Mobile QR Code

1. (Department of Information and Communication Engineering, Changwon National University / 20 Changwondaehak-ro, Uichang-gu, Changwon-si, Gyeongsangnam-do 51140, Korea )

isible light communication, Optical camera communication, Intelligent transport system

## 1. Introduction

With the recent advances in solid-state lighting technology, light emitting diodes (LEDs) offer greater energy efficiency, a longer lifespan, more durability, and excellent illumination quality compared to conventional incandescent lamps. Due to these exclusive properties, LEDs are used for lighting purposes almost everywhere. Additionally, they are switchable at various light intensity levels at a very fast rate, and thus, data can be encoded using an LED driver and can be used for information transmission. Recently, visible light communication (VLC) [1-3] that allows simultaneous communication between any two devices equipped with LEDs and a photodetector is widely addressed in various research areas.

VLC can achieve data rates up to several gigabits per second, which is comparatively faster than conventional radio frequency (RF) communications. Although there are several approaches to efficient resource allocation and reuse of the available RF spectrum, its limited availability makes it difficult to meet the increasing demands of cellular traffic in today’s world. Therefore, VLC can be a viable solution to overcome the drawbacks of RF communication and designing high-capacity wireless networks.

Optical camera communication (OCC) [4,5] is a subset of VLC that operates in the visible light frequency band with an LED and a camera acting as a transceiver pair. To establish the communication link, OCC technology can utilize readily available LED lights with off-the-shelf cameras, digital cameras, rear cameras in vehicles, surveillance cameras, etc. Because OCC is feasible for a large body of applications, it can be applied to the Internet of Things (IoT), to indoor localization, and to vehicular communication in an intelligent transport system (ITS). In an OCC system, information is sent from the transmitter by varying the intensity of the LED. Likewise, the image sensor used at the receiver captures the LED state and position, thereby decoding the data from multiple-pixel information in the captured image.

The massive number of road accidents occurring every day costs thousands of lives, mostly due to irrelevant and misunderstood information between vehicles. Therefore, to diminish the number of casualties, there is increasing demand for efficient and relevant communication between vehicles and road-side units (RSUs) such as digital signage and display boards. Several vehicular communication technologies, including conventional RF [6,7] and vehicular ad-hoc networks (VANETs) [8], are widely considered in the ITS field. However, they cannot provide robust solutions to several complications, such as quality of service (QoS), high connectivity, and high-bandwidth requirements. Besides, vehicular communication using RF technology is not feasible in some places, including electromagnetically sensitive areas and certain architectures, such as tunnels and underground roads.

OCC-based ITS [9,10] takes advantage of LEDs and cameras equipped in both vehicles and road infrastructures. The headlights and rear lights of a vehicle can be used as data transmitting devices. Traffic lights and digital signage on the roads can also broadcast OCC data containing ITS information via the visible light spectrum. Dashboard cameras and surround-view cameras mounted on vehicles can be used to receive the transmitted OCC data. Likewise, surveillance cameras used for license plate recognition and lane detection also serve as a receiver for OCC data. Considering the easy availability of transmitter/receiver pairs, it is efficient to implement an OCC system for vehicular communications. Besides, the installation cost is comparatively less than RF, and the OCC system utilizes 300 THz of license-free bandwidth with no network usage cost. This makes a vehicular communication system with OCC much more reliable, easier, and efficient to implement in the real-world ITS scenario.

Several challenges must be solved for OCC technology to be used in vehicular communication. To establish an OCC link, visible lights must be within the field of view (FOV) of the receiver's camera, and depend not only on the lens but also on the sensor. Visual information challenges, including the impact of natural conditions, such as snow and rain [11,12], smoke and fog [13,14], in the optical channel have a negative impact on the performance of a vehicular communication system. Despite these challenges, OCC-based vehicle communication is expected to be in high demand, because it reduces the cost of establishing a ubiquitous and stable communication system for an ITS.

In OCC-based ITS technologies, vehicle-to-vehicle (V2V) communication [15-20] considers information transmission from one vehicle to a nearby following vehicle. Similarly, in vehicle-to-infrastructure (V2I) communication [10, 21-27, 35], RSUs and LED traffic lights transmit information to an onboard receiver camera equipped on a vehicle. In vehicle-to-everything (V2X) communication [28-33], the vehicle communicates with every digital entity with a camera, such as drones and pedestrians wearing smart glasses.

In this paper, we summarize the recent vehicular communication technologies using OCC. The contributions of our paper are as follows.

· We explain the various types of recent technologies explored in the OCC-based vehicular communication model.

· We explore the various OCC system designs and the impact of the transmission channel on system performance.

The rest of the paper is organized as follows. In Section 2, the system model for OCC-based vehicular communication is discussed. Section 3 presents various types of vehicular communication technologies that use an LED and a camera for data communication. In Section 4, we discuss the channel characteristics and design challenges of the system. Finally, we present some suggested future work in Section 5, followed by the conclusion in Section 6.

## 2. System Model

The concept of OCC-based vehicular communication is illustrated in Fig. 1. The candidate for the transmitter can be traffic LED lights, vehicle tail lights, or any commercial digital display unit on the road. Likewise, the receiver can be a dashboard camera mounted on the front of the vehicle. In V2V communication, vehicles exchange driving information like speed, braking conditions, and upcoming traffic information. Likewise, in V2I communication, roadside infrastructure with LEDs, such as traffic lights and digital signage, transmit road-state information, such as public broadcasts and accident warning messages, to vehicles nearby. And, in the V2X scenario, the vehicle communicates with every digital entity in the surrounding area to send and acquire road safety information.

