Measurements and Characterization for the Vehicle-to-Infrastructure Channel in Urban and Highway Scenarios at 5.92GHz

Yuanyuan Fan,Yi Feng,Liu Liu,*,Shuoshuo Dong,Zhaoyang Su,Jiahui Qiu,Xiaobo Lin

1 School of Electronic and Information Engineering,Beijing Jiaotong University,Beijing 100044,China

2 China United Network Communications Group Co.,Ltd.,Beijing 100033,China

Abstract:In the future smart transportation system,reliable vehicle-to-infrastructure(V2I)communication is very important to ensure vehicle driving safety and to improve vehicle driving efficiency.In this paper,V2I channel measurements at 5.92 GHz are conducted in typical urban and highway scenarios.The frequency and bandwidth of transmission,as well as the deployment of the RSU(roadside unit)and the OBU(on board unit),are selected by considering the recommendation proposed by 3GPP TR 36.885.Then,based on the measured data,the key channel characteristic parameters of the V2I channel are extracted,including path loss,root-mean-square delay spread,stationarity distance,and Doppler spread,etc.Also,the statistical characteristics of the parameters,including time-varying and Doppler characteristics,are investigated and characterized.The work in this paper helps researchers design technology and communication systems in similar scenarios.

Keywords:vehicle-to-infrastructure;channel characterization;measurement;urban scenario;highway scenario

I.INTRODUCTION

Internet of Vehicles(IoV)refers to the realization of a comprehensive network connection of vehicle to everything(V2X)with the help of a new generation of information and communication technology.It can improve the intelligent level and autonomous driving ability of vehicles,thus improving traffic efficiency,building new formats of vehicles and transportation services,and providing intelligent,comfortable,and efficient comprehensive services for users.V2X communication technology is used to realize information sharing between vehicles and the outside world and promote the evolution of IoV to the direction of intelligence and cloud[1,2].According to different business models,V2X communication is divided into the following four categories:vehicle-to-vehicle(V2V),vehicleto-infrastructure(V2I),vehicle-to-person(V2P),and vehicle-to-internet(V2N)[3].

In recent years,the diversified application of 5G mobile communication technology in vertical industries has become a hot topic of inquiry,one of which is the IoV.The introduction of 5G new technology[4,5]has improved the information processing,data storage and computing capabilities of IOV communication[6],which can further reduce road congestion and accidents,and improve the efficiency of traffic infrastructure.

At present,the main technologies for V2X communication in the world include dedicated short-range communication(DSRC)technology based on IEEE 802.11P standard and cellular-V2X(C-V2X)technology based on 3GPP standard[7].From an industrial perspective,the widespread deployment of DSRC requires significant investment in network infrastructure[8].3GPP designed C-V2X to address the shortcomings of DSRC in testing and industrial applications,and completed the formulation of the first stage standard(3GPP Release 14)in 2016[9].3GPP series standards have more excellent system designs and better long-distance transmission performance,with the ability of subsequent evolution.The current C-V2X standard is V2X based on the LTE system,which is called LTE-V2X(Long Term Evolution-Vehicle to X).In 2018,3GPP proposed version R15,which is enhanced over version R14,known as LTE-eV2X[10].LTE-V2X and its evolution(5G NR-V2X)could meet the needs of IoV for the next 10 years[11,12].In the same year,the Ministry of Industry and Information Technology of China allocated dedicated band 5905 5925 MHz to LTE-V2X technology.In 2019,the Federal Communications Commission(FCC)of the United States allocated 20 MHz dedicated spectrum in the 5.9 GHz frequency band to C-V2X technology,which marked the official abandonment of DSRC in the United States.

