Vehicular Channel in Urban Environments at 23GHz for Flexible Access Common Spectrum Application

With the development of the vehicular network, new radio technologies have been in the spotlight for maximizing the utilization of the limited radio spectrum resource while accommodating the increasing amount of services and applications in the wireless mobile networks. New spectrum policies based on dynamic spectrum access technology such as flexible access common spectrum (FACS) have been adopted by the Korea Communications Commission (KCC). 23GHz bands have been allocated to FACS bands by the KCC, which is expected extensively for vehicular communications. *e comprehensive knowledge on the radio channel is essential to effectively support the design, simulation, and development of such radio technologies. In this paper, the characteristics of 23GHz vehicle-to-infrastructure (V2I) channels are simulated and extracted for the urban environment in Seoul. *e path loss, shadow factor, Ricean K-factor, root-mean-square (RMS) delay spread, and angular spreads are characterized from the calibrated ray-tracing simulation results, and it can help researchers have a better understanding of the propagation channel for designing vehicular radio technologies and a communication system in a similar environment.


Introduction
Recently, flexible access common spectrum (FACS) has been considered to have an important role in accommodating the fast-growing spectrum demands in vehicular communications, which will ramp up the development of mobile and wireless vehicular communication technologies to advance safety and convenience on the roads [1,2]. Among them, vehicle-to-infrastructure (V2I) robust connectivity is the key enabler for enhancing traffic safety, reliability, and efficiency [3,4]. e channel characteristic is a critical role in the design and performance evaluation of V2I connectivity networks. For example, the realistic large-scale fading channel parameters are indispensable for efficient network deployment and optimization; the fidelity small-scale fading channel parameters are crucial in physical layer design, such as optimal modulation, coding, diversity, and protocol scheme development [5].
Most of the previous works on vehicular propagation channels are focused on the 5 GHz band, e.g., [6][7][8][9][10][11][12][13][14][15][16], which use the lower frequency band to increase the V2I and vehicle-to-vehicle (V2V) link ranges. In [6], based on the narrowband channel measurement at 5.2 GHz, the propagation loss (PL) parameters of the vehicular channel are studied under a highway and under urban, suburban, and village environments. In [7], a large number of vehicular narrowband channel measurements are carried out at 5.9 GHz. A dual-slope PL model is proposed for the suburban environment. Meanwhile, it is found that, in such an environment, the small-scale fading of the channel obeys the Nakagami distribution, and the effect of speed and relative distance of the transmitting and receiving vehicles on the Doppler spectrum and quasistationary time is discussed. Based on the wideband multi-input multioutput (MIMO) channel measurement at the 5.3 GHz band, the authors in [8,9] discuss the nonstationary delay spread (DS) and small-scale fading characteristics (following Ricean distribution) in vehicular channels. In [10], under a highway environment, the PL, delay distribution, and Doppler characteristics of the V2I channel are studied according to the MIMO channel measurement. Similarly, the dispersion of multipath components (MPCs) in delay and Doppler domains is analyzed for V2I channels under urban and rural environments, and the geometry of scatterers is also being constructed to illustrate the MPC distribution [11,12]. Moreover, in [13], the authors present a three-dimensional (3D) distribution estimation of MPCs and discuss the power ratio of different propagation mechanisms associated with the actual propagation environment. e work in [14][15][16] describes the MPC dispersion in delay and angle domains for the non-line-of-sight (NLOS) environments and explores MPC distribution and its propagation mechanisms in different propagation environments. All of those important vehicular channel characteristics show that the vehicular channel exhibits severe fading and statistical nonstationarity.
e propagation mechanisms and spatial distributions of MPCs are mainly determined by the vehicular propagation environment.
For the 20-60 GHz band, the studies in [17,18] show that the large spectrum in a higher frequency band provides a possibility of high data rate, but the channel is much different from that in the 5 GHz band, and directional antennas are adopted by transceivers in such vehicular communication systems to overcome the heavy loss at those high frequencies [19,20]. But the obvious limitation by using directional antennas for the higher frequency applications is the difficulty of providing an "always connected" V2I or V2V link. In [21], the two-ray channel model with random reflected paths is used to represent 60 GHz vehicular channel, and the system performance has been theoretically evaluated by varying modulation formats, coding complexity, and propagation characteristics for vehicular communications. e similar works in [22,23] analyze the characteristics of MPC propagation and verify the availability of the vehicular communication system at 60 GHz. Based on channel measurement with rotating directional antennas, the DS, path loss exponent (PLE), and angular spread of arrival (ASA) are compared between line-of-sight (LOS) and NLOS conditions at 60 GHz under outdoor environments in [24,25]. e channel impulse response and scattering functions, as well as the DS, are obtained from 38 GHz to 60 GHz channel measurements in [26]. Such measurements show that the MPC in the channel typically contains LOS, road reflection paths, and reflection paths from the guard rails. However, the two distinguishing features of vehicular channels at higher frequency bands, e.g., directionality and blockage, are not fully explored in the work, which are the significant challenge remaining in the design of such vehicular networks.
An overview of the previous work is to study the typical behavior of vehicle lanes, identifying missing features in currently available channel models to design and evaluate vehicle radio links in a safe and efficient environment.
In this paper, intensive simulations at the 23 GHz band with 1 GHz bandwidth in a realistic V2I urban environment are performed by a calibrated ray-tracing (RT) simulator to complement these missing characteristics, e.g., directionality and blockage. Besides, different road widths, traffic flows, vehicle types, antenna heights, and traffic signs are considered in the RT simulation to fully explore the V2I channel behaviors under different conditions. e time-varying power delay profile (PDP), path loss (PL), shadow factor, DS, Ricean K-factor, and angular spreads (ASs)-azimuth angular spread of arrival (ASA), azimuth angular spread of departure (ASD), elevation angular spread of arrival (ESA), and elevation angular spread of departure (ESD)-are analyzed and extracted from the simulation results. In addition, the propagation mechanism in different environments is analyzed that can illustrate the impact of the actual environment on the channel characteristics. Furthermore, the extracted parameters can be input to the channel generator, like QuaDRiGa or METIS, to generate similar channels, which can be used to evaluate or verify the performance of the system-or link-level design. e rest of this paper is organized as follows: In Section 2, the realistic vehicular environments are reconstructed and introduced. In Section 3, extensive RT simulations are performed in all cases. en, based on the RT results, the vehicular channels are comprehensively characterized in Section 4. Finally, the conclusions are drawn in Section 5.

