Distributed Antenna System for Mitigating Shadowing Effect in 60 GHz WLAN

Transcription

1 Distributed Antenna System for Mitigating Shadowing Effect in 60 GHz WLAN Citation for pubished version (APA): Wang, Q., Debbarma, D., Lo, A., Cao, Z., Niemegeers, I., & Heemstra, S. (2015). Distributed Antenna System for Mitigating Shadowing Effect in 60 GHz WLAN. Wireess Persona Communications, 82(2), DOI: /s DOI: /s Document status and date: Pubished: 01/05/2015 Document Version: Pubisher s PDF, aso known as Version of Record (incudes fina page, issue and voume numbers) Pease check the document version of this pubication: A submitted manuscript is the version of the artice upon submission and before peer-review. There can be important differences between the submitted version and the officia pubished version of record. Peope interested in the research are advised to contact the author for the fina version of the pubication, or visit the DOI to the pubisher's website. The fina author version and the gaey proof are versions of the pubication after peer review. The fina pubished version features the fina ayout of the paper incuding the voume, issue and page numbers. Link to pubication Genera rights Copyright and mora rights for the pubications made accessibe in the pubic porta are retained by the authors and/or other copyright owners and it is a condition of accessing pubications that users recognise and abide by the ega requirements associated with these rights. Users may downoad and print one copy of any pubication from the pubic porta for the purpose of private study or research. You may not further distribute the materia or use it for any profit-making activity or commercia gain You may freey distribute the URL identifying the pubication in the pubic porta. If the pubication is distributed under the terms of Artice 25fa of the Dutch Copyright Act, indicated by the Taverne icense above, pease foow beow ink for the End User Agreement: Take down poicy If you beieve that this document breaches copyright pease contact us at: openaccess@tue.n providing detais and we wi investigate your caim. Downoad date: 07. Apr. 2019

2 Wireess Pers Commun (2015) 82: DOI /s Distributed Antenna System for Mitigating Shadowing Effect in 60 GHz WLAN Qing Wang Diptani Debbarma Anthony Lo Zizheng Cao Ignas Niemegeers Sonia Heemstra de Groot Pubished onine: 27 December 2014 Springer Science+Business Media New York 2014 Abstract The 60 GHz unicensed frequency band has been adopted to support high data rate WLAN appications. The probem of this frequency band is that the wireess signa is vunerabe to the shadowing of objects. Especiay in indoor scenarios, humans may frequenty bock the signa paths that cause disconnections of devices. This probem can be mitigated by using mutipe antennas that can create rich signa paths between the devices. The idea presented in this paper is to empoy a distributed antenna system (DAS) architecture enabed by radio-over-fiber technoogy to aeviate the shadowing probem. In the DAS architecture, mutipe remote antenna units (RAUs), each of them equipped with an antenna array, are connected with the same WLAN access points via optica fibers. But there are questions that need to be answered, incuding the beamforming strategy with mutipe RAUs, and the pacement and the required number of the RAUs. These questions are reated to factors ike room size, popuation density, etc. We discuss these questions in this paper and present our findings Anthony Lo is currenty with Huawei Technoogies. The work was done whie he was with Deft University of Technoogy. Q. Wang (B) D. Debbarma Z. Cao I. Niemegeers S. Heemstra de Groot COBRA Research Institute, Eindhoven University of Technoogy, Den Doech 2, 5612 AZ Eindhoven, The Netherands e-mai: Qing.Wang@tue.n D. Debbarma e-mai: D.Debbarma@tue.n Z. Cao e-mai: Z.Cao@tue.n I. Niemegeers e-mai: I.G.M.M.Niemegeers@tue.n S. Heemstra de Groot e-mai: S.HeemstradeGroot@tue.n A. Lo Deft University of Technoogy, Deft, The Netherands e-mai: Anthonyo@ieee.org

3 812 Q. Wang et a. through theoretica modes and extensive simuations, which can be referred to for practica impementations. We find that the proposed 60 GHz DAS can significanty improve the ink connectivity. Even when ony two RAUs are used, the signa coverage can be improved to an acceptabe percentage in heavy shadowing scenarios. Keywords 60 GHz Beamforming Radio-over-fiber Shadowing WLAN 1 Introduction Recenty, the 60 GHz unicensed band has been adopted to support emerging short-range and high data rate WLAN appications ike uncompressed video streaming, high-speed fie transfer, high-definition wireess dispay, etc. These appications produce data traffic in the order of muti-gigabits per second, which is difficut to achieve in the congested ow WLAN frequency bands ike 2.4 or 5 GHz. On the contrary, there is a 5 GHz continuous spectrum bock gobay avaiabe in the 60 GHz band [1]. This suggests that gigabits per second communication can be achieved even when simpe moduation schemes are appied. The signas in this high frequency band experience strong path oss which imits the communication distance to a few meters [2]. This characteristic makes the 60 GHz band suitabe for indoor networks where most of the high data rate appications are presumed to be ocated. An important issue of 60 GHz communication is that the signa is susceptibe to the bockage of obstaces ike humans and was due to its short waveength [3]. Therefore, 60 GHz communication is usuay constrained to ine-of-sight (LOS) condition. Especiay in indoor scenarios, waking humans may frequenty bock the wireess inks, which generates more probems than stabe objects. It is reported that the human body can attenuate the signa by more than 20 db [4], which is hard to overcome by using ony beamforming techniques, particuary when considering power reguations and potentia heath consequences. For instance, it has been shown in [4] that the channe is thus unavaiabe for about 1 2 % of the time in the presence of one to five persons in a room. Severa approaches have been reported in the iterature for aeviating the shadowing effect of humans. One practica soution is to transport the signas through refective paths induced by arge surfaces ike was. Athough the refective paths wi introduce more attenuation, typicay db [5] or even ower for some refective materias [6], beamforming may be used to compensate this oss. But there is aso the possibiity that no refective paths can be found since the refections strongy depend on the physica environment, ike the refectivity of the was and furniture. So, [7] proposes to pace artificia meta refectors on the was to create more refections. It is caimed that the connectivity can be significanty boosted by pacing a few refectors at optimized positions. Another promising approach is to use reays to forward the signas of the bocked devices to the destination which is reported in, e.g. [8]. For WLAN, mutipe access points (APs) can be used to cover the same area, such that the mobie stations (STAs) can find aternative APs when experiencing a bockage [9] proposes to pace a few synchronized APs to serve the same room and shows that the visibiity can be improved sufficienty with two APs. A simiar method was recenty presented in [10], but aternativey it uses a centra controer to seect the best visibe AP to continue the bocked transmissions. Due to the ack of channe information at the controer, the seection has to be done based on packet transmission faiure reports from the APs. As stated in the paper, this may cause additiona deay to the packet deivery. Besides these techniques, a new promising approach that we propose in this paper is to expoit spatia diversity. Spatia diversity can increase the robustness of the wireess ink

