Intelligent Reflecting Surface Aided Wireless Energy Transfer and Mobile Edge Computing for Public Transport Vehicles
IIntelligent Reflecting Surface AidedWireless Energy Transfer and Mobile EdgeComputing for Public Transport Vehicles–
A Communication Eco-System For Transport Vehicles of 6G-era.
Shan Jaffry
School of Internet of Things,Xi’an Jiaotong-Liverpool [email protected]
Abstract —In the forthcoming 6G era, Future Publictransport vehicles (F-PTV), such as buses, trains etc. willbe designed to cater the communication needs of thecommuters that will carry numerous smart devices (smart-phones, e-bands etc.). These battery-powered devices needfrequent recharging. Since recharging facilities are notreadily available while commuting, we envision F-PTVsthat will provide in-vehicle recharging via Wireless EnergyTransfer (WET) through in-vehicle Access Points. F-PTVwill also be internally coated with Intelligent ReflectingSurface (IRS) that reflect incident radio waves towardsthe intended device to improve signal strength at thereceiver, for both information and energy transmissions. F-PTVs will also be equipped with Mobile Edge Computing(MEC) servers to also serve multiple purposes, includingreduction in devices’ task completion latency and offeringin-vehicle Cloud services. MEC-server offloading will alsorelieve smart devices’ processors off intensive tasks topreserve battery power. The challenges associated withthe IRS-aided integrated MEC-WET model within F-PTVare also discussed. We present simulation to show theeffectiveness of such a model for F-PTV.
Index Terms —6G, Intelligent Reflective Surface, Wire-less Energy Transfer, Wireless Power Transfer, MobileEdge Computing.
I. I
NTRODUCTION
Modern applications running on smart devicesand gadgets are increasingly becoming sophisti-cated and consume excessive computation resourcesand battery power. On the other hand, the availablecomputational and battery resources are limited forsuch devices, mainly due to their small form factor.Replenishing devices’ battery levels on-the-go isbecoming a serious challenge. In particular, mobileusers that are on the move, for example, commutersinside public transport vehicles (PTVs), such asbuses, trains, subways etc., are more vulnerable tobattery drainage as they have limited rechargingfacilities available. These users collectively carrya large number of devices and engage in online or offline activities during the commute. Such activi-ties (e.g. audio/video calling, gaming, web-surfing,etc.) drain extensive amount of devices’ battery.On the other hand, it has been proposed thatfuture PTVs (F-PTVs) will have a dedicated in-vehicle small cell to provide services to commutingusers [1]. Primarily, Access Points within F-PTV(PTV-AP) are aimed to enhance in-vehicle cellularuser’s quality of service (QoS). The commuterswithin F-PTV will be effectively decoupled fromthe out-of-vehicle macro-cell base station whichsignificantly reduce cellular network load. Further-more, F-PTV will also reduce the group handoverinto a collective hand-off by the vehicle [2]. Inte-grating in-vehicle wireless energy transfer (WET)methods with PTV-APs can enable wireless charg-ing of commuters’ energy-constrained devices andgadgets while on the move.However, due to hostility of air interface towardselectromagnetic radio signals, the magnitude ofpower delivered though WET remains effectivelyvery low. To make matter worse, signal qualitydegrades drastically in a crowded public trans-port as signal passes through a human bodies.This body-shadowing effect further diminishes theprospect of any energy gains with WET-enabledPTV-AP. To bolster wireless channels within F-PTV, we further envision installing the IntelligentReflecting Surface (IRS) within the vehicle. TheIRS are passive reflecting elements that change thephase and amplitude of wireless signals and caneffectively undo the signal attenuation by reflectingincident waves towards the receiver device.On the other hand, to reduce increasing com-puting requirements from the smart devices, re-searchers have focused on offloading services andtasks to the near-by external servers which are a r X i v : . [ c s . N I] F e b ig. 1. A depiction of a communication-friendly futuristic Public Transport Vehicle for 6G era(An idea). relatively more powerful than smart devices andgadgets. These powerful computing nodes, alsoknown as MEC-servers, are located at the networkedge, for example with the base-stations and APs.These MEC-servers can be collocated with PTV-APto provide in-vehicle Cloud services and augmentcomputing requirements of commuting users insideF-PTV via task offloading. The task offloadingcan be exploited for dual, albeit closely linked,purposes. First, offloading reduce tasks’ processingtime as MEC-server’s Central Processing Units(CPUs) are more powerful than smart devices.Second, the task offloading also preserves extensiveamount of device’s power that is otherwise con-sumed during the local computations.This envisioned IRS-aided MEC-WET systemcreates a cellular-friendly eco-system within F-PTVin the 6G era. In the rest of this article, we firstprovide a brief overview of the envisioned com-bined IRS-aided integrated MEC-WET system forF-PTVs. We then discuss potential challenges andpossible solutions associated with the envisioned F-PTV. Later, Simulation and numerical results areprovided to validate the effectiveness of the design.II. IRS-A IDED
MEC-WET
WITHIN P UBLIC T RANSPORT V EHICLE
In this section we will first discuss about theWET technology followed by IRS and MEC systemwithin F-PTV.