The system architecture for OCC-based vehicular communication is presented in Fig. 2. The input data to be transmitted are in binary form, are subjected to channel coding, and are followed by modulation to convert them into optical signals. The optical signals are transmitted from vehicular LEDs, and the visible light spectrum of license-free bandwidth carries OCC information with no network usage costs. When the transmitted signal passes through the optical channel, several types of noise occur, such as shot noise from photon emissions of the LEDs and additive white Gaussian noise (AWGN). Besides, various natural conditions, such as smoke, fog, and rain, impair the transmitted OCC signal. At the receiver, the image region that includes the OCC signal is extracted from the received image through image processing. Then, the received signal is demodulated, followed by channel decoding to retrieve the originally transmitted binary data.

## 3. Intelligent Transport Systems (ITS) using OCC

The concept of the ITS is to introduce a cooperative driving environment with exchanges of many kinds of vehicular and roadside information via wireless channels between vehicles and the nearby digital infrastructure. The goal of this system is to provide real-time interaction with drivers, providing information from surrounding RSUs to solve traffic congestion problems, reduce road fatalities, and increase driving safety.

For efficient management of the road transportation system, RF-based communication technologies such as dedicated short-range communication (DSRC), wireless access in vehicular environments (WAVE), and continuous air interface, long and medium range (CALM), now referred to as communications access for land mobiles, have been considered. In [6], state-of-the-art VANET technology was discussed, including the detailed architecture and network topology modeling for an ITS. Here, various routing protocols that can withstand the variable and dynamic VANET topology were discussed. Similarly, Kosmanos et al. [7] proposed the relative speed estimation algorithm (RSEA) for a moving vehicle that approaches a transmitter/receiver pair. However, these RF-based approaches lead to problems related to spectrum scarcity, and they result in high computational costs, signal latency in a dense crowd, and signal degradation due to multipath propagation, which are severe detriments.

Given the limitations of RF technology and the associated inconvenience in congested traffic scenarios, OCC systems can serve as a viable solution to overcome these shortcomings of the ITS. Using the vast and freely available visible light spectrum, an OCC-based ITS solves the problem of channel congestion, and allows handling multiple users at the same time.

There are various scenarios for vehicular communication in an ITS, where vehicles communicate with each other and with RSUs. In V2V communication, only the vehicles communicate with each other to share useful traffic information. In V2I, communications are held between vehicles and the road infrastructure. Similarly, in V2X scenarios, vehicles exchange information between, for example, pedestrians wearing smart glasses, bicycles, and unmanned aerial vehicles (UAVs).

### 3.1 Vehicle-to-vehicle

In an OCC-based V2V communication system, information is exchanged between any two vehicles, where the preceding vehicle transmits information from its LED rear lights, and the following vehicle receives that information with an onboard camera. The common types of information usually shared in V2V communication scenarios are speed, braking conditions, lane transfer information, etc. In V2V communication, the two vehicles communicating with each other should be within range, which is determined by the type of camera and the illumination properties of the LED light source. Fig. 3 presents a diagram of a typical V2V scenario where three vehicles on the road communicate with each other. The two preceding vehicles, B and C, with different field-of-view (FOV) angles ${\theta}$$_{1} and {\theta}$$_{2}$ (where, ${\theta}$$_{2}> {\theta}$$_{1}$), are at distance $\textit{d}$$_{1} and \textit{d}$$_{2}$ ($\textit{d}$$_{1} < \textit{d}$$_{2}$), respectively, from the following vehicle, A. Although vehicle C is at shorter distance from vehicle A, its communication performance is degraded slightly due to the small FOV angle. And although vehicles A and B are in good FOV alignment, their performance is altered due to the long transmission distance. Therefore, in an OCC-based V2V environment, angular alignment and communication distance between the following and preceding vehicles are crucial factors to be considered for robust communication performance.

In [15], a real-world driving experiment was conducted using OCC, where two vehicles were driven a total of 108~km. By utilizing different software and hardware techniques with orthogonal frequency division multiplexing (OFDM), the proposed system reliably achieved a working range of 45 meters. Through extensive experiments, the study showed that multipath propagation had little effect on error performance. Instead, the transmission distance and angle were the main factors determining received power.

A low-complexity and high-reliability prototype for vehicular communication was designed where the receiver in the following vehicle can detect hard braking of a preceding vehicle from a distance of 20 meters [16]. Visible light signals from a brake light are used when emergency messages for hard braking need to be transmitted. Whenever hard braking is detected, the system immediately notifies a vehicle following at speeds of up to 80 km/h, and reduces the chance of an accident.

In real-world vehicular communication, vehicles might suffer dynamic and uneven motion due to erratic road conditions. This might adversely affect the overall communication performance of the V2V scenario. To examine this issue, Ashraf et al. [17] presented a mobility characterization study, where they set up a V2V communication scenario using a constantly illuminated transmitter on the preceding vehicle and a multi-camera setup on the following vehicle. Through extensive analysis of the experimental data, they found that the typical range of horizontal and vertical motion of vehicles is on the order of 40 pixels in a 1920 ${\times}$ 1080 resolution video at 30 frames per second (fps).

