Toward Location-aware In-body Terahertz Nanonetworks with Energy Harvesting
Filip Lemic, Sergi Abadal, Aleksandar Stevanovic, Eduard Alarcón, Jeroen Famaey
PPREPRINT 1
Toward Location-aware In-body TerahertzNanonetworks with Energy Harvesting
Filip Lemic, Sergi Abadal, Eduard Alarcón, Jeroen Famaey
Abstract —Nanoscale wireless networks are expected to revolu-tionize a variety of domains, with significant advances conceivablein in-body healthcare. In healthcare, these nanonetworks will con-sist of energy-harvesting nanodevices passively flowing throughthe bloodstream, taking actions at certain locations, and com-municating results to more powerful Body Area Network (BAN)nodes. Assuming such a setup and electromagnetic nanocom-munication in the Terahertz (THz) frequencies, we propose anetwork architecture that can support fine-grained localizationof the energy-harvesting in-body nanonodes, as well as their two-way communication with the outside world. The main novelties ofour proposal lie in the introduction of location-aware and Wake-up Radio (WuR)-based wireless nanocommunication paradigms,as well as Software-Defined Metamaterials (SDMs), to THz-operating energy-harvesting in-body nanonetworks. We arguethat, on a high level, the proposed architecture can handle(and actually benefits from) a large number of nanonodes,while simultaneously dealing with a short range of THz in-bodypropagation and highly constrained nanonodes.
Index Terms —In-body nanonetworks, terahertz, ultrasound,software-defined metamaterials, localization, energy harvesting; I NTRODUCTION R ECENT developments in nanotechnology are bringinglight to nanometer-size devices that will enable a va-riety of groundbreaking applications. In-body healthcare isamong many of the interesting domains where nanotechnologyis expected to be beneficial. Nanotechnology is envisionedto enable molecular-level detection of viruses and bacteria,high-precision drug delivery, targeted monitoring, and neu-rosurgery [1]. For enabling such applications, nanodevicescomprising an in-body nanonetwork will flow through thebloodstream, take actions upon commands at target locations,and communicate results to a more powerful Body AreaNetwork (BAN) [2], [3]. Given the small sizes of thesenanodevices, harvesting surrounding energy (e.g., from bloodcurrents) is expected to be their sole powering option [4]. Dueto their constrained energy and tiny form factors, these nanode-vices are anticipated to be passively flowing, i.e., without thepossibility of mechanical steering toward the targeted location.To support controlling the nanodevices upon reaching theirtarget locations, there is intuitively a need for knowing theircurrent locations. There is also a need for communicationbetween the outside world (i.e., BAN) and the nanodevice(e.g., for issuing control commands), as well as between thenanodevice and the outside world (e.g., for delivering device’s
F. Lemic and J. Famaey are with Internet Technology and Data ScienceLab (IDLab-Antwerpen), University of Antwerp - imec, Belgium, e-mail:{name.surname}@uantwerpen.beF. Lemic, S. Abadal, and E. Alarcón are with NaNoNetworking Centerin Catalunya (N3Cat), Universitat Politècnica de Catalunya, Spain, e-mail:{surname}@ad.upc.edu readings). One of the promising enablers for communication insuch a scenario is to utilize electromagnetic signals in the Ter-ahertz (THz) frequencies. This is because the communicationin these frequencies allows for tiny transceiver form-factors,the prime requirement for in-body nanodevices. However, theTHz band has its peculiarities, primarily pertaining to highscattering and spreading losses. Combined with the limitedcomputational and storage resources, low range of in-bodyTHz propagation (i.e., in the order of 1 cm), and constrainedpowering of nanonodes relying only on energy harvesting [3],communication between the BAN and the in-body nanonodesis currently unclear. The main challenges include i) mitigatingthe high attenuation of in-body THz propagation, ii) maintain-ing a low energy profile and complexity of the nanodevices,while iii) supporting unprecedented network scalability, andiv) enabling fine-grained localization of the nanonodes – arequirement for many of the envisioned applications [5].Current research efforts in this direction are sparse, mostlyfocusing on addressing only some of the above-stated chal-lenges. In terms of in-body localization, the existing ap-proaches focus on coarse-grained localization, resulting ina detection of the body region in which a nanonode islocated [3], which is not enough for some applications suchas neurosurgery. In terms of communication, current researchmostly focuses on random channel access-based link-layerand flooding-based network-layer protocols, both known toincrease the interferences in the system, making the envisionednanonetworks with a large number of nanonodes infeasiblein practice [5]. Finally, many approaches assume that, if thenanonetworking