Introduction
A. Overview of V2X Wireless Technologies
In recent years, wireless access technologies applied to connected and autonomous vehicles (CAVs) have emerged as a promising paradigm for enhancing safety, traffic efficiency, and reducing pollutant emissions through V2X communications. By improving situational awareness, optimizing traffic flow, and decreasing emissions, they have the potential to revolutionize transportation, leveraging communication between vehicles and all the other road actors (including pedestrians, signals, and infrastructures), through vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), vehicle-to-network (V2N), and vehicle-to-pedestrian (V2P) interactions. Consequently, Cooperative Intelligent Transport Systems (C-ITS) have evolved, supporting dynamic, real-time information exchange. This also provides solutions to challenges encountered in advanced autonomous driving technologies, such as perception blind spots, computational capacity limitations, and decision-making difficulties, thereby further advancing the development of intelligent transportation systems.
C-ITS systems are mainly enabled by two key wireless communication technologies: Cellular-V2X (C-V2X) and ITS-G5 (the European version of IEEE 802.11p) [1]. ITS-G5, derived from the Wi-Fi standard (IEEE 802.11) and also known as mobile Wi-Fi, operates at 5.9 GHz band with a typical Effective Isotropic Radiated Power (EIRP) of 23 dBm (0.2 W) and a maximum of 33 dBm (2 W), primarily facilitating short-range V2V and V2I communications. Meanwhile, the C-V2X technology, developed by the Third Generation Partnership Project (3GPP), encompasses LTE-V2X and 5G NR-V2X (released in 2018), supporting frequency range 1 (FR1) bands at 5.9 GHz (both LTE-V2X and 5G NR-V2X) and frequency rang 2 (FR2) bands (5G NR-V2X only), with a maximum EIRP of 23 dBm.
To implement C-ITS systems, various devices, especially RSUs and OBUs, have been developed. RSUs, as a critical component of vehicle-road cooperative technology, primarily gather road and traffic condition information and, through communication with roadside sensing devices, traffic signals, and electronic signage, allow real-time vehicle-road connectivity and traffic signal exchange, enhancing driving safety and efficiency. OBUs, on the other hand, engage in V2X communication with RSUs and can independently support various types of V2X communications, such as short-range V2V and V2I communication (ITS-G5), or connect to cloud services or traffic management centers via legacy cellular networks (C-V2X). It makes every connected car as a network node.
While the application of C-ITS and V2X technologies is expected to significantly improve road safety and traffic fluidity, the introduction of complex communication scenarios and new devices not only poses challenges for the development of communication technologies, such as channel modeling in high-speed mobility environments and electromagnetic compatibility issues [2], but may also lead to new health concerns. As C-ITS systems and V2X technologies could be widely deployed, road users will be continually exposed to EMF generated by various communication technologies, raising public concerns about the health impacts of EMF exposure in V2X communication scenarios.
B. Literature Review and Study Objectives
To the best of the authors’ knowledge, research on the level of EMF exposure for road users under V2X communication scenarios is currently limited. Existing studies have primarily focused on RF EMF exposure inside vehicles, centering on radiation sources from mobile communications and personal infotainment devices operating from 900 MHz to 2.5 GHz frequency bands [3], [4], [5], [6], [7], [8], [9], [10]. These studies do not encompass the novel frequency bands utilized in V2X communications and intelligent vehicular sensing devices, particularly ranging from 5.9 GHz to 24-80 GHz.
A preliminary work highlighting the importance of exposure studies at V2X frequencies has been proposed by Tognola et al. [11] and provides a comprehensive survey of RF EMF exposure from multiple technologies in connected cars, including V2X communication, automotive radars, and intra-car IoT devices. They discussed the frequency ranges, potential exposure levels, and compliance with existing exposure limits for each technology. The study highlights the necessity for an in-depth investigation into the real-world EMF exposure scenarios associated with 5.9 GHz V2X communications, particularly the lack of data on cumulative exposure effects when these systems operate concurrently with other technologies.
Regarding 5.9 GHz V2X technologies, there are studies [12], [13], [14], [15] that explore external antennas for V2V communication, focusing on antenna performance optimization and channel modeling, but they do not cover EMF exposure assessments in V2X connectivity. For example, Tognola et al. [1] employed numerical simulations to assess the Specific Absorption Rate (SAR) impact on vehicle occupants by roof-top antennas operating in the IEEE 802.11p technology’s 5.9 GHz frequency band for V2V connectivity. The results indicated that, even in the worst-case exposure scenarios, the SAR values in various human tissues remained significantly below the ICNIRP’s basic constraint. Schilling et al. [16] measured the electric field strength produced by ITS-G5, Bluetooth, and Wi-Fi technologies inside vehicles, finding that the maximum exposure level came from the ITS-G5 antenna on the front windscreen but did not exceed 15.1% of the ICNIRP reference level. Nevertheless, these studies mainly focused on the exposure levels inside vehicles without addressing the exposure assessment of pedestrians and others in V2X communication scenarios.
Recently, Bonato et al. [17], [18] conducted studies using numerical and stochastic dosimetry to investigate the electromagnetic exposure of pedestrians near vehicles equipped with 5G-V2X antennas operating at 3.5 GHz, finding that even under the influence of antenna 3D beamforming, the SAR values in all evaluated scenarios remained below the basic restrictions for public exposure. In Benini’s studies [19], [20], the EMF exposure of adults and children near vehicles with V2V technology has been assessed through simulation, considering two monopole antennas mounted on the roof operating at 5.9 GHz, reaching similar conclusions. In their most recent study [21], the authors employed ray tracing methods to simulate the EMF exposure levels (the Whole Body SAR (WBS) as metric) of road users in a simplified urban environment due to vehicles and infrastructure equipped with 5.9 GHz V2X antennas. The findings concluded that the WBS in this scenario was exceedingly low, generally around the level of 10−4 W/kg. Tognola et al. [22] used a hybrid deterministic and stochastic approach to evaluate the variability of RF exposure in urban V2V communication at the 5.9 GHz band across various human body models. The study revealed that even with multiple vehicles and antennas transmitting at maximum power, the median SAR values remained at lower levels. These studies highlight the application of numerical simulation in assessing EMF exposure but also point out the lack of experimental measurement data.
