Fang Li, Wei Chen, Yishui Shui, Lida Xu, Junyi Yu,＊, Changzhen Li, Kun Yang, Fuxing Chang, Yi Liu

1 School of Automation, Wuhan University of Technology, Wuhan 430070, China

2 Faculty of Information Technology and Electrical Engineering, Norwegian University of Science and Technology, Trondheim 7491, Norway

3 School of Information Engineering, Wuhan University of Technology, Wuhan 430070, China

4 Information Technology & Decision Sciences, Old Dominion University, Norfolk, VA 23529, USA

5 Super Radio AS, Oslo 1208, Norway

I. INTRODUCTION

Armed with the new vehicle networking and autopilot technology, wireless communication technology for V2V has been widely concerned recently due to its huge potentials to improve road safety and avoid a traffic jam,for instance, vehicles can exchange information through the Internet of Vehicles to make decision previously [1]. Thus, that is one of the important technologies in Intelligent Transportation System (ITS). Federal Communications Commission allocated 5.9 GHz for the ITS to protect public safety in 1999 [2]. And the ministry of Industry and Information Technology of China had also designated 5.905-5.925 GHz as long-term evolution (LTE)-vehicle to everything (V2X) test band for ITS [3].

Previous V2X measurement campaigns had been carried out in multiple scenarios, which mainly in Europe and America [4]. Paper [5]conducted measurements both on highways and in cities at 5.12 GHz. Paper [6] gave the results of channel measurement campaigns under multiple environments (highways, tun-nels, bridges, street crossings etc.) with vehicles driving in the same direction at 5.6 GHz.There were also some measurement campaigns carried on 5.9 GHz in recent years. Paper [7]showed a GPS-enabled sounding system to measure V2X channels. Besides, wide band(75 MHz) V2V channel measurements were conducted at 5.9 GHz [8]. To model the path loss at 5.9 GHz, on-road vehicular testbed was used to measure the outdoor V2V propagation channel in highway and rural environments[9]. As an important parameter in the signal energy field of the Ricean distribution, the K-factor has been widely studied in the recent years. Ricean K-factor analysis in wide band indoor channels was provided in [10]. As for outdoor measurements, [11] studied the K-factor of V2V radio channels of the street crossings in the open suburban area. [12] have modeled K-factor for High-Speed Railway(HSR) radio channels. [13] studied K-factors in detailed under a U-shape cutting environment for HSR scenario, while paper [14]focused on Rician K-Factor of the millimeter wave. For delay and Doppler spreads, paper[15] presented multi-path propagation measurements under suburban, highway and rural areas. Considering effects of a traffic jam on channel radio, paper [6] studied traffic congested scenario at 5.6 GHz, where the delay spread and Doppler spread were mainly analyzed. However, as a relevant factor, analysis of small-scale amplitude fading characteristics was not provided. Worth mentioning is the fast velocities (4.2 m/s to 8.3 m/s) in the paper [6].From our investigation, we leamed that the average speed of 400 cities in China is generally limited to 6.89 m/s. And the density of pedestrians, vehicles and street buildings are quite different from the environment in Europe.

Most of the measurements afore-mentioned above are carried out at different locations, the discontinuity in measurements would impact the accuracy of radio channel characteristics analysis under different traffic states. To fill these gaps, this study conducted a measurement campaign at 5.9 GHz in a typical big city- Wuhan, China. The measurements were carried out on the same expressway in the same region but different lanes of opposite directions. All of the measurement campaigns were under the normal drive-by scenario, and the traffic state was the way as it used to be. With these experimental setups, measurements were expected to provide a more convincing way to study the propagation channel characteristics.Furthermore, the low-velocity in the congested state ranged from 0.05 m/s to 3.76 m/s in this work, which is a common congested speed in big Chinese cities. In specific, we emphasized on the difference channel properties between two common status on city expressway: congestion and non-congestion.

In addition, power delay profile (PDP),small-scale amplitude fading characteristics,delay spread and Doppler shift have also been estimated. Since Chinese cities have similar urban layout and traffic conditions, the obtained parameters in Wuhan is capable of providing valuable references to the other cities in China.

The remainder of this study is organized as follows: In section II, measurement campaigns have been described in detail, both of them were undertaken on the same avenue but in different situations. Section III presents the differences of channel properties between these two traffic states. Finally, we draw the conclusions in section IV.

