Measurements of Ultra Wide Band In-vehicle channel - statistical description and TOA positioning feasibility study

Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 DOI 10.1186/s13638-015-0332-3 RESEARCH Open Access Measu emen s o ul a wide band in- ehicle channel - s a is ical desc ip ion and TOA posi ioning easibili y s udy Ji i Blumens ein1*, Ales P okes1, Tomas Mikulasek1, Roman Ma salek1, Thomas Zemen2 and Ch is oph Mecklenb äuke 1,3 Abs ac This pape epo s on a eal-wo ld wi eless channel measu emen campaign o in- ehicle scena ios in he UWB equency ange o 3 o 11 GHz. The e ec s o an enna placemen in he ehicle’s passenge compa men as well as he e ec s due o he p esence o passenge s a e s udied. The measu emen s ha e been ca ied ou in he equency domain, and he co esponding channel impulse esponses (CIRs) ha e been es ima ed by in e se Fou ie ans o m. The in luence o a speci ic band g oup selec ion wi hin he whole UWB ange is also gi en. S a is ical analysis o he measu ed channel ans e unc ions gi es a desc ip ion o he wi eless channel s a is ics in he o m o a gene alized ex eme alue p ocess. The co esponding pa ame e se s a e es ima ed and documen ed o all pe mu a ions o an enna placemen and occupancy pa e ns inside he ehicle’s passenge compa men . Fu he , we ha e ca ied ou a easibili y s udy o an in- ehicle UWB-based localiza ion sys em based on he TOA. The posi ioning pe o mance is e alua ed in e ms o a e age e o and s anda d de ia ion. Keywo ds: UWB; In- ehicle en i onmen ; Channel model; Posi ioning; TOA 1 In oduc ion The onboa d elec ical powe dis ibu ion, communica- ion, and ne wo king unc ionali ies a e ealized by cable bundles in oday’s ehicles. We obse e a end owa ds inc easing numbe s o senso s, ac ua o s, con ol uni s, and in o ainmen sys ems in ca s and ucks. As a di ec esul , he weigh o he wi ing in all ypes o ehicles inc eases. Mo eo e , hei lexible ins alla ion and eli- abili y ep esen a challenging and cos ly ask [1]. The weigh o he wi ing becomes e en mo e se ious when he ehicles a e powe ed ully elec ically. In [2,3], he au ho s conclude ha ul a wide band- wi h (UWB) echnology wi h i s a o able adio en i on- men cha ac e is ics o indoo a eas such as low ansmi powe and obus ness agains mul ipa h ading could be ex apola ed e en o he ehicula passenge compa - men . Na u ally, a emp s o eplace cable bundle s a *Co espondence: [email p o ec ed] 1Depa men o Radio Elec onics, B no Uni e si y o Technology, Technicka 12, 612 00, B no, Czech Republic Full lis o au ho in o ma ion is a ailable a he end o he a icle up wi h in- ehicle adio channel measu emen s we e pe - o med by au ho s in [4-9] and by channel modeling in [3,10], and a clus e ing app oach o in a-bus channel modeling is s udied in [11,12]. A emp s o build a p o- o ype o an UWB-based wi eless senso ne wo k wi hin a ehicle, bo h in he passenge and he engine com- pa men s, a e published in [13,14]. In [15], he opic o wi eless in- ehicle communica ion links based on LTE is discussed while eckoning wi h speci ic in- ehicle impulse noise. In [16], he UWB channel inside a ehicle is s ud- ied om a spa ial s a iona i y poin o iew. The necessi y o de ailed knowledge o he channel cha ac e is ics is o highes impo ance o he p ope physical laye design o any wi eless communica ion sys em. Toge he wi h his mo i a ion o subs i u e a leas pa o he ehicle’s cable bundles by wi eless links, a wi eless localiza ion se ice wi hin he ehicle is desi able. Fu u e applica ions o such a localiza ion se ice include emo e keyless en y and igni ion sys ems, ad anced child pas- senge sa e y, and beams ee ing o in- ehicle high-speed In e ne access. © 2015 Blumens ein e al.; licensee Sp inge . This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (h p://c ea i ecommons.o g/licenses/by/4.0), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly c edi ed. Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 2 o 10 In [14,17], a localiza ion se ice u ilizing UWB is s udied and i is concluded ha in gene al, hanks o he high ime esolu ion o UWB impulses, he ime o a i al (TOA) echnique is capable o p o iding su - icien spa ial esolu ion o a a ie y o applica ions. Al hough he TOA wo ks eliably in en i onmen s wi h LOS, i can be used wi h some es ic ion also in NLOS scena ios. In he mul ipa h en i onmen , he impo an componen o anging based on he TOA echnique is he di ec ay, which p opaga es om he ansmi - e owa ds he ecei e . When he beam pene a es some obs acles whose a enua ion does no a oid he beam de ec ion, he TOA echnique is applicable. No e ha , o example, in he US, he equency ange o 1.99 o 10.6 GHz is de egula ed o communica ions and wall-pene a ing ada s, enabling looking in o o h ough non-me allic ma e ials [14,18]. Thus, he p e- sump ion is ha e en he ha sh in- ehicle ambiance wi h OLOS p opaga ion may p o ide su icien posi ioning accu acy. In [4,6,7,19], he pa h-loss, sea ma e ial, and occupancy in luences a e p esen ed o he equency ange o 3 o 8 GHz. Since he posi ioning se ice deploymen is no seen as he aim o [4,6,7,19], he placemen o ans- mi an ennas is inapp op ia e om ha poin o iew. Thus, esul ing pa ame e s could di e om pa ame- e s ob ained by measu emen campaigns which ake he posi ioning in o accoun in he i s place. 1.1 Con ibu ion o he pape While aking in o accoun he in luence o he occupancy, an enna placemen , and he in luence o a speci ic e- quency band g oup selec ion, in his pape , we add ess he ollowing: •In a- ehicle channel measu emen and s a is ical e alua ion ia GEV. This allows a ep oducibili y o he measu ed esul s o 90 selec ed wi eless links wi hin a passenge ca compa men . •S a is ical analysis and he in- ehicle posi ioning in he UWB ange o 3 o 11 GHz. The aim o he a icle is o gi e a gene al o e iew o he achie able accu acy o anging ega dless o LOS and NLOS scena ios. The pape is o ganized as ollows. In Sec ion 2, we p o- ide an o e iew o ou measu emen si e including a ha dwa e desc ip ion. In Sec ion 3, we p esen ou chan- nel measu emen ools including ou conical monopole an enna design [20] and we de ine he sough channel pa ame e s. In Sec ion 4, he easibili y o he posi ion- ing se ice deploymen wi hin a ehicle compa men is assessed, while he conclusion in Sec ion 5 sums up he pape . 2 Measu emen se up 2.1 Measu emen bandwid h and dynamic ange Theschemeo hemeasu emen se upisshownin Figu e 1. The complex CIRs (below in oduced by (1)) co esponds o he s41,s42,ands43 sca e ing pa ame e s which a e measu ed in he equency domain o wo di - e en equency bands, 3 o 11 GHz (en i e UWB band wi h a bandwid h o B= 8 GHz ) and 3.3168 o 4.752 GHz ( i s band g oup wi h a bandwid h o B=1.58GHz),u i- lizing a ou -po ec o ne wo k analyze Agilen Tech- nologies E5071C (VNA; Agilen Technologies Inc., San a Cla a, CA, USA) (Figu e 2). The spa ial placemen o he ecei ing (RX) and ansmi ing (TX) an ennas inside he ehicle is depic ed in Figu e 3. The dynamic ange o he measu emen se up is highe han 90 dB (Pou VNA = 5 dBm, IF bandwid h = 100 Hz). The chosen equency s ep o 10 MHz esul s in 801 e- quency poin s in he case o he en i e UWB band and 159 equency poin s in he case o he i s band g oup. In o de o a oid a deg ada ion o he measu ed phase accu acy due o mo emen s o he RX an enna, phase- s able coaxial cables we e used and included in he cal- ib a ion p ocess. The measu emen is ca ied ou in he Skoda Oc a ia 1.8 TSI ca . 