Realistic Vehicular Moving Network Environments
In order to obtain realistic behavior of vehicular channels in the moving network (MN), the influence of vehicles must be considered in the propagation model [27,28]. In this paper, to map the MN architecture defined by 3GPP TR 37.885 to the real propagation environments, the 3D models of a set of comprehensive vehicular environments are reconstructed.

Urban Environments.
e 3D details of the considered environment are built by OpenStreetMap (OSM). OSM is a collaborative project to create a free editable map of the world, and in addition to street-level map information, it can provide 3D building information. In order to make it easier to access the OSM data, a plugin for SketchUp (3D modeling software) is developed in this work. e plugin can import both street and building information from OSM data. It is the open source software that can be obtained from http:// www.raytracer.cloud/software. As shown in Figure 1, the typical urban area in Seoul is selected.
is urban environment is universal and can represent most of the vehicular propagation environment in the urban area. In addition to OSM, KakaoMap, the Korean map software, is also used to reconstruct the real environment more completely, which has a 360-degree street view, and some landmark buildings such as cafes and restaurants will be specially marked. Combined with OpenStreetMap and KakaoMap, the urban environments in Seoul can be reconstructed by the SketchUp software, which is shown in Figure 2. Considering the impact of different neighborhood environments on the vehicular channel, two typical urban streets in the city are selected as the simulation environments. As shown in Figure 1, the selected streets were marked with red lines on the maps. Figures 2(b) and 2(c) show the 3D environment models of these two areas. e difference between the two areas is the width of the road, characterized by the number of lanes.

