Channel Characteristics of High-Speed Railway Station Based on Ray-Tracing Simulation at 5G mmWave Band

In order to satisfy the increasing demand for the higher transmission capacity of “smart station”, millimeter wave (mmWave) technology is expected to play a signiﬁcant role in the high data rate communication system. Based on the ray-tracing simulation technology, this paper would study wireless channel characteristics of the three-dimensional (3D) model of high-speed railway station at the mmWave band. Key parameters such as path loss exponent, shadow fading factor, delay spread, Rician K -factor, angular spread, power angle spectrum, and spatial correlation are extracted and investigated. These channel characteristics are of value for the selection of antenna arrays and even the design of future 5G communication networks in the railway environment.


Introduction
Nowadays, with the convenience and flexibility of highspeed railway (HSR), more and more people prefer to take rail traffic for travel or for work.In order to meet these goals in regard to safety, convenience, and efficiency, the investigations about the fifth-generation (5G) wireless communication system have become a trend [1][2][3][4] in different railway scenes.At present, the first 5G smart railway station has been established at Shanghai Hongqiao Railway Station, which is expected to have the ability to guarantee the 5G network depth coverage in this year.Smart railway station would show the wisdom of railway technology to the public from the perspective of operation and service, including indoor navigation, face identification, and 5G-guided robot.In other words, smart station has been the vital application environment in intelligent transportation systems (ITSs) [5,6] and the analysis of the wireless channel model in the railway system for the 5G communication system is essential.
To meet the demand for seamless high data rate wireless connectivity for railway services, the efficient transmission technologies like massive multiple-input multiple-output (MIMO) is proposed [7] to improve system capacity and data rate.e large-scale fading characteristics of different railway scenarios have been summarized for the first time in [8].Based on [8], it is noted that railway station [9,10] is significantly different from common public network scenarios or other high-speed railway scenarios (such as viaducts, cuttings, tunnels, etc.) [11].In [11], the authors also explored channel characteristics in different HSR scenarios by the 3D ray tracing (RT) [12][13][14] which could compensate for the shortcoming of traditional models and accurately analyze the special channel characteristics in 5G systems.Previous simulations focused on the single-input singleoutput (SISO) system and MIMO channel for indoor and outdoor communication [15][16][17], but they missed the research of multipath parameters such as spatial correlation in the railway system.As a result, there is still an urgent demand for a complete investigation by considering large-scale parameters, small-scale parameters, and spatial parameters [18,19] for station scenario in MIMO systems.
is paper mainly focuses on the medium HSR station, which is the most common station scenario with two platforms and four lines.First of all, channel parameters such as path loss exponent, shadow fading factor, multipath delay, and angular spread can be obtained based on RT. e experiment will be carried out at the HSR station where base station antennas are arranged on the tower.On the contrary, the strengths of the massive MIMO system to compensate for the high path loss of the mmWave band have made high carrier frequencies become an integral component of the upcoming 5G networks.Because of the available larger transmission bandwidths, the 5G network is able to supply high data rates and enhance the quality of experience.Although propagation signals at higher frequencies have the higher path loss and are more susceptible to rapid signal degradations caused by moving or fixed obstacles [20], higher frequencies can allow the deployment of smaller antennas which enables the function of multiantenna technology to be integrated into the antenna array.In this paper, not only have the channel characteristics that can guide 3D channel modeling in 5G environment been discussed but optimized deployment recommendations have also been provided for antenna configuration based on the spatial correlation in massive MIMO systems at 37.0-42.5GHz mmWave band.e rest of the paper is organized as follows.e defined high-speed railway station scenario and simulation parameters are presented in Section 2. e wireless channel model for the station scenario is provided in Section 3.According to simulation results, we point out the main channel characteristics to discuss the antenna deployment in Section 4. Conclusions are drawn in Section 5.