This paper focused on the differences of 5.9 GHz V2V channel characteristics between two traffic states - congestion and non-congestion on the same road under same environment in Wuhan, China.

II. MEASUREMENT CAMPAIGN

2.1 Measurement setup

Fig. 1. TDM channel sounder and the measurement vehicles. (a) is the TDM sounder; (b) is the TX part; (c) is the RX part.

The V2V channel measurement campaigns were conducted by using the TDM Channel Sounder provided by Super Radio AS, which was developed by Norwegian University of Science and Technology (NTNU) (see figure1.(a)). The sounder emitted chirp signals with the center frequency of 5.9 GHz. The frequency bandwidth was 100 MHz with 10 ns delay resolution. For transmitter (TX) part and receiver (RX) part, we employed two standard compact cars to carry them during these measurements. The red Fiat car (Figure1.(b)) loaded TX part and RX part was in black Volkswagen (Figure1. (c)). Two Omnidirectional antennas with 2 dBi gain were mounted on the roof of vehicles (shown in figure1). The height of TX equaled to 1.57 m, whereas the height of RX was 1.50 m. In order to get the frequency and time synchronization of the two parts, two global positioning system (GPS)were fixed beside the antennas on the top of the cars. The GPS also recorded the velocities and positions in real-time. A 12V uninterruptible power supply supported power for each part. And the transmitted power was roughly 0.4 W. In our measurement, the RX part received 1933 chirps per second, and the chirp interval was 517.24 μs. Its TX power was set at 16 dBm.

All of the measurement data and position information were recorded by laptops in real-time. We have collected about 255 seconds of trace data in the measurements. The raw data that we got from the channel sounder is frequency domain signalsH(f,t). Then, Inverse Fourier Transform would be used to obtain complex channel impulse responses (CIRs,h(t,τ)). For details, parameters of measurements had been summarized in TAB.I.

2.2 Measurement description

To study vehicular channel characteristics in the same environment continuously, we chose the city expressway from Xiongchu road to the Jiedaokou as the test site in Wuhan city. The satellite map of the two measurements was shown in figure2. The expressway from south to north was the main way to the downtown,and it was always in a traffic jam all day long,while traffic in the opposite direction was quite smooth. Along the expressway side there located dense buildings, trees, and road signs.And the measurement procedure has been recorded by both video recorders and GPS on the car.

Congested scenario:Firstly, two vehicles went through the right lanes from south to north (see figure2) and all of the two tests cars were stuck in a traffic jam with a low speed around 2 m/s.

Non-congested scenario:After passing the traffic jam, the two test cars turned around at the front traffic circle immediately, and then they went in the opposite direction on the left side lanes, which was non-congested (shown in figure2). The lanes were separated by a 0.5 meters high concrete wall.

Moreover, TAB.II showed the speeds of two cars, which had always maintained car-following status. When the test cars were stuck in a traffic jam, the speeds were very small - an average speed of the two cars was all below 2m/s (see TAB.II), while the minimum distance between the two vehicles was 8.59m, the maximum distance was 20.43m(shown in TAB.III). In the non-congested situation, the minimum distance between the two vehicles was 26.46m, the maximum distance was 34.23m, and fluctuations were kept very small (TAB.III), the average speeds of two cars were around 10 m/s.

Consequently, the effects of congestion and non-congestion on the radio channel in the same region but under different situations could be studied.

III. COMPARISON OF RADIO CHANNEL CHARACTERISTICS

According to Wide-Sense Stationary and Uncorrelated Scatterers (WSSUS) assumption,we view the channel as a steady state within 10-wavelength sliding window. And in this work, our radio channel characteristics analyses are based on the WSSUS assumption of[16, 21], including the power delay profile,small-scale amplitude fading distribution, delay spread and Doppler frequency shift etc.

3.1 Power delay pro file

PDP is a time-delay function is used to describe the received signal power between (τ,τ+dτ)[22]

whereh(t,τ) in Equation 1 denotes CIR. On the basis of PDP, the average power delay pro file(APDP) is calculated by Equation 2 [11].

Fig. 2. Google map of the second ring road in Wuhan, China.

Table I. Measurement parameters.

