Exposure to RF Electromagnetic Fields in the Connected Vehicle: Survey of Existing and Forthcoming Scenarios

Future vehicles will be increasingly connected to enable new applications and improve safety, traffic efficiency and comfort, through the use of several wireless access technologies, ranging from vehicle-to-everything (V2X) connectivity to automotive radar sensing and Internet of Things (IoT) technologies for intra-car wireless sensor networks. These technologies span the radiofrequency (RF) range, from a few hundred MHz as in intra-car network of sensors to hundreds of GHz as in automotive radars used for in-vehicle occupant detection and advanced driver assistance systems. Vehicle occupants and road users in the vicinity of the connected vehicle are thus daily immersed in a multi-source and multi-band electromagnetic field (EMF) generated by such technologies. This paper is the first comprehensive and specific survey about EMF exposure generated by the whole ensemble of connectivity technologies in cars. For each technology we describe the main characteristics, relevant standards, the application domain, and the typical deployment in modern cars. We then extensively characterize the EMF exposure scenarios resulting from such technologies by resuming and comparing the outcomes from past studies on the exposure in the car. Results from past studies suggested that in no case EMF exposure was above the safe limits for the general population. Finally, open challenges for a more realistic characterization of the EMF exposure scenario in the connected car are discussed.


I. INTRODUCTION
The automotive field is experiencing a fast and pervasive technological innovation that is pushing towards the realization of the new concept of the connected car [1]. Modern and forthcoming scenarios of connected cars comprise (i) vehicles capable to communicate and exchange data with other vehicles, the infrastructure, and pedestrians to share real-time traffic information and alert signals (e.g., in case of car accidents, road interruptions, or obstacles), (ii) vehicles capable to sense outside and inside the cabin to provide driving assistance and to monitor the driver's alertness and vital signs of car passengers, and (iii) vehicles that are equipped with wireless sensors and actuators capable to connect and exchange data with each other and the car's electronic control unit (ECU) through an intra-vehicle network of Internet of Things (IoT). The technologies used to operate such services generate an electromagnetic field (EMF) at different frequencies in the radiofrequency (RF) range, from a few hundred MHz, such as in intra-vehicle IoT communication deployed with ultra-high frequency (UHF) [2], to hundreds of GHz as in radars used for in-vehicle occupant detection [3].
Many of these technologies, such as radars used for driving assistance, parking aid, and collision warning, IoT for intra-vehicle sensor network, and wireless devices used for electronic toll collection (ETC), already have a widespread use and are nowadays present as standard equipment in all new cars, whereas other technologies will see widespread use in the coming years [4]. People in a car and in the car vicinity are thus daily exposed to EMF generated by devices in the car or mounted on the car body and by devices mounted on cars in the vicinity.
The present work gives for the first time a comprehensive survey of RF EMF exposure in the specific scenario of the connected car in the RF range, from 100 MHz to 200 GHz. In our survey, we grouped the technologies according to their purpose of use, namely in (i) technologies for vehicle-toeverything (V2X) communication, (ii) technologies for car sensing, and (iii) technologies for intra-vehicle wireless communication. For each group of technologies, we describe the EMF they generate and the impact they might have on EMF exposure of car passengers and people in the car vicinity, as derived from currently available studies. Open challenges and required RF EMF characterization in the connected car are identified and discussed. If not explicitly indicated, the maximum allowed transmitted power set by the standards is the EIRP calculated as A + G + 10 log [1/DC], where A (dBm) is the measured power output of the device, DC is the duty cycle, and G is antenna assembly gain (dBi).
V2X is operated through two main wireless access technologies, both working in the Intelligent Transport Systems (ITS) band at 5.9 GHz, namely: WiFi for mobility based on the IEEE 802.11p standard used in the US [5] and denoted as ITS-G5 in the European Cooperative Intelligent Transport Systems (C-ITS) initiative [6] and cellular technology for V2X (C-V2X) [7]- [11].
The IEEE 802.11p protocol (Table I, first row) supports medium range (under 1 km), low latency (~2-10 ms) and high reliability communications also in adverse weather conditions (e.g. rain, fog, snow) [12]. Communication is fully distributed among vehicles and/or road side units (RSUs), without the intervention of any infrastructure, neither for resource allocation. Communication is operated in the ITS 5.9 GHz band with a channel bandwidth of 10 MHz; the maximum power that can be transmitted by the device's antenna, as set by the standards, is 33 dBm both in EU and in the US, with 44.8 dBm also allowed in the US for government services.
C-V2X indicates an ensemble of technologies standardized by the 3rd Generation Partnership Project (3GPP) and includes V2X operated through both the Long-Term Evolution (LTE-V2X) [7]- [9] and 5G communication protocol (5G-V2X) [10], [11] and can operate via the infrastructure by using the Uu interface, i.e., the logical interface between the user equipment and the base station (e.g., to handle V2N use cases) or over the PC5 interface, enabling direct communications (also called sidelink communication) between vehicles (i.e., V2V) or between vehicles and other road users (i.e., V2P). In V2N, vehicle connectivity is obtained through the conventional cellular network to enable cloud services, such as infotainment and latency-tolerant road safety messages (e.g., information on longer-range road hazards and traffic conditions).
