Evolution of Short-Range Optical Wireless Communications

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 41, NO.
4, FEBRUARY 15, 2023 1019
Evolution of Short-Range Optical Wireless

Communications
Ke Wang , Member, IEEE, Tingting Song, Member, IEEE, Yitong Wang, Chengwei Fang , Jiayuan He,
Ampalavanapillai Nirmalathas , Senior Member, IEEE, Christina Lim , Senior Member, IEEE,
Elaine Wong , Senior Member, IEEE, and Sithamparanathan Kandeepan , Senior Member, IEEE
(Invited Tutorial)
Abstract—The optical wireless communication (OWC) technol- I. INTRODUCTION

ogy explores the broad unregulated optical spectrum to provide
IGH-SPEED wireless communications are highly de-
high-speed wireless communications, and the visible, ultraviolet
and near-infrared wavelength ranges all have been investigated.
Compared with the conventional radio frequency (RF) band, the
H manded nowadays, particularly driven by bandwidth-
hungry applications such as high-definition video-on-demand,
spectrum in OWC systems is much less congested and the in- virtual reality and augmented reality. However, due to the
terference is much lower. The OWC technology also provides
enhanced physical layer security due to the natural confinement
congested microwave spectrum, conventional radio-frequency
of optical beams that makes it difficult to intercept the transmit- (RF) wireless communication systems are struggling to meet the
ted signal. Hence, the OWC technology has been widely studied bandwidth and speed demands [1], [2]. To solve this issue, elec-
and considered as a promising candidate in future beyond-5G tromagnetic waves with higher frequency have been proposed
communications. The OWC technology has been considered for and studied to realize high-speed wireless connectivity, such as
both long-range and short-range applications, and in this paper
we will introduce the fundamentals and recent developments of
the millimeter-wave and terahertz bands, and gigabit-per-second
short-range OWC systems. In particular, we will focus on the key (Gbps) data rates have been demonstrated [3], [4], [5].
short-range OWC applications in indoor personal or local area Compared with the millimeter-wave and terahertz frequency
communications, underwater wireless communications, wireless regions, even broader unregulated bandwidth is available in the
data center networks, and vehicular communications. We will optical frequency range, enabling the possibility of high-speed
also introduce some recent widely studied advanced techniques to
boost the performance of short-range OWC systems, including the
wireless data transmissions even with simple modulation for-
spatial domain diversity, multiplexing and modulation principles mats [6]. To explore this frequency range, the optical wireless
to improve the system robustness and data rate, and the machine communication (OWC) technology has been studied, which
learning algorithms and hardware accelerators to suppress both achieves the wireless transfer of information via optical waves
linear and nonlinear effects to improve OWC signal quality and directly, where the data-carrying optical signal propagates via
bit-error-rate performance.
the free-space from the transmitter to the receiver. It is also
Index Terms—Optical wireless communications, optical MIMO, widely known as the free-space optical (FSO) communication,
machine learning, underwater optical wireless communications, and it has been widely studied recently as a promising solution
optical wireless vehicular communications. to realize high-speed wireless connections and as a promising
alternative to congested conventional RF systems [1], [7], [8].
In addition to the broad bandwidth and high speed features,
OWC systems are also immune to electromagnetic interference.
Manuscript received 12 June 2022; revised 19 August 2022 and 24 September Therefore, they can be deployed in RF-sensitive environments
2022; accepted 13 October 2022. Date of publication 19 October 2022; date of such as hospitals or complicated RF environments [9]. Further-
current version 16 February 2023. This work was supported by the Australian more, since the optical beam is naturally confined to a certain
Research Council (ARC) under Grant DP170100268. (Corresponding author:
Ke Wang.) area and cannot penetrate opaque objects, it is difficult to inter-
Ke Wang, Yitong Wang, Chengwei Fang, and Sithamparanathan Kandeepan cept the optical signal, providing inherent security in the physical
are with the School of Engineering, RMIT University, Melbourne, VIC 3000, layer [10]. Therefore, the OWC technology has been widely con-
Australia (e-mail: [email protected]; [email protected];
[email protected]; [email protected]). sidered as a promising RF-alternative option for future beyond-
Jiayuan He is with the School of Computing Technologies, RMIT University, 5G (also called 6G) communications. For example, OWC links
Melbourne, VIC 3000, Australia (e-mail: [email protected]). have been proposed to replace the optical fiber links to provide
Tingting Song, Ampalavanapillai Nirmalathas, Christina Lim, and Elaine
Wong are with the Department of Electrical and Electronic Engineering, high capacity mobile fronthaul and backhaul connections, which
University of Melbourne, Melbourne, VIC 3010, Australia (e-mail: tingt- are easier to establish without the need of laying optical fibers
[email protected]; [email protected]; [email protected]; in advance [11], [12]. Another example is the enhanced mobile
[email protected]).
