Fundamentals, Key Technologies and Networking of Ultraviolet Non-Line-of-Sight Scattering Communication: A Review

Abstract

Traditional wireless communication signals are often susceptible to physical obstructions and background noise in complex geographical environments or adverse weather conditions, hindering stable and reliable data transmission. Ultraviolet communication (UVC) offers a compelling solution; its unique scattering mechanism and low background noise characteristics facilitate robust communication under non-line-of-sight (NLOS) conditions. At present, there remains a relative lack of comprehensive reviews spanning UVC, including fundamental theory, physical devices, channel models and networking technologies. This review synthesizes the current state of global research, providing a systematic overview of the background, advantages and application scenarios of UVC. It examines the hardware characteristics of light sources and detectors, evaluates NLOS scattering channel models, analyzes key signal processing techniques, including modulation/demodulation, coding/decoding and multiple-input multiple-output technology. Furthermore, this review conducts an in-depth analysis of multi-user networking protocols and three-dimensional topology control mechanisms. Finally, it identifies the prevailing technical challenges and outlines promising directions for future development.

1. Introduction

With the advent of the information age, the explosion of daily human information has driven the rapid development of communication technologies. While modern systems have achieved significant milestones in transmission speed and coverage, maintaining robust, high-quality connectivity in complex or hostile environments remains a formidable challenge. Conventional radio frequency (RF) communications are increasingly hindered by electromagnetic interference (EMI), stringent licensing requirements, and the escalating scarcity of available spectrum [1]. Similarly, although visible light communication (VLC) offers high bandwidth and capacity, its efficacy is inherently limited by a strict reliance on line-of-sight (LOS) links and high susceptibility to ambient light interference and physical obstructions. Infrared (IR) communication, while cost-effective and hardware-simple, suffers from intense solar interference in outdoor settings, restricting its utility primarily to short-range, indoor LOS scenarios [2]. In contrast, ultraviolet communication (UVC)—leveraging its unique low solar background noise and potent scattering properties—presents unparalleled advantages for robust, non-line-of-sight (NLOS) connectivity in challenging environments. Consequently, UVC is rapidly emerging as a pivotal research frontier for next-generation wireless communication systems [3,4,5,6].

Ultraviolet (UV) radiation encompasses electromagnetic waves with wavelengths ranging from 10 nm to 400 nm. Current academic research predominantly focuses on the deep ultraviolet (UV-C) band, specifically the 200–280 nm range [7,8], as illustrated in Figure 1. Due to the intensive absorption by the stratospheric ozone layer, solar radiation in the UV-C spectrum is effectively filtered before reaching the Earth’s surface. This phenomenon creates a “solar-blind” window where background noise is negligible, significantly enhancing signal-to-noise ratios [9]. Furthermore, the interaction between UV light and atmospheric constituents, such as gas molecules and aerosols, induces strong scattering effects. This mechanism facilitates NLOS communication, enabling robust data transmission even in the presence of physical obstructions and eliminating the requirement for precise pointing and tracking. As depicted in the schematic of Figure 2a, even over distances of several kilometers, the inherent scattering properties of atmospheric particles enable UV signals to bypass obstacles. Simultaneously, the pronounced atmospheric attenuation limits signal propagation to localized areas, inherently providing a low probability of interception (LPI) and a low probability of detection (LPD). These physically rooted characteristics endow UVC with its core competencies: exceptional security and high operational flexibility. Moreover, UVC facilitates robust connectivity in electromagnetically sensitive or silent states, such as underground mines, where traditional radio frequency (RF) signals are restricted or prone to severe interference. It also demonstrates significant potential for collaborative formation networking, as illustrated by the application scenarios in Figure 2b,c.

Figure 1. Solar spectrum.

Figure 2. Typical application scenarios of UVC: (a) Schematic of solar-blind UVC based on atmospheric scattering for NLOS connectivity; (b) UVC implementation in electromagnetically sensitive underground mines with RF constraints; (c) Deployment of UVC for collaborative formation networking and coordination among high-mobility platforms.

Based on the aforementioned attributes of NLOS transmission, robust security, and high interference immunity, UVC holds immense potential across a diverse range of critical sectors. Within industrial paradigms, such as smart factories and large-scale workshops, UVC facilitates stable connectivity for autonomous systems, remaining impervious to intense electromagnetic interference and physical obstructions [10]. In contested or radio-denied environments, unmanned aerial vehicle (UAV) swarms can leverage UVC networking to ensure real-time data synchronization and the reliable exchange of flight commands [1,11]. Furthermore, in Intelligent Transportation Systems (ITS), UVC enables vehicles to rapidly exchange kinematic data including velocity and heading, thereby mitigating collision risks through enhanced situational awareness. In the defense sector, UVC stands as a premier candidate for covert tactical communications. Personnel equipped with portable or helmet-mounted UVC terminals can securely transmit intelligence and operational directives while maintaining strict radio silence, thereby minimizing the risk of compromising their tactical positions. Figure 3 is a schematic diagram of the deployment of UVC in cities.

Figure 3. Schematic diagram of UVC application scenarios.

Despite its technical merits, radiation in the UV-C band presents inherent photobiological hazards to human eyes and skin. Consequently, the deployment of UV light sources must strictly adhere to exposure limits prescribed by international regulatory bodies, such as the International Commission on Non-Ionizing Radiation Protection (ICNIRP) and the International Electrotechnical Commission (IEC). In UVC system design, a critical challenge lies in reconciling the need for maximized transmission power and extended range with stringent safety constraints. To resolve this trade-off, communication performance can be optimized via adaptive power control strategies or the integration of ultra-sensitive photodetectors [12], ensuring robust connectivity while maintaining a safe operational environment.

The remainder of this paper is structured as follows. Section 2 reviews the evolution and global research status of UVC. Section 3 presents the specifications and performance of advanced light sources and detectors. Section 4 investigates NLOS scattering channel models and key influencing factors. Section 5 focuses on physical layer optimizations, comparing modulation, coding, and equalization techniques for UV signals. Section 6 extends to multi-user networking, covering multiple access protocols and 3D topology control. Section 7 concludes by identifying technical bottlenecks and outlining future directions for next-generation UVC systems.

2. Current Research Status of Ultraviolet Communication

2.1. International Research Status

The origins of UVC research can be traced back to approximately 1940, when researchers in the United States initiated fundamental theoretical studies on ultraviolet light sources, detectors, and optical filters [9]. Over the subsequent half-century, spurred by rapid technological evolution and the strategic exigencies of modern warfare, the U.S. Navy began exploring the practical utility of ultraviolet light for robust communication systems.

Research on hardware systems is the core premise underpinning UVC systems, and this section begins with the introduction of this part. In 1968, researchers at the Massachusetts Institute of Technology (MIT) constructed an experimental ultraviolet scattering communication system [13]. The transmitter leveraged a xenon flash lamp paired with a searchlight reflector, while the receiver employed a photomultiplier tube (PMT) and a pulse counter. Theoretical analysis and system calibration were conducted over a 26 km link. The study demonstrated that approximately 300 photons per pulse could be detected under clear weather conditions, thereby validating the feasibility of UV scattering for medium-range communication and establishing a rigorous system calibration framework.

In 1986, Geller et al. conducted field trials at the San Diego Naval Base, establishing a UVC link over an initial distance of 0.75 km [14]. The system was designed to support both LOS and NLOS communication modes. Experimental evaluations demonstrated that at a data rate of 2400 bps and a bit error rate (BER) of 10⁻⁵, the system could achieve a maximum communication range of 3 km in LOS configuration and 1 km in NLOS scenarios.

Entering the 21st century, GTE Corporation introduced a mercury-lamp-based UVC system (253.6 nm) utilizing PMT detection [15]. This architecture enabled NLOS links of 1–2 km and directional LOS links of 5–10 km via atmospheric scattering. The research highlighted the system’s robust anti-interference and covertness properties, maintaining a BER of 10⁻⁶ at 4800 bps for distances up to 10 km, further proving the technical viability of UVC for long-range tactical applications.

The Defense Advanced Research Projects Agency (DARPA) catalyzed significant advancements by initiating the Semiconductor Ultraviolet Optical Sources (SUVOS) and Deep Ultraviolet Avalanche Photodetectors (DUVAP) programs in 2002 and 2007, respectively [9]. In 2004, BAE Systems developed a solar-blind UVC system utilizing LEDs developed under the SUVOS program and a quantum-noise-limited photon-counting receiver. Experimental results indicated that the system could achieve NLOS communication over several hundred meters at data rates in the hundreds of kbps with a BER below 10⁻⁷, while LOS links reached kilometer-scale distances at Mbps rates. Subsequently, researchers at MIT conducted a series of trials using 274 nm solar-blind LEDs and channel photomultiplier (CPM) tubes [16]. To overcome the power limitations of individual devices, a large-scale LED array comprising 10 modules (totaling 240 units) was employed in 2005; this configuration achieved an optical output power of 45 mW, facilitating a 200-bps link over a 100 m range.

In 2007, at the MILCOM annual conference held in Florida, BAE Systems unveiled a handheld UVC prototype based on ultraviolet light-emitting diode (UV LED) arrays [16]. The device was powered by eight AA batteries and was capable of facilitating point-to-point (P2P) voice communication, demonstrating the potential for portable, tactical ultraviolet networking in the field. During the same period, the U.S. Army Research Laboratory (ARL), in collaboration with the University of California, conducted extensive research on LED-based solar-blind NLOS communication. In 2008, Chen et al. from the University of California constructed an experimental solar-blind NLOS system utilizing deep UV LED sources and PMT receivers [17]. Their study systematically investigated various elevation angle configurations, including symmetric pairs $({\theta }_T, {\theta }_R)$ of $(30°, 30°)$ and $(45°, 45°)$, as well as asymmetric settings of $(45°, 30°)$ and $(30°, 45°)$. The experimental results demonstrated that a BER below 10⁻⁴ could be achieved within a 100 m range at a signal-to-noise ratio (SNR) of 15 dB. This work validated the feasibility of LED-based solar-blind UVC for short-range, covert NLOS applications.

In 2010, Chen and Xu at the University of California developed an experimental platform to measure the NLOS UVC channel impulse response (CIR) [18]. The setup employed a 266 nm Q-switched fourth-harmonic Nd:YAG pulsed laser as the source, paired with a high-speed PMT receiver and a solar-blind filter. Their research systematically quantified the effects of transceiver elevation angles, beam divergence, receiver field-of-view (FOV), and baseline distance on pulse broadening. The experimental results indicated that the pulse width increases approximately linearly with elevation angles, FOV, and communication distance, while remaining relatively insensitive to variations in the beam divergence angle. Specifically, over distances ranging from 10 m to 100 m, as the elevation angles increased from 10° to 90°, the full width at half maximum (FWHM) of the pulse expanded from approximately 20 ns to 1100 ns. This corresponded to a significant reduction in channel bandwidth from 50 MHz to 0.9 MHz, providing foundational channel characterization data for UVC system optimization. Subsequently, the team further extended their research to explore medium access control (MAC) protocols for outdoor wireless optical networks [19].

In 2020, researchers at the King Abdullah University of Science & Technology experimentally validated the feasibility of NLOS optical communication in turbid underwater channels using a 377 nm UV laser and a PMT receiver [20]. The study demonstrated that in water with an extremely high attenuation coefficient, a data rate of 85 Mbit/s could be achieved. Notably, even under conditions of complete transceiver misalignment, the system maintained a throughput of 72 Mbit/s. These results prove that UV scattering can effectively mitigate the stringent pointing and tracking requirements typically associated with underwater wireless optical communication (UWOC).