4 Distributed Antenna System 813 by taking advantage of the spatia separation of mutipe antennas, which is aso appicabe for 60 GHz communication. 60 GHz ink mandates LOS paths, as opposed to the ower frequencies that can aso count on non-ine-of-sight (NLOS) paths. So, the antennas need to be separated by an adequate distance which shoud be much arger than the carrier waveength, such that the mutipe transmitters can have different shadowing properties. As pointed out in [9] which addresses muti-ap diversity, this distance needs to be as arge as severa meters. A simiar requirement hods for the spacing of antennas for spatia diversity against shadowing effect. This requirement makes the spatia diversity approach ess feasibe for devices that are constrained in size ike ce phones, than for arge devices ike TV sets. However, for WLAN access network design, this imitation is ess important. Particuary, distributed antenna systems (DASs) are aready empoyed for the ower frequency WLANs [11,12], where the antennas of an AP are paced far apart. The same architecture can be empoyed for 60 GHz WLAN. The difference woud be the high frequency and arge bandwidth of the 60 GHz signa. So optica fiber is more suitabe than other wired mediums ike coax cabe, for connecting the remote antenna units (RAUs) with an AP, thanks to its huge bandwidth and ow oss [13]. The way of transporting radio signas via optica fiber is termed radioover-fiber (RoF) [14]. In comparison with other anti-bockage approaches, DAS architecture has the advantages of: No coordination cost for controing the additiona antennas whie this is usuay necessary for reaying schemes and muti-ap diversity More LOS paths can be provided instead of NLOS paths that have arger power oss The contribution of this paper is to examine the DAS concept in 60 GHz WLAN for mitigating the shadowing probem. This invoves buiding theoretica modes and conducting extensive simuations to investigate the beamforming approaches with mutipe RAUs, and the pacement and the necessary number of RAUs. The rest of the artice is organized as foows: Section 2 introduces the RoF based DAS system for 60 GHz WLAN communication and derives the mathematica beamforming mode. Section 3 describes the 60 GHz physica channe mode and its simuation. Section 4 presents the setup of our simuations. The resuts and discussions are given in Sect. 5. Section 6 concudes the paper. 2 System Mode In this section, we wi describe the RoF network and derive the mathematica beamforming mode of the 60 GHz DAS. The concept of RoF for indoor networking has been demonstrated in, e.g. [12,15]. Due to the sma scae of indoor scenarios, the required fiber engths are usuay in the order of hundreds of meters. Therefore, the signa quaity oss, ike the SNR drop of the radio signa is margina in comparison with the wireess channe. Thus, it is ignored in this paper. One can consider the RoF ink between the RAUs and the AP as an idea wired connection. 2.1 RoF System for 60 GHz WLAN The RoF system for 60 GHz WLAN communication is depicted in Fig. 1. It is intended to provide wireess access for the whoe indoor area. As mentioned previousy, 60 GHz communication is isoated by the was, which makes it necessary to depoy at east one RAU in each room. Mutipe WLAN APs are are co-ocated in the Centra Station (CS). The APs are connected with the RAUs through the radio-optica interface and the optica network. The

5 814 Q. Wang et a. Externa network Centra Station AP 1 AP 2 AP 3 Radio-Optica Interface Optica ink Optica Network RAU RAU RAU Room 1 AP X RAU Fig. 1 An radio-over-fiber based distributed antenna system for 60 GHz WLAN 1,1 1,2 RAU 1 1, 1 2 AP,1 Channe STA,2 RAU K, Fig. 2 Beamforming mode of distributed antenna system function of the radio-optica interface is to convert the signas between eectrica and optica domains such that the radio signas can be transported between the APs and RAUs through the optica network (e.g. with a star topoogy [15]). As proposed in this paper, mutipe RAUs are connected with the same AP for a singe room to mitigate the human shadowing effect. E.g. two RAUs are depoyed in Room 1 in Fig. 1. This resuts in the aforementioned DAS architecture [16]. Since beamforming is needed to enhance the signa strength for extending the signa transmission range, each RAU is equipped with mutipe antennas, i.e. an antenna array. These RAUs can be utiized in different ways. In this paper, we consider two strategies: banket transmission and seective transmission. In banket transmission, a the RAUs are used for transmitting signas; whie ony the RAU that gives the best signa quaity is used in seective transmission. 2.2 Beamforming Mode of Distributed Antenna System We derive the beamforming modes by focusing on the DAS of a singe AP which is iustrated in Fig. 2 in the form of a MIMO system. Assume that there are K RAUs depoyed in a room. Each RAU has a panar antenna array with a tota of M antenna eements arranged on a rectanguar grid with equa spacing aong both x and y dimensions. We assume haf waveength spacing in both dimensions. Denote the number of antenna eements in the two dimensions with M x and M y, respectivey, and M = M x M y. The mobie stations are assumed to have the same configuration with N antenna eements in a singe antenna array. The wireess signa uses OFDM moduation with L subcarriers and a tota bandwidth of B Banket Transmission For banket transmission, a the RAUs are used in beamforming. The sum transmission power of a RAUs is fixed and is equay aocated to a subcarriers. The foowing derivations are

