PROPAGATION MODELS FOR V2V COMMUNICATION IN VEHICULAR AD-HOC NETWORKS

ROAGATION MODELS FOR V2V COMMUNICATION IN VEHICULAR AD-HOC NETWORKS 1 of. VAISHALI D. KHAIRNAR, 2 D. KETAN KOTECHA 1 Reseach Schola, Depatment of Compute Science and Engineeing, Nima Univesity, INDIA 2 Diecto, Depatment of Compute Science and Engineeing, Nima Univesity, INDIA E-mail: 1 khaina.vaishali3@gmail.com, 2 dketankotecha@gmail.com ABSTRACT Thee is lot of eseach wok going in the aea of vehicula ad-hoc netwoks (VANETs). Because these systems often compise seveal tens, hundeds o even thousands of vehicle nodes, a eal wold test is vey costly and time consuming opeation. Vehicula Ad-hoc Netwoks (VANETs) eseach is caied out using diffeent mobility and netwok simulatos like MOVE, SUMO, TaNS, NS2 vesion 2.34 etc., because it allows fo fast and cheap evaluation of potocols and applications in a contollable and epoducible manne. Simulation study helps us to use models in ode to make a judgment on eal-wold poblem viability. Models should eflect eality using adio popagation model fo vehicle-to-vehicle communication, hence accuacy is an impotant equiement fo popagation models. Keywods: VANETs, V2V Communication, opagation models, Simulations etc. 1. INTRODUCTION Intelligent Tanspotation System (ITS) applications ae being defined to impove highway oad taffic safety, efficiency and comfot. Many applications ely on communication povided by vehicula ad-hoc netwoks (VNAETs). In vehicleto-vehicle (V2V) communication on highway oad takes place between vehicle nodes which meet by chance. A VANET occus as soon as two o moe vehicle nodes ae within the communication distance/ange. No infastuctue is involved, vehicula ad-hoc netwoks (VANETs) ely heavily on distibuted measues to egulate access to wieless channel. otocols fo andom access, TDMA (Time Division Multiple Access) and flooding ae implemented and evaluated in simulatos. How well such a potocol will fae once deployed in a ealwold test bed may diffe geatly fom the simulation esults [1], as the simulato may be ovely optimistic [2]. Reality povides oppotunities fo two vehicle nodes to exchange infomation which would not have been possible in simulato due to a simplistic popagation model [3]. Radio popagation model also has a stong impact on the pefomance of a potocol [4] because the popagation model detemines the numbe of vehicle nodes within one collision domain, an impotant input fo contention and intefeence. This has a diect effect on a vehicle node s ability to tansmit a packet to anothe node, which can esult in diffeent values fo metics such as thoughput, dopped packets, medium load and latency. The mobility often involved in vehicula ad-hoc netwoks (VANETs) cause s vehicle nodes to move in and out of each othe s tansmission ange. Depending on the adio popagation model a vehicle node may shae a collision domain with tens o hundeds of othe vehicle nodes, o with only a handful because the model accounts fo buildings [5]. This pape povides the diffeent popagation models which can be used in vehicula ad-hoc netwoks (VANETs) eseach, specifically in simulation studies. 2. SIMULATION Netwok simulato i.e. NS2 is used in Vehicula Ad-hoc Netwoks eseach often povides a stack of potocols on top of which the potocol o application unde test is implemented. A component with possible connections between vehicle nodes often woks in conjunction with the popagation model in ode to evaluate which vehicle nodes ae affected by a tansmission. Results could be a node coectly eceives a message o eceives gabled bits due to a collision. 686

Mobility simulatos i.e. MOVE, SUMO can be used to move the vehicle nodes on highway oads as in geneally the case in a VANET eithe based on measued o geneated taffic taces [6], an embedded mobility model [7,8] o a coupling with taffic simulation tool [9,10]. A simulation can have two goals:- 1) efom a statistical exploation to gain insight in how a system will wok in a geneic envionment, o 2) efom a site-specific evaluation of a system to gain insight in the opeational popeties in a specific envionment. 