At the 2022 International Conference on Unmanned Aircraft Systems (ICUAS), Horyna from the Czech Technical University proposed and implemented a scheme integrating optical communication into a relative positioning system for unmanned aerial vehicles (UAVs) [21]. The architecture utilizes UV LEDs as transmitters and a UV-sensitive camera—equipped with a wide-angle lens and a UV bandpass filter—as the receiver. By employing a 4D Hough Transform for the simultaneous processing of multiple signals, the system achieves integrated local-area communication and positioning without relying on radio frequency (RF) links. Experimental results in both simulated and real-world environments demonstrated that the system maintains a data rate of at least 15 bps over a range of 17–20 m, with the BER kept within 20%. Furthermore, the framework successfully enabled heading consensus and collaborative mission execution in multi-UAV formations through optical signaling.

At the 2023 IEEE Research and Applications of Photonics in Defense Conference (RAPID), Kirste et al. presented high-performance UV-C optoelectronic devices based on the AlGaN material system grown on aluminum nitride (AlN) substrates [22]. This suite of hardware includes 265 nm UV LEDs with output powers exceeding 100 mW and solar-blind avalanche photodiodes (APDs) featuring a quantum efficiency (QE) of 70–80% and a gain of approximately 10⁵. Their research suggests that, by leveraging high-power emission, high-sensitivity detection, and intrinsic solar-blind characteristics, these devices are poised to enable NLOS UVC over ranges of 2–5 km. Consequently, this work establishes a critical hardware foundation for the development of covert, interference-resistant short-range wireless optical communication systems.

In 2025, simulation studies at the University of Rome mainly focused on the performance disparities between underwater UVC and VLC [23]. The simulation results indicated that in a pure seawater channel with a 30 m link distance and 10 W transmit power, the Signal-to-Interference-plus-Noise Ratio (SINR) of ultraviolet light at 360 nm was 4–15 dB higher than that of VLC. Furthermore, the UVC system sustained a spectral efficiency of 1 b/s/Hz using binary modulation at a distance of 170 m, effectively demonstrating its inherent solar-blind advantage against ambient sunlight interference in underwater environments.

Channel modeling is the core theoretical foundation for accurately characterizing UV optical transmission characteristics and quantifying system performance boundaries. In 2012, Kashani at McMaster University, Canada, conducted a theoretical analysis and numerical optimization of free-space optical (FSO) communication systems employing serial and parallel relaying schemes [24]. Based on the lognormal turbulence channel model, the study derived analytical expressions for outage probability and proposed an optimization methodology for relay node positioning. These findings provided important theoretical insights into enhancing the reliability and coverage of optical links in turbulent environments.

In 2017, Uysal’s research team in Turkey investigated multi-hop NLOS UVC systems employing decode-and-forward (DF) relaying [25]. Utilizing a log-normal turbulence channel model, they derived—for the first time—the closed-form probability density function (PDF) for turbulence fading and the analytical expression for the outage probability in NLOS links. The research demonstrated that for a 1000 m link at a BER of 10⁻⁶ and constant transmit power, adding a single relay node yielded a 7.4 dB performance gain, whereas a second relay provided negligible further improvement. Consequently, the study concluded that increasing the number of relay nodes does not inherently guarantee performance enhancement, as overall system efficacy is governed by other configuration parameters.

2.2. Research Status in China

Compared with international research, research on UVC in China started relatively late. Major contributors to this field include the Beijing Institute of Technology, National University of Defense Technology, Chongqing University, Xi’an University of Technology, Beijing University of Posts and Telecommunications, Nanjing University of Posts and Telecommunications, and the University of Science and Technology of China. Key research institutes, most notably the Institute of Semiconductors of the Chinese Academy of Sciences, have also played a pivotal role. Domestic research focuses predominantly on the theoretical foundations of NLOS UVC, the development of atmospheric scattering models, and the characterization of hardware components.

Research on UVC hardware systems in China has undergone rapid development. At the beginning of the 21st century, Beijing Institute of Technology pioneered the field of UVC in China by developing one of the first hardware platforms [15]. Utilizing low-pressure mercury lamps as emitters and GBD-152 photomultiplier tubes (PMTs) as detectors, they achieved NLOS connectivity over distances up to 1330 m. In 2012, Guo et al. constructed a UVC system using a UV laser source and implemented a real-time 4-PPM (Pulse Position Modulation) scheme via a Field-Programmable Gate Array (FPGA) [26]. Their experimental results showed that despite a low pulse repetition frequency (PRF) of only 20 kHz, the system successfully achieved a data rate of 40 kbps with a BER under 10⁻⁴, effectively overcoming the bandwidth limitations typically imposed by low-PRF sources.

In 2015, Huang et al. established an omnidirectional UVC testbed using UV LED arrays and PMTs [27]. By designing specialized omnidirectional optical antennas for both transmission and reception, the system achieved a data rate of 115.2 kbps over a 100 m range with a BER superior to 10⁻⁵. Notably, the measured received SNR showed high consistency with channel model simulations based on the Monte Carlo method, further validating the accuracy of domestic scattering models.

The National University of Defense Technology began the study on UVC in 2000. Jia leveraged frequency loading technology and a half-bridge series resonant circuit to achieve an integrated design for light source excitation and information modulation in 2003 [28]. The experimental results demonstrated that this system successfully increased the baud rate to 9600 Baud, effectively overcoming the limitations of conventional designs where excitation and modulation were separate. This work laid a crucial foundation for the practical deployment of UVC. Two years later, to address the low communication rates caused by the “self-trapping” effect in mercury lamps [29], Jia proposed a high-speed UVC system based on low-pressure iodine lamps. The system achieved a communication rate of 48 kbps over short distances, ten times that of the GTE system reported in 2000, representing a leading performance in the industry at that time.

Chongqing University has made notable contributions to the advancement of UVC hardware systems. In 2007, Lan developed a solar-blind UVC system utilizing an 8 W quartz UV germicidal lamp as the light source [30]. The modulation circuitry was designed around an L6574 electronic ballast chip, incorporating a 4N25 optocoupler for signal isolation and level shifting. Experimental results demonstrated successful data transmission at 1200 bps over an outdoor range of 50 m, establishing a foundational framework for research into solar-blind UV voice communication. Subsequently, Professor Xiao Shali’s team developed a novel solar-blind UV voice communication architecture [31]. In 2010, the group introduced a UV LED-based system with a 10 m communication range. Two years later, they implemented an FPGA-based demodulation method that facilitated connectivity over 5 m [15]. This system achieved a bit rate of 9.6 kbps with a BER of 1% for coded communication, and reached 115 kbps for uncoded transmission.

Xi’an University of Technology, led by Zhao Taifei’s research group, has conducted extensive and productive investigations into UVC technologies. In 2011, Zhao developed a comprehensive wireless optical communication system utilizing solar-blind UV LEDs and OOK modulation [32], followed by rigorous performance evaluations across diverse environmental conditions. The findings indicated that the system BER correlates positively with transmission distance and data rate, while lower background interference significantly enhances performance; furthermore, the study confirmed that relaying mechanisms effectively extend the communication range.

In 2017, the University of Science and Technology of China achieved a significant breakthrough in UVC hardware. Wang et al. developed a real-time NLOS UVC system based on a 266 nm solid-state laser with an output power of 200 mW [33]. This experimental setup successfully attained a data rate of 400 kbps over a 500 m transmission range, while maintaining a frame error rate (FER) below 10⁻⁵. This work represents a landmark achievement, effectively pushing the boundaries of both transmission distance and data throughput in UVC systems.

Nanjing University of Posts and Telecommunications has made significant progress in UVC hardware. In 2023, Qi et al. developed a full-duplex deep ultraviolet (DUV) communication system using a 275 nm DUV LED array as the light source [34], complemented by solar-blind filters and a central processing unit. The transmitter and receiver were integrated into a single package, enabling simultaneous and independent data transceiving. Experimental results demonstrated that at a 10 Mbps transmission rate over a 7 m distance, the system achieved a coverage area of 46 m² with a packet loss rate (PLR) of 1.28% in daylight environments, successfully supporting real-time video transmission and file sharing.

Two years later, Qi designed and implemented a Mobile Full-Duplex Deep Ultraviolet Communication (MDUVLC) system featuring integrated visual tracking [11]. The system utilized a 275 nm DUV LED array for the transmitter and an Avalanche Photodiode (APD) with a solar-blind filter for the receiver. RS coding and OOK modulation were implemented via FPGA, while dynamic target tracking was achieved through a camera and a 3-axis gimbal. Experimental results indicated that the system maintained real-time video communication between a UAV and a ground vehicle over a 12 m distance at 10 Mbps, with a PLR of 1.51%. This work successfully established an all-optical communication network integrating underwater, terrestrial, and space domains, validating its anti-interference and full-duplex capabilities on mobile platforms.

In 2025, Wang et al. from the University of Science and Technology of China developed a LOS UVC system utilizing a wide field-of-view (FOV) LED array as the light source [4]. The transmitter consisted of nine 370 nm UV LED chips with a beam angle of 60°, while the receiver incorporated a PMT and a 370 nm bandpass filter. Indoor experimental results verified that the BER remained below the forward error correction (FEC) threshold within a transceiver angle range of ±60°. In outdoor field tests, the system achieved a data rate of 1.25 Mbps over a distance of 1131 m using OOK modulation and FEC coding. With zero bit errors recorded over a transmission of 65,536 bits, the study demonstrated the feasibility of reliable, kilometer-level UVC without the need for precise alignment, facilitated by wide-FOV LED arrays.

Research institutions in China have also conducted significant studies on UVC. In 2024, Zhao Yingkai from the Institute of Semiconductors, Chinese Academy of Science, proposed and validated a UV weak-light communication system based on 4-level frequency shift keying (4FSK) modulation and a photon-counting algorithm [35]. The system utilized a 254 nm low-pressure mercury lamp as the transmitter and a Hamamatsu H10720-210 PMT as the detector. By employing an FPGA for photon counting, frequency detection, and clock synchronization, the design addressed the high sensitivity requirements of conventional FSK schemes. Experimental results showed that in NLOS scenarios with a 20° elevation angle and 3–4 m obstacles, the system achieved 5 kbps at 65 m with a BER of 7.52 × 10⁻⁴. In LOS conditions, the BER remained below 10⁻⁵ within 400 m and was approximately 10⁻⁴ at 700 m.

In the same year, Jiao Yu from the Institute of Semiconductors, Chinese Academy of Science, proposed a scheme to optimize OOK modulation by adjusting the signal duty cycle to mitigate the increased BER caused by the LED trailing effect [36]. A channel model incorporating path loss and the LED trailing effect was established, and BER expressions for various duty cycles were derived. Using a 100 mW 265 nm UV LED array, a 262 nm filter, and a PMT receiver, field tests were conducted at distances of 500–700 m and rates of 150–300 kbps. The results indicated optimal duty cycles of 48%, 47%, and 45% for 150 kbps, 200 kbps, and 300 kbps, respectively. The study verified that reducing the duty cycle effectively suppresses BER degradation due to the trailing effect in high-speed communication, enhancing overall UV system performance.

In the field of channel modeling, researchers in China have also achieved a series of relevant research accomplishments. Beijing University of Posts and Telecommunications has also made significant contributions to the physical layer technologies and channel modeling of UVC. In 2012, Han et al. developed a dual-PMT diversity reception experimental system using an array of 36 UV LEDs (265 nm) [37]. By employing an equal gain combining (EGC) algorithm, they theoretically modeled and experimentally validated diversity reception for NLOS UVC at data rates ranging from 2.4 to 9.6 kbps. Their experimental results showed that at a BER of 10⁻⁴, the dual-diversity system achieved communication distances of 70–160 m, yielding a performance gain of approximately 2.5 dB over a single-receiver setup and effectively enhancing system transmission performance.

Building on this, Han proposed a photon-tracing-based Monte Carlo simulation method in 2013 to establish an NLOS UVC channel model that accounts for multiple scattering effects [38]. Using a 265 nm source with 43 mW transmit power at 5 kbps, the simulation indicated that a dual-receiver configuration could reduce path loss by approximately 3 dB compared to a single receiver. At a constant BER, the dual-receiver setup significantly extended the communication range. The high degree of consistency between the simulation results and outdoor experimental trends further validated the reliability of the MC model for predicting diversity-reception channel characteristics.

