ULTRAWIDEBAND (UWB) radio has mainly been considered

Transcription

1 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 7, SEPTEMBER UWB Channe Modeing in Roadway and Indoor Parking Environments Joon-Yong Lee, Member, IEEE Abstract In this paper, the characteristics of utrawideband (UWB) channes on outdoor roadway and indoor parking environments are investigated. A set of propagation measurements were conducted on roadways and in underground parking garages, and the resuting propagation data consist of approximatey 1200 measured signas. Mutipath properties are characterized using tapped-deay-ine and Saeh Vaenzuea (S V) modes. Poweraw-type decay patterns are observed in the ray-energy decay, as we as in the power deay profie. Resuts are compared with existing UWB modes, incuding the IEEE a channe mode. Index Terms Channe mode, custer mode, outdoor environments, path oss, tapped-deay-ine (TDL) mode, utrawideband (UWB). I. INTRODUCTION ULTRAWIDEBAND (UWB) radio has mainy been considered to be a viabe soution for various short-range indoor appications. However, outdoor appications of UWB, such as sensor networks, search and rescue, and vehice-tovehice communications, are aso of great importance. As compared to the extensive work conducted on indoor UWB channe characterization, very itte work has been reported on outdoor channe modeing [1] [5]. These outdoor modes are based on measurements or ray-tracing simuations for farm [1], suburban [2], outdoor office [3], outdoor parking [4], and university campus [5] areas. In this paper, we characterize UWB channes in specia types of environments, reated to transportation systems and vehicuar appications, specificay, outdoor roadways and underground parking areas. These environments are unique because of the presence of vehices, whose density can vary depending on traffic conditions. Outdoor roadways might be ess open than farm areas but ess compex than outdoor office environments. Underground parking has different channe characteristics from conventiona indoor environments such as office and residentia environments. Narrow-band channe modes for vehice-tovehice communication scenarios have previousy been studied [6] [9]; however, no UWB channe mode for these environ- Manuscript received Juy 10, 2009; revised November 14, 2009 and January 21, 2010; accepted February 5, Date of pubication March 4, 2010; date of current version September 17, This work was supported in part by the Daegu Gyeongbuk Institute of Science and Technoogy and in part by HGU The review of this paper was coordinated by Prof. T. Kuerner. The author is with the Schoo of Computer Science and Eectrica Engineering, Handong University, Pohang , Korea (e-mai: joonee@ handong.edu). Coor versions of one or more of the figures in this paper are avaiabe onine at Digita Object Identifier /TVT ments has been proposed. Understanding UWB propagation channe characteristics in these environments wi be usefu for the potentia appication of UWB for intervehice communication or ranging. We performed a set of measurements in the time domain at various ocations in the city of Pohang, Korea. Measurement setups and pans are described in detai in Section II. Using the database, we first derived a path oss mode (see Section III) and then modeed the mutipath fading statistics (see Section IV). We aso compared the derived mode with other existing UWB channe modes in Section V, which is another important contribution of this paper. II. MEASUREMENT CAMPAIGN A. Measurement System Measurements were performed in both the outdoor and the indoor environments. The PusON 210 evauation kit manufactured by the Time Domain Corporation was used to obtain channe responses. The scanning receiver system consists of a transmitter and a receiver, and each unit is controed by a notebook computer. The receiver empoys mutipe in-phase and quadrature correator pairs to synchronize and scan the waveform [10], [11]. The tracking correators have the capabiity of synchronizing with and tracking the received puse train. This enabes the capture of the received waveform without a wired connection between the transmitting and receiving sides, which serves as a common trigger signa. The scanning correators sampe the received signa reative to the