New Trends in Radio Network Positioning

Transcription

1 New Trends in Radio Nework Posiioning Kamiar Radnosrai, Fredrik Gunnarsson, Fredrik Gusafsson Deparmen of Elecrical Engineering, Linköping Universiy, Linköping, Sweden Ericsson Research, Linköping, Sweden, Absrac Posiioning in radio neworks is a well esablished research area. The dominaing approach has been ha posiioning algorihms are implemened in he higher levels of he communicaion sysem based on posiion relaed informaion derived in he lowes (physical) layer. Examples of measuremen include received signal srengh (RSS), ime of arrival (TOA), angle of arrival (AOA), and fusion and filering is a sraighforward ask. The echnical driver for posiioning has been E911 and for commercially driver comes from locaion based services and logisics managemen. These demands are fundamenal in he developmen of posiioning in fuure radio neworks sandards. There is oday a rend for accuracy demand ha goes beyond wha can be achieved wih odays measuremens. Anoher rend is ha measuremens and posiioning algorihms are approaching each oher, so some pars of he posiioning are performed on he chip-ses (lowes layer) and low-level measuremens are available o he operaing sysem (highes level). The purpose of his survey is o describe his rend in more deail, wih examples of developmens in cellular neworks as well as WiFi and Blueooh. I. INTRODUCTION Awareness of he posiion of a device, eiher in absolue erms or relaive o a reference locaion, is becoming increasingly imporan. Use cases include emergency calls posiioning, navigaion, gaming, auonomic vessels, logisics, flee managemen, proximiy services, locaion-based services, nework managemen o menion a few. Up o dae, i has mainly been emergency call posiioning ha has driven much of he work in cellular neworks due o regulaory requiremens. However, some use cases can also be addressed via crude posiioning such as cell ID associaion. The emergency call posiioning requiremens by he Federal Communicaions Commission (FCC) in he Unied Saes have been refined several imes, iniially wih requiremens on nework-based posiioning, and subsequenly wih igher requiremens on mobile-assised posiioning [1], [2]. Recenly [3], FCC has ye again refined he requiremens o give paricular aenion o requiremens for posiioning of indoor devices. These requiremens are presened as a roadmap wih sricer requiremens over ime, and considering all mobiles, boh oudoors and indoors. The requiremen is a horizonal accuracy corresponding o a dispachable address or wihin a radius of 50 meers for 40 percen of all wireless 911 calls wihin wo years, gradually ighened o 80 percen of he wireless 911 calls wihin six years. Furhermore, for verical posiioning informaion, compaible mobiles shall deliver baromeric pressure informaion wihin hree years. In addiion, operaors commi o develop a specific verical locaion accuracy meric ha would be used as he sandard for any fuure deploymen, and o be generally adoped wihin eigh years. An alernaive, or a complemen o pressure repors, is a plausible naionwide Naional Emergency Address Daabase (NEAD) conaining locaions of WiFi access poins and Blueooh beacons. Wireless nework posiioning is based on measuremens gahered eiher a he base saions, a he mobile saions and repored o he nework, or a combinaion. I may be based on snapsho measuremens or ime series of measuremens. The survey aricles [2], [4] [8] provide furher informaion abou posiioning in wireless neworks and associaed sandardizaion. This paper aims a exending he measuremen survey in [8] o include recen advances in sandardizaion and echnology. The ouline of he paper is as follows. Secion II presens a general sensor fusion framework for posiioning based on generic measuremens, and Secion III describes available measuremens for posiioning in wireless neworks, separaed in differen key caegories and associaed wih accuracy saisics. Moreover, Secion IV addresses some posiioning aspecs in consideraions of he presened framework, models and measuremens, followed by some conclusions in Secion V. y() II. POSITIONING FRAMEWORK AND SYSTEM MODEL Physical Layer Spaial Fusion Modaliy Fusion and Temporal RSS α Fingerprining/ Filering Trilaeraion TOA τ Doppler β Tri/ Mulilaeraion Velociy and Posiion Hybrid Localizaion/ Filering Fig. 1: Differen layers of fusion for posiioning in radio neworks. Figure 1 provides an overview of he informaion flow in posiioning algorihms as hey appear in lieraure oday. This secion will summarize he differen levels of fusion in erms of signal models. The underlying noaion is as follows. Denoe he hreedimensional mobile posiion a ime by p = (X, Y, Z ) T, and he known reference poin (base saion) posiions by p()

