Localization Technologies for Indoor Human Tracking

Localization Technologies for Indoor Human Tracking Da Zhang , Feng Xia, Zhuo Yang , Lin Yao School of Softw are Dalia n Uni versit y of T ec hnolo gy Dalian 116620, Ch ina f.xia@ieee.org Wenhong Z hao Colle ge of Me cha nica l En gine eri ng Zhejiang University of Technology Hangzh ou 310032, Ch ina zwh2010@ sina.com Abstract —The proliferation of wireless lo caliza tion technologie s prov ides a promis ing fu ture for ser ving hum an b eings in indoor scenarios. Their applicatio ns include real-ti me tracking, activity recognit ion, healt h care, navi gation, e mergence detecti on, and target-of-interest monitoring, among others. Additionally, indo or localizati on technolog ies address the i nefficiency of GPS (Global Posi tioni ng S yst em) insi de buil ding s. S ince p eopl e spe nd mos t of their time in in door enviro nments, ind oor tracking se rvice is in great pub lic demand . Based on th is observat ion, this paper aims to provide a better understa nding of state-of-t he-art techn ologies and stimulate new research efforts in t his field. For these purposes, ex isting locali zation technologi es that can be used for tracki ng in divi duals in i ndoor enviro nments are review ed, along with some further discussio ns. Key words — indo or loca lizati on; hu man track ing; si gnal measurement; p ositioning al gor i thms; wireless networking I. I NTRODUCT I ON In the p ast decades, wirele ss localizatio n technolo gies have under gone co nsiderabl e progre ss. They gr adual ly play an important role in all as pects of people' s daily lives [ 1], includ ing e.g. living assi stant, naviga tion, emer gency de tection, surveill ance /track ing of ta rget- of-in terest , an d man y ot her locati on-b ased s erv ices. Reli able, accu rate , and real -ti me indoor tracki ng servic es are req uired by people even more strongl y th an ever. For ex ample, w ith th e severely incr easin g number of elder p eople, t he aging popula tion has beco me a burni ng issue for all moder n societies around t he world. It has conseq uently beco me an urgent problem how to monit or those old people eff ectively w hen th ey are at hom e or in si de ot her buildi ngs [2]. I n additio n, mo re and more att ention ha s been paid to con text-aw are a ppli cati ons w hich can mak e our lif e easier and conv enien t [3]. The realiza tion of these a pplic ation s is essentia lly bas ed on l ocat ion inf orm ation. For outdoo r environments, GPS (Globa l Positioni ng System) play s a dom inant r ole in l ocalizat ion [4] . How ever, it does not w ork w ell in in door scen arios. This in effici ency is due to the w eakness of sign als emitte d by GPS an d thei r disabil ity to pene trate most b uilding materi als. There fore, GPS doe s not fit well in indoor e nvironments where people spend most of their t ime. Even tho ugh GPS devices are beco ming more and more pr omising and po nderab le in t he future and are able to provi de su ff icient precis ion f or out do or use, oth er ef fecti ve technol ogies are demanded for indoor human/objec t tracking. To fulfil l this requirement , vario us indoor local ization technol ogies have bee n developed in the lit erature [5]. However , due to the co mplexity of i ndoor environments , the de velopment o f an indoor local ization te chnique is al ways accom panie d w i th a se t of chal leng es, e.g. NLO S (none l ine of sight), m ultipa th ef fect , an d noise inte rference . These challen ges resu lt m ainly from the influ ence of o bstacl es (e.g . walls, e quipments, and human b eings) on the prop agatio n of electrom agnetic w aves. F or i nstance, th e m obility of people incurs ch anges in phys ical conditi ons of the envir onm ent, which might si gnificant ly affect t he behavior of wireless radio prop agation. Al though the se negati ve effect s can not be elimin ated com pletely , in rece nt years rese arches are cons tantly going o n to improve t he performance of indoor (human/objec t) trackin g. Th ere are seve ral s urv ey pap ers in the lit eratu re of indoor lo caliza tion, e.g. [1,3, 6]. To inspire new re search effort s in the field, there is still a need of better understanding of state- of-t he-ar t lo caliza