Validation of a Non-Line-of-Sight Path-Loss Model for V2V Communications at Street Intersections

Transcription

1 Valiation of a Non-Line-of-Sight Path-Loss Moel for V2V Communications at Street Intersections Taimoor Abbas, Anreas Thiel, Thomas Zemen, Christoph F. Mecklenbräuker, an Frerik Tufvesson Department of Electrical an Information Technology, Lun University, Lun, Sween Forschungszentrum Telekommunikation Wien (FTW), Vienna, Austria Delphi Deutschlan GmbH, Ba Salzetfurth, Germany Institute of Telecommunications, Vienna University of Technology, Vienna, Austria taimoor.abbas@eit.lth.se Abstract In this paper a non-line-of-sight (NLOS) path-loss an faing moel evelope for vehicle-to-vehicle (V2V) communication at 5.9 GHz is valiate with inepenent an realistic measurement ata. The reference NLOS moel is claime to be flexible an of low complexity, an incorporates specific geometric aspects in a close-form expression. We valiate the accuracy of the moel with the help of realistic channel measurements performe in selecte street intersections in the city of Lun an Malmö, Sween. The moel fits well, with a few exceptions, to the measurements. Those are in turn mae in ifferent intersections having variable geometries an scattering environments. It is foun that the moel can be mae more general if an intersection epenent parameter, that epens on the property an number of available scatterers in that particular intersection, is inclue in the moel. I. INTRODUCTION Intelligent transportation systems (ITS) have gaine consierable attention in recent years, because they have the potential to enhance location awareness an safety of vehicles. Safetyrelate applications, e.g., cross-traffic assistance an traffic conition warnings, will help to reuce the rate of accients. The iea is to utilize vehicle-to-vehicle (V2V) an vehicle-toinfrastructure (V2I) communications for such applications. A lot of research efforts an fiel trails focusing on V2V safety applications have been one an are still ongoing both in acaemia an the vehicular inustry. There are a number of challenges an use cases that are not yet completely unerstoo an require further investigation. Among those, cross-traffic assistance in urban street intersections is one of the most critical use cases as visual line-of-sight (LOS) is often blocke by builings at the corners of the streets. When the LOS is not available then scattering, i.e., reflection, iffraction an refraction, of raio waves can enable signal reception. There exist a number of V2V measurement an ray-tracing simulation base stuies to characterize the raio channel in urban street intersections [1] [6]. In these stuies it was foun that the geometry of the street intersections, i.e., street with an alignment, structure of builings, an antenna height have great impact on the NLOS reception in an urban intersection at 5.9 GHz frequency. Moreover, it is foun that V2V ray-tracing simulation moels often unerestimate channel parameters like, e.g., elay sprea an Doppler sprea, for certain streets ue to limite geographical information. Thus, a generalize an well valiate NLOS moel is neee for urban intersections. Base on an extensive channel measurement campaign Mangel et al. in [7] have evelope a flexible an generalize NLOS path-loss moel name VirtualSource11p at 5.9 GHz. The channel measurements were performe in many ifferent street intersections in the city of Munich, Germany, which were selecte carefully such that they may represent a majority of urban street intersections. However, it is anticipate that the given VirtualSource11p NLOS path-loss moel can be limite in its valiity to the intersections where the measurements were conucte. Thus, it woul be interesting to valiate the moel using an inepenent measurement ata set an stuy the generalizability of the moel. In this paper, we valiate the afore-mentione NLOS pathloss moel using inepenent an realistic channel measurement ata recore in a measurement campaign calle DRIVEWAY [8]. To the author s best knowlege there has been no such valiation stuies of the NLOS path-loss moel an large-scale faing of [7] using inepenent measurement ata. The DRIVEWAY measurements were conucte in various traffic situations that are of particular interest for safety-relate ITS applications. All test sites are locate in an aroun the cities of Lun an Malmö in southern Sween. The aim of our measurements was to unerstan an moel the properties of V2V raio channels with a realistic measurement setup in realistic traffic conitions. However, in this paper, the measurement ata recore only at urban street intersections is use for the analysis. The rest of the paper is organize as follows; section II summarizes the DRIVEWAY V2V measurement campaign where the measurement setup, scenarios escription an measurement ata evaluation are iscusse. Section III explains very briefly the VirtualSource11p NLOS path-loss moel followe by Section IV in which the moel is valiate using measurement ata. Finally, section V summarizes the whole iscussion an conclues the paper. II. V2V MEASUREMENT CAMPAIGN The measurements were performe using stanar hatchback style cars with roof mounte four-element antenna arrays,