A. Suitable WET Technology for PTV
WET is not an entirely new idea. In fact, itis more than century old concept [3]. Famousinventor Nikola Tesla conducted wireless powertransfer experiments in the late 19th century [4].However, modern improvements in microelectronic circuit design have sparked a new interest in WETfor smart devices, gadgets, and sensor networks.We broadly categorized WET technologies intotwo types. The first type, called Inductive CouplingWET (IC-WET), exploits near-field electromag-netic energy transmission phenomena. IC-WET ispopular for charging smart gadgets, electric vehi-cles, implanted medical devices etc. More recently,key smart device manufacturers have built practicalwireless charging accessories based on IC-WET.However, IC-WET works only for very short dis-tances and in most cases the charging adapter andthe rechargeable device needs to be within near-physical contact. Due to this limitation, IC-WET isnot a suitable fit for F-PTV use case.The second type of WET uses radio frequency(RF) signals to transmit energy and offers flexibilityof charging devices at a relatively larger distances(few meters). This makes RF-WET a more suitableoption for the use case within F-PTV. However,electromagnetically radiated RF signals withers sig-nificantly in-air before reaching the receiver. Thehighly attenuated signal received at the harvestingdevice is significantly low, and even ineffectiveat times, to provide any useful energy gains. Thesignal attenuation exacerbates within F-PTV be-cause of human-body shadowing due to presenceof passengers.
B. IRS Coating within Vehicle
The poor wireless propagation environmentwithin the F-PTV can be alleviated by internallycoating passive IRS elements within the vehicle.An IRS can virtually reconfigure RF propagationenvironment by controlling the phase and ampli-tude of incident waves [5]. Interestingly, unlikelegacy relaying systems, IRS do not require externalower source to boost signal strength. Instead, IRScomprise of array of low-cost passive reflectingelements that can be controlled in real-time tofocus the direction of transmitted wave towardsthe receiving device, much like a mirror. An in-vehicle IRS controller is responsible for adjustingamplitude and phase of incident waves. The wavescan be added constructively to significantly improvethe signal strength at the desired receiver. The IRS-aided constructive interference can be achieved bothfor information and energy harvesting signals. Thesame mechanism can be used to induce destructiveinterference at unwanted or eavesdropping receiversto nullify the undesirable interference effect.Due to passive nature of IRS elements, additionalpower for signal decoding or amplification is notrequired. Furthermore, unlike active relays IRS alsodo not impose additional white noise as a conse-quence of decoding or amplification [6]. The IRS’sRF reflecting elements can be conveniently coatedover the existing structures such as on the walls orwindows of the vehicle which significantly reducesthe deployment complexity [5].To fully exploit IRS benefits within F-PTV, con-trolling reflective elements is a challenging task.The IRS controller must have real-time knowledgeof factors such as channel state information (CSI),number of wireless links, optimal phase and ampli-tude etc [7]. However, even a fixed or sub-optimaltuning of reflectors have shown to significantlyimprove the quality of signal-to-noise-ratio (SNR)or harvested energy at the receiver [6], [8].Despite IRS-aided improved channel conditions,effective energy harvesting gain for the user devicesremain limited due to low RF-to-DC power con-version efficiency and other hardware constraints.Hence, next we discuss MEC offloading to furtherimprove energy preservation within F-PTV.
C. Edge Computing for Energy Preservation
When it comes to WET, RF-to-DC power con-version efficiency is practically much lower thantheoretically anticipated [9]. This is in part truedue to device’s size limitation and hardware energyconversion constraints. Hence only limited energygain could be achieved by relying solely on WET,even after IRS installation. On the other hand, asignificant amount of devices’ energy consumptionis due to processing requirements of applications’tasks. This is true because a device’s CPU’s powerconsumption is proportional to the operating fre-quency and the operational voltage of the CPU. Byfrequently forcing device’s CPU into idle mode, sig-nificant energy savings can be achieved. One way of relieving device’s CPU off continuous operations isby offloading tasks to the MEC-servers collocatedwith PTV-AP. This MEC offloading serves two pur-poses. Firstly, it allows faster completion of tasksas CPU(s) at MEC-servers are more powerful thanuser devices. Secondly, it preserves a significantamount of the device’s battery power, courtesy totask offloading. Furthermore in [8] researchers haveshown that IRS-aided wireless environment alsoreduce system-level latency for MEC systems byimproving SNR at the end receivers and also overallimproves energy efficiency.The IRS-aided MEC-WET system within F-PTVdoes not come without challenges which are dis-cussed next.III. D
ESIGN C HALLENGES AND F UTURE R ESEARCH D IRECTIONS
In this section we discuss some of the chal-lenges associated with envisioned in-vehicle IRS-aided MEC-WET system.