1.1 Mobility VANETs ae subset of Mobile Ad-hoc Netwoks (MANETs) with seveal diffeences between them. Mobility is usually constained, because the vehicle nodes follow highway oads accoding to some physical vehicle model. Speed is geneally high in VANETs, but can diffe geatly, e.g. V2V communication between stopped vehicle nodes o vehicle nodes passing in opposite lanes. Vehicle nodes in a vehicula ad-hoc netwok geneally do not have stict weight, size and powe consumption limits. VANET nodes can safely be assumed to have access to cetain peipheals such as positioning and navigation hadwae. Anothe impotant diffeence is a vehicle may easily tavel outside an aea coveed by a cetain legislatue. Vehicles fom multiple vendos will need to be able to coopeate; such standadization is an impotant accepts which is geneally not consideed when evaluating a MANET application. 1.2 opagation Envionment The wieless channel is a highly chaotic and unpedictable system [3]. It s a way fom tansmitte to eceive a signal is being eflected, scatteed and absobed by the objects in the popagation envionment. As such its magnitude is alteed, but due to multiple paths it can also intefee with itself o with signals sent in othe fequency anges. With context of VANET s comes also a typical adio wave popagation envionment. Vehicles geneally move on oads, but othe sceney can vay fom open famlands to foests to lage uban canyons and bidges. VANET popagation envionment is the pesence of lage metal objects which ae continuously changing position in the envionment, namely the vehicles themselves, such envionment is highly dynamic. 1.2.1 Lage-scale effects Lage-scale effects on adio wave popagation ae the following thee phenomena:- 1) Reflection: It occus when a wave encountes a lage suface with cetain optical popeties. In models eflection is often tanslated to a path loss exponent, such as the 2 in (2) and 4 in Eq. (3) 2) Diffaction: This phenomenon is explained by Huygens inciple, which states that evey point on a wave font acts as the seed fo a seconday wave font. This enables waves to popagate aound edges o though holes. This can be modeled with the knife-edge diffaction model [11], which can be used fo site-specific modeling of popagation ove mountains and lage buildings. 3) Scatteing: A adio wave scattes when it encountes an object which is small compaed to the wavelength, speading the waves in all diections. This can account fo a eceived signal which is stonge than would have been pedicted by eflection and diffaction alone. 1.2.2 Small-scale effects Small-scale effects on adio wave popagation ae often efeed to as fading. At the eceive multiple vesions of the oiginal signal aive; they can be eflected and diffacted and aive with time and phase diffeence. These multipath waves intefee with each othe, which can cause lage fluctuations in signal quality with appaently small changes in time o eceive location. This elative motion causes fequency modulation because each multipath will have a diffeent Dopple Shift; the esulting fequency change is deived as follows: v f λ cosθ d = (1) Hee v is the elative velocity, λ the wavelength and θ the angle between the signal path and the diection of movement. 1.3 Channel aametes Mobile channel can be chaacteized with channel paametes. The eception of multipath components can be seen as a sample which can be expessed by means of statistical quantities. Delay Spead is the standad deviation of the aival times. Dopple Spead measues the spectal boadcasting caused by elative motion of tansmitte and eceive. 1.4 Radio Technologies Many communication technologies ae used in VANETs, such as infaed [12] and shot ange adio. Shot ange adio technologies used is Wi-Fi, 687

but some eseach has done in 900MHz band [13] and in the millimete ange (60-78Hz) [14]. VANET eseach conveges to IEEE 802.11p [15], a Wi-Fi used fo communication in the vehicula envionment pat of the Wieless Access in Vehicula Envionments (WAVE) standad [16], [17]. IEEE 802.11p builds upon the poven and matue 802.11 standads, poviding elatively cheap but poweful and communication devices. It povides low latency access to the medium - nodes do not fist have to associate and authenticate with base stations - and is optimized fo the ad hoc domain. IEEE 802.11p opeates on 7 channels in the 5.8-5.9GHz band (as shown in Figue.1) [18] and is expected to have a maximum communication ange in the ode of 1km. 1.6 Implementation In Simulatos Implementation of popagation model in a simulato usually takes the following steps, illustated in Figue.2: 1) Fo evey node n within a elevant distance, pefom a calculation of the eceived signal stength. The eceived signal stength is calculated using a popagation model. 