Ray-Tracing Simulation
2.1.Scenario Modeling.As shown in Figure 1, the 3D scenario model with a 450 m (length) high-speed railway station is established by SketchUp in the three-dimensional Cartesian coordinate system.ere are 6 types of objects and 6 materials (Table 1) in the scenario model.e LOS paths, scattering, and up to 2nd-order reflections are considered, and the initialized dielectric parameters are listed in Table 2.
en, referring to the CRH380A, the high-speed train model is built in Figure 2.With simulation and measurement results, the effect of adjacent train compartments almost can be ignored for the received signal, which indicates the front of the train could be considered individually in the simulation environment.

Ray-Tracing Simulation Parameters.
In Figure 1, the height of the single-tube tower next to the high-speed railway station is 40 m [21].e distance from the center line of the tower to the edge of station platform is 10 m, and the      e omnidirectional vertical polarization antenna is employed in this work.As the transmitter (Tx) antenna, the base station antenna is deployed on a platform of tower [21] whose height is 38.5 m and vertical distance to the center line of the tower is 0.8 m.Hence, the coordinates can be set to (30.5, 225, 38.5).
In the top view of Figure 1, simulating on Line 1, the process of passing the train through the station on Line 1 is divided into Area I where the train is approaching the tower and Area II where the train is leaving the tower.e top antenna is the receiver (Rx) which is located at the middle line on the top of train, and the vertical distance to the rear section of the train is 13 m.e Rx antenna moves along the positive direction of Y-axis, and the height is 4.16 m which is the sum of the distance to the top of train (0.15 m), the height of the train (3.70 m), and the height from the bottom of the train to the ground (0.31 m). e Y-axis coordinate of Rx antenna is 0 : 2:450 on Line 1. e key configurations of ray-tracing simulation are listed in Table 3.

Wireless Channel Model
In signal processing and wireless communication, the channel transfer function H(f) at frequency f is expressed as the coherent sum of different multipath components H n (f) coupling with the polarimetric Tx and Rx antenna radiation patterns (G Tx , G Rx ) [22].
where N denotes the number of multipath, a n and Ψ n are the signal amplitude and phase, respectively, ϕ AoD , ϕ AoA , θ AoD , and θ AoA are the azimuth angle of departure/arrival and elevation angle of departure/arrival, respectively [23], all of which are for the n-th multipath component.In addition, in the wireless channel model, the line-ofsight (LOS) paths and the non-line-of-sight (NLOS) paths are usually separated because the LOS paths are obvious, whereas the NLOS paths are based on the simulation transmission modes of the RT platform including reflection, scattering, and diffraction.

Wireless Channel Characteristics
In this section, the simulation results of high-speed railway station on Line 1 are presented and analyzed.Furthermore, suggestions on optimized antenna array deployment are also provided.

Path Loss and Shadow
Fading.Path loss refers to the loss caused by the diffusion of electromagnetic wave energy when the wave propagates.From the macroscopic point of view, it is reflected in the function of the received signal power changing with distance.
By analyzing the channel data obtained by the RT simulation platform, with the change of Y-axis coordinate of the Rx antenna, the path loss is shown in Figure 3.
In Figure 3, the path loss is almost symmetrical with respect to the Y-axis coordinate of Tx, whose maximum is 132.2 dB when the Y-axis coordinate is 22 m.In [24], assuming that the receiving sensitivity of the Rx antenna is − 125.23 dBm and combining the parameters of Table 3, the maximum allowable of path loss is about 148.73 dB considering penetration loss.
In this paper, the fitting results of path loss, by the model in (2) that combines the path loss and shadow fading [25], are shown in Figure 4: where A is the interception, d is the distance between the Tx antenna and the Rx antenna (unit: m), and n is the path loss exponent.X σ indicates the zero mean Gaussian random variable with a standard deviation σ which is the shadow fading factor reflecting the large-scale fading that is caused by obstacle occlusion on the wireless channel.From Section 2, Line 1 is divided into Area I and Area II.e fitted parameter values are shown in Table 4.According to Figure 4 and Table 4, on Line 1, as a result of a few obstacles (such as canopy) in the scenario, the fitting path loss exponent (n � 2.26) is slightly larger than that in the free space path loss (FSPL) model (n � 2).In Area I, n is 2.69 because of more obstacles.It should be noted that n is 1.83 in Area II on account of the superposition of the LOS path, more reflection, and scattering which can be found in [11] regarding rich multipath propagation at mmWave channels.In addition, all shadow fading factors σ in the scenario are close to 6 dB which is the value of the LOS environment for rural macro (RMa) scenario in 3GPP [26].