2 = (X i, Y i, Z i ) T. The reference poins can in general move in ime, as is he case in some ad-hoc neworks and sensor neworks problems. A generic measuremen y i relaive o reference poin i is a funcion h ype (p, p i ) of boh mobile posiion and reference posiion posiion, and i is subjec o an uncerainy e i. Hence p i A. Level 1: Radio Measuremens y i = h ype (p, p i ) + e i (1) On he physical layer, a pilo symbol s i () is ransmied, and he receiver samples he signal n z i () = α ik s ( ) β ik ( τ ik + eik (). (2) k=0 Here, we have inroduced The impulse response α ik of he (muli-pah) channel. The ime delay per incoming pah τ ik. In case of pure line-of-sigh, his is equal o p p i /c, where c is he speed of ligh. The Doppler shif β ik, which scales ime. Suppose ha hese parameers can be esimaed accuraely in he receiver. Then, we can define he following posiion relaed measuremens: TOA (ime of arrival) corresponds o τ ik, which yields he high-level measuremen y i,toa = p p i + e i,toa (3) where he error e i,toa capures boh he esimaion error and he model error due o mulipah. The LTE posiioning evaluaion in [9] presens a Gaussian esimaion error of variance 8.5 meers based on simulaions wih a realisic receiver. If here are several synchronized ransmiers and he receiver in un-synchronized, we ge a TDOA (ime difference of arrival measuremen) by compuing pair-wise differences of τ, yielding y ij,tdoa = p p i p p j + e i,toa e j,toa (4) Round rip ime (RTT) measuremens are basically he sum of wo TOA measuremens in uplink and downlink, respecively, y i,toa = 2 p p i + e i,toa,uplink + e i,toa,downlink. (5) AOA (angle of arrival) can be compued o comparing differen delays τ o muliple anennas in he receiver, and he high-level measuremen is y i,aoa = arcan(x X i, Y Y i ) + e i,aoa (6) For wo nearby receivers, separaed in disance D meers (less han half a wavelengh), we can use he simple formula y i,aoa = arccos(cτ/d). There is a rich lieraure on array processing wih much more sophisicaed algorihms, see [10]. Anoher more coarse mehod o esimae AOA is based on he anenna lobe diagram, ha usually provides a secor of widh beween 60 and 180 degrees. This can be refined [11] based on non-coheren power measuremens of muliple secors a he same sie o abou degrees in he secor overlap regions. RSS (received signal srengh) is basically he oal energy n k=0 α2 ik. The log-disance model saes ha y i,rss = P b log ( p p i ) + e i,rss. (7) Here, P is he ransmied power (which migh be known) and b is he pah loss exponen (usually beween 2 and 3). The Doppler parameer β provides a measuremen of he relaive speed y i,doppler B. Level 2: Spaial Fusion = d p p i d + e i,doppler. (8) Given N ransmiers, we have a se of equaions y i,ype = h ype ( p p i ) + e i,ype, i = 1,..., N. (9) where ype is eiher TOA, TDOA, AOA, RSS or Doppler. Given a sufficien large N, his forms a sysem of equaions which under quie general condiions has a unique soluion in p. There is a vas lieraure describing he principles of rilaeraion (TOA), mulilaeraion (TDOA), riangulaion (AOA), rilaeraion or fingerprining (RSS), and muli-saic radar (Doppler). C. Level 3: Modaliy Fusion and Temporal Filering Using measuremens of differen modaliy (kind) is no a problem and is covered in he same nonlinear se of equaion framework as (9). The only difference is ha we lis all ypes available. Here, also oher sensor informaion can be enered. One complemenary sensor ha resolves he ricky verical posiion problems is baromeric pressure, ha provides y i,baro = Z Z i + e i,baro. (10) The inerial sensor uni in smar phones is oday used o compue a lo of moion relaed parameers, for insance he moion sae (sill, walking, running, vehicle, ec.) and sep deecions. These can be used on he device for posiioning, bu also ransmied o he nework. Furher, i is common knowledge how o add a moion model of he kind x +1 = f(x, v ). (11) A nonlinear filer gives a smooh inerpolaion of he posiion esimaes, and can also be used o predic fuure posiions.