Table II. Speed information of measurements

Table III. Distance between tx and rx

Figure3 shows the PDP from 20 s to 109 s under the situation of congestion. A strong path which is roughly parallel to the X axis is due to the small change of the relative distance between the two vehicles (shown in TAB. III).The powerful reflection phenomenon marked at 40s and reflections which are approximately parallel to the strong path are derived from cars nearby as well as buildings or facilities beside the road.

However, for the non-congested condition, many approximately parallel reflection paths (shown in figure4) are approaching the strong path gradually. From 15 s to 20 s, the excess delays of these paths decrease about 290 nanoseconds, while the TX and RX move 50.10 m. Therefore, their reflectors are fixed.In addition, on the basis of Delay-doppler spectrum (Figure10 (b)) in Section III Part-D,the positive of these reflection paths prove that these reflectors are located at the front of the moving TX and RX.

To be specific, according to the measured data, we find out that the reflection paths in PDP present a big difference between the two tests, i.e., most of the reflection paths are roughly parallel to the strong path under congested state, while the reflection paths are gradient and closing to the strong path gradually in non-congested state. In other words, PDP of congested state is relatively more stable than that in noncongested scenario. These phenomena are mainly caused by different movement status.

Fig. 3. PDP in the congested situation.

Fig. 4. PDP in the non-congested situation.

3.2 Small-scale amplitude fading characteristics

on the basis of the same sliding window(10-wavelength), small-scale amplitude fading characteristics will be studied in this section.Lognormal, Rayleigh, Ricean, Nakagami-m and Weibull distributions are commonly used to describe small-scale amplitude fading distribution [17] [18]. In order to acquire optimal distribution, we employed Akaike’s Information Criteria (AIC) algorithm [21]

where,nin the sample setxnvaries from 1 toNrepresents the probability density function(PDF) of thei-th distribution, andθiis the distribution parameter vector of the measured data.is calculated from the maximum likelihood estimate ofθi.

Akaike weightwi[21] represents the matching degree between measured data and the models to be chosen. Bigger values of the Akaike weight means better fits. The sum ofwiis 1.

where,Iis the number of distribution candidate. And the AIC differences Φiequals to

Fig. 5. Small-scale amplitude fading distribution determination and the results of K-factors estimation. (a) AIC determination result for congested environment; (b) AIC determination result for non-congested scenario; (c) K-factors estimation result for congested condition; (d)K-factors estimation outcomes for the case of non-congested condition.

The results of two considered scenarios are shown in figure5. For congested condition, Ricean distribution occupies a dominant position (95.05%), Nakagami-m distribution accounted for 4.40 % and Weibull distribution accounted for 0.55%, while both Lognormal and Rayleigh distribution fitting results are 0.00%. As a result, this scenario can be defi ned as Line-of-sight (LOS) environment (average K-factors equals to 4.03 dB, figure5 (c)).

For non-congested situation, Ricean distribution, Nakagami-m distribution, Weibull distribution, Rayleigh distribution and Lognormal distribution account for 51.4426%, 30.3308%,17.8044%, 0.42%, 0.00%, respectively. The K-factors fluctuate from -8 to 6 dB.

According to the recorded video, a white car blocked in the middle of the two cars occasionally. The Fresnel Zone is then calculated to find whether the white car is a shelter or not. The radius of the Fresnel Zone can be calculated by [16]

where, thed1andd2are shown in figure6 (a),respectively. And the First Fresnel Zone (FFZ)is often used to judge whether the diffraction loss should able to be ignored [18]. Therefore,nis usually equal to 1.

Using geometric information recorded during measurements, we can observe two antennas distance is 30 m, and the three cars are almost the same height, the white car is in the center of two antennas, through calculating the Fresnel radius of 0.62 meters, if the antenna height was lower than 0.62 m the white car would cause an obstructed-line-ofsight (OLOS) phenomena [20]. The height of the antenna is about 0.1m in this test (shown in figure1). Hence 0.1m ＜0.62m, so this white car blocked the V2V signal.

Fig. 6. White car blocked the signal. (a) Analysis of fi rst Fresnel zone;(b) Snapshot taken from TX car.

Fig. 7. K-factors of Ricean distribution, m-factors of Nakagami distribution and shape parameter (β) of Weibull distribution for non-congested scenario.