For the 24-80 GHz mmWave frequency band, commonly used in vehicles’ Advanced Driver-Assistance Systems (ADAS) or Vehicle Occupant Detection (VOD) radar, there is currently a scarcity of research on the RF EMF exposure from vehicular mmWave radars in real-world scenarios. Existing studies on mmWave exposure include the work by Ruddle et al. [23], which assessed the exposure levels inside vehicles generated by external ADAS radars (24/46.8/77 GHz). It was found that the exposure produced by the highest frequency radar, 77 GHz, was a maximum of 0.76 W/m2, well below the ICNIRP reference level of 10 W/m2. Other studies [24], [25], [26] explored the temperature rise in eye and ear models under plane wave exposure in the mmWave frequency band through numerical simulation. Vermeeren et al. [27] measured the power density at distances of 3-30 mm from vehicle radars operating at frequencies of 60/79 GHz with maximum output power (10 dBm) and full duty cycle, finding that the antenna exposure levels (calculated as average power density) were below the basic limits set by ICNIRP. Thus, these studies have only analyzed exposure levels under highly simplified scenarios or laboratory environments. They have not conducted evaluations within actual vehicular contexts.
On the contrary, this study aims to comprehensively assess the impact of RF EMF exposure on road users when V2X communication technologies are adopted through a combination of numerical simulation and experimental measurement, thus filling the existing research gaps. Our research goals are multi-fold.
First, we focus on assessing the impact of EMF radiation induced by ITS-G5 and C-V2X technologies employed by RSUs and OBUs on adults and children near RSUs, using numerical simulation methods with SAR values as the evaluation metric.
Subsequently, we evaluate the exposure levels of RSUs in laboratory (in-lab) and real-world (in-situ) environments as well as OBUs installed on vehicles through experimental measurements of the electric field strength or EMF power density.
Finally, our research measures the exposure levels of infotainment, advanced sensing, and IoT devices inside and around modern smart cars (with a Tesla Model S as the study object), with particular attention to Bluetooth, integrated LTE modules, and mmWave radar.
The work is organized as follows: Section II details the simulation and measurement methods employed, Section III presents and comprehensively discusses the simulation and measurement results, and finally, Section IV summarizes the findings of this study.
Equipments and Methodologies
A. Rsu Equipment Setup and Rsu Exposure Simulation
The RSU device involved in this study consists of a large electrical cabinet, which houses all the different modules of the RSU. These include modules for wireless communication, modules for local processing on the RSU, and modules that allow the RSU to be managed and recovered remotely when needed. The antennas of the communication modules are mounted on top of the RSU, as shown in Fig. 1 (left). The RSU is mobile, it can be installed outdoors during the testing period and positioned as desired in customs sites. Fig. 1 (right) shows one of the RSUs deployed along the E313 highway in Antwerp on top of the gantry.
RSU equipment: front view of the RSU at the left and its deployment on the highway at the right.
Table. 1 lists the three commercially available wireless communication devices mounted on the RSU, including Cohda Wireless MK5 and MK6c, and Ettus USRP N310 (a Software Defined Radio (SDR), hereafter referred to as SDR), along with their respective implemented communication protocols, especially noting their maximum transmitted ERIP power which will be used in our subsequent simulations and experimental measurements. It is evident that RSU’s technology stack covers the current mainstream V2X communication technologies. It is important to note that while the ITS-G5 technology implemented by Cohda MK5 specifies a maximum transmitted EIRP power of 33 dBm (for non-government services), the actual maximum transmission power of this device is only 23 dBm. The maximum transmission power of Cohda MK6c is 23 dBm, consistent with the standard specification of the C-V2X technology (Table. 1). The SDR implementation of C-V2X technology through the Uu interface, including LTE-V2X and 5G NR-V2X technologies, has not been deployed in real-world RSUs. Therefore, no related measurements of this device were involved in experimental measurements. However, in the numerical dosimetry investigation, the SDR’s transmitting antennas for each band (as specified in Table. 1) are still included as radiation sources in this work to consider the worst-case scenario.
As far as the simulation is concerned, the exposure scenarios are illustrated in Fig. 2. It includes an overall model of the RSU, consisting of an electrical cabinet, a base plate for mounting antennas of each communication module, and a human body model (adult or child) placed sequentially at four marked positions, which namely: “POS 1”, “POS 2”, “POS 3”, and “POS 4”. Specifically, we considered two types of Virtual Population (ViP) anatomical human models, namely the Duke model (an adult male, 34 years old) [38] and the Billie model (a female child, 11 years old) [39].
Simulation scenario: a full RSU setup, featuring all the transmitting antennas, with tri-band: 0.8-1.8-2.6 GHz (marked as triangles), 3.5 GHz (square), and 5.9 GHz (circles) antennas for each technology, along with a human model (Duke or Billie) placed beside each of the RSU’s four sides.
Fig. 2 also depicts the layout of the transmitting antennas, it includes three ITS 5.9 GHz antennas: the Cohda MK5 (ITS-G5), Cohda MK6c (C-V2X, PC5 Interface), and SDR N310 (C-V2X, Uu Interface), which are denoted by blue, green, and yellow circles, respectively. Additionally, the RSU’s SDR is connected to antennas operating at other frequencies. This includes two tri-band antennas for LTE technology operating at 0.8, 1.8, and 2.6 GHz, marked with green triangles, and a 5G NR FR1 band antenna at 3.5 GHz, indicated by a yellow square. An active Global Navigation Satellite Systems (GNSS) antenna is also connected, although it is not modeled in this study due to its irrelevance to our study.
The simulations were conducted by Sim4life (ZMT Zurich Med Tech AG, Zurich, Switzerland) [40], which utilizes the Finite-Difference Time-Domain (FDTD) method for its solver. The antennas installed on the RSU were modeled as dipole arrays matching the dimensions of commercial antennas actually used in practice (an example of a modeled 5.9 GHz antenna is depicted in Fig. 2), with antenna and electrical cabinet materials set as perfect electric conductors (PEC) to enhance simulation speed; the base plate was modeled with plexiglass (relative permittivity
To compare with the results of experimental measurements, we first calculated the spatial electric field within 2 m around the RSU on the azimuthal plane (i.e. E-plane of the antennas) of the antennas with both Cohda MK5 and MK6c switched on, excluding the human model. Both ITS-G5 and C-V2X utilize Orthogonal Frequency Division Multiplexing (OFDM) and Quadrature Amplitude Modulation (QAM) modulation, meaning that the signals transmitted by the two antennas are not always in phase; the phase difference varies randomly over time. This leads to fluctuating exposure patterns from the two antennas. To better understand the variations in EMF exposure levels due to multi-signal source interactions, we set the phase of one antenna’s source to 0° and varied the phase of the other from 0° to 360° (excluding 360°), at intervals of 45°, including randomly selected intermediate values, resulting in 16 distinct phase differences: 0° (in phase), 21.9°, 45°, 66.2°, 90°, 107.6°, 135°, 167.4°, 180°, 205.6°, 225°, 246.2°, 270°, 290.2°, 315°, and 338.4°. All results normalized to their actual transmission power of 23 dBm.