tion t echn ologi es. This pape r is an attempt to se rve f or this pu rpose. We pr esent a b rief overview of existi ng localizat ion techniqu es and methods, i ncluding signal m easurem ent m ethods, positionin g alg orith m s, networking t echniq ues and systems, which c an be used for indoor human tracking. The rest of this p aper is org anize d as fol low s. Sec tion II will first i llust rate basi c concepts in in door locali zation . The whole lo caliz ation p rocess i s divi ded int o tw o phases , i.e. signal m easu remen t and p osition cal culati on. St ate- of-th e-art localiza tion m ethods and alg orith ms u sed in these tw o phas es are rev iew ed in Se ction s I II an d IV res pect ively . In Secti on V, severa l popular ne twork techniq ues used in the filed are discussed. Some well-kno wn existing loc alization systems are also com pared . Sect ion V I con cludes th e pap er w ith a discussion on open issues. II. P ROBL EM S TA TEMENT Indoor localizatio n system, as Dempse y [7] defined , is a system th at can d eterm ine the position of som ething or someon e in a ph ys ical sp ace such as in a hosp ital, a gym nasium , a sch ool, etc. con tinu ously and in rea l tim e. Based on this con cept , cons ider a ty pical s cena rio of in door h um an trac king. First, eac h reference se nsor node (with known This w o rk is par tial ly suppo rted by Nat ural S cien ce F ound atio n of Chin a under Gran t No. 60903153 and Zhejian g Provin cial Nat ural Sci ence Fo undatio n of Chi na unde r G rant No . Y10 868 5. position) sen ds a ran ging re quest to th e com patible m obile device attach ed t o th e targ et (i.e . peo ple t o be loca ted). Th is device cou ld be for exam pl e a cell phon e or a PDA. Then the mobile d evice perceives the re quest sig nals an d issues a rangi ng reply to t he refer ence se nsor. To thi s end, the sensor could calcu late the tran sm ission tim e be tw een the sen so r a nd the mobi le device. Ne xt, the sensor forward s the calculate d time t o a calc ulat ion center . Usua lly the calcul atio n center coul d be a bas e stati on (BS ) or a pers onal com pute r (PC). Wi th powerfu l computat ional ca pability , the cal culati on cen ter proce sses the received data using so me positioni ng algorithm to obtain th e pos ition of th e ta rget. From th e a bove desc ript ion, w e can s ee th at in ord er to obta in the physica l position of the target-o f-inter est in indoor enviro nments, t wo steps are usuall y needed [3,8 ]: first, some posi tion -r elat e d sig nal par amet ers co rres pon ding to w irele ss comm unications b etween the t arget and the sen sor ar e measu red; and then , the ph ysical position o f the target is calculat ed based on these s ignal param eters. Signal Measurement Coordinates of Reference Nodes First Phase Second Phase Signals Position Calculation Position of the Target Signal Parameters Fig ur e 1 . Two phases in localiz ation As s hown in Fig. 1, the w hole l o cali zati on pr ocess can generally be divided into two phases : signal measur emen t and position cal culation. In th e first phase , som e signals are transm itted betw een the t arget node (representing the commu nication entity attache d to the people ) and a num b er of refer ence (sensor) nod es. During this pro cess, some proper ties of these signal s, such as arrival time, signal stre ngth, and dire ction, are captu red by the receiv ers. As such , certain sig nal param eters, such as TOA (Time of Arriv al), TDOA (Tim e differ ence of Arriv al), RSS (R eceived Signal S trength), a nd AOA (A ngle of A rrival ) , w ill be extr acted. Va rious m ethods used in th is phase w ill be cove red in th e next secti on. In the secon d phase, the phy sical position of th e target node will be dete rmined based on the signal paramete rs obtained in the firs t phase. The m ost com mon techn ique us ed he re is bas ed on rangi ng, whereb y distance or angle appro ximations are obtained [3]. In this contex t, geometric a pproaches will be emplo yed to cal culate the posi tion o f the target node as the intersect ion of position lin es obtaine d from the positi o n-rela ted para meters at referenc e nodes. Tril ateratio n and tria ngulation are two most popular geo metric