2 RX RX (a) Intersection-1 (N , E ) (b) Intersection-2 (N , E ) RX RX (c) Intersection-3 (N , E ) () Intersection-4 (N , E ) Fig. 1. Google Earth TM [9] aerial image of the four measure street intersections in Lun an Malmö. In each intersection the an RX vehicles were approaching the intersection center while riving along the yellow lines. All the images have same map scale. specifically esigne for V2V communication [10]. During the measurements the channel transfer function H(f, t) was recore using the RUSK Lun channel souner that performs switche-array MIMO measurements [11]. Each transmitter () an receiver (RX) vehicle logge their GPS coorinates, an vieos were taken through the winshiel of the /RX cars. In the post-processing these ata were also use in combination with the measurement ata. The most important measurement parameters are liste in Table I. A. Scenarios During the DRIVEWAY measurement campaign several measurements were recore in many ifferent propagation environments. The selecte measurement scenarios are of particular interest for many safety relate intelligent transport systems (ITS) applications [8]. In this paper we are mainly intereste in NLOS propagation, especially in the intersection scenario. The reference NLOS path-loss moel takes into account the inter-builing istance, i.e., the street withs w t an w r, on the transmitter as well as on the receiver sie, Parameter TABLE I MEASUREMENT PARAMETERS CONFIGURATION Value Center frequency [GHz], f c 5.6 Measurement banwith [MHz], BW 240 Test signal length [µs], τ max 3.2 Snapshot repetition time [µs], t rep Number of snapshots in time, N t or Number of samples in frequency, N f 769 Recoring time [s], t rec 10 or 20 Number of antenna elements 4 Number of RX antenna elements 4 /RX antenna height [m] 1.73 Average plus RX antenna gain [B] 6.7 Approximate system loss [B] 4 istance of the an RX from the intersection center an position on the street relative to sie walls. To be able to analyze the effect of these variations, we have selecte 9 measurements (2 suburban an 7 urban) recore in the four ifferent street intersections of variable street withs an

3 propagation conition, shown in Fig. 1. The spee of the an RX cars were varying between 7 11 m/s (25 40 km/h), epening on the street intersection an yiel signs. Intersection-1: (N , E ) can be escribe as suburban in which the builings are situate to the left, to the right is an open surrouning with some vegetation an a power line. In both measurements the an RX vehicles are approaching the intersection, in 18 m an 20 m wie orthogonal streets, from south-to-north an from westto-east, respectively (see Fig. 1(a)). Intersection-2: (N , E ) can be escribe as urban with wier geometry where multi-story builings are situate at each corner of the intersection. In the measurement the an RX vehicles are approaching the intersection, in 43 m an 23 m wie orthogonal streets, from southwest-to-northeast an from northwest-to-southeast, respectively (see Fig. 1(b)) Intersection-3: (N , E ) can be escribe as urban with relatively narrow geometry where 3 4 story builings are situate at each corner of the intersection. In both measurements the an RX vehicles are approaching the intersection, in 20 m an 24 m wie streets, from westto-east an from south-to-north, respectively. Streets are not perfectly orthogonal. (see Fig. 1(c)). Intersection-4: (N , E ) is the same as intersection-3. The only ifference is that the an RX vehicles are approaching the intersection, in 24 m an 23 m wie, from north-to-south an east-to-west, respectively. (see Fig. 1()). B. Data Evaluation an Results To analyze the measurement ata we first erive the instantaneous time-varying power-elay-profile (PDP) of each measurement using the channel transfer functions at each time instant. These PDPs are average over the time samples that correspon to a movement of 10 wavelengths in orer to remove the effect of small scale faing (for etails see [1]). The temporal variations, impact of scatters as a function of propagation istance, the LOS component an other specular/non-specular components can be visualize in the average-pdp (APDP) plots. APDPs of two selecte measurements taken at the intersection-1 an intersection-2 are shown in Figs. 2(a) an 2(b). From