As far as LTE-V2X is concerned (Table I, second row), it supports both 10 MHz and 20 MHz channels, with a maximum transmitted power of 23 dBm. LTE-V2X can operate following two resource allocation schemes, called mode 3 and mode 4 for V2X sidelink communications, i.e., for direct communication through PC5 interface. In both modes the communication between vehicles or road actors is direct, but in mode 3, the cellular infrastructure manages the resource allocation and it works just in coverage, whereas in mode 4 vehicles autonomously select, manage and configure the communication, that can, thus, works also out of coverage.
As anticipated above, C-V2X can be operated also through the 5G NR (New Radio) communication protocol (Table I, third row). The use of 5G NR is required to enhance V2X services for future autonomous driving, which require ultrareliable low-latency communications with high data rate and spectral efficiency. To satisfy the larger bandwidth needs of forthcoming advanced V2X use cases (e.g., in autonomous driving), 5G-V2X has been designed to operate not only in the ITS 5.9 GHz band (like in V2X operated through the LTE protocol) but also in the frequency range 1 (FR1, 410 MHz -7.125 GHz) and the mmWaves frequency range 2 (FR2, 24.25-52.6 GHz).
Similarly to LTE-V2X, also 5G-V2X defines two new modes (modes 1 and mode 2) for the selection of sub-channels in 5G-V2X sidelink communications. These two modes are  the counterparts to modes 3 and 4, however, LTE-V2X only  supports broadcast sidelink communications while 5G-V2X  supports broadcast, groupcast, and unicast sidelink  communications. Specifically, as reported in 3GPP Release 16 [10], 5G-V2X sidelink can be realized through the PC5 interface in the ITS 5.9 GHz band, in the 3.5 GHz (for time division duplex (TDD) devices) or 2 GHz (for frequency-division duplex (FDD) devices) bands for operation at FR1, and in the 28 GHz band for operation at FR2. Although 5G-V2X sidelink supports both FR1 and FR2, no specific optimization has been deployed for FR2 yet and most of the sidelink design refers to FR1. Indeed, we expect the sidelink design to be reengineered when considering the mmWave spectrum of FR2, due to its peculiarities. The channel bandwidth is 20 MHz for 5G-V2X operated at 5.9 GHz and can be as high as 100 or 400 MHz when the service is operated in FR1 and FR2, respectively. Like in LTE-V2X, the maximum transmitted power of the device is limited to 23 dBm.
Finally, ETC (Table I, last row) is a short-range radiocommunication technology between a roadside infrastructure and a vehicle or a mobile platform [13]. In addition to electronic toll collection per se, applications of ETC technology include parking payment, gas (fuel) payment, in-vehicle signing, traffic information, management of public transportation and commercial vehicles, fleet management, weather information, electronic commerce, probe data collection, highway-rail intersection warning, tractor-to-trailer data transfer, other content services, border crossing, and electronic clearance of freight. ETC is operated in the 5.795-5.815 GHz band, with a 0.5 MHz channel bandwidth; the maximum transmitted power of an ETC on-board unit (OBU) ranges from 14 to 21 dBm.

III. TECHNOLOGIES FOR CAR SENSING
Second, we review technologies used by automotive radars mounted on the car body or in the car cabin. Radars mounted on the car body are used in advanced driver-assistance system (ADAS) applications to detect the presence of objects in the vicinity of the vehicle. Recently, radars are being used also inside the vehicle for vehicle occupant detection (VOD) applications, which aim to detect the presence of people inside the car and warn the driver of passengers left in the rear seats when the driver exits the car.
The main characteristics of car sensing technologies, including a description of their application, the relevant standards, and the characteristic of the radiated field (i.e., the operating frequency band, bandwidth, and maximum transmit power) are reported in Table II. In ADAS, radars are mounted on the car body to sense the surroundings of the car and acquire information, such as the distance, velocity, direction, and angular position of objects that are in the radars' range. This information is processed by the central processing unit or field-programmable gate array of the car to provide vehicle control corrections, collision warnings, and to prevent vehicle from accelerating into the front of vehicles or pedestrians ahead. ADAS is the core technology of the forthcoming fully autonomous vehicle. Depending on the specific application, ADAS is deployed through radars with a range from 1 to 250 m. ADAS radars are operated in the 24, 77, and 79 GHz band and can use pulse Doppler or frequency-modulated continuous wave (FMCW) technology [14]- [17]. The limit for the transmit power is typically around 20 dBm; however, depending on the type of technology, it can be as high as 50 dBm.  42 dBA/m. b ERP (effective radiated power): it is the power radiated in the direction of the maximum radiated power. a Specifically indicated for automotive applications [42]. b H-field strength limit at a distance of 10 m.
Recently, automotive radars are being used also in VOD applications (see, e.g., [3], [18]- [26]) to detect how many passengers are in the car and which seats are occupied by passengers. Most VOD applications are able to recognize also the type of passenger in the car, distinguishing adults, little children, or even animals, and can record passenger breathing movements, generating warnings when these movements are different from what is expected from a healthy subject.
Although radars for VOD applications are already available on the market (see e.g., [19]- [26] in last row of Table II), their use is not yet formally standardized. This lack of standards / regulations specific to the use of radars inside a vehicle means that there is not yet a consensus among the manufacturers on the band at which the service is operated nor on the maximum transmitted power. From the review we made of the systems on the market, it appears that VOD radars are typically operated in the 60 GHz [19]- [26], 77 GHz [19], [20] and 140 GHz band [3]; usually, they are operated at significantly lower transmission power than in ADAS application, typically at 12 dBm.