app roache s, which will be introdu ced in S ection IV. In add ition, since s ignal meas uremen ts in real sy stems are on ly accu rate t o some exte nt (espec ially in indoor enviro nments), op timizatio n-based statist ical techni ques are of ten used t o fil ter measurem ent noise and im prove ac curacy of the res ult. III. S IGN AL M EA SURE MENT In this section , we ela borate on various measurem ent method s involved in the first p hase of loc alization (s ee Fig. 1). More spe cifically, our focus is on the three m ost popular catego ries of meth o ds for this ph ase: one categ ory is time based meth ods; anoth er is th e angle b ased m ethod (i.e . AOA); the third is th e received signa l strength based method ( i.e. RSS). In the fo llowing, we de scribe releva nt techno logies belongi ng to each of these c ategories , with s o me re lated w ork revi ew ed. A. Time-bas ed Methods 1) Time-of-Arrival (TOA) With TOA, the distan ce between the transm itting node and the receivin g node is deduce d from the transm ission time delay and the co rresponding spe ed of signal as follo ws: R time speed =× ( 1 ) where speed denotes the traveling speed of the sig nal, time the amou nt of time spen t by the signal tr avelling f rom the transm itting t o the receiving node, an d R the dist ance between the tr ansm itting node and the receiving no d e. Sin ce speed can be regard ed as a known consta nt, R can be com puted by ob serving time . One of the most w idespread techniques u ses TOA jointly with UWB tec hnology to achieve hi gher precisio n [9]. An overvi ew of this co mbinat ion has been given i n [10]. It has been r ecognized th at TOA tech nique can best deal w ith fine time reso lution with the help of UWB technology [11]. While TOA tec hnique has a restr ict requirement for synchroniz ation, this ineff iciency can be c o mpen sated by UWB due to its nat ure of sensitivity to time. UWB technology uses sh ort pulse duration to filter out the s ignals cau sed by reflect io n t o improve the ove r all p erform ance. As a cons equence, th e combination of these tw o te chnologie s is predominant i n the indoor locali zation fiel d at pr esent. 2) Time Diffe rence-of-Arrival (TDOA) This technology uses two different kinds of transmitting signals. T he time diffe rence bet ween these two kind s of signals is used t o reconstru ct the tr ansmittin g node's position . Th e calculat ion is based on the follow ing eq uation: 12 12 RR tt cc −= − ( 2 ) In (2 ), c 1 denot es the speed o f one kind of signal, c 2 the speed of anothe r kind of signal, t 1 and t 2 the tim e for these two signals travell ing from one node to the ot her resp ectivel y, and R still the distanc e betw een the tr ansm itting node and the rece iving node. A con side rable number of w orks hav e expl ored TDOA-bas ed m ethods. F or instan ce, Taka bayas hi et a l . [12] emplo y TDOA te chniq ue to realize t arget tr ack ing. This technology is based on EKF (Extended Kalman Filter), FDOA (Freq uency D iffer ence o f Arriva l) and TDOA technolo gies. Unlike co nventiona l methods which req uire enough sensor s to estimate th e positi on of target , the authors only use utiliz able sensors to calc ulate the po sition. T herefore, the a pproac h is pret ty suitable in e nvironments where the numb er of sensor s is not suff icient. 3) Round Trip Tim e (RTT) This measu rement m ethod emerges with the goal of solving the problem of synchroniza tion incurre d by TOA. With RTT, the dis tance is calcul ated as f ollow s: () 2 RT tt s p e e d R −Δ × = ( 3 ) where t RT denote s the amount of time need ed for a signal to travel f rom one n ode to the oth er and ba ck again, t Δ the pred etermine d time delay requi red by the ha rdware devi ce to oper ate at the rece ivi ng node, and spe ed the speed of the transm itting signal. It is cl ear th at RTT is a recipr ocal technol ogy [13,14] . Instead of using two lo cal cloc ks in both nodes t o calculat e the delay (a s TOA techn o logy does), i t us es only one n ode to record the transmittin g and arrival ti me. There fore, to some ext ent, this tec hnology sol ves the problem of synchro nization. Although t ime-based measurement metho ds are now in widesp read us e, they are lim ited by r estrict