6 Distributed Antenna System 815 given with unit sum transmission power. The compete channe matrix of the th subcarrier between the K RAUs and the STA is expressed as [ ] H = H (0), H (1) (K 1),...,H (1) which is a N KM matrix. And each eement H (k) is the channe matrix of the kth RAU and the STA of size N M. So, the received signa of the th subcarrier at the STA is x y = H w + n (2) KM where x is the symbo transmitted on the th subcarrier with unit power, i.e. E( x 2 ) = 1. n is the N 1 Gaussian noise vector with zero mean and E(n n H ) = δ 2 I. w is the transmit beam steering vector for a the antenna eements at the K RAUs, which satisfies w H w = KM (3) ( such that the tota transmission power is normaized to 1, i.e. E w x 2) = 1. Each KM component of w is a compex number that modifies both the ampitude and phase of the signa. The transmission power of the RAUs is therefore controed according to the channe states. The decision variabe after receive beamforming is then given by x r = c H y = c H H w + c H n = H c,w [] + c H n (4) KM KM where c is the beamforming vector at the receiver side and c H c = N; H c,w [] =c H H w is the effective channe gain after beamforming. Then, we can obtain the noise power of the th subcarrier as The SNR of this subcarrier is E ( c H n 2 ) = Nδ 2 (5) ( Hc,w ) E [] KM x 2 SNR,banket (c, w) = Nδ 2 = H c,w[] 2 KMNδ 2 (6) To achieve the highest SNR, we need to find the beamforming vectors, w and c,tomaximize the effective channe gain H c,w []. Ideay, the beamforming vectors can be optimized for each subcarrier. Assuming perfect channe knowedge, these vectors can be found through singuar vaue decomposition (SVD) of the channe matrix H [16]. That is, the channe matrix of the th subcarrier is first decomposed into H = U D V H (7) where U and V are unitary matrices of size N N and KM KM, respectivey, and D is a N KMmatrix with the ( j, j)th eement, denoted by λ, j for j = 1, 2,...,min(N, KM), being the singuar vaues and the other eements are zero, and λ, j = λ,k for j > k. Then, the optimum beamforming vectors are derived as c = Nu,1, w = KMv,1 (8) where u,1 and v,1 are the first coumn of U and V. Finay, the maximum SNR can be obtained as γ = max c,w x c H H w 2 MNδ 2 = λ,max 2 δ 2 (9)

7 816 Q. Wang et a. Since each subcarrier may have a different SNR, the average SNR shoud be taken as the ink performance. We empoy in this paper the exponentia effective SNR mapping (EESM) scheme which is widey adopted, ike in [16,17], to characterize the OFDM signa quaity. The reation between the effective SNR and the SNR of the subcarriers is expressed as [18] γ eff = β n ( 1 L 1 L =0 ( exp γ ) ) (10) β where β is a parameter depending on the moduation and coding scheme (MCS). The vaues of β can be found in [18] and it is set to two in this paper, which is a typica vaue for QPSK signaing Seective Transmission In seective transmission, the RAU with the best channe condition is used to transmit signas with fu power. For exampe, if the kth RAU is chosen, the received symbo of the th subcarrier after beamforming is given by r (k) So the SNR of the received signa is = c H H (k) w x + c H n = H c,w (k) [] x + c H n (11) M M SNR (k) = H (k) c,w[] 2 MNδ 2 (12) The optimum beamforming is performed by using the foowing beamforming vectors c = Nu (k),1, w = Mv (k),1 (13) where u (k),1 and v(k),1 are the first coumns of U(k) of H (k), i.e. H (k) = U (k) V (k) kth RAU is derived as D (k) And the effective SNR is thus expressed as ( γ eff (k) 1 L 1 = βn exp L and V (k), which are derived from the SVD. Then, the maximum SNR of the th subcarrier using the λ (k) γ (k),max 2 = δ 2 (14) =0 ( γ (k) β Then, the RAU with the highest effective SNR is seected, i.e. the effective SNR of seective transmission is given by γ eff,seective = max k )) (15) γ eff (k), k = 0,...K 1 (16) The above derivations of beamforming are based on the assumption that the wireess channe is known at both the transmitter and the receiver side. Normay, the transmitter can send probing signas to the channe and then the receiver cacuates the channe information and feeds it back to the transmitter. Due to the arge overhead of the feedback procedure, perfect channe knowedge is usuay not avaiabe in practice. However, we woud ike to eucidate the potentia of 60 GHz DAS, so fu channe knowedge is assumed for the anaysis

8 Distributed Antenna System 817 to get more genera resuts. For a practica beamforming approach, one can refer to other studies ike [17,19], which use imited channe information feedback or codebook based beamforming GHz Channe Mode and Simuation Method The beamforming modes introduced in the ast section require the channe information of a antenna pairs at arbitrary positions. It is necessary to simuate 60 GHz channes in a 3D space so as to incorporate factors ike the physica position of the RAUs and the shadowing effect of humans, which have significant impact on the wireess channe. Because of the physica characteristics of 60 GHz signa, the wireess channe simuation is different from the ower frequencies. This section wi describe the 60 GHz mathematica channe mode and our simuation method based on 3D ray tracing, which shoud give acceptabe accuracy GHz Channe Mode In the ower WLAN frequency bands ike 2.4 and 5 GHz, the wireess channe modes are usuay separated into arge scae and sma scae fading modes. However, this approach is not appicabe for 60 GHz due to the appication of high directiona steerabe antennas or beamforming agorithms, which eads to the fitering out of a singe custer of the propagation channe [20]. The path oss and channe impuse response (CIR) significanty depend on the characteristics of this custer, which in genera does not aow deveoping independent modes for them. In this paper, we use a genera discrete-path channe mode, which characterizes the signa paths or rays with parameters incuding the path gain, the time deay, the ange of arriva (AoA), and the ange of departure (AoD). This aows us to examine the performance of beamforming at both the transmitter and the receiver sides. The 3-dimensiona CIR mode is given as foows and is aso depicted in Fig. 3. L 1 h(t,ϕ t,θ t,ϕ r,θ r ) = α δ(t τ )δ(ϕ t ϕ t, )δ(θ t θ t, )δ(ϕ r ϕ r, )δ(θ r θ r, ) (17) =0 where α is the compex gain (incuding ampitude and phase) of the th path with AoD [ϕ t,,θ t, ],AoA[ϕ r,,θ r, ], and time of arriva (ToA) τ.thetermδ( ) represents the Dirac deta function. ϕ represents the eevation ange which is in the range of [ π/2,π/2]; θ is the azimuth ange in the range of [0, 2π]; L is the number of mutipath components. Notice Fig. 3 Iustration of channe impuse response mode Refector 1 ( 1, 1,,1,,1,,1,,1) ( 0, 0,,0,,0,,0,,0) Tx ( 2, 2,,2,,2,,2,,2) Rx Refector 2