ock spot estabished by the tracking correators. Verticay poarized omnidirectiona dipoe antennas are used in the system. Measurements were made on roadways and underground parking garages at various distances. The transceivers were mounted 60 cm above the ground, which is the approximate height of a vehice s grie or bumper. UWB puses were emitted by the transmitter every 10 μs, and the received waveform was averaged over 1024 traces to increase the dynamic range. The resuting dynamic range of our measurement system with averaging is greater than 60 db. Each waveform was captured at a samping rate of GHz. Fig. 1 shows a normaized pot and the spectra density of the UWB puse received at 3 m with a cear ine of sight (LoS). The frequencies of 10-dB emission points are 3.28 and 5.03 GHz. This waveform is used as a correator tempate signa w(t) and is modeed by w(t) =A exp( at 2 )sin(ωt), (1) where a =5.55, ω =26.15, and A is a constant /$ IEEE

2 3172 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 7, SEPTEMBER 2010 Fig. 1. (Left) Tempate waveform and (right) its normaized spectra density. Dotted ine in the figure on the eft is the mathematica mode for the waveform. B. Measurement Pan Typica channe environments where the measurements performed are cassified into the foowing three channe modes [12]: 1) CM 1: outdoor roadways/ m; 2) CM 2: outdoor roadways/30 80 m; 3) CM 3: underground parking garages/ m. A measurements were made with a cear LoS, and sampes were coected with as itte movement as possibe to approximate a stationary channe. With regard to the tests for CM 1 and CM 2, we chose 67 fixed sites on roadways in the city of Pohang, Korea, incuding the campus driveways of Handong University (HU) and Pohang Coege. At each site, we took sampes at different ocations with a 2.5-m separation between each ocation, whereas the receiver was fixed at one ocation. For safety purposes, we seected roadways where traffic was very ight and, instead, paced vehices around the transceivers to simuate traffic situations. Exempary test pans are shown in Fig. 2. For CM 3, sampes were taken in the underground parking garages of five apartment compexes. The number of measured signas in CM 1, CM 2, and CM 3 are 425, 388, and 371, respectivey. Exempary measured signas with normaized ampitude are shown in Fig. 3. From this figure, we can observe that the first path arrives at approximatey 10 ns in a cases, regardess of range. This is because the time of each sampe is defined reative to the initia ock point, which is estabished near the first arriving path [10], [11]. III. PATH-LOSS MODEL A genera mode for path oss in decibes at distance d, i.e., PL(d), is we known to be [13] ( ) d PL(d) =PL 0 +10n PL og + X PL (2) where n PL is the path oss exponent, and X PL is a random variabe that is often modeed as a zero-mean Gaussian. Constant PL 0 is the path oss at a reference distance d 0 =1m. Fig. 4 shows the scatter pots of path oss reative to 1 m, which are obtained by computing the tota received power of each measured signa. The path osses for CM 1 and CM 2 are potted together because they are measured in simiar environments, with different distances. The vaues of n PL are 1.60 for CM 1 and CM 2 and 1.50 for CM 3. As shown by the comparisons d 0 Fig. 2. (a) Exempary test pan for CM 1. Sampes were taken at eight ocations with a 2.5-m separation. (b) Measurement photographs captured at a campus driveway of Pohang Coege and in the underground parking structure of an apartment compex. in Section V, these vaues are very cose to those given by the IEEE a channe mode for open outdoor environments. For CM 1, it is observed that path oss decreases with an increase in distance between 12.5 and 20 m. The same phenomenon is aso observed for CM 3 at simiar distances. This phenomenon may indicate the existence of a geometrica factor that affects the tota received power at certain distances and does not dramaticay vary with a change in environment. This is very unusua, considering the fact that the number of measurement sites is arge enough to provide significant spatia diversity. We beieve that ground refection is a strong candidate cause for this phenomenon. Reca that most of our measurements were made on pain asphat or concrete grounds.