requi rement of synch r oni zation [15]. Th at m eans it is ne cessary to s et synchroni zed cl ocks to b oth the transmitting nodes and the rece iving node s. A consequenc e is that it would be ver y costly in orde r to install the system and maintain the accu racy at runti me. The RTT based techniqu es are ab le to solve the proble m of synchro nizatio n to a certa in extent. Ho wever, t hey increase the complex ity of the whole ( r efer ence) sens or system to O ( n 2 ). In a sensor system consi sting of n no d es, it takes every node n times to locate it s position through message exchang ing. Additi onally, w ith the R TT tech nology, o ther uncer tain fact ors (e.g. noise) coe xist duri ng the proce ss of time meas urement. T heref ore, th e pro blem of sy nchroni zati on deser ves further inve stigatio n. B. Angle-of-Arri val (AOA) Wit h respect to AOA-ba sed techniq ues [5,6,8,1 6], the reference nodes or the targ et node has the capability o f measuri ng the angle o f arrival based o n informatio n obtained. For thi s purpose, techni ques like a ngle diver sity may be utilized in orde r to exploit t he directi onality of the recei ver. Usual ly, dire ction fi nding can be acco mplishe d by either wit h direc tional a ntennas or with an arra y of ante nnas. The mai n princi ple behind the AOA measurem ent via antenna arrays consists in that diffe rences in arriv al times of an in coming signal at differe nt antenna e lements inc lude t he angle inform ation given that the array g e ometry is know n. With AOA, no ti me synchronizatio n between node s is required. AOA-ba sed techniq ues have been widel y used in the literatu re. For example, Yan g et al. [1 7] com b ine the AOA technol ogy with a nother T OA-based technique , i.e. TD OA, to achi eve higher accuracy. I n [18], ba sed on the basi c AOA concep t, the aut hors develo p a local izatio n techniqu e using cooper ative AOA approach. I nstead of requiring se ts of acous tic m odel array s and antenna array s in ea ch n ode like other convent ional AOA ba sed techniq ues, the appro ach onl y need s one set of acousti c model array and ant enna arra y in each node by int roduci ng the concep t of super nod es which are act ually virt ual AOA-cap able nodes. Unsurpri singly, AOA-base d technique s have their limitations . Since AOA -based methods are highly sensitive to multi-path and NLOS, it is n ot suitable for indoor localizati o n sometim es. As th e distan ce in creases, th e localizati on prec ision will dec rease. In a ddition, te chn ologies based on A OA requ ire addition al antenn as with the ca pacity to m easure the angles . This in creases the cost of the w hole sy stem. C. Received Signal Str ength (RSS) For t he RSS b ased tec hniq ues, the di sta nce is mea sured based on the at tenuati on intr oduce d by th e propagati on of t he signal from the transm itting node to the rece iving node. A n empiric al math ematical model t o calcu late the d istance accordi ng to signal propa gation is as follows [19 ,20]: 0 0 () () ( ) 1 0 l o g ( ) ( ) nW WAF nW C R pR pR n CW A F n W C R ×< ⎧ =− − ⎨ ×≥ ⎩ (4) The atten uation form ula can be expressed in (4) , w here R denotes th e distance between the t ransmitter an d the receiv er, R 0 a referen ce distanc e, p ( R ) and p ( R 0 ) the signal str e ng th receive d at R and R 0 respec tively, nW the num ber of o bstacles between the tr ansmitter an d the rece iver, WAF the attenu ation factor of the w all, C th e maxi m um n u mb er o f o b sta cl e s between the transm itter and the receiver, and n th e r outin g atte nuation fac tor which could be determined by bo th theore tical and em pirical c alculations . Based on the RSS techno logy, severa l methods have bee n proposed to estim ate the pos ition of th e target- o f-inter est. For exam ple, the fingerprint- based soluti o n [ 21] for ta rget positioning is the m ost typical a pplication of RSS tech nology. In g eneral , we can divi de th e fing erprint meth odol ogy into tw o steps: s ampling (off line) and m atching (on line). In the sampling step, a data b ase is creat ed offline to store the ra dio sign al map c onsistin g of th e geogr aphica l pos itions and th e corres ponding s igna l strengt hs. These s ignals m