To address the research gap in channel modeling for NLOS UVC in mobile scenarios, Song Peng from Xi’an Polytechnic University proposed a single-scattering channel model based on the Monte Carlo method in 2022 [39]. This model was designed to evaluate the impact of the relative movement speed and direction of communication terminals on the system impulse response. Simulation results—conducted under parameters of a 500 m transmission distance, 30° elevation angles for both transmitter and receiver, a 3° beam divergence half-angle, a 15° receiver FOV half-angle, and speeds ranging from 5 to 100 m/s—indicate that transmission distance remains the dominant factor influencing system response. This conclusion holds true across coplanar, non-coplanar, and varying height scenarios, providing a solid theoretical foundation for the design and analysis of mobile UVC systems.

Networking is the core link to promote UV optical communication technologies from point-to-point link verification to large-scale practical scenario deployment. In 2024, Zhao introduced a UVC-assisted coalition capture algorithm for UAV formations [40]. By integrating an optimal coalition generation strategy with a Voronoi partition-based area minimization strategy, the algorithm leverages onboard hemispherical UV LEDs to facilitate secure and interference-resistant information sharing and cooperative decision-making within the swarm. Simulation results conducted in a 1 km × 1 km × 1 km 3D space demonstrated that, compared to traditional area minimization strategies, this algorithm reduces average energy consumption by 12.73% and average iterations by 27.49% during multi-target capture missions, significantly enhancing the operational efficiency of collaborative UAV formations.

Ultraviolet Positioning (UVP) and Integrated Ultraviolet Communication and Positioning (IUCaP) represent cutting-edge research directions in the field of UV. Yuan proposed a UV NLOS positioning method based on a two-photon counting receiver [41]. In this method, the transmitter azimuth angle is first obtained by solving the intersection line between the transceiver coplanar plane and the ground plane, using the receiving azimuth angle corresponding to the maximum received power; subsequently, the transceiver distance, as well as the elevation angle and azimuth angle of the transmit beam, are solved based on the power constraint equation. Performance verification results demonstrate that, under a typical transceiver geometric configuration within a 100 m range, the proposed method achieves a positioning distance error of less than 2 m and a transmitter azimuth estimation error of less than 2°. Subsequently, the same team upgraded the two-photon counting receiver to a linear array consisting of three or more receivers, and proposed corresponding positioning methods for two typical spacing configurations between adjacent receivers [42]. The results show that when the transmit elevation angle is greater than 30° and the spacing between adjacent receivers exceeds 2 m, the proposed method can still maintain the aforementioned performance. In addition, the adoption of three receivers can completely eliminate the multi-solution problem in positioning and achieve acceptable positioning accuracy.

However, existing studies on UVC and UVP have largely been conducted independently of one another. In [43], the authors systematically established the technical framework of Integrated Ultraviolet Communication and Positioning (IUCaP), comprehensively elaborated its system-level gain and collaborative gain, and devised a universal system architecture for IUCaP. The results demonstrate that the communication-centric Integrated Transceiver Geometric Design (ITGD) scheme, which adopts a low transmit elevation angle and vertical receiving architecture, enables multi-user high-speed communication with low path loss. In contrast, the positioning-centric ITGD scheme, which employs a vertical transmit configuration and low receive elevation angle architecture, achieves omnidirectional uniform signal coverage and superior positioning accuracy. UVC frame-level a priori information assistance can proportionally reduce positioning latency by narrowing the angle search space, while UVP-assisted predictive link tracking can significantly mitigate the link misalignment probability of mobile nodes.

This section systematically reviews the Chinese and international research progress in the field of solar-blind UVC. Existing studies have achieved significant improvements in core performance indicators such as communication rate, transmission distance, and system reliability. However, scattered experimental data fail to intuitively reflect the inherent logic of technological iteration and performance evolution. To clearly present the performance evolution context of representative systems, this paper standardizes and compares the core parameters of the aforementioned key studies, as shown in Table 1.

Table 1. Performance Parameter Comparison of Representative Systems.

Year	Core Technical Solution	Transmission Distance (m)	Communication Rate	BER Threshold	Ref
2008	Deep UV LED sources, PMT receivers, various elevation angle configurations	100	—	<1 × 10−4 (SNR = 15 dB)	[17]
2010	266 nm Q-switched fourth-harmonic Nd:YAG pulsed laser, high-speed PMT receiver, solar-blind filter (channel impulse response measurement)	10–100	—	—	[18]
2023	275 nm DUV LED array, solar-blind filters, integrated transceiver, full-duplex communication	7	10 Mbps	Packet Loss Rate (PLR) = 1.28% (daylight environment)	[34]
2024	254 nm low-pressure mercury lamp, PMT detector, 4FSK modulation, photon-counting algorithm, FPGA	NLOS: 65; LOS: 700	5 kbps	NLOS: 7.52 × 10−4; LOS: <1 × 10−5 (400 m), ~1 × 10−4 (700 m)	[35]
2025	275 nm DUV LED array, APD receiver, RS coding, OOK modulation, FPGA	12	10 Mbps	PLR = 1.51%	[11]

It should be noted that the performance metrics presented in the comparative table are highly dependent on specific experimental conditions. Key influencing factors include types of light sources and detectors, atmospheric conditions (e.g., daylight background), and modulation and demodulation schemes. To ensure the comparability of the data presented in this review, we have explicitly distinguished between LOS and NLOS scenarios, BER and PLR, and annotated all special experimental conditions in the table. Readers are advised to fully consider these contextual factors when interpreting and comparing the performance of different systems.

3. Ultraviolet Light Sources and Detectors

The preceding section reviewed domestic and international research progress in UVC, covering theoretical studies, channel modeling, transmission attenuation characteristics, and signal processing schemes. These investigations have established a solid foundation for understanding the physical mechanisms of UVC. However, for the outstanding characteristics described in theoretical models, such as NLOS performance and strong resistance to interference, to truly move beyond the laboratory and adapt to complex and ever-changing real-world environments, a transition from theory to practice must be achieved [44]. The performance ceiling of an efficient and stable UVC link is often directly constrained by the manufacturing processes and physical specifications of its core components. Therefore, Chapter 3 shifts focus from abstract theoretical derivations to concrete engineering implementations, analyzing the core physical hardware architecture of UVC systems and exploring the evolution of UV light source technologies at the transmitter end, as well as performance breakthroughs in high-sensitivity detectors at the receiver end.

A typical UVC system consists of a transmitter module and a receiver module. In consideration of the inherent characteristics of UVC, the transmitter design should account for transmission power, form factor, thermal performance and portability, while the receiver design prioritizes high sensitivity and compactness [45]. Because UV light undergoes significant attenuation as it propagates through the atmosphere, the transmission power at the transmitter cannot be increased indefinitely without exposing the human body to excessive UV radiation. Therefore, optical antennas must be installed at both the transmitter and receiver ends to extend the communication range. The design of the transmitter’s optical antenna should minimize the divergence angle as much as possible to reduce losses, while the design of the receiver’s optical antenna should maximize the field of view to enhance gain. In addition, the receiver should incorporate a filter tailored to the light source’s wavelength band to block out background light and improve the signal-to-noise ratio.

Figure 4 shows a schematic diagram of a UVC system. The transmitter end employs a transmitter information processing module to preprocess raw source data, including format conversion and source coding, to generate a digital baseband signal suitable for transmission, which is then modulated onto a carrier waveform. The signal is subsequently amplified by an LED driver to power the UV light source in the transmission optical system, converting the electrical signal into a UV light signal, which is emitted into free space as UV light that is invisible to the human eye. The receiver collects and filters out background noise through its optical system, extracting the valid UV light signal. This signal is then converted into a weak electrical signal by the photoelectric conversion module, which also amplifies and shapes it. Subsequently, the receiver’s information processing module performs demodulation, channel decoding, and format restoration to recover the original digital baseband signal. Finally, the restored source information is output, completing the entire ultraviolet light communication process.

Figure 4. Schematic diagram of a UVC system.

3.1. Light Source

UV light sources are key to the development of UVC technology. Traditional UV light sources include xenon flash tubes, low-pressure mercury lamps, and UV lasers. The light radiation from a xenon lamp is produced by thermal excitation of a high-temperature arc and the recombination of positive and negative ions, resulting in a very intense continuous spectrum, with UV radiation covering the wavelength range of 200–400 nm [46]. Conventional gas-filled mercury lamps generate UV radiation through mercury vapor discharge. High-pressure gas-filled mercury lamps feature high power density, a wide power range, and large output power; their spectral energy distribution spans the mid-ultraviolet, long-ultraviolet and visible light regions. In contrast, the UV spectral energy output of low-pressure gas-filled mercury lamps is largely concentrated at 254 nm, within the mid-ultraviolet region. UV lasers typically employ fourth-harmonic Nd:YAG lasers, sapphire lasers, or copper vapor lasers [47]. UV lasers generally have a power output of less than 1 W and exhibit excellent directional radiation characteristics, with their energy concentrated solely in a specific beam direction. However, high-power UV lasers are not only expensive and bulky, but their core technologies are still in the developmental stage. Therefore, for free-space communication, which involves LOS and directional constraints, UV lasers are currently unable to meet the required performance standards. Most of these light sources are bulky, inefficient, and have narrow bandwidths, making them unsuitable for use as communication light sources.

Semiconductor UVC light-emitting diodes (LEDs), with their compact size, low cost, and high bandwidth, have greatly advanced the development of UVC technology [12]. Since the beginning of the 21st century, countries and regions such as the United States, China, Japan [48], and Europe have conducted extensive research on ultraviolet light sources.

Compared with conventional mercury lamps and UV lasers, UV LEDs exhibit remarkable differences in four key performance dimensions. In terms of modulation bandwidth, UV LEDs can reach the megahertz (MHz) level. By contrast, mercury lamps only achieve a bandwidth in the hertz (Hz) to kilohertz (kHz) range, while UV lasers deliver a bandwidth spanning from kHz to MHz [49]. Regarding operational lifetime, UV LEDs can reach 20,000 to 50,000 h, whereas mercury lamps have a service life of approximately 8000 to 12,000 h [50], and UV lasers mostly feature a lifetime ranging from several thousand hours up to 10,000 h. In terms of power consumption, the wall plug efficiency of UVC LEDs in the 275–285 nm wavelength range is approximately 10–15%, that of low-pressure mercury lamps is about 30–40% [51], and that of lasers is about 0.2–5%. In terms of cost, mercury lamps have the lowest upfront purchase cost. The total cost of ownership (TCO) of UV LEDs continues to decline yet remains slightly higher than that of mercury lamps, while UV lasers incur the highest cost along with expensive maintenance overheads.

The optical output power and bandwidth of a UV light source are key parameters that determine the performance of communication systems. The light output power (LOP) of the light source largely determines the communication distance and is directly related to the external quantum efficiency (EQE). The EQE represents the proportion of charge carriers injected into the device that are converted into UVC photons emitted by the LED; it can be calculated as the product of the internal quantum efficiency and the light extraction efficiency [52]. The EQE of conventional UVC LEDs typically ranges from 1% to 3%, with luminous intensity at the milliwatt level.

A single 15 mW 285 nm micro-LED can achieve a data rate of 6.5 Gbps over a transmission distance of 10 m, while maintaining a data rate above 1 Gbps at a distance of 116 m. In contrast, multi-milliwatt UVC LEDs only support low-speed transmission within a 20 m range [53]. On the other hand, increasing the −3 dB bandwidth from 150 MHz to 600 MHz can boost the communication data rate from 1 Gbps to 4.6 Gbps [53]. Therefore, high output power serves as the fundamental basis for realizing long-distance, large-coverage communication, and high bandwidth is the prerequisite for achieving high-rate data transmission. The two parameters must be collaboratively co-optimized to meet the requirements of practical applications.

A summary of related research on traditional UV LED-based communication is provided in Table 2 [54].

Table 2. Performance of Conventional LEDs for Ultraviolet Communication.