In [18], a mirrorless, vehicle-based autonomous driving concept for V2V communication was investigated. In this scenario, the camera sensor is located in the rear-view mirror in the leading vehicle and can provide a much broader view for improving driving safety. Here, a flicker-free LED was used in the following vehicle for simultaneous illumination and data transmission. With its headlight LEDs, the following vehicle can transmit future-action data, such as overtaking, cutting in, caution signals, and vehicle type. The lead vehicle receives the OCC data by using high-speed cameras, and can determine the safest driving route depending on the future-action data transmitted from the following vehicle. With extensive experiments, the proposed system proved to have robust V2V communication performance during both daytime and night-time.

In [19], an optical V2V communication system where the receiver camera employed a special complementary metal oxide semiconductor (CMOS) image sensor called an optical communication image (OCI) sensor was introduced, that consists of pixel arrays suitable for imaging and communication [26]. By using the OCI-based camera, the receiver obtained a 10~Mbps optical signal, and was capable of accurately detecting the LEDs using photodetector (PD) pixels. Also, they achieved real-time LED detection in outdoor lighting conditions by using a flag image for LED detection. The use of a flag image effectively eliminates unnecessary objects in the captured image. Likewise, in [20], the authors presented test results for OCC-based V2V and V2I communication. Flicker-free LEDs were used as light-emitting sources, with high-speed cameras at the receiver. In real-world testing, they obtained successful reception of absolute location data sent by the tail lights of the forward vehicle.

### 3.2 Vehicle-to-infrastructure

In V2I communication, vehicles communicate with nearby infrastructure to exchange information. In OCC-based V2I communication, the transmitter candidates can be vehicular LED lights, digital signage along the road, LED traffic lights, digital lane markings, street lamps, parking meters, digital milestones, etc. Receivers, on the other hand, are cameras installed on the dashboard of the vehicle, surveillance cameras, and special-purpose cameras for lane tracking or license plate number detection.

In either V2I or infrastructure-to-vehicle (I2V) communication, information such as public safety messages about construction work ahead, emergency weather alerts, availability of parking spaces, road accident warnings, and vehicular traffic information are transmitted. The vast number of RSUs with LEDs for displaying vehicular safety information can serve as OCC transmitters.

Fig. 4 shows an OCC-based V2I communication scenario where three vehicles are communicating independently with digital signage and traffic lights on either side of the road. Real-time communication is influenced by various factors, such as vehicular speed, angle between transmitter and receiver, and the range between them. The speed at which a vehicle is approaching a digital RSU also determines performance in OCC-based V2I communication. For a slow vehicle approaching the receiver, it is relatively easy to capture the LED lights of the transmitter. However, for a high-speed vehicle, a high-speed camera capable of capturing multiple frames per second is required to effectively capture the transmitted content.

In [10], a V2I broadcast system that integrates traffic lights with the existing ITS architecture was presented. Here, a prototype with a 200 mm custom-designed LED traffic lighting system was used. A robust modulation technique based on direct sequence spread spectrum (DSSS) and sequence inverse keying (SIK) was used to minimize the effect of noise. They compared their scheme with existing modulation schemes, such as L pulse position modulation (L-PPM), inverted L-PPM, and on-off keying (OOK). They found that their scheme provided better performance than conventional ones in terms of BER and maximum achievable data rate. Furthermore, through extensive experiments under bright sunlight during the daytime, they demonstrated robust communication performance at a transmission range of more than 40 m.

Note that the capture rate of the camera plays a significant role in efficient and fast V2I communication. In general, there is a mismatch between the transmit frame rate and the capture rate of the camera, and thus, synchronization techniques [34] should be considered for robust V2I communication. Usually, LEDs are capable of transmitting data at a very high frame rate. Therefore, a moving vehicle in V2I communication with a stationary infrastructure requires very high-speed cameras with high capture rates of more than 30 fps to achieve good data transmission speed. In [21], a V2I experiment on the road was conducted to address the issue of mismatch frames by using high frame-rate cameras at the receiver. An LED array with an emission rate of 500 Hz was used in the transmitter. At the receiver, the emission patterns as a receiving data were captured with an on-vehicle high-speed camera at a capture rate of 1000 fps. The captured images were processed to find the transmitter, to track that transmitter, and to obtain its LED lighting pattern. Extensive experiments were conducted by increasing the transmission distance from 20 m to 70 m. For distances as low as 20-30 m, no errors were observed. However, upon increasing the transmission distance, an error of up to 10$^{-1}$ was incurred.

In [22], a transmitter detection algorithm was proposed for visible-light road-to-vehicle communication. The transmitter and receiver candidates were LED traffic lights at a crossroad and an on-vehicle high-speed camera, respectively. The frame subtraction technique was used to detect the approximate position of the transmitter, which reduced the data by cutting out detected areas. The exact portion of the transmitter was detected by template matching, and data were acquired from the image. Through the experiments, a vehicle at a speed of 30 km/h was able to correctly detect most of the frames, and achieved good communication with high accuracy at distances of less than 35 m.

A study in [23] investigated the I2V and V2I concept for connected vehicles at road intersections with traffic lights. They proposed a smart-vehicle lighting system capable of combining multiple functions, such as illumination, signaling, and positioning. To transmit data, on-off code was used where the encoded data contained a unique emitter ID. At the receiver, the message was decoded and resent to either another vehicle or to the traffic light within the intersection.

Similarly, in [24], a robust range estimation technique for I2V communication was presented by using an LED array and a high-speed camera as transmitter and receiver, respectively. To avoid vehicular vibrations from irregularities in the road, range estimation was achieved by using the phase-only correlation (POC) technique. This technique is a pattern-matching algorithm where the amplitude components of the Fourier-transformed images are replaced with a constant value. They used this technique for position displacement detection of the same object, and achieved a range accuracy of 0.3 meters with a measuring time of two milliseconds on a rough road.