Color versions of one or more figures in this article are available at broadband (EMBB) use case, which in 5G can support a peak
https://doi.org/10.1109/JLT.2022.3215590. data rate of 20 Gb/s for downlink and 10 Gb/s for uplink, support
Digital Object Identifier 10.1109/JLT.2022.3215590
0733-8724 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: National Institute of Technology Patna. Downloaded on October 28,2024 at 07:05:18 UTC from IEEE Xplore. Restrictions apply.
1020 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 41, NO. 4, FEBRUARY 15, 2023
Fig. 1. Application of optical wireless communications in beyond-5G.
high mobility of 500 kmph to 1000 kmph, have an end-to-end systems. In this paper, “short-range” is used to refer to all these
delay in a few milliseconds, and realize less than 1 ms of OWC systems.
mobile interruption time. The EMBB plus in beyond-5G will In this tutorial, we will focus on short-range OWC tech-
have more stringent requirements, and it is widely expected that nologies and systems, introducing key principles and con-
the OWC technology can offer high-speed wireless connectivity cepts. Whilst the long-range and ultra-long-range OWCs are
to end users without the congested spectrum and complicated also considered as essential in beyond-5G, such as the inte-
interference limitations to satisfy these needs [11], [12]. grated terrestrial-satellite networks to provide high-speed global
Taking the advantages discussed above, the OWC technology coverage, they are out of scope and are only briefly mentioned
has been investigated for various types of applications, as shown for completeness. Interested readers can refer to the references
in Fig. 1. Here we divide these applications into four categories above or recently published relevant survey papers, such as [1],
according to the communication distance, namely ultra-long [7], for more details. Compared with long-range systems, short-
range, long range, medium range and short range scenarios. range OWC systems require simpler architecture, lower-cost
For the ultra-long range applications (i.e., space-related appli- components and lower-complexity signal processing. In addi-
cations), OWC systems have been studied and used in satellite tion, the simple intensity modulation/direct detection (IM/DD)
communications, including satellite-to-ground links and deep- scheme is also preferred over the more complicated and ex-
space links, where the wireless communication distance can pensive coherent transmission. Furthermore, since short-range
exceed 2.5 astronomical units (about 3.7 × 108 km) [13], [14], OWC systems are typically deployed closer to the end user, the
[15], [16]. In particular, recently due to the intensive interests transmission power is normally more constrained due to safety
in low-Earth orbit (LEO) satellite and LEO constellations, the considerations, particularly when laser transmitters are used.
OWC technology has also been widely investigated for inter- Therefore, short-range OWC systems face unique constraints
satellite links (ISLs), which are critical to enable high-speed and challenges. Here, we will first introduce major types of
satellite communications with global coverage [17]. OWC sys- short-range OWC systems, including in-building and indoor
tems have also been investigated for long range applications, OWC systems, V2V and V2I OWCs, OWC-based wireless
such as the high-capacity wireless mobile fronthaul and back- data-center networks, and underwater OWC systems. With the
haul links and the wireless connections to unmanned aerial wide availability of cameras in mobile phones and vehicles, the
vehicles (UAVs), with distances ranging from a few kilometers use of readily available cameras as the receivers has attracted
to tens of kilometres [18], [19]. In addition to ultra-long and considerable attention recently [29]. This results in the optical
long distance scenarios, the OWC technology has also attracted camera communication (OCC) technology, which will also be
interests in short range and medium range applications. Here we discussed. We will focus on the fundamental working principle,
define short range OWCs as those with transmission distances typical system architecture, and key advantages and challenges
up to several meters, including in-building and indoor wireless of these short-range OWC systems. In addition, in this tuto-
local area networks [2], [20], [21] and wireless data center rial we will further discuss recent more advanced studies and
interconnects and networks [22], [23], [24], and define medium techniques to improve the performance of short-range OWC
range OWCs as those up to a few hundreds of meters, includ- systems, focusing on the spatial domain diversity, multiplexing
ing vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) and modulation principles, and the machine-learning (ML) and
communications [25], [26], [27] and underwater wireless sensor artificial intelligence (AI) techniques.