IV. TECHNOLOGIES FOR WIRELESS INTRA-VEHICLE COMMUNICATION
Finally, in this group we review the technologies for wireless connectivity within the car and their typical applications, as listed in Table III.
Current cars are typically equipped with around 60-100 different on-board sensors to monitor the health of vehicle parts and to measure and control the vehicle's asset and performance. The number of on-board sensors is expected to grow in the forthcoming years due to the increasing demand of more efficient and sustainable cars, supporting a higher degree of automation. Traditionally, on-board sensors have been connected to each other and the car ECU exclusively through cables; today, wire-only connection is no more sustainable as the cables needed to connect such a great number of sensors would increase the car weight and costs and diminish the fuel efficiency.
Wireless technology is frequently used in modern cars as a viable solution to replace (whenever possible) the sensors' cables, as for example to replace with a wireless link the cables between the car windows, the mirrors, and the ECUs and to allow monitoring moving and difficult to access parts that can not be reached through a wired link, such as the tires.
Typical applications of intra-vehicle wireless connectivity are vehicle diagnostics, smart car access, in-vehicle control and personalization, infotainment, and in-cabin multimedia transmission [27]- [36]. As to vehicle diagnostics, wireless technology is used for example to: measure the temperature of the brake discs through wireless sensors mounted directly on the wheels of the car; measure in tire pressure monitoring system (TPMS) applications the tire's air pressure and temperature through a wireless sensor mounted on the tire valve; measure the level and flow of the fuel, the strain and vibration of the chassis, the torque of drive train, the engine's valves displacement, the vehicle orientation and dynamics, the acceleration and displacement of the suspension system; in case of an accident, to measure the severity and the location of the impact on the vehicle.
In addition to vehicle diagnostics, wireless intra-vehicle connectivity is used for in-vehicle control and personalization such as in passive entry passive start (PEPS) modules that enable to unlocking the car and starting the engine with a smartphone, key fob or a smart card holding a digital key and for in-vehicle control services that allow the car to automatically recognize the driver's smartphone and activate interior and/or exterior lighting, adjust seating, ventilation and air conditioning settings.
Finally, a common use of intra-vehicle wireless connectivity is for multimedia transmission within the vehicle cabin (e.g., for displaying multimedia contents to the screens of the rear passengers) and for pairing the smartphone to the car's infotainment central unit to access to navigation, music and phone apps through the car dashboard while driving.
The main characteristics of intra-car wireless communication technologies, including a description of their application, the relevant standards, and the characteristic of the radiated field (i.e., the operating frequency band, bandwidth, and maximum transmitted power) are listed in Table III. As reported in Table III, intra-car wireless communication is deployed on current vehicles through Bluetooth low energy (BLE), UHF short range communication, ultra-wide band (UWB) communication, and near-field communication (NFC).
BLE is a short range communication developed by the Bluetooth Special Interest Group and is characterized by ultra-low power consumption and transmission efficiency [37]- [41]. BLE is operated in the 2.4 GHz industrial, scientific and medical (ISM) band and supports up to 100 mW (+20 dBm) transmitted power.
UHF short range communication [42]- [46] is used in automotive connectivity for enabling TPMS and remote keyless entry (RKE) services; it is operated in the 315, 434, and 868 MHz bands and supports up to 10 or 25 mW transmitted power, depending on Regional regulations and operating band.
UWB is a short range communication that uses a relatively large bandwidth of 500 MHz or more and/or a bandwidth that is at least 20% the carrier frequency [47]- [49]. It is operated in the unlicensed 3.1-10.6 GHz band and supports a mean power spectral density of -41.3 dBm/MHz.
Finally, NFC is a short range contact-less technology [45], [50], [51]; by using magnetic induction, it enables the exchange of data and transfer of power between devices by bringing them into close proximity to a distance of a few centimeters. NFC is operated at 13.56 MHz and supports a magnetic field (H-field) limit of 42 dBA/m or 60 dBA/m, depending on the bandwidth of the device.