Other studies [25-27] addressed the use of OCI for V2I communication. Yamazato et al. [25] proposed an OCI-based automotive system for both V2V and V2I applications. The system uses an LED array as a transmitter and a high frame-rate image sensor as a receiver. The LED array used as an LED traffic light is set on horizontal ground, and that a high-speed camera is mounted on the dashboard of the vehicle. Multiple field trials were conducted by driving a vehicle at speeds of 30~km/h from distances of 70 m and 30 m towards the receiver. Through the experiment, a data rate of 32 kbps was achieved for real-time audio signal transmission consisting of driving safety information. Similarly, in [26], an OCC system was developed for automotive applications targeting both V2V and V2I communications by using an OCI image sensor. This system incorporated the OCI chip for high-speed signal reception and accurate LED detection. The OCI chip consists of a specialized communication pixel, which is capable of prompt responses to variations in optical intensity. This automotive application-friendly OCC system achieved a data rate of 15 Mbps, and 20 Mbps, with and without real-time LED detection, respectively· Similarly, in [27], the authors proposed an advanced OCI-based OCC system in order to reach the IEEE standardized maximum data rate of 54~Mbps for vehicular communication. Here, an OFDM-based V2I communication system was presented, and the frequency response characteristics and induced circuit noise were analyzed for enhancing the signaling design. This OCI sensor-based vehicular communication system achieved a BER of 10$^{-5}$ and a data rate of up to 55 Mbps when tested at a fixed distance of 1.5 m between the transmitter and receiver.

In [35], a cooperative system using both I2V and V2V communications was proposed, where a commercial LED-based traffic light was used as the RSU. The proposed system consists of a commercial LED traffic light as a transmitter, a transceiver system for receiving the message and retransmitting it to the vehicle behind it, and a second receiver for receiving the message from the transceiver pair. With extensive experiments, the proposed system achieved a BER of 10$^{-7}$ without using any complex error correction codes. The experimental results proved that RSUs can communicate with vehicles outside the service area by practicing multi-hop connections.

### 3.3 Vehicle-to-everything

In V2X communication, vehicles can communicate with any nearby digital entity, including cyclists, motorcyclists, and pedestrians, which are all termed vulnerable road users (VRUs) [36]. Along with the safety concerns of four-wheeled vehicles, safety requirements for VRUs are also an essential part of an ITS. To enhance the safety of VRUs, any vehicle on the road must be promptly available to communicate with them. Owing to the evolution of smart, wearable glasses with cameras and augmented reality (AR) features, pedestrians can now communicate with vehicular LED lights that provide information to alert people about accidents in real time. Fig. 5 depicts a simple V2X environment where all digital entities equipped with either an LED or a camera communicate with each other. The UAV has a camera as a receiver to detect vehicular data, which can be a convenient way of communicating with vehicles in areas with an absence of road-side infrastructure. Similarly, the smartphone camera can be used by the pedestrian to retrieve vehicular information instantly from nearby vehicular LEDs. Therefore, in OCC-based V2X, for vehicles to communicate with every digital entity, they must be equipped with LEDs and cameras.

Several wireless technologies, such as Bluetooth, Wi-Fi, and near field communication (NFC) [37,38], play a non-trivial role in establishing reliable communication in the V2X scenario. These systems can establish communication with almost everything under IoT, such as vehicle-to-home (V2H) communication [39] to control home appliances, vehicle-to-broadband (V2B), and vehicle-to-cloud (V2C) communication [40]. Nevertheless, there are several problems associated with these technologies, such as security concerns, module installation costs, and high power consumption, which create hindrances to using them efficiently in daily life. Therefore, OCC can be a feasible solution to accomplishing efficient and reliable communication in a V2X scenario.

In [29], a V2X prototype was presented that uses a 32$\times$ 32 LED array as a transmitter and a high-speed image sensor as a receiver capable of capturing images at 1000~fps. Here, pulse width modulation (PWM) was used to modulate the LEDs by decreasing the transmission distance from 70 m to 30 m. In the experiment, error-free data transmission was achieved at up to 55 Mbps, which is faster than the RF-based DSRC technique. Similarly, in [30], a color-independent, color space-based visual MIMO was proposed for V2X communication. The objective of the proposed system is to maintain the original color and brightness while sustaining seamless communication at the same time. Two scenarios were considered: multipath transmission with visual MIMO, and multi-node V2X communication. By dividing the communication area on a node basis, simultaneous multi-node communication was achieved without any interference. Also, the authors demonstrated numerical improvements in the symbol error rate (SER) through noise compensation where the reference colors are sent in the transmitted networking information.

Existing studies regarding vehicular communication with UAVs or drones [28,31] fall under V2X communication. In [28], an approach combining OCC with UAVs was proposed to solve the problem of power minimization under communication and illumination constraints. The problems are modeled as the smallest enclosing disk problem and a minimum-size clustering problem. To solve these problems and obtain the optimal location of the UAVs, a randomized incremental construction technique was applied, and a greedy method was applied for obtaining a sub-optimal cell association. The obtained UAV locations and cell associations were iteratively optimized several times to reduce power consumption. The numerical results from the experiments showed more than a 53.8% improvement in power efficiency, compared to conventional algorithms. A small-scale air-to-ground communication channel based on OCC was considered in [31]. That study was based on an investigation of achievable capacity on the downlink channel, where UAVs provided multiple access to a cluster of vehicles on the ground. The camera attached to the UAV communicates with the vehicles where a line-of-sight (LOS) channel is provided between them.