9 818 Q. Wang et a. that we misuse L to represent both the number of subcarriers and signa paths, which shoud be identifiabe in accordance with the context. For an OFDM system, the channe gain of each subcarrier between the mth Tx antenna and the nth Rx antenna, is denoted as H n,m [],where = 0,...,L 1 represent the indices of subcarriers. It is formay obtained through the foowing steps [16]. First, the CIR between the two antennas, h n,m (t), is convoved with the puse shaping fier c(t), which is typicay a raised cosine fiter; and then samped at the symbo rate, which resuts in the discrete-time CIR (DT-CIR), h n,m [n]. Then, H n,m [] is derived from the L-point DFT of h n,m [n]. The N M channes between the kth RAU and the STA are grouped as H (k) with the (n, m)th eement given by H n,m[]. (k) This channe matrix can then be used in the beamforming process presented in Sect GHz Channe Simuation with 3D Ray Tracing Because of the miimeter-wave characteristics of 60 GHz signa, ray tracing is accepted as a more accurate way to simuate the physica channe in comparison with the empirica modes [8,21]. So, we use a 3D ray tracing software caed Radiowave Propagation Simuator (RPS) [5,22] to obtain the CIRs. RPS can predict, with verified accuracy, the paths between arbitrary transmitter and receiver positions and antenna patterns. The information of a the possibe mutipath components, incuding path gain, time deay, AoD and AoA can be recorded and exported into text fies. This aows us to re-construct the CIRs in the exact form of Eq. (17). For a singe antenna simuation, the channes can be directy simuated by RPS. However, the RAUs and STAs a have mutipe antennas that resuts in a arge number of channes from the combination of them, which makes it too time-consuming to obtain a the CIRs by RPS. Aternativey, we obtain the CIRs through the foowing method. First, we use RPS to simuate the CIRs by using isotropic antennas at both the transmitter and receiver sides for a predefined positions. Then, cacuate the CIRs of the channes of a antenna eements by adding the phase shifts caused by the geometric position of the antenna eements. The phase shifts are cacuated for each mutipath component. So finay, from a singe simuated channe between isotropic antennas, we can construct the mutipe channes between a antenna eements mathematicay. This approach is based on the fact that the antennas are coocated with sma spacing (haf carrier waveength) in between, so the refectors can be assumed to be in the far fied. The path difference of different antenna eements can be simpified into the phase difference induced by their reative distance. The mathematica description for obtaining the spatia CIRs can be found in [23]. We formuate this theory in the foowing in combination with the channe mode provided previousy. Given the CIR between the first antenna eement pair of the transmitter and receiver, say L 1 h nx =0,n y =0,m x =0,m y =0(t) = α δ(t τ )δ(ϕ t ϕ t, )δ(θ t θ t, )δ(ϕ r ϕ r, )δ(θ r θ r, ) =0 (18) The coordinate of the (m x, m y )-th Tx antenna eement is ((m x 1)λ/2,(m y 1)λ/2, 0). Its reative distance to the first eement (0, 0) in the direction of (ϕ t,θ t ) is d mx,m y (ϕ t,θ t ) = (m x 1) λ 2 sin θ t sin ϕ t + (m y 1) λ 2 sin θ t cos ϕ t (19)

10 Distributed Antenna System 819 which introduces a phase shift of φ t (ϕ t,θ t ) = d m x,m y (ϕ t,θ t ) 2π (20) λ where λ is the waveength of the carrier frequency. The phase shift at the receiver can aso be derived simiary, which is denoted as φ r (ϕ r,θ r ). The phase shifts are then added to a the paths of the first antenna eement pair and then derive the channe between the (m x, m y )th and the (n x, n y )th eement pair as L 1 h nx,n y,m x,m y (t) = α exp ( j φ t (ϕ t,,θ t, ) exp ( j φ r (ϕ r,,θ r, ) =0 δ(t τ )δ(ϕ t ϕ t, )δ(θ t θ t, )δ(ϕ r ϕ r, )δ(θ r θ r, ) (21) These cacuated CIRs are then used in the simuations in the next section. 4 Simuation Setup Intuitivey, the performance of the DAS beamforming system is reated to severa factors incuding room size, popuation, the number of RAUs that can be used, the position of the RAUs, and the transmission strategy adopted. We then set up a number of scenarios to discover their impacts. We define three reaistic indoor scenarios, iving room, office room and conference room, to represent different room sizes. The engths, widths and heights of the rooms are given beow: Living room: 6 m 6m 3m Office room: 10 m 6m 3m Conference room: 10 m 10 m 4m For a the rooms, one side of the was is made of concrete with a window of 4 m 1m size in the center, which represents the exterior wa. The other three sides are interior was covered with pasterboard. A door is ocated on the opposite side of the exterior wa with the size of 1.5 m (width) 2 m(height) (cosed in simuation). An exampe is shown in Fig. 4, showing the top-down view of the iving room setup. Humans are modeed as absorptive rectanguar prisms with the same height (1.75 m) and width (0.5 m), random facing direction, and random positions. In each snapshot, which represent an independent RPS simuation, the humans are randomy paced with uniform distribution in a room. For each simuation scenario, we take 20 snapshots and then average the coected resuts. The EM properties of the different materias used in the RPS simuations are isted in Tabe 1. A the RAUs are paced on the ceiing which we think is a suitabe pace for practica instaations. Moreover, higher heights can provide better visibiity to the stations [8]. The RAU positions are divided into two groups, i.e. edge positions and center positions. The coordinates of the positions are the geometrica center and edge of the ceiing pane with 0.2 m resoution. The exact position of them wi be scaed with respect to the room size. These positions are defined in this way aso because they are easy to ocate in reaity. The numbers in Fig. 4 represents the indices of the RAU positions. The receivers are paced on the horizonta panes of three different heights: 0, 0.5 and 1 m, with a grid spacing of 0.5 m, to represent the possibe receiver positions inside a room. This grid spacing is chosen because

11 820 Q. Wang et a. L y W Window Door x Edge posi ons Center posi ons Receiver posi ons Fig. 4 Top-down view of the simuation scenario Tabe 1 Reative permittivities of various materias at 60 GHz [5,24] Name Materia Thickness (m) Rea Imaginary Exterior wa Concrete Interior wa Pasterboard Ceiing Concrete Ground Wood Window Soda borosiicate gass Door Wood Human we assume the human width to be the same vaue such that the human shadowing effect can be sufficienty captured. The RPS simuation is done at 60 GHz center frequency with 2 GHz bandwidth. The tota bandwidth is divided into 512 subcarriers in the OFDM system. The tota transmission power is imited to 10 db m. The noise figure is assumed to be 6 db. An exampe of a iving room scenario in RPS simuation is given in Fig. 5. The spheres are the transmitters equipped with isotropic antennas. The red one represents the active transmitter the others are inactive. The ines between the transmitters and the receiver position represent the signa paths and the coors of the ines indicate different signa strengths. 5 Simuation Resuts and Discussions In this section, we conduct different simuations to investigate the effect of the above mentioned factors. Specificay, we are trying to answer the foowing questions: How do mutipe RAUs improve the connectivity of 60 GHz devices? What is the performance difference of banket and seective transmission strategies? Where shoud the RAUs be paced? How many RAUs shoud be used? Ceary, these questions are interconnected but they provide different aspects of understanding the proposed 60 GHz DAS.