3 III. IMPROVED MEASUREMENTS This secion coninues he brief overview of radio nework measuremens in Secion II, and provides a pracical survey similar o [8], bu exended wih recen measuremens and sandards. Lower layer echniques for providing hese measuremens are no addressed, and insead we refer o [5], [12] for 2G, [2], [13] for 3G and, [14], [15] for 4G/WiFi respecively, while [16] provides an overview of error sources in general sysems. Table I provides ypical accuracy levels of posiion measuremens covering various wireless sysems. Firs, he iming measuremens are based on synchronizaion echniques. Addiionally, i is assumed ha measuremens are performed in LOS condiions. LTE-relaed values are provided from he bes case scenario, i.e. highes frequency in iming measuremens, o wors case condiions. Moreover, in mos of he measuremens, PRS pilo signal of he LTE sysem, which mainly gives he bes performance compared o oher pilo signals, is invesigaed. A. Received signal srengh Consider he RSS measuremen (7). The reference (ransmission) power is assumed eiher o be known or broadcased hrough he nework. Having he ransmied and received powers known o he sysem, he channel aenuaion can be compued. In addiion o he measuremen noise inroduced in 7, one migh also consider he diffracion facor. This way, 7 can be re-wrien as: y i,rss = P b log ( p p i ) + e i,rss + d i,rss. (12) where d i is he diffracion. The ypical measuremen noise (e ) is on he order of 3 db in sandard deviaion. Propagaion also feaures diffracion effecs which resembles as shadow fading ha is a lowpass spaial process. A number of mehods exiss o deal wih he diffracion error. One way is o lump i ogeher wih he measuremen error. This way, sd(e +d ) 6-10 db. Anoher approach is o capure hese variaions in a model/daabase which essenially forms he fingerprining mehod. A hird way is o assume ha he shadow fading is only presen in he inermediae o far field from anenna, bu no in he near field. This way, in he near field, he only source of error is he measuremen noise. The approach inroduced below is based on he inheri feaure of diffracion. Tha is, by aking advanage of he high emporal correlaion feaure of diffracion, one can inroduce he difference equaion as below resuling in d o be discarded. y i,rss y i,rss 1 = b log ( p p i ) + b log ( p 1 p i ) (13) + e i,rss e i,rss 1 In addiion o RSS-based channel aenuaion compuaion, I is also possible o uilize a prediced or measured spaial digial map wih RSS values. B. Time of arrival and round-rip ime Handshake procedures generae RTT/ToF/range measuremens according o (5). [11] provides a novel mehod for RTT calculaions in LTE sysems on he uplink iming alignmen mechanism. Performance analysis performed by [17], inroduce differen levels of accuracies based on he pilo symbols used by he LTE sysem and also he frequency. The bes esimaion can be achieved in he 20 MHz sysem, as expeced, while PRS pilo signal is used. In his scenario, he σ CRLB of measuremen lies beween 0.1 ns o 10 relaing o various SNR values, i.e. 20 E s /σ 2 [db] 20. Keeping he pilo signal as PRS bu changing he sysem frequency o 1.4 MHz resuls in a σ CRLB approximaely beween 1 ns o 100 ns in he same range of SNR. In GSM, as anoher example, a 26 bi known raining sequence in each burs is found by correlaion in iming measuremens. The bi duraion is abou 554 m, bu using coninuous ime echniques, ime synchronizaion down o a fracion (say 100 m) is possible. Similar figures hold for he 3G sandard universal mobile elecommunicaions sysem (UMTS), where he ravel ime is esimaed using he firs deeced ray in he RAKE algorihm from he known pilo symbols. C. Baromeric