ay be e.g . sound, l ight, color , and human move ment, among o thers. In the matchin g step, the r e levant signals collected f o r the tar get (nod e) are compa red agai nst the pre -stored re cords of the geographic- signal map. By doing so , it will be able to dete rmin e where the target is, as long as any r ecord in the data base is m atched. IV. P OS ITION C ALCULATION Based on the sig nal pa rameter s m easured in the firs t phas e and t he k now n coordin ates of refe ren ce n odes, it is then poss ible t o calcul ate th e phy sical pos ition (i.e. co ordin ates) of the targ et in the s econ d phas e of local ization (F ig. 1). To do this, the trilaterati o n an d triangulati on techniques are comm only us ed. In addition , statis tical techni ques coul d be employ ed to impr ove the so lution accuracy by coping w ith measu rement n o ise. In this r egard, w e will intro duce a very popular pa rametric appr oach: m axi mum likelihood estim ation, though there are many other appr oaches in th e literatu re. A. Trilateration As illust rated in Fig . 2, the t rilateration based posi tioning algorithm uses three fixed non-collinear refere nce nodes to calculat e the phy sical position of a target n ode (in 2D ). R 1 R 2 R 3 A B C T Fig ur e 2 . Tr ilate ration- based pos itio ning Based o n the c oordina tes of thr ee re ference nodes: A ( x 1 , y 1 ), B ( x 2 , y 2 ), an d C ( x 3 , y 3 ), an d the c orrespon ding distanc es from each r eference n ode to th e target n ode: R 1 , R 2 , and R 3 , w e can obtain th e follow ing equati ons: 22 2 11 1 22 2 22 2 22 2 33 3 () ( ) () ( ) () ( ) xx y y R xx y y R xx y y R ⎧ −+ − = ⎪ −+ − = ⎨ ⎪ −+− = ⎩ ( 5 ) where ( x , y ) denote s the (unknown) coor dinate s of the target T . Based o n the tri lateratio n algorit hm, Han et a l. [22 ] further impro ve loca lizati on per formance b y taking into account the layout of the thre e refere nce node s. The work has appr oved tha t the tri lateration algorithm can best d emon strate its advanta ges when the three ref erence no des are de ployed in th e vertices of equi lateral triangle s. Yang and Liu [23] co nsider t he effect of noisy envi ronments, and use differe nt confiden ce coeffic ients for thr ee nodes t o guaran tee the quality of tril aterati on. B. Triangul ation When AO A measurement s are availa ble, tria ngulat ion can be us ed to det ermin e the posit ion of t he targ et node. Instea d of meas uring dis tances b etween no d es as trilat eration does, triangul ation-b ased po sitioning i s ba sed on the measure ment of angles , though they work in a similar manner . In most situat ions, tria ngulation ca n be tra nsformed to tri lateratio n since th e distance be tween node s can be reconstr ucted from the bearin gs betw een them . How e ver, c ompared t o trilatera tion, only two re ferenc e nodes are needed for triangul atio n (in 2D), inst ead of thr ee. T 1 θ 2 θ A B Fig ur e 3 . Tr iangulat ion-base d positio ning Wit h triangula tion, t he posit ion of the target nod e can be determ ined by th e inters ection o f sev eral pai rs of an gle dire ction lines [ 6]. As s hown in Fig . 3 wh ere A a nd B represen t refer ence node s, after obt aining the a ngles 1 θ , and 2 θ , the phy sical positi on of T (re presenting th e target to be located) could the n be calculated ba sed on the prede termined coordi nate s of the r efer ence nodes. C. M aximum Likelihood Esti mation (M LE) MLE is a popular stat istical me thod used for addre ssing the problem of measu rement unce r tainty in local ization. In t his subsecti o n, w e w ill describe MLE in the contex t of tri lateration-ba sed position ing. Suppose the MLE metho d [20] uses n referenc e node s to calcula te the tar get node ’s coordi nates ( genera lly 3 n ≥ ). The re levant equatio ns ar e given belo w: 22 2 11 1 22 2 22 2 22 2 () ( ) () ( ) () ( ) nn n xx y y R xx y y R xx y y R ⎧ −+ − = ⎪ −+ − = ⎪ ⎨ ⎪ ⎪ −+ − = ⎩ # ( 6 ) In (6 ), using ev ery equ ation to subt ract the su bsequen t one, we w ill get: 22 2 2 2 2 11 1 1 1 22 2 2 2 2 11 1 1 1 2( ) 2( ) 2( ) 2( ) nn n n n nn nn n n n n n n xx x x x y y y y y R R xx xx x y y y y y R R −− − − − ⎧ − − − +−− − = − ⎪ ⎨ ⎪ −− − + − − − = − ⎩ # Let 11 11 2( ) 2( ) 2( ) 2( ) nn nn n n xx y y A