the APDP plot it is evient that the scattering environment is very ifferent in both intersection-1 an 2. Intersection-2 has many MPCs incluing a couple of strong MPCs, ue to wier streets an presence of builings at the corners. Whereas in intersection-1 an open surrouning to one sie result in fewer MPCs with less receive power in the NLOS. This ifference in the receive power ue to the geometry of the intersections can be best appreciate by looking at Fig. 2(c), in which the receive power is plotte as a function of combine istance from to RX via the intersection center ( c ). The negative an positive c inicate that the an RX vehicles are approaching an leaving the intersection center, respectively. The effect of the antenna gain an cable losses have been remove from the measure channel gains. Moreover, the transmit power (P T X ) of 23 Bm is use that is ae in the measure channel gain in orer to calculate the receive power (P RX ). To observe the overall ifference in the receive power ue to the variable geometry an the scattering environment the receive powers of the measurements at each intersection are presente together in Fig. 2(). As a first observation, it can be seen that the smallest power is receive at the Intersection- 1 which is logical as it has an open surrouning an fewer scatterers available that can ai NLOS reception. Moreover, Intersection-2 (wier), 3 an 4 (narrow) have 3 4 story builings at each corner of the intersection with more or less similar traffic conitions, which means similar scattering environments. Ieally, the receive power in a wie intersection shoul be higher than that in a narrow intersection given that the scattering environment is similar. However, this is not always the case, the receive power in a narrow intersection can be higher than that in a wie intersection if the scattering environment in the narrow intersection is rich. This can be seen in Fig. 2() where Intersection-2, which is wier, has lower receive power than Intersection-4. After a etaile investigation of the measurement results an having a close look at the geography of the intersections it is foun that there are a couple of very strong reflecte MPCs in Intersection-4 unlike in Intersection-2. These strong reflections arise from the metallic winow structures of a builing at one corner of the intersection an give rise to the receive power. Such an unexpecte behavior in the power epens not only on the number of scatterers but on their material properties an geographic location. If we wish a moel to be universal or most generalize then it shoul be able to reflect such unexpecte behavior. In the subsequent analysis we woul like to fin out how well the NLOS path-loss moel fits our measurement ata an possible unerlying limitations. It is specifie in the IEEE p stanar [12] that a 10 MHz frequency banwith will be use at 5.9 GHz for V2V communication systems but the DRIVEWAY channel measurements were performe over a 240 MHz banwith. For the measurements this larger banwith was selecte to get a fine resolution in elay an to get better statistics of the channel, which is important for a etaile unerstaning. In the analysis of the measurement ata, no ifference in the receive power is foun when consiering the total 240 MHz of measure banwith or selecte portions of 10 MHz measure banwith. III. REFERENCE NLOS PATH-LOSS MODEL Before the comparison it is important to explain briefly the reference NLOS path-loss moel. Mangel et. al. in [7] has performe an extensive measurement campaign an presente a realistic an a well valiate NLOS path-loss moel at 5.9 GHz frequency name VirtualSource11p. The moel is consiere to be of low complexity, thus, enabling largescale packet level simulations in intersection scenarios. The VirtualSource11p moel is obtaine from a NLOS path-loss moel propose in [13] for cellular communications, which is slightly moifie to correspon well to V2V measurements. The erivation of the VirtualSource11p moel can be foun

4 in [7] where the corresponing NLOS path-loss equation to calculate the receive power at a certain istance is given as follows, P L( r, t, w r, x t, i s ) = C + i s L SU ( ( ) ) EL 10 log E T t 4π r 10 (x tw r) E S λ, if r b + ( ( ) ) EL (1) 10 log E T t 10 (x tw r), if E S r > b 4π 2 r λ b (a) APDP of Intersection-1 where t / r enotes the istance of the /RX to the intersection center, w r is with of the RX street, an x t is the istance of the to the wall, respectively. As the NLOS path-loss at a certain istance can be calculate by the moel, the NLOS receive power P RX () for a certain P T X () can be calculate using the following expression, P RX () = P T X () S Loss +G a P L( r, t, w r, x t, i s ) (2) Intersection 1 Intersection 2 (b) APDP of