3 LEE: UWB CHANNEL MODELING IN ROADWAY AND INDOOR PARKING ENVIRONMENTS 3173 Fig. 3. Typica measured responses in each channe mode. The ampitudes are normaized to the peak. Therefore, it is highy probabe that the differentia path ength between the direct path and the ground refection path is amost constant for a fixed distance, regardess of changes in the surrounding environment. Assuming that the ground refection occurs at the midpoint between the ocations of the transceivers on the ground, the interarriva times between the direct path and the ground refection signas corresponding to and 20.0-m ranges are and ns, respectivey. Since the main-obe width of the tempate signa is approximatey ns, the ground refection signa at these ranges can constructivey be added to the direct path signa, increasing the tota received power. Furthermore, because of the openness of CM 1, a cose mutipath component other than the ground refection signa is not ikey to exist, which makes its constructive fading action more notabe. This aso hods for CM 3, athough it is an indoor environment. This is because the geometrica compexity of CM 3 is very ow as compared with other indoor environments, which wi be verified by channe parameters presented in Section IV-B3. IV. MULTIPATH MODEL To characterize sma-scae effects, we adopted two different cassica approaches. One approach is to use the tapped-deayine (TDL) mode [13], [14], where the path arriva time is quantized into discrete sampes whose spacing satisfies the Nyquist criterion. Using this mode, we statisticay characterized the received power in each individua time bin, as we as the average power decay pattern. The other approach is to use the Saeh Vaenzuea (S V) mode, which is a typica custer mode. In this mode, the paths are assumed to be grouped into custers. Using this mode, we extracted some auxiiary channe parameters, as we as the S V mode parameters. A. TDL Mode One of the conventiona ways to represent a channe in the TDL mode is to define the power deay profie (PDP) [15] such that p(t) = k p k δ(t τ k ), τ k = kτ w k 0. (3) Fig. 4. Scatter pots of the path oss versus the range. The coefficient p k denotes the received power in the nth time bin, and it is normaized such that k p k =1. The parameter τ w must be chosen to be ess than or equa to the reciproca of the bandwidth of w(t). In this paper, it was chosen to be 1.4 ns, which is the approximate width of the puse represented by (1). Fig. 5 shows the scatter pots of p k vaues in each channe mode. It is generay known that the PDP of UWB channes foows an exponentia decay pattern [14] [16]. However, it exhibits a power-aw decay. Thus, we can mode it by 10 og 10 p k = η 0 10n pdp og 10 (τ k )+X pdp, k>0 (4) where n pdp is the power decay exponent. Parameters η 0 and X pdp are a constant and a random variabe, respectivey. Parameter n pdp is determined by finding a inear east square fit to the scatter pot shown in Fig. 5. Since (4) is not vaid

4 3174 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 7, SEPTEMBER 2010 Fig. 6. Averaged PDPs in each channe mode. Geometric average was taken for each coefficient p k. Fig. 5. Scatter pots of the PDPs and their power-aw fits. TABLE I COMPARISON BETWEEN POWER-LAW AND EXPONENTIAL FITS FOR THE PDPs PDPs with power-aw decay have aso been reported in a 2.6-GHz propagation mode for LoS street microce environments in urban areas [17]. It is interesting to note that these environments are simiar to our environments. However, it is difficut to make a generaized statement that a certain type of environment, for exampe, the PDPs of outdoor LoS, wi exhibit a power-aw decay. It shoud be noted that a simiar decay pattern is observed in CM 3, which is indoors, as we as in CM 1 and 2, which are outdoors. At this point, it seems difficut to identify the exact geometrica factors that introduce a power-aw decay rather than a pure exponentia decay. Fig. 6 shows the averaged PDPs. These were obtained by finding the geometric average of sampes of each coefficient p k, which were coected from different channe reaizations in each channe mode. It is observed that decay patterns up to 10 ns, corresponding to approximatey 3 m of differentia path ength, are very simiar in a channe modes. This indicates that the spatia structures, which generate a cose mutipath, are not very different across channe modes. Now, we mode the normaized power coefficients {p k } as ognorma random variabes such that 10 og 10 p k N ( μ k,σ 2 k). (6) for τ 0 =0, p 0 is not incuded for fitting. We compared the goodness of fitness of the power-aw and exponentia fits using the mean square error (MSE); Tabe I shows the resuts of this comparison. Parameter γ pdp is the power decay factor in the foowing exponentia mode: 10 og 10 p k = 10τ k γ pdp n 10 + Y pdp (5) where Y pdp is a random variabe. Notice that the power-aw fit exhibits fitness that is db better than the exponentia fit. The parameter μ k is aready shown in Fig. 6, and σ k is shown in Fig. 7. To examine the goodness of fitness, a chi-square test with 30 DOF was performed for the first 36 time bins (within the first 50 ns). The resuting chi-square vaues for CM 1, which significanty vary with excess deay, are shown in Fig. 8. The percentage of the tested bins that passed the hypothesis test at 5% and 10% significance eves are summarized in Tabe II, and the resuts indicate a very good fit. The cumuative distribution functions (cdf s) of the received power of some exempary time bins in CM 1 are shown in Fig. 9. B. Custer Mode The custer mode is one of the most widey used modes for UWB channes. In this mode, the deay parameters are