xx y y −− −− ⎡ ⎤ ⎢ ⎥ = ⎢ ⎥ ⎢ ⎥ −− ⎣ ⎦ ## , 222 2 22 11 1 22 2 2 2 2 11 1 nn n nn n n n n xx yy R R b xx y y R R −− − ⎡ ⎤ −+− + − ⎢ ⎥ = ⎢ ⎥ ⎢ ⎥ −+ − + − ⎣ ⎦ # , and x X y ⎡ ⎤ = ⎢ ⎥ ⎣ ⎦ , then we have A Xb = . By adopt ing the mini mum varia nce estimati on method , the coordi nates ( x , y ) of the targe t can be cal culate d based on t he follo wing equatio n: 1 () TT XA A A b − = ( 7 ) Beside s estimating t he coordina tes of the target , the authors of [24] a nd [25] also use MLE to solve t he probl em of synchroni zation by predicting the uncertain param eters in ti me bias. Par ticular ly, Tian et al. [24] thoroughly anal yzed the source fa ctors causing time bias in diffe rent transmission sta ges. This an alysis cont ribut es to further research on resolvi ng diffe rent aspe cts of t he synchro nization pr oblem. V. N ETW ORKI NG T ECHNIQUES AND S YSTEMS In this s ection, we f irst outline th e signal tech nologies that are co mmonly used, di scussing thei r pro s and cons; then classify existing system s into several groups and make a comparis on among st them. It is worthy to note that n ot only t he meas urem ent method an d positi oning alg orithm but also th e signal tec hnology of a localizat ion system can have a heavy impact on the accuracy of loca lization. A. Infrared ( IR) Based Systems The m ost prominent advantage of IR is its w ide avail ability since m any devi ces ar e equippe d w ith IR sources , such as mobi le phon es, TV , printe r, PDA s, and s o fort h. In a dditi on, since the whole inf r astru cture is very simple, it does not need costly in stallation and m aintenance. H o w e ver, due to its requi rement of line-of-s ight an d its inabil ity to pen e trat e opaq ue obstacles, i t can not be applied to some kind s of indoor scenarios in which the environm ent is pretty c omplex. B esides, it is subject to interf erence of oth er s o urces of IR devic es. Several systems are based on this techno logy, includi ng Active Bad ge [26], Fire fly [27], and OPTOT RAK [28], for e xample. B. Radio Freque ncy (RF) Bas ed Systems Sy stems desig ned base d on R F can cov er la rger dis tance since i t uses elect romagnetic transmis sion, w hich is able to penet rate opaq ue objec ts such as people and walls. Beside s, a RF system can uniq uely id entify peop le or object s tracked in the system. In t he literature , triangula tion and fingerp rint techniques are widespread used in RF based system s. Based on this tec hnology, RFID (Rad io Frequency Ide ntification) , WLAN ( W ireles s Loca l A rea Netw ork), Bluetooth , w ireless sensor ne tworks, UWB (Ultra Wide Band) ar e creat ed. In addi tion, RF based technologies are furthe r divided i nto narro w band ba sed technolo gies (RFID , Bluetooth a nd WLAN) and wide ba nd based tec hnologies ( UWB). Amongst these technologies, UWB is the most accurate and fault-t olerant sys te m th at has a widespre ad usage in indoo r local ization. C. U ltras ound Based Systems Although the systems based on ul trasound tec hnology is relative ly cheap, th e pr ecision is lower in co mpa r ison w ith IR-bas ed system s due t o the re flect inf luenc e. Addit ionally, this kind of system s is alway s associated with RF te chnology t o fulfill th e synchroniz ation requ irement , wh ich may incre a se the cost of the whole system. Active B at [29] and Cricke t [30] are example applicat ions of ultrasound technolo gy. In Table I [1,5,6], w e make a comparison between some major localization systems in various aspects, including accuracy, adv a ntages and disadv antages, networking technologies and localization methods. VI. C ONCLUS ION In thi s paper, we pre sented a br ief overvie w of state- of-th e-art localization technologi es for t r ack ing individ uals in indoor environments. So me related works are review ed. Des p ite th e great progress made in recent y ears, there a re a num ber of o pen iss ues that need t o be ad dress ed. Exampl es inc lude e. g. conti nuously tr acking peop le tra velling bet ween indoor and outdoor e nvironments, solving synchro nization prob lems, red ucing the impac t of noise interfe rence, an d impr oving en ergy eff iciency. A lthough som e previo us technolo gies are concerned with t hese issues, t hey might