Year	Wavelength/nm	Power/mW	Bit Error Rate	Rate/kbps	Line of Sight	Distance/m	Ref
2006	274	4.5 (10 piece)	—	2.4	NLOS	11	[55]
2009	270	—	10−4	100		11.46	[56]
2011	270	0.32 (Single)	7.0469 × 10−4	115.2		4.5	[32]
2012	—	43 (36 piece)	10−4	2.4	LOS	—	[37]
2014	265	1–10	10−3	500	LOS	27.5	[57]
2015	—	30	<0.01	—	NLOS	24	[57]
2017	294	190 μW (7 V)	2.8 × 10−4	71,000	LOS	—	[58]
2018	280	3–5 (mW/cm2)	—	2,080,000	LOS	1.5	[59]
2021	276.8	0.854	2.86 × 10−3	2,000,000	LOS	0.5	[60]
2023	266	16 (4 piece)	<10−3	100,000	NLOS	40	[61]
2025	275	—	<10−3	2000	LOS	122	[62]

Reference [59] utilized commercial 280 nm LEDs combined with Tomlinson-Harashima Precoding (THP) and 4-level Pulse Amplitude Modulation (PAM-4) to achieve error-free transmission at rates up to 1.6 Gbps over a 1.5 m LOS link. The system’s 3 dB bandwidth was 153 MHz, primarily limited by the LED devices. In the absence of solar interference, the effective data rate reached 2.08 Gbps. Even under direct noon sunlight, the receiver maintained an effective rate of 1.66 Gbps with a BER below the FEC threshold. These experimental results demonstrate that solar-blind UVC possesses exceptional anti-interference capabilities in strong background light environments, offering a viable, low-cost, and high-speed solution for outdoor optical communication in the 5G era.

3.2. Detector

Since the energy of UV radiation is extremely weak by the time it reaches the detector after traveling long distances and undergoing multiple scatterings, PMTs with a UV window are detection devices with ultra-high sensitivity; traditional PMTs offer the advantages of low dark count rates and high gain, but they are bulky. Solid-state detectors, such as AlGaN-based and SiC-based APDs [63], can meet basic requirements for sensitivity, selectivity, speed and stability. Additionally, other alternative materials, such as MgZnO and Ga₂O₃, are available.

PMTs are the earliest and most widely used detectors in UVC applications. They feature extremely high sensitivity, low dark current, high gain and a large detection area, making them particularly suitable for detecting extremely weak light signals [3]. However, PMTs are bulky, fragile, and expensive, requiring the use of solar-blind filters, which hinders system miniaturization. Moreover, their typical bandwidth is only around 50 MHz, making it difficult to satisfy high-speed communication requirements. Xing et al. constructed a 2 × 2 MIMO system using two independent PMTs with a bandwidth of 5 MHz [64]; Meng et al. constructed a four-receiver SIMO system using equal-gain combining (EGC) technology, which significantly improved BER performance under weak signal conditions [64].

With advancements in semiconductor technology, Si- and SiC-based PIN diodes and APDs are gradually replacing PMTs [3]. SiC PIN detectors exhibit high responsivity and ultra-low dark current within the 200–400 nm range, making them ideal for UV detection. Meanwhile, Si APDs offer high bandwidth (above 400 MHz) and high gain, suitable for high-speed communication links. AlGaN-based APDs, leveraging the tunable bandgap of the material, possess an inherent solar-blind advantage without requiring additional filters. Furthermore, the Separate Absorption and Multiplication (SAM) design enables high gain and low-noise performance, showing great potential for integration. Shao et al. realized an AlGaN APD with a gain as high as 1.2 × 10⁴ and a peak responsivity of 0.15 A/W at 280 nm [65].

In the field of emerging detectors, materials such as Ga₂O₃, ZnO-Ga₂O₃ heterostructures, and MgZnO have become intensive research focal points. With a wide bandgap of 4.9 eV, Ga₂O₃ is inherently suitable for solar-blind detection, and its heterostructure-based devices have already demonstrated high responsivity [3]. Specifically, an Au-Ga₂O₃-Au detector fabricated by Hu et al. achieved its peak responsivity at 255 nm [66]. Furthermore, ZnO-Ga₂O₃ core–shell APDs have exhibited exceptional responsivity and detectivity with microsecond-level response times. For instance, Zhao et al. developed a core–shell APD with a peak responsivity of 1.3 × 10³ A/W, a record-high detectivity of 9.91 × 10¹⁴ cm·Hz¹/²/W, and a response time of 20 μs [67].

For MgZnO, the bandgap can be precisely tuned by modulating the Mg content. Metal-Semiconductor-Metal (MSM) structured devices have shown responsivities in the order of 10⁴ A/W at 276 nm, indicating superior detection performance. Alema et al. reported a device with a responsivity as high as 1.8 × 10⁴ A/W at 276 nm, attributed to the asymmetric barriers within the MSM structure [68]. Collectively, UVC detectors are transitioning from traditional PMTs to semiconductor-based alternatives (e.g., SiC PIN, Si APD, and AlGaN APD) to satisfy the systemic demands for miniaturization, high bandwidth, and low power consumption.

In the performance analysis of photodetectors, dark current noise, shot noise, and thermal noise collectively determine the SNR and the minimum detectable optical power. PIN photodetectors have no internal gain, where thermal noise is dominant under low optical power conditions, thus requiring pairing with a low-noise amplifier (LNA) [69]. APDs suppress thermal noise via avalanche multiplication, yet their excess noise factor amplifies shot noise and dark current. The optimal multiplication factor must be selected with a deliberate trade-off, and it exhibits strong temperature sensitivity [70]. PMTs feature ultra-high gain with an excess noise factor close to unity, and can operate in the single-photon counting mode under ultra-weak light conditions. However, their key limitations lie in dark current (especially ion feedback in the ultraviolet band) and the requirement for a complex high-voltage power supply [71].

The performance of UV detectors is poised for further optimization, enabling the design and fabrication of devices that surpass the capabilities of traditional bulk-material-based counterparts. The key performance metrics for these detectors are summarized in Table 3. The parameters were measured under standard laboratory conditions. Currently, catalyzed by innovative device architectures based on low-dimensional solid materials, self-powered photodetectors, and plasmonic enhancement technologies, a multitude of novel UV photodetection concepts is emerging [72]. Furthermore, emerging wide-bandgap materials, represented by Ga₂O₃ and MgZnO, have demonstrated significant advantages in solar-blind detection. Although their fabrication processes are still in the nascent stages, these materials are expected to become the ideal choice for future UVC detectors.

Table 3. Performance of Ultraviolet Detectors.

Detector	Spectral Range (nm)	Responsivity (A/W)	Response Time (ns)	Dark Current (nA)	Ref
PMT	110–1100	~105	1–15	2–30	[3]
SiC PIN	200–400	0.085–0.13	~10	—	[3]
Si PIN	200–1100	~0.38	6–50	~10	[3]
AlGaN PIN	220–280	~0.15	~6.5	—	[73]
Si APD	260–1100	0.24–0.5	—	1–100	[3]

This section reviewed the evolutionary history and current research status of UV light sources and detectors, delineating their technological trajectories and contemporary research frontiers. By comparing the critical performance parameters of various devices, this work establishes a comprehensive theoretical foundation for component selection and system design in subsequent ultraviolet communication developments.

4. Ultraviolet Non-Line-of-Sight Scattering Channel Models

The UV-C band within the solar spectrum is heavily absorbed by the stratospheric ozone layer as it penetrates the atmosphere. This absorption ensures that negligible solar radiation reaches the Earth’s surface, creating a solar-blind region. Consequently, UVC systems can achieve a superior SNR even with relatively weak signals due to the nearly non-existent background noise. However, the NLOS UVC channel is highly complex, with transmission quality strictly dictated by atmospheric physical conditions. Therefore, a systematic analysis of the fundamental elements and physical characteristics of channel modeling is essential to accurately evaluate communication performance.

4.1. Factors Influencing Channel Modeling

The troposphere contains fluctuating concentrations of gaseous molecules and aerosol particles. Under diverse meteorological conditions, such as clear sky, rain, fog and dust, the concentration and physical state of these aerosols vary significantly, directly modulating the scattering distribution as well as attenuation magnitude of UV light during propagation. The atmospheric impact on solar-blind UV light is primarily characterized by the synergistic effects of absorption and scattering induced by both gaseous constituents and aerosol particles.

Simultaneously, atmospheric turbulence is a critical factor. Turbulence triggers random fluctuations in the refractive index, leading to intense intensity fluctuations at the receiver, known as the scintillation effect [5]. Experiments have shown that models integrating classical turbulence theory with Monte Carlo simulations can effectively characterize irradiance fluctuations under both strong and weak turbulence conditions using probability density functions.

4.1.1. Atmospheric Absorption

Atmospheric absorption is primarily caused by gas molecules. In the UV-C band, strong ozone absorption is the main reason for the extremely low background noise near the surface and the rapid attenuation of UV signals along the transmission path. The attenuation of light due to absorption is classically described by the Beer-Lambert Law [74]:

$$I(d)=I_0{e}^{-k_{a}d},$$

(1)

where $I_0$ is the initial transmitted intensity, $I(d)$ is the intensity after transmission distance $d$, and $k_a$ is the atmospheric absorption coefficient. On a microscopic level, the absorption coefficient is determined by the absorption cross-sections and number densities of the various absorbing gases. In UV channel modeling, the absorption coefficient and the total scattering coefficient together constitute the total atmospheric extinction coefficient.

4.1.2. Atmospheric Scattering

Scattering is the physical basis for NLOS UVC. Based on the relative size of suspended particles compared to the UV wavelength, scattering is categorized into Rayleigh scattering and Mie scattering.

Rayleigh Scattering: Caused by gas molecules (e.g., $O_2$, $N_2$) much smaller than the UV wavelength. Its energy distribution is relatively uniform and symmetric between the forward and backward directions. The normalized scattering phase function is given by [75]:

$$P_{Ray}(\theta )={\displaystyle \frac{3[1+3\gamma +(1-\gamma ){\cos }^2\theta ]}{16\pi (1+2\gamma )}},$$

(2)

where $\theta$ is the scattering angle. $\gamma$ denotes the molecular depolarization factor for atmospheric Rayleigh scattering. Under standard atmospheric conditions, $\gamma \in [0,0.035]$ holds for both general theoretical research and practical engineering implementations. Specifically, $\gamma =0$ corresponds to the theoretical limit of classical Rayleigh scattering theory for ideal isotropic spherical molecules, whereas $\gamma =0.035$ represents the upper bound of this parameter in extreme atmospheric environments with high water vapor content and abundant polar molecules. For research in the field of UVC and UVP, the widely adopted typical value of $\gamma$ is 0.017. Rayleigh scattering is strongest in the forward and backward directions and weakest at the sides.

Mie Scattering: Caused by aerosol particles (e.g., fog, dust and pollutants) with sizes comparable to or larger than the UV wavelength. Unlike Rayleigh scattering, Mie scattering exhibits strong forward-scattering characteristics. Due to the complexity of the rigorous Mie theory based on Maxwell’s equations, the Henyey-Greenstein (H-G) model is commonly used for approximation [76]:

$$P_{Mie}(\theta )={\displaystyle \frac{1-g^2}{4\pi }}[{\displaystyle \frac{1}{{(1+g^2-2g\cos \theta )}^{3/2}}}+f{\displaystyle \frac{0.5(3{\cos }^2\theta -1)}{{(1+g^2)}^{3/2}}}],$$

(3)

where $g$ is the asymmetry parameter $(g\in [-1,1])$. $g>0$ indicates dominant forward scattering (typically 0.6–0.9 for atmospheric aerosols), $g=0$ represents isotropic scattering, and $g<0$ indicates dominant backward scattering. $\theta$ is the scattering angle.