5 LEE: UWB CHANNEL MODELING IN ROADWAY AND INDOOR PARKING ENVIRONMENTS 3175 Fig. 7. Standard deviation for ognorma fading σ k versus excess deay. Fig. 9. CDFs of exempary time bins in CM 1. Fig. 8. Resuting chi-square vaues of the fitness test for each time bin in CM 1. TABLE II PERCENTAGE OF THE BINS THAT PASSED THE CHI-SQUARE TEST Fig. 10. Mean number of significant paths greater than the given threshod. assumed to be continuous, and a custering effect is considered. Standardization groups such as IEEE a [18] and IEEE a [1] have adopted the modified S V mode, which is a typica exampe of the custer mode. To examine the channe parameters, channe impuse responses (CIRs) are first obtained. They can be obtained by deconvoving the tempate waveform w(t) from the measured signas. The CLEAN agorithm [19], which is a deterministic maximum-ikeihood approach, is used to estimate the channe parameters. A stopping threshod is conventionay determined reative to the peak power [20], [21] or noise foor [3], [5]. In this paper, we chose the threshod to be 30 db of the peak power. 1) Number of Major Paths and RMS Deay Spread: The number of significant paths and the RMS deay are often used as auxiiary parameters indicating the channe characteristics, and both are reated to the spread of the channe. They vary with the threshod used in the CLEAN agorithm, and Figs. 10 and 11 show their mean vaues with varying threshods. It is interesting to note different increasing patterns with decreasing threshods. The number of paths exponentiay increases with a decrease in the threshod, whie RMS deay spread ineary increases. This is probaby because of the fact that, as the threshod decreases, the contribution of newy detected paths in increasing the deay spread decines because of the decrease in their ampitudes. Both parameters are smaer in CM 1 than in CM 2 and CM 3. This may indicate the presence of more refectors in CM 2 and CM 3 than in CM 1. Geometricay, it seems to make sense that more major refection paths exist at ong ranges than at short ranges and more exist indoors than outdoors.

6 3176 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 7, SEPTEMBER 2010 Fig. 11. Mean RMS deay spread versus the threshod. 2) Path Arrivas: In the custer mode, the CIR h(t) is generay represented by [1] h(t) = L =0 k=0 K α k, δ(t T τ k, ). (7) The parameter T is the deay of the th custer, and τ k, is the deay of the ray reative to the custer arriva time. The parameter α k, denotes the path strength. The subscript {k, } indicates that the quantities depend on the th custer and the kth ray. We do not consider shadowing in this paper. To examine the statistics of these channe parameters, we first need to identify custers from the CIR. There exist some heuristic approaches to custer identification [22] [24]; however, it is very difficut to design a robust estimation agorithm. Instead, we carried this out manuay using a technique simiar to the one used in [19]. We first carried out a siding correation between the squared version of the CIR and a 2-ns-wide rectanguar window, and then, we used visua inspection [1], [20] to identify custers. Some exempary custer maps are reproduced in Fig. 12. Custer and ray arrivas are modeed as Poisson processes with different arriva rates, namey, Λ and λ, respectivey. They can be obtained by measuring the mean intercuster and interray arriva times. 3) Ampitude Statistics: In the IEEE a channe mode [1], the path strength is assumed to foow ognorma fading, with its mean energy exponentiay decreasing with excess deay, i.e., E [ α k, 2] =Ω 0 e T /Γ e τ k,/γ where Γ and γ are the custer and ray decay factors, respectivey, and Ω 0 is a constant that accounts for shadowing. We first attempted the exponentia fits for ray and custer decays, as in (8). We determined the ray decay factor γ using channe coefficients coected from a measured signas in each channe mode. The excess deay was caibrated reative to the first arriving ray in each custer. Instead of normaizing the ray ampitudes within the custer with respect to a specific path such (8) Fig. 12. Exempary custer maps in each channe mode. Each figure shows an output of a siding correator with a 2-ns rectanguar window. Fig. 13. Determination of the scaing factor β (m) from the ray energies. This is determined by the vertica vaue corresponding to the horizonta vaue T (m) in the inear regression ine for the th custer. as the peak or first arriving path [20], we scaed them such that the resuting squared error was minimized. That is, the optima vaue for parameter γ, i.e., ˆγ, was determined by [ ˆγ =arg min min} γ {β (m) m k α (m) 20 og 10 β (m) + k, ] (m) 2 10τ k, (9) γ n 10 where the superscript (m) indicates that the parameter is datasampe dependent. Parameter β (m) accounts for the scaing of the th custer of the mth measured signa. In other words, it denotes the amount of vertica shift of data points {(τ (m) k,,α (m) k, )} in the scatter pot, which is required to minimize the squared error (see Fig. 13).