2.2 An enna placemen As depic ed in Figu e 3, he RX an enna is placed a a i- ous loca ions inside he ca compa men (on all sea s and in he boo ) and he TX an ennas a e placed on he le and igh sides o he dashboa d, op co ne s o he windshield, and a he ea pa o he ceiling. The channel measu emen s a e ca ied ou o bo h LOS and NLOS scena ios. NLOS is caused by he back es o he sea s, he dashboa d, and/o pe sons si ing inside he ehicle. Figu e 1 Measu emen se up con aining he ec o ne wo k analyze Agilen Technologies E5071C and he ca Skoda Oc a ia 1.8 TSI. Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 3 o 10 Figu e 2 Images o he conical monopole an enna, measu emen posi ion, and ou -po VNA. [le ] De ail o he conical monopole an enna moun ed on he on windshield. [middle] One measu emen posi ion inside he ehicle. [ igh ] Fou -po VNA connec ed wi h an ennas inside he measu ed ehicle. Since he adia ion pa e n o he conical monopole an enna [20] is e y close o he omnidi ec ional adia- ion pa e n, we we e able o cap u e a maximal numbe o mul ipa h componen s ( e lec ed wa es). In Figu e 4, he conical monopole has an omnidi ec- ional H-plane adia ion pa e n which is in a ian in he equency band o in e es . Due o a a iable gain in he lowe hal E-plane adia ion pa e n (ele a ion angle om 90° o −90°), he an ennas we e placed in he ca compa men so ha he uppe hal E-plane adia- ion pa e n (almos cons an ) was used. I means ha when he an enna was placed a he cabin ceiling, i was si ua ed bo om up. Thus, he LOS and NLOS a e min- imally a ec ed by he adia ion pa e n; howe e , wi h he e lec ed wa es a i ing om he TX an enna o inciden on he RX an enna, he lowe ele a ion angle migh be a ec ed by he non-ideal adia ion pa e n o he an ennas. 3 Channel pa ame e s The CIR desc ibes he wi eless channel. We u ilize an in e se disc e e Fou ie ans o m o he windowed sca - e ing pa ame e se ies, exp essed as: hα(n)= N−1  k=0 w(k)sα ζ(k)ejkn2π/N,(1) whe e sα ζ(k)co esponds o he k h measu ed sca e ing pa ame e (as desc ibed in Sec ion 2) and w(k) ep e- sen s he Blackman window. Pa ame e αdeno es he spa ial posi ions o he ansmi and he ecei e an enna in he measu ed ehicle and ζ∈{41, 42, 43}. Fo p ac i- cali y in he ollowing s a is ical p ocessing, we a bi a ily me ge indices αand ζin o one measu emen numbe α∈{1, ...,90}.Hence,in he ollowing,i isno possible o assign he speci ic measu ed da a o he ac ual spa ial posi ions. Figu e 3 The posi ions o ansmi ing ( ed) and ecei ing (blue) an ennas. We employ wo possible ecei e an enna placemen pa e ns. As seen on he le pa , he an ennas 1a and 2a occupy he le and igh op co ne s o he windshield, while on he igh pa , he an ennas 1b and 2b a e posi ioned in he lowe co ne s. Please no e ha all measu emen s ha e been measu ed o a ious passenge layou s. We ha e conside ed (1) emp y ehicle and (2) d i e and wo o h ee passenge s. Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 4 o 10 Figu e 4 Measu ed gain pa e n o he conical monopole an enna [20]. The numbe o measu ed equency poin s N=801 o heen i eUWBo N=159 o he i s band g oup. Since he in- ehicle channel is assumed o be ime in a ian , we pe o med one epe i ion o he sca e ing pa ame e measu emen . The ela ionship be ween disc e e ime delay nand con inuous ime delay τis gi en by: τn=n1 B,(2) whe e 1/Bs ands o he ime esolu ion (see Equa ion 7). Fo a s a is ical cha ac e iza ion o he UWB channel, we use he MDP which is de ined as he magni ude o complex CIR: A(τ) =|h(τ)|.