4.1.3. Composite Phase Function

In real atmospheric environments, neither Rayleigh nor Mie scattering alone can accurately describe photon angular deflection because the atmosphere is a mixed medium. Therefore, a composite phase function must be introduced. This function uses the scattering coefficients of both particles as weights to perform a weighted linear summation [77]:

$$P_{comp}(\theta )={\displaystyle \frac{k_s^{Ray}}{k_s}}{P}_{Ray}(\theta )+{\displaystyle \frac{k_s^{Mie}}{k_s}}{P}_{Mie}(\theta ),$$

(4)

where $k_s^{Ray}$ is the Rayleigh scattering coefficient, $k_s^{Mie}$ is the Mie scattering coefficient, and $k_s=k_s^{Ray}+k_s^{Mie}$. This mechanism provides strong environmental adaptability: in clear high-altitude conditions, $k_s^{Ray}$ dominates and the function degrades to a Rayleigh-like symmetric distribution; in adverse weather like fog or sandstorms, $k_s^{Mie}$ dominates, resulting in a sharp forward-scattering peak.

4.1.4. Atmospheric Turbulence

Irregular fluctuations in local atmospheric physical properties—specifically temperature, pressure, and wind speed—induce stochastic variations in the atmospheric refractive index [78]. As signal light propagates through such an atmospheric channel, turbulence causes phenomena such as beam divergence, distortion, and wandering [79], which leads to intensity scintillation and jitter at the receiver [5,80]. Under weak turbulence conditions, the Rytov solution of the wave equation is employed to characterize these scintillation effects based on the log-amplitude variance of the received wave.

For a plane wave propagating over a channel of length $L$, the log-amplitude variance ${\sigma }_{\chi }^2(L)$ is defined as [81]:

$${\sigma }_{\chi }^2(L)=0.56{\left({\displaystyle \frac{2\pi }{\lambda }}\right)}^{7/6}{\displaystyle {\displaystyle \underset{0}{\overset{L}{\int }}}{C}_n^2(x){(L-x)}^{5/6}dx},$$

(5)

where $\lambda$ represents the wavelength and $C_n^2$ denotes the atmospheric refractive index structure constant.

For a spherical wave, the log-amplitude variance is given by [81]:

$${\sigma }_{\chi }^2(L)=0.56{\left({\displaystyle \frac{2\pi }{\lambda }}\right)}^{7/6}{\displaystyle {\displaystyle \underset{0}{\overset{L}{\int }}}{C}_n^2(x){(\frac{x}{L})}^{5/6}{(x-L)}^{5/6}dx}.$$

(6)

In short-range weak turbulence scenarios, a higher refractive index structure constant (indicating increased turbulence intensity) surprisingly results in a lower system BER. This phenomenon occurs because intensified atmospheric turbulence enhances the scattering capability of the medium, allowing a greater fraction of scattered signals to be captured by the receiver. This effectively reduces path loss and subsequently improves the BER performance.

4.2. Analytical Channel Models

Analytical models formulate NLOS scattering into closed or semi-closed integral expressions, typically categorized by single-, double-, or triple-scattering orders. These models offer transparent physical insights and high computational efficiency, often providing path loss estimates within seconds. However, their applicability is frequently constrained by specific geometric configurations and the inherent complexity of higher-order scattering. In particular, as second- and third-order scattering involve mathematically intensive multiple integrals, many analytical channel models primarily focus on single-order scattering to maintain a balance between computational tractability and fundamental accuracy.

4.2.1. Single-Scatter Model in Cartesian Coordinates

The single-scatter model is defined by scenarios where a photon is received by the detector after undergoing exactly one scattering event. Figure 5 illustrates the parameter definitions for the single-scatter geometric model. ${\phi }_1$ and ${\theta }_1$ represent the divergence angle and elevation angle of the transmitter (Tx) beam, respectively; ${\phi }_2$ and ${\theta }_2$ denote the field of view (FOV), divergence angle, and elevation angle of the receiver (Rx); $V$ signifies the common volume where the FOV and beam overlap; and $r$ is the baseline distance between the transmitter and receiver.

Figure 5. Parameter definition of a single-order scattering geometry model.

Let $\delta V$ be an arbitrary differential volume element within the common volume $V$, be the distance from the Tx to $\sigma V$, $r_2$ be the distance from $\delta V$ to the Rx, ${\theta }_s$ be the angle between $r_1$ and $r_2$, and $\zeta$ be the angle between $r_2$ and the FOV central axis. The optical power received by the Rx, scattered from all differential elements within the common volume, can be expressed as [82]:

$$P_{r,1}={\displaystyle \underset{V}{\int }\frac{P_t{A}_r{k}_{s}P(\cos {\theta }_s)\cos \zeta }{{\Omega }_t{r}_1^2{r}_2^2}}{e}^{-k_e(r_1+r_2)}\delta V,$$

(7)

in this expression, $P_t$ is the transmitted optical power, $A_r$ is the active area of the receiver, ${\Omega }_t$ is the solid angle of the transmitter, $k_s$ is the atmospheric scattering coefficient, $k_a$ is the absorption coefficient, and $k_e=k_s+k_a$ is the extinction coefficient. $P(\cos {\theta }_s)$ represents the phase function jointly determined by Rayleigh and Mie scattering [83].

4.2.2. Single-Scatter Model in Ellipsoidal Coordinates

Following the analysis of scattering models in Cartesian coordinates, the single-scatter model based on ellipsoidal coordinates offers an alternative perspective. In 1979, Reilly established a transmission model for single scattering of wireless UV light through the atmospheric channel using ellipsoidal coordinates [84]. Building on this, Luettgen utilized the focal properties of ellipses to propose an NLOS coplanar single-scattering model, as illustrated in Figure 6. Here, $F_1$ and $F_2$ serve as the two foci of the ellipse with a focal length of $r$, ${\beta }_s$ represents the scattering angle, and ${\psi }_1$ and ${\psi }_2$ are the two focal angles of the ellipsoid.

Figure 6. Parameter definition of a single-order scattering geometric model under ellipsoidal coordinates.

The parameters $\xi$ and $\eta$ are defined as follows [85]:

$$\xi ={\displaystyle \frac{r_1+r_2}{r}},1\le \xi \le \infty ,$$

(8)

$$\eta ={\displaystyle \frac{r_1-r_2}{r}},-1\le \eta \le 1.$$

(9)

Consequently, within the ellipsoidal coordinate system, the focal angles can be expressed through these parameters as:

$$\cos {\psi }_1={\displaystyle \frac{1+\mathsf{\xi }\eta }{\mathsf{\xi }+\eta }},$$

(10)

$$\sin {\psi }_1={\displaystyle \frac{\left[({\xi }^2-1)\right.{\left.(1-{\eta }^2)\right]}^{\frac{1}{2}}}{\xi +\eta }},$$

(11)

$$\cos {\psi }_2={\displaystyle \frac{1-\mathsf{\xi }\eta }{\mathsf{\xi }-\eta }},$$

(12)

$$\sin {\psi }_2={\displaystyle \frac{\left[({\xi }^2-1)\right.{\left.(1-{\eta }^2)\right]}^{\frac{1}{2}}}{\xi -\eta }},$$

(13)

$$\cos {\beta }_s={\displaystyle \frac{2-{\xi }^2-{\eta }^2}{{\xi }^2-{\eta }^2}}.$$

(14)

The elegance of this model lies in placing the transceiver at the foci of the ellipse. Assuming the scattering point lies on the ellipse significantly simplifies the calculation of the system’s impulse response, facilitating a more convenient derivation of path loss.

In [86], the authors propose a non-direct-view ultraviolet single scattering path loss model based on spherical coordinates, which can handle non-coplanar geometric scenarios with arbitrary orientation of the transmitter and receiver. Simulation results show that the model is highly consistent with the Monte Carlo results under parameters such as wavelength 260 nm, communication distance 10–100 m, divergence angle 30°, and field of view 40°.

4.2.3. Single-Order Scattering Model with Reflection from Obstacles

To enhance the practical utility of channel modeling, researchers have integrated obstacle effects into idealized single-scatter frameworks. In [87], the authors established an analytical diffused-LoS model accounting for factors such as obstacle dimensions, positioning, and transceiver height differentials; this approach reduced computational time by approximately 93% compared to Monte Carlo methods while demonstrating that diffused-LoS scenarios provide superior path loss and BER performance over traditional NLOS links. Furthermore, for NLOS environments involving rectangular obstacles, a simplified single-scatter path loss algorithm was proposed in [88]. Simulation results confirm that obstacle reflections significantly mitigate path loss, and the simplified model maintains high precision with a computational complexity far lower than that of conventional Monte Carlo photon-tracing techniques.

4.3. Probabilistic Channel Models: Monte Carlo Method

The Monte Carlo simulation model builds probability models for photon emission, transmission, scattering, and reception. By tracing a large number of photons, it calculates the path loss and impulse response [89].

Figure 7 illustrates the parameter definitions for the Monte Carlo geometric model [89]. In this configuration, the receiver (Rx) is positioned at the origin $(0, 0, 0)$, while the transmitter (Tx) is located on the y-axis at $(0, r, 0)$. The angles ${\theta }_0$ and ${\phi }_0$ denote the zenith and azimuth angles, respectively, of a photon at the moment of emission from the Tx. The distance traveled by the photon after leaving the i-th scatterer (where i = 0 corresponds to the photon leaving the emitter) is given by the following equation:

$$d_i=-{\displaystyle \frac{\ln (1-rand(1))}{k_s}}.$$

(15)

Figure 7. Parameter definitions for the Monte Carlo geometric model.

Assuming a uniform light source distribution, the initial emission direction of a photon is determined by the following equations [89]:

$${\theta }_0=\arccos \{1-rand(1)[1-\cos \left({\displaystyle \frac{{\beta }_T}{2}}\right)]\},$$

(16)

$${\phi }_0=2\pi rand(1),$$

(17)

where $rand(1)$ is independent random variables uniformly distributed on $[0, 1]$.

The direction of the photon changes after being scattered by the scatterer. Let $({\theta }_i,{\phi }_i)$ represent the angle by which the photon deviates from its original propagation direction after the $i$ scattering, and satisfy the probability density function $f({\theta }_i,{\phi }_i)=P({\theta }_i,{\phi }_i)\sin \alpha$, where $P({\theta }_i,{\phi }_i)$ is the phase separation function. Then $({\theta }_i,{\phi }_i)$ can be generated by the following two equations [89]:

$${\theta }_i=F^{-1}(rand(1)),$$

(18)

$${\phi }_i=2\pi rand(1),$$

(19)

where $F({\theta }_i)$ is the marginal probability distribution function of $f({\theta }_i,{\phi }_i)$ with respect to ${\theta }_i$.

After each scattering, if the photon is within the receiving field of view (FOV), the probability of the photon being received by the detector is calculated. Assuming the solid angle formed by the $i$ scattering point and the receiving surface is ${\Omega }_i$, the coordinate vector of the scattering point is $r_i$, and the angle between the original direction of the photon and $r_i$ is ${\theta }_{r_i}$, then the $i$ order scattering and receiving probability of the photon at this time is [89]:

$$p_{r,i}=\left\{\begin{array}{l}{p}_{s,i-1}{e}^{-k_a{d}_{i-1}}P({\theta }_{r_i},0){\Omega }_i{e}^{-k_a\left|r_i\right|},r_i\in FOV\\ 0,r_i\notin FOV\end{array}\right.,$$

(20)

where $p_{s,i-1}$ is the survival probability before the $i$ scattering. The survival probability of the photon is updated after each reception judgment as [89]:

$$p_{s,i}=p_{s,i-1}{e}^{-k_a{d}_{i-1}}(1-P({\theta }_{r_i},0){\Omega }_i).$$

(21)

The Monte Carlo process of a single photon is shown in Figure 8.

Figure 8. Flowchart of the Monte Carlo process for a single photon.

Figure 9 presents the experimental results obtained via the aforementioned Monte Carlo simulation, with the core benchmark parameters adopted in the simulation listed in Table 4. The simulation results demonstrate that the path loss of the UV NLOS communication link exhibits a significant monotonically increasing trend with the increase in communication distance. For a fixed communication distance, both increasing the transmitter beam divergence angle and widening the receiver FOV can effectively reduce the link path loss.