A. Crowded Channels within PTV
Due to relatively higher device-density (devicesper square meter) within public transport and instal-lation of reflectors within F-PTV, in-vehicle radiochannels may get over-crowded in case excessiveinformation and energy signals are exchanged be-tween PTV-AP and commuters. Even in the absenceof online activity, users may compete for channelsto harvest RF energy or to upload and downloadtasks to and from the MEC-servers, respectively.Given the insufficient number of available channelsfor both information and energy transmission, fairchannel scheduling will be challenging for IRSaided environment within PTV.Alternatively, the mmWave band, which is suit-able for indoor communication, offers wider band-width and more radio channels to alleviate conges-tion problems. However, due to large propagationlosses, mmWave band may not be suitable forenergy signal transmission. One possible solution toresolve above stated issues is to use a combinationof mmWave and sub-6 GHz (e.g. 2.4 GHz and 3.5GHz) bands within F-PTV such that informationsignals are carried over mmWave frequencies andenergy harvesting is carried over sub-6 GHz band.
B. Limited MEC-WET Resources within PTV
As compared to Cloud storage, MEC servershave relatively limited storage and computing re-sources. In particular, MEC servers within PTVswill have further reduced capabilities due to phys-ical size limitations. Hence, optimally distributing ig. 2. A simplistic decision grid for scheduling computing andenergy resources scheduling among devices. the limited computing resources among several in-vehicle devices would be another challenge.Furthermore, energy signal transmission willbase on time-sharing scheme. For a given Trans-mission Time Internal (TTI) τ , energy signal istransmitted for the period τ v = γτ , where γ ∈ [0 , .The remaining time τ i = (1 − γ ) τ , is dedicatedfor information transmission ( τ i > τ v ). Given that τ v is the small proportion of τ , allocating slotsfor numerous devices need efficient time schedul-ing. PTV-APs with Multiple-Input-Multiple-Output(MIMO) antenna may allow simultaneous transmis-sion of multiple energy signals via spatial diversity.However, scheduling energy resources for largenumber of users is still a challenge.To address computing and energy resource man-agement and scheduling, smart optimization tech-niques are needed. One simplistic scheme could beto rely on the tasks’ completion deadline ( Z n ) andthe battery level ( E n ) for each device. Based onthis information, PTV-AP can allocate MEC andWET resources for individual user. For example, asshown in Figure 2 if a device has high battery leveland lower deadline constraint, then that the devicetask can be performed locally. On the contrary, ifa device has low battery levels then it is givenhigher priority for both MEC and WET resourceallocation. C. Active and Passive Beamforming
The internal structure of most public transportis generally suitable for coating IRS elements. Forexample, the roof or the upper sidewalls could becovered with the reflecting elements. Given thepassive nature of IRS elements additional poweris not required to improve signal quality as theIRS controller adjust the phase and amplitude ofincident waves striking the meta-material surface.The installation of PTV-AP and IRS would fol-low Line-of-Sight (LoS) link. Active beamformingat the PTV-AP and passive beamforming at the IRS elements would guide the direction of thetraveling waves. As mentioned earlier, the pres-ence of large number of passengers would increasebody-shadowing effect within PTV. In the down-link direction, the combined active and passivebeamforming can alleviate this effect. However,active beamforming may not always be possiblefor the uplink transmission due to device (suchas smartphones, watches, e-bands etc.) limitations.Still, researchers in [6], [10] have shown that in anIRS aided environment, even the signals transmittedin random direction provide better quality than non-IRS environment.