Small-Scale Structures.
In addition to the large buildings, the common small-scale structures such as roadside trees, traffic lights, traffic signs, and bus stations are considered in the environment model. ese smallscale structures are on the same order of magnitude as the wavelength, and the most relevant propagation mechanism of them in vehicular environments is scattering [29]. Figure 3 shows the 3D models of small-scale structures in urban environments.  [30], which are defined as follows:  In general, the width of the line is 3.5 m. In Area 1 (shown in Figure 2(b)), the road is with four lanes. Considering the traffic congestion in the actual environments, the speed and the direction of vehicles in each lane are shown in Figure 5.

Vehicle Types and Mobility
e width of the road in Area 2, as shown in Figure 2(c), is more extensive than that in Area 1, and the eight-lane road is designed in Area 2. e vehicle speed on such a road is shown in Figure 6. According to the recommendations of 3GPP TR 37.885, the distribution of vehicle types in the urban environment is determined: (i) 60% Type 1 vehicles and Type 2 vehicles (passenger car) (ii) 20% Type 3 vehicles (bus) (iii) 20% Type 4 vehicles (delivery van) e vehicles in each lane are randomly generated according to vehicle distribution. Vehicles do not change their direction at the intersection.

Antenna Model.
e antennas of the base station (BS) and the user equipment (UE) of the MN system defined are with the same antenna pattern [30], as shown in Figure 7.
e locations and heights of the transmitter (Tx) at the BS and receiver (Rx) at the UE are given in Table 1. e settings follow the recommendations of 3GPP TR 37.885.

RT Simulations in Realistic Vehicular Scenarios
In this section, in order to characterize the 23 GHz vehicular channels, RT simulations with different antenna height setups are developed in different realistic environment cases. First, the calibration and verification of the RT simulator are discussed, and then the RT configuration and different propagation mechanisms caused by the surrounding environment in various environment cases are presented.

RT Simulation Verification.
According to realistic urban 3D maps and design specifications for urban infrastructure, the vehicular environment features, and typical small-scale structures is presented in the above section. Based on which, a large number of statistical consistent environment models can be generated to simulate urban vehicular channels. Before this, RT needs to be calibrated via propagation measurement, and then the intensive close-to-real channel simulations can be conducted with standard-defined configurations.
e validations of the RT simulator against various measurements have been presented in various propagation environments and different frequency bands. For example, RT is calibrated and verified in the tunnel environment at 25 GHz [31], in the urban environment at 28 GHz [28], in the indoor environment at 26 GHz [32], in the viaduct environment at 93 GHz [33], and so on. After the calibration, the corresponding material parameters are obtained. e calibrated materials basically cover all the material composition of the considered propagation environments in this paper, e.g., concrete, brick, granite, marble, glass, and metal. Note that applying appropriate material parameters in RT is a precondition to obtain practical channel results. In general, the material parameters are frequency dependent. e material parameters are calibrated in similar propagation environments at the adjacent frequency band, i.e., 25 GHz and 28 GHz [28,31]. In Rec. ITU-R P.1238-7 [34] and Rec. ITU-R P.2040 [35], such material parameters for urban and indoor environments can be obtained for 1 GHz-100 GHz, indicate an insignificant difference, and are almost independent of the frequency in the range from 20 GHz to 30 GHz.