Figure 9. Path loss curves versus communication distance based on Monte Carlo simulation: (a) Fixed divergence angle and field of view (FOV); (b) Different divergence angles; (c) Different FOV.

Table 4. Core Simulation Parameters of UV NLOS Communication Channel.

Symbol	Parameter Description	Value
${\theta }_T$	Transmitter Zenith Angle	30°
${\theta }_R$	Receiver Zenith Angle	30°
${\beta }_T$	Transmitter Beam Divergence Angle	7°
${\beta }_R$	Receiver Field of View (FOV)	20°
$\lambda$	Optical Wavelength	266 nm
$k_a$	Atmospheric Absorption Coefficient	0.802 × 10−3 m−1
$k_{s\_ray}$	Rayleigh Scattering Coefficient	0.266 × 10−3 m−1
$k_{s\_mie}$	Mie Scattering Coefficient	0.284 × 10−3 m−1
$g$	Mie Scattering Asymmetry Parameter (Henyey-Greenstein)	0.72
$\gamma$	Rayleigh Scattering Anisotropy Factor	0.017
$f$	Mie Phase Function Correction Factor	0.5

This section systematically analyzes the critical issues of UV channel modeling. First, it elucidates core physical mechanisms such as atmospheric absorption and scattering to clarify the underlying patterns of transmission loss and perturbation. Subsequently, it categorizes and discusses mainstream modeling methodologies, including analytical approaches and probabilistic Monte Carlo simulations. The established analytical framework and modeling taxonomy provide a robust theoretical foundation for subsequent link performance evaluation, transmission scheme optimization and simulation verification.

5. Ultraviolet Signal Processing Technology

5.1. Modulation and Coding Technologies

In UVC systems, modulation techniques are critical for determining data transmission rates and bandwidth requirements [90]. Selecting an appropriate modulation scheme optimizes signal attenuation characteristics, power utilization and spectral efficiency [75]. Essentially, optical modulation involves embedding information onto optical waves at the transmitter, which is subsequently recovered through detection and demodulation at the receiver [91].

On-Off Keying (OOK) modulation is a special form of Amplitude Shift Keying (ASK) and a fundamental method of optical intensity modulation. Compared with OOK modulation, pulse-phase modulation (PPM) is threshold-free, exhibiting stronger anti-interference performance against inter-symbol interference and enabling further interference suppression via soft demodulation. Based on the number of pulses, it can be divided into single-pulse (LPPM), multi-pulse (MPPM), and differential pulse (DPPM) position modulation. Differential pulse position modulation (DPPM) transmits information by the number of time slots between adjacent pulses, while digital pulse interval modulation (DPIM) uses the number of empty time slots between adjacent pulses to transmit information.

Figure 10 compares the average normalized transmit power of four modulation modes—OOK, DPPM, DPIM, and PPM—at different modulation orders M. The normalized power of OOK modulation is constant at 1.0, independent of the modulation order; the normalized power of DPPM, DPIM, and PPM all decrease significantly with increasing modulation order M; when the modulation order is 6, PPM has the lowest average normalized transmit power. This figure provides a power efficiency basis for selecting modulation schemes in optical communication systems.

Figure 10. Comparison of normalized transmit power mean for different modulation modes.

In [92], the limited transmission distance of NLOS UVC is addressed by theoretically modeling and numerically analyzing the node isolation probability in serial multi-hop networks. Using a single-scattering path loss model together with OOK and PPM modulation formats, the study derives analytical expressions for the node isolation probability and investigates the effects of node density, data rate, transmit power, and transceiver geometry on network connectivity. Under the conditions of 50 mW transmit power, 100 kbps data rate, 10⁻³ BER performance, 30° transceiver elevation angle, and 10° beam divergence angle, as well as receiver field of view, covering a 10 km service range requires approximately 1400 nodes (OOK) and 700 nodes (PPM), respectively. This demonstrates that PPM modulation significantly reduces the required node density, providing a theoretical basis for the deployment of ultraviolet multi-hop networks.

In [93], BER expressions for OOK and PPM modulations under weak turbulence are derived based on the NLOS UVC model, and a simulation model for helicopter landing guidance using UV light is established. Simulation results show that PPM significantly outperforms OOK in terms of BER, with higher modulation orders yielding better performance. The system BER decreases with increasing transmit power, beam divergence angle, and receiver field of view, but increases with longer communication distance and larger transceiver elevation angles. This study provides theoretical and simulation support for enhancing the anti-interference capability of UVC systems for helicopter landing assistance in atmospheric turbulence.

References [58,94] adopt an Orthogonal Frequency Division Multiplexing (OFDM) modulation scheme combined with QAM, where the OFDM signal is superimposed with a DC bias to drive a 294 nm UVB LED. The system uses 2048-point OFDM symbols with 25 information subcarriers and adaptively selects the QAM order based on channel conditions. Under perfect alignment between the transmitter and receiver, 8 QAM is employed, achieving a data rate of 71 Mbit/s. When an angle exists between the transmitter and receiver, the system switches to 4 QAM or 2 QAM to maintain the BER below the forward error correction threshold (3.8 × 10⁻³).

OOK, PPM, Frequency Shift Keying (FSK) and OFDM are four widely employed modulation schemes in ultraviolet UVC systems. In terms of BER performance, PPM eliminates the need for threshold-based detection, achieving the optimal BER performance under identical transmit power and channel conditions. OFDM effectively mitigates frequency-selective fading through orthogonal subcarrier multiplexing, while FSK conveys information via multi-frequency carriers, both delivering moderate BER performance. OOK relies on threshold-based detection, resulting in the poorest BER performance under the same operating conditions.

In terms of implementation complexity, OOK features a simplified architecture, with only on-off control of the optical source required at the transmitter and threshold-based detection at the receiver, thus presenting the lowest hardware and software implementation complexity. PPM requires a stringent time slot and symbol synchronization, and FSK imposes high requirements on optoelectronic devices, leading to moderate implementation complexity for both schemes. OFDM necessitates a complete set of modules for Inverse Fast Fourier Transform/Fast Fourier Transform (IFFT/FFT) processing, carrier synchronization, channel estimation and equalization, along with matched peak-to-average power ratio (PAPR) suppression algorithms, resulting in significantly higher implementation complexity than the other three schemes.

In terms of spectral efficiency, OFDM fully leverages the limited modulation bandwidth of UV LEDs via orthogonal subcarrier multiplexing, achieving the highest spectral efficiency among the four schemes. OOK has no redundant time slot overhead, outperforming PPM and FSK in spectral efficiency at the same symbol rate. FSK requires reserved guard bands for multi-carrier transmission, leading to relatively low spectral utilization. PPM encodes information through the time slot position of optical pulses, where redundant time slot overhead grows exponentially with the increase in modulation order, causing a sharp degradation in spectral efficiency and rendering it the lowest among the four modulation schemes.

For the core characteristics of ultra-low received optical power and scarcity of valid photon counts in the weak photon-counting scenario of UVC, there are distinct differences in the applicability characteristics of Reed-Solomon (RS) codes, Low-Density Parity-Check (LDPC) codes, and Polar codes. RS codes can effectively combat burst errors induced by atmospheric turbulence and photon count fluctuation, yet exhibit inferior performance to LDPC codes and Polar codes in single-photon-level ultra-low SNR scenarios. LDPC codes can fully exploit the soft count information from single-photon detection via iterative soft decoding, and still achieve remarkable coding gain under ultra-low received optical power, while suffering from the drawback of high decoding complexity at large code lengths. Polar codes outperform LDPC codes in error correction performance in single-photon-level ultra-low SNR scenarios, while their performance advantage is insignificant at short code lengths, and their maturity in engineering applications lags behind that of LDPC codes.

In UVC systems, pulse broadening and noise severely degrade system performance. In [95,96], the authors introduce LDPC codes under the single-scattering model, significantly improving both BER performance and communication distance under the same modulation scheme; at a BER of 6.5 × 10⁻⁶, the coded sequence achieves approximately twice the communication distance of the uncoded case. In [97], the authors employ a 36-element 265 nm UV LED array and FPGA-based coding modulation to compare RS(18,10) and LDPC(960,480) codes. Experimental results show that at a BER of 10⁻³, RS codes and LDPC codes extend the communication distance by an average of 32% and 78%, respectively. At a transceiver elevation angle of 30°, the uncoded distance is about 6 m, increasing to about 9 m with RS coding and about 11 m with LDPC coding, demonstrating the significant gain of LDPC codes. In [98], the authors develop a Monte Carlo-based UV scattering channel model and a 2 × 2 UV MIMO experimental platform to verify 2/4/8-PPM schemes. It compares the BER performance of 1 × 1, 2 × 1, and 2 × 2 PSP-OSTBC + 4-PPM systems under various transmission distances and transceiver elevation angles. The results indicate that the proposed scheme features low decoding complexity, significantly enhances channel capacity, diversity gain, and coding gain, and makes the system insensitive to path loss and elevation angle variations. The 2 × 2 configuration achieves the best performance, and higher-order 8-PPM outperforms 4-PPM and 2-PPM. This work provides experimental and theoretical support for the practical implementation of optical MIMO in UV scattering channels.

5.2. MIMO Technology

Multiple-Input Multiple-Output (MIMO) is a technology that uses multiple antennas simultaneously at the transmitting and receiving ends. It has two core functions. First, it transmits the same information through multiple antennas and combines diversity combining technology to achieve spatial diversity, effectively mitigating channel fading, lowering the BER, and improving system reliability. Second, it achieves spatial multiplexing by transmitting different data streams in parallel through multiple antennas, significantly increasing the system transmission rate and capacity without expanding bandwidth or raising transmission power [99].

In UV NLOS scattering channels, channel spatial correlation imposes a significant limitation on the performance gain of MIMO systems. UV NLOS communication relies on atmospheric scattering to achieve signal transmission, where the angular spread of the channel is determined by the geometric parameters of the transceiver. When the system is configured with small antenna element spacing, narrow beam divergence angle or narrow FOV and large transceiver elevation angles, strong correlation is presented among multi-antenna sub-channels, resulting in a marked degradation of the MIMO system’s diversity gain, BER performance, and channel capacity gain. In practical deployment of UVC systems, increasing antenna spacing and optimizing optical parameters can effectively mitigate channel spatial correlation and enhance MIMO system performance.

In [100,101], spatial diversity techniques for NLOS UVC are studied from a theoretical perspective. The BER performance of SISO, SIMO and MISO schemes under a log-normal turbulence channel model is analyzed in [100]. The results show that at a communication distance of 1000 m and the BER of 10⁻⁶, using two and three receivers provides signal-to-noise ratio gains of 4 dB and 5.5 dB, respectively, outperforming the multi-transmitter schemes. In [101], the authors address Gamma-Gamma turbulence fading by employing a dual-branch switch-and-stay combining (SSC) diversity technique, deriving analytical expressions for outage probability and average symbol error probability (ASEP). It is found that dual-branch SSC improves system performance by approximately 7 dB compared to no diversity, and that channel correlation degrades performance, while a higher switching threshold reduces diversity gain.

Hardware implementation and experimental validation are addressed in [37,102]. In [102], the authors adopt spectral-amplitude coding on a broadband UV light source, and employ two PMTs at the receiver for linear combining and demodulation, achieving a favorable trade-off between performance and computational complexity. In [37], the authors build an experimental platform with a 36-element 265 nm UV LED array and dual PMTs, validating the feasibility of equal-gain combining (EGC) diversity reception through outdoor experiments. The measured results indicate that diversity gain depends on receiver spacing and transceiver elevation angles, and are compared with theoretical predictions.