For dosimetric assessment, simulations at each frequency were conducted with all antennas transmitting at that frequency activated at the same input power and in phase to simulate the worst-case scenario. Following the guidelines of the ICNIRP [42], the simulations calculated the peak spatial-average SAR averaged over 10 g (psSAR10g) of tissue and the WBS, across all frequency bands implemented by the RSU at four specific positions. Specifically for the 5.9 GHz frequency band of interest in this study, the SAR10g distribution was calculated and analyzed within six typical human tissues: skin, subcutaneous adipose tissue (SAT), fat, muscle, bone, and blood. These results were normalized to an input power of 1 W (EIRP: 30 dBm) for convenience in rescaling the SAR values at different input powers.
B. RSU and OBU Exposure Measurements Setup
For the experimental measurement of EMF exposure levels from RSUs and OBUs, we measured the root mean square (RMS) value of electric field strength E in V/m (or the power density S in W/m2) of the received signals. The equation for the interconversion between S and E is [42]:\begin{equation*} S = \frac {|E|^{2}}{Z_{0}} \tag {1}\end{equation*}
The exposure levels of the RSUs have been evaluated both in laboratory (in-lab) settings and on-site next to the highway (in-situ). Fig. 3 shows the in-lab measurement setup. Precautions were taken to avoid pedestrians to prevent any impact on the measurements. The measurement points were aligned to the simulation, situated in the directions of the four sides of the RSU. The center of the probe was positioned at a height of 1.16 m above the ground, and the distance between the probe and the RSU was defined as the distance from the center of the probe to the edge of the RSU antenna base plate. The measurement points in each direction from the RSU were at distances of 0.5 m, 1 m, and 2 m. It is important to highlight that we configured all the devices to transmit continuously to allow us to measure the worst-case conditions, so both MK5 and MK6c were configured to continuously transmit their real-time Global Positioning System (GPS) information at a bandwidth of 10 MHz, maximum byte rate of 12 Mbps, and maximum transmit power of 23 dBm during the measurements.
RSU in-lab measurement experiment setup. The probe, positioned 1.16m above the ground, is placed at various distances around the four sides of the RSU.
For in-situ measurements alongside highways where RSUs are deployed, the upper part of Fig. 4 shows seven RSUs positioned along the E313 highway in Antwerp as part of the Smart Highway testbed for developing C-ITS applications [43]. The wireless communication modules installed in these RSUs and their technical characteristics are identical to those used in the RSU for the in-lab experiments described in this study and thus are not reiterated here. In this work, we selected two of these seven deployment sites for evaluating the exposure levels in the vicinity of their RSUs, hereafter referred to as Site 1 and Site 2, respectively. The lower part of Fig. 4 depicts the RSU at Site 1 and its installation position on the gantry top along the E313 highway. We used the Duke human model from the ViP [38] as a reference to define the measurement positions. We selected measurement points at heights of 1 m (basin (pelvis) position) and 1.64 m (eye level) above the ground, as well as the maximum height reachable by the probe at 2 m. At Site 1, the exposure levels at these points were measured at distances of 1 m, 5 m, and 10 m from the gantry. Conversely, at Site 2, the scope of measurement was expanded, conducting continuous measurements at six positions extending from 13 m away from the gantry, which was installed with RSU, approximately every 9 m. Still, vertical measurement points were confined to heights of 1 m and 2 m. For both in-lab and in-situ measurements, the SA was set to frequency domain measurement (default mode) in max-hold mode (with RMS detector) to record the highest field levels received until the signal stabilized. Fig. 5 (a) and Fig. 5 (b) respectively present the experimental setups for the in-situ measurements conducted at Site 1 and Site 2.
RSUs deployment along the E313 highway in Antwerp (upper) and RSU used for measurements and its installation position on the top of a gantry (lower).
The RSU in-situ static measurement setups at (a) Site 1, (b) Site 2, and (c) the on-road in-car measurement setup.
Furthermore, we dynamically assessed the exposure levels at passenger seats in vehicles traveling on highways equipped with RSUs (so-called on-road measurements). Specifically, vehicles were driven past two RSUs with the SA probe placed in the passenger’s seat at a height of approximately the human chest. Fig. 5 (c) depicts the setup of this on-road in-car experimental measurement. In Fig. 5 (c), the gray car was equipped with an OBU antenna on the roof, but this OBU was only used to verify the RSU’s work status in this experiment and was not used for signal transmission.
Critically, all in-situ measurements were preceded by a control measure, where all wireless communication devices on the RSU were turned off to observe the spectrum of our measured band and ensure no signals from other radiation sources were present. All in-situ measurements were conducted in full compliance with local traffic safety laws and regulations.
This study also assessed the exposure levels from a Cohda MK5 OBU based on ITS-G5 technology, aiming to measure the electric field or power density received both outside and inside the vehicle. Measurements used an Audi A5, where the Cohda MK5 OBU installed shared the basic characteristics with the Cohda MK5 on the RSU, with the difference that the OBU’s maximum transmission power could be set to the maximum transmission power of 33 dBm following ETSI standard [30]. Hence, to accommodate worst-case scenarios, all our measurements were conducted with the OBU transmitting at this power level. The OBU was equipped with a shark fin antenna such that the transmitted EIRP was equal to 33 dBm with a gain of 5 dBi. As shown in Fig. 6 (a), we considered two antenna configurations, symmetrically placed on the vehicle’s roof at the front and rear windscreens proximity, indicated by the blue dots in the figure. To comprehensively measure the electric fields received from the front/rear antenna both outside and inside the vehicle, we predefined five measurement positions each for the exterior and interior. For the exterior, the five positions and their respective distances from the front/rear antenna are also indicated in Fig. 6 (a). Similar to the RSU’s in-situ measurements, we referenced the Duke and Billie models from the ViP model for positioning the probe at the height of the head (eye) and basin (pelvis) for measurements, with Fig. 6 (a) also showing these heights. As for the in-car measurements, shown in Fig. 6 (b), the predefined positions were determined based on the five passenger seats within the vehicle cabin, namely: driver, front passenger, back left, back central, and back right seat. At each seat, electric field strength was measured at both the head and basin positions (Fig. 6 (b)).