protocols are carefully designed, the energy-harvesting nanonodes could experience perpetual operation,which, we will argue, is impossible in practice.Based on the shortcomings identified above, we outline anetwork architecture consisting of a Software-Defined Meta-material (SDM)-based BAN, a Machine Learning (ML)-basednetwork controller, and a large number of energy-harvestingnanonodes flowing through the bloodstream. Furthermore, wepropose a high-level approach for localizing the nanonodesusing THz signals without the need for surgically implantedlocalization anchors, as the localization procedure can be basedon utilizing solely the SDM-based BAN nodes. Moreover,we propose the usage of a directional nanoscale Wake-upRadio (WuR) based on ultrasound (hence easily penetrat-ing the human body) for location-based wake-up of onlya desired set of nanonodes. This enables noise-limited (incontrast to interference-limited) nanocommunication betweenthe nanonodes and the SDM-based BAN nodes, which in turnbenefits the energy consumption of the nanonodes. Finally,the ML-based network controller serves for estimating the a r X i v : . [ c s . ET ] J a n REPRINT 2
Figure 1: High-level architectureTABLE I: Summary of requirements that healthcare applications areexpected to pose on the supporting nanonetworks [5]
Requirements In-body healthcare Considered?
Network size 10 to 10 (cid:88) Node density > nodes per cm (cid:88) Latency ms to sNetwork throughput 1-50 MbpsTraffic type bidirectional (cid:88)
Reliability very high (cid:88)
Energy consumption very low (cid:88)
Mobility high (cid:88)
Addressing individual (cid:88)
Security very highAdditional features localization & tracking (cid:88) quality of the localization, as well as the energy levels of thenanonodes, which is then used for optimizing the selectionof the nanonodes to be awaken, resulting in an optimizedselection of the communication paths between the nanonodesand the outside world.L
OCATION -A WARE T ERAHERTZ - OPERATING I N - BODY N ANONETWORKS WITH E NERGY H ARVESTING
The authors in [5] discuss the requirements that in-bodyhealthcare applications are likely to pose on the supportingnanonetworks, which are summarized in Table I with theindication of whether they are considered this work. Withthe design requirements in mind, the high-level networkarchitecture is depicted in Figure 1. A block diagram ofthe main components of a nanodevice is also depicted inFigure 1. The nanodevice consists of a module for harvestingenvironmental energy, with the most promising approaches forsuch harvesting at the nanoscale based on vibrational cyclesof ZnO nanowires generated by blood currents or heartbeats,or Radio Frequency (RF) power transfer [4]. The nanodevicealso features modules for computation, data storage, and appli-cation support, as indicated in the figure. Application supportcan be defined as sensing and/or actuation functionalities thatthe nanodevice is envisioned to enable, for example detectionand localization of cancer cells (sensing example) or targeteddrug release (actuation example). These functionalities arewell-established in the existing literature, with more detailsavailable in e.g., [6]. In the context of this work, the most interesting modulesare the THz radio for localization and data communication,and the WuR used for the wake-up of the nanonodes. Fromthe technological point of view, a THz-operating transceivercan be implemented at the nanoscale by utilizing graphene,which for low range communication could theoretically featureextremely high bandwidths (in the order of hundreds ofGigahertz (GHz)) for short distances of roughly 1 cm [7].On the physical layer we consider the usage of time-spread ON-OFF keying (TS-OOK), as it is a de-facto stan-dard communication scheme for nanocommunication in THzfrequencies. In TS-OOK, very short pulses (i.e., 100 fs long)represent logical “1”s, while logical “0”s are represented bysilences [8]. The time between the transmission/reception oftwo consecutive bits of a packet is characterized by 𝛽 andgenerally much longer (i.e., two to three orders of magnitude)than the duration of a TS-OOK pulse, which is a featureof TS-OOK allowing for unsynchronized transmission withlow chance of interference. The energy consumption modelingin this scheme has usually been carried out by attributingcertain consumed energy to transmission and reception of aTS-OOK pulse. However, one could argue that, from moretraditional wireless networking approaches such as WirelessSensor Networks (WSNs), we know the energy consumptionof any wireless transceiver is dominated by transmission,reception, and idling . This argument has been outlined in [9],where the authors consider idling energy in the overall energyconsumption of a THz-operating TS-OOK-based nanonode,in addition to the energies consumed in transmission andreception. They show that such nanonodes will mostly not haveenough energy for communication, which is a direct result ofthe long idling time 𝛽 between the transmission/reception ofconsecutive pulses or silence, as well as (more importantly)the long expected time between transmission/reception ofconsecutive packets for the expected case of low data-ratecommunication.Based on the work presented in [9], we argue that theenergy-harvesting nanonodes will have to be asleep most ofthe time in order not to deplete their energy levels in idling.Therefore, there is intuitively a need for their wake-up whenthey are needed, which is in our design of a nanodevice (Fig- REPRINT 3