5.3. Equalization Technology

In the field of UVC, equalization technology is a critical technical approach to break through the inherent bandwidth bottleneck of UV LED devices and mitigate the dispersion effects of atmospheric scattering channels. In [60], targeting the high-frequency attenuation characteristics of a 276.8 nm DUV micro-LED, the authors proposed a frequency-domain single-tap pre-equalization scheme tailored for 16-ary quadrature amplitude modulation orthogonal frequency division multiplexing (16-QAM-OFDM) systems. Through inverse compensation of the measured full-link frequency response and optimization of weighting coefficients, this scheme achieves an optimal trade-off between high-frequency gain compensation and noise amplification. Experimental results demonstrate that this scheme extends the available system bandwidth from 340 MHz to 500 MHz, achieves a maximum transmission rate of 2 Gbps over a 0.5 m free-space link (a 47% improvement compared with the non-pre-equalized scheme), and maintains support for 0.82 Gbps high-speed communication at a 3 m transmission distance. These findings fully validate the remarkable performance gain enabled by pre-equalization technology in high-speed UVC systems.

While mainstream linear adaptive equalization algorithms, including least mean square (LMS), recursive least square (RLS), and minimum mean square error (MMSE), can only compensate for linear channel distortion, they exhibit extremely weak fitting capability for the strong nonlinear characteristics of UVC scattering channels. Accordingly, modern equalization technologies have increasingly advanced into the field of deep learning. In [103], the authors proposed an end-to-end blind equalization scheme based on a hybrid long short-term memory-deep neural network (LSTM-DNN). With an integrated design of “temporal feature extraction and nonlinear distortion compensation”, this equalizer achieves significant performance enhancement, with an average BER reduction of 67.8% and an average mean square error (MSE) reduction of 70.8% over the entire SNR range of 0 dB to 20 dB. This scheme provides critical theoretical and experimental support for the design of high-speed, high-reliability UVC systems.

6. Networking Technologies

As research progresses, UVC is evolving from traditional point-to-point physical-layer transmission towards multi-node, large-scale self-organising networks [104]. The focus of research into UV networking technologies has shifted from simple channel modelling to protocol design and system architecture optimisation at the network layer. Overall, current research primarily concentrates on addressing three key challenges: first, the issues of spatio-temporal desynchronization and deaf nodes caused by directional transmission and reception; second, collision conflicts arising from high path loss in scattering channels and multi-node concurrency; third, the maintenance of three-dimensional dynamic topologies posed by mobile nodes such as UAVs. To address these challenges, researchers have achieved significant breakthroughs in three key areas: neighbour discovery mechanisms, media access control (MAC) protocol design, and three-dimensional mobile node architectures.

6.1. Multi-User Channel Modeling

The mathematical modeling of multi-user systems can be generally divided into four stages [105]: three-dimensional geometric modeling of multi-user systems, link channel path loss modeling, quantitative receiver-side modeling of multi-user interference (MUI), and performance boundary modeling of multi-user systems. In the first stage, the spatial coordinates of each user node are uniformly defined, and a unified spatial geometric framework for multi-node networking scenarios is constructed. In the second stage, the path loss of all links is calculated. In the third stage, the calculated path loss of the target link and each interference link is substituted into the receiver photon counting model, respectively, to complete the quantitative conversion from single-link attenuation to total received interference. In the fourth stage, the optimal decision threshold at the receiver is solved, and the system BER is calculated.

It is assumed that OOK modulation is adopted, and the receiver is equipped with a photon-counting PMT, where the number of photons arriving at the receiver within each modulation symbol interval follows a Poisson distribution. The first two stages are consistent with the modeling process of conventional single-link systems. In the third stage, the average photon number of the target signal at the receiver is defined by the formula [105]: ${\lambda }_S={\displaystyle \frac{{\eta }_1{\eta }_2{P}_t}{L_{Tx}{R}_{bTx}h{v}_S}}$, where ${\eta }_1$ is the photoelectric conversion efficiency of the PMT, ${\eta }_2$ is the optical filter efficiency, $P_t$ is the transmit power of the target transmitter (Tx), $R_{bTx}$ is the bit rate of the target link, $h$ is Planck’s constant, and $v_S$ is the optical frequency of the target signal. The average photon count of the total multi-user interference (MUI) at the receiver is defined by the formula: ${\lambda }_i={\displaystyle {\sum }_{j=1}^K\frac{{\eta }_1{\eta }_2{P}_{ij}}{L_{Ixj}{R}_{bIxj}h{v}_i}}$, where $K$ is the total number of interfering user nodes in the network, $P_{ij}$ is the transmit power of the j-th interfering node, $L_{Ixj}$ is the path loss from the j-th interfering node to the receiver (Rx), $R_{bIxj}$ is the bit rate of the j-th interfering node, and $v_i$ is the optical frequency of the interfering signal. In the fourth stage, the optimal decision threshold for the receiver to distinguish between binary “0” and “1” symbols is defined by the formula: $m_T=\left[{\displaystyle \frac{{\lambda }_S}{In(1+{\lambda }_S/{\lambda }_i)}}\right]$. The system BER is constructed by the formula: $P_e={\displaystyle \frac{1}{2}}{\displaystyle {\sum }_{k=0}^{m_T}\frac{{({\lambda }_S+{\lambda }_i)}^k{e}^{-({\lambda }_S+{\lambda }_i)}}{k!}}+{\displaystyle \frac{1}{2}}{\displaystyle {\sum }_{k=m_T+1}^{\infty }\frac{{\lambda }_i^k{e}^{-{\lambda }_i}}{k!}}$.

In multi-user UVC, MUI has emerged as a core challenge. Conflicts originally confined to the time-frequency domain have been extended to the spatial collision domain formed by the overlapping coverage of communication links. Meanwhile, the analytical scope has been expanded from point-to-point BER to network coverage and connectivity, while the receiver architecture has evolved from single-signal demodulation to array-based space division multiple access (SDMA) and multi-user detection (MUD). This gives rise to the critical challenge of effectively suppressing multiple access interference and coordinating multi-node access over strong-scattering, power-constrained channels, and has spurred a series of technological innovations, including collision-free medium access control (MAC) protocols integrating graph theory and cross-layer design, receiver array-based SDMA with successive interference cancellation (SIC) detection, as well as beamforming and power optimization schemes.

6.2. Ultraviolet Neighbour Discovery and Link Establishment Mechanisms

Neighbour discovery is the first step in ultraviolet self-organising networks. As ultraviolet networks typically employ directional transmission, omnidirectional reception, or directional/directional modes, misalignment of transmission and reception beams between nodes is common, resulting in prolonged neighbour discovery delays. Wang proposed a leader-based neighbour discovery protocol [106]. By designating a node as the leader to sequentially perform neighbour discovery, this approach avoids the conflict issues associated with random access. Furthermore, to enhance fairness, the protocol balances the number of times each node acts as the leader via a leader table and incorporates a counter-based mechanism to determine when to terminate the discovery process. Verification via OPNET simulation demonstrated that this protocol can reduce neighbour discovery time by over 90%. The study also analysed the impact of parameters such as the number of nodes and the number of transmission directions on discovery performance, providing an important reference for UV network design.

With the maturity of UAV combat technology, UAVs in practical operational scenarios are frequently subjected to the adverse impacts of multiple factors, including complex atmospheric conditions and EMI, resulting in degraded intra-formation communication efficiency. Zhao designed a spherical ultraviolet LED communication node model [107], as shown in Figure 11, based on the characteristics of near-line-of-sight ultraviolet communication during the initial phase of three-dimensional formation assembly for UAV fleets. This paper proposes a handshake interaction framework incorporating random probabilities and a token-derived random avoidance algorithm to address the issues of channel conflicts and neighbour discovery efficiency during the neighbour discovery process. Compared to traditional single-token or simple random backoff schemes, this algorithm achieves clock synchronisation through dynamic token derivation, thereby significantly balancing channel collisions and discovery efficiency. The proposed algorithm offers a novel solution for neighbor discovery in UAV formations.

Figure 11. Model of nodes with non-uniform longitudinal and lateral distribution: (a) side view; (b) top view; (c) physical schematic diagram.

Wang, starting from the characteristics of the physical layer, proposed a UVCN–NDA-based method for ultraviolet cooperative networks [108]. By dynamically adjusting the bit transmission time to establish a neighbour table and selecting the shortest transmission path between nodes for frame transmission, this approach avoids unnecessary forwarding, thereby reducing the network’s power consumption. In a 16-node simulation, the UVCN–NDA protocol achieved a neighbour discovery probability of 1 and a channel collision probability of 0 by the 16th time slot. Under full network load and a maximum hop count of 9, UVCN–NDA reduced average power consumption by approximately 86.2% and increased overall throughput by approximately 52% compared to BST. In a physically constructed 7-node network, a 265 nm wavelength LED (single-chip optical power 100 mW, six-sided emission) and an H10721-110 photomultiplier tube (paired with a 10 nm bandwidth, OD4 cutoff depth filter) were used. with a terminal in a hexahedral configuration (diagonal 20 cm, height 10.5 cm, side length 10 cm), and a maximum communication distance of 50 m. Only R1 participates in data relaying; R2 and R4 receive data but do not forward it; R3, R5 and R6 have no data input whatsoever, fully realising the shortest-path relaying.

Five neighbor discovery algorithms, including asynchronous probabilistic scanning, are collectively discussed in [107,108]. Figure 12 presents a comparison of their key performance metrics. The bar chart represents the average neighbour discovery time, while the line chart represents the channel collision probability. It can be seen that the newly proposed token derivation and bit-sharing techniques significantly reduce collisions whilst improving the success rate.

Figure 12. Comparison of key metrics for different neighbour discovery mechanisms (discovery time vs. conflict probability).

6.3. Multiple Access and Media Access Control (MAC) Protocols

During concurrent transmission by multiple nodes, scattered UV signals are highly prone to overlap, leading to decoding failures. The design of the MAC protocol directly determines the network’s throughput and resource utilisation.

The CSMA-based UVOC-MAC protocol developed by the University of California (Riverside) is tailored for NLOS ultraviolet transmission, full-duplex communication, and multi-rate operation [19]. It fully exploits the potential of spatial multiplexing, reducing the collision probability by 50% compared to traditional protocols and increasing throughput by up to fourfold. Support for full-duplex operation results in a 377.26% increase in throughput compared to pure half-duplex mode (full-duplex is supported only when the beam angle is ≤40°). Without UVOC-MAC, the collision probability is 70.59% with an average of 7 neighbours and an SIR of 10 dB; after adopting the protocol, the collision probability under high-rate priority sorting drops to 19.83%, and under path loss priority to 23.61%. Compared to traditional protocols, the collision probability is reduced by approximately 50%, although the conflict rate remains extremely high under heavy load.

Li from the Engineering University of the PLA Army proposed an Ultraviolet Lossless Contention MAC (UVLLC-MAC) protocol [109]. This MAC protocol is based on the superposition logic of ultraviolet power, transforming traditional destructive collisions into analysable lossless contention. It effectively reduces collision losses in multi-node contention access and fundamentally alters the collision handling mechanism, with the network’s peak normalised throughput increasing by 77.8% compared to Slot ALOHA. In a practical physical setup with four nodes, full-duplex communication with one-hop reachability was achieved, with an effective communication rate of approximately 115 kbps, which is essentially consistent with the 100 kbps set in the simulation, thereby validating the protocol’s actual throughput performance. Latency is inversely proportional to node priority. Under light load, the average network latency stabilises at 25 ms; under medium load with a cache capacity of 5: the average latency for nodes A, B and C is below 200 ms, whilst the lowest-priority node D has a latency of 410 ms; Under heavy load: the rate of delay increase for low-priority nodes is significantly higher than that for high-priority nodes, and can only meet the requirements of low-real-time communication. Although this protocol can achieve lossless contention to a certain extent, with stable performance for high-priority nodes suitable for the transmission of critical information, network fairness is poor. The throughput of low-priority nodes decreases significantly, latency rises sharply, and the data loss rate is relatively high.