V. EMF EXPOSURE ASSESSMENT IN THE CONNECTED VEHICLE
In the current Section V, we describe the main evidences from currently available literature on the assessment of the exposure field and the dose absorbed by passengers of cars equipped with the technologies previously described.

A. THE EXPOSURE SCENARIO
Due to its partially closed structure, the car is a peculiar exposure scenario, which can generate standing waves and a loss of the power transmitted in the cabin from a source external to the vehicle (e.g., an antenna mounted on the car roof).
As to standing waves, Hirata and Ida [52] found that the electric field (E-field) induced inside the car by an external plane wave at 10 MHz-1 GHz was enhanced at ~120 MHz, due to the standing waves generated inside the vehicle cabin. Standing waves were suppressed in the 100-200 MHz range when a human body was present inside the vehicle, due to the power absorbed by the car occupant [52]. Also, in the frequency region of standing wave suppression (i.e., at 100-200 MHz), the dose of EMF absorbed by the car occupant at the whole-body level was lower inside the car than in free space [53]. Vice versa, at frequencies outside the 100-200 MHz range, the presence of a car occupant had only a marginal effect on the suppression of the standing waves [52]. The frequency region at which standing waves were suppressed depends on the dimension of the vehicle cabin and the number and size of car windows. The presence of passengers in the car generated a similar suppression effect of also the exposure field: as observed in [54], the average in-vehicle E-field was lower in the presence of passengers than in an empty car.
The car body also causes a loss in the penetration inside the car of fields generated by antennas mounted on the roof of the car or mounted on the road infrastructure. The amount of penetration loss depends on the size of the car and the number and size of the car windows. As observed in [55], RF penetration loss due to the car body could be as high as 3.2-23.8 dB at 600-2400 MHz. As a consequence, it is expected that EMF exposure generated by sources external to the vehicle would be lower inside the vehicle cabin than outside.

B. THE LIMITS FOR EXPOSURE IN THE RF RANGE 100 KHZ -300 GHZ
Absorption of RF EMFs can generate a temperature rise in the body. The exposure limits recommended by the International Commission on Non-Ionizing Radiation Protection (ICNIRP) [56] and IEEE [57] were set to keep the local and core body temperature rise to a safe level. Namely, compliance with these latter limits would provide a protection against potential adverse health effects that are observed when the core body temperature increases over 1 °C and local body temperature increases more than 5 °C for Type-1 tissues (all tissues in the upper arm, forearm, hand, thigh, leg, foot, pinna and the cornea, anterior chamber and iris of the eye, epidermal, dermal, fat, muscle, and bone tissue) and 2 °C for Type-2 tissues (all tissues in the head, eye, abdomen, back, thorax, and pelvis, excluding those defined as Type-1 tissue) [56].
As to core body temperature rise, the basic restrictions for exposure in the RF range 100 kHz -300 GHz [56], [57] are set in terms of the specific energy absorption rate (SAR) at the whole body, that is the power absorbed per unit mass of the entire body. For the general public, the whole body SAR exposure limit is equal to 0.08 W/kg, averaged over 30 minutes of exposure [56], [57]. As to local body temperature rise, the basic restrictions for exposure fields within the 100 kHz -6 GHz range are set in terms of the local SAR averaged over 10 g of mass, with a limit of 2 W/kg and 4 W/kg for local exposure at the head/torso and the limbs, averaged over 6 minutes [56], [57]. For frequencies within the >6 GHz -300 GHz range, as the RF energy is deposited mainly at superficial tissues and not in deeper tissues as for lower frequencies, the basic restrictions are set in terms of the local absorbed power density (Sab), that is the density of the power absorbed over a square 4-cm 2 surface area of the body; the limit is set to 20 W/m 2 , averaged over 6 minutes [56], [57].
In case in which it is not feasible to measure the power absorbed in the body, it is possible to assess compliance with the so-called reference levels that are based on exposure quantities that shall be measured outside the body, that is the incident E-field and H-field strength and the incident power density. For exposure in the far-field zone at frequencies 2 GHz, compliance shall be assessed with either the E-field or H-field or the incident power density reference levels; these latter reference levels depend on the frequency of the emitting source [56], [57]. For frequencies >2 GHz, like in automotive radars and V2X communication, compliance shall be assessed with the reference level based on the incident power density only, which is equal to 10 W/m 2 averaged over 30 minutes, at any frequency >2 GHz [56], [57].  (Table IV) and in the 24-100 GHz band of automotive radars (Table V).