(3) 3.1 S a is ical desc ip ion o he ecei ed signal 3.1.1 Independen iden ically dis ibu ed (IID) phase In his chap e , he phase s a is ics o he measu ed CIRs a e p esen ed. As isible in Figu e 5 [ igh ], acco ding o he ecd e alua ed o each measu ed CIR, he phase α(τ) is uni o mly dis ibu ed. Mo eo e , u ilizing he Pea son co ela ion coe icien ρα,β, we e alua e he mu ual dependence be ween phases o all measu ed posi ions deno ed as αand β.ThePea - son co ela ion coe icien is gi en as: ρα,β=E(α(τ) −ξα)β(τ) −ξβ ℘α℘β,(4) whe e ℘αdeno es he s anda d de ia ion and ξα he mean o α(τ). The e alua ion o he Pea son co ela ion coe - icien is isible in Figu e 5 [le ] showing unco ela ed beha io o α(τ). The ope a o E[·] deno es he expec ed alue. Figu e 5 Pea son co ela ion coe icien and ecd cu es. [le ] The Pea son co ela ion coe icien ρα,βe alua ion o he measu ed α(τ).[ igh ] The ecd cu es o α n. The close he blue ecd cu e o he ed line, he close he p obabili y dis ibu ion o α(τ) o he uni o m dis ibu ion. Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 5 o 10 Acco ding o he esul s p esen ed in Figu e 5, we con- clude ha α(τ) is iid uni o mly dis ibu ed wi h espec o he measu emen numbe α. 3.1.2 S a is ics o he ecei ed signal magni udes and GEV U ilizing MLEs [21], we ha e ound a s a is ical model o ecei ed signal magni udes. As seen in Figu es 6 and 7, he ecei ed signal magni udes can be app oxima ed using he GEV dis ibu ion [22] wi h he PDF gi en by: (x|0, μ,σ) =1 σexp{−z−exp(−z)},whe ez=x−μ σ, (5) wi h μbeing he loca ion pa ame e and σ he dis ibu- ion scale pa ame e . Equa ion 5 ep esen s he GEV ype I dis ibu ion, also known as log-Weibull dis ibu ion, whe e he shape pa ame e de ined in he egula GEV is se o ze o. This app oach is jus i ied in Sec ion 3.1.3. In o de o cap u e he s a is ical na u e o he en i on- men , we ha e pe o med 90 measu emen s pe mu ing bo h he TX and RX an enna placemen s as well as he in- ca sea occupancy. In Figu e 6, we can see he CDF cu es o all pe mu a ions o he an enna placemen and occu- pancy,whileeachcu eis i edbyaGEV ypeI andom p ocess ob ained by he MLE i ing. 3.1.3 GEV pa ame e s as a andom p ocess Acco ding o he obse a ions o esul ing GEV pa am- e e s, we app oxima e he co esponding μ,k,andσ pa ame e s wi h he s a is ical model ob ained by MLE. Figu es 8, 9, and 10 compa e he CDFs o he measu ed μ,k,andσpa ame e s wi h andom p ocesses o co e- sponding dis ibu ions. Figu e 7 PDF o he ecei ed signal magni udes in dBm o one measu emen se up. The measu ed PDF is i ed wi h he GEV ype I p ocedu e. The loca ion pa ame e μ ollows he logno mal dis i- bu ion gi en as: (x|η,ν) =1 xη√2πexp −(lnx−ν)2 2η2,(6) whe e νis he mean and η ep esen s he s anda d de i- a ion. The ex ac ed shape pa ame e kis o signi ican ly low alues; he e o e, ou choice o he GEV ype I (also known as log-Weibull) cha ac e ized by k=0 is app o- p ia e (see Equa ion 5). The scale pa ame e σis no mally dis ibu ed. Figu e 6 The CDF o ecei ed signal magni ude in dBm o all pe mu a ions o an enna placemen and occupancy. Measu ed cu es a e i ed wi h he GEV ype I p ocedu e. Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 6 o 10 Figu e 8 The CDF plo o loca ion pa ame e μ i ed wi h logno mal dis ibu ion (wi h 95% con idence in e al). A summa ized o e iew, including he speci ic alues o μand η, is gi en in Table 1. Thanks o a high numbe o pe o med measu emen s, he abula ed alues ep esen ypical da a o an in- ehicle channel which also applies o ehicles o simila size, sea con igu a ions, and ma e ials u ilized o i s manu ac u e. A co ela ion be ween he de i ed pa ame e s μand η (k=0) exhibi s a e y weak posi i e co ela ion alue o 0.35 wi h a p alue below 6 ×10−4. Thus, o ec ea e he ecei ed signal magni udes, one can a bi a ily choose he pa ame e s μand ηacco ding o Table 1. −0.2 −0.15 −0.1 −0.05 0 0.05 0.10.