networks [28]. For underwater, V2V and V2I OWCs, although It also worth noting that in addition to this tutorial paper
the data transmission distances are longer, they are also included that focuses on fundamentals, recently a number of related
here since many of their technological aspects, such as the survey papers have also been published. Here we summarize
preference of low-complexity intensity modulated transmitters a few, aiming to provide a quick reference to interested read-
and direct detection, are similar with other short-range OWC ers. Pathak et al. [30] provides a detailed summary of indoor
WANG et al.: EVOLUTION OF SHORT-RANGE OPTICAL WIRELESS COMMUNICATIONS (TUTORIAL) 1021
VLC channel models, modulation techniques, medium access

control algorithms, as well as testbeds. A more recent survey
on VLC is available in [31], where recent developments such
as non-orthogonal multiple access (NOMA) have been further
reviewed, together with a deep analysis within the concept of
future Internet of Light. A recent VLC survey is also available
in [32], which focuses more on the application potentials of
the VLC technology in beyond-5G networks. In addition to
VLC, a review of near-infrared in-building and indoor OWCs
is provided in [2], which provides a summary of beam steering,
OWC-based localization, and hybrid RF/OWC systems. [8] also
presents a review of recent progresses in near-infrared indoor
OWCs, focusing on the access network based on OWC physical
links. In addition to the surveys focusing on in-building and
indoor OWCs, [28] presents a comprehensive overview of recent
progress in UOWC systems, including the fundamentals of light
wave propagating through the water channel, the UOWC chan-
nel modelling, modulation and coding techniques in UOWC,
typical transceiver architectures, as well as recent experimental
demonstrations. In the area of OWC-based vehicular communi-
cations, [33] provides a well-structured review. The basic OWC
V2V and V2I system architecture, transceiver design, use cases,
key challenges and potential solutions are presented. Based on
it, [27] further focuses on the camera-receiver based V2V and
V2I communications, where recent studies on the key properties Fig. 2. (a) Wired and wireless in-building and indoor communications [2];
and (b) principle of in-building and indoor OWC systems [8].
of camera as data transmission receiver, region-of-interest (ROI)
detection and signalling, modulation techniques adapted by the
camera receiver, and the impact of vehicle mobility have been
opto-electronic components and the higher safe transmission
summarized.
power in this band. Taking advantage of the broad transmitter
bandwidth, over Tb/s data rate has recently been demonstrated
II. SHORT-RANGE OWC SYSTEMS [6]. The rapid and significant progress of in-building and indoor
OWC technologies have also benefited other types of short-range
Short-range OWCs have been an active research field since
OWC systems in the past two decades, such as for underwater
late 1970s [34], [35]. In the early generation of short-range
wireless connections [28], wireless data-center networks [43],
OWCs, most studies focus on in-building and indoor wire-
and vehicular communications [33]. In this section, we will
less communications and the 850 nm near-infrared wavelength
introduce the OWC technology in these typical short-range
band [36]. Both direct and diffusive links are studied, aiming
scenarios.
to achieve high energy efficiency and sufficient signal cover-
age, respectively. The channel modelling, advanced multi-spot
transmitters, and various types of diversity receivers have been A. In-Building and Indoor OWC Systems
investigated [37], [38], [39]. With the advancement and widely Broadband access networks have been widely deployed and
deployment of LEDs, the the visible light communication (VLC) high-speed network access has been delivered to the doorstep.
technology has then been proposed in early 2000 s [40] and has Therefore, high-speed in-building and indoor communications
since attracted intensive interests, thanks to the capability of are highly demanded. As shown in Fig. 2(a), the multi-mode fiber
providing simultaneous illumination and wireless communica- (MMF) and plastic optical fiber (POF) can provide high-speed
tion. With advanced techniques, over Gb/s data transmission in wired access. With the popularity of mobile devices such as
VLC systems have been demonstrated [41], [42]. In parallel with mobile phones and tablets, the wireless option is preferred.
short-range OWCs relying on the visible range and LED trans- Currently, WiFi has been widely used to provide wireless con-
mitters, near-infrared OWCs using laser transmitters have also nectivity in local and personal area networks, particularly in
been widely investigated during the last a few years. Compared working and living spaces. As a mature RF technology, low-
with LEDs, laser transmitters have broader modulation band- cost fully integrated WiFi transceivers are widely available.