12 Distributed Antenna System 821 Tx -70 dbm -78 dbm Rx -86 dbm Human -94 dbm -102 dbm -110 dbm Fig. 5 RPS simuation mode of iving room 5.1 Coverage Improvement with Mutipe RAUs This subsection simuates the coverage improvement with mutipe RAUs. The simuations are conducted in the iving room scenario with 1 4 RAUs. Each RAU and STA is equipped with an isotropic antenna. The tota transmission power is set to 10 dbm for a simuations and is equay aocated to the RAUs. A RAUs are used for transmitting signas simutaneousy. The position of the RAUs are configured as: 1 RAU 13; 2 RAUs (11, 15); 3 RAUs (11, 13, 15); 4 RAUs (3, 11, 15, 23), which are indicated by the position indices shown in Fig. 4. The received power across the 1 m-height receiver pane (with 0.2 m resoution) is given in Fig. 6. It can be observed that the shadowed receiver positions have more than 10 db ower power than the visibe positions. These shadowing positions are eiminated when more RAUs are used. The improvement is a resut of the RAU diversity that offers richer signa paths, especiay the LOS paths that take the major portion of the received signa power. Thus, when some RAUs are obstructed by objects, the other RAUs may sti be reachabe such that the connectivity can be sustained. With beamforming, the signa strength of the visibe positions can be enhanced to achieve higher data rates. However, the shadowed positions that have much ower signa power, may require other approaches to reach the desired power eve. 5.2 Banket and Seective Transmission Strategies As stated in Sect. 2, there are two strategies to use the RAUs in a DAS. Different beamforming strategies use different antennas which impies that the requirement of channe information and beamforming compexity are different. The seective transmission strategy has ower beamforming compexity due to its simpicity in cacuating the beamforming vectors. This eads us to compare their performance under different conditions so as to understand the differences. Since it is a necessity for 60 GHz devices to use directiona antennas to reach onger distances to cover the entire room, we assume a panar antenna array with 4 isotropic eements at each RAU and each STA. The ocations of the RAUs are arranged as foows (simpified in consideration of the geometrica symmetry of the rooms): 1 RAU: ocated at the center of the ceiing 2 RAUs: ocated at the symmetric center or edge positions of the ceiing

13 822 Q. Wang et a. Fig. 6 Power distribution with respect to different number of RAUs with equa power aocation. The white pixes are the positions occupied by humans. Red dots represent the position of the RAUs a 1RAU,b 2RAUs, c 3RAUs,d 4RAUs 3 RAUs: one ocated at the center of the ceiing and the other two at the symmetric center and edge positions of the ceiing 4 RAUs: a RAUs ocated at the symmetric center or edge positions of the room So, there are severa candidate positions for a specific number of RAUs. A the possibe positions are isted in Tabe 2 whie the symmetric positions are omitted. For each scenario with various number of RAUs, a the possibe RAU ocations are simuated. In each simuation, ten humans are randomy paced in a room. Then, the received SNR of each STA position is cacuated with different beamforming approaches. This process is repeated for 20 times. In the end, the RAU positions that give the highest average SNR of the 20 simuations are seected, to remove the effect of the RAUs position. The averaged outage probabiity curves of different scenarios with different number of RAUs and SNR threshods are potted in Fig. 7. The outage probabiity is defined as the percentage of area within a room that is ower than the SNR threshod. Due to the discretized receiver positions, there are severa receiver positions occupied by humans that cannot be covered, so the ower bound of outage probabiity is not equa to zero. The confidence intervas of the data are not

14 Distributed Antenna System 823 Tabe 2 RAU positions Number of RAUs (a) Center positions (7, 19) (8, 18) (12, 14) 3 (7, 13, 19) (8, 13, 18) (12, 13, 14) 4 (7, 9, 17, 19) (8, 12, 18, 14) Number of RAUs (b) Edge positions 1 2 (3, 23) (2, 24) (1, 25) (6, 20) (11, 15) 3 (3, 13, 23) (2, 13, 24) (1, 13, 25) (6, 13, 20) (11, 13, 15) 4 (3, 11, 15, 23) (2, 4, 22, 24) (1, 5, 21, 25) (6, 16, 10, 20) potted in the figure for the sake of carity, but the widest 95 % confidence interva is [ 2.65, 1.93], which suggests acceptabe accuracy. For a scenarios, a singe isotropic antenna cannot provide sufficient signa strength to cover the entire room for a arge range of SNR threshods. With a singe RAU that has a 4-eement antenna array, the distance range is extended but sti eaves a arge portion of receiver positions in outage. By using mutipe RAUs, this outage probabiity is significanty owered, even when using ony one additiona RAU. From the figures, we can observe two effects of using the 60 GHz DAS. One effect is the mitigation of human shadowing as a resut of signa path diversity. This effect is obvious when the SNR threshod is ow, which can be reached by LOS paths and strong NLOS paths. Human shadowing is the main reason of ow receiving power. In a scenarios, the outage probabiity drops dramaticay in both banket and seective transmission as a consequence of the DAS. But they do not show much performance difference. However, the performance differences between different number of RAUs are obvious. Therefore, for soving the human bockage probem, the number of RAUs is the main factor rather than the beamforming strategies. The other effect is a consequence of the distributed signa transmission scheme in the DAS that aso improves the coverage. This phenomenon is more evident at higher SNR threshods, for which shadowing is not the ony cause of ow signa strength. The distance between the transmitters and the receivers aso has a notabe impact on signa attenuation. The DAS architecture is beneficia from another two perspectives. First, it shortens the average distance between transceivers by spreading the antennas over the space. The propagation oss is therefore reduced, which significanty increases the signa quaity of the distant positions. This is an essentia outcome of empoying DAS in ower frequency bands for coverage improvement [25]. Respectivey, both banket and seective transmission experience reduced outage probabiity at the higher SNR threshods and the coverage increases with more RAUs. Second, a the antennas at the RAUs are used in banket transmission that resuts in high beamforming gain, and even arger than seective transmission that uses ony one RAU. This aso boosts the signa quaity, and expains why banket transmission performs better than seective transmission at the higher SNR threshods. It can be deduced that this effect is more observabe for room sizes arger than the ones given in this paper, as the dimension of the room wi impact more evidenty on the signa strength due to arger distance between tranceivers.