su ffer from various l imitations, e . g. in crease in the cost of the w hole sy stem, pre cision defici ency, an d sev ere compu tationa l overhead . Innovative rese arch efforts are expected to tackle these issues in the n ear future . TABL E I . L OCALIZATION S YSTEMS System Network Accur a cy Method Overall Evaluat ion (A : Advanta ge; D: Disad vantage) Whe reNet [3 1] RFI D 2m to 3m TDOA A: Uniq uely identify equi pment and pe rson. D: N eed numer ous i n frastr ucture com p one nts RADAR [ 32] WLAN 2.26 m out of 312m 2 Triang ulation A: Re use the e xisting WL AN inf rastr ucture. D: L ow level accur acy, no co nsiderat ion of pr ivacy EK AHAU [ 33 ] WLA N 1m RS S I A: Low c os t a n d p owe r l ev el o f t h e b at te ry. D: L ow level accuracy an d only pr ovide 2D loca tion i nformatio n. COMPASS [3 4] WLAN 1 .65m out of 312m 2 Fingerp rin t A: Co nsid er th e orient ation i mpact of the u ser. D: O nly consi der sing le use r. Ubis ense [ 35] UWB Tens of cen timet ers TD OA and AOA A: No requir ement of l ine-o f-s ight; la rge cover age are a; 3D locatio n; high accur acy D: The p ric e of th e syst em is high. Activ e Badge [ 26] I nfrared Room level RSS A: A ddress pr ivacy D: Low accuracy; l ong transm issio n perio d; influe nced by fluo resce nt light and sunl ight Fir efly [ 27] Inf rare d 3.0mm Not avail able A: H igh level accuracy; sm all measur ement de lay of 3 ms D: Use wire to c onn ect tags and th e coverag e area i s lim ited to 7m. OPTO TRAK [ 2 8] Inf rar ed 0 .1m m to 0.5mm Not availabl e A: Hig h accuracy ; able to measure relative motions o n the dif ferent part s of one ob ject. D: L imited by line- of-sight r equireme nt. Sonit or [36] Ultras ound Room lev el Not avai lable A: Energ y effic ient D: L ow lev el accuracy IR IS_ LP S [3 7 ] In fr ar ed 16 cm ou t of 100m 2 Triang ulation A: L arger co vere d area D: Sub ject to int erferen ce from f lor esc ent li ght and sunli ght Active Bat [29] Ult rasou nd 3c m out of 1000m 2 Multi lat eration A: Cover large a rea; p rovid e 3-D posit ion. D: Sub ject to ref lecti ons of obst acles; us e a lar ge numb er of trans mitte rs on the ceil ing. Crick et [30] Ult rasou nd, R F 10cm TOA a nd triang ulation A: A ddress pr ivacy; low cost, de central ize d admi nistratio n. D: Mo re ene rgy cons umptio n R EFEREN CES [1] Yan ying Gu, Anth ony Lo, Ignas Ni emegeers, “ A Survey of Ind oor Positioning Systems for W ireless Personal Networks”, I EEE Commu nications Survey s & Tutor ials, vol . 11, no. 1, 2009, Pag es:13-3 2 [2] Jason W .P. Ng, “ Ubiqu itous Heal thcare L ocalisat ion Sche mes”, 7th Interna tiona l Workshop on Enterprise n etworkin g and Comp uting in Heal thcare I ndustr y, H EALTHCO M, June 20 05, Page (s):156 - 161 [3] Isaac A mundson an d Xenofo n D . Koutsouko s, “A S u rvey on Local ization for Mobile Wirele ss Sensor Ne tworks”, R. Fulle r and X.D. Koutsou kos (E ds.): ME LT 2009, LNCS 580 1, 2009 , Pages: 235 -254 [4] C. Frits che , A. Kl ein, “On t he Perf ormance o f Hybr id GPS/G SM Mobil e Terminal Trac k ing”, I EEE ICC Wo rkshops, June 200 9 Page(s ):1- 5 [5] Holge r Linde, “ On Aspe cts of I ndoor L ocalizatio n”, T hesis, Un iversity of Dortmu nd, Augu st 2006 [6] Hui L iu, H. Da rabi , P. Ba nerjee , Jing L iu, “Surv ey of W irele ss Indoor Positioni n g Te ch niques and Sy stems, ” IEEE Trans. Sy stems, Man, an d Cybe rnetics, P art C: Appl ications and Re view s, Vo l.37, No.6 , Octo ber 2007, pp. 1067-108 0. [7] M. De psey, “I ndoor Positioning Sy stem s in He althcare ”, Radians e Inc.Wh ite Pap er, 2003 . [8] S. Gezici, “A survey on wireless position estim ation”, Wirele ss Person al Commu nications , vol. 4 4 , no . 3, pp . 263- 282, F eb. 2008 [9] C. Falsi, D. Darda ri, L. Mu cchi, M. Z.Win, “Time of Arri val Est ima tion for UWB Localiz ers i n Reali stic E nviron men ts”, EURASIP Jou rn al on Appl ied Sign al Pr ocessing , 2006, A rticle ID 32 082, 13 pag es [10] I. Guvenc , C. -C. Chong, “A Survey on TO A Based Wirel ess Local ization and NL OS Mitigatio n Technique s”, I EEE Communications Surveys an d Tutori als, vol. 1 1, no. 3, 2009, Pages : 107-12 4. [11] P. Cheong, A. Rab bachin, J.