C. METHODS FOR RF EXPOSURE ASSESSMENT
For each study listed in Table IV and Table V, we report all the relevant details, namely the frequency and the power radiated by the investigated RF source, the method used to assess the exposure inside the car, the use-case (i.e., exposure scenario) addressed, the estimated / measured exposure quantity, and the main outcomes. For the sake of clarity, we preferred to put all the analytical details of the studies in Table  IV and Table V, while we resumed in the following paragraphs only the main outcomes by grouping the studies in clusters with similar research aims and setups.
As described in Table IV and Table V, in-cabin field exposure assessment was typically performed through experimental measurements and numerical simulations. In experimental studies (see [62], [64]- [69] in Table IV and [80], [81] in Table V), the aim was to measure inside the cabin of a real vehicle the field exposure generated by one or more RF device placed either outside (e.g., on the car roof) or inside the car (e.g., at the front dashboard). Measurements were done in the empty car (i.e., with no car occupants inside) at typically the driver and passengers' seats to assess compliance with exposure limits in the most critical positions, that is the positions occupied by car passengers. All experimental studies ( [62], [65]- [69], [80], [81]) but [64], measured the E-field: as such, for these former studies, the compliance with exposure limits can be assessed by comparing the measured E-field against the reference E-field level in [56] and [57]. Instead, the latter study [64] measured the SAR inside an anatomicallyrealistic adult human phantom that consisted of a fiberglass shell filled with a material of the same dielectric properties of the human muscle tissue. The phantom was seated inside a real car; SAR was measured by placing the measuring probe in the phantom's head. For this latter study, the compliance with exposure limits can be assessed by comparing the measured SAR against the basic SAR restrictions in [56] and [57].
In numerical simulation studies (see [58]- [61], [63], [70]- [79] in Table IV and [58], [82]- [85] in Table V) the aim was to estimate the exposure field using numerical methods capable to calculate electromagnetic quantities, such as the E-field and H-field, as generated by a simulated source in a simulated exposure scenario.
The numerical simulation approach can be purely analytical or can use computational electromagnetic methods. Any analytical approach relies on a strong simplification of the exposure scenario and does not allow taking into account the potential effects on the exposure field of the real 3D geometrical shape neither of the car nor the emitting antenna. For example, in [58] an analytical approach -the power balance method [87] -is applied to estimate the average E-field strength inside a vehicle cabin in the 0.9-5.8 GHz frequency range. The application of the power balance method relies on the assumption that at frequencies above ~1 GHz, the vehicle cabin can be approximated as an electrically large cavity where the average internal E-field strength is a function of basically only the windows' size and glazing materials [58]. The output of an analytical approach is typically the E-field strength and the power density; compliance with exposure limits shall thus be assessed using reference levels in [56] and [57]. Being based on strong approximations, the analytic approach is useful to gain a first insight on the exposure field when it is not possible to implement more computational demanding methods capable to model and take into account the 3D geometry of the exposure scenario (i.e., the car, the emitting antennas, and the car's occupant(s)).
In addition to the analytical approach described above, numerical simulation of the exposure field can be done also using computational electromagnetic approaches capable to solve Maxwell's equations directly, as the Finite-Difference Time-Domain (FDTD) method [88]. Examples of application of computational electromagnetics for field exposure assessment in the car are in [59]- [61], [63], [70]- [79] in Table  IV and [82]- [85] in Table V. In computational electromagnetics, the exposure scenario is modelled using 3D geometries; the scenario is then discretized in terms of grids and Maxwell's equations are solved at each point in the grid. The scenario typically includes the 3D geometric model of the emitting antenna(s) (that could be as schematic as a simple monopole/dipole antenna or more complex as a 3D model of the entire emitting device, case included), the 3D geometric model of the car (simplified or a realistic CAD model) and the 3D human phantom (simplified as a spheroid or detailed and anatomically realistic, including all organs and body tissues). Computational electromagnetic approaches allow the computation of electromagnetic quantities both in the space around the human phantom and, most important, inside the phantom (and in every tissue/organ), being thus a consolidated and robust methodology to assess the SAR and the Sab in the body and its tissues. The drawback of computational electromagnetic approaches is that they are computationally demanding, especially when the simulated exposure scenario occupies large volumes in space, like in automotive scenarios.  Table IV in three clusters based on the scenario they addressed, that is: specific V2X exposure scenarios (Section D.1) and generic in-car connectivity exposure scenarios generated either by antennas external to the car (Section D.2) or antennas placed inside the car (Section D.3). person at the driver position.