width to achieve ultra-high data rates and are compatible with WiFi signal, which has much longer wavelength compared
the widely available optical fibers for optical signal distribution, with the optical signal in OWC systems, has the capability to
whilst the higher cost and more stringent transmission power penetrate walls and other channel blockers, and hence, wide
(i.e., limited signal coverage area) are the major limitations. signal coverage and robust wireless communications can be
Different from the early generation using the 850 nm band, provided even without the direct line-of-sight (LOS) channel,
most recent studies focus on the 1550 nm band [2], mainly due which also enables full mobility support. However, as a RF
to the availability of high-performance and mature optical and technology relying on the 2.4 GHz or 5 GHz frequency band,
WiFi only has highly limited bandwidth. Thus, it is challenging in indoor environments, such as the reflections from walls and
to realize ultra-high-speed wireless communications exceeding the ceiling, reflected signals have different optical paths and
a few Gb/s even with advanced technologies such as OFDM and reach the receiver with different time delays. Hence, this type of
multi-input-multi-output (MIMO) configurations. In addition, signal introduces multipath dispersion, and the NLOS channel
due to the dense deployment of WiFi access points, WiFi signals bandwidth is normally limited. Both theoretical and experimen-
also experience large interference, further limiting the data trans- tal studies have been conducted and results show that the channel
mission quality. On the other hand, in the OWC technology, large bandwidth of NLOS OWC link is limited to tens of MHz [37],
unregulated bandwidth is available and it is also immune to elec- [44]. Whilst the data rate supported by the NLOS signal is much
tromagnetic interference. Therefore, achieving higher wireless lower than the LOS case, LOS signals are vulnerable to possible
communication speed becomes possible. In addition, the OWC channel blocking that leads to communication interruptions,
technology can also be used in RF-restricted environments, such particularly for mobile users. Therefore, the NLOS signal can
as in hospitals, where WiFi is restricted due to interference be explored to maintain the wireless connectivity when the LOS
considerations, and in underwater environments, where WiFi channel is unavailable.
is limited due to the high propagation loss of RF signal through In addition to the multipath dispersion caused by signal
the water channel. Furthermore, due to the natural confinement reflections, there are two more major limiting factors on the
of the light signal within a certain area, OWC also enables performance of indoor OWC systems. First, the change of local
better communication security in the physical layer. Due to these temperature and humidity levels lead to turbulence, such as those
advantages, the OWC technology for personal and local area caused by heaters, air conditioners and humidifiers. In indoor
networks has been proposed and widely studied. environments, due to the short wireless channel, no significant
The general principle of in-building and indoor OWC sys- turbulence effect is expected. However, the turbulence may lead
tems is shown in Fig. 2(b). At the transmitter side, the data to the change of polarization state of the data-carrying light
to be transmitted is first modulated onto the optical carrier signal arriving at the receiver, particularly when lasers are used
using an electrical-to-optical (E-O) conversion device, where as the light source. Due to the polarization dependent loss (PDL)
both direct modulation and external modulation can be used. of photo-detector (PD), the performance of data transmission is
The data-carrying optical signal then goes through an optical affected. Typically, the PDL is smaller than 0.5 dB, and hence,
assembly before being emitted to the free-space channel, which this impact is relatively minor. The second major performance
controls the divergence, profile and orientation of the beam. The limiting factor is the background light, due to both sunlight and
emitted optical beam then propagates to the receiver side, where indoor lamps for illumination purposes collected and detected
another optical assembly collects and focuses the signal beam. by the receiver. This leads to the background light induced noise,
Then the signal light is converted back to the electrical domain which is an additional noise compared with conventional optical
by an optical-to-electrical (O-E) conversion device for further fiber communication systems that further limits the performance
processing and detection, where both PIN and avalanche photo- of indoor OWC systems. Results show that the impact of unique
diode (APD) can be used for the O-E conversion. Considering background light induced noise in indoor OWC systems is more
the low-cost and low operation voltage requirements, PIN is dominant at lower data rates, caused by the smaller noises
typically used in indoor OWC systems. contributed from other sources (e.g., the preamplifier noise
Two types of E-O converters have been widely used in indoor and the thermal noise), which highly depend on the operation
OWC systems as light sources, including LEDs and lasers. bandwidth [45].