To study the impact of obstructions on vehicular propagation channel, figure7 shows the K-factor, M-factor and shape parameter (β)respectively. It can be seen that the K-factor of Ricean distribution is generally small and M-factor of Nakagami-m roughly equals to 1,while β factor of Weibull is equal to 2 at most of the time. According to the definition, whenm= 1 Nakagami-m distribution becomes Rayleigh distribution and when β = 2, Weibull distribution will also change to Rayleigh distribution [22]. This can also correspond to the OLOS phenomenon analyzed by FFZ.

We can observe that the values of K-factor, M-factor and β significantly increased infigure7, mainly at the period between 14s and 18s denoting a short LOS phenomenon. The reason for this phenomenon is that the distance between the cars is larger, while the white car blocks into the two testing cars sometimes(seen from figure6 (b)), and the occlusion will cause OLOS phenomenon, when the obstacle is removed, the LOS phenomenon is back.This variation can also consistent with the previous work in Paper [20], which noted that the probability of OLOS condition appearance would become bigger as the distance increases, on the contrary, the probability of LOS phenomenon appearance decreases.

From the analysis above, it is obvious that vehicular channels are more stable in congested state compared with non-congested condition. Transitions between LOS and OLOS happen occasionally under non-congested condition.

The short LOS phenomenon has been marked as Area 1 in figure5 (d), the percentage of Ricean distribution for Area 1 marked in figure5 (b) and figure5 (d) is 87.88 % (Nakagami-m:5.57%; Weibull: 6.57%; Average K-factors equals to 2.40 dB). Therefore, both congested scenario and Area 1 are in the case of LOS condition. In order to guarantee the uniqueness of conditions (LOS condition),we chose 90s of the data for congestion and 5s (Area 1) for non-congestion. The compar-ison of statistical properties for K-factors,delay spreads and Doppler spreads (shown in Section III Part-C) are all based on it. OLOS phenomena appear beside the Area 1 (Average K-factor＜0 dB).

For Ricean channel in congested scenario and Area 1 (non-congested scenario), the statistical characteristics of K-factors (Kfin dB scale) are depicted in figure8 (a) and figure8(b), respectively. The mean value of K-factor(4.03 dB) in congested scenario is 1.63 dB bigger than average K-factor value (2.40 dB)in non-congested scenario. Both K-factor values are matched well with Gaussian distribution. These observed results are consistent with [17, 19].

3.3 Delay spread

Delay spread is a statistical description of delay characteristics for multi-path channel,which is related to environmental factors[21]. A large number of experimental results show that the root mean square (RMS) delay spread will have a direct impact on the error floor caused by the delay dispersion [17] [23].Therefore, it is necessary to study the delay spread in delay domain. Mean delay τmis defined as the normalized fi rst-order moment of APDP (shown in Equation (6)), while RMS delay spread - τrmsis the normalized second-order central moment [16] (shown in Equation(7)).

As shown in figure9 (a) and figure9 (b)in congested situation, CDF of RMS delay spread fits well with lognormal distribution with mean value - 104.34ns. Meanwhile, the CDF of RMS delay spread in non-congested situation also presents a good fit with lognormal distribution with mean value - 96.83 ns.The statistic data of RMS delay spread of two conditions has a small difference - mean value in congested state is 7.51 ns bigger than that in the non-congested state. That means different density and speed of cars have little impact on the statistic characteristics of delay domain.

Fig. 8. Statistical characteristics of K-factor. (a) is the CDF of K-factor in congested condition; (b) displays the CDF of K-factor in non-congested condition

It is worth noting that the two results are all much bigger than the measured results in paper [6](mean value of τrmsfor traf fi c congestion scenario is around 30ns). It is also much bigger than that in Germany [25] (mean value of τrmsfor urban scenario: 47 ns). This is possibly because that the typical scenarios of Chinese city expressway are generally surrounded by dense, tall buildings, to some extent like a street canyon, which is different from the urban environment in Europe. In this case, it makes sense that the τrmsis much bigger in our measurement, which is caused by the richness multipath components (MPCs).

Fig. 9. CDF of RMS delay and RMS Doppler. (a) and (c) is CDF of RMS delay and RMS Doppler in congested state, respectively. (b) and (d)is CDF of RMS delay and RMS Doppler in non-congested state (LOS region, Area 1 in Fig.5), respectively.