15 824 Q. Wang et a. Fig. 7 Outage probabiity of different transmission strategies. a Living room, b office room, c conference room Outage probabiity (%) Omni 40 1RAU 2RAUs Banket 30 2RAUs Seective 3RAUs Banket 20 3RAUs Seective 10 4RAUs Banket 4RAUs Seective SNR threshods (db) (a) Outage probabiity (%) Omni 40 1RAU 2RAUs Banket 30 2RAUs Seective 20 3RAUs Banket 3RAUs Seective 10 4RAUs Banket 4RAUs Seective SNR threshods (db) (b) Outage probabiity (%) Omni 40 1RAU 2RAUs Banket 30 2RAUs Seective 20 3RAUs Banket 3RAUs Seective 10 4RAUs Banket 4RAUs Seective SNR threshods (db) (c)

16 Distributed Antenna System 825 Tabe 3 Outage probabiity with 6dB SNR threshod Number of RAUs Center (%) Edge (%) Center (%) Edge (%) Center (%) Edge (%) (a) Banket transmission Living room Office room Conference room (b) Seective transmission Living room Office room Conference room Position of the RAUs The positions of the RAUs are of great importance since they not ony impact the avaiabiity of LOS paths but aso the propagation oss between the STAs and RAUs. The optimum position of the RAUs shoud be reated to the room size and the transmission strategies. These reationships are eucidated through the foowing simuations. We divide the previousy defined RAU positions into two groups, i.e. center positions and edge positions. The simuation parameters are the same as in the ast subsection. We then coect the outage probabiities of the best RAU positions in each group, for banket and seective transmission strategy, respectivey. Notice that, the foowing discussion is under the assumption of a specific human density. We aso investigate at both ow and high SNR threshods. We first set the SNR threshod to 6 db as it can be achieved in a scenarios under LOS condition but not under a NLOS conditions (observed from Fig. 7) when using a singe RAU, so that we can evauate the performance of human shadowing aeviation. The resuts are shown in Tabe 3. The data shows that, the best edge positions perform better than the best center positions in most simuation setups. This impies that arger RAU separation can provide better LOS visibiity against human bockage. Furthermore, the outage probabiities of the best center positions are times higher than the best edge positions, which may suggest that optimizing the positions of the RAUs can improve the performance to some extent. Above that, there is aso a notabe performance difference for the RAU positions inside the center and the edge groups. We find that the argest outage probabiities are [1, 58 %] and [27, 128 %] higher than the owest ones in the two groups, respectivey. We understand that, in the edge groups, the performance has stronger dependence on the ocation of the RAUs. Center positions have ower variations due to smaer position difference. We then give the indices of the best RAU positions in Tabe 4, indicated by the position indices of Tabe 2. Interestingy, we find that diagona center positions seem to be the most optimum center position in a configurations. The optimum edge positions, however, change in different scenarios and for different transmission strategies. Hence, if we want to pace the RAUs at edge positions, it is necessary to choose the positions according to the environment, particuary the room size, in order to have optimum performance. This requirement is reaxed if we choose the center positions, which can reduce the instaation compexity. For this reason, we might suggest to pace the RAUs at the diagona center positions in practice, athough the performance may not be the optima.

17 826 Q. Wang et a. Tabe 4 Best RAU positions of different transmission strategies Number of RAUs Center Edge Center Edge Center Edge (a) Banket transmission Living room Office room Conference room (b) Seective transmission Living room Office room Conference room Tabe 5 Outage probabiity with 13dB SNR threshod Number of RAUs Center (%) Edge (%) Center (%) Edge (%) Center (%) Edge (%) (a) Banket transmission Living room Office room Conference room (b) Seective transmission Living room Office room Conference room Another case is the performance at higher SNR threshods. We then increase the threshod to 13 db and coect the outage probabiities in Tabe 5. As mentioned previousy, the signa propagation oss dominates the reasons of outage at higher SNR threshods. The RAUs positions become a more important factor of performance. Except that the center positions seem superior when using two RAUs, we do not see a reguar pattern for 3 and 4 RAUs in the two tabes. This suggests that it is crucia to optimize the position of the RAUs at higher SNR threshods, which is understandabe since we need to optimize the distances between the RAUs and the STAs in the room. In addition, when using more RAUs, the probabiity of outage drops significanty for a cases because the propagation oss is compensated by beamforming. However, for overcoming path oss, the easier and more cost-effective way is to increase the antenna array size instead of depoying more RAUs. A practica 60 GHz system may cover a arge range of SNR threshods for different moduation and coding schemes (MCSs). So, the cases of ow and high SNRs may both exist. A baance is required for pacing the RAUs to compromise most appications. Since most rooms have simiar sizes as the ones given in this paper, we beieve that the ack of LOS ink avaiabiity is the main probem. So, more attention shoud be paid to improving LOS ink visibiity, which is equivaenty the ower-snr case discussed here.

18 Distributed Antenna System 827 Outage probabiity (%) Center positions 1 RAU 2 RAUs 3 RAUs 4 RAUs Outage probabiity (%) Edge positions 1 RAU 2 RAUs 3 RAUs 4 RAUs Number of humans (a) Number of humans Outage probabiity (%) Center positions 1 RAU 2 RAUs 3 RAUs 4 RAUs Outage probabiity (%) Edge positions 1 RAU 2 RAUs 3 RAUs 4 RAUs Number of humans (b) Number of humans Fig. 8 Outage probabiity versus the number of humans. a Banket transmission, b seective transmission 5.4 The Number of RAUs As mentioned earier, the number of RAUs has a significant roe in soving the bockage probem. It is reported in [6] that the visibiity probabiity can be significanty improved by ony two APs. A simiar resut is aso found in our anaysis of DAS systems with a sma number of RAUs. We simuate a conference room scenario with different numbers of humans (5 30) and different RAU configurations. The RAUs are paced at the optimum positions given in Tabe 4. For simpifying the simuation, we ony simuate the receiver ayer at 1m height. For each case, we take ten snapshots. The outage probabiity curves of 6dB SNR threshod for different setups are potted in Fig. 8.The error bars indicate the 95 % confidence intervas of the averaged data. Notice that the 1-RAU case is the same in a figures.