D. Electromagnetic Radiation Concerns
Regular in-vehicle transmissions and additionalIRS reflections inside public transport increaseelectromagnetic radiation inside the vehicle whichraises health concerns for passengers. Hence, theseconcerns must be addressed by suitable researchersin line of standardized guidelines outlined by thegoverning bodies.To demonstrate the feasibility of IRS-aidedMEC-WET installation within PTV, numerical re-sults are presented in the next Section.IV. S
IMULATION R ESULTS AND E MPIRICAL O BSERVATION
The model diagram for the IRS-aided MEC-WETsystem within PTV is shown in Figure 1. WEconsider a PTV-AP that is capable of transmittingdata and energy signals. The MEC server is co-located with the PTV-AP and hence AP-to-serverlatency is negligible. The IRS’ reflecting elementsare coated on the vehicle’s interior as shown inFigure 1. A single-antenna PTV-AP and a singledevice are considered for the simulation to showthe effectiveness of the system. Figure 3 showsthe geometrical representation of PTV-AP, IRS, anddevice with default locations and distances.A practical non-linear energy harvesting model[10] is considered for the simulation in whichthe harvesting is successful if the received signalpower is above a threshold P th , zero otherwise. Theharvested power cannot exceed an upper thresholdlimit of P max . Parameters considered in [10] areused in our simulation as well, unless otherwisestated.The channel model consist of both small andlarge-scale fading. In particular, exponentially dis-tributed channels with zero mean and unit vari-ance are considered in both downlink and uplinkdirection. Channels are independent and identicallydistributed. The IRS placement is considered suchthat PTV-AP and IRS elements follows LoS link ig. 3. Geometrical representation of elements within PTV used forsimulation. with pathloss parameter α o = 2 . . The pathloss pa-rameter for PTV-to-User and IRS-to-User channelsis α n = 3 . .At the start of time-frame of length τ , userdevice generates l Kbits of data per unit of time,where l is distributed uniformly between 40 Kbitsand 50 Kbits. If the device wants to offload itstask, it sends a message with relevant informationto the PTV-AP which makes decision regardingoffloading or local computing. In case of offloading, l o = µl Kbits of data is offloaded to MEC serverand l c = (1 − µ ) l Kbits is computed locally bythe device, where µ ∈ [0 , . Device’s CPU takes ψ = 10 cycles to compute one bit of raw dataand energy consumed by CPU to compute l c bits is ξψ l c mJoule, where ξ = 1 × − is the effectiveCPU capacitance [10]. After offloading l o Kbits, thedevice saves ξψ l o mJoule energy. We consideredPTV-AP transmit power P t = 1 Watt. For theIRS, amplitude reflection coefficient is set to 1 [8]and phase shift coefficient ( Φ ) is set to π . Eachsimulation is iterated 1000 times and the averagedresults are reported in the following. A. Numerical Results
First we discuss Figure 4 and Figure 5 in whichdevice’s energy levels E [ t ] are shown with respectto time t . Device’s new energy level in the sub-sequent time steps is E [ t + 1] = E [ t ] + E g [ t +1] − E c [ t + 1] , where E g [ t ] = E v [ t ] + E o [ t ] is theinstantaneous energy gain at time t and E v [ t ] , E o [ t ] denote the harvested energy and energy preserveddue to offloading, respectively. The energy con-sumption is calculated as E c = E ckt + E lc + E tr ,where E ckt , E lc , E tr denote device circuit’s constantenergy consumption, energy utilization for localcomputing, and energy needed to offload task toMEC server. The mobile device’s energy level isnormalized to 1 mJoule at the start of the simulation(i.e. t = 0 ). Fig. 4. Device’s energy levels with respect to time P max = .004Watts, µ = 0.75 , τ v = 0.25 Both figures (Fig. 4 and Fig. 5) show the con-sumption of device’s battery energy levels undernormal scenario. There is some energy gain dueto WET, however it is not very significant particu-larly due to poor channel condition between PTV-AP and the device. Even though the IRS aidedtransmission significantly improves signal strengthat the receiving device, the benefits reaped from thereflecting surface are still limited due to device-level saturation [11]. This is because in a practi-cal WET system, a device cannot harvest energybeyond its saturation limit due to hardware con-straints. To fully exploit the IRS-aided MEC-WET,the maximum allowed harvesting power thresholdat the receiving device should be improved. Thisproblem require hardware-level solution from thedevice manufacturers and is out of scope of theo-retical research. As the maximum harvesting powerthreshold in Figure 