7 LEE: UWB CHANNEL MODELING IN ROADWAY AND INDOOR PARKING ENVIRONMENTS 3177 Custer decay was characterized by fitting to the scatter pot of custer energy. Instead of the peak or averaged energy [1], [20] of each custer, we defined the energy of the th custer of the mth signa, i.e., ν (m), by the quantity satisfying ν (m) =arg min ν k [ α (m) 20 og 10 ν+ k, 10τ (m) k, ˆγ n 10 ] 2. (10) In fact, ν (m) is equa to β (m), provided that γ is chosen to be its optima vaue in (9). Simiar to ray decay characterization, the custer energies of the mth signa are scaed by the constant θ (m), such that the squared error is minimized. The optima vaues of Γ can thus be found by [ ] ˆΓ=arg min min ν (m) (m) 2 Γ θ (m) 10T +. (11) Γn10 {θ (m) } m We aso attempted power-aw fits for ray- and custer-energy decays, assuming that E [ α k, 2] =Ω 0 T n custer τ n ray k, (12) where n custer and n ray are the custer and ray decay exponents, respectivey. A power-aw decay of custer and ray energies has been reported in a UWB channe mode for industria environments [25]. A normaization technique simiar to that used for exponentia fits was used to determine the optima vaues for n custer and n ray by ˆn ray = arg min n ray min } {β (m) [ α (m) 20 og 10 m ˆn custer = arg min n [ custer [ min {θ (m) } where κ (m) m >0 [ = arg min κ k>0 k>0 κ (m) k, ] 2 β (m) +10n ray og 10 τ (m) k, θ (m) ] ] 2 +10n custer og 10 T (m) [ α (m) 20 og 10 k, κ +10ˆn ray og 10 τ (m) k, ] 2 ] (13) (14) (15) denotes the custer energy. A coefficients with k =0 were excuded in (13) and (15) because they are not vaid for τ 0, =0. For a simiar reason, (14) was not evauated for =0. Figs. 14 and 15 show the scatter pots of ray and custer energies of CM 3 with different fits. It shoud be noted that the data points appearing in each scatter pot are different Fig. 14. Scatter pot of the ray energies of CM 3 and the (upper) exponentia and (ower) power-aw fits. for exponentia and power-aw fits (this is more noticeabe in Fig. 15). This is because the scaing factors were chosen in an optima manner reative to the fitting function. Tabe III shows a comparison between the resuting MSE of the two fits. It is interesting to note that ray decay is coser to a power-aw fit, whereas custer decay is coser to an exponentia fit. The fitness gains are around 1 db for both cases. The standard deviations of ognorma fading terms for the custer and ray (σ custer and σ ray for exponentia fit and σ custer and σ ray for power-aw fit) were then evauated on the basis of the determined optima parameters. Tabe IV summarizes a the resuting fading parameters. V. C OMPARISON OF OUR EXTRACTED MODEL WITH OTHER OUTDOOR ULTRAWIDEBAND CHANNEL MODELS In this section, we compare the channe mode extracted in this paper with the previousy reported outdoor and indoor UWB modes for LoS scenarios. Outdoor modes are comparabe to CM 1 and/or CM 2 of our mode, which are named in the foowing HU mode, and possiby, indoor modes may be comparabe to CM 3. Stricty speaking, some channe

8 3178 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 7, SEPTEMBER 2010 TABLE IV FADING PARAMETERS FOR THE MODIFIED S V MODEL Fig. 15. Scatter pot of the custer energies of CM 3 and the (upper) exponentia and (ower) power-aw fits. TABLE III COMPARISON BETWEEN POWER-LAW AND EXPONENTIAL FITS FOR THE CLUSTER AND RAY ENERGY DECAYS parameters proposed in different works may not directy be comparabe because of inherent differences in the modeing methodoogy, for exampe, the custer identification method and threshod empoyed in the CLEAN agorithm. However, simiarities and contrasts observed in different modes can hep eucidate how environments affect the channe characteristics. The IEEE a standardization group has proposed UWB channe modes for various environments. One of them is proposed for an open outdoor environment (farm area) and is not based on measurements but on ray-tracing simuations [1]. Some of our measurement sites for CM 1 and CM 2 are simiar to open outdoor environments, whereas others are near buidings with surrounding vehices. Di Francesco et a. [2] proposed an outdoor mode based on measurements in a suburban area, around which cars and was were present. 1 The sampes were taken in the time domain using the same measurement system we used in this paper. Another outdoor mode proposed by Kim et a. [3] is based on a set of frequency-domain measurements in the midde of two connected buidings. 2 We beieve that this mode invoved a much more compex mutipath structure than our mode. In their study, LoS measurements were performed at a distance of 3 24 m, which is comparabe with CM 1, and sma-scae effects were modeed by the modified S V mode. The study conducted by Choi et a. [4] at Seou Nationa University mainy focused on characterizing path oss properties and frequency dependence of channe parameters for outdoor environments. Among their measured sampes, LoS data taken at a parking ot, namey, environment 1, with a range of ess than 28 m is comparabe with CM 1. Souza et a. [5] carried out severa measurement campaigns on the Federa Fuminense University (UFF) campuses and examined PDPs, the tempora dispersion parameters, and path osses. 1 This study was conducted at the University of L Aquia (UDA). 2 This study was conducted at the Institute for Infocomm Research (IIR).