A reliable and realistic channel model serves as the enabling foundation for communication system design and performance evaluation in IoV.So far,the V2V channel has been extensively studied in[13–17],of which deterministic and statistical models have been proposed.In contrast,the research on V2I channel characteristics is far less than that on V2V channel characteristics.This gap in research may be explained by the similarity between V2I channels and cellular communication channels where the transmitter is fixed and only the receiver is moving.However,the propagation characteristics of the V2I channel are quite different from those of cellular channels[18],especially in terms of the carrier frequency,communication range,antenna height,etc.Therefore,the establishment of the V2I channel model is of great importance to fill the research gaps mentioned above,as well as to forecast the coverage of V2I communication and analyze the interference.

The existing V2I model mainly focus on statistical models and geometry-based deterministic models(GBDM)(e.g.ray tracing modeling approach).The statistical model is built by analyzing the measured data,so the channel measurement is necessary.In Europe and America,most of the research on measurements of V2I channel were based on IEEE 802.11p standard[19,20].Although references[21,22]all conducted channel measurement based on 3GPP standard,the Doppler characteristics of V2I channel were not analyzed.Authors in[23]measured V2I channel in urban,suburban and highway scenarios,respectively.But they only analyzed the small-scale fading characteristics without estimating path loss and shadow fading.Authors in[24]studied the large-scale fading characteristics and small-scale fading characteristics of different V2I communication links,but the transmitting frequency and bandwidth are 2.4 GHz and 2 MHz respectively,which are lower than the required frequency carriers and bandwidth for actual deployment.A GBDM model of V2I channel is proposed in[25],which can accurately predict the intensity of the received signal according to the variation of the interval between receiver(Rx)and transmitter(Tx)antennas at different antenna heights.However,it is too computationally heavy to be used in large-scale simulation of vehicle self-organization networks.In addition,some scholars have used the ray tracing method to simulate the V2I channel in the 5.9GHz[26],millimeter wave[27,28]and terahertz[29]bands,respectively,and analyze its channel characteristics.The ray tracing method needs to build an accurate environmental database and has high requirements for computing power.To sum up,there is no relevant work that not only conducts V2I channel measurement based on 3GPP standard,but also analyzes the large-scale fading(LSF)characteristics and smallscale fading(SSF)characteristics of V2I channel.

In this paper,V2I channel measurements at 5.92 GHz are carried out in typical urban and highway scenarios in Shanghai,China.The frequency,bandwidth,power of the transmitted signal as well as the deployment of the Tx and Rx,are selected based on the 3GPP TR 36.885[9].Path loss(PL),shadow fading(SF),power delay profile(PDP),root mean square(RMS)delay spread,K-factor,Doppler power spectral density(DPSD),RMS Doppler spread and stationary distance(SD)are extracted in this paper.The statistical characteristics of channel parameters are analyzed and modeled.The analysis and parameters presented in this paper will help researchers understand the channel of propagation in order to design technical and communication systems in similar scenarios.

The rest of this paper is organized as follows.Section II introduces the V2I channel measurement scenarios,measurement systems and data processing methods.Key channel parameters are analyzed and modeled in Section III.Finally,Section IV draws the conclusions.

II.MEASUREMENT CAMPAIGN

2.1 Measurement Scenarios

The measurements are conducted in typical urban and highway scenarios in Shanghai,China,as shown in Figure 1.The measurement location of the urban terrain is on Yunjuan Road in Pudong New District,Shanghai.The length of this road is about 440 m,and the vehicle traveled for 50 seconds.In the urban terrain,the Tx antenna was set up near the traffic lights at the intersection,and the Tx antenna height was about 4.6 m.The measurement location of the highway scene is in Shanghai Intelligent Connected Vehicle Comprehensive Test Demonstration Zone.The length of this road is about 500 m,and the vehicle traveled for 60 seconds.In the scene of highway,the Tx antenna is set up to the roadside,and the height of the Tx antenna is about 5 m.In both scenarios,the vehicle at Rx accelerates from a stationary state to a speed at 9 m/s.During the measurement,the vehicle moves towards Tx and then away from Tx.The Rx antennas are placed at the middle position of the top of the vehicle,with a height of about 2 m from the ground.In addition,both Tx and Rx antennas are oriented perpendicular to the ground.Seen as in Figure 1(a),there are a large number of scatterers in urban,including buildings,street lights,vegetation,vehicles,etc.Compared with the urban environment,the site of the highway is more open,surrounded only by vegetation and street lamps,as shown in Figure 1(b).Because there is no obvious area of occlusion,both scenarios belong to line-of-sight(LOS)transmission.