In [32], an artificial intelligence (AI)-based vehicular OCC system was presented that effectively employs a You Only Look Once (YOLO) object detection algorithm for region-of-interest (ROI) tracking under city and highway night-driving conditions. YOLO is capable of tracking multiple light sources on vehicular surroundings from possible transmitters that include nearby vehicles and VRUs. After detecting the objects, the camera can extract essential ROI information from them. With the integration of a neural network-based data decoding system, and AI-based error correction techniques, the proposed technique effectively improved the data decoding rate in terms of BER by up to 10$^{-4}$. Likewise, a study in [33] proposed a new system integrating OCC and convolutional neural networks (CNNs) [46] for an intelligent internet of vehicles (IoV). In the IoV platform, a vehicle can receive ITS information from any entity connected to the internet. In this research, the purpose of using the CNN is for precisely detecting and recognizing the LED patterns, regardless of a long transmission distance and weather extremities. Here, the authors proposed an algorithm for extracting the region of interest from the captured LED patterns, and they applied stereo vision techniques to find the desired targets in the distance. The simulation results on a real-time video yielded good BER performance, and demonstrated the efficiency of the algorithm for automotive applications.

## 4. Channel Characteristics and Challenges

In a vehicular communication environment with OCC, the optical wireless network topology is related to various factors, including FOV angle and the distance between transmitter and receiver. Since communication is completely outdoors, the performance will be influenced by several natural factors, such as rain, smoke, fog, etc. Additionally, the communication link between two vehicles should be held in the presence of ambient light sources such as sunlight during the daytime. Lights from roadside infrastructure and from other vehicles during night-time can have a particularly great influence on communication performance. The presence of ambient light overexposes the image sensor of the receiver camera, and hence, it disturbs capturing complete information from the LED transmitter. Therefore, the performance of OCC-based vehicular communication is dynamic, and greatly depends upon the channel characteristics.

Unlike RF, visible light does not suffer from multipath propagation. In contrast, it undergoes a narrow beam path of LED lighting, providing a very high signal-to-noise ratio (SNR). Besides, the OCC channel is secure and not affected by electromagnetic interference. Because OCC operates on an optical wireless channel, it provides a lot of bandwidth for data transmission, which is entirely free of cost. On the other hand, when the transmission rate of the LED is higher than the capture rate of the camera, the synchronization problem [41] affects the performance of an OCC-based vehicular system. Typically, the capture rate of the commercial image sensor is 30 fps. Although that is enough for capturing pictures and videos, it cannot achieve high data-rate communication. Flickering [42] is another challenge in OCC, which arises when the transmission rate of the LED transmitter is significantly low. When the transmission rate is not fast enough, the light signals are easily detected by the human eye as flickering.

LED lighting is vulnerable to natural factors, such as snow, fog, smoke, and rain, which create a severe performance drop in the OCC system, even within a short transmission distance. Fog contains small water droplets that disturb the passage of light due to reflection, refraction, or scattering [13]. Furthermore, in snowy and rainy weather, the moisture in the air reduces visibility and creates disturbances while capturing data with the receiver camera. A study in [14] addressed this issue by simulating these environmental factors under LOS or non-LOS (NLOS) conditions with respect to distance and angular variations. Extensive experiments showed reliable data transmission and a fair SNR in the presence of different environmental deterrents. In [43], an OCC system was emulated inside a laboratory to model atmospheric turbulence by using heat-induced turbulence. They found that turbulence does not significantly affect the system, but rather, the attenuation caused by fog was responsible for degrading the signal quality. Therefore, by increasing the gain in the receiver camera by 16 dB, the optical power loss and quantization noise were reduced.

Vehicular communication with OCC is also subjected to multiple types of interference, including shot noise emitted by photons, natural lighting conditions, and ambient lighting, which influence the communication channel. Extreme exposure of the receiver camera to sunlight creates high saturation problems, and eventually blinds the photoelement, so the camera cannot capture accurate positioning of the LEDs. Likewise, thermal noise and shot noise associated with photon emissions from LEDs also attenuate the signal strength of vehicular communication. To study issues concerning the negative effects of noise on the transmitted signal, Cailean et al. [44] proposed the use of Manchester and Miller coding approaches with OOK. Here, the effect of noise on the pulse width of the data for both those coding approaches was analyzed. With extensive simulations, by encoding messages with two different codes for different noise levels, they found that the Miller-coded pulse is less affected by distortions caused by noise.

## 5. Future Work

Optimization of the vehicular communication model with OCC in terms of transmission range and accuracy is still a field of study to explore. To achieve higher data rates, using high frame-rate cameras is usually costly to implement. With the prevalence of AI, we can adopt several machine learning technologies [45] for signal modulation and coding, which will help to maintain a reasonable data rate, despite a low frames-per-second capture rate in the camera. CNNs, a deep learning (DL) framework have the ability to extract features and patterns, such as lines and gradients, by processing structured arrays of data such as images. Therefore, they can be employed for classification and segmentation, where ROI segmentation of the captured raw image from LED light is required. Normally, for a long transmission distance in OCC-based vehicular communication, the LED patterns of the transmitter cannot be correctly identified, and require a high-end CMOS image sensor camera with complex image processing. In that scenario, a multi-layer deep CNN with the capability of extracting vital and more significant features from captured data can be used for accurately identifying the LED patterns, even for long-distance transmission. Similarly, in an ITS, an image consisting of unrecognizable LED positions captured from a very small (or large) FOV can be enhanced by using high-end image processing techniques with computer vision algorithms [47]. Furthermore, in OCC-based vehicular communication, the signal power at the receiver is usually attenuated due to several noise factors present on the transmission channel. Therefore, unsupervised learning techniques such as autoencoders [48], which learn efficient data representations by training the network to ignore signal noise, can be suitable for SNR enhancement.