Figure 2: Software-Defined Metamaterial (SDM) architecture ure 1) enabled through the WuR. Specifically, we envision theutilization of an ultrasound-based and, therefore, human body-penetrable nanoscale WuR. The design and implementationof one such system have been presented in [10], which isbased on a piezoelectric Nanoscale Ultrasound Transducer(pNUT) generating a voltage signal by picking up pressurechanges due to incident ultrasound. The authors demonstratethe WuR is able to maintain robust data transfer over a range of0.5 m when operating with a 40 kHz carrier signal modulatedat 250 Hz, which intuitively would suffice for reaching thenanonodes in any region inside the human body. With itssizes of roughly 100 µ m × µ m × OCALIZATION IN T ERAHERTZ - OPERATING E NERGY - HARVESTING I N - BODY N ANONETWORKS
The goal of localization is to estimate the locations of theTHz-operating energy-harvesting in-body nanonodes. As such,localization can be considered as the first (also mandatory)phase in the operation of the proposed architecture. Theproposed approach in enabling localization is grounded in thework by Slottke [15] (n.b., an excellent PhD dissertation).In [15], Slottke envisions a sensor network consisting ofhundreds or thousands of wireless micro-nodes with sub-millimeter dimensions. For such a setup, Slottke shows that themicro-nodes can be accurately localized by utilizing inductivecoupling between them and the anchors. He demonstratesthat such localization can be performed iteratively, so thatthe "range" of localization is increased. Iterative localizationimplies first localizing the micro-nodes close to the skin (i.e.,where there is a sufficient number of localization anchors),followed by utilizing (some of) the newly localized micro-nodes as virtual anchors in the following iterations. In thiscase, a relatively large number of, ideally optimally positioned,anchors (either real or virtual) is required for maintaining goodlocalization performance, which might be infeasible in prac-tice, argues Slottke. Finally, Slottke shows that the accuracyof such iterative localization improves with an increase in thenetwork density.Although the outlined approach is not directly applicable atthe nanoscale, it provides several guidelines on how nanoscalelocalization inside of the human body could be performed.First, it is important to realize that in-body nanonetworks areenvisioned to be extremely dense for many of the envisionedapplications (this could also be considered as a design require-ment for our system), suggesting that the accuracy of local-ization could be very high. Second, although Slottke statesthat Time of Flight (ToF) estimation cannot be performeddue to the narrowband nature of the wireless communicationcandidates considered in his work, at THz frequencies forsmall transmission distances of up to a few centimeters that isnot the case. In contrast, for such distances and THz-operatinggraphene antennas, even THz-wide nanocommunication can
REPRINT 4
Figure 3: Iterative localization scheme be achieved [14]. This implies that highly accurate two-wayToF estimation between metamaterial elements and the nanon-odes, as well as among the nanonodes, can be performed.Assuming 1 THz of available bandwidth, the raw resolutionof such sampling would result in a mm-range ranging error,which would roughly translate to the localization accuracy ofthe same order of magnitude, assuming localization in a 3-Dimensional (3D) space and minimum required number ofanchors (i.e., four). Note that two-way ToF estimation doesnot require tight synchronization between the devices, whichwould be complex to achieve for the considered setup. Alsonote that, though 1 THz of bandwidth (implying comparablesampling frequency at the nanonode level) is not practicallyfeasible, different approaches exist for two-way ToF samplingthat can mitigate this limitation and effectively result in two-way ToF ranging error that could be achieved for the THz-frequent sampling (more details in [14]).Slottke’s argument about the need for a large number of(ideally optimally positioned) localization anchors provides thejustification for the utilization of a body-worn SDM containingsuch anchors. The utilization of graphene-based THz signalsin the localization process implies tiny transceiver sizes, whichin turn implies that many potential localization anchors couldbe embedded in the smart