Intersection-2 LOS: Intersection 2 LOS: Intersection (c) Receive power of Intersection-1 an 2 Intersection 1 Intersection 1 Average Intersection 2 Intersection 2 Average Intersection 3 Intersection 3 Average Intersection 4 Intersection 4 Average 120 () Receive power in all intersections Fig. 2. Evaluation steps: a,b) APDP of 2 selecte measurements at Intersection-1 an 2. The color bar inicates the receive power in Bm, c) Receive power as a function of combine istance of /RX from the intersection center ( c ) for both measurements, ) Receive power as function of c an curves showing their mean power average over every 10 meters for each intersection intersection. where S Loss an G a are the system loss an the combine /RX antenna gain, respectively. For the comparison both of these parameters are assume to be zero as the effect of average antenna gain an cable losses have been remove from the measure channel gains before calculating the receive powers. The NLOS moel parameters an their values are liste in Table II. TABLE II NLOS PATH-LOSS MODEL PARAMETERS Parameters Values Curve shift, C (B) 3.75 Sub-urban loss, L SU (B) 2.94 Urban loss factor, i s 0 Sub-urban loss factor, i s 1 Loss exponent, E L 2.69 Small-scale faing, σ (B) 4.1 Street exponent, E S 0.81 istance exponent,e T Break even istance, b (m) 180 IV. VALIDATION OF THE NLOS MODEL In this section, the NLOS path-loss moel is compare to the channel measurements ata. It is esire to investigate whether the overall behavior of the receive power in NLOS, both for the moel an measurements, changes in accorance to the geometry of the respective intersection. The NLOS P RX (), obtaine from the measurements an moel, as a function of c for Intersections-1 to 4 are presente in Fig. 3(a) to 3(). The P RX () from the moel is calculate by using the measure istances. Thus, the powers from each measurement an their respective powers from the moel are plotte together to fin out the ifference. From Figs. 3(a), 3(b) an 3(c), it can be seen that the measurement ata fit very well to the NLOS moel for Intersection-1, 2 an 3. However, in Intersection-2, the ifference in the moel an measurement starts to grow for

5 Intersection 1: Meas. 1 Intersection 1: Meas. 2 (a) Intersection-1 Intersection 2: Meas. 1 (b) Intersection-2 Intersection 3: Meas. 1 Intersection 3: Meas. 2 (c) Intersection-3 Intersection 4: Meas. 1 Intersection 4: Meas () Intersection-4 Fig. 3. Receive power P RX () as a function of combine istance of /RX from the intersection center ( c ) for Intersection-1 to Intersection-4: measurement an moel. c > 130 m. The reason behin this is the curve street. The street start to ben at an approximate istance of 60 m from the intersection center (see Fig. 1(b)). Due to that many significant MPCs, which are visible at c < 130 m, an start to isappear graually for c > 130, which in turn continuously reuce the total P RX. On the other han, the moel assumes the intersecting streets to be straight an perpenicular to each other, thus we o not see such power egraation in the moel curve in Fig. 3(b) an that is one of the limitations in the moel, urban streets are not always straight an perpenicular. Another, interesting ifference can be foun in Fig. 3(), where the ifference in the measurement ata an the moel is quite large. After taking a closer look at the intersection an oing some visual inspection with the help of vieo recorings it is foun that there are metallic winow structures at the corner of the builings that are situate at the corner of the street intersection right in front of the an RX. The strength of the reflecte MPCs from these metallic winow frames is very large, giving an unexpecte rise to the receive power. From the Fig. 3() we can see that the NLOS moel is unable to capture this unusual behavior in the receive power which is ue to the material properties an the placement of scatterers. Although Mangel et. al. in [7] trie to select most representative street intersection in the city of Munich but still they may not be representative enough for many other cities. This causes another limitation in the moel that the NLOS moel cannot capture unusual behaviors ue to the variety in the scattering environment of street intersections. From this valiation stuy it is foun that it coul be beneficial to a another intersection epenent gain parameter (G ID ) in the moel of (1) to compensate for the loss or gain in the power ue to variations in the scattering as follows, P L( r, t, w r, x t, i s ) = C + i s L SU + G ID ( ( 10 log E T t 10 + ( ( 10 log E T t 10 (x tw r) E S (x tw r) E S 4π r λ 4π 2 r λ b ) EL ) ) EL, if r b ) (3), if r > b To fin the correct values of such a parameter, available measurement ata is not sufficient thus it