Figure 1.Measurement scenarios.(a)Urban scenario.(b)Vehicle route of urban scenario.(c)highway scenario.(d)Vehicle route of highway scenario.

2.2 Measurement System

The channel measurement system established in this paper is shown in Figure 2(a),including a vector signal generator,a spectrum analyzer,Tx/Rx antennas and a synchronization unit.A frequency domain channel sounder is used,which includes a Tx with vector signal transmitter R&S SMW200A as the core component and a Rx composed of vector signal analyzer R&S FSW43.A single-input multiple-output(SIMO)mode is used in this measurement,with two antennas at Rx.Each antenna is omnidirectional with the gain of 6 dBi.The distance between the antennas of Rx is 2λ(wavelength),about 10 cm.GPS-tamed rubidium clocks are used at both the Tx and Rx to provide 10 MHz reference clocks.The channel sounder transmits and receives multi-carrier symbols,each one of which contains 2560 subcarriers.There are 256 zero-padding sub-carriers(for sideband protection)in frequency domain,as shown in Figure 2(b).The frequency of the transmitted signal is 5.92 GHz and the effective bandwidth is 10 MHz.In addition,the power of the transmitted signal is 27 dBm,which contains the gain of the power amplifier and antenna.Both the transmitter and receiver configurations are in accordance with LTEV2X standard,which is based on 3GPP TR 36.885[9].The specific measurement parameter configuration is shown in Table 1.

Table 2.Parameters of large-scale fading.

Figure 2.Channel measurement system and transmission signal configuration.

Figure 3.Large-scale fading characteristics of urban scenario.

Table 1.Configuration of the measurement system.

2.3 Data Processing

2.3.1 Extraction of Channel Impulse Response After the measurement is completed,the channel impulse response(CIR)extraction is the most important step,which determines the accuracy and reliability of the channel characteristic parameters.CIR contains large-scale fading information and small-scale fading information of the channel,from which many important channel parameters such as path loss,delay spread,and Rice K factor can be extracted.

The original CIR can be obtained by sliding correlation between the baseband signal and the transmitted OFDM symbol,where the baseband signal is obtained by down-conversion and low-pass filtering of the measured intermediate frequency(IF)signal.In order to combat the problem of CIR energy leakage,the original CIR in this paper needs to be processed by coarse synchronization,fine synchronization and adding window function.First,rough synchronization is performed to find the beginning of an OFDM frame.Then,fine synchronization is carried out to correct the phase deviation.Next,channel estimation is performed by the least squares(LS)method,and the channel frequency response(CFR)can be obtained by using the received signal and the transmitted signal.Finally,the channel frequency response is added with Hanning window[30]and the CIRh(t,τ)is obtained by inverse fast Fourier transform(IFFT)of CFR,which is difined as

wheretandτare time and delay,Lis the number of distinguishable multipaths,al(t)is the amplitude of path,which is determined by the large-scale fading and the small-scale fading,fD(t)is Doppler frequency shift andτl(t)is the delay ofl-th path.After extracting the CIR,the effective multipath components(MPCs)needs to be identified.

2.3.2 MPCs Identification

In this paper,effective MPCs identification includes two steps: denoising and multipath search.Generally speaking,the measured CIR contains effective MPCs and ineffective noise components.In order to improve the estimation accuracy of channel characteristic parameters,it is necessary to set an appropriate threshold value to distinguish MPCs and noise components.For the noise threshold,if a fixed constant value is used,the accuracy of the decision will be affected when the SNR(signal-to-noise ratio)is low or the noise fluctuation is large.Therefore,on the basis of the reference[31],this paper dynamically selects the threshold according to the change of noise in the measurement results.