Besides, the mean delay (10% - 90%: 50.48- 105.8 ns) of congested condition is much smaller than the outcomes in non-congested condition (10% - 90% mean delay: 189.6 ns- 228.3 ns). That is because the separated distance between TX and RX in non-congested condition is much longer (shown in TAB.III).More detailed results of delay spread can be observed in TAB. IV.

3.4 Doppler shift

Delay-doppler spectrum is widely used for frequency domain analysis [2] [24]. On the basis of Fourier transform,PrB(v,τ) is obtained by formula (8) [16]

Here,denotes scattering function.?rrepresents doppler resolution. In congested scenario,while in non-congested scenario ?r= 25.43 Hz for ?38≤s≤38. Fig.10 (a) depicts the delay-doppler spectrum of congested scenario at the 40th second. Most of the bright points such as strong path and powerful reflection path marked in Fig.10 (a) tend to 0 Hz, which consistent with channel characters in the congested scenario. For non-congested scenario,Fig.10 (b) presents the delay-doppler spectrum at the 20th second.

Doppler spread is an important part for wireless channel characteristics. According to the delay-doppler spectrums, the average Dop-pler frequency shiftvmand the RMS doppler spreadSv[16] are given by

In Equation (9),PrB(v) is the integration of scattering function [16]. The 90 % values ofvmaverage Doppler frequency shift and 90%values ofSvin congested scenario are 11.01 Hz and 51.35 Hz, respectively. The corresponding values for non-congested scenario arevm=0.90 Hz andSv= 125 Hz. More detailed results are summarized in TAB. IV. In addition, the statistical properties of RMS doppler spread in two scenarios are shown in figure10 (c) and figure10 (d). Unlike delay spreads, the statistic data of RMS Doppler spread of two conditions has a bigger difference - mean value in congestion is 48.82 Hz bigger than that in non-congestion.

We observe that lognormal distribution has a good matching level with the CDF of RMS doppler spreads under both congested and non-congested scenarios (LOS region, Area 1 in figure 5). The distribution and parametersof RMS doppler spreads (Lognormal distribution, 10% - 90%: 40.64 Hz - 51.35 Hz) for congestion are similar with the results in [25]for urban environment at 5.9 GHz in Germany(Lognormal distribution, 10% - 90%: 30.4 Hz- 51.1 Hz). As for non-congestion, our measurement results (Lognormal distribution, 10%- 90%: 75.51 Hz - 125 Hz) are much bigger than that in [25].

Table IV. Statistical characteristics

Fig. 10. Delay-doppler spectrum congested and non-congested scenarios. (a) 40th second in congested scenario. This figure is a local enlarged image, X-axis: 1 - 5120 ns, Y-axis: -416 - 416 Hz;(b) 20th second in non-congested scenario. This figure is a local enlarged image,X-axis: 1～25600 ns,Y-axis: -416 Hz- 416 Hz.

Hence, the effect of expressway on radio channel characteristics is quite different from the measurement results of urban at low speed on highway in [25].

Fig. 11. Measured LCR and AFD.(a) Measured LCR of two traffic states. (b) Measured AFD of two traffic states.

Table V. Measured lcr and afd.

3.5 Level crossing rate and average duration of fades

To obtain fast fading characteristics, the wide band signal should be transformed to the equivalent narrow band signal [26].

In (13),h(t) denotes the equivalent narrowband time-variant IRs, andkpresents the delay index (k= 1,2,…,K),where,K= 2560 samples. In this case, level crossing rate (LCR)and average duration of fades (ADF) are all based on this.

LCR helps us to know the fluctuation in narrow band, which represents the occurrence rate that the received signal crosses a preset fading depth level within one second. Increasing LCR leads to a heavier fluctuation along with a bigger bit error rate [18]. ADF expresses the average time that the received signal stays below the certain threshold. The measured LCR and ADF results are shown in figure11 and TAB.V.

As we can see from TAB.V, LCR of non-congested scenario is 91.32 larger than that in congested situation. Reference [22]noted that LCR of Ricean process is related to both LOS component power and MPCs power, the Doppler shift also has a great impact on it. As it always observed in [6], cars that were driving beside our two test cars can’t be taken into consideration as scatterers. In our measurements, antennas were fixed on the top roof of the test cars as the paper [6] did, which is slightly above the other cars besides. Big buses or trucks are relevant to MPCs, but from the video records we found there was no big buses or trucks near the two test cars.