4 is P max = 0 . Watts, WETdoes not offer any significant energy gains. On theother hand, when P max is increased to 0.04 Watts,moderate WET-enabled energy gain is observed inFigure 5, even without IRS.Figure 4 and Figure 5 also show that the signifi-cant gain in energy occurs due to MEC offloading.For example, for µ = 0 . and τ . , negligibleenergy losses are observed for the MEC-only case.This is due to the fact that MEC server takes the toilof computation as result of task offloading whichallows energy preservation at user-device end. Incase of IRS-aided MEC-WET scenario, the device’sbattery gains significantly more energy than thatconsumed. In fact, device is able to actually storeenergy above the starting level of 1 mJoule. It canbe observed in Figure 5 that by increasing number ig. 5. Device’s energy levels with respect to time P max = .04 Watts, µ = 0.75 , τ v = 0.25 of reflecting elements from 64 to 200 the gain dueto energy harvesting improved considerably.In Figure 6 we show the energy gain to consump-tion ratio for different amount of data generatedby user device. Considering µ = 0 . , we canobserve that the ratio for MEC-only scenario andremains below 1 for WET-only and IRS-aided WETscenarios when data size is greater than 30 Kbits.For smaller data size (e.g. 24 Kbits), energy gainin case of IRS-aided WET is twice that of energyconsumed by the device. On the other hand, theratio is significantly higher for MEC and IRS-aided MEC-WET scenario, particularly when datasize are low. The ratio decreases as the size ofdata increases because device need more power foroffloading and computing remaining data locally.V. C ONCLUSION
In this paper we present an idea for the futuristicpublic transport vehicle that is equipped with in-vehicle AP, Mobile Edge Computing and Wire-less Energy Transfer facilities and use IntelligentReflecting Surface to improve in-vehicle channelcondition. Since preserving device battery levels forcommuting users is a serious challenge, future F-PTVs will replenish commuters’ smart devices byemploying IRS aided MEC-WET. The simulationresults for IRS-aided integrated MEC-WET systemshowed significantly increased device battery’s en-ergy levels. Some challenges associated with theproposed F-PTV and their solutions are also pre-sented in this paper. Such a model is still open forresearch. As our future work, we are working onjointly optimizing τ v , µ and phase shift coefficient Φ within the proposed framework, along with de-signing MEC-WET resource scheduling scheme formultiple user scenario. Fig. 6. Ratio of energy gain-to-consumption ratio N=64, P max =0.04 Watt, N = 64 , τ v = 0 . , µ = 0 . . VI. A
CKNOWLEDGMENT
This work is submitted to IEEE Vehicular Tech-nology Magazine for publication. It is supportedby XJTLU’s Research Development Fund: RDF-20-01-15. R
EFERENCES [1] S. Jaffry, S. F. Hasan, and X. Gui, “Making a case for the mov-ing small cells,” in , 2016.[2] S. Jaffry, R. Hussain, X. Gui, and S. F. Hasan, “A compre-hensive survey on moving networks,”
IEEE CommunicationsSurveys & Tutorials , 2020.[3] N. Shinohara, “Power without wires,”
IEEE Microwave mag-azine , vol. 12, no. 7, pp. S64–S73, 2011.[4] N. Tesla, “The transmission of electric energy without wires(the thirteenth anniversary number of the electrical world andengineer),” 1904.[5] E. Basar, M. Di Renzo, J. De Rosny, M. Debbah, M.-S.Alouini, and R. Zhang, “Wireless communications throughreconfigurable intelligent surfaces,”
IEEE Access , vol. 7, pp.116 753–116 773, 2019.[6] C. Pan, H. Ren, K. Wang, M. Elkashlan, A. Nallanathan,J. Wang, and L. Hanzo, “Intelligent reflecting surface aidedmimo broadcasting for simultaneous wireless information andpower transfer,”
IEEE JSAC , 2020.[7] Q. Wu and R. Zhang, “Towards smart and reconfigurable envi-ronment: Intelligent reflecting surface aided wireless network,”
IEEE Comms. Magazine , vol. 58, no. 1, pp. 106–112, 2019.[8] T. Bai, C. Pan, Y. Deng, M. Elkashlan, A. Nallanathan, andL. Hanzo, “Latency minimization for intelligent reflectingsurface aided mobile edge computing,”
IEEE JSAC , vol. 38,no. 11, pp. 2666–2682, 2020.[9] A. S. Almansouri, M. H. Ouda, and K. N. Salama, “A cmosrf-to-dc power converter with 86% efficiency and- 19.2-dbmsensitivity,”
IEEE Transactions on Microwave Theory andTechniques , vol. 66, no. 5, pp. 2409–2415, 2018.[10] F. Zhou and R. Q. Hu, “Computation efficiency maximizationin wireless-powered mobile edge computing networks,”
IEEETrans. on Wireless Communications , vol. 19, no. 5.[11] S. Wang, M. Xia, K. Huang, and Y.-C. Wu, “Wirelessly pow-ered two-way communication with nonlinear energy harvestingmodel: Rate regions under fixed and mobile relay,”