D. IN-CABIN FIELD EXPOSURE ASSESSMENT TO FREQUENCIES USED IN V2X, INTRA-CAR AND GENERIC IN-CAR WIRELESS COMMUNICATION
places inside the cabin.
Numerical simulation using a 3D model of the car, a human phantom seated inside the car, and a phone.
Internal: phone in hands-free use (placed to the right of the steering wheel) and hand-held use (phone held by the front passenger, close to the right side of the head) SAR evaluated over three planes: at the heart, thighs, and eye level. f In hands-free use, the maximum SAR of 56 mW/kg was at heart plane level, in the right arm; in hand-held use, the maximum SAR of 2 W/kg was at the eye plane level, at the right side and 15 mm toward the interior of the head.
Numerical simulation using a 3D model of a car and 4 human phantoms as car passengers.
Internal: one antenna mounted inside the vehicle cabin.
Whole-body and local SAR: the highest SAR was at 400 and 900 MHz (equal to 9.9% and 8.6% of basic restrictions, respectively); SAR at 1800 and 2400 MHz was lower (4.6 % and 3.6% of basic restrictions, respectively). The head and trunk were the regions with the highest SAR at 900, 1800, and 2400 MHz; at 400 MHz, the highest SAR was at the whole-body level.
Harris et al. [ Numerical simulation using a 3D model of a car and human phantoms of adult and child passengers in the car.
Internal: the emitting devices were placed inside the vehicle cabin at the driver and passenger seats. Single-user (i.e., only one device operated at a time) and twouser (two devices operated at the same time) scenarios were evaluated.
Whole-body SAR: the highest value was obtained in the two-user scenarios with two UMTS devices and was equal 2.13 mW/kg.
Local SAR10g: the highest value in the head/trunk was obtained with UMTS and was equal to 25.3 mW/kg; in the limbs, the highest value was obtained with the WiMax device and was equal to 392.6 mW/kg.
The contribution of the Bluetooth device to the SAR is not significant. In all conditions, the highest SAR was observed in the configuration with 1 adult + 3 children as car passengers.
Numerical simulation using a 3D model of a car and a human phantom at the driver position.
Internal: a mobile phone antenna at 2.5 cm from the phantom head. phantom, only the head was modelled. a The paper does not report the radiated power of the analyzed antenna. b The devices under analysis are the built-in Bluetooth module that allows to pair the smartphone with the car dashboard and a smartphone operating in the GSM/UMTS band. No indications are given on the actual power radiated by these devices. c All values reported in this Table were measured during the phone call, with also the car's Bluetooth activated. d GSM and UMTS transmitters were operated in real conditions, i.e., with transmitted power dynamically set by the power control unit of the device. e The simulated devices (mobile phones) are operated at their maximum power level, that is at 600 mW for AMPS, 250 mW for GSM-900, 125 mW for GSM-1800, and 200 mW for CDMA. f SAR computed in a cell of 27 mm 3 of volume, corresponding to about 27 mg of mass. It is not a SAR averaged over 10 g.

1) IN-VEHICLE EXPOSURE FROM SPECIFIC V2X COMMUNICATION
To the best of the authors' knowledge, there are only a very few studies ( [58], [59], [60]) on EMF exposure in specific V2X scenarios, whereas most of the past studies focused on in-vehicle exposure from generic wireless communication used in cars, such as mobile communication (e.g., Global System for Mobile Communications-GSM, Universal Mobile Telecommunications Service-UMTS, and LTE), Bluetooth and WiFi.
As to exposure assessment in V2X scenarios, a first attempt was done by Ruddle [58] (Table IV) that used a simplified analytical approach for evaluating the field coupled inside a car from an external ETC device (a toll beacon) at 5.8 GHz. The ETC device was simulated at the maximum allowable Effective Isotropic Radiated Power (EIRP), at 5 m from the car. Results evidenced that the power density coupled inside the vehicle was 0.005 W/m 2 , that is well below the safety reference level of exposure of 10 W/m 2 for the general population [56], [57].
A more realistic and complex exposure scenario was addressed for the first time by Tognola et al. [59], [60] (Table  IV) who assessed RF exposure in a realistic anatomical human phantom seated at the driver position inside a 3D model of a real city car equipped with V2V external antennas mounted on the roof. The dose absorbed by the driver at the typical V2V ITS 5.9 GHz band was quantified as the SAR. As observed in [59], [60], the dose was mainly absorbed at the most superficial tissues of the body -the skin -and in the head. In the worst-case scenario (that consisted of four antennas operated at the same time and at the maximum EIRP) the dose absorbed by the whole body (0.008 W/kg), at the head/torso (1.58 W/kg), and the limbs (0.76 W/kg) was well below the basic restriction limit for EMF exposure of the general population at 100 kHz-300 GHz, which is equal to 0.08 W/kg for the whole body, 2 W/kg in 10 g tissue for the head/torso, and 4 W/kg in 10 g tissue for the limbs [56], [57].