After the noise threshold is determined,the local maximum method is used for multipath search to determine MPCs.Because the signal above the decision threshold is not all MPCs signal.Therefore,the peak scanning and traversal of the signal whose power is higher than the noise threshold are carried out,and each value obtained by searching is the value of each MPCs.Specific judgment rules are as follows:

Step 1:If there is only one peak value in the component region(a group of continuous sampling points),then the peak value is determined as MPCs.

Step 2: If there are multiple peaks,it should be determined by the maximum power difference ΔP.When ΔPis less than ΔPmin(for example,ΔPmin=2dB),the peak value is determined to be the noise component.If ΔPis greater than ΔP,it is judged as MPCs.When all ΔPvalues are less than ΔPmin,one of the peaks is randomly selected as the MPCs.

Through denoising and multipath search,MPCs can be obtained to analyze the large-scale fading characteristics and small-scale fading characteristics of the two scenarios.

III.CHANNEL CHARACTERISTICS

3.1 Path Loss and Shadow Fading

Large-scale fading characteristics are highly significant for the analysis of V2I channel availability,wireless network planning and interference analysis.Since the measured CIR contains both LSF information and SSF information,in order to remove the SSF resulting from multipath and Doppler frequency shift,the received powerPRx(m)at timemis denoted as

whereWLS=20λis recommended in[32,33].When the carrier frequency is 5.92 GHz and the average speed is 9 m/s,WLScorresponds to 1.0 m or 550 snapshots.Based on transmitted power and received power,PL can be calculated.The empirical path loss in decibels is given in logarithmic scale as

wherePLis the path loss,dis the actual distance between the Tx antenna and the Rx antenna,d0represents the reference distance,and in this paper is 10 m.A0represents thePLat the reference distanced0,andnis thePLexponent,which determines how fast the path loss changes with distance.By fitting the measured data with the root mean square error(RMSE)of the log-distance model,A0andncan be obtained.XSFis shadowing factor(SF),which can be obtained by subtracting the interpolated modelPLfrom the re-alistic measurement results.SFis a normal distribution with a mean of 0 and a standard deviation ofXSF.Figure 3 shows the results of statistical analysis of urban and highway scenarios.

Figure 3(a)and Figure 4(a)depict the trend of path loss with logarithmic distance.The shadow fading is shown in Figure 3(b)and Figure 4(b).The parameters of large-scale fading are shown in Table 2.WINNER II B1 model is applicable to urban micro(Umi),while WINNER II D1 model is applicable to rural macro(Rma)[34].Therefore,WINNER II B1 model and WINNER II D1 are compared with urban and highway scenarios,respectively.According to the Figure 3(a)and Figure 4(a),it is not difficult to find that the fitting effect of WINNER II model and measured data is not ideal.Moreover,thePLexponent of WINNER II B1 and WINNER II D1 models is 2.27 and 2.15 respectively,which are lower than thePLexponent of urban and highway measured in this paper,which is 2.50 and 2.20.As the height of the base station in a traditional cellular network is about 10˜35 meters,and there are no obvious scatters around Tx.However,the height of the Tx antenna in V2I is about 4˜6 meters,which is generally set up near traffic lights or street lamps,so the signal will be more affected by scatters around Tx.As a result,thePLexponent of the V2I channel will be greater than that of the cellular network model in the same scenario.In conclusion,the traditional cellular network model is not suitable for describing the large-scale fading characteristics of the V2I channel.

Figure 4.Large-scale fading characteristics of highway scenario.

Figure 5.Power delay profile of urban and highway scenarios.

Figure 6.Number of MPCs for urban and highway scenarios.