A lot of research based on DL approaches [49,50] has been conducted for error rate decrement, for reducing flicker, to sustain constant illumination levels, for signal detection, and for channel estimation in VLC systems. These types of techniques can be adopted in ITS applications for OCC systems. Also, DL-based, end-to-end, multi-colored VLC transceivers [51], and binary VLC transceivers [52] that support universal dimming constraints, can be adopted to design end-to-end transceiver networks in OCC systems for vehicular communication. Besides, in an OCC-based ITS, the information for vehicle safety is publicly broadcasted, and there is the chance of a privacy leak, of eavesdropping, and of unauthorized access to transmitted information, which is a threat to security [53]. Likewise, Rohner et al. [54] studied various techniques for physical layer security in visible light communication. However, this has not been widely explored in the OCC-based vehicular communication environment yet. Therefore, these aspects can be included as research of interest, and can be properly explored as upcoming work in the area of vehicular communication with OCC.

## 6. Conclusion

In this paper, we reviewed recent OCC-based vehicular communication technologies in the ITS. Different types of OCC scenarios in an ITS were addressed, including V2V, V2I, and V2X. The OCC channel characteristics and their effect on the performance of vehicular communication were also explored. We identified possible future work for the advancement of vehicular communication in an ITS. We believe that OCC with high bandwidth and information-carrying capacities will make a crucial contribution to the future of autonomous vehicles and smart traffic environments, offering cheap, reliable, and efficient communications.

### ACKNOWLEDGMENTS

This research was supported by a National Research Foundation of Korea (NRF) grant funded by the Korean government (NRF-2019R1A2C4069822).