19 828 Q. Wang et a. With arger popuation, the frequency of human bockage shoud be heavier. From the figures, we see that the outage probabiity increases amost ineary with the number of humans. However, when the popuation increases, we observe that the transmission strategies and RAU positions do not affect the performance much since the outage probabiity differences are reativey sma. On the contrary, the number of RAUs pays the main roe with respect to popuation. There is a cear improvement when more RAUs are invoved. With more than three RAUs, the outage probabiity is reduced cose to the ower bound (cose to the vaue of the factor of popuation and the number of receiver positions in the simuations), especiay for the cases of using four RAUs. The argest outage probabiity drop happens when two RAUs are used. By using three or four RAUs, ony a few percentages of outage probabiity are decreased, even when the popuation grows to 30. Therefore, two RAUs coud be a practica choice in the sense cost-effectiveness for a moderate popuation. 6 Concusions The paper has proposed to improve 60 GHz communication performance by empoying a DAS architecture. Particuary, 60 GHz DAS empoying mutipe RAUs can significanty mitigate the shadowing effect of obstaces. We have vaidated this idea by means of buiding mathematica modes and conducting extensive simuations. The questions of how to utiize the mutipe RAUs for beamforming, where to pace them and how many shoud be empoyed, have been answered in the paper. We found that, for aeviating the shadowing effect, the number of RAUs pays a very important roe. For a reasonabe human density, two RAUs coud be a practica soution that dramaticay increases the 60 GHz ink connectivity. They shoud be paced sufficienty far apart from each other. For a coarse instaation, they can be paced at the diagona centers of the ceiing, which is appicabe to most practica room shapes. And higher performance can be achieved by optimizing the RAUs positions according to the ayout of a room. However, banket and seective transmission do not exhibit significant difference between each other, in terms of soving shadowing probem. This may suggest that seective transmission is more preferabe due to its ower compexity. The studies in this paper assumed that the beamforming methods are optima, but this is not feasibe in practice. So, further investigations, ike considering codebook based beamforming that is currenty empoyed by 60 GHz WLAN/WPAN standards, are of great importance. However, the resuts presented in this paper shoud give meaningfu insights for these future works. Acknowedgments The work in this paper is supported by Dutch IOP Gencom MEANS project. References 1. Smuders, P. (2002). Expoiting the 60 GHz band for oca wireess mutimedia access: Prospects and future directions. IEEE Communication Magazine, 40(1), Rappaport, T. S., Murdock, J. N., & Gutierrez, F. (2011). State of the art in 60-GHz integrated circuits and systems for wireess communications. Proceedings of the IEEE,,99(8), Zhou, L., & Ohashi, Y. (2012). Efficient codebook-based MIMO beamforming for miimeter-wave WLANs. In Proceedings of the 2012 internationa symposium on persona indoor and moobie radio communications (PIMRC 2012) (pp ). 4. Coonge, S., Zaharia, G., & Zein, G. E. (2004). Infuence of the human activity on wide-band characteristics of the 60 GHz indoor radio channe. IEEE Transactions on Wireess Communications, 3(6),

20 Distributed Antenna System Genc, Z., Rizvi, U. H., Onur, E., & Niemegeers, I. (2010). Robust 60 GHz indoor connectivity: Is it possibe with refections? In Proceedings of 2010 IEEE 71st vehicuar technoogy conference (VTC Spring) (pp. 1 5). 6. Sato, K., Manabe, T., Ihara, T., Saito, H., Ito, S., Tanaka, T., et a. (1997). Measurements of refection and transmission characteristics of interior structures of office buiding in the 60-GHz band. IEEE Transactions on Antennas and Propagation, 45(12), Sawada, H., Takahashi, S., & Kato, S. (2012). Disconnection probabiity improvement by using artificia muti refectors for miimeter-wave indoor wireess communications. In Proceedings of 2012 IEEE 75th vehicuar technoogy conference (VTC 2012-Spring) (Vo. 61, pp ). 8. Genc, Z., Ocer, G. M., Onur, E., & Niemegeers, I. (2010). Improving 60 GHz indoor connectivity with reaying. In Proceedings of 2010 IEEE internationa conference on communications (ICC 2010) (pp. 1 6). 9. Sato, K., & Manabe, T. (1998). Estimation of propagation-path visibiity for indoor wireess LAN systems under shadowing condition by human bodies. In Proceedings of the th IEEE vehicuar technoogy conference (VTC 1998) (Vo. 3, pp ). 10. Zhang, X., Zhou, S., Wang, X., Niu, Z., Lin, X., Zhu, D. & Lei, M. (2012). Improving network throughput in 60GHz WLANs via muti-ap diversity. In Proceedings of 2012 IEEE internationa conference on communications (pp ). 11. Crisp, M. J., Li, S., Watts, A., Penty, R. V., & White, I. H. (2007). Upink and downink coverage improvements of g signas using a distributed antenna network. Journa of Lightwave Technoogy, 25(11), Sauer, M., Kobyakov, A., & George, J. (2007). Radio over fiber for picoceuar network architectures. Journa of Lightwave Technoogy, 25(11), Guiory, J., Meyer, S., Sianud, I., Umer-mo, A. M., Charbonnier, B., Pizzinat, A., et a. (2010). Radioover-fiber architectures. IEEE Vehicuar Technoogy Magazine, 5(3), Heath, R., Peters, S., Wang, Y., & Zhang, J. (2013). A current perspective on distributed antenna systems for the downink of ceuar systems. IEEE Communications Magazine, 51(4), Beas, J., Castanón, G., & Adaya, I. (2013). Miimeter-wave frequency radio over fiber systems: A survey. IEEE Communications Surveys and Tutorias, 15(4), Yoon, S., Jeon, T., & Lee, W. (2009). Hybrid beam-forming and beam-switching for OFDM based wireess persona area networks. IEEE Journa on Seected Areas in Communications, 27(8), Lee, H.-H., & Ko, Y.-C. (2011). Low compexity codebook-based beamforming for MIMO-OFDM systems in miimeter-wave WPAN. IEEE Transactions on Wireess Communications, 10(11), Bankenship, Y., & Sartori, P. (2004). Link error prediction methods for muticarrier systems. In: Proceedings of 2004 IEEE 60th vehicuar technoogy conference (VTC 2004-Fa) (Vo. 6, pp ). 19. Ramachandran, K., Prasad, N., Hosoya, K., Maruhashi, K., & Rangarajan, S. (2010). Adaptive beamforming for 60 GHz radios: Chaenges and preiminary soutions. In Proceedings of the 2010 ACM internationa workshop on mmwave communications: From circuits to networks (mmcom 10) (pp ). 20. Akeyama, A. (2004). Channe modes for 60 GHz WLAN systems. IEEE /0334r Smuders, P. (2004). 60 GHz indoor radio propagation comparison of simuation and measurement resuts. In Proceedings of the 2004 IEEE 11th symposium on communications and vehicuar technoogy (SCVT 2004) (pp. 2 5). 22. Deiner, J., Hbner, J., Hunod, D., & Voigt, J. (2008). RPS radiowave propagation simuator user manua version 5.4. Actix GmbH. 23. Erte, R. B., Cardieri, P., Sowerby, K. W., Rappaport, T. S., & Reed, J. H. (1998). Overview of spatia channe modes for antenna array communication systems. IEEE Persona Communications, 5(1), Langen, B., Lober, G., & Herzig, W. (1994). Refection and transmission behaviour of buiding materias at 60 GHz. In Proceedings of 1994 IEEE 5th internationa symposium on persona, indoor and mobie radio communications (PIMRC 1994) (pp ). 25. Winters, J., Kobyakov, A., & Sauer, M. (2011). Picoces with MIMO and ce bonding for WLANs. IEEE Journa on Seected Areas in Communications, 29(6),