2) Fo a tansmission instance (e.g. the tansmission of message x) all signal stengths fom concuent tansmissions othe than x eceived at node n ae added as noise. 3) Based on the Signal-to-Intefeence and Noise Ratio (SINR) and Bit Eo Rate (BER) a decision is made whethe the message is coectly eceived o has bit eos. If the SINR is below a cetain theshold it is impossible to detect the signal in the eceived noise, and a collision has occued. Most popagation models in simulatos conside nodes to be stationay fo the duation of one tansmission. 3. ROAGATION MODELS Figue.1:WAVE Channel Assignments A node listens to the Contol Channel (CCH) at least a cetain amount of time. On the CCH announcements fo sevices can be tansmitted, these sevices can then be povided on the Sevice Channels (SCH). The WAVE standad does not define if one adio should listen to channels in time slots o if multiple adios can be used to obseve seveal channels simultaneously. The channel access is defined in IEEE 1609.4 [19]. So fa, most ITS-elated VANET eseach focuses on applications opeating on a single channel as if in isolation. Figue.2: Geneic Model To Evaluate Reception 1.5 Signal aametes The fequency at which a adio technology opeates geatly impacts its popagation popeties. Besides its caie fequency, othe metics ae the tansmitted powe, the bandwidth and the symbol time, these ae esults of the modulation scheme, a combination of signal and channel paametes can lead to diffeent kinds of fading. This fading is often chaacteized by a pobability distibution and appopiate paametic assumptions [20]. The popagation envionment in the simulato is used to judge the effects of popagation of electomagnetic waves though the medium, usually this medium is ai. In its most abstact fom, this defines success o failue of eception of a message fo a cetain node. opagation models can be classified in lage scale and fading o small-scale models. Fom an implementation point of view they can be eithe deteministic o pobabilistic. 3.1 Deteministic Models A deteministic model allows computing the eceived signal stength, based on actual popeties of the envionment such as the distance between tansmittes T and a eceive R. These models ange fom simple to vey complex whee they also account fo multipath popagation in the envionment modeled exactly as the aea of deployment. 3.1.1 Fee space model Which is sometimes also efeed to as Fiis model, afte its invento [21]? It models a single. Unobstucted communication path [20]. The eceived powe depends only on the tansmitted powe, the antenna gain and the distance between the sende and the eceive, as shown in Figue. 3.a). As a adio wave tavels away fom an (Omnidiectional) antenna, the powe deceases with the squae of the distance. 688

2 G t tgλ ( d) = (2) 2 α (4π ) d L Whee t is the tansmitted powe, G t and G ae the gains of the tansmitte and eceive antenna gains and λ is the wavelength. α is the path loss exponent and is 2 in Fee Space. L is the system loss. Often, G t, G and L ae set to 1 fom a topology point-of-view; this model egads the nodes as floating in fee space. 3.1.2 Two-ay gound model The two-ay gound model also accounts fo a eflection via the gound, given the dielectic popeties of the eath in addition to the diect line of sight (LOS). Nodes ae positioned on a plane as depicted in Figue.3.b). This model gives moe accuate pedictions at longe ange than the Fee Space model [11] and is given as follows: 2 2 G t tght h ( d) = (3) 4 d L Whee h t and h ae the heights (in metes) of the tansmit and eceive antennas espectively. Eq. (3) shows a faste powe loss than (2), but does not give good esults fo shot distances because of oscillation caused by the constuctive and destuctive combination of the two sepaate paths. Eithe (2) o (3) ae used based on the magnitude of d, the T-R sepaation. 