LEDs have been widely used in VLC, and they are particularly 1) Visible Light Communication Systems: In VLC systems,
attractive due to the simultaneous illumination and data one key limitation is the limited bandwidth of LED transmitters.
communication capability and the low-cost feature. From the To achieve high-speed wireless communications, various meth-
optical beam point of view, LEDs emit incoherent light signals ods have been proposed and studied. For example, since the
with a relatively divergent radiation pattern, which can be light emitted by LED has a broad wavelength spectrum, the use
modelled as the Lambertain profile. The modulation bandwidth of wavelength division multiplexing (WDM) has been explored
of LEDs is also typically limited to tens of MHz. Hence, VLC to increase the data rate via parallel transmissions. The RGB
systems normally provide a moderate wireless communication type of LED is naturally fit [46], where up to 1.35 Gb/s data
speed, but a broad signal coverage area. On the other hand, lasers rate over 30 cm transmission is demonstrated using the three
have been widely used as the light source in near-infrared indoor colors of one LED. In addition, advanced modulation formats
OWC systems. Due to the much broader bandwidth, higher data to use the available bandwidth more efficiently have also been
rates can be achieved using laser transmitters. However, the studied in VLC systems, such as the carrierless amplitude and
beams are normally narrow with small divergence, which limits phase (CAP) and the discrete multitone (DMT) formats [47].
the coverage area using near-infrared coherent laser signals. Furthermore, recently LEDs with higher modulation bandwidth
In indoor OWC systems, two types of optical signals can have also become available. One example is the GaN-based
be collected and detected at the receiver side. The first type micro-LED with 980 MHz bandwidth, where over 10 Gb/s data
is the LOS signal, which has ultra-broad channel bandwidth and rate has been realized in VLC systems when combined with
hence, can be explored for high-speed wireless connectivity. The the DC-bias optical OFDM modulation format [48]. LEDs with
second type is the reflected non-line-of-sight (NLOS) signal that even broader bandwidth have also been demonstrated, such as
reaches the OWC receiver. Due to the typical diffusive reflection the micro-LED based on InGaN quantum well with 1.3 GHz
To solve these limitations, advanced transmitter technologies

have been proposed and studied, and one example is the
multi-beam diffusive configuration shown in Fig. 3(b), where
an OWC transmitter points at the ceiling and generates multiple
light spots in the ceiling. The beams are then reflected by the
ceiling and walls to collectively provide full signal coverage.
Therefore, multiple diffusive reflected links are available to
improve the channel bandwidth and the SNR at the receiver
side. By optimizing the angular distribution of multiple light
spots and the allocation of power, up to 10 Gb/s data rate
has become possible [51]. However, the transmission power
required is still relatively high beyond the safety level.
To provide high-speed wireless communication and user mo-
bility while satisfying the laser safety requirement, the use of
beam steering has become a popular solution in near-infrared
indoor OWC systems. In this type of systems, as shown in
Fig. 3(c), the transmitter is located at the ceiling facing the
user, the user localization function is used and then a laser beam
with controlled and limited divergence is adopted to cover the
Fig. 3. Typical configurations of near-infrared indoor OWC systems. (a)
Diffusive link configuration; (b) multi-beam diffusive configuration; (c) beam limited area surrounding the user, providing limited user mobil-
steering OWC configuration; and (d) silicon integrated optical phased array for ity within the coverage area. When the user moves out of the
beam steering in indoor OWC systems. signal coverage area, which can be identified by the localization
function, the beam steering function is activated to track the
bandwidth [49]. Therefore, multi-Gb/s wireless communication user and maintain the wireless connectivity. Through this way,
becomes a reality in VLC systems. For more details on VLC, the direct LOS link and safe transmission power are used, and
recent review and survey papers [30], [31], [32] provide in-depth high speed data transmission is achieved. In particular, due to the
and systematic discussions. use of relatively narrow beam, only a limited number of users are
2) Near-Infrared Indoor OWC Systems: Another major type connected simultaneously, enabling a large capacity per user.