21 830 Q. Wang et a. Qing Wang received his B.Eng. degree in Teecommunication Engineering from Xidian University, China, in In 2011, he obtained his M.E. degree in Information and Communication Engineering from Nationa University of Defense Technoogy, China. He is currenty working towards his Ph.D. degree in COBRA Research Institute at Eindhoven University of Technoogy, The Netherands. His research interests incude wireess and mobie communications, wireess indoor networks, WLAN and radio-overfiber technoogies. Diptani Debbarma was born in Agartaa, India in He received his B.E from Sathyabama University, India in Eectronics and Teecommunication in 2004 and graduated as one of the University Topper. He obtained his Masters (M.E) from the prestigious Indian Institute of Science (IISc), India in Teecommunication, 2010, and was the recipient of the Ministry of Human Resource and Deveopment (MHRD) schoarship. Currenty he is pursuing his PhD at Eindhoven University of Technoogy (ECO group) working for the Dutch IOP-GenCom project MEANS. His research interest incudes Wireess Communication, Wireess Network, Stochastic Network Contro, Network Coding impementation in wireess MAC, Information Theory, MU-MIMO, Cooperative and Distributed MIMO. Anthony Lo [SM] received combined B.S. and B.E., and Ph.D. degrees in 1992 and 1996 from La Trobe University, Austraia. He is currenty a Principa Scientist at Huawei Technoogies R & D Sweden. He was an assistant professor at Deft University of Technoogy and, prior to the academic appointment, he was a research engineer at Ericsson EuroLab working on UMTS/HSPA and LTE. His research interests incude M2M Massive MIMO, communications, inteigent transportation systems, smart grids, and the wireess network coud.

22 Distributed Antenna System 831 Zizheng Cao received his Bacheor of Science degree on eectronic information science and technoogy (awarded Outstanding Graduate of Hunan Province, 2007) from Hunan Norma University, Changsha, China. And he obtained his Master of Engineering on teecom engineering (awarded Outstanding thesis of master degree of Hunan Province, 2010) from Hunan University, Changsha, China. He is currenty working towards the Ph.D. degree at Eindhoven University of Technoogy, Eindhoven, The Netherands supervised by Prof. Ton Koonen. Funded by NWO project SOWICI, he started to work on energy efficient access/in-home optica networks empowered by integrated optics, ow-compexity digita signa processing, and fexibe optica network design. In SOWICI, the broadband optica mm-wave beam steering system based on integrated optica circuit was buit for a hybrid optica-wireess network. Further, the energy efficient and broadband operation in such network were optimized by the dedicated physica optica ayer design and the impementation of advanced DSP. These research activities produce a series of interesting scientific resuts. Zizheng Cao has pubished 15 firstauthor peer-reviewed IEEE/OSA journa artices, incuding an invited paper in JLT. He aso has an invited tak about the integrated optica radio beam steering system in PIERS His research artices have been cited for 448 times, with a H-index of 12 and a i10 factor of 14 (googe schoar). His research interests incude modeing and design of integrated optica circuit, microwave photonics, advanced DSP, and physica ayer design of optica network. He is a student member of the IEEE Photonics Society. He serves as an active reviewer for many journas, incuding Journa of Lightwave Technoogy, Photonics Technoogy Letters, Photonics Journa, Journa of Optica Communications and Networking, Optics Communications, Opitcs Express, and Optics Letters. Ignas Niemegeers got an MSc degree in Eectrica Engineering from the University of Gent, Begium in 1970, an MSc in 1972 and a PhD degree in 1978 in Computer Engineering from Purdue University, USA. From 1978 to 1981 he was a system designer at Be Teephone Mfg. Cy, Antwerp, Begium. From 1981 to 2002 he was professor at the University of Twente, The Netherands. From 1995 to 2002 he was Scientific Director of the Centre for Teematics and Information Technoogy (CTIT) of the University of Twente. From 2002 unti 2012 he was chairman of the Teecommunications Department and professor in Wireess and Mobie Communications at Deft University of Technoogy. Since August 2012 he is emeritus professor at Deft University of Technoogy and advisor to the Centre for Wireess Technoogy at Eindhoven University of Technoogy, The Netherands. He was invoved in many European research projects and reviewer for many projects. His present research interests are 5G, Radio-over-fiber networks, 60 GHz networking, energy-harvesting networks. Sonia Heemstra de Groot hods M.Sc. degrees in Eectrica Engineering from Universidad Naciona de Mar de Pata, Argentina and Phiips Internationa Institute/NUFFIC, The Netherands. She obtained the Ph.D. degree in Eectrica Engineering at the University of Twente, The Netherands, in Since 2012 she is a fu professor at Eindhoven University of Technoogy where she hods the parttime chair in Heterogeneous Network Architectures. Before she has had assistant and associate professor positions at the University of Twente and a fuprofessor position at the Deft University of Technoogy in Persona and Ambient Networking. After having worked some years as a senior researcher at Ericsson EuroLab, The Netherands, she cofounded the Twente Institute for Wireess and Mobie where she has been Chief Scientist from 2003 to Her expertise and interests are in the areas of wireess and mobie communications, wireess networks, Internet of Things, wireess security, persona and ambient networks (co-inventor