number of RIS elements in X-band. It should be noted that the elevation angle is π . A c h i e v ab l e R a t e ( b i t s / s / H z ) No RISRIS BroadcastingRIS Beamforming (N = 4)3 /8 /2456 10 -7 (a) A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting3 /8 /20.020.040.06 (b) A c h i e v ab l e R a t e ( b i t s / s / H z ) Beamforming (N = 1024)Beamforming (N = 512)Beamforming (N = 1)Broadcasting /21234 10 -4 (c)Fig. 6. Uplink achievable rate versus elevation angle ϕ for (a) comparison between non-RIS case and RIS-assisted satellites, (b) the various number of RISelements in C-band, and (c) the various number of RIS elements in X-band. It should be noted that the transmit power is W. V. O
PEN I SSUES AND R ESEARCH D IRECTIONS
Even though there are numerous open issues for RIS-assisted wireless communications, we discuss three of themost crucial issues for RIS-assisted satellites for IoT net-works, which are the channel estimation, RIS deployment onsatellites, and simultaneous wireless information and powertransfer (SWIPT), with a focus on physical layer.
A. Channel Estimation
Since RISs can change the amplitude and/or phase of theincident electromagnetic wave, they can largely eliminatesthe randomness of the propagation medium. However, itcertainly needs high-quality channel state knowledge to obtainthe maximum performance. Channel estimation thus plays acritical role in satellite communications for IoT networks tosatisfy the desired QoS. Deep learning may be used for thispurpose in order to achieve high efficiency under convincingchannel conditions in channel estimation. Recently, for thechannel estimation in RIS-assisted backhaul communications,we proposed a channel estimation framework based on graphattention networks (GATs) in [23]. By taking unseen nodesinto consideration, the GAT can reduce computational com-plexity and improve learning performance. In the trainingprocess, the obtained signal samples and known pilot samplesare assigned to the nodes and vertices of graphs, respectively.Since IoT devices lack the hardware infrastructure required for training and using the deep learning model, it is reasonableto deploy the proposed GAT model to LEO satellites ratherthan IoT devices. By utilizing channel reciprocity, the channelcoefficients can be obtained through the uplink pilot signaling.Without the need for downlink pilot signaling, phase adjust-ment can be performed by RIS by utilizing the coefficientsfound for the uplink channel. However, it is worth notingthat this channel estimation cannot be employed in frequencydivision duplex communications for uplink and downlink.On the other hand, when using time division duplex, thesignal processing can be performed through the propagationenvironment and RIS rather than IoTs, thus, it extends batterylife.
B. RIS Fabrication and Deployment
In the design of RISs, the environmental conditions must beconsidered. RISs that can be resilient to temperature variationsbetween day and night should be developed. Although spacemay be assumed as a vacuum, it should be noticed that theentire space is filled with plasma and particles emitted fromthe Sun. The electronic elements of the communication unitmay be affected by charged particles encountered by spacecraftin the Van Allen Belts. As a result, it is crucial to considerradiation-resistant RIS and any instrumentation that allows RISoperational.How to implement the RISs into the LEO satellites is unde-niably the most fundamental and first-glance question. Taking into account the mechanical systems of satellites, it is obviousthat it is necessary to address a significant design challenge.Especially, depending on the variations in their locations andelevation angles, the device itself is likely to shadow any orall of the meta-atoms. Based on the development of theirstructures, the authors envisage that conformal metasurfacescan be used in the coating of objects with irregular surfacesand arbitrary shapes. It is worth to remember that satellite sys-tems have already utilized reflectarray antennas to reflect theincident beam with a constant phase [24]. These reflectarrayscan be replaced by RISs to make satellites smart. As a result,satellites can reflect or scatter the incident wave with variousphase configuration. Interdisciplinary studies can make novelRIS designs possible. For example, it may be possible to placemany RIS elements in a wide area by covering the lower facesof the satellite solar panels with RIS. Moreover, the antennasubsystems currently on satellites might be replaced by RISs.