of indoor OWC systems that have been widely studied is near- The beam steering function is critical for the near-infrared
infrared systems with laser transmitters. Early generations of indoor OWC systems discussed above, and it has been realized
near-infrared indoor OWC systems have mainly used the 850 nm through different mechanisms. The MEMS-based steering mir-
and 980 nm wavelength regions [36]. With the maturity of optical ror has been utilized, where the beam steering is realized via
fiber communication devices, recent near-infrared OWC studies simple signal reflection [20]. High steering efficiency can be
have mainly focused on the 1550 nm band [2], [20], mainly achieved as well via mirror coating. However, the achievable
due to the larger eye-safe transmission power, lower background steering angular range is typically limited, precise control is
light, and more efficient opto-electronic devices. Compared with required due to the inherent analog steering, and relatively
VLC systems, laser transmitters in near-infrared indoor OWC high operation voltage is needed. The spatial light modulator
systems have much broader modulation bandwidth, and hence, (SLM) has also been explored for beam steering, which is
ultra-high speed wireless data transmission can be realized even realized via signal diffraction and interference [52]. In addition
without advanced modulation formats and complicated signal to beam steering, the SLM is also capable of generating multiple
processing. beams, and hence, it can provide simultaneous OWC to multiple
Whilst high data rate can be easily achieved in near-infrared users. However, the beam steering efficiency is normally limited,
indoor OWC systems, one key challenge is the limited signal particularly for large beam steering angles due to the use of
coverage area due to the use of narrow laser beams. To solve high-order diffraction, and the control algorithm is complicated.
this limitation, the use of diffusive links enabled by diffusive In addition, the beam steering in near-infrared indoor OWC
signal reflections from walls and the ceiling have been proposed systems has been realized using the silicon integrated optical
and studied, as shown in Fig. 3(a), where the entire indoor space phased array (OPA) as well [53], as shown in Fig. 3(d). By
can be covered to provide full user mobility. Strict alignment changing the phase difference between adjacent elements in the
between the transmitter and the receiver is not required either phased array, the optical beam can be generated in different free-
and the robustness against channel blocking can be provided. space directions. Compared with other beam steering solutions,
However, the channel bandwidth is highly limited to tens of the silicon integrated OPA leverages recent rapid developments
MHz [50], and wide field-of-view (FOV) receivers need to be of silicon photonic integration technology, and hence, it can
used, which leads to strong background light and reduced signal- use the advanced CMOS fabrication facilities for high density
to-noise ratio (SNR). In addition, a high transmission power is integration and low-cost mass manufacturing. However, to avoid
required to provide full coverage. Hence, the energy efficiency inter-channel interference between adjacent elements inside the
is low and more importantly, it causes the laser safety concern OPA, a moderate channel spacing is required. This leads to the
in personal areas. generation of side lobes that may lead to signal crosstalk in
the free-space, the data-carrying light then propagates to the

receiver. Similar with other beam-steered OWC systems, the
LoS channel is required to achieve high data rates and the
communication is vulnerable to possible physical blockages.
Combining user localization, limited signal coverage and
beam steering, ultra-high speed wireless communications have
been realized in near-infrared indoor OWC systems. Using the
PAM-4 modulation format, over 100 Gb/s has been experi-
Fig. 4. Passive beam steering in near-infrared indoor OWC systems. (a) mentally demonstrated [55]. The WDM technology has been
Diffraction grating-based method; and (b) quasi-passive method based on optical incorporated as well to further boost the data rate, where over
latching switches. Tb/s OWC has been achieved in indoor environments [6]. In
addition to ultra-high data rate, indoor OWC systems employing
steerable narrow beams also have the advantages of high energy
indoor OWC systems, which require further study and device efficiency, large capacity per user, minimal interference among
optimization. users and enhanced physical layer security.
The beam steering mechanisms discussed above are all active 3) Light-Fidelity (LiFi): In the last decade LiFi, as a short-
methods, where power supply is required at the ceiling mounted range OWC technology, has seen significant progress, which
near-infrared OWC transmitter. To eliminate the need of local explores the optical spectrum for high-speed wireless commu-
power supply, the passive beam steering method has also been nications [58]. LiFi is similar to VLC since both explore LEDs
proposed and demonstrated [54]. This method is based on the as transmitters, but LiFi further extends the concept to achieve
use of diffraction grating and signals with different wavelengths, a high speed, secure, bi-directional and complete wireless net-
and the basic principle is shown in Fig. 4(a). Two cross-aligned working system. In addition to point-to-point connections, LiFi
gratings can steer a beam in two dimensions according to supports multiuser communication, involves multiple access
wavelength, and hence, high-speed wireless connections can be points with seamless handover to form a network, and also
provided over a large area. This passive beam steering principle enables full user mobility [59]. As a complete wireless network,
has been further extended by using a high port count arrayed LiFi not only supports the downlink data transmission, but also
waveguide grating router (AWGR) together with two dimen- the uplink, where the near-infrared spectrum is normally utilized
sional fiber array [55]. Signals with different wavelengths are since it is invisible to users [60]. LiFi networks normally use
routed to different optical fiber in the fiber array via AWGR, very small optical attocells to provide high throughput. However,
which are then emitted to the free space in different direc- the dense deployment of access points also leads to substantial
tions. Compared with the diffraction grating-based approach, interference between cells that affects the achievable data rate.