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 k Cumula i e p obabili y k: shape pa ame e o GEV, measu ed Logis ic dis ibu ion i Figu e 9 The CDF plo o shape pa ame e k i ed wi h a logis ic dis ibu ion (wi h 95% con idence in e al). Since he alues o he k pa ame e a e e y small, in he ollowing we se k=0; hus, he GEV becomes a GEV ype I (o log-Weibull) dis ibu ion. Figu e 10 The CDF plo o scale pa ame e σ i ed wi h a no mal dis ibu ion (wi h 95% con idence in e al). Due o a high lexibili y o he GEV i , which is gi en by h ee inpu pa ame e s as opposed o usual wo pa ame- e s, he MLE me ic ecommends he GEV dis ibu ion. On o he hand, au ho s in [23] claim ha he e is no heo- e ical explana ion o encoun e ing his dis ibu ion ype. We may, howe e , add ha he GEV con ains he accep ed log-Weibull dis ibu ion as a special case o k=0. 4 Localiza ion One o he o en discussed UWB applica ions is p ecise anging and localiza ion especially when he TOA ech- nique is used. As men ioned abo e, his is because he la ge UWB bandwid h allows excellen ime esolu ion (see Equa ion 7) and MPC sepa a ion. Because we had measu ed he channel ans e unc ion o many di e - en an enna posi ions, we wan ed o ge some insigh in o a ainable anging accu acy. Ou es ima ion o he dis- ance esul s om he CIR calcula ed om he complex ans e unc ion. This app oach gi es some limi a ions compa ed o a di ec channel sounding in he ime domain whe e some ad anced echniques such as he ma ched il- e ing o he known Gaussian pulses o a well-co ela ed bina y sequences can be used [24]. We calcula ed he an enna dis ance using he TOA ech- nique based on he de ec ion o he i s ay ansmi ed Table 1 Summa iza ion o GEV ype I pa ame e s cha ac e izing in- ehicle en i onmen o 90 pe mu a ions o an enna placemen and ca sea occupancy μσk Dis ibu ion ype Logno mal No mal Logis ic Mean ν48.37 5.48 −0.08 Va iance η31.05 0.26 0.002 Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 7 o 10 om a pa icula an enna. The p oposed h eshold-based sea ch algo i hm compa es indi idual signal samples o he CIR wi h a ce ain h eshold in o de o iden i y he ampli ude peak co esponding o he i s MPC. This app oach allows o calcula e dis ance also in he NLOS scena io because he i s ay may no be he s onges ay. Howe e , pene a ion o he obs acles can cause some measu emen accu acy deg ada ion (see below). The aim o his chap e is o gi e a basic idea abou accessible a e age e o and s anda d de ia ion o he measu ed dis- ances o he en i e UWB band and o he i s band g oup and also o he emp y and occupied ca . Fo mo e in o ma ion abou he measu emen s and he dis ance calcula ion, see [25]. Because all he pa icula measu e- men s we e done o h ee TX an ennas and one RX an enna, we also calcula ed RX an enna posi ion using he 2D localiza ion echnique in o de o assess whe he i co esponds a leas oughly o eali y, i.e., whe he i is possible o example o eliably de ec a de ice on a pa icula sea . No e ha mos o he abo e men ioned applica ion does no need an accu a e localiza ion bu only ough es ima ion o he de ice posi ion. 4.1 Basic sys em pa ame e calcula ion The equency band Bde e mines he ime esolu ion o measu emen by [24]: T =1 B=1 FU−FL ,(7) whe e FUis he uppe equency and FLis he lowe e- quency o he band. The p opaga ion dis ance esolu ion is hen D =T c,(8) whe e c=2.998 ×108m/s is he speed o he ligh . The maximum measu able p opaga ion dis ance depends on he numbe o measu ed equency alues NMinside he equency band, i.e., on he equency s ep sacco ding o Dmax =c BNM=c s .