OBU measurement setup: (a) two blue dots on the car roof indicate the two OBU antenna positions; five predefined measurement positions around the car, labeled CONF 1-5, along with the distances from each position to the two antenna; the lower part of the figure shows the heights from the ground for the four measurement points based on the ViP Duke and Billie human models. (b) illustrates the five seats inside the vehicle and the defined measurement points at each seat (i.e., ‘Head’ and ‘Basin’).
C. Tesla Devices Measurements Setup
In this study, we investigated the exposure levels inside and around a modern smart vehicle (Tesla Model S) from Bluetooth technology used by infotainment devices, integrated LTE wireless communication technology, and mmWave radars used by advanced sensing systems. The locations of the exposure sources of the various technologies installed on the investigated vehicle are illustrated in Fig. 7 (a). This work involved measuring the electric field strength/power density and DC (%) of the signals emitted by these technologies, with DC defined as the fraction of the actual transmission time of these signals \begin{equation*} DC = \frac {t_{trans}}{t_{tot}}\times 100 \;(\%)\,. \tag {2}\end{equation*}
(a) Top view of Tesla Model S and position of each wireless communication device under investigation; (b) measurement setup for the radar in the headlights, involving an FSV-3030 SA with a mmWave probe held very close to the radar; (c) measurement setup for the in-vehicle Bluetooth device, involving an SRM3006 SA and a probe placed on the height of the chest of the driver.
We measured the electric field strength and DC of the Bluetooth devices inside the vehicle under different operating states. It should be noted that while the tire pressure sensors are equipped with Bluetooth Low Energy (BLE) modules, their transmission only occurs when the tires are rotating, thus their measurement was omitted. This study considered the electric field levels received at the driver and passenger positions under different operating modes of the Bluetooth device, i.e., when the vehicle is turned off, idle, searching, and when connected to a phone transmitting data continuously (playing Spotify). To this aim, the probe was placed at the torso positions of the driver and passenger, as illustrated in Fig. 7 (c). Notably, when the Bluetooth device was connected to a phone for continuous data transmission, the phone was placed between the two seats, measuring the EMF exposure levels from both the vehicle’s Bluetooth and the phone’s Bluetooth simultaneously.
Finally, we measured the exposure level from the vehicle’s integrated LTE communication module, with its antenna integrated inside the rearview mirror housing, as shown in Fig. 7 (a). We considered two states of the vehicle: turned off and startup. In the initial measurement, the probe was directly in contact with the rearview mirror housing (probe diameter: 12 cm), referred to as the 0 cm distance. Subsequent measurements were taken at distances of 15 cm and 50 cm from the rearview mirror to measure the electric field strength and DC of the signals emitted by the module.
Results and Discussion
A. Rsu Measurements and Simulations
To the best of the authors’ knowledge, this study represents the first assessment of human RF EMF exposure in V2X communication scenarios implemented with ITS 5.9 GHz technology within RSU-based V2I cooperative systems. We conducted numerical simulations, in-lab, and in-situ measurements. An RSU with 5.9 GHz, LTE, and 5G NR antennas was modeled using FDTD to calculate the electric field, which was then validated by in-lab measurements. Exposure levels were evaluated using ViP anatomical models (Duke and Billie) for SAR10g and WBS. In-lab and in-situ experiments measured exposure at various distances from the RSU.
Fig. 8 shows the electric fields around the RSU as a function of distance in every 4 directions along the blue dash line showed in Fig. 2 extracted from simulation results, alongside the exposure levels at different distances measured in the in-lab experiment. Fig. 8 illustrates the simulation results with three curves showing the electric field distribution: one for ITS-G5 and C-V2X antennas under in-phase excitation, and the other representing the maximum and minimum field distributions obtained with different phase differences between the two antennas. Simulation results revealed that in the POS 1 and POS 2 areas, the in-phase excitation produces the highest field strength among the 16 different phase shifts tested. Conversely, in the other areas, there are configurations with different phase differences that result in higher field levels. Focusing on the maximum field strength distributions obtained from the simulation, in the reactive near-field region close to the antennas, the field strength is higher with noticeable fluctuations. However, the field rapidly decreases as the distance from the antenna increases, especially at POS 3 and 4, reaching up to 20 V/m within 0.5 m from the RSU but then quickly dropping below 5 V/m and continuing to decrease inversely with the square of the distance. In the study by Tognola et al. [1], FDTD simulation was used to assess the exposure level to occupants inside a vehicle from ITS-G5 technology operating at 5.9 GHz in terms of SAR within the human body. Although the transmission power, antenna types, antenna installation positions, and exposure scenarios used in [1] are different from ours, making a direct quantitative comparison unfeasible, a qualitative observation can be made that the field generated inside the vehicle by an ITS-G5 monopole antenna mounted on the vehicle’s exterior also decreases very rapidly with distance, showing a more significant attenuation rate compared to other communication technologies operating at lower frequencies (LTE, Bluetooth, etc., operating frequencies within 0.8-2.6 GHz). This indicates the relatively higher path loss of this frequency band. Despite the absence of transmitting antennas near POS 1 and 2, the field is maximally superimposed at these two positions due to the symmetrical distribution of the two antennas relative to these positions. Even though the exposure level is not as high as at POS 3 and 4 within close proximity to the RSU (within 0.5 m), the overall decreasing trend is not as rapid, hence the highest exposure level measured in the in-lab experiment was 5.8 V/m at POS 1, 0.5 m from the RSU. It can also be seen that although other antennas of different frequency bands are installed on the RSU, their impact on the field at POS 3 and 4 at 5.9 GHz can be disregarded, with the electric field at these positions being considered symmetrically distributed.
Simulated (normalized with input power of 23 dBm) versus measured (input power of 23 dBm) electric field (RMS value) generated by the RSU antennas as a function of the distance from the antenna. The simulation results present three curves: one for ITS-G5 and C-V2X antennas under in-phase excitation, and two representing the maximum (Max E-field) and minimum (Min E-field) field distributions with varying phase differences between the antennas.