Besides,according to[35],the two-ray model is used to fit the measured data,and the correction factor of the two scenarios are-4 dB and 0 dB respectively.Particularly,we find that the fluctuation trend of the urban scene is similar to the two-ray model,that is,the smaller the distance is,the smaller the fluctuation range is,and the larger the distance is,the larger the fluctuation range is,which indicates that vehicles in urban scene receive a large number of scattering components from the ground.The measurement frequency of V2I channel is 5.8 GHz in[36]and the measurement place is a road along the lake in the urban area of Brazil.Its surrounding environment is a resemblance to that of the highway measured in this paper,which is very open,mainly vegetated,without dense buildings.Therefore,the PL exponent is 2.197 measured in[36],which is very close to the 2.20 measured in this paper.By contrast,the PL exponent of the urban scene measured in this paper is larger,because there are more scatterers in the urban and the propagation environment is worse.

3.2 Power Delay Profile

Time dispersion characteristics of wireless channels are usually characterized by power delay profile.In order to reduce the influence of LSF,PDP can be obtained through the statistics of the discretized CIR,which is caculted as

whereWavrepresents the length of the average window,and the value ofWavneeds to make the channel meet the wide sense stationary(WSS)condition[32].In this paper,there isWav=20λ,which is about 1 meter.

The PDP of urban and highway scenarios is shown in Figure 5,which depicts that as the train approaches the transmitter,the transmission delay of the strongest path gradually decreases,while the power increases significantly.The received power of the urban and the highway reaches its maximum at 35 s and 40 s,respectively,which means that the vehicle meets the transmitter.Then,as the train moves away from the transmitter,the delay of the strongest path increases gradually,and the received power decreases gradually.Moreover,the highway scenario changes faster than the urban scenario.Due to the fact that the horizontal and vertical distance from the Rx antenna to the Tx antenna is greater in the highway scenario than that in the urban scenario.In addition,the shadow area around the strongest path in the urban scenario is obviously larger than that of the highway scenario,which indicates that the time dispersion phenomenon is more serious in the urban scenario.According to the content in Section 2.3,the statistical distribution of the number of MPCs at different locations can be obtained,as shown in Figure 6.The X-axis is the distance between the vehicle and the intersection point(IP),and the positive direction of the X-axis is the moving direction of the vehicle.The IP refers to the position of the vehicle when the distance between Tx and Rx is the closest.According to the number of MPCs,the driving Area of the vehicle is divided into three types:Toward Area(TA),Close Area(CA)and Arrival Area(AA).As can be seen from Figure 6,the number of MPCs in highway site is mainly 1-2,while the number of MPCs in urban site is mainly 2-3.

Figure 7.RMS delay spread for different regions.

3.3 RMS Delay Spread

In the process of multipath propagation,the received signals have dispersion in the delay domain due to the different duration of each multipath propagation.Delay spread is usually used to describe the dispersion degree of signal in the time delay domain,which has an important effect on carrier modulation.RMS delay spreadστis generally used to represent the time dispersion characteristics,which is usually expressed as

where E(·)is the symbol for calculating the mean,and E(τ2)and E2(τ)are defined as

P(τk)is the power of thek-th ray with delayτk.According to Eq.(5)-(7),when there is only one effective MPC,στis 0.The RMS delay spread for both scenarios are shown in Table 3.It’s clear that the mean and standard deviation of RMS delay spread in urban and highway scenarios are respectively: 69.21 ns,224.97 ns,36.71 ns and 58.97 ns.The mean and standard deviation of RMS delay spread of WINNER II B1 model and WINNER II D1 model are respectively: 36.31 ns,149.62 ns,15.85 ns and 891.2 ns.Therefore,the existing standard models of cellular networks cannot accurately describe the delay characteristics of V2I channel in urban and highway scenarios.

We use a boxplot to represent the changes of RMS delay spread in three regions of two scenarios,as shown in Figure 7.In the boxplot,the bottom and top edges of each box represent the 25th and 75th percentile,respectively,the horizontal line in the center of the box represents the median,and the circle represents the mean.The dotted line extends to the farthest data point which is not an outlier,and the data points outside the box represented by “-” are the maximum and minimum values respectively(“-” above the box is the maximum and “-” below the box is the minimum).Outliers are drawn separately with a“+”sign.Because the value of individual outlier points is too large,in order to better view the changes in each area,Figure 6 only shows the result of[0,300]ns,which already contains 99.5%of the data.