2) IN-VEHICLE EXPOSURE FROM GENERIC IN-CAR WIRELESS COMMUNICATION -EXTERNAL ANTENNAS
The remaining studies listed in Table IV investigated in-vehicle field exposure generated at the frequencies used in generic wireless communications by external (described in this current Section D.2) and internal antennas (Section D.3).
As to external antennas, studies [61]- [64] in Table IV  addressed [56], [57]. Similarly, the dose absorbed by the passengers in the car was below the basic restrictions for the general public at all bands mentioned above. As a general remark, the highest exposure was observed at the whole body and in the head region [63], [64].

3) IN-VEHICLE EXPOSURE FROM GENERIC WIRELESS IN-CAR COMMUNICATION -INTERNAL ANTENNAS
Studies [58], [65]- [79] in Table IV addressed EMF exposure from antennas and devices operated inside the vehicle, such as mobile phones (GSM 900 and 1800 MHz and UMTS 2100 MHz) and Bluetooth, WiFi, and ZigBee devices.
Field exposure inside the vehicle cabin measured experimentally [65]- [69] or assessed with numerical simulations [58], [63], [70]- [79] was again below the reference level of exposure, at all tested frequencies. The dose absorbed by the passenger closer to the emitting device(s) slightly increased with the number of car occupants (see e.g., [74]) and the number of devices simultaneously used in the car (see e.g., [75], [76]). In any case, the dose of exposure was always below the basic restriction limits [56], [57]. Table V summarizes the outcomes from studies on field exposure at radar frequencies, i.e., 24-100 GHz.

E. IN-CABIN FIELD EXPOSURE ASSESSMENT TO FREQUECIES USED IN AUTOMOTIVE RADAR APPLICATIONS
As a general remark, we did not find any study addressing realistic scenarios of automotive applications of radars. The best approximation of a realistic automotive radar scenario was in [80] and [81]. Study [80] assessed the exposure from automotive radars by applying a horn antenna to measure the increase of the superficial temperature of the human skin and porcine eye. The horn antenna was fed with a continuous wave at 1-10 mW/cm 2 power density at 77 GHz. The measured temperature rise was well below the safe limit of 2-5 °C even at an incident power density of 10 mW/cm 2 (=100 W/m 2 ), which is ten times greater than the reference exposure limit [56], [57]. In [81], a 79 GHz automotive radar is analyzed to measure the power density emitted at 3-30 mm distance in the worst case scenario of 100% duty cycle and maximum output power. The study evidenced that the power density averaged over 1 cm 2 was below the limits for exposure [56], [57]. Numerical simulation using 3D anatomical model of only the eye. The car is not modelled.
Plane wave in air directed towards the eye model.
Energy absorption: the energy is absorbed in the eye tissues in the same way for UWB pulses and CW excitation. At 22-29 GHz and 57-64 GHz the energy is mostly absorbed by the cornea. a The devices are operated at the maximum average power density allowed at a distance of 3 m, as established in [86], that is: 0.0009 W/m 2 for 24 GHz side/rear radar; 0.3 W/m 2 for 46.8 GHz side/rear radar; 0.6 W/m 2 for 46.8 GHz ACC radar; 0.88 W/m 2 for 77 GHz side/rear/ACC radar. b The amplitude of the excitation signal was adjusted to the general public exposure recommendations of a maximum incident power density of 10 W/m 2 [56], [57]. c The papers do not specify the output power of the tested antennas.
Another quite realistic scenario was addressed in [58] that analytically estimated the effect of the car body on the field coupled inside the vehicle cabin. In [58], it is observed that the power density coupled inside the vehicle from a radiating antenna at 24, 46.8 and 77 GHz at the maximum EIRP was 0.76 W/m 2 at a distance of 3 m from the antenna, that is well below the reference exposure limit of 10 W/m 2 [56], [57].
The remaining papers [82]- [85] in Table V addressed extremely simplified exposure scenarios to estimate the temperature rise in 3D numerical models of the ear and the eye in the range of frequencies used in automotive radars. The excitation source was a pulsed plane wave in the air at the ICNIRP maximum power density of 10 W/m 2 [56]. Results evidenced that the temperature rise was of the order of 0.5 °C, that is below the local exposure safe limit of 5 °C for Type-1 tissues and 2 °C for Type-2 tissues [56], [57].