Fig. 8 also presents the in-lab measurement results of exposure levels (RMS values of the electric field) at different distances around the RSU, with the uncertainty of isotropic measurements: ±3 dB displayed as error bars in the figure. As mentioned, the highest measured exposure level was 5.8 V/m, which is 9.5% of the ICNIRP’s reference level of 61 V/m for the 5.9 GHz field. Comparing simulation and in-lab measurement results reveals good agreement, particularly at distances of 1 m and 2 m from the RSU in various directions. However, field strength measurements at 0.5 m from the RSU are consistently lower than those predicted by simulations, most notably in the POS 3 and POS 4 regions, with deviations not exceeding 3 dB in the POS 1 and POS 2 regions, which remains within a reasonable range. Three primary factors contribute to these deviations: firstly, the phase differences in signals causing significant exposure are transient, and our use of an RMS detector to prevent systematic overestimation of exposure levels results in the under-recording of higher field strengths; secondly, there is uncertainty in the placement of the measurement probe; and lastly, differences between the simulated and actual commercial antennas used, as well as the uncertainty in the input power of each device.
Fig. 9 presents the in-situ static measurement results at Site 1 for ITS-G5 and C-V2X technologies at three different heights selected with reference to the ViP Duke human model, showing the exposure levels at various distances from the gantry equipped with the RSU. It is observed that due to the far-field pattern of the antennas, i.e., the omnidirectional dipole (array) antenna, the closest position to the RSU in the horizontal direction, specifically 1m from the gantry, is not where the highest exposure level is found since this position is almost immediately below the RSU. Similarly, the closest vertical position to the RSU does not always correspond to the highest exposure level. For the ITS-G5 technology, the maximum exposure level was measured at 81.2 mV/m at the eye level of the Duke human model (i.e., Duke Head, 1.64 m) 10 m away from the gantry, significantly below the ICNIRP’s reference level of 61 V/m, representing less than -58 dB of the reference value. For C-V2X technology, the position of the maximum exposure level was found at 5m from the gantry at the basin region of the human model (Duke Basin, 1 m), measuring 75.2 mV/m, also not exceeding −58 dB of the ICNIRP reference level.
Site 1 RSU measurements: Electric field (RMS value) of ITS-G5 and C-V2X (both with 23 dBm input power) as a function of distance for three heights: 1 m (“Duke Basin”), 1.64 m (“Duke Head”), and 2 m (“Max. Height”).
Fig. 10 presents the exposure levels measured at Site 2 for two technologies. Considering the measurement results from both sites, it becomes clear that the exposure levels of both technologies do not monotonically decrease with increasing distance from the radiation source. Instead, significant fluctuations were observed at some locations, with the maximum variation exceeding 8 dB between two adjacent measurement points at the same height, and a maximum difference of 7.7 dB between two measurement points at different heights but the same horizontal position. This reflects that due to their higher frequency (5.9 GHz), the spatial variation in exposure levels for both technologies is pronounced, and the complexity of real-world RF propagation. Nevertheless, the overall trend shows a decay in exposure levels with increasing distance, with the maximum exposure level measured at Site 2 (65.5 mV/m) being lower than that at Site 1, but the rate of decrease is not linear, which could be due to environmental factors, interference, or multi-path propagation effects.
Site 2 RSU measurements: Electric field (RMS value) of ITS-G5 and C-V2X (both with 23 dBm input power) as a function of distance for two heights: 1 m and 2 m.
For the real on-road in-car measurements, the maximum exposure levels recorded at the passenger position for ITS-G5 and C-V2X technologies were 12.4 mV/m and 10.3 mV/m, respectively, which is 74 dB lower than the ICNIRP reference level (61 V/m). It is important to note, as mentioned in [49], that the number of occupants inside the vehicle can affect the exposure levels. The experiment included two occupants in the vehicle, including the driver.
Since the SAR values are directly proportional to the input power, we normalized the calculated human SAR values for an antenna transmit power of 1 W. This normalization facilitates the scaling of SAR values for different input powers. Table. 2 shows that the highest psSAR10g in both Duke and Billie occurs when the human body is positioned at POS 4, exposed to the 2.6 GHz electric field. This is due to the fact that the human body is the closest to the transmitting antenna (LTE band 0.8-2.6 GHz antenna, marked as the green triangle in Fig. 2) at this location. The maximum psSAR10g values for the two cases are 248.7 mW/kg (Duke) and 218.5 mW/kg (Billie), respectively, both are 8 times below the 2 W/kg limit of local exposure (torso) in the 100 kHz-6 GHz range set by ICNIRP. Nevertheless, the maximum transmit power in this 2.6 GHz LTE band is limited to 23 dBm, so the results here normalized to an input power of 1 W overestimate the SAR. Rescaling the results by the EIRP of 23 dBm, the values of the two maximum psSAR10g above become 49.7 mW/kg and 43.7 mW/kg, respectively, which is 40 times lower than the ICNIRP limit.
For the ITS 5.9 GHz band of interest in this study, we analyzed the SAR10g distribution in human tissues at this frequency. Fig. 11 compares the distribution of SAR10g in various tissues (Skin, SAT, Fat, Muscle, Bone, and Blood) within the two human models (Duke and Billie) across different exposure scenarios at 5.9 GHz. It is important to note that SAR averaging is generally considered an average over a cubical mass of tissue. However, for thin-layered tissues (e.g., skin), the averaging volume may substantially deviate from a cube due to the complex algorithm used for SAR averaging [50]. The results in Fig. 11 show that the distribution of SAR10g in all scenarios and tissues exhibits a positive skewness, meaning that in various scenarios, the SAR10g values for most tissues are distributed at lower exposure levels, with only a few samples reaching relatively high SAR10g values. In all exposure scenarios, the skin obtains the highest SAR10g values, significantly higher than those of other tissues, i.e., the radiation energy was predominantly absorbed by the skin. The maximum values occur when Billie is at POS 1 and Duke at POS 3 (Fig. 2), with 264 mW/kg and 223 mW/kg respectively, both considerably below the ICNIRP’s basic restriction for general public exposure of local SAR10g (torso), which is 2 W/kg [42]. These positions also represent the closest proximity of the two human models to the antenna. Fig. 12 shows the distribution of SAR10g over the skin of the whole human models in each exposure scenario at 5.9 GHz, normalized by their peak values. Due to differences in body height, for Duke, the SAR10g is primarily distributed in the abdominal area, whereas for Billie, it is mainly in the chest region. Finally, Table. 2 lists the WBS for different human body models in each exposure scenario. The maximum WBS at 5.9 GHz for Duke and Billie are 0.60 mW/kg and 1.06 mW/kg, respectively, which are both below the 0.08 W/kg limit for the general public exposure [42] (−18 dB of the ICNIRP). Additionally, in most exposure scenarios, Billie’s WBS is higher than Duke’s despite Billie having a smaller body volume. This is because, although Billie absorbs slightly less EMF power compared to Duke, the difference in energy absorption is smaller than the difference in weight between Billie and Duke, resulting in higher WBS for Billie.