Figure 7 reveals the straight line connecting the mean value shows a downward trend,which indicates that in each scenario,the mean value of the TA region is the largest,followed by the CA region and the AA region.This is related to the number of MPCs in each region,while TA region has the largest number of MPCs and AA region has the smallest number of MPCs.In each region,the mean RMS delay spread of highway is smaller than that of the urban.Moreover,the minimum RMS delay spread of each region is 0,which indicates that each region contains a case where there is only one path.Only the AA region does not have outliers,due to the fact that the signal in the AA region mainly comes from the LOS path,and there is not much deviation.

The CDF of RMS delay spread is fitted based on Gaussian distribution.The probability density function(PDF)f(x)of the Gaussian distribution[37]is defined as

whereais the coefficient,bis the mean value of the Gaussian distribution,andcis the deviation.Threedifferent gaussian functionsf1(x1),f2(x1)andf3(x1)are used to fit the cumulative distribution function(CDF)of RMS delay spread,that isCDFSD(x)=f1(x1)+f2(x1)+f3(x1).The CDF of RMS delay spread for urban scenes and high-speed scenes calculated in this paper are shown in Figure 8,and Table 4 shows the model parameters when Gaussian fitting is used.

Figure 8.CDF of RMS delay spread.

Figure 9.K factor in urban and highway scenarios.

Table 3.Small-scale fading characteristic parameters.

3.4 K-Factor

In order to reduce the influence of large-scale fading,RMS normalization is required for the original CIRh(m,τ)before calculating K-factor,which is expressed as

wherehss(m,τ)is the CIR after RMS normalization.

Rice K-factor is a vital parameter used to measure the intensity of LOS path and NLOS path.The Rice K-factor can be calculated as

where there isµ2=E(r2)andµ4=E(r4),andris the amplitude of CIR after RMS normalization.

Table 3 shows the calculation results,which demonstrate that the mean value and standard deviation of K factor in urban and highway scenes are 9.44 dB,4.47 dB,15.28 dB,and 5.91 dB,respectively.The mean value and standard deviation of K factor of WINNER II B1 model and WINNER II D1 model are 9 dB,6 dB,7 dB,and 6 dB,respectively.Therefore,the existing standard models of cellular networks cannot accurately describe the K-factor characteristics of V2Ichannel in urban and highway scenarios.Figure 9 shows the variation of K-factor with distance.Generally,the K-factor varies with distance and the influence of other MPCs increases in power domain.However,the fluctuation of K-factor at far distance increases for all the scenarios.As the vehicle moves away from the IP point,some indistinguishable MPCs are judged to be LOS path,which results in an increase in LOS path energy.Therefore,the K factor is larger in the highway scenario when it is far from the IP point.

Figure 10.CDFs of K-Factor.

Table 4.Parameter of RMS delay spread model.

In this paper,two different Gaussian functions are used to fit the CDF of K-factor.Figure 10 indicates that the Gaussian function can better fit the CDF of K factor in urban areas and highway.It is evident that the mean K-factor in urban areas is less than that in highway.Reflection and scattering components in urban scenarios weaken the influence of LOS paths,which makes channel fading more severe.Table 5 shows the model parameters when Gaussian fitting is used.

Table 5.Parameter of RMS delay spread model.

Table 6.Parameter of stationarity distance model.

Figure 11.Doppler power spectral density in urban and highway scenarios.

3.5 Doppler Power Spectral Density and Doppler Spread

On account of the Doppler effect,the channel causes signal dispersion in the frequency domain.The frequency dispersion characteristics of wireless channels are usually characterized by Doppler power spectral density.Different incidence angles produce different Doppler shifts,so the superposition of all MPCs forms the DPSD.The DPSD can be calculated by the discrete fourier transformation(DFT)of the autocorrelation function of the CIR,which is defined as

wheremis time,vis Doppler frequency shift,F {·}is the DFT operation,h(k)is the amplitude of the effective MPCs,(·)*is the conjugate operation,andWFFTis the length of the DFT window.The length of the DFT window is 64 snapshots.