is left for future investigations. After this investigation a nee of new an larger amount of measurement ata is felt to fin a better an more generalize moel for urban intersections. In the comparison plots in Fig. 3(a) to 3() measure P RX ata points eviate slightly from the moel curves, which is ue to the large-scale faing. Whereas the moel curves are showing a istance epenent mean path-loss only, without effect of any faing. As mentione earlier in the APDP erivation, the effect of small-scale faing has been remove by averaging the APDP but the effect of largescale faing is not remove an thus the measure P RX ata points fluctuate aroun a istance epenent path-loss curves. Therefore, variations in the large scale faing for all intersections with respect to the moel are investigate. The large-scale faing for each intersection is approximate as lognormal an the stanar eviation (σ) of zero mean Gaussian istribute large-scale faing is parameterize in Table III. TABLE III LARGE-SCALE FADING PARAMETER Large-scale faing Intersection-1 (B) 6.26 Intersection-2 (B) 5.5 Intersection-3 (B) 4.16 Intersection-4 (B) 3.2 σ

6 V. SUMMARY AND CONCLUSIONS We present a valiation of the NLOS path-loss moel name VirtualSource11p evelope for vehicle-to-vehicle communications at 5.9 GHz. A large amount of measurement ata collecte at selecte intersections in the city of Munich Germany is use to evelop the moel. In this paper the moel is valiate with the help of an inepenent measurement ata set, recore uring the DRIVEWAY measurement campaign in the city of Lun an Malmö in southern Sween. It is intene to test the moel for its accuracy an flexibility. It is foun that the moel is very flexible an fits well to most of our measurements in ifferent intersections. However, there are some cases in which the moel oes not fit well to the measurements ue to rich scattering in some intersections. The results in this stuy suggest to introuce an intersection epenent gain parameter in the reference NLOS such that it can cope with the varying richness of scattering. REFERENCES [1] J. Kareal, F. Tufvesson, T. Abbas, O. Klemp, A. Paier, L. Bernaó, an A. F. Molisch, Raio channel measurements at street intersections for vehicle-to-vehicle safety applications, in IEEE VTC 71st Vehicular Technology Conference (VTC 2010-spring), May [2] T. Mangel, M. Michl, O. Klemp, an H. Hartenstein, Real-worl measurements of non-line-of-sight reception quality for 5.9 GHz IEEE p at intersections, Communication Technologies for Vehicles, Springer Berlin Heielberg, vol. 6596, pp , [3] M. Schack, J. Nuckelt, R. Geise, L. Thiele, an T. Kürner, Comparison of path loss measurements an preictions at urban crossroas for C2C communications, in 5th European Conference on Antennas an Propagation (EuCAP), April [4] T. Abbas, F. Tufvesson, an J. Kareal, Measurement base shaow faing moel for vehicle-to-vehicle network simulations, ArXiv e-prints, Mar [5] J. Nuckelt, T. Abbas, F. Tufvesson, C. F. Mecklenbräuker, L. Bernao, an T. Kürner, Comparison of ray tracing an channel-souner measurements for vehicular communications, in 2013 IEEE 77th Vehicular Technology Conference: VTC2013-Spring, IEEE, [6] T. Abbas, J. Kreal, an F. Tufvesson, Measurement-Base Analysis: The Effect of Complementary Antennas an Diversity on Vehicle-to- Vehicle Communication, IEEE Antennas an Wireless Propagation Letters, vol. 12, no. 1, pp , [Online]. Available: [7] T. Mangel, O. Klemp, an H. Hartenstein, 5.9 GHz inter-vehicle communication at intersections: a valiate non-line-of-sight path-loss an faing moel, EURASIP Journal on Wireless Communications an Networking, vol. 2011, no. 1, p. 182, [8] A. Paier, L. Bernao, J. Kareal, O. Klemp, an A. Kwoczek, Overview of vehicle-to-vehicle raio channel measurements for collision avoiance applications, in Vehicular Technology Conference (VTC 2010-Spring), 2010 IEEE 71st, 2010, pp [9] Google earth v (2013). [Online]. Available: [Accesse: 2013/08/15] [10] A. Thiel, O. Klemp, A. Paier, L. Bernaó, J. Kareal, an A. Kwoczek, In-situ vehicular antenna integration an esign aspects for vehicle-tovehicle communications, in EUCAP, Apr [11] R. Thoma, D. Hampicke, A. Richter, G. Sommerkorn, A. Schneier, U. Trautwein, an W. Wirnitzer, Ientification of time-variant irectional mobile raio channels, Instrumentation an Measurement, IEEE Transactions on, vol. 49, no. 2, pp , apr [12] IEEE, Part 11: Wireless lan meium access control (MAC) an physical layer (PHY) specifications: Amenment 7: Wireless access in vehicular environment, IEEE p,2010, [13] H. El-Sallabi, Fast path loss preiction by using virtual source technique for urban microcells, in IEEE 51st Vehicular Technology Conference Proceeings (VTC 2000-Spring) Tokyo., vol. 3, 2000, pp vol.3.