textile, providing an unprecedenteddegree of freedom in the optimization of localization in terms of number and locations of the localization anchors. Weargue that the above considerations illustrate the potential ofTHz-, two-way ToF-, and SDM-based localization of in-bodynanonodes.The envisioned localization setup is given in Figure 3. Inthe initial step (not depicted), the SDM-based anchors are tobe localized, in case their locations are not fixed and known.For localizing the anchors one can utilize the localizationapproach proposed in [14], which is comparable to the oneoutlined here (i.e., two-way ToF-based trilateration), with thedifferences being the propagation environment (free-space vs.in-body) and iterative nature of localization. In the secondstep, the anchors are envisioned to perform two-way ToFestimation with the nanonodes, once these nanonodes areawoken using ultrasound-based WuR. The wakening shouldbe done for all the nanonodes close to the skin and, therefore,in direct communication range with multiple anchors. In thesubsequent step, some of nanonodes whose locations are nowestimated can become “virtual” anchors in the localization ofthe nanonodes deeper inside the body while others can becomedata relays toward the outside world, as depicted in the figure.Intuitively, the nanonodes should be awoken in order i) to belocalized, ii) to serve as virtual anchors in localization, iii)to serve as relays in delivering the data to the outside world(for both localization and communication), and iv) to serve
REPRINT 5 the application purposes (e.g., initiating sensing or actuationand communicating the results or success of such actions).Seemingly, this four-states level of control can be achieved bytransmitting carefully designed sequences of WuR signals.
Selected Open Challenges
We assert that the outlined localization approach intuitivelyaddresses the issues stemming from a low range of THz in-body propagation under the assumption of high number anddensity of anchors and nanonodes. We view the approach assuitable for energy-harvesting nanonodes, primarily becauseit is based on the exchange of only a small number of TS-OOK pulses and does not require tight synchronization amongthe nanonodes, nor between the nanonodes and the anchors.Nevertheless, there are open challenges to be resolved inorder to fully demonstrate the feasibility of the proposedapproach for fine-grained localization of energy-harvestingTHz-operating in-body nanonodes. Below we discuss severalof them, which we consider to be the most important.
Multiple Responses:
As numerous nanonodes are expected,their near-simultaneously retransmitted TS-OOK pulses willhave to be distinguished at the anchor or controller level, sothat the nanonodes’ locations can be accurately estimated. Wehypothesize that, in this regard, the best way forward is toconsider random back-offs, as well as trilateration constraintsaccounting for the fact that all nanonodes are in the body.In other words, the fact the TS-OOK specifies the parameter 𝛽 much larger than the pulse duration could be utilizedfor embedding the retransmissions from multiple nanonodesthrough random back-off. Hence, the chances of multipleresponses can be reduced to an extent, though the questionon the maximum number of retransmissions that could bedistinguished in the duration of one 𝛽 remains open. Note thatthis number has to be limited by the ability to time discriminatearrival of retransmitted TS-OOK pulses, so that the anchors areable to distinguish two different responses based on the delaybetween them, accounting for minimum differences in theback-off times and times needed for the signals to propagatethrough the medium. In case of multiple signals that cannot bedistinguished based on the delay between them (i.e., randomback-off), one could aim at the estimation of nanonodes’locations based on all combinations of the signals received bydifferent anchors, and discarding the ones that are estimatedto be outside of the body (i.e., trilateration constraints). Performance of Iterative Localization:
It is well-knownthat iterative localization approaches as the one outlined abovewill result in compounded localization errors. The errors com-pound as a function of the increase in the number of iterationsand, therefore, also as a function of the distance between theBAN nodes and the nanonodes (Figure 4). This suggests thatthe localization performance will be worse for the deeper partsof the body. However, it is unclear how much the performanceof the approach will degrade, given the unprecedented numberof anchors and density of in-body nanonodes, as well asthe unprecedented amount of bandwidth available (suggestinghighly accurate localization in the first iteration, in turn alsosuggesting low level of compounded localization errors). In