3.1.3 Ray tacing model Ray tacing is a technique often used to pedict popagation fo cellula systems. Modeling the popagation envionment plays a citical ole in the development, planning and deployment of, fo instance, UMTS/IMT2000 cellula systems [22]. Because fo these systems not only coveage but also bandwidth is an impotant issue, caeful site planning is in ode. Ray tacing models can take into account the exact position, oientation and electical popeties of individual buildings in the envionment in which the system is to function. Using the ules fo eflection, diffaction and scatteing all ays emanating fom the souce taveling towads a eceive can be modeled, as shown in Figue.3.c). As a esult, a complex impulse esponse h(t) can be calculated as the sum of all contibutions [23]: N h( t) = = A δ ( t τ )exp( υ ) (4) n 1 n n j n The eceived signal h(t) has N time-delayed impulses (ays), each of which is an attenuated and phase-shifted vesion of the oiginal tansmitted signal. Amplitude A n, aival time T n and phase υ ae calculated fo each ay using Snell's laws, the unifom geometical theoy of diffaction (UTD) and Maxwell's equations. All objects in the envionment need to be modeled with chaacteistics such as pemittivity, conductivity and thickness. This method also allows to use antenna adiation pattens. Basically, ay tacing models ae computed using 3-D vecto mathematics. Evaluating evey ay individually fo a fixed antenna position is feasible, as it is used in cell planning. In VANET multiple tansmittes and multiple eceives ae moving in a continuously changing envionment and h(t) will need to be ecomputed upon a change in the envionment. Ray tacing popagation models ae not often used in VANET [24]. Figue.4: obabilistic opagation. 3.2 obabilistic Models obabilistic models allow a moe ealistic modeling of adio wave popagation [3]. A pobabilistic model takes a deteministic model as one as its input paametes in ode to get a mean tansmission ange. Fo evey individual tansmission the eceived powe is then dawn fom a distibution, as shown in Fig. 4. The esult is a moe divese distibution of successful eceptions. It can happen with a cetain pobability that two nodes close to each othe cannot communicate, although it can also happen with a cetain pobability that two nodes beyond the deteministic tansmission ange can communicate. The distibution of these effects depends on the pobabilistic model and its paametes. n 689

3.2.1 Log-nomal shadowing The Log-Nomal Shadowing model uses a nomal distibution with vaiance σ to distibute eception powe in the logaithmic domain: ( d; σ 2 ) ~ LN( det ( d), σ = (5) Whee det is a deteministic model such as Equation (.2) o (3). As such the eceived powe is given as: d d) = t L( do ) + 10α log X σ d o ( (6) + Hee α is a path loss exponent like the 2 in Equation (2) and the 4 in Equation (3). L( d o) is a efeence path loss measued close to the tansmitte. Equation (6) can be ewitten as L( d ) = ( d ) 10 det o (7) 2 ) Whee the owe Loss facto is defined by: d L( d) = α log10 (11) d o 3.2.3 Longley-ice The Longley-Rice model (o Rice model) [3] models the eception powes following the Rayleigh distibution but additionally takes into account the positive effects of a LOS path with a cetain scale facto k [25]: L( d ) ( d) = ( d ) 10 det o (12) ( d) = ( d) F( d) (13) Rice With L(d) as given in Equation (11) and F(d) defined as a Ricean DF with a nomal distibution: F( d) d 2 2 = c( N( ( d)1, ) + 2k ) + N( ( ), 1) (14) Which gives a eceived powe by multiplying the deteministic eceived powe with a owe Loss scale facto in db? With c defined as 1 2( k + 1). d L( d) = 10α log10 + X σ d (8) o 3.2.2 Rayleigh The Rayleigh popagation model [11] models the situation when thee is no LOS, and only multipath components exist. This model incopoates intensive vaiations in eceived signal powe because multiple paths can eithe combine constuctively o destuctively. The amplitude, delay and phase shift of these components geatly depends on the envionment. Like the Log-Nomal shadowing model in Equation (5), the Rayleigh model also depends on a deteministic model to which a cetain vaiation is applied: Rayleih ( d) ~ Rayleigh( ( d)) (9) det This can be ewitten to ead: L( d) ( d) = ( d ) 10 log( 1 unif(0,1)) (10) det o 3.2.4 Nakagami The Nakagami model is highly geneic. Reception powe follows a gamma distibution: ( d) d; m) ~ Gamma m, m det ( (15) The paamete m specifies the intensity of fading effects. Nakagami includes othe models, such as: ~Rayleigh fo m=1 ~Fee space fo lim, m Yet it is pobabilistic [26]. This model has been poven to eflect cetain envionmental conditions and the consequences on eception powe. 4. SIMULATION RESULTS To find the popagation model that best chaacteize VANETs channel, two kinds of 690