RT Simulation Configuration.
Considering the vehicular MN in the urban environment, not only the bus equipped with the Rx can move but also other vehicles such as cars, delivery vans, and other buses can move around as the moving scatterers. Aiming at supporting these features, the workflow is shown in Figure 8. Dynamic mobility patterns of moving objects and the Tx/Rx should be defined prior to simulation. After all the models and configurations are uploaded to the RT simulator, the geometry and visibility relationship of the objects are updated for each snapshot. e carrier frequency of V2I simulation is 22.6 GHz with a bandwidth of 1 GHz. Both Tx and Rx beam patterns are the same as the Tx antenna beam pattern of the MHN-E system shown in Figure 7. e deployment of the Tx and Rx is shown in Figure 9; the Tx is placed at the top of the building with a height of 25 m or installed on the pillar of the traffic light with a height of 5 m. e Rx is placed on the top of the bus. e travel distance of the bus where the Rx is located is 500 m. e RT simulation includes two simulation granularity settings: (i) e low-resolution simulation: with the sampling interval 1.38-1.67 m, it is used to get the channel profile and determine the propagation zones where the high-resolution simulations will take place. Different propagation zones indicate different propagation mechanism combinations.
(ii) e high-resolution simulation: for every propagation zone in every case, one CIR will be simulated per OFDM symbol duration (which is much shorter than half of the wavelength), in order to feed into the link-level simulator.
e simulation setups are based on the RT frequencydomain method for the ultra-wideband channel, and the material parameters are extracted from measurements or literature. e involved material parameters are summarized in Table 2, where ε r ′ is the real part of the relative permittivity, tan δ is the loss tangent, and S and α are the scattering  coefficient and scattering exponent of the directive scattering model [36]. Particularly, the parameters of wood are calibrated in [37,38], and the parameters of concrete are calibrated in [39]. erefore, with the provided calibrated material parameters, various V2I Tx/Rx deployments can be simulated by an RT simulator.
Propagation mechanisms in the simulations are LOS, scattering (multiple scattering theory for vegetation and single-lobe directive model for others), diffraction (uniform theory of diffraction (UTD)), reflection (up to 2 orders), and transmission. e detailed configuration of the RT simulated channels in urban environments is summarized in Table 3. e propagation mechanisms for different objects in the simulation environment are summarized in Table 4.

Propagation Mechanism Analysis.
In this work, 3 different traffic flows are considered: (i) Full traffic flow: 100% randomly generated vehicles on the road (ii) Half traffic flow: 50% randomly generated vehicles on the road (iii) Low traffic flow: 10% randomly generated vehicles on the road e considered simulation cases are summarized in Table 5 for clarity, and the exemplified simulation results are shown in Figure 10. Figures 10(a) and 10(b) show the MPC distribution of the typical snapshots for Case 9 and Case 10, respectively; moreover, the time-varying PDPs are also given to distinguish the MPC distribution in the delay domain intuitively. According to different MPC distribution and propagation mechanisms, the "zone" concept is proposed. Different zones for different cases are used for channel characterization in the next section. For Case 9, three zones can be detected, and two zones are marked for Case 10. According to the RT simulation results for each case, there are 12 typical divided zones in Table 6, in which the details of propagation mechanisms for different zones are listed. For example, in Zone 2, the Rx is out of the Tx main lobe, the LOS path and reflected and scattered paths (from urban furniture and other vehicles) exist, and no ground reflection exists, as shown in Figure 10(c); in Zone 9, the Rx is in the Tx main lobe, LOS and ground reflected paths exist, and no reflection or scattering from urban furniture or other vehicles is seen, as shown in Figure 10(d); in Zone 11, the Rx is in the Tx main lobe and the LOS path is blocked by a vehicle, as shown in Figure 10(e). Note that the rays whose power is more than 80 dB lower than that of the strongest ray are cut off to get the dominant propagation mechanisms.
Extensive high-resolution simulations are conducted in each propagation zone for different cases. All simulated channel data can be imported to the V2I link-level simulator to guide the construction of the vehicular infrastructure for better road services.

Key Channel Parameters for Link-Level Simulation
In order to support the link-level simulation, for every propagation zone, the spatial sampling interval is set to 80λ for high-resolution simulation. e delay resolution is

Path Loss.
e PL in dB is modeled by the A-B model, which is expressed as    Table 7, where A < 0 is under the conditions that the Tx and Rx antenna beams are misaligned at the beginning and that the moving Rx gradually approached the main lobe of the Tx. e similar result is presented in the previous work [31]. It can be found that the absolute value of A in Zone 1 to Zone 4 (outside the main lobe of the Tx and LOS condition) and Zone 11 and Zone 12 (within the main lobe of the Tx and NLOS condition) is larger than that in Zone 7 to Zone 10 (within the main lobe of the Tx and LOS condition).
is indicates that the main lobe and the