9 LEE: UWB CHANNEL MODELING IN ROADWAY AND INDOOR PARKING ENVIRONMENTS 3179 TABLE V COMPARISON OF CHANNEL PARAMETERS OF PREVIOUSLY REPORTED UWB MODELS TABLE VI COMPARISON OF FADING PARAMETERS OF PREVIOUSLY REPORTED UWB MODELS We aso incuded a few indoor modes for comparison. One of them is the IEEE a mode for indoor residentia environments [1], which is based on measurements with a range of 7 20 m. Another mode for indoor residentia areas, which was proposed by Chong and Yong [20], extensivey anayzed the measurement data coected in various types of high-rise apartments with both arge-scae and sma-scae statistics. 3 Muqaibe et a. [26] examined channe parameters incuding path oss exponents and deay spreads on the basis of the measurements performed in buidings on the Virginia Poytechnic Institute and State University campus. Tabe V shows a comparison between the path oss exponent and some auxiiary parameters of the channe modes mentioned previousy. Each mode is named after the organization where the research was conducted. The vaues of the path oss exponents are simiar in both the outdoor and the indoor modes, ranging from 1.5 to 1.8; however, arger vaues were reported in the UDA and UFF modes. In the case of the UDA mode, the fact that it is based on non-los and LoS measurement data might be the cause for a arger n PL. The mean number of paths and the RMS deay spread refect the compexity of the mutipath structure. In both the HU and the IEEE a modes, these vaues are greater in indoor environments than in outdoor environments. The vaues of these parameters in our resuts refect the ow compexity of our environments. In particuar, the vaue for τ RMS in the HU mode is very sma compared with the vaues of other modes. Unike N p, the mean number of custers is around three in both the indoor and the outdoor modes, except in the IIR mode. In the IIR mode, the mean number is 13.6, which may refect the fact 3 This study was conducted at the Samsung Advanced Institute of Technoogy (SAIT). that the measurement site was fanked by two buidings and, possiby, other major refectors. Tabe VI shows a comparison between the fading parameters. In the HU and IEEE a modes, the vaues for the custer arriva rate are around in both the outdoor and the indoor environments. A arger Λ = is reported in the SAIT mode. This mode is based on the measurement data taken in furnished and empty apartments; therefore, it is probabe that major refectors were reativey cosey spaced. If we assume that the custer arriva rate is reated to the spacing of major refectors, a arger Λ makes sense. The ray-arriva rates are aso simiar in outdoor and indoor modes, i.e., around , except that they are ower in the IEEE a mode. It is interesting to note that the custer decay time constant is greater in outdoor modes than in indoor modes. We can interpret this as foows. Considering the differences in the geometric structures of indoor and outdoor environments, it is more highy probabe that, in indoor cases, a majority of rays within the custer at a certain deay have been formed by more mutipe refections than in outdoor cases. Since mutipe refections introduce a arge power oss, the custer decay increases in indoor environments. In CM 3 of the HU mode, Γ = is sti arge, even if it is an indoor environment. This is possiby because the underground parking areas are more open than other indoor environments, such as residentia environments. The ray-decay time constants of outdoor and indoor modes are not noticeaby different, except that γ = in the SAIT mode is somewhat arge, as compared with those of the other modes. VI. CONCLUSION In this paper, we have characterized UWB propagation channes in roadways and indoor parking areas on the basis of timedomain propagation data. The channe characteristics we have examined incuded a path oss mode, PDP, time dispersion, path arrivas, and sma-scae fading statistics. The measured path-oss exponent is 1.6 on roadways and 1.5 in underground parking garages, which are typica vaues reported in other LoS UWB modes. The ow compexity of CM 1 and CM 2