Box plot of the 10g average SAR (normalized with input power of 1 W) distribution in each tissue (skin, SAT, fat, muscle, bone, and blood) of the two human models for the four exposure scenarios at 5.9 GHz. The orange line in the center of the box indicates the median; the top and bottom edges of the box represent respectively the 75th and 25th percentiles; the upper whisker is 1.5 times the height of the box i.e. 99.3% of 10g average SAR value, the lower whisker is the minimum value.
Distribution of normalized 10g average SAR across the whole body skin of different human models (Duke and Billie) for each RSU exposure scenario at 5.9 GHz (normalized to their respective maximum values). Color map range from 0 to −25 dB.
Comparing our results with prior studies helps to establish a clearer picture of the exposure in the V2X communication scenario. Although recent studies have primarily focused on exposure assessments within V2V, i.e., OBU communication scenarios, they all utilize ITS-G5/C-V2X technologies. By cross-comparing exposure levels across different endpoint devices within V2X communication scenarios, we can gain a more intuitive understanding of the exposure levels of RSUs, especially in relation to OBUs. Therefore, conducting a qualitative comparison of these studies is meaningful. In a recent study, Benini et al. [21] employed the ray tracing method to simulate exposure assessments in urban environments, using WBS as a metric. Their research utilized a simplified city model and five vehicle models, with 5.9 GHz antennas having a gain of 0 dBi (i.e., half-wave dipole) and a maximum transmission power of 33 dBm. They found that the exposure levels were generally very low (on the order of 10−4 W/kg), even in the worst-case scenario where antennas on five vehicles and traffic light RSUs were transmitting simultaneously, the maximum WBS was only
Other studies by the same authors [19] and [20] conducted numerical dosimetry assessments of electromagnetic exposure levels for adults and children near vehicles equipped with 5.9 GHz antennas for V2V connections. In these studies, two monopole antennas identical to those in [1] were placed at the front and rear positions on the vehicle roof, each with an input power of 1 W, the same as in our study. Different human models were positioned around the vehicle in two orientations. The highest local SAR10g values were likewise found in the skin tissues, with the adult model (Ella) having a maximum SAR10g in the head skin tissue of 34.7 mW/kg, and for the child model (Dizzy), it was 9 mW/kg. Both are below the maximum SAR10g values found in adult and child skin tissues exposed to the 5.9 GHz RF field in our study. The reasons are twofold: firstly, although the human models in [19] and [20] were placed very close to the vehicle, the shortest horizontal distance from the models to the actual antenna installation positions was still 0.5 m, especially the child model, which due to its shorter stature, was further from the antennas, overall greater than the distance between the human models and the antennas in this study; secondly, compared to the two independent monopole antennas in [19] and [20], each transmitting antenna in this study consists of an array of three dipoles, offering higher directionality and gain towards the human models. For WBS, the maximum values in [19] and [20] for adult and child models were 0.19 mW/kg and 0.18 mW/kg, respectively. These values are not only lower than those found in our study but also indicate that the WBS for children is less than for adults, due to the lower height of children causing them to be further away from the antennas compared to the adult model.
In [1], an “asymmetric” worst-case exposure scenario was considered, moving the originally symmetrically mounted antennas on the vehicle roof and rear/left mirrors directly to the nearest point to the head of the human model inside the vehicle and simultaneously activating these four antennas with the maximum transmission power of 44.8 dBm (30 W) as specified by the IEEE 802.11p protocol (for U.S. government services). They found the maximum local SAR10g in the head’s skin issue to be 1.58 W/kg, and the WBS to be 0.008 W/kg. Scaling our results to the input power of 30 W, the values for the adult model become 6.7 W/kg and 0.018 W/kg, respectively; for the child model, they would be 7.9 W/kg and 0.032 W/kg. Hence, the local exposure levels in our findings not only exceeded those in [1] but also surpassed the general public exposure limits set by ICNIRP [42] and IEEE [51], while the WBS remains below the limit and is higher than in [1]. Although [1] used four antennas, in the “asymmetric” scenario, only two antennas were within 0.5 m of the human model, and the vehicle’s metal body (PEC) partially shielded the occupants. It is noteworthy that these local exposures exceeded general public exposure limits because they used the maximum transmission power reserved for government services, and such close proximity between the human body and the RSU typically occurs in occupational exposure scenarios. Fig. 4 shows an example of a real RSU deployment location, situated at the top of gantries, a position inaccessible to pedestrians. Therefore, assessing them against occupational exposure limits is more reasonable, and these local SAR values are all below the ICNIRP and IEEE occupational limits (for the torso), which is 10 W/kg. Moreover, if these antennas were implemented with C-V2X technology (including 5G-V2X), with the protocol’s maximum input power set at 0.2 W (EIRP: 23 dBm), the exposure levels would be even lower.
B. OBU Measurements
In this study, we measured the exposure levels inside and around an Audi A5 vehicle equipped with ITS-G5 technology, implemented through a Cohda MK5 OBU mounted on the vehicle. A 5.9 GHz antenna connected to the OBU was placed at two locations: near the front windshield and on the rear part of the vehicle roof. Fig. 13 summarizes the exposure levels at five positions around the vehicle at four different heights when the antenna was located at the front/rear roof positions, respectively. The highest exposure level occurred at position CONF 3 at the head height of the Billie model when the rear antenna switched on, measuring 2.4 V/m, which is only 4% of the ICNIRP reference level. When the front antenna was activated, the highest exposure level was found at position CONF 2, at the same height as the antenna installation, the head height of the Billie model (i.e., 1.36 m), measuring 1.8 V/m, which was the closest measurement point to the antenna. Compared to the rear antenna, higher exposure levels were often found at lower heights (Billie head or Duke basin) because the front antenna installation position is lower than the rear antenna, on the windscreen below the roof level. The shielding effect of the vehicle roof resulted in lower exposure levels at lower heights for the human model, i.e., the basin position, for the rear antenna.