(9) I is ob ious om he equa ions abo e ha na owing he bandwid h dec eases he dis ance esolu ion and he educ ion in he measu ed equency poin s sho ens he measu able p opaga ion dis ance. 4.2 Ranging and localiza ion o he ecei ing an enna As men ioned abo e, he main aim o his sec ion is calcula ion o he a e age e o and s anda d de ia ion o he anging and e i ica ion o he RX an enna posi ion (Table 2). The p ocessing o he measu ed da a consis s in he ollowing s eps: •Calcula ion o he CIR •De ec ion o he i s inciden ay •Calcula ion o he RX - TX1 o TX3 dis ances and anging e o s •RX an enna localiza ion The CIR was calcula ed using he IFFT in combina- ion wi h a Blackman window (see Figu e 11) applied o all 801 equency esponse poin s. Be o e e o s a is- ics calcula ion, we es ed a ew windows ( ec angula , Hann, Hamming, la op, Blackman, and Kaise -Bessel). Al hough some windows (e.g., ec angula ) a e gene ally ecommended o he applica ions whe e good compo- nen sepa a ion is equi ed, hese windows could be inap- plicable in ou case as hey may p oduce la ge side lobes ha c oss he h eshold and cause inco ec i s ay de ec- ion. Expe imen ally, we ound ha he bes esul s gi ing hedis ancescloses o eali ya egi enby heBlackman window. The second bes esul s can be hen ob ained using he Hann window. The h eshold o he i s ay de ec ion is gene ally de e mined by he noise loo . I s alue is equal o he le el o he peaks o noise, i.e., o he maximum ampli ude o he CIR whe e he mul ipa h componen ampli udes a e below noise le el. I is ob ious ha he p oposed algo- i hm wo ks eliably in bo h LOS and NLOS scena ios, bu i ails in some NLOS cases when he i s (di ec ) ay is s ongly a enua ed and d owned in noise. The dis ance o RX and TX an ennas is gi en by he o - mula DA=cTD,whe eTDis he i s de ec ed ay a i al ime. The e o s a is ics we e calcula ed sepa a ely o he emp y and occupied ca . I was expe imen ally disco e ed ha he wo o h ee passenge s si ing in he ca compa - men cause e y simila esul s, and he e o e, hese cases we e joined in o one se o esul s. Fo he RX an enna localiza ion, he ila e a ion echnique [24] was applied. Using he h ee calcula ed dis ances, his echnique allows 2D localiza ion. Calcula ion o he a e age e o and s anda d de ia ion o he measu ed dis ances is summa ized in Table 3 ( o Table 2 The pa ame e s used o he anging Bandwid h F eq. s ep Time esolu ion Dis ance Max. p opag. Max. measu emen [GHz] [MHz] [ns] esolu ion [cm] dis ance [m] ime [ns] UWB 8 10 0.125 3.750 30 100 Fi s band g oup 1.58 0.633 18.987 Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 8 o 10 Figu e 11 Magni udes o channel ans e unc ions and channel impulse esponses (RX an enna was placed on igh ea sea ). he en i e UWB) and Table 4 ( o he i s band g oup). The imein e alsused o henoisepeakde ec ionwe e 0 o 1.25 ns (be o e ecei ing o he i s MPC) and 80 o 100 ns (whe e he MPC can be neglec ed). These ime in e als co espond o he ollowing dis ances: 0 o 37.5 cm (minimum dis ance o RX-TX an ennas in all scena ios is 50 cm) and 24 o 30 m. The e e ence an enna dis- ances we e measu ed by a ule . We compa ed 15 ×3 dis ances wi hou passenge s and 15 ×3dis anceswi h wo o h ee passenge si ing on he sea s su ounding he RX an enna. An example o peak de ec ion o he emp