simulations ae conducted, one by using Netwok simulato ns-2.34. 4.1 NS2 simulation The netwok simulato ns2.34 is used, which is a well known simulato in both academic and industial fields in simulating and analyzing VANET s envionment. The simulato has been extended to model VANETs by utilizing the IEEE 802.11p technology. The simulated netwok is mapped as cicula bidiectional highway with a diamete of 2000m (6283m length) with 4 lanes in each diection. Thee ae 600 vehicles on this highway segment and all of them equipped with DSRC and GS technologies. The vehicles speed anges fom 70 to 120Km/h and thei movements follow a micoscopic mobility model whee the instantaneous speed is influenced by font vehicle s speed and has to change lane if it decides to bypass anothe vehicle. Each vehicle is configued to boadcast a status message of size 250Bytes peiodically and all vehicles within its ange ae possible ecipients. All configuation paametes ae listed in Table.1. At the end we compae and analyze the diffeent popagation models based on the packet delivey atio and the time delay in eceiving an emegency message. In the fist simulation scenaio, only one vehicle is boadcasting its status message; all othe vehicles ae potential ecipients. We ae inteested in the successful atio of the eceived messages at diffeent distances fom the tansmitte. Fo the Shadowing popagation model, we used 2.8 as the path loss exponent and 4 as a standad deviation as specified in [27] fo the highway scenaio. Fo the Nakagami popagation model, we used the paametes specified by [28]. Figue.5 shows the packet successful eception ate vesus distance. It is obvious that diffeent popagation models give vey diffeent esults fo the same setup. This means that choosing the popagation model in any simulation setup is a main facto to judge on the validity of the esults. Figue.5: The success atio vs Distance In the second simulation scenaio, we use the same paametes as in the fist scenaio except fo two: the tansmission powe is inceased to 0.002W and all vehicles ae tansmitting thei status messages peiodically. One vehicle is configued to send an emegency safety message to all behind vehicles. We ae inteested in the time till the waning message eaches a distance of 2000m. Figue.6 shows the time delay until the emegency message eaches the intended distance vesus the status messages sending ate (taffic load). It is obvious that the Two-Ray model suffes fom high delay in a high taffic situation since all nodes within the ange ae competing to use the channel. While in the pobabilistic models (Shadowing and Nakagami) not all nodes eceive the signal successfully and so the numbe of nodes competing fo the channel is less. It can be seen also that diffeent popagation models give diffeent esults fo the same scenaio. This is a vey seious issue in VANET especially in an accident situation whee safety messages have to be popagated to all vehicles behind the accident in a shot time. Using a simple model which assumes that all vehicles in the ange eceive the message successfully while in eality they ae not, may esult in fatal consequences. Figue.6 Time delay vs Taffic load 691