RMS values of the electric field E around the car (human beside) with antenna (input power of 33 dBm) positioned on the rear and front of the car roof.
Considering the worst-case scenario, with both front and rear antennas operating simultaneously, combining the two results yields the highest exposure measurement point still at the Billie head position at CONF 3, measuring 4.2 V/m or 6.9% of the ICNIRP limit. This result is similar to the RSU measurement results, where the highest exposure level was measured at the center position, equidistant from the two antennas, with the fields from both antennas superimposing to create a relatively larger field value at this position.
Fig. 14 compares the exposure levels at selected positions (head and basin) inside the vehicle at the five passenger seats for both front and rear antennas. Compared to the measurements outside the car, the exposure level inside the car is substantially lower due to the shielding effect of the car body. In a comparison of the maximum values of electric field strength measured inside and outside the vehicle, the former is only one-third of the latter. It can be observed that in most scenarios, the head position always receives a higher exposure level, and the exposure levels at the head of rear passengers are similar for both front and rear antennas, making the head exposure level almost independent of the passenger’s position. The maximum exposure was also found at the driver’s seat head position when the front antenna was fed, measuring 0.68 V/m, and at the rear right passenger’s head position when the rear antenna switched on, measuring 0.69 V/m, both less than 1.2% of the ICNIRP limit. It is also noted that for the basin position, exposure levels are lower in the front seats but relatively higher in the rear seats regardless of which antenna was operating, with the highest found in the back central seat at 0.55 V/m (0.9% of ICNIRP). Overall, back passengers receive higher electric field strength, especially considering the worst-case scenario, i.e., both front and rear antennas being excited simultaneously. The exposure levels at the heads of rear passengers could be twice that of front passengers and could reach up to six times at the basin position, yet still not exceeding 2% of the ICNIRP limit.
Peak RMS values of the electric field E inside the car with antenna (input power of 33 dBm) positioned on the rear and front of the car roof.
To facilitate comparison with our previous measurements of RSU exposure levels, when rescaling the RSU measurement results based on an antenna transmission power of 33 dBm, the maximum exposure level measured in close proximity to the RSU (at 0.5 m) was found to be 18.3 V/m. In real deployment scenarios for RSUs, the maximum exposure received by pedestrians and inside the vehicle was 256.8 mV/m and 39.2 mV/m, respectively. It is evident that, in scenarios close to the radiation source, humans could receive higher exposure levels from the RSU than from the OBU. However, such scenarios are more common in occupational exposures and are not prevalent in general public exposure settings. In real deployment cases for both RSU and OBU, the exposure levels generated by RSU to road users are significantly lower than those from the OBU, especially for exposure levels inside the vehicle, where RSU contributes only 6% (−25 dB) of that from the OBU. Therefore, we can conclude that in novel V2X communication scenarios utilizing the ITS 5.9 GHz band, the RF EMF exposure to road users is predominantly contributed by OBUs.
Comparing recent studies, the placement of human models around the vehicle in the simulations by [19] is similar to the setup in this work. However, different from our findings, [19] found the highest SAR10g in the head’s skin tissue of human models at the rear of the vehicle, the position closest to the rear antenna. This discrepancy arises because, in our work, the CONF 5 position is not the closest to the rear antenna. Still, the location of maximum exposure in our study also is not the closest to the front/rear antennas. This reflects the complexity of 5.9 GHz field propagation in real-world scenarios, which is significantly influenced by the vehicle’s size and the antenna’s installation position. In [16], the exposure levels of Bluetooth, WiFi, and ITS-G5 technologies at nine different predefined points inside a vehicle were measured, with the ITS-G5 technology’s two antennas located at the front windscreen and rear part of the roof, both with a maximum transmission power of 33 dBm. Reference [16] found that the antenna on the front windscreen of ITS-G5 produced the highest field at various points inside the vehicle, with the maximum exposure level found at the driver’s seat trunk (defined as the basin in this study) being 3.3% of the ICNIRP reference level, significantly higher than the results measured at a similar position in this work (0.18% for the front antenna) and also higher than the maximum value found in this study (1.2%). This deviation could be because the front antenna in [16], a Di-patch antenna [15], might have a larger gain inside the vehicle, and the vehicle size also impacted the distribution of the field inside the car. Meanwhile, the exposure levels from the rear antenna in [16] were lower at all measured points inside the vehicle, not exceeding 0.2% of the ICNIRP reference level, which is less than the results found in this study. This is due to the installation position of the rear roof antenna in [16], which resulted in the vehicle body providing effective shielding for this antenna. It is also important to observe that variations in the measurement results are expected when considering different cars with different dimensions and shapes of windows and bodywork of the car.
C. Tesla Measurements
Table. 3 summarizes the measurements of the maximum exposure levels \begin{equation*} E_{avg} = \sqrt {DC}\cdot E_{max}\;({\mathrm {\text {V}/ \text {m}}})\,. \tag {3}\end{equation*}
Frequency spectrum of the signal emitted by the mmWave radar at the headlight of Tesla Model S.
For the Bluetooth devices on the dashboard inside the vehicle, we found the highest exposure levels at the driver and passenger positions when the Bluetooth device was connected and communicating with a phone (playing Spotify), measuring 2.2 V/m (3.6% of the ICNIRP limit) and 1.2 V/m (2% of the ICNIRP), respectively, over twice the maximum values in other conditions and lower in the passenger’s seat than in the driver’s seat. However, it is important to note that in the Bluetooth connected case, the field is not solely from the vehicle’s Bluetooth device but also from the Bluetooth antenna of the phone placed between the two seats. Bluetooth signals were detectable even when the Tesla was off, with signal strength and DC nearly at the same level as when the car was on and the Bluetooth device was idle. During the pairing search, although signal strength remained almost unchanged from the idle state, the DC was the highest at 7.8%, even exceeding that in the connected state, doubling the average exposure level from the idle state but still lower than in the connected state.