10 3180 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 59, NO. 7, SEPTEMBER 2010 is refected by very ow vaues for tempora dispersion and the sow decay of custer energy. This might be because most measurements have been performed when ony a few cars were in the studied environments. In the presence of heavy traffic, the compexity woud probaby have increased. Characterization of the channe for underground parking environments (CM 3) is a unique contribution of this paper. By comparing the vaues of N p and τ RMS for each channe mode, we found that the geometrica compexity of CM 3 was higher than that of CM 1 and CM 2. This makes sense because CM 3 is an indoor environment. However, the vaues of these parameters for CM 3 are sti sma as compared with those of other indoor modes, which indicates the reative openness of CM 3. It is interesting to note that the PDP foows a power-aw decay rather than an exponentia decay. Furthermore, ray-energy decay aso exhibits a power-aw pattern, whereas custer-energy decay foows an exponentia pattern. Further investigation is required to examine the geometrica factors that introduce different types of power decay patterns. ACKNOWLEDGMENT The author gratefuy acknowedges the vauabe comments and suggestions of Dr. J.-H. Lee and S.-D. Kim. The author aso wishes to thank J.-H. Jeong, W.-H. Kim, G.-Y. Ha, C.-S. Jung, J.-E. Oh, D.-W. Ha, S.-J. Kong, and E.-Y. Ahn for conducting propagation measurements and assisting data anaysis. REFERENCES [1] A. F. Moisch, D. Cassioi, C.-C. Chong, S. Emami, A. Fort, B. Kannan, J. Kareda, J. Kunisch, K. Siwiak, and M. Z. Win, A comprehensive standardized mode for utrawideband propagation channes, IEEE Trans. Antennas Propag., vo. 54, no. 11, pp , Nov [2] A. Di Francesco, M. Di Renzo, and R. Minutoo, Sounding and modeing of the utra wide-band channe in outdoor scenarios, in Proc. 2nd Int. Workshop Netw. With Utra Wide Band Workshop Utra Wide Band Sensor Netw., Ju. 2005, pp [3] C. W. Kim, X. Sun, L. C. Chiam, B. Kannan, F. P. S. Chin, and H. K. Garg, Characterization of utra-wideband channes for outdoor office environment, in Proc. IEEE WCNC, Mar. 2005, vo. 2, pp [4] J. Choi, N.-G. Kang, Y.-S. Sung, and S.-C. Kim, Empirica utra wide band channe mode for short range outdoor environments, in Proc. VTC Spring, Apr. 2007, pp [5] C. F. Souza and J. C. R. Da Beo, UWB signas transmission in outdoor environments for emergency communications, in Proc. 11th IEEE Int. Conf. Comput. Sci. Eng., Ju. 2008, vo. 2, pp [6] A. S. Akki, Statistica properties of mobie-to-mobie and communication channes, IEEE Trans. Veh. Techno., vo. 43, no. 4, pp , Nov [7] L. Cheng, B. E. Henty, D. D. Stanci, F. Bai, and P. Mudaige, Mobie vehice-to-vehice narrow-band channe measurement and characterization of the 5.9 GHz dedicated short range communication (DSRC) frequency band, IEEE J. Se. Areas Commun., vo. 25, no. 8, pp , Aug [8] I. Sen and D. W. Matoak, Vehice-vehice channe modes for the 5-GHz band, IEEE Trans. Inte. Transp. Syst., vo.9,no.2,pp , Jun [9] A. G. Zajic and G. L. Stuber, Space-time correated mobie-to-mobie channes: Modeing and simuation, IEEE Trans. Veh. Techno., vo. 57, no. 2, pp , Mar [10] P. Withington, R. Reinhardt, and R. Staney, Preiminary resuts of an utra-wideband (impuse) scanning receiver, in Proc. MILCOM, Nov. 1999, vo. 2, pp [11] Time Domain, Inc. [Onine]. Avaiabe: products/p2101e2evk.pdf [12] J.