Figure 4: Compounding errors in iterative localization this regard, quantification of the accuracy as a function of thedistance between the anchors and nanonodes is needed.If the proposed approach eventually cannot meet the accu-racy posited by applications, one could resort to introducingadditional anchors at strategic places inside the body. Thoughthis is ideally to be avoided, in case it is needed it is an openquestion on where to position such anchors, which should beapproached simultaneously from the localization (i.e, to opti-mize the localization performance) and surgical perspectives(i.e., to minimize the complexity of the surgical procedure). Itis also interesting to observe that, in the proposed localizationsystem, multiple subsystems can be specified based on thebody regions where the SDMs are mounted. Hence, potentiallymultiple iterative localization procedures could be initiatedsimultaneously (e.g., a number of regions can be specifiedin the torso area, which seems to be the most challengingto localize). In that sense, one could simultaneously obtainmultiple estimates (i.e., one from each localization subsystem)of a single nanonode, which could then be used for enhancingthe accuracy of localization, especially for the nanonodespositioned deep in the body. Note that the same approach couldbe used for scalability or reliability-related enhancements.Finally, iterative localization, combined with communica-tion with the outside world upon the execution of eachiteration, could result in high localization latencies, whichwould in turn reduce the accuracy as the controller wouldonly be aware of the outdated locations of the passivelyflowing nanonodes. In this regard, further research is neededin terms of establishing and potentially minimizing the latencyof communication, primarily the one required for localization.C
OMMUNICATION IN TH Z - OPERATING E NERGY - HARVESTING I N - BODY N ANONETWORKS
As mentioned, the location estimates will feature certainlocalization errors, which will be higher for the estimates"deep" inside the body compared to the ones closer to theSDM-based anchors. In the following, we will assume thelocations of the nanonodes in a given time instance areestimated and their quality (characterized by the expected levelof error in localization) of such estimates is known. Alongthese lines, we postulate that the location-aware wirelesscommunication paradigm is a strong candidate for enabling
REPRINT 6
Figure 5: Location-based wake-up process two-way communication between the outside world and thein-body nanonodes, which will be demonstrated in this section.There are two main communication-related aspects con-sidered in this work, i.e., addressing and routing. In termsof addressing, we posit that the nanonodes can be awokenand, therefore, addressed individually, if their locations canbe estimated and if the quality of such estimation can beobtained. Our high-level approach in waking up the nanonodesis based on issuing directional WuR signals from the SDM, asindicated in Figure 5. Specifically, by utilizing an SDM onecould sequentially transmit multiple directed WuR signals insuch a way that only the nanonode(s) to be awoken receive(s)all of them. By doing that, one could wake up only a desirednanonode and, therefore, achieve individual addressing of thatnanonode based on its practically-relevant location. Moreover,by waking up only desired nanonodes, in contrast to wakingup all of them in a given region, one could create a multi-hoprouting path for data delivery from a particular nanonode tothe outside world, for both localization and communication.Existing literature provides a variety of approaches forlocation-based link quality estimation for the purposes ofDevice-to-Device (D2D) link establishment, positioning of amobile relay, handover between base stations or network tech-nologies, etc. Though these approaches are usually outlined formacro-scale wireless networks, some of them are interestingin this context as they assume the availability of erroneous location information, where the levels of average localizationerrors can be quantified (e.g., [16]). Based on [16] we canoutline, on a high-level, our approach for location- and locationquality-based route selection. Specifically, as we show in Fig-ure 6, based on the distances between nanonodes, the expectederrors of estimated locations of the nanonodes, and some high-level knowledge about the path-loss in the environment ofinterest, a metric characterizing the expected link quality (e.g.,expected Signal-to-Noise Ratio (SNR)) between two devicescan be established. The metric can then be used for selectingthe nanonodes to be awoken based on the WuR approachoutlined before. Note that the mentioned approaches can beused in case of both devices featuring a localization error, aswell as when the location of one device is perfectly accurate,which is intuitively suitable for all cases envisioned in thiswork (i.a., anchors’ locations are known or can be estimatedwith a certain level of accuracy, while nanonodes’ locationestimates always feature a certain localization error).