C. Simultaneous Wireless Information and Power Transfer
As IoT nodes are power-limited devices, energy harvestingcan be considered in SWIPT framework. RIS-aided SWIPThas been proposed to achieve significant performance gainin energy harvesting [25]. SWIPT can be a solution to keeppower-constrained IoT systems running for a long time. But,its applications in satellite-IoT systems are lacking in theliterature. Considering the power transfer capacity of spacesolar satellites with microwave waves [26], thr joint utiliza-tion of RISs and rectenna arrays can further improve theenergy-efficiency of IoT networks. To shortly and preciselyemphasize, RIS-assisted satellites for SWIPT in IoT networksdeserve more attention in the future.VI. C
ONCLUDING R EMARKS
Ubiquitous connectivity and user-centric communicationsare the strict requirements of 6G networks. As the number ofthe connected devices and Internet of Things (IoTs) networksare expected to increase exponentially, LEO satellite networksgain attention to serve as the backhaul/fronthaul connectivitysolution to densely deployed IoT sensors. However, the currentsystems under consideration either have a very high pathloss or require steerable antennas in IoT devices. Consideringthese drawbacks of the previously proposed system models,we introduce an overview of the RIS-assisted LEO satelliteframework for energy-efficient IoT networks in this study. Themotivation behind the RIS communications in satellite-IoT aregiven using extensive numerical results throughout the study.The potential gain in terms of the transmission powers oflightweight IoT nodes are quantified. Furthermore, open issuesregarding the proposed system model are also discussed.R
EFERENCES[1] Huawei, “Touching an intelligent world,” https://bit.ly/38KHhmT, 2019,(Accessed on 01/07/2021).[2] McKinsey Global Institute, “The Internet of Things: Mapping the valuebeyond the hype,” https://mck.co/3igecTk, June 2015, (Accessed on01/08/2021).[3] Deloitte, “The Internet of Things: a technical primer,” https://bit.ly/35HPPc8, 2018, (Accessed on 01/08/2021). [4] S. Kota and G. Giambene, “Satellite 5G: IoT use case for rural areasapplications,” in
International Conference on Advances in Satellite andSpace Communications , 2019, pp. 24–28.[5] C. A. Hofmann and A. Knopp, “Direct access to GEO satellites: AnInternet of Remote Things technology,” in
IEEE 5G World Forum , 2019,pp. 578–583.[6] R. Barbau, V. Deslandes, G. Jakllari, J. Tronc, J.-F. Chouteau, and A.-L.Beylot, “NB-IoT over GEO satellite: performance analysis,” in
AdvancedSatellite Multimedia Systems Conference and the Signal Processing forSpace Communications Workshop , 2020, pp. 1–8.[7] Z. Qu, G. Zhang, H. Cao, and J. Xie, “LEO satellite constellation forInternet of Things,”
IEEE Access , vol. 5, pp. 18 391–18 401, 2017.[8] S. Cluzel, L. Franck, J. Radzik, S. Cazalens, M. Dervin, C. Baudoin,and D. Dragomirescu, “3GPP NB-IoT coverage extension using LEOsatellites,” in
IEEE Vehicular Technology Conference , 2018, pp. 1–5.[9] E. Basar, M. Di Renzo, J. De Rosny, M. Debbah, M.-S. Alouini, andR. Zhang, “Wireless communications through reconfigurable intelligentsurfaces,”
IEEE Access , vol. 7, pp. 116 753–116 773, 2019.[10] W. Tang, M. Z. Chen, X. Chen, J. Y. Dai, Y. Han, M. Di Renzo,Y. Zeng, S. Jin, Q. Cheng, and T. J. Cui, “Wireless communications withreconfigurable intelligent surface: Path loss modeling and experimentalmeasurement,” arXiv preprint arXiv:1911.05326 , 2019.[11] L. Dai, B. Wang, M. Wang, X. Yang, J. Tan, S. Bi, S. Xu, F. Yang,Z. Chen, M. Di Renzo et al. , “Reconfigurable intelligent surface-basedwireless communications: Antenna design, prototyping, and experimen-tal results,”
IEEE Access , vol. 8, pp. 45 913–45 923, 2020.[12] K. Tekbıyık, G. K. Kurt, A. R. Ekti, and H. Yanikomeroglu, “Reconfig-urable intelligent surfaces in action for non-terrestrial networks,” arXivpreprint arXiv:2012.00968 , 2020.[13] Q. Wu and R. Zhang, “Intelligent reflecting surface enhanced wirelessnetwork via joint active and passive beamforming,”
IEEE Trans. onWirel. Commun. , vol. 18, no. 11, pp. 5394–5409, 2019.[14] 3GPP, “Study on Narrow-Band Internet of Things (NB-IoT)/enhanced Machine Type Communication (eMTC) supportfor Non-Terrestrial Networks (NTN),” 3rd Generation PartnershipProject (3GPP), Technical Report 36.763, 01 2021, release17. [Online]. Available: https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3747[15] C. Huang, A. Zappone, G. C. Alexandropoulos, M. Debbah, andC. Yuen, “Reconfigurable intelligent surfaces for energy efficiency inwireless communication,”
IEEE Trans. on Wirel. Commun. , vol. 18,no. 8, pp. 4157–4170, 2019.[16] S. W. Ellingson, “Path loss in reconfigurable intelligent surface-enabledchannels,” arXiv preprint arXiv:1912.06759 , 2019.[17] C. A. Balanis,
Antenna Theory: Analysis and Design
Satellite communications systems:systems, techniques and technology arXiv preprint arXiv:2010.12004 , 2020.[24] J. Huang, “Reflectarray antenna,”
Encyclopedia of RF and MicrowaveEngineering , 2005.[25] Q. Wu and R. Zhang, “Weighted sum power maximization for intelligentreflecting surface aided SWIPT,”
IEEE Wireless Commun. Lett. , vol. 9,no. 5, pp. 586–590, 2019.[26] C. A. M. Bergsrud, R. Bernaciak, S. Chaieb, and J. Casler, “Rectennaarray equipped on satellites,”