Table.1 Value Of aametes Used In Simulation 5. DISCUSSION VANET is mostly modeled as a cluste of vehicle nodes on a highway oad in a simulato. This can be accounted fo by simply using a path loss exponent α 2 in the fee space o two-ay gound model, depending on the envionment and by changing othe paametes such as deviation σ when using a pobabilistic model. When using the Nakagami o Rice model, the stength of a LOS component can be set with the m-paamete o the k-facto espectively. opagation model s used, it still needs to be paameteized coectly. In [30] the Log Nomal Shadowing model was paameteized with α=2.56 and σ =4 wee used, based on eal-wold measuement data. In [26] a ealistic set of paametes is povided fo Nakagami model. Measuements pefomed at 900MHz [13] povided input fo a set of paametes fo the Rice model [31-33]. Model can be paameteized coectly; these paametes ae aveages of eal-wold data-mixing measuements of a highway. Choosing a set of paametes ceates a homogenous popagation envionment inside the simulato. Thee is no VANET simulato which allows fo sectoised popagation models, these scenaios could be simulated sepaately, but boundaies and tansitions fom one aea to anothe may be of inteest. Deteministic models ae often used in VANET eseach. They can geatly incease the untime pefomance of a simulation but it is easoned they descibe eal conditions insufficiently [3]. A pobabilistic model could bette account fo the vaiance in eal wold situations, which enables vastly diffeent communication between two vehicle nodes having the same T-R sepaation. Anothe obsevation is that in VANET simulation, vehicle nodes themselves ae often dimensionless. The vehicle nodes have no influence on adio popagation. It seems easonable though, that in pactice the lage metal bodies of vehicles povide a wide ange of effects on popagation: Vehicle nodes often block LOS between communicating vehicle nodes, making multipath components dominant. Vehicle nodes can function as waveguides o as eflectos, theeby inceasing he tansmission ange beyond what could be expected based on fee space popagation. 6. CONCLUSION AND FUTURE SCOE opagation model used in vehicula ad-hoc netwoks simulation has lage influence on the esults. It impacts which nodes ae able to communicate and the pobability of coect eception ange. It can influence the speed at which messages popagate though the netwok, diectly influencing end-to-end delay in a multi-hop highway oad scenaio. The pobability distibution of coect eception also influences the ovehead with espect to collisions and medium utilization. The eal-wold implementation could behave diffeent fom the simulation, so cae must be taken when mapping model and paametes to the taget envionment. REFRENCES: [1] K. Ramachandan, M. Gutese, R. Onishi, and T. Hikita, Expeimental analysis of boadcast eliability in dense vehicula netwoks,"vehicula Technology Magazine, IEEE, vol. 2, no. 4, Dec. 2007, pp. 26-32. [2] U. M. Colesanti, C. Cociani, and A. Vitaletti, On the accuacy of omnet++ in the wieless senso netwoks domain: Simulation vs. testbed," in E-WASUN '07: oceedings of the 4th ACM wokshop on efomance evaluation of wieless ad hoc, senso,and ubiquitous netwoks. New Yok, NY, USA: ACM, 2007, pp. 25-31. [3] A. Kuntz, F. Schmidt-Eisenloh, O. Gaute, H. Hatenstein, and M. Zittebat, Intoducing pobabilistic adio popagation models in omnet++ mobility famewok and coss validation check with ns-2," in Simutools '08: oceedings of the 1st intenational confeence on Simulation tools and techniques fo 692

communications, netwoks and systems & wok- shops. ICST, 2008, pp. 1-7. [4] I. K. Eltahi, The impact of di_eent adio popagation models fo mobile ad hoc netwoks (manet) in uban aea envionment," Aug. 2007, pp. 30-30. [5] R. Nagel and S. Eichle, Efficient and ealistic mobility and channel modeling fo vanet scenaios using omnet++ and inetfamewok," insimutools '08: oceedings of the 1st intenational confeence on Simulation tools and techniques fo communications, netwoks and systems & wokshops. ICST, 2008, pp. 1-8. [6] V. Naumov, R. Baumann, and T. Goss, An evaluation of inte-vehicle ad hoc netwoks based on ealistic vehicula taces," in MobiHoc '06: oceedings of the 7th ACM intenational symposium on Mobile ad hoc netwoking and computing. New Yok, NY, USA: ACM, 2006, pp. 108-119. [7] R. Ba, Z. J. Haas, and R. van Renesse, Scalable Wieless Ad hoc Netwok Simulation, in Handbook on Theoetical and Algoithmic Aspects of Senso, Ad hoc Wieless, and ee-to- ee Netwoks. CRC ess, 2005. [8] H. Dytkiewicz, S. Soka, V. Handziski, A. Koepke, and H. Kal, A mobility famewok fo omnet++," Jan 2003. [9] A.Wegene, M. i_okowski, M. Raya, H. Hellb uck, S. Fische, and J.