Figure 6: Location-based route selection process
Selected Open Challenges
We believe our approach addresses the infeasibility offlooding-based routing for energy-harvesting nanonetworks,as well as enables individual addressing of energy-harvestingnanonodes. Nonetheless, there are certainly open challenges,with selected ones outlined below.
Communication Reliability:
The question of the level ofcommunication reliability the proposed approach can achieveremains open. In this context, the reliability has to be con-sidered jointly with the latency of data delivery. This isprimarily because the nanonodes are envisioned to be passivelyflowing, thus it is rather important to deliver desired controlcommands to initiate certain actions at an appropriate time, sothat the actions can be executed at a target location. However,this interplay between nanonodes’ (and potentially anchors’)mobility on the one hand, and the reliability and latency ofdata delivery on the other is yet to be investigated. Note thatone could, if needed, consider specifying multiple routes fordata delivery in order to increase the reliability of communi-cation along the Ultra-Reliable Low-Latency Communication(URLLC) paradigm.
Network Control:
The proposed communication systemrelies on the estimated locations of the nanonodes, and shouldintuitively account for the localization errors, especially giventhe iterative localization process where the errors are com-pounded with each new iteration. This implies that there is aneed for estimating the quality of location estimates, ideallyon a per-estimate basis. We are aware of only one suchattempt [17], where the authors propose a ML-based systemfor estimating localization errors on a per-estimate basis usingsolely the raw data (i.e., Received Signal Strength (RSS))utilized for generating location estimates. They follow bydemonstrating a comparatively higher accuracy of estimationof localization errors than the one that can be obtained throughregion-based benchmarks. Nevertheless, the proposal in [17]is focused on the macroscale and considers a different local-ization approach (i.e., fingerprinting vs. trilateration) and rawdata (i.e., RSS vs. two-way ToF), and traditional networkingtechnologies (i.e., SigFox and LoRa vs. in-body nanonetwork)and frequencies (i.e., sub-GHz vs. THz). Hence, the utility ofsuch a system for the problem at hand is yet to be established.
Nanonode’s Energy Level Estimation:
Finally, accountingfor the energy levels of the nanonodes in their wakening
REPRINT 7 and route selection is of prime importance, given that thealternative could result in an unnecessary wake-up of nanon-odes whose energy levels are nearly depleted. To the bestof our knowledge, proposals for such modeling do not existfor the nanonodes assumed in this work (Figure 1). Weenvision this functionality to be supported by the networkcontroller, together with the discussed functionality of esti-mating the quality of localization. Moreover, we believe theenergy level modeling should be based on the region in whicha nanonode is located at a given moment. Based on thenanonode’s location one could postulate the recently harvestedamount of surrounding energy. For example, the nanonode’samounts of harvested energy would intuitively increase asthe nanonode moves closer to the heart, assuming that theenergy is harvested from the heartbeats. One could also aimat utilizing localized historical information for nanonode’senergy level characterization. Specifically, if one would beable to characterize that in a relatively recent time-frame therewas very little activity for the nanonodes in a given region,one could argue the energy levels of the nanonodes in theregion are high, hence these nanonodes could be awoken moresuccessfully. These considerations again demonstrate severalbenefits of location-aware nanonetworks in this context.C
ONCLUSION