Pressure Lack of accuracy and reliabiliy on op of limied availabiliy o oudoor environmens of GPS-based elevaion esimaion, is he main moivaion behind his mehod. An example of a possible use of a baromeer sensor in verically oriened aciviies is presened in [18]. Generic measuremen of aliude via a pressure sensor relaive a known pressure informaion a a reference aliude provides he aliude measuremen (10). Sources of error in his mehod are horizonal and verical disances o he reference poin(s) and he ime gap. Three sors of reference poins exis. Meeorological saions for weaher forecas already deployed by he naional meeorological agencies. These saions have coarse spaial densiy on he ampliude of ens of kilomeers and low updae frequency of almos once an hour. The elevaion of a person wih a smar phone in oudoor environmen goen from Digial Elevaion Model (DEM)-map based on his curren locaion is called a DEM reference. The hird reference poin is based on an ad-hoc fashion of smar phones wihin he sysem. In case a reference pressure is no available, [19] presens a framework ha does no depend on any special infrasrucure and provides accurae elevaion measuremens using only smar phones. I is considered ha hree sources of errors inroduced above, follow N(0, σ 2 hs ), N(0, σ2 vs), and N(0, σ 2 ) respecively. The final Accuracy obained applying he sysem presened in [19] is less han 5 meers in %90 of he cases and less han 3 meers in % 75 of imes.

4 TABLE I: measuremens MEASUREMENT NOTATION Accuracy RSS y i = h RSS( p p i ) + e i 4-12 db TOA y i = p pi + e i = hrange(p, pi ) + e i M Baromeric Pressure y i = P (p pi ) + e i = h bp(p p i ) + e i 1 M TDOA y i,j = p p i p p j + e i ei = h T DOA(p, p i, p j ) + e i ei 1-60 M AOA y i = h AOA(p, p i ) + e i 1-20 RF Fingerprining y = h MAP (p, p i ) + e (RSS MAP, 3 db) Posiion Esimaion y = p + e 4-20 M D. Time difference of arrival Taking ime differences of TOA as in (4) provides TDOA measuremens ha eliminaes he clock bias nuisance parameer. I is a pracical mobile measuremen relaed o relaive disance. The measuremens are repored o he nework, which performs necessary compuaions and i is no necessary o communicae he nework synchronizaion nor he reference poin locaions o he mobile. As for TOA, he synchronizaion accuracy deermines he performance, bu also he base saion locaions. The observed TDOA accuracy requiremen for locaion purposes in WCDMA is 0.5 chip [2] which means an error of abou 40 m (σ e 20m). Similarly, a TDOA accuracy requiremen of 0.5 chip in cdma2000 (advanced forward link rilaeraion - A-FLT) means 120 m (σ e 60m) due o he lower chip rae. On he oher hand, saellie navigaion sysems have a much higher chip rae, so for insance assised GPS can provide TDOA measuremens wih σ e 1m. In case of LTE, [17] presens an accuracy of 3.4 ns (corresponds o 1 m) for 20 MHz sysem a he confidence level %68. In a LTE sysem running a 1.4 MHz ha is apparenly a criical scenario for he LTE posiioning, when all available pilos (PRS, CRS, PSS and SSS) are used, an accuracy of 72 ns is achieved. E. Angle of arrival The use of direcional sensiive anennas provide angle of arrival (AOA) informaion as in. I is oday mainly available as a very crude secor informaion (e.g 120 o for a hree secor anenna as illusraed in Figure 2). Wih he use of group anennas his will be improved o abou 30 o beam widh (σ e 8 degrees), and perhaps even beer. Geomerically, he spaial resoluion of he inersecion of wo perfecly complemening AOA measuremens is limied o 2D sin(α/2), where α is he angular resoluion and D he disance beween he anennas. For α = 30 o, his means 36% of D. In LTE sysems, MIMO pre-coder index is fed back o provide a beer performance in AOA-based localizaion. Reference [20] presens he dispersion rms values for wo differen environmens. 