Moreover, based on the analysis in section III Part - C, the differences of RMS delay spread between the two traffic states turned out to be very small (seen TAB.V). Considering our measurements environment were same under two different traffic states, that means the MPCs powers of the two considered situa-tions present almost the same level.

Thus, the difference of LCR is mainly caused by the Doppler shift. From TAB.IV,we can see that RMS doppler spread differs a lot under these two traffic conditions. For the same reason, the ADF of non-congestion shows an obvious faster rising tendency than congestion.

IV. CONCLUSION

This paper focused on the differences of 5.9 GHz V2V channel characteristics between two traffic states - congestion and non-congestion on the same road under same environment in Wuhan, China. The comparisons on PDP,small-scale amplitude fading characteristics,delay spread, Doppler Shift and LCR were analyzed, respectively.

Compared with previous wireless channel measurement works, the measurement of this study was only different in the traffic conditions, i.e., different traffic density and speeds.The measurements were conducted on the same road in the dense urban area, and the test vehicles and the weather conditions were all the same. Therefore, this paper could derive a more convincing result of the impacts on the wireless vehicular channels, which was under different traffic situations.

For PDP, delay of reflection path presents a bigger change in non-congested state, which was caused by the different traffic speed. On the basis of the same sliding window (10 wavelengths), small-scale amplitude fading distribution was determined by using AIC algorithm. The above analysis showed that the possibility of LOS appearing decreased with increasing distance. In delay domain and frequency domain, it is found that different intensity and speed of vehicles contribute little on delay spread, but played an important part in frequency domain. The mean value of RMS delay spread only has a difference of 7.51 ns,while the mean value of RMS Doppler Spread differed by 48.82 Hz. Considering both the environment and delay domain, MPCs relevant with vehicular channel mainly came from surrounding buildings, trees and metal street signs. Thus, the variation of LCR and ADF mainly due to the Doppler shift, i.e., speed was the main factor that influenced LCR and ADF.

Nevertheless, compared with previous works, our continuous measurement kept the environment unchanged, but with different traffic density, which is different from the test launched in European cities. We found the radio channel characteristics were different from the outcomes measured under similar scenarios. Therefore, the propagation channel comparison in the same environment but different traffic state was indeed valuable. The measurement results in this study can support the V2V communication system design in ITS.

ACKNOWLEDGEMENTS

This study has been performed in the Project“MAMIME - Massive MIMO radio communication solutions in maritime scenarios”supported by Norwegian Research Council(No. 256309), and partly supported by an International Cooperation Project: 5G-Channel Measurement and Channel Modeling for Ocean Scenario (No.20172h0046), Hubei college excellent young science and technology innovation team project: Fast Varying Channel Modeling and Analysis (No.T201736), Young Scientists Found of National Natural Science Foundation of China (No.61701356) and Fundamental Research Funds for the Central Universities (No.2017-JL-004). The authors would like to express their sincere thanks to(China Scholarship Council) CSC agency for funding and the Super Radio AS. and Ningbo Allmeas Tech. Co. Ltd. for their support of measurement equipment.

References

[1] Guoqiang M., “5G Green Mobile Communication Networks,”China Communication, vol.14,no. 2, 2017, pp:183-184.

[2] Telematics Industry Application Alliance (TIAA)[Online]. http://www.tiaa .org.cn/default.aspx.

[3] Molisch A, Tufvesson F, Karedal J, et al. “A survey on vehicle-to-vehicle propagation channels,”Wireless Communications2009, 16(6):12-22.

[4] Sen, I, Matolak D W. “Vehicle - Vehicle Channel Models for the 5-GHz Band,”IEEE Transactions on Intelligent Transportation Systems2008,9(2):235-245.

[5] Bernado, Laura, et al. “Delay and Doppler Spreads of Nonstationary Vehicular Channels for Safety-Relevant Scenarios.”IEEE Transactions on Vehicular Technology63.63(2013):82-93.

[6] Tan, Ian, et al. “Measurement and analysis of wireless channel impairments in DSRC vehicular communications.”Communications, 2008.ICC’08. IEEE International Conference on.IEEE,2008.

[7] Nilsson, Mikael G., et al. “Measurement Uncertainty, Channel Simulation, and Disturbance Characterization of an Over-the-Air Multiprobe Setup for Cars at 5.9 GHz.”IEEE Transactions on Industrial Electronics62.12(2015):7859-7869.