Ren from the University of Science and Technology of China evaluated the system through network model-based simulations, taking into account node mobility, node tracking and UV link gain, and combined this with beam tracking to optimise multiple access for directional UVC [110]. Under high transmission demand, CSMA-A’s throughput growth virtually stagnates due to the significant resources consumed by control frames, whilst CSMA-B and CSMA-C continue to show steady growth. CSMA-C, lacking RTS/CTS control frames, exhibits the greatest increase in throughput. Furthermore, CSMA-C directly replaces control frames with data transmission and PN broadcast relaying, whilst utilising PN relaying as implicit acknowledgement, eliminating the need for additional ACK frames and thereby further reducing channel overhead. As CSMA-C lacks the pre-emptive RTS/CTS collision avoidance mechanism, direct collisions occur when multiple stations transmit data simultaneously; the resource overhead per collision is slightly higher than in CSMA-A/B (though the overall resource overhead is lower due to the absence of control frames).

Li from the Naval Aviation Engineering College, based on a vertical transceiver model for ultraviolet light communication, selected central nodes to radiate to adjacent cells using a weighted algorithm that takes power factors into account [111]. By employing a segmented time-slot method to allocate different time-slot segments to adjacent cells, they enabled nodes in more distant cells to reuse time slots, thereby resolving the conflict between an increasing number of nodes and a limited number of time slots. Compared to TDM, this method reduces message queuing delay by 4 ms when the number of network nodes is 40, increases link utilisation by 2.7%, and boosts network throughput by 1.8 kbps.

Pan designed and demonstrated an NLOS UVC network [112]. He developed a multi-node transmission protocol based on time-division multiple access (TDMA) that supports beacons, as well as protocols for beacon transmission, beacon reception, time compensation and slot conversion on a field-programmable gate array (FPGA) board based on a master-slave architecture. The master node periodically sends beacon signals to the slave nodes to implement the TDMA protocol. During outdoor field tests within a transmission range of 110 m × 90 m, the real-time communication throughput of this UVC network reached 800 kbps; the system utilised OOK modulation with a base transmission symbol rate of 2 Mbps, and the throughput was 40% of the symbol rate, primarily due to transmission resources being occupied by guard intervals in slot allocation and beacon control slots. Although the synchronisation error was reduced to the 50-nanosecond range after time compensation, the system was still unable to completely eliminate synchronisation errors, necessitating the design of a relatively long guard interval (29,124 symbols) to avoid slot collisions; this is also a significant cause of channel resource wastage.

Li proposed the Ultra-Violet Random Access Cooperative Networking Protocol (UVCN-TDMA) based on Time Division Multiple Access (TDMA), which integrates the advantages of synchronous and asynchronous MAC protocols and utilises the characteristics of the ultraviolet physical layer to design a cross-layer frame structure [113]. Through Markov chain analysis of node buffer queue changes, they derived formulas for queue length and average queuing delay. Simulations demonstrate that, through dynamic slot allocation, this protocol achieves higher throughput, lower latency, and lower packet loss rates than traditional TDMA. The study also investigates the impact of maximum buffer capacity and burst traffic on network performance.

Wang conducted research targeting the critical challenges in UVC, including severe path loss, the restriction of existing MAC protocols to only small-scale UV networks, and the scarcity of tailored protocols for large-scale distributed networks [114]. Based on a lossless contention mechanism and a bit-synchronous transmission relay scheme, a neighbor discovery algorithm with low control overhead was designed to enable all nodes to construct a network-wide topology matrix. By innovatively decoupling channel contention from data transmission, a space division multiplexing (SDM)-enabled MAC protocol via dynamic subnet division was proposed, which forms multiple interference-free subnets through lossless contention and achieves parallel data transmission. This study analyzed the impacts of factors including node density on network performance, and indicated that further optimization of the protocol is required in conjunction with mobile scenarios and the energy efficiency requirements of UAV platforms.

As shown in Figure 13, based on the quantitative comparison results of the core performance metrics (throughput, delay, and control overhead) of the MAC protocols with normalized processing, it can be seen that the throughput of the traditional CSMA/CA protocol increases first and then decreases with rising load, while its delay and control overhead deteriorate significantly with the increase in node number. The traditional TDMA protocol presents balanced performance but suffers from insufficient flexibility in topology adaptation. Lossless contention protocols including the improved TDM and UVLLC-MAC can achieve low-delay and high-throughput transmission, but are prone to resource waste in scenarios with large-scale node deployment. Relying on space division and sectorized directional transmission, the SDM-MAC protocol maintains optimal throughput, delay and control overhead performance across the entire load range, exhibiting the most prominent comprehensive performance advantages.

Figure 13. Performance comparison of different MAC protocols: (a) normalized throughput versus normalized load; (b) normalized delay versus normalized load; (c) normalized control overhead versus number of nodes.

In terms of scenario adaptation, for low-to-medium mobility scenarios with small-to-medium node scale, including static sensor networks and mobile emergency rescue networking, the improved TDM or UVLLC-MAC protocol is applicable, which can strike a balance between implementation complexity and transmission reliability. For high-mobility, large-scale node deployment scenarios such as high-density UAV swarms and space-air-ground heterogeneous collaborative networking, the SDM-MAC protocol can accurately adapt to dynamic topology changes and effectively avoid channel collisions, making it the optimal adaptation scheme. For confidential short-range networking scenarios, multi-channel encryption scheduling can be integrated based on the SDM-MAC protocol to simultaneously meet the requirements of communication concealment and multi-user parallel transmission.

6.4. Three-Dimensional Topology Control and Dynamic Mobile Architecture

The transition from early two-dimensional sensor networks to real-world three-dimensional UAV networks places stringent demands on node hardware architecture and topological dynamics.

Zhao systematically evaluated the impact of parameters such as transmission power [115], transmit/receive elevation angles and data rates on the coverage, connectivity and survivability of three-dimensional UAV networks (UVNNS), providing theoretical support for non-uniform deployment. The 3D region spans 200 × 200 × 200 m³ and is divided into 20 × 20 × 20 sub-regions, categorised into four priority levels based on deployment difficulty. Compared with the Random Deployment Algorithm (RDA), under conditions of Pe = 10⁻¹ (OOK + Poisson model), a transmission vertex angle of 60° and a reception vertex angle of 70°, UVNNS coverage improved by an average of over 5.01% and deployment costs were reduced by an average of over 3.11%. The convergence times for both methods increased with rising node density, and the average convergence time for UVNNS was significantly longer than that of RDA.

Although UV light offers the advantage of NLOS communication, beam pointing adjustments are still required in mobile networks. Song designed a mobile node device that combines transmission and reception functions, integrating an Acquisition, Pointing and Tracking (APT) method [116]. Through Monte Carlo simulations, the intricate relationship between node rotation speed and acquisition time was revealed, providing critical insights for the hardware architecture design and networking protocols of high-mobility platforms. Experimental results from outdoor NLOS UVC experiments indicate that the code rate curves obtained in the experiments follow the same trend as the simulation results. When the transmission power is 50 mW, the maximum deflection angle for NLOS communication is set to 6°, and the BER is 10⁻⁵, the transmission code rate at the transmitter can reach 1.64 × 10⁶ Baud.

Based on [115,116], a comparison chart was produced showing how the probability of connectivity varies with node movement speed under different node architectures. The results indicate that as node movement speed increases, the probability of network connectivity decreases gradually for all three schemes. Furthermore, at the same speed, the probability of connectivity is highest for directional mobile nodes using APT tracking, followed by nodes with three-dimensional omnidirectional LED arrays, whilst traditional two-dimensional/unidirectional nodes exhibit the lowest probability. Figure 14 shows the results of a comparison of connectivity performance in three-dimensional mobile self-organising network topologies.

Figure 14. Comparison of connectivity performance in 3D mobile ad hoc network topologies (probability of connectivity as a function of mobility speed/distance).

Unlike the omnidirectional static coverage described above, Ren adopted a model in which secondary nodes track the primary node unidirectionally [110]. Directional protocols under this dynamic topology not only reduce the interference range but also significantly enhance link gain, representing another efficient evolutionary direction for mobile ad hoc networks.

Based on a comprehensive analysis of the latest literature, the development of wireless ultraviolet light communication networks has moved beyond the stage of verifying fundamental principles. Future trends clearly point towards: integrating lossless optical power superposition and spatial division multiplexing (SDM) at the physical layer; utilising dynamically derived tokens or beacons at the data link layer to enable ultra-low-latency neighbour discovery and MAC scheduling; and at the network layer, relying on intelligent nodes equipped with APT tracking and three-dimensional transmission-reception models to adapt to highly dynamic mobile platforms such as UAVs. The deep, cross-layer integration of these three elements will form the core technological foundation of future tactical-level, all-weather covert communication networks.

7. Challenges and Future Trends

This study provides a systematic review of the evolution, underlying physical mechanisms, and core network architectures of UVC. As a highly promising specialized communication modality, UVC offers irreplaceable advantages in environments characterized by severe electromagnetic interference or requirements for strict covertness. These advantages stem from the negligible background noise in the solar-blind band, robust NLOS scattering characteristics, and an inherent low LPI. Currently, the field is in a pivotal transition phase, moving from theoretical validation toward engineering applications in complex, real-world scenarios.

7.1. Key Challenges

While UV communication demonstrates unique strengths in specific scenarios, transitioning from theoretical proof-of-concept to large-scale field deployment still faces several rigorous challenges.

• Hardware Limitations: Current mainstream semiconductor UV light sources, such as UV LEDs, continue to suffer from low photoelectric conversion efficiency and limited luminous power. Breakthroughs in novel wide-bandgap semiconductor materials and advanced device packaging structures are urgently required. Simultaneously, there remains significant potential to improve the detection sensitivity of existing UV detectors, including APDs and PMTs.
• Channel Modeling Complexity: Although numerous researchers have proposed various single and multiple scattering models, achieving an ideal balance between computational efficiency and accuracy remains difficult in extreme meteorological conditions (e.g., thick fog, heavy haze or dust) or field environments containing irregular dynamic obstacles. Furthermore, stochastic fluctuations in the refractive index caused by atmospheric turbulence lead to intense scintillation effects at the receiver. More refined models are necessary to accurately describe the transient influence mechanisms of turbulence across different altitudes and weather conditions.
• Signal Processing Constraints: NLOS UVC channels are characterized by severe path loss and multipath delay spread. Under such harsh transmission conditions, conventional modulation and channel coding techniques often lack sufficient spectral efficiency and error correction capabilities [117]. Additionally, further advancements are needed in weak signal acquisition, equalization algorithms to mitigate inter-symbol interference (ISI), and fundamental signal processing technologies.
• Networking and Multi-node Coordination: In collaborative multi-node networks, multiple access technologies (including TDMA, SDMA, CDMA, FDMA and WDMA) tailored to the wide-angle scattering properties of UV light remain in the exploratory stage. The absence of a mature protocol architecture limits network concurrent capacity and the ability to suppress co-channel interference. Moreover, most existing network protocols are based on static or quasi-static topologies, providing insufficient support for nodes with high 3D mobility, such as UAV swarms. Because UVC relies on scattering, reducing mutual interference between nodes in dense distributed networks at both the protocol and physical layers remains a formidable challenge.

7.2. Future Outlook

The unique NLOS, high-security and anti-electromagnetic interference properties of UVC position it for significant growth within future ubiquitous networks. The technology is advancing from theoretical simulation to engineering implementation and from isolated links to intelligent heterogeneous networks.

In terms of application scope, UVC is expected to expand into diverse fields, including emergency lifeline support in extreme environments, secure inter-platform communication for aircraft, and high-precision assisted landing systems for aerial vehicles. Regarding architectural evolution, cross-band heterogeneous fusion will become a dominant trend. By leveraging the synergy between UV, RF and VLC, researchers aim to construct resilient, all-weather “Space-Air-Ground-Integrated” systems.

Furthermore, the integration of deep learning and artificial intelligence (AI) algorithms for adaptive equalization of complex nonlinear channels and intelligent topology optimization will grant UVC systems enhanced self-sensing as well as self-healing capabilities. These advancements will enable UVC to play a critical strategic role in the future special-purpose communication landscape of the Internet of Everything.