RMS Delay Spread.
Figure 5 shows the mean root mean square (RMS) delay spreads and cumulative distribution functions (CDFs) of RMS delay spreads.International Journal of Antennas and Propagation In Figure 5(a), the RMS delay spread curve is almost symmetric.
e RMS delay spreads near Tx antenna are larger than the two ends of the station.
e mean RMS delay spread 0.28 ns on Line 1 is less than the measurement result of delay spread 0.8 ns in rural [4] because the semiclosed station space limits the range of multipaths [11].Simultaneously, due to the high-speed railway station is a semilimited type, the rays are occluded and reflected by various buildings and the train body during the propagation process.LOS paths provide the primary energy, and the RMS delay spreads are relatively small.In Figure 5

Rician K-Factor.
From the analysis of previous chapters, there are LOS paths and a large number of NLOS paths on Line 1.So it is indispensable to analyze the relationship between LOS path and NLOS path with the Rician K-factor [27].
With the Y-axis coordinate position of Rx antenna changing, the Rician K-factor is shown in Figure 6(a).Additionally, Figures 6(b)-6(d) demonstrate CDFs and fitting results of Rician K-factor.e mean values of the Rician K-factor are summarized in Table 5.
As shown in Figure 6(a), with the Rx antenna close to Tx antenna (Area I), the overall process of Rician K-factor is a decreasing trend.Rician K-factors near Tx antenna are the smallest, whose fluctuation is relatively serious.Conversely, in Area I and Area II, the Rician K-factors become larger with the distance between Tx antenna and Rx antenna increasing, which denotes the fading is getting weaker.
In Table 5, the mean of Rician K-factor on Line 1 is 28.72 dB that is far greater than 0 dB, which means LOS path contributed the main energy.
e similar observation at mmWave band could be found in rural scenario [4] regarding the mean value 25.7 dB.As shown in Figures 6(b)-6(d), it should also be noted from the fitting results that Rician K-factors obey the Gaussian distribution owing to the wideband channel characteristic of the 5G communication system.Furthermore, when the bandwidth is wider, the multipath resolution capability that leads to a decrease in the power of the NLOS component of the main path gets stronger, so Rician K-factors become greater in the 5G wideband channel system.In Figure 7, the ESA is relatively larger, and most (≥80%) values are greater than 10.7 °, while the 80% of ASA are less than 7.8 °, which explains there are more scatterers from the elevation domain at the Rx and the train body plays a major role.en, the ESD is relatively smaller, and most ( ≥80%) values are less than 1.7 °whereas the ASD is relatively larger and 80% are greater than 5.5 °, which explains there are less scatterers at the elevation domain at the Tx.Besides, the measurement scenario in [28] is similar to the station scenario in the paper, and the results in [28] agree with the  International Journal of Antennas and Propagation simulation results in the paper.Since there is a similar environment that more reflectors and scatterers in both scenarios, the scenarios have the same characteristic that the maximum of angular spread does not exceed 70 °.

Power Angle Spectrum.
In Figure 7(a), the angular spreads at two ends of the station are relatively small.at is why we need to focus on the power angle spectrum (PAS) [29] at two ends of the station.In this section, Figure 8 shows the PAS of AoA and AoD when the Y-axis coordinates of the Rx antenna are 6 m and 444 m, respectively.In Figure 8(a), at the Y-axis coordinate 6 m, the main beam at Rx could be horizontal because values of normalized power are relatively the largest as the elevation angle of AoA is about 90 °.Fortunately, elevation angle directions of the main beam are also distributed around 90 °at the Y-axis coordinate 444 m.Furthermore, the relatively large ESA and horizontal beam at Rx show that the linear Rx antenna array should be perpendicular to the top of the train in order to achieve the low correlation among beams.
In Figure 8(b), at the 6 m, while the azimuth angle of AoD is about − 20 °, the normalized powers are the largest, which implies that the azimuth angle direction of the main beam at Tx could be about − 20 °.Conversely, at the 444 m, the azimuth angle directions of main beam are mainly distributed around 20 °, but the elevation angles always approach 100 °. en, the ASDs of Tx antenna are greater than ESDs in Figure 7(a), which means the antenna array