[8] Cheng, Lin, et al. “Highway and rural propagation channel modeling for vehicle-to-vehicle communications at 5.9 GHz.”Antennas and Propagation Society International Symposium,2008. AP-S 2008. IEEE.IEEE, 2008.

[9] Park J, Kim M D, Chung H K, et al. “Characteristics of Ricean Kfactor in wideband indoor channels at 3.7 GHz.”International Conference on Signal Processing and Communication Systems. IEEE,2010:1-5.

[10] Bernado L, Zemen T, Karedal J, et al. “Multi-Dimensional K-Factor Analysis for V2V Radio Channels in Open Sub-urban Street Crossings.”Personal Indoor and Mobile Radio Communications (PIMRC). IEEE 2010 IEEE 21st International Symposium on,2010:58 - 63.

[11] Zhou T, Tao C, Liu L, et al. “Ricean K-Factor Measurements and Analysis for Wideband Radio Channels in High-Speed Railway U-Shape Cutting Scenarios.”Vehicular Technology Conference IEEE. IEEE, 2015.

[12] Tariq S A M, Despins C, Affes S, et al. “Experimental Results of Rician K-Factor and Co-Polarization Ratio of 60 GHz Wireless Channel in an Underground Mine Gallery.”International Conference on Ubiquitous Wireless Broadband.IEEE, 2015.

[13] He, Ruisi, et al. “Measurements and Analysis of Propagation Channels in High-Speed Railway Viaducts.”IEEE Transactions on Wireless CommunicationsIEEE.12.2(2013), pp.794-805.

[14] Cheng, Lin, et al. “Multi-path propagation measurements for vehicular networks at 5.9 GHz.”Wireless Communications and Networking Conference, 2008. WCNC 2008. IEEE.IEEE, 2008.

[15] Molisch, Andreas F.Wireless communications.Vol. 34. John Wiley & Sons, 2012.

[16] He, R., et al. “Vehicle-to-Vehicle Radio Channel Characterization in Cross-Road Scenarios,IEEE Transactions on Vehicular Technology(2016):1-1.

[17] Guan, Ke, et al. “Propagation measurements and analysis for train stations of high-speed railway at 930 MHz,IEEE Transactions on Vehicular Technology63.8 (2014): 3499-3516.

[18] Bultitude, Yvo de Jong, and Terhi Rautiainen.“IST-4-027756 WINNER II D1. 1.2 V1. 2 WINNER II Channel Models, 2007.

[19] Abbas, Taimoor, and F. Tufvesson. “Line-of-Sight Obstruction Analysis for Vehicle-to-Vehicle Network Simulations in a Two-Lane Highway Scenario.”International Journal of Antennas &Propagation 2013.4(2013), pp.455-461.

[20] He, Ruisi, et al. “Measurements and analysis of propagation channels in high-speed railway viaducts.”IEEE Transactions on Wireless Communicationsvol.12, no. 2, 2013, pp.794-805.

[21] Patzold, Matthias.Mobile radio channels, Wiley Publishing, 2012

[22] Matolak D W, Sen I, Xiong W, et al. “5 GHz wireless channel characterization for vehicle to vehicle communications.”MILCOM 2005 2005 IEEE Military Communications Conference.IEEE,2005:3016-3022 Vol. 5.

[23] Paier, Alxender, et al. “Car-to-car radio channel measurements at 5 GHz: Pathloss, power-delay pro file, and delay-Doppler spectrum,2007 4th International Symposium on Wireless Communication Systems. IEEE, 2007: 224-228.

[24] Kunisch, Jurgen, and Jorg Pamp. “Wideband car-to-car radio channel measurements and model at 5.9 GHz.”Vehicular Technology Conference, 2008. VTC 2008-Fall.IEEE 68th. IEEE, 2008.

[25] Renaudin, Olivier, et al. “Wideband measurement-based modeling of inter-vehicle channels in the 5 GHz band.”European Conference on Antennas and PropagationIEEE,2013:2881-2885.

[26] Guan, Ke, et al. “Propagation Measurements and Analysis for Train Stations of High-Speed Railway at 930 MHz.”IEEE Transactions on Vehicular Technology63.8(2014):3499-3516.