-P. Montill et, K. Y u and I. O ppermann , “Synch roni zation, TOA and Position Esti mationfor Low-comp lexit y LDR U WB Device s”, I EEE I nternatio nal Confe rence on Ultr a-Wideban d (ICU ), Sept.200 5, Page s: 480-48 4 [12] Y. Takaba yashi , T. Matsuza ki, H. Kamed a, and M. Ito, “Ta rget Tra ck ing Usin g TDOA/FDOA Meas uremen ts i n the Di stribu ted Sensor Network ”, SI CE Annual Co nfere nce, Augus t 2008 , J apan. Pag es: 344 1-3 446 [13] S. Bar telma os , K. Abed -Mera im , R. Leyma n, “Gen eral se lect ion crit eria to mit igat e th e impact of NLoS erro rs in R TT measu remen ts f or mobile positioning. ”, IEEE I C C, June 2 007, Page s: 4674-4679. [14] A mgad Ze itoun Zhi heng Wang Sugi h Jam in, “RT Tome t er: Measuring Path Mi nimum RT T with Confide n ce” I EEE Workshop on I P Oper ations an d Manage ment (I POM), 200 3 , P age s:12 7-134 [15] L aurence Mail aender, “ Comparing Geol ocation Bounds fo r TOA, TDOA and . Round-Trip TOA”, IEEE PIMRC , 2007, Pages :1-5. [16] Dr agos¸ Nicule scu and Badri Na th, “A d Hoc Po sition ing S ystem (APS) Using A OA”, IEEE I NFOCO M, March 2003, Pag es: 1734-1743 [17] Chunhu a Yang, Yi Huang, Xu Zhu, “Hybrid TDOA/AOA meth od for indoor p ositi onin g systems” , IET Semi na r on Locati on Techn ologies, Dec.2 007, Pa ges:1-5 [18] Hui T ian, Shua ng Wang, H uaiyao Xie , “Localiz ation using Coope rative AOA Approach”, I EEE WiCOM'0 7, Se p t. 2007, P ages: 2416-19 [19] S.Y . Seidel and T.S. Rap port, “9 14 MHz path los s pre diction Mo del fo r Indoo r Wirele ss Communica tions in Multi- floored build i ngs”, I EEE Trans. on Anten nas & Propa gation, Feb.1992 , 40(2): 2 07-217 . [20] Jun Zh ao, Zero-C onfigu rati on Indoor Posit ionin g System B ased on RF Sign al St rength , Mas ter Th esis , Zhejia ng Un iversi ty, M ay 2007 . [21] Mart in A zizyan, I onut Cons tand ache, Rom it Roy Choud hury , “Sur roundSense : mobil e pho ne local ization via ambiance fingerp rint ing”, ACM MobiCom, Sept ember 20 09, Pages: 2 61-272 [22] G. Han, D. Choi, W. Lim, “A Novel R eferenc e Node S electi on Alg orithm Base d on Tr ilateratio n for I ndoor Se n sor N etw orks”, IEEE Intl Conf. on Compute r and Infor mation Tech nology , 2007, Pages:10 03-10 08 [23] Zhe ng Yang, Y unhao L iu, “Quality of T rilateratio n: Confide nce bas ed Iterativ e Localiz ation”, I EEE ICDCS, Ju n e 200 8 , Page s: 446-453. [24] Xianz hong T ian, Yong gang M iao, T ongsen H u, Bojie Fan, Jian Pa n, Wei X u, “Max imum Like lihoo d Estimat ion Base d on T ime Sy nchronizat ion Alg orithm for Wir eless Se nsor Ne t wo rks”, IS ECS Interna ti onal Colloq uiu m on Computin g, Communi cation , Control, and Manage ment ( CCCM) , August 2009, Page s: 416-42 0. [25] Pau Cl osas, Carle s Fernánde z-Prades, and J uan A. F ernán dez-Rubio , “Maximum L ikelihood Estimatio n of Positio n in GNSS”, I EEE Signal Pro cessing L etters, v ol. 14, no . 5, Ma y 2007, P ages:359- 362. [26] Roy Want, Andy Hopper , Veronica Falcao, J onathan Gib bons, “T he Activ e Badge L ocation Sy stem”, A CM Transact ions on Informatio n Sy stems (T OIS) , vo l.10, no. 1, Janu ary 199 2, Page s:92-102 . [27] Firefly Moti on Tracki ng System Us er’s guid e, 1999, http:/ /www. ges turecentral .com/fire fly/Fir efly UserGuide .pdf. [28] Northen Digita l Inc., Opt otrak , http:/ /www.nd igita l.com/ [29] The Ba t Ultra soni c Locati on Syst em, Ca mbrid ge Uni versi ty Com puter L aborato r y , http://w ww.cl .cam.ac.uk/ rese arch/dtg/attar chive/ba t/ [30] N. P riy antha, A. Chakr aborty, a nd H. Bal akrish n an, “ The cri cket locat ion -support system” , ACM M obiCom, 2000. [31] Ze bra Techno logy Company , http://w ww.w herenet.co m/ [32] P. Ba hl and V. Padm anabhan, “RADA R: An in-b uilding RF based us er locatio n and trac k ing sy stem”, I EEE I N FO COM, vol . 2, March 2 000, pp. 775- 784. [33] Ekaha u, http://w ww.e kahau.com / [34] T. K ing, S. K opf, T . Haense lmann, C . Lubbe rger and W. Effe lsberg , “COMPASS: A Prob abi listic Indoor Posit ionin g System Ba sed on 802.11 and Digital Compas ses”, ACM W i NTECH, Los A n geles, USA, Sept. 2006, Pag es: 34- 40. [35] Ubise nse, htt p://w ww.ubise nse.net /en [36] Sonito r, http: //ww w.sonitor .com/ [37] E. A itenbichl er, M. Mühlhäus er, “ An IR L ocal Pos itioning S ystem for Smart I tems a nd Devic es”, IEEE ICDCS Workshop s, May 2003, Page s:334-3 39.