### REFERENCES

1
Pathak P. H., Feng X., Hu P., Mohapatra P., 2015, Visible Light Communication, Networking, and Sensing: A Survey, Potential and Challenges, IEEE Communication Survey Tutorials, Vol. 17, No. 4, pp. 2047-2077
2
Salian P. P., Prabhu S., Amin P., Naik S. K., Parashuram M. K., 2013, Visible light communication, Proc. - 2013 Texas Instruments India Educ. Conf. TIIEC 2013, pp. 379-383
3
Cevik T., Yilmaz S., 2015, An Overview of Visible Light Communication Systems, International Journal of Computer Networks and Communications, Vol. 7, No. 6, pp. 139-150
4
Saeed N., Guo S., Park K. H., Al-Naffouri T. Y., Alouini M. S., 2019, Optical camera communications: Survey, use cases, challenges, and future trends, Physical Communication, Vol. 37, pp. 1-17
5
Lain J., Yang Z., Xu T., Dec. 2019, Experimental DCO-OFDM Optical camera communication systems with a commercial smartphone camera, IEEE Photonics Journal, Vol. 11, No. 6, pp. 1-13
6
Hossain E., Mamun N., Faisal M. F., 2017, Vehicle to Vehicle Communication Using RF and IR Technology, 2nd International Conference on Electrical Electronics Engineering (ICEEE), Rajshahi, pp. 1-5
7
Kosmanos D., Argyriou A., Maglaras L., Nov. 2019, Estimating the Relative Speed of RF Jammers in VANETs, Security and Communication Networks, Vol. 2019, pp. 1-18
8
Ur Rehman S., Khan M. A., Zia T. A., Zheng L., 2013, Vehicular Ad-Hoc Networks (VANETs) - An Overview and Challenges, EURASIP Journal on Wireless Communications and Networking, Vol. 2013, No. 3, pp. 29-38
9
Masini B. M., Bazzi A., Zanella A., 2018, Vehicular visible light networks for urban mobile crowd sensing, Sensors, Vol. 18, No. 4, pp. 1177
10
Kumar N., Lourenço N., Terra D., Alves L. N., Aguiar R. L., 2012, Visible light communications in intelligent transportation systems, 2012 IEEE Intelligent Vehicle Symposium, pp. 748-753
11
Hasirlioglu S., Riener A., 2017, Introduction to rain and fog attenuation on automotive surround sensors, 2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC), Yokohama, pp. 1-7
12
Zaki R. W., Fayed H. A., El Aziz A. A., Aly M. H., 2019, Outdoor visible light communication in intelligent transportation systems: Impact of snow and rain, Applied Sciences, Vol. 9, No. 24, pp. 5453
13
Kim Y. H., Cahyadi W. A., Chung Y. H., 2015, Experimental Demonstration of VLC-Based Vehicle-to-Vehicle Communications Under Fog Conditions, IEEE Photonics Journal, Vol. 7, No. 6, pp. 1-9
14
Joshi K., Roy N., Singh G., Bohara V. A., Srivastava A., 2019, Experimental Observations on the Feasibility of VLC-Based V2X Communications under various Environmental Deterrents, 2019 IEEE International Conference on Advanced Networks and Telecommunications Systems (ANTS), GOA, India, pp. 1-4
15
Shen W., Tsai H., 2017, Testing vehicle-to-vehicle visible light communications in real-world driving scenarios, 2017 IEEE Vehicular Networking Conference (VNC), Torino, pp. 187-194
16
Siddiqi K., Raza A. D., Muhammad S. S., 2016, Visible light communication for V2V intelligent transport system, 2016 International Conference on Broadband Communications for Next Generation Networks and Multimedia Applications (CoBCom), Graz, pp. 1-4
17
Ashraf K., Islam S. M. T., Hosseini A. S., Ashok A., 2019, Motion Characterization for Vehicular Visible Light Communications, 2019 11th International Conference on Communication Systems & Networks (COMSNETS), Bengaluru, India, pp. 759-764
18
Jin S. Y., Choi D., Kim B. W., 2018, Optical Vehicle to Vehicle Communications for Autonomous Mirrorless Cars, Journal of Multimedia Information System, Vol. 5, No. 2, pp. 105-110
19
Takai I., Harada T., Andoh M., Yasutomi K., Kagawa K., Kawahito S., 2014, Optical vehicle-to-vehicle communication system using LED transmitter and camera receiver, IEEE Photonics Journal, Vol. 6, No. 5, pp. 1-14
20
Kim B. W., Jeon H. J., Yun S. K., Jung S. Y., 2016, Test Results for V2V and V2I Optical Camera Communications, The Fifth International Conference on Advances in Vehicular Systems, Technologies and Applications, pp. 20-21
21
Premachandra H. C. N., et al. , 2011, Road-to-vehicle Visible Light Communication Using High-speed Camera in Driving Situation, Vol. 459, pp. 459-464
22
Iwasaki S., Premachandra C., Endo T., Fujii T., Tanimoto M., Kimura Y., 2008, Visible light road-to-vehicle communication using high-speed camera, IEEE Intelligent Vehicles Symposium, pp. 13-18
23
Vieira M. A., Vieira M., Louro P., Torre Q., Caparica M., Vieira P., 2018, Vehicular Visible Light Communication I2V2V2I connected cars, The Ninth International Conference on Sensor Device Technologies and Applications, pp. 175-180
24
Yamazato T., et al. , 2016, Range estimation scheme for integrated I2V-VLC using a high-speed image sensor, 2016 IEEE International Conference on Communications Workshops (ICC), Kuala Lumpur, pp. 326-330
25
Yamazato T., et al. , 2014, Image-sensor-based visible light communication for automotive applications, IEEE Communication Magazine, Vol. 52, No. 7, pp. 88-97
26
Takai I., Ito S., Yasutomi K., Kagawa K., Andoh M., Kawahito S., Oct. 2013, LED and CMOS Image Sensor Based Optical Wireless Communication System for Automotive Applications, IEEE Photonics Journal, Vol. 5, No. 5, pp. 6801418-6801418
27
Goto Y., et al. , 2016, A New Automotive VLC System Using Optical Communication Image Sensor, IEEE Photonics Journal, Vol. 8, No. 3, pp. 1-17
28
Yang Y., Chen M., Guo C., Feng C., Saad W., 2019, Power Efficient Visible Light Communication With Unmanned Aerial Vehicles, IEEE Communication Letters, Vol. 23, No. 7, pp. 1272-1275
29
Yamazato T., Dec 2017, V2X Communications with an Image Sensor, Journal of Communications and Information Networks, Vol. 2, No. 4, pp. 65-74
30
Kim J. E., Kim J. W., Park Y., Kim K. D., 2016, Color-Space-Based Visual-Mimo for V2X Communication, Sensors, Vol. 16, No. 4, pp. 591
31
Amantayeva A., Yerzhanova M., Kizilirmak R. C., UAV Location Optimization for UAV-to-Vehicle Multiple Access Channel with Visible Light Communication, 2019 Wireless Days (WD), Manchester, United Kingdom, Vol. 2019, No. April, pp. 1-4
32
Pham T. L., Shahjalal M., Bui V., Jang Y. M., 2020, Deep Learning for Optical Vehicular Communication, in IEEE Access, Vol. 8, pp. 102691-102708