VI. OPEN ISSUES
Despite the massive and pervasive use in modern vehicles and the resulting potential impact on the health of car occupants, we could find only a very few studies that addressed the specific scenario of EMF exposure in the connected car. The majority of past studies focused on the use of generic personal wireless communication technologies, such as mobile phones, Bluetooth and WiFi devices. Only a few studies ( [58], [59], [60]) addressed technologies specific to car connectivity, such as V2V. As for car sensing, we found only studies addressing extremely simplified exposure scenarios. Finally, we could not find any study on the exposure generated by IoT sensors and actuators, specifically used in intra-vehicle wireless networks.
Research on EMF exposure in the connected car shall go deeper in the forthcoming years to address more realistic scenarios to consider aspects not yet investigated, such as: 1) The effect of the combined use of technologies operating at different frequency bands. As described in the current paper, a connected car is a kind of 'ecosystem' where a variety of different wireless technologies are used at the same time, e.g., ADAS radars, antennas for V2X connectivity, intra-car wireless connectivity and infotainment that are operated at different frequency bands. Thus, a realistic assessment of RF exposure in car passengers shall take into account the peculiarity of this multi-source and multi-frequency scenario. At the moment, current studies addressed only single-frequency scenarios. Instead, in situations of simultaneous exposure to fields at different frequency bands (like in the connected car), it is important to assess the compliance with exposure limits not only in each separate frequency band but also as a whole to account for possible additive effects of multiple exposure [56], [57]. As a matter of fact, recommendations in [56], [57] provide specific formulae to assess cumulative exposure by assuming worst-case conditions (i.e., pure additive effects) among the fields from multiple sources.
2) The effect of the number of devices simultaneously used in the car. Differently from use-cases that are already extensively addressed in the literature where the exposure scenario consists of a user and a single source of EMF, e.g. in the assessment of the dose absorbed by using a smartphone or a tablet, the typical exposure scenario in the car is intrinsically a multi-source one. V2X connectivity, for example, requires multiple antennas to be mounted on the car roof or embedded in the windscreen or the external mirrors; in intra-car wireless connectivity, multiple IoT sensors and actuators are placed on different parts of the car; similarly, a typical ADAS implementation requires multiple radars mounted on the car body, e.g., on the bumpers and at the sides of the vehicle. Even by considering the simplest scenario of a single technology (i.e., a single frequency band), the exposure field inside the car varies with the number of devices as it is influenced not only by pure additive effects but also by resonance and interference effects generated by the partially closed structure of the vehicle cabin [52]- [54].
3) The effect of the variability of the exposure scenario, e.g., the effect of the size and shape of the car, the age (children vs. adults and pregnant women) and size of passengers (height, weight, body composition), the position of the devices and passengers in the car (EMF exposure depends on the distance between the source of the field and the person). For example, as to the effect of age, it is well known that the dose of EMF absorbed by a person varies with the person's age as a result not only of the different total body mass (adults are bigger and heavier than children and neonates) but also the different body composition (muscle and fat tissues have different dielectric properties and thus they absorb the field in a different way). As a matter of fact, previous studies (see, e.g., [89], [90]) observed that whole body and local SAR could have higher levels in children than adults, for identical exposure conditions. Because of the massive number of computationally demanding simulations that would be required, characterization of such variability is nearly unfeasible using deterministic dosimetry (as done by the studies in the current survey) and standard techniques of uncertainty propagation, such as Monte Carlo method [91]. Recently, advanced statistic approaches such as stochastic dosimetry and Machine Learning were applied to build computationally efficient surrogate models to assess EMF exposure in complex and uncertain scenarios [92]- [94]. It is thus recommended that future assessment of EMF exposure in the car would address uncertainty and variability of the exposure scenario with such innovative approaches. 4) Last but not least, exposure in the connected vehicle shall address the new forthcoming scenarios that will make use of innovative communication technologies such as 5G and 6G [95] at frequencies scarcely investigates up to now, e.g., mmWaves.

VII. CONCLUSION
This paper summarizes the characteristics and application domain of the main technologies used in the connected car, ranging from technologies for vehicle-to-everything connectivity to technologies for car sensing and intra-vehicle wireless connectivity. The paper also extensively describes the exposure field and the dose of EMF absorbed by passengers of cars equipped with such technologies, including the generic technologies for in-car personal connectivity (e.g., smartphones, tablets, etc.), as derived from current literature. All studies analyzed in the current survey evidenced that in no case the exposure field and the dose absorbed in car passengers were above the safe limits of exposure for the general population. Nevertheless, research on EMF exposure in the connected car shall address in the forthcoming years still open challenges to address more realistic scenarios and to consider aspects not yet investigated, such as simultaneous and combined exposure to field from multiple sources at different frequencies, the effect of the variability of the exposure scenario and the impact of new 5G and 6G technologies and very high frequencies in the mmWave band on EMF exposure in the car. Engineering, National Research Council of Italy, Rome, Italy. Her research interests include the study of the computational modeling of noninvasive brain and spinal stimulation techniques, the design and the optimization of biomedical technologies based on electromagnetic fields (EMF) for diagnostic and therapeutic applications, and the computational modeling of the interactions between EMF and biological systems.