In this work, we have outlined a nanonetworking archi-tecture for in-body THz-operating energy-harvesting nanonet-works. This has been done with the goals of enabling local-ization of the energy-harvesting nanonodes, as well as theirtwo-way communication with the outside world. The proposedarchitecture can handle a large number of nanonodes andvery short range of THz in-body propagation, and simulta-neously be supported through highly constrained nanonodes.We have posited the architecture assuming passively flowingnanonodes. Nevertheless, we believe the architecture to beapplicable for medical nanorobots for enabling their navigationand communication with the outside world [18].In summary, the envisioned applications pose extreme re-quirements on the supporting nanonetworks, among othersin terms of scalability, energy consumption, reliability, andfeatures such as localization. We argue that, to meet theseunprecedented requirements, the nanonetworks will have to behighly optimized. In the optimizations, one should considerand potentially utilize the available sources of informationabout the network and its performance. In that regard, weenvision location-, energy-, or generally speaking contextual-awareness in combination with machine learning as onepromising way forward. Our future work will consist ofinvestigating this hypothesis in more details and aiming ataddressing some of the open issues underlined in this paper.A
CKNOWLEDGMENT
This work was received funding from the European Unionvia the Horizon 2020 Marie Curie Actions (grant nr. 893760)and Future Emerging Topics calls (grant no. 736876). R
EFERENCES[1] N. Agoulmine et al. , “Enabling communication and cooperation inbio-nanosensor networks: Toward innovative healthcare solutions,”
IEEE Wireless Communications , vol. 19, no. 5, pp. 42–51, 2012.[2] I. F. Akyildiz et al. , “Nanonetworks: A new communicationparadigm,”
Computer Networks , vol. 52, no. 12, pp. 2260–2279, 2008.[3] F. Dressler et al. , “Connecting in-body nano communication withbody area networks: Challenges and opportunities of the internet ofnano things,”
Nano Communication Networks , vol. 6, no. 2, pp. 29–38,2015.[4] G. Piro et al. , “On the design of an energy-harvesting protocol stackfor body area nano-networks,”
Nano Communication Networks , vol. 6,no. 2, pp. 74–84, 2015.[5] F. Lemic et al. , “Survey on terahertz nanocommunication and net-working: A top-down perspective,” arXiv preprint , 2019.[6] S. Canovas-Carrasco et al. , “On the nature of energy-feasible wirelessnanosensor networks,”
Sensors , vol. 18, no. 5, p. 1356, 2018.[7] J. M. Jornet and I. F. Akyildiz, “Graphene-based plasmonic nano-antenna for terahertz band communication in nanonetworks,”
IEEEJournal on selected areas in communications , vol. 31, no. 12, pp. 685–694, 2013.[8] J. M. Jornet et al. , “Femtosecond-long pulse-based modulation forterahertz band communication in nanonetworks,”
IEEE Transactionson Communications , vol. 62, no. 5, pp. 1742–1754, 2014.[9] F. Lemic et al. , “Idling energy modeling and reduction in energyharvesting terahertz nanonetworks for controlling software-definedmetamaterials,”
IEEE Journal on Emerging and Selected Topics inCircuits and Systems , vol. 10, no. 1, pp. 88–99, 2020.[10] P. Simeoni et al. , “Long-range ultrasound wake-up receiver with apiezoelectric nanoscale ultrasound transducer (pnut),” in , IEEE, 2020, pp. 849–852.[11] J. Chen et al. , “Deep-subwavelength control of acoustic waves inan ultra-compact metasurface lens,”
Nature communications , vol. 9,no. 1, pp. 1–9, 2018.[12] C. Liaskos et al. , “A new wireless communication paradigm throughsoftware-controlled metasurfaces,”
IEEE Communications Magazine ,vol. 56, no. 9, pp. 162–169, 2018.[13] S. Walia et al. , “Flexible metasurfaces and metamaterials: A review ofmaterials and fabrication processes at micro-and nano-scales,”
AppliedPhysics Reviews , vol. 2, no. 1, p. 011 303, 2015.[14] F. Lemic et al. , “Localization in terahertz-operating energy harvestingsoftware-defined metamaterials,” arXiv preprint , 2020.[15] E. Slottke, “Inductively coupled microsensor networks: Relay enabledcooperative communication and localization,” Ph.D. dissertation, ETHZurich, 2016.[16] F. Lemic, A. Behboodi, J. Famaey, and R. Mathar, “Location-baseddiscovery and vertical handover in heterogeneous low-power wide-area networks,”
IEEE Internet of Things Journal , vol. 6, no. 6,pp. 10 150–10 165, 2019.[17] F. Lemic, V. Handziski, M. Aernouts, T. Janssen, R. Berkvens, A.Wolisz, and J. Famaey, “Regression-based estimation of individualerrors in fingerprinting localization,”
IEEE Access , vol. 7, pp. 33 652–33 664, 2019.[18] M. T. Nistor et al. , “Nanorobots with applications in medicine,”in