Spatial Correlation.
As we all know, the diversity gain and multiplexing gain of the MIMO system [30,31] are directly related to the spatial characteristics of the channel.e above is the investigation of some spatial characteristic parameters at 5G mmWave band.In addition, as the important aspect of MIMO channels [32], the spatial correlation among beams must be discussed to achieve the suitable antenna spacing.
In this section, in order to improve the performance of the location where there is worser signal coverage, the angular spreads at that position are selected for correlation calculation to ensure the user experience of the overall scenario.
According to  International Journal of Antennas and Propagation e wavelength at 40 GHz is approximately 7.5 mm.From Figure 9(a), the relative distance of Rx antenna array elements is at least 68 times to ensure that correlation coefficient between adjacent antenna elements is less than 0.1, but the antenna element spacing is around 51 cm that is too large to meet the engineering requirements.If the correlation coefficient can be relaxed to 0.5, it indicates the antenna element spacing only will be about 2.25∼3.75cm that might meet the requirement for application at the Rx.
In Figure 9(b), when the correlation coefficient of adjacent Tx antenna elements is less than 0.5, antenna spacing can be set to about 3∼30 times of wavelength which means around 2.25∼22.50cm.

Conclusion
In this paper, channel characteristics of 5G mmWave band have been investigated at the high-speed railway station.Based on the measurement and simulation results, key channel parameters, such as path loss, delay spread, Rician K-factor, angular spread, PAS, and spatial correlation, are all analyzed.Some important results and suggestions are summarized as follows: (1) When Tx antennas are installed on the tower, the maximum path loss on Line 1 is about 132.2 dB, which is so large that multiantenna technology is expected to ensure signal coverage strength and increase system capacity.(2) e high-speed railway station is a semilimited space where the values of RMS delay spreads are almost less than 1.2 ns on Line 1, which is reasonable and acceptable in 5G systems.(3) In the scenario, the Rician K-factors obey the Gaussian distribution which is owing to the large bandwidth in 5G systems.(4) When the Tx and Rx both adopt uniform linear antenna arrays, the antenna spacing could be 2.25∼22.50cm at Tx, whereas the Rx antenna spacing might be set to 2.25∼3.75cm.(5) e Rx antenna array should be vertical to the top of the train, while the horizontal Tx antenna array should be rotated counterclockwise or clockwise horizontally to obtain low correlation due to the propagation direction of the main beam.( 6) By analyzing the train's moving speed and the angle between the train's moving direction and incident wave direction, it can be found that the angle is approaching 90 °as the Rx reaches the Tx, which makes Doppler shift decline.Due to simulation on Line 1 where the train needs to stop at the platform and the speed is relatively low, this paper does not focus on the Doppler effect and more related research could be conducted in future work.

8
International Journal of Antennas and Propagation 1 : tr a in m o v in g d ir e c ti o n

Figure 1 :
Figure 1: High-speed railway station scenario.Line 1 is divided into two parts: Area I (the area of Rx approaching the Tx) and Area II (the area of Rx leaving the Tx).

2
International Journal of Antennas and Propagation distance to the X-axis is 225 m.
(b), all values of RMS delay spread on Line 1 are less than 1.2 ns, which mean most of the powerful multipath components are concentrated around LOS path in the time delay domain.e mean values of RMS delay spread are shown in Table5, and values of two areas are almost the same which

4. 4 .
Spatial Parameters 4.4.1.RMS Angular Spread.On Line 1, the RMS angular spreads of angle of arrival (AoA) and angle of departure (AoD) are shown in Figure 7(a).ASA, ESA, ASD, and ESD are angular spreads of the azimuth angle of arrival, the elevation angle of arrival, the azimuth angle of departure, and the elevation angle of departure, respectively.e CDFs of angular spreads are illustrated in Figure 7(b).

Table 5 :
Mean value of parameters.