-. Hubaux, Taci: an inteface fo coupling oad ta_c and netwok simulatos," in CNS '08: oceedings of the 11 th communications and netwoking simulation symposium. New Yok, NY, USA: ACM, 2008, pp. 155-163. [10] M. i_okowski, M. Raya, A. L. Lugo,. apadimitatos, M. Gossglause, and J.-. Hubaux, Tans: ealistic joint ta_c and netwok simulato fo vanets," SIGMOBILE Mob. Comput. Commun. Rev., vol. 12, no. 1, pp. 31-33, 2008. [11] T. S. Rappapot, Wieless Communications: inciples and actice. Communications Engineeing and Emeging Technologies. entice Hall, 2001. [12]. Fenandes and U. Nunes, Vehicle communications: A shot suvey," July 2007. [13] J. S. Davis, Vehicle to vehicle f popagation measuements", Jan 1996. [14] T. Ohyama, S. Nakabayashi, Y. Shiaki, and K. Tokuda, A study of eal-time and autonomous decentalized DSRC system fo inte-vehicle communications," Intelligent Tanspotation Systems, 2000. oceedings, pp. 190-195. [15] IEEE daft amendment to standad fo infomation technology - telecommunications and infomation exchange between systems - local and metopolitan netwoks - speci_c equiements - pat 11: Wieless LAN medium access contol (MAC) and physical laye (HY) speci_cations: Amendment:Wieless access in vehicula envionments," IEEE 802.11p Daft 2.0, Novembe 2006. [16] T. Weil, Wieless access in vehicula envionments (WAVE) - the emeging IEEE 1609 standad," Nov 2007. [17] IEEE 1609 - family of standads fo wieless access in vehicula envionments (WAVE)," http://www.standads.its.dot.gov/fact sheet.asp?f=80, Jan 2006. [18] D. Jiang, Vikas, A. Meie, W. Holfelde, and R. Hetwich, Design of 5.9 GHz DSRCbased vehicula safety communication," Wieless Communications, IEEE [see also IEEE esonal Communications], vol. 13, no. 5, Octobe 2006, pp. 36-43. [19] IEEE tial-use standad fo wieless access in vehicula envionments (wave) - multi-channel opeation," IEEE Std 1609.4-2006, Novembe 2006. [20] K. Laasonen, Radio popagation modeling," Sept 2003. [21] H. T. Fiis, A note on a simple tansmission fomula," o- ceedings of IRE, vol. 34, May 1946, pp. 254-256. [22] G. E. Athanasiadou, A. R. Nix, and J.. McGeehan, A micocellula ay-tacing popagation model and evaluation of its naow-band and wide-band pedictions," Selected Aeas in Communications, IEEE Jounal on, vol. 18, no. 3, Ma 2000, pp. 322-33. [23] G. Tuin, F. Clapp, T. Johnston, S. Fine, and D. Lavy, A statistical model of uban multipath popagation," Vehicula Technology, IEEE Tansactions on, vol. 21, no. 1, Feb 1972, pp. 1-9. [24] S. Mose, F. Kagl, and A. Kelle, Inteactive ealistic simulation of wieless netwoks," Sept. 2007, pp. 161-166. [25] A. Kuntz and F. Schmidt-Eisenloh, Senso netwok extensions fo the omnet++ mobility famewok," http://www.tm.uka.de/sne4omf. 693

[26] V. Taliwal, D. Jiang, H. Mangold, C. Chen, and R. Sengupta, Empiical detemination of channel chaacteistics fo dsc vehicle-tovehicle communication," in VANET '04: oceed- ings of the 1st ACM intenational wokshop on Vehicula ad hoc netwoks. New Yok, NY, USA: ACM, 2004, pp. 88-88. [27] Netwok Simulato 2 (ns2), Discete Event Netwok Simulato, http://nsnam.isi.edu/nsnam/index.php/main age, Apil 2009 [28] M. Toent-Moeno, S. Cooy, F. Schmidt- Eisenloh, and H. Hatenstein, IEEE 802.11- based one-hop boadcast communications: undestanding tansmission success and failue unde diffeent adio popagation envion ments, oceedings of the 9th ACM intenational symposium on Modeling analysis and simulation of wieless and moile systems, 2006, pp- 68-77. [29] Discete Event Netwok Simulato, http://nsnam.isi.edu/nsnam/index.php/main age, Apil 2009. [30] F. Kanadi, Z. H. Mo, and K. chan Lan, Rapid geneation of ealistic mobility models fo vanet,"mach 2007, pp. 2506-2511. [31] J. S. Davis, Statistical chaacteization of ician multipath e_ects in a mobile-to-mobile communication channel", Jan 1996. [32] B.N. Singh, Bhim Singh, Ambish Chanda, and Kamal Al-Haddad, Digital Implementation of an Advanced Static VAR Compensato fo Voltage ofile Impovement, owe Facto Coection and Balancing of Unbalanced Reactive Loads, Electic owe Enegy Reseach, Vol. 54, No. 2, 2000, pp. 101-111. [33] J.B. Ekanayake and N. Jenkins, A Thee-Level Advanced Static VAR Compensato, IEEE Tansactions on owe Systems, Vol. 11, No. 1, Januay 1996, pp. 540-545. 694

FIGURE.3: DETERMINISTIC ROAGATION: A) FREE SACE, B)TWO-RAY GROUND, C)RAY TRACING. 695