1-5 of AOA rms in rural posiions, which ges worse in microcells wih rms values of 5-20 Reference [12] e.g. repors angular dispersion rms values of 1 5 degrees in rural environmens and 5 20 degrees in microcells. These figures are roughly of he same size as hose presened in example 1. Deailed performance comparisons Cell A Cell C Cell B Evaluaion Area Fig. 2: Use of Direcional Sensiive Anennas in AOA can be obained e.g. by using he Cramer-Rao lower bound of [4]. F. RF Fingerprining Fingerprining describes a class of algorihms ha removes deerminisic componens from measuremens bu using a digial map of hese creaed off-line. Such a digial map conains for insance RSS measuremens relaive he reference poins eiher prediced or provided via dedicaed measuremen scans in he service area. The former is conduced in he nework deploymen phase using graphical informaion sysems dedicaed for nework planning G. Posiion esimaes A direc posiion esimae may be available, from insance from GPS. Typical accuracy wihou differenial suppor is in he order of 5 10 m. According o [21] lowes bandwidh of LTE, i.e. 1 MHz resuls in 12 meers of posiion errors in 67% of he ime. This improves o 4 meers in case of 18 MHz signal bandwidh of LTE. IV. TRENDS Secion II described he area of posiioning in radio neworks as a more or less closed research area, where only incremenal conribuions have been seen laely. However, here are some imporan rends ha will change his picure.

5 A. Trend 1: Tigher Informaion Exchange One rend ha informaion and algorihms are shared beween differen layers in he classical OSI model of a communicaion sysem opens up for leaps in developmen. To moivae wih some low hanging fruis, consider he power delay profile (PDP) in Figure 3. PDP can be seen as he esimaed squared impulse response αik 2 for τ = 0. Tha is, he receiver sars o look for he ransmied symbol when i is known i is ransmied, and he correlaion a each ime is compued. The rue cτ is indicaed by he red line, bu how should his be esimaed from he PDP? The alernaives include hresholding wih a fixed or relaive hreshold, firs peak, sronges peak and so on. Furher, RSS is he inegral over a verical slice. From he figure, we can conclude he following: There are segmens in he es where here is non line of sigh (NLOS) condiion. There are many muli-pah componens. There is a srong spaial correlaion in ha nearby PDP s look quie similar. All hese effecs give rise o ambiguiies in he esimaed RSS and TOA. However, he PDP conains useful informaion abou he qualiy (variance of e i,ype ), he emporal correlaion beween e i,ype and e i,ype 1 and he modulariy correlaion beween e i,toa and e i,rss. If he PDP was available in he higher levels, more sophisicaed algorihms aking hese facs ino accoun could easily be derived. The boom line of his moivaion is ha he whole chain of informaion should be considered, no being resriced of he classical OSI layers. Timing Measuremens: IEEE v implemens iming measuremens as an opional managemen for STAions(STA). Those STAs who do no suppor his procedure, shall ignore a received iming measuremen frame. Reference [22] presens he work flow of various Wireless Nework Managemen(WNM) procedures of IEEE v sandard including iming measuremens ha is also presened in Figure 4. Iniiaion or sopping an ongoing procedure akes place by a Reques frame sen by STA. The value of he rigger field dicaes if i s an iniiaive frame or a sop one. IEEE Sd 802.1AS defines a proocol for clock synchronizaions beween STAs. Timing Measuremen