The DPSD of urban and highway scenes is shown in Figure 11.It is obvious that the DPSD of bothscenes has the characteristics of rapid variation,which means it changes rapidly from the maximum positive frequency to the minimum negative frequency.Moreover,the measured maximum Doppler frequency shift is basically consistent with the theoretical calculation results(vmax= 197.3 Hz).In addition,we can find that the DPSD of both urban scene and highway scene shows a certain degree of Doppler spread from Figure 11.

Doppler spread is caused by the superposition of multipath signals with frequency drift phenomenon,which causes the signal spread in the frequency domain.It is usually characterized by RMS doppler spreadσv,which is expressed as

Gamma distribution[38]is used to fit the measured RMS Doppler spread,and the PDF of gamma distribution is defined as

whereαandβrepresent shape parameters and scale parameters of gamma distribution respectively,Γ(·)is the gamma function.

Because the multipath effect causes Doppler spread,the more multipaths there are,the greater the value of Doppler spread will be.The number of multipaths in the highway scene is obviously less than that in the urban scene,so the mean value of Doppler spread in the highway scene is less than that in the urban scene.Figure 12 shows the CDF of the RMS Doppler spread and the fitting results of the gamma distribution.The RMS Doppler spread of the urban scenario(mean value 11.99 Hz,standard deviation 7.82 Hz)is greater than the results of the highway scenario(mean value 5.52 Hz,standard deviation 19.63 Hz).In addition,the gamma distribution can fit the results of urban and highway RMS Doppler spread well.The parameters of Gaussian fitting areα= 25.96 andβ= 0.03 for the urban scenario,andα= 72.62,β= 0.01 for highway scenario.

Figure 12.CDF of RMS doppler spread.

3.6 Stationarity Distance

Mobile channel is widely considered to be nonstationary[39–41],and the degree of non-stationary channel is different in urban and highway scenarios.In this paper,temporal PDP correlation coefficient(TPCC)is used to measure the channel stability.This indicator quantifies the temporal similarity of PDP over time.According to the definition in[42],the TPCC between the PDPs at timetiandtjcan be computed as

The value of TPCCs is normalized from 0 to 1.We calculate the TPCCs between each snapshot and all snapshots,and then form a symmetric matrix.Obviously,the diagonal is going to be 1.

TPCCs for urban and highway scenarios are plotted in Figure 13(a)-(b),respectively.According to the physical significance of TPCC,a high TPCC value indicates that the channel has higher similarity betweentiandtjof the channel is.In other words,if there are many high-value TPCC in a time region,it can be con-sidered that there is no rapid change in the channel during this period.As shown in Figure 13(b),obviously,more high-value TPCC regions can be found in this figure compared with the urban scenario as shown in Figure 13(a).Because of the large number of trees on either side of the highway,the channel changes slowly in the vicinity of these areas.In addition,when the distance between Tx and Rx is closer in the same scenario,the TPCC value is larger,since the PDP received in this condition is very similar.For instance,in the urban scenario,besides higher TPCCs in the 20 s and nearby time,there is also a higher TPCC between the 20 s and 50 s.Since the vehicle reaches IP at 35 s,the 20 s and 50 s are both about 150 meters away from IP.

Figure 13.TPCCs for urban and highway scenarios.(a)TPCCs for the urban scenario.(b)TPCCs for the highway scenario.(c)CDFs of SDs.

Based on the relationship between time,distance and velocity,the stationarity distance(SD)is further calculated.In this paper,according to[15]and[41],0.8 is used as the threshold to estimate SD .In other words,it can be considered that whenC(ti,tj)is greater than the threshold value of 0.8,there is no significant change in the channel between timetiand timetj.Similarly,this paper uses three different Gaussian functions to fit the CDF value of SD,and the fitting results are shown in Figure 13(c).Figure 13(c)indicates that the Gaussian function can better fit the CDF value of SD in urban and highway scenes,while their mean values are 5.89 m and 11.86 m,respectively.Table 6 shows the model parameters when Gaussian fitting is used.