In this study, we measured the exposure levels at various distances from the integrated LTE module installed inside the vehicle’s rearview mirror. Through broadband frequency-domain measurements, we observed the LTE1800 signal on the spectrum, including the uplink (UL) at 1720.3 MHz and the downlink (DL) at 1815.3 MHz. The maximum signal amplitude measured was 3.17 V/m (5.5% of the ICNIRP limit), with the probe in direct contact with the rearview mirror housing, i.e., 0 cm, then decreasing with distance. The DC of the LTE signal measured in various cases was between 5-8%, yielding a maximum average exposure level of 0.91 V/m (1.5% of the ICNIRP limit), only a third of the maximum exposure level. Similar to in-vehicle Bluetooth devices integrated LTE modules continue to transmit signals with unchanged amplitude and DC when the vehicle is off. So pedestrians around the vehicle or passengers inside could potentially be exposed to the RF EMF from these devices.
In [27], the authors measured the spatial average power density of the vehicle’s 60 GHz mmWave radar, finding around 5 W/m2 at 30 mm from the radar, significantly higher than 0.29 W/m2 measured 20 cm from the front lights in this study, possibly due to the larger path loss at mmWave frequencies. This difference is also due to [27] setting the radar to transmit with an input power of 10 dBm with a 100% DC for worst-case scenarios, whereas the actual maximum transmission power of the radar is considered here. Reference [52] measured the electric field distribution inside a vehicle with a case of the phone turned off and another case which is during a Bluetooth-connected call. The maximum exposure level found at a measurement point above the driver’s seat for the 2.4 GHz-2.8 GHz frequency band used by Bluetooth technology was 218.1 mV/m, corresponding to the Bluetooth call case. This result is not on the same order of magnitude as the highest field strength found in this study, likely due to differences in the devices’ maximum transmission power among manufacturers, the considered configurations, and devices not necessarily operating at maximum transmission power in real scenarios, thus only a qualitative comparison is made with [52]’s results. Similar to our findings, [52] found that switching from off (i.e., the idle state in our study) to Bluetooth calling, the exposure level nearly tripled. Reference [16]’s maximum measurement of the field produced by Bluetooth technology at the driver’s trunk was 0.61 V/m, lower than our findings at the driver’s position and similar to our measurements at the passenger’s position. It is important to note we do not know the operating state of the in-vehicle Bluetooth device during [16]’s measurements.
D. Cumulative Exposure Levels
Our study has found that individual devices —whether they be RSU communication modules, OBU systems, or Tesla’s onboard technologies —consistently operate below the ICNIRP reference levels for general public exposure. However, the real-world scenario often involves the simultaneous functioning of multiple devices which could potentially amplify the cumulative exposure. Therefore, it is necessary to assess the cumulative exposure levels of road users in V2X scenarios, across various technology frequencies, based on our experimental measurements. To facilitate the assessment of this cumulative exposure, we have defined a cumulative exposure index \begin{equation*} I_{cum} = \sum ^{\mathrm {2~\text {G}\text {Hz} }}_{i\gt {\mathrm {30~\text {M}\text {Hz} }}}\left ({{\frac {E_{max,i}}{E_{RL,i}}}}\right)^{2}+\sum ^{\mathrm {300~\text {G}\text {Hz} }}_{i\gt {\mathrm {2~\text {G}\text {Hz} }}}\left ({{\frac {S_{max,i}}{S_{RL,i}}}}\right) \tag {4}\end{equation*}
It is evident that higher exposure levels across frequencies result in a higher cumulative
Fig. 16 presents the cumulative exposure indices
Cumulative exposure indices for drivers, passengers, and pedestrians for each wireless technology, where RSU data are used from in-situ measurements.
Our analysis highlights the significant role that mmWave radars play in the cumulative electromagnetic exposure within the smart connected car ecosystem. However, this study did not assess the electric field strength produced inside adjacent vehicle cabins by the mmWave radars. The study by Ruddle [23] employed the power balance method to estimate the electric field strength at frequencies of 24/46.8/77 GHz, coupling through car windows into the vehicle interior. In a worst-case scenario, where a vehicle is irradiated by radars on all four sides (i.e., front and rear, left and right), at the frequency of 77 GHz with an EIRP of 50 dBm, the coupled power density inside the vehicle was found to be 3.55 W/m2. Considering all different frequency radiation sources both inside and outside the vehicle (e.g., 0.9-2.1 GHz mobile phones, 2.4 GHz Bluetooth, 5.8 GHz toll beacons), the 77 GHz radars contributed 44.5% to the total cumulative exposure inside the vehicle. Although the [23]’s assumption of uniformly distributed and normal incidence of external sources at the car windows may overestimate the exposure levels, the prominent role of mmWave radars in exposure assessments both inside and outside vehicles is evident.
Importantly, the advent of 5G/6G technologies for V2X communications will prominently feature the mmWave bands, such as the 5G NR’s FR2 band range from 24.25 to 52.6 GHz. Moreover, the integration of Multiple-Input Multiple-Output (MIMO) antennas for these innovation technologies, which employ sophisticated beamforming techniques, poses new challenges for the assessment of EMF exposure. These antennas’ ability to direct power more precisely may alter exposure patterns, necessitating refined measurement and modeling techniques to quantify EMF in dense traffic environments accurately. Future studies should therefore focus on developing new methodologies for reliably assessing the EMF exposure contributed by these advanced communication technologies. The integration and cumulative effects of multiple sources of radiation in the V2X ecosystem should also be considered.
Conclusion
This study proposes a comprehensive assessment of RF exposure from ITS-5.9 GHz V2X technologies, including RSUs, OBUs, and integrated systems in a Tesla Model S including DC and power levels. Our combined approach of numerical dosimetry and experimental measurements confirms that RF exposure levels are well below ICNIRP guidelines across all scenarios investigated, with the highest local SAR10g and WBS at 5.9 GHz not exceeding −8 dB and −18 dB, respectively, of the limits specified by ICNIRP. Table. 4 summarizes the maximum exposure levels found across all measurements of this study, compared against the ICNIRP reference levels (in terms of dB of ICNIRP limits). The exposure from Tesla is predominated by its mmWave radar, which is well within safe limits, underscoring the safety of these technologies. These findings validate the effectiveness of current V2X communication systems and support their safe integration into intelligent transportation systems. Future work will consist of exposure assessment from forthcoming 5G and 6G technologies for connected car mobility.