acion frames hen will be sen by a sending STA. Wih he firs acion frame, boh sides capure imesamps. Transmission ime of he acion frame (1) and arrival ime of ACK response (4) is sored by he sending side. Meanwhile, receiving STA capures acion frame arrival ime (2) and he ACK response sen ime (3). The sending STA hen ransfers is capured imesamps (1 and 4) o he oher STA. Assuming he wireless channel o be symmeric, he offse of he clock a he receiving STA relaive o he sending STA is given by he relaive iming offse [(21)(43)]/2. Fine Timing Measuremens: This feaure is supposed o be officially published by lae However, he revision [23] is publicly available. More over, The proposal [24] ha discusses new feaures for an enhanced indoor posiioning, consider various candidaes in his regard. In his proposal, i is menioned ha UEs should repor FTM also o he radio nework in order o enable accurae range esimaions. Alhough i is nearly he same as is predecessor iming measuremen, some enhancemens are expeced. Timesamp resoluion incremen from 10 ns o 100 ps is one example. Fig. 3: Example of power delay profile (PDP) from an LTE deploymen as a funcion of ime. Red line indicaes he rue disance p p i according o a posiion reference sysem. B. Trend 2: New and Beer Informaion 1) New iming measuremens: Two algorihms are inroduced below. Fig. 4: IEEE sd v Timing Measuremen Implemenaion 2) Massive MIMO: Classic array processing wih MIMO (muliple inpu muliple oupu) anennas as surveyed in [10] is focused on accurae DOA esimaion. Though he heory

6 is raher maure and ha all cellular sandards since GSM are prepared for MIMO, i is so far no a big commercially success. One reason may be ha he capaciy does no scale well wih he addiional cos. On he oher hand, massive MIMO, where he number of anenna elemens is an order of magniude larger han he number of communicaion links hey serve, scales very favorably. This and may oher advanages are described in [25]. However, DOA is no compued explicily in conras o classic MIMO. This is an area for fuure research wih large poenial for super-resoluion DOA esimaion. 3) Ad-hoc neworks: Localizaion services ha are applicable on hese neworks mus mee differen demands such as low power consumpion, availabiliy, and reliabiliy. Tha is why some exising services such as GPS canno be employed on wireless ad-hoc neworks. To address his issue, one alernaive is o use shor-range single-hop localizaion sysem. However, here are cases in which reference nodes are no in he range of unknown ones. Then, muli-hop echniques mus be aken ino accoun. In hese scenarios, beacon posiions is broadcased over muliple hops. This allows esimaion of he disance o beacon nodes by calculaing hop sizes and number of hops while no direc communicaion is required where nodirec-communicaion is an inheri feaure of ad-hoc neworks. C. Trend 3: New Infrasrucure The infrasrucure illusraed in Figure 5 conains differen eniies ha each of hem can affec he measuremen resoluion drasically. All he devices a he lowes layer are conneced o heir upper layer devices via a shor-range echnology such as Blueooh, ZigBee, ec. A he mean ime, devices in he middle layer could vary from a simple User Equipmen(UE) acing as a gaeway o a Machine Type Communicaion (MTC) device [26]. Differen ypes of access of he middle devices could be an IP-conneciviy o anoher gaeway, cellular access o he AP or even an inra connecion o anoher device of he same layer via a shor range echnology. his is furher elaboraed in [27] 1) BLE beacons: Blueooh Low Energy (BLE) beacons can be low cos iny compuers equipped wih Blueooh radios. More complex hand-held devices such as smar phones can also provide he same funcionaliy. The generic idea is ha hese devices emi shor-range signals ha can be decoded by anoher BLE-enabled device. The disance o he receiving beacon can hen be esimaed. The possibiliy of idenificaion of muliple beacons simulaneously in parallel wih relaive disance calculaions of each beacon, locaion awareness of he device becomes possible. A rend in he BLE beacon indusry is o only use proximiy for he LBS, ha is, a coarsely quanized posiion. 