3.7 Spatial Correlation and Spatial Consistency

Spatial correlation describes the correlation of signals received by antennas at different positions.It can be expressed by the pearson correlation coefficientρ1,2[43]of the envelope of the received signal:

where,|h1,i|and|h2,i|represent the amplitude of the received signal envelope of the two antennas,which is of thei-th snapshot.respectively represent the mean amplitudes of all received signals of the two antennas in a time window.A time window in this paper contains 1000 snapshots,approximately 2 meters.According to the calculation in section 3.6,V2I channel remains time stationary in this time window.

Figure 14 illustrates that the correlation coefficient of the two antennas at Rx changes with the movement of vehicles in the urban and highway scenarios,which shows that although the distance between the two antennas is constant,they have dissimilar fading characteristics at different positions.What’s more,the antenna correlation coefficient is maximum when the receiver and transmitter meet.

Figure 14.Antenna correlation coefficient.

Spatial consistency is used to describe the process of continuous and smooth evolution of channel characteristics when Tx or Rx moves,which means that two communication links with close location have similar channel characteristics[44].In other words,spatial consistency means that the channel parameters are correlated within the continuous time window.The autocorrelation of channel parameters can describe the correlation degree of channel parameters between continuous time windows.In this paper,the autocorrelation coefficient of RMS delay spread is calculated to describe the correlation:

where,ρi,jrepresents the autocorrelation coefficient of the i-th and j-th time windows.|σik|is the value ofkth RMS delay spread of thei-th time window.is the mean value of all RMS delay spreads in thei-th time window.Lis the number of snapshots per time window,which is 20 in this paper.

The X-axis represents the distance from the starting point,and the Y-axis represents the autocorrelation coefficient between different time windows and the first time window in figure 15.Figure 15 illustrates that channels at different locations have a certain degree of correlation,and the spatial correlation degree of highway scenes is greater.

Figure 15.Autocorrelation coefficient of RMS delay spread.

IV.CONCLUSION

Based on 3GPP TR 36.885 standard,V2I channel measurements in urban and highway scenarios are carried out in this paper.We describe the characteristics of V2I channel from large-scale fading and smallscale fading by using the measured data.In the aspect of large-scale fading,the characteristics of path loss and shadow fading are analyzed,and a suitable largescale propagation model is provided.In the aspect of small-scale fading,the time nonstationary characteristics of V2I channel are studied by PDP,and different regions are divided according to the statistical results of MPCs.Then,the calculation results of RMS delay extension are given,and the variation of RMS delay spread in different regions is analyzed by boxplot.Next,the relationship between K factor and distance is studied,and the K-factor model is established.In addition,the Doppler fast variation phenomenon mentioned in the standard model is verified by the measured Doppler power spectrum,and the RMS Doppler spread is counted,which provides a reference for the study of the Doppler characteristics of V2I scenes.Finally,the stationary distance is calculated by using TPCC,and the correlation coefficient is used to study the spatial correlation and spatial consistency.

Through comparison and analysis,we find that WINNER II model cannot accurately describe the channel characteristics of V2I,which is the standard model of V2X proposed by 3GPP.In other words,the traditional cellular network is not suitable as the standard channel model for V2I.The analysis of the parameters and characteristics measured in this paper will help researchers understand the propagation channel in order to design technology and communication systems in similar scenarios.In the future,V2I channels based on LTE-V2X or NR-V2X in multiple scenarios deserve further research.

ACKNOWLEDGEMENT

The research was supported by National Natural Science Foundation of China(NSFC)under grant of 61931001.The authors would like to thank the editors and the reviewers for their valuable comments that helped to improve the quality of this paper.