RMS DS and Ricean K-Factor.
e Ricean K-factor is defined as the ratio of the power of the strongest MPC to the power of the sum of the remaining MPCs in the received signal [40]. Traditionally, the Ricean K-factor is calculated from the narrowband channel sounding results by using a moment-based method [41]. However, the ultra-wideband (UWB) channel sounding results in this measurement have high resolution in the time domain. us, the Ricean Kfactor can be calculated according to its definition as follows: where KF denotes the Ricean K-factor and P strongest and P remaining denote the power of the strongest MPC and the power of the remaining MPCs, respectively. RMS DS is used to quantify the dispersion effect of the wide channel. It is defined as the square root of the second central moment of the PDP [42]: where σ τ denotes the RMS delay spread and P n and τ n denote the power and the excess delay of the n-th multipath. e fitting parameters for KF and σ τ are listed in Table 8.
As it can be seen from Table 8, the zones with the LOS path are outside the main lobes of transceivers (as shown in Figure 11), the Ricean K-factor is smaller than that of the zones with the LOS path in such main lobes, yet the delay spread is larger than that in the zones with the LOS path in the main lobes. e reason is that the MPCs in the main lobes would have higher gain than the LOS path which gives rise to increasing P remaining with a long delay. Another finding is that, in the case of the same environment, i.e., the same Tx height and the same zone, the mean value of Ricean K-factor in full traffic flow is smaller than that under the half and low traffic flow conditions; meanwhile, the mean value of DS is larger. As shown in Figure 12, there are more MPCs caused by more vehicles in the case of full traffic flow, resulting in an increase in power to diffuse MPCs. As shown in Figure 13, the Ricean K-factor in some cases of the NLOS channel with reflection is larger than that in other cases of the NLOS channel without reflection. e channel with the strongly reflected path is close to the cases of the LOS channel. It has been found that such strong reflected paths are mainly from a metal vehicle or the ground, and power of them is almost as strong as the LOS path.

Angular Spread.
According to the 3GPP definition, the conventional AS calculation for the composite signal is given by where σ AS denotes the AS, P n denotes the power of the n-th ray, and θ n,μ is defined by where θ n is the ASA/ASD/ESA/ESD of the n-th ray and μ θ is Generally speaking, ASD and ASA are larger than ESD and ESA, implying that most of the MPCs come from the horizontal direction. is reflects the fact that more objects (scatterers) are located on the two sides of the Rx vehicle in the urban environment than above/under the Rx vehicle (as shown in Figure 14). Figure 15 shows the variation of the AS with Tx-Rx distance for the urban environment. It can be seen that, in the case of the LOS channel, the mean value of angular spread within the main lobe is generally smaller than that outside the main lobe. is indicates that when the Rx is within the main lobe, the distance between the Rx and the Tx is larger, and the object that produces reflection and scattering is closer to the Tx than the Rx. Moreover, the main lobe filters low-energy rays, resulting in the reduced AS.

Environment
Zone  Another finding is that when the Rx vehicle is within the main lobe, the AS for NLOS channels is much larger than that for LOS channels.

Blockage Loss.
ere are buildings, small-scale structures, and moving vehicles in the analyzed urban environments. When the vehicle on which the Rx is located is moving on the road, the directed ray may be blocked by these objects. As shown in Figure 16, the maximum blockage loss and the corresponding blockage time for each segment are calculated and fitted by the normal distribution with the mean value μBL, σBL_t, and standard deviation μBL, σBL_t, respectively. e parameters are summarized in Table 10.

Conclusion
In this paper, the 23 GHz V2I channels are characterized by measurement-validated RTsimulations for different antenna heights in various propagation environments. e PL, shadow fading, Ricean K-factor, DS, AS, and blockage loss of the channels are characterized for each zone in different cases. In the considered 3D Seoul urban environments, all the main buildings and small-scale structures, such as traffic lights, traffic signs, bus stations, and trees, are included with their typical geometries and materials in reality. e fundamental channel parameters analyzed are summarized in tables, which can be imported to standard channel models for generating realistic channels for similar environments. e study of this paper and the provided suggestions will be useful to guide the design and deployment of vehicular communication systems with FACS technologies.
Data Availability e environment models and simulation data of this study are available from the corresponding author upon request.