2) IoT: Inerne of Things (IoT) can be seen as a grea poenial in many lines of research and developmen. However, massive signaling raffic produced by numerous objecs ha updae heir locaions, arises new challenges ha need o be addressed. Thus, here is a need for appropriae soluions ha provide accurae locaion informaion while keep he signaling level low. Fig. 5: 5g 3) M2M: Machine o Machine (M2M) neworks conain a number of devices such as RFID, sensors, ags, ec. This ype of nework is employed in differen locaion-based applicaions ranging from healh monioring o balefield surveillance. M2M communicaion neworks are self-configurable wih he feaure of being accessed remoely. The efficiency of approaches for locaion esimaion of M2M nework devices can be defined by scalabiliy, wheher or no is dependen o GPS sysems, range-based or range free propery, and error handling capabiliies. 4) 5G and Fuure Radio Nework Sandards: Firs of all, 5G should no be seen as an evoluion of 4G. The communicaion capaciy will probably increase, bu more imporan is he indusrial applicaion and for embedded sysems as IoT and M2M. Posiioning is one of he mos imporan design specificaions. V. CONCLUSIONS Posiioning in devices and gadges is currenly in ransformaion from nice o have o a mus. Firs, we have safey legislaions giving ough specificaions on he posiion informaion in emergency calls. Then, we have he rapid developmen of locaion based services (LBS) which requires posiion in siuaions where saellie navigaion sysems do no work (indoors, underground, ec). Furher, a rapidly increasing number of devices conneced o he cellular nework are no operaed by humans. We have he rends of Inerne of Things, machine o machine communicaion, auonomous vehicles and sysems, ec, where communicaion and posiioning will be he key enabler for fuure funcions and services. The purpose of his survey was o describe he over-all picure of how sae of he ar is organized oday (see Figure 1), recen advances in how he fundamenal measuremens are compued in recen sandards, and poining ou new rends. The inenion is o provide he fusion communiy wih background informaion o make relevan simulaions

7 and performance analysis, as well as ideas for new research direcions. VI. ACKNOWLEDGEMENT This work is funded by he European Union FP7 Marie Curie raining program on Tracking in Complex Sensor Sysems (TRAX). REFERENCES [1] D. N. Hafield, A repor on echnical and operaional issues impacing he provision of wireless enhanced 911 services, Federal Communicaions Commission, Technical Repor, [2] Y. Zhao, Sandardizaion of mobile phone posiioning for 3G sysems, IEEE Communicaions Magazine, vol. 40, no. 7, July [3] Wireless e911 locaion accuracy requiremens, fourh repor and order, Federal Communicaions Commission, Technical Repor 15-9, [4] J. J. Caffery and G. L. Suber, Overview of radiolocaion in CDMA cellular sysems, IEEE Communicaions Magazine, vol. 36, no. 4, April [5] C. Drane, M. Macnaughan, and C. Sco, Posiioning GSM elephones, IEEE Communicaions Magazine, vol. 36, no. 4, April [6] G. Sun, J. Chen, W. Guo, and K. J. Ray Liu, Signal processing echniques in nework-aided posiioning: A survey, IEEE Signal