-Y. Lee, J.-H. Lee, S.-D. Kim, J.-H. Jeong, W.-H. Kim, G.-Y. Ha, C.-S. Jung, J.-E. Oh, D.-W. Ha, S.-J. Kong, and E.-Y. Ahn, UWB propagation measurements in vehicuar environments, in Proc. IEEE RWS, Jan. 2009, pp [13] S. S. Ghassemzadeh, R. Jana, C. W. Rice, W. Turin, and V. Tarokh, Measurement and modeing of an utra-wide bandwidth indoor channe, IEEE Trans. Commun., vo. 52, no. 10, pp , Oct [14] D. Cassioi, M. Z. Win, and A. F. Moisch, The utra-wide bandwidth indoor channe: From statistica mode to simuations, IEEE J. Se. Areas Commun., vo. 20, no. 6, pp , Aug [15] S. S. Ghassemzadeh, L. J. Greenstein, T. Sveinsson, A. Kavčić, and V. Tarokh, UWB deay profie modes for residentia and commercia indoor environments, IEEE Trans. Veh. Techno., vo.54, no.4,pp , Ju [16] L. J. Greenstein, S. S. Ghassemzadeh, S.-C. Hong, and V. Tarokh, Comparison study of UWB indoor channe modes, IEEE Trans. Wireess Commun., vo. 6, no. 1, pp , Jan [17] S. Ichitsubo, T. Furuno, T. Taga, and R. Kawasaki, Mutipath propagation mode for ine-of-sight street microces in urban area, IEEE Trans. Veh. Techno., vo. 49, no. 2, pp , Mar [18] J. Forester, Channe modeing sub-committee report (fina), IEEE, Piscataway, NJ, Tech. Rep. P /368r5-SG3a, IEEE P Working Group for Wireess Persona Area Networks (WPANs), Dec [19] J. M. Cramer, R. A. Schotz, and M. Z. Win, Evauation of an utrawideband propagation channe, IEEE Trans. Antennas Propag., vo. 50, no. 5, pp , May [20] C.-C. Chong and S. K. Yong, A generic statistica-based UWB channe mode for high-rise apartments, IEEE Trans. Antennas Propag., vo. 53, no. 8, pp , Aug [21] B. M. Donan, D. R. McKinstry, and R. M. Buehrer, The UWB indoor channe: Large and sma scae modeing, IEEE Trans. Wireess Commun., vo. 5, no. 10, pp , Oct [22] O. H. Woon and S. Krishnan, Identification of custers in UWB channe modeing, in Proc. IEEE VTC Fa, Sep. 2006, pp [23] J. Chuang, S. Bashir, and D. G. Micheson, Automated identification of custers in UWB channe impuse responses, in Proc. CCECE, Apr. 2007, pp [24] M. Corrigan, A. Waton, W. Niu, and J. Li, Automatic UWB custers identification, in Proc. IEEE RWS, Jan. 2009, pp [25] J. Kareda, S. Wyne, P. Amers, F. Tufvesson, and A. F. Moisch, A measurement-based statistica mode for industria utra-wideband channes, IEEE Trans. Wireess Commun., vo. 6, no. 8, pp , Aug [26] A. Muqaibe, A. Safaai-Jazi, A. Attiya, B. Woerner, and S. Riad, Pathoss and time dispersion parameters for indoor UWB propagation, IEEE Trans. Wireess Commun., vo. 5, no. 3, pp , Mar Joon-Yong Lee (S 99 M 02) was born in Seou, Korea, in He received the B.S. degree in eectrica engineering from Hong-Ik University, Seou, in 1993 and the M.S. and Ph.D. degrees in eectrica engineering from the University of Southern Caifornia (USC), Los Angees, in 1997 and 2002, respectivey. From 1998 to 2002, he was a Research Assistant with the Utra-Wideband Radio Laboratory (Utra- Lab), USC, where he primariy worked on the design of the utrawideband (UWB) ranging system. In the summer of 2000, he was an Intern with Time Domain Inc., Huntsvie, AL, deveoping a UWB precision ocation system. In 2002, he joined the facuty of the Schoo of Computer Science and Eectrica Engineering, Handong University, Pohang, where he is currenty an Associate Professor. He has consuted for the Ministry of Information and Communication, Repubic of Korea. His research interests incude the design of UWB communication and ranging systems and the characterization of UWB propagation.