y ca is shown in Figu e 12 (uppe pa o he en i e UWB and lowe pa o he i s band g oup), while Figu e 13 depic s he 2D localiza ion esul also o he UWB and i s band g oup. 4.3 Posi ioning esul s and sou ces o e o I is ob ious ha he ough dis ance esolu ion in he case o he i s band g oup measu emen causes ma kedly highe a e age e o and s anda d de ia ion compa ed o he measu emen o he en i e UWB band. The calcula ed Table 3 A e age e o and s anda d de ia ion o he measu ed dis ances o he i s band g oup TX1 TX2 TX3 To al A e age e o wi hou passenge s [cm] 6.76 6.30 5.75 6.27 A e age e o wi h wo o h ee passenge s [cm] 11.83 10.37 7.62 9.94 S anda d de ia ion wi hou passenge s [cm] 7.49 6.86 2.10 5.87 S anda d de ia ion wi h wo o h ee passenge s [cm] 11.13 9.28 8.95 9.80 dis ances exhibi no iceable posi i e bias caused by a ew phenomena: • Exis ence o di e ence be ween he calib a ion plane and phase cen e o he an enna. The coaxial in e aces o he an ennas (line be ween he connec o and phase cen e o he an enna) we e no included when he VNA was calib a ed. They we e applied only du ing channel measu emen and inc eased he o al an enna dis ance. • Inaccu a e e e ence measu emen . Dis ance measu ed be ween he an ennas by he ule was pe o med be ween he cen e s o he op o cones which a e no iden ical o he phase cen e s o an ennas. In many cases, he measu ed dis ance we e sligh ly sho e (when he TX an enna was upside down wi h ega d o RX an enna) . • Time lag in he i s ay de ec ion. The i s ay (peak) de ec ion abo e he h eshold exhibi s andom delay in he in e al 0 o D due o he disc e e na u e o he CIR ime axis. Recei ed ay canno be gene ally de ec ed in ad ance. Table 4 A e age e o and s anda d de ia ion o he measu ed dis ances o he i s band g oup TX1 TX2 TX3 To al A e age e o wi hou passenge s [cm] 25.85 21.04 14.3 20.39 A e age e o wi h wo o h ee passenge s [cm] 34.73 23.50 10.82 23.02 S anda d de ia ion wi hou passenge s [cm] 20.76 13.88 8.74 15.67 S anda d de ia ion wi h wo o h ee passenge s [cm] 20.13 12.63 7.88 17.26 Blumens ein e al. EURASIP Jou nal on Wi eless Communica ions and Ne wo king (2015) 2015:104 Page 9 o 10 Figu e 12 Fi s peak de ec ion o CIRs ( o en i e UWB) [uppe ] o emp y and [lowe ] occupied ehicle. • Inco ec MPC componen de ec ion. La ge a enua ion o some obs acles in he ca may a oid co ec de ec ion o he di ec ay. In his case, he o he e lec ed MPC which a els on a longe pa h is ega ded as he i s ay. • Lowe wa e p opaga ion eloci y in media. The eloci y o an elec omagne ic wa e pene a ing an obs acle is less han ha in ee space, and i depends on he obs acle ma e ial cons an s. The i s phenomenon is sys ema ic and can be sub- ac ed (i is abou 2 cm oge he o wo an ennas). The wo las phenomena occu only in he NLOS scena io. In he las case, he eloci y in some ma e ial can be calcu- la ed acco ding he o mula p=c/√ε μ ,whe e pis he eloci y o p opaga ion in m/s, μ is he ma e ial el- a i e pe meabili y, and ε is he ela i e pe mi i i y. I is easy o ind ha when, o example, he wa e passes he 10-cm- hick plas ic obs acle (ε =2 o3,μ = 1 [26]), he p opaga ion ime delays a e in he in e al 0.138 o 0.244 ns which esul s in he dis ance bias om 4.1 o 7.3 cm. 5Conclusions We pe o med an ex ensi e UWB measu emen campaign o he ehicula passenge compa men . The measu ed -300 -200 -100 0 100 200 300 -200 -100 0 100 200 300 400 X coo dina es [cm] Y coo dina es [cm] x=36.3685cm y=131.665cm x=31.0792cm y=135.2535cm 8 GHz bandwid h 1.58 GHz bandwid h Co ec posi ion Figu e 13 Localiza ion o he RX an enna using TOA (RX an enna was placed on on passenge sea ).