Processing Magazine, vol. 22, no. 7, July [7] A. H. Sayed, A. Tarigha, and N. Khajehnouri, Nework-based wireless locaion, IEEE Signal Processing Magazine, vol. 22, no. 7, July [8] F. Gusafsson and F. Gunnarsson, Possibiliies and fundamenal limiaions of posiioning using wireless communicaions neworks, IEEE Signal Processing Magazine, vol. 22, no. 7, July [9] H. Ryden, S. Modarres Razavi, F. Gunnarsson, S. M. Kim, M. Wang, Y. Blankenship, G. A., and A. Busin, Baseline performance of LTE posiioning in 3GPP 3D MIMO indoor user scenarios, in Proc. Inernaional Conference on Localizaion and GNSS, June [10] Krim and Viberg, Two decades of array signal processing research: he parameric approach, IEEE Signal Processing Magazine, vol. 13, no. 4, July [11] F. Gunnarsson, F. Lindsen, and N. Carlsson, Paricle filering for nework-based posiioning erresrial radio neworks, in Daa Fusion Targe Tracking 2014: Algorihms and Applicaions (DF TT 2014), IET Conference on, April 2014, pp [12] Y. Zhao, Overview of 2G LCS echnologies and sandards, in Proc. 3GPP TSG SA2 LCS Workshop, London, UK, January [13] J. J. Caffery, Wireless Locaion in CDMA Cellular Radio Sysems. Kluwer, [14] J. Del Peral-Rosado, J. Lopez-Salcedo, G. Seco-Granados, and F. Zanier, Achievable localizaion accuracy of he posiioning reference signal of 3gpp le, in Proc. Inernaional Conference on Localizaion and GNSS, June [15] K. Keunecke and G. Scholl, Deriving 2d oa/doa ieee g/n/ac locaion accuracy from an experimenally verified fading channel model, Ocober [16] K. J. Krizman, T. Biedka, and T. Rappapor, Wireless posiion locaion: Fundamenals, implemenaion sraegies, and sources of error, in Proc. IEEE Vehicular Technology Conference, June [17] W. Xu, M. Huang, C. Zhu, and A. Dammann., Maximum likelihood oa and odoa esimaion wih firs arriving pah deecion for 3gpp le sysem, Transacions on Emerging Telecommunicaions Technologies, November [18] K. Muralidharan, A. J. Khan, A. Misra, R. K. Balan, and S. Agarwal, Baromeric phone sensors: More hype han hope! ser. HoMobile 14. New York, NY, USA: ACM, 2014, pp. 12:1 12:6. [19] G. Liu, M. Iwai, Y. Tobe, D. Maekenya, K. M. A. Hossain, M. Io, and K. Sezaki, Beyond horizonal locaion conex: Measuring elevaion using smarphone s baromeer, in Proceedings of he 2014 ACM Inernaional Join Conference on Pervasive and Ubiquious Compuing: Adjunc Publicaion, ser. UbiComp 14 Adjunc. New York, NY, USA: ACM, 2014, pp [20] A. Kangas and T. Wigren, Angle of arrival localizaion in LTE using MIMO pre-coder index feedback, IEEE Communicaions Leers, vol. 17, no. 8, Augus [21] J. Del Peral-Rosado, J. Lopez-Salcedo, G. Seco-Granados, F. Zanier, and M. Crisci, Analysis of posiioning capabiliies of 3gpp le, Sepember [22] IEEE Sandard for Informaion echnology Telecommunicaions and informaion exchange beween sysems Local and meropolian area neworks Specific requiremens Par 11: Wireless LAN Medium Access Conrol (MAC) and Physical Layer (PHY) Specificaions, Sd., Rev. Revision of IEEE Sd , March [23] Ieee p revmc/d3.3, Tech. Rep., Ocober [24] Poenial enhancemens for indoor posiioning, Ericsson, Tech. Rep. R , February [25] F. Rusek, D. Persson, B. K. Lau, E. Larsson, T. Marzea, O. Edfors, and F. Tufvesson, Scaling up mimo: Opporuniies and challenges wih very large arrays, IEEE Signal Processing Magazine, vol. 30, no. 1, pp , January [26] D. Asely, E. Dahlman, G. Fodor, S. Parkvall, and J. Sachs, Le release 12 and beyond, Communicaions Magazine, IEEE, vol. 51, no. 7, pp , July [27] J. Sachs, E. P. a. M. J. Beijnar, N. and, F. Miliano, and P. Salmela, Capillary neworks a smar way o ge hings conneced, Ericsson Review, vol. 91, Sep 2014.