33
A.Islam , Hossan M.T., Jang Y.M., Apr. 2018, Convolutional Neural Network Scheme-Based Optical Camera Communication System for Intelligent Internet of Vehicles, International Journal of Distributed Sensor Networks, Vol. 14, No. 4, pp. 1-19
34
Kwon T. H., Kim J. E., Kim K. D., 2017, Synchronization Method for Color-Independent Visual-MIMO Communication, 2017 IEEE International Conference on Consumer Electronics ICCE 2017, pp. 458-459
35
Cailean A., Cagneau B., Chassagne L., Topsu S., Alayli Y., Dimian M., 2013, Visible light communications cooperative architecture for the intelligent transportation system, 2013 IEEE 20th Symposium on Communications and Vehicular Technology in the Benelux (SCVT), Namur, pp. 1-5
36
Matsumoto S., Ohhigashi N., Hasuike T., 2016, Developing a Transportation Support System for Vulnerable Road Users in Local Community, Proc. - 2016 5th IIAI International Congress on Advanced Applied Informatics, pp. 797-800
37
Gheorghiu R. A., Cormos A. C., Stan V. A., Iordache V., 2017, Overview of Network Topologies for V2X Communications, 2017 9th International Conference on Electronics, Computers and Artificial Intelligence (ECAI), Targoviste, pp. 1-6
38
Arena F., Pau G., Jan, 2019, An Overview of Vehicular Communications, ~ Future Internet, Vol. 11, No. 2, pp. 27
39
Haines G., McGordon A., Jennings P., 2009, The Simulation of Vehicle-to-Home Systems-Using Electric Vehicle Battery Storage to Smooth Domestic Electricity Demand, Proceedings on Ecologic Vehicles/Renewable Energies, No. 2009, pp. 2-10
40
Pillmann J., Sliwa B., Schmutzler J., Ide C., Wietfeld C., 2017, Car-to-Cloud Communication Traffic Analysis Based on the Common Vehicle Information Model, 2017 IEEE 85th Vehicular Technology Conference (VTC Spring), Sydney, NSW, pp. 1-5
41
Hu W., Gu H., Pu Q., 2013, Lightsync: Unsynchronized Visual Communication over Screen-Camera Links, MobiCom13: Proceedings of the 19th annual international conference on Mobile computing & networking, pp. 15-26
42
Thummaluri U., Kumar A., Natarajan L., 2018, Flicker Mitigating High Rate RLL Codes for VLC with Low Complexity Encoding and Decoding, 2018 IEEE International Symposium on Smart Electronic Systems (iSES) (Formerly iNiS), Hyderabad, India, pp. 209-214
43
Matus V., Eso E., Teli S. R., Perez-Jimenez R., Zvanovec S., 2020, Experimentally Derived Feasibility of Optical Camera Communications Under Turbulence and Fog Conditions, Sensors, Vol. 20, No. 3, pp. 757
44
Cailean A., Cagneau B., Chassagne L., Popa V., Dimian M., 2014, Evaluation of the noise effects on Visible Light Communications using Manchester and Miller coding, 2014 International Conference on Development and Application Systems (DAS), Suceava, pp. 85-89
45
Khan F. N., Fan Q., Lu J., Zhou G., Lu C., Tao Lau A. P., 2020, Applications of Machine Learning in Optical Communications and Networks, 2020 Optical Fiber Communications Conference and Exhibition (OFC), San Diego, CA, USA, pp. 1-91
46
Albawi S., Mohammed T. A., Al-Zawi S., 2017, Understanding of a convolutional neural network, 2017 International Conference on Engineering and Technology (ICET), Antalya, pp. 1-6
47
Yu C., Wang J., Peng C., Gao C., Yu G., Sang N., 2018, BiSeNet: Bilateral Segmentation Network for Real-Time Semantic Segmentation, In: Proceedings of the European Conference on Computer Vision (ECCV), Vol. 11217, pp. 325-341
48
Bhowick D., Gupta D. K., Maiti S., Shankar U., 2019, Stacked Autoencoders based Machine Learning for Noise Reduction and Signal Reconstruction in geophysical data, arXiv: Computational Engineering, Finance, and Science, pp. 1-21
49
Ulkar M. G., Baykas T., Pusane A. E., 1 Nov.1, 2020, VLCnet: Deep Learning Based End-to-End Visible Light Communication System, in Journal of Lightwave Technology, Vol. 38, No. 21, pp. 5937-5948
50
Amran N. A., Dehghani Soltani M., Yaghoobi M., Safari M., 2020, Deep Learning Based Signal Detection for OFDM VLC Systems, 2020 IEEE International Conference on Communications Workshops (ICC Workshops), Dublin, Ireland, pp. 1-6
51
Lee H., Lee I., Lee S.H., Mar. 2018, Deep Learning based Transceiver Design for Multi-colored VLC Systems, Optics Express, Vol. 26, No. 5, pp. 6222-6238
52
Lee H., Quek T. Q. S., Lee S. H., Feb 2020, A Deep Learning Approach to Universal Binary Visible Light Communication Transceiver, IEEE Transactions on Wireless Communications, Vol. 19, No. 2, pp. 956-969
53
Schaub F., Ma Z., Kargl F., 2009, Privacy Requirements in Vehicular Communication Systems, 12th IEEE International Conferrence on Computer Science Engineering CSE 2009, Vol. 3, No. march, pp. 139-145
54
Rohner C., Raza S., Puccinelli D., Voigt T., 2015, Security in Visible Light Communication: Novel Challenges and Opportunities, Sensors & Transducers, Vol. 192, No. 9, pp. 9-15

## Author

##### Lakpa Dorje Tamang

Lakpa Dorje Tamang received his B.Eng. in Electrical and Electronics Engineering from Kathmandu Univer-sity, Nepal, in 2018. He is currently an MSc student in Information and Communication Engineering at Changwon National University, Changwon-si, South Korea. His research interests include optical camera communication, display-to-camera communication, digital image processing, deep learning, and artificial intelligence.

##### Byung Wook Kim

Byung Wook Kim received his BSc from the School of Electrical Engi-neering, Pusan National University, Pusan, South Korea, in 2005, and an MS and a PhD from the Department of Electrical Engineering, KAIST, Daejeon, South Korea, in 2007 and 2012, respectively. He was a Senior Engineer with the Korea Electrotechnology Research Institute, Changwon-si, South Korea, from 2012 to 2013. He was an Assistant Professor with the School of Electrical and Railway Engineering, Kyungil University, Gyeongsan-si, South Korea, from 2013 to 2016. He was an Assistant Professor with the Department of ICT Automotive Engineering, Hoseo University, from 2016 to 2019. He is currently an Assistant Professor with the Department of Information and Communication Engineering, Changwon National University, Changwon-si, South Korea. His research interests include wireless communications, visible light communications, machine learning, and deep learning.