the extended method with AWGR removes the need for highly- To solve this issue, the hybrid LiFi/WiFi networks have been
efficient gratings with low polarization dependency and the need proposed and explored. This type of system combines the wide
for highly-precise alignment of two gratings. coverage and high robustness against physical shadowing advan-
Whilst the passive beam steering mechanism eliminates the tages of WiFi signals and the ultra-high data rates in localized hot
need of local power supply, wavelength tuning is used, and spots capability of LiFi signals. The LiFi hot spots are used when
hence, precisely controlled and expensive tunable laser sources available to achieve high throughput, and the system switches
are required. To solve this limitation, recently a quasi-passive to the WiFi signal if high-speed LOS links are blocked or the
beam steering method has been proposed [56], where optical LiFi local hot spots are unavailable. Due to the use of different
latching switches (OLSs) are explored as shown in Fig. 4(b). spectra, there is no interference between WiFi and LiFi, and at
By combining two OLSs with 3 dB optical power splitters, the same time, the additional WiFi network helps to guarantee
the data-carrying optical signal can be routed to different out- equally high quality of service through the whole coverage
puts pointing at different free-space directions to realize the area. To further improve the transmission performance, the load
beam steering function. Due to the latching property, the cross- balancing [61] and the smart handover [62] in the hybrid network
connection status can be maintained without any power to reduce have been further investigated.
the power consumption. The optical signal can also be divided In LiFi, the mobility support is important and highly desired.
into multiple parts and transmitted to different directions to In order to support use mobility, the full communication area
support multi-user access and broadcast/multicast. Furthermore, needs to be covered, and this can be achieved in various ways,
a remote-powering scheme has been further proposed [57], such as using a relatively large number of distributed OWC
where an optical pump is remotely transmitted from the home access points, beam steering to extend the effective signal cover-
gateway to the ceiling mounted OWC access point together age, NLoS links, or a combination of these methods. A receiver
with the communication signal light. At the access point, the with large effective FOV is equally important for user mobility
optical pump is converted to the electrical domain to provide the in order to collect and detect signals from various directions.
energy required for OLS reconfiguration (i.e., beam steering). This can be achieved with a single wide FOV receiver, an
Therefore, similar with the wavelength tuning-based method, the angle-diversity receiver consisting of multiple narrow FOV ele-
beam steering function can be achieved without the need of local ments, or a steerable receiver. In addition, the efficient handover
power supply, eliminating the need for extra wiring to facilitate between OWC access points is needed to support mobility, and
the design of indoor OWC systems. After being launched into in the case of hybrid LiFi/WiFi system discussed above, the
vertical handover between the WiF signal with large coverage

and moderate speed and the LiFi hot spot with limited coverage
and high speed is also needed.
It also worth noting that in addition to the conventional indoor
personal and local areas, LiFi supports much broader application
scenarios, such as for industrial wireless networks [63] and
outdoor communications [64]. For example, the high capacity
OWC link with physical security is highly attractive in industrial Fig. 5. Typical link configurations of UOWC systems. (a) Point-to-point
environments, and it has been considered as a promising solution UOWC link; (b) diffusive UOWC link; and (c) NLOS UOWC link.
to the wireless ICT infrastructure enabling future Industry 4.0.
4) Summary: In this part, we have introduced the basics of in-
building and indoor OWC technologies, including both the VLC by underwater OWC systems is much higher than the acoustic
and near-infrared laser-based OWCs. In general, VLCs have the counterpart, where much higher than Gb/s underwater wireless
advantages of wide signal coverage, larger signal transmission connectivity has been demonstrated [65]. Using RGB lasers
power due to the use of incoherent light, lower cost, and the (i.e., WDM) with two-stage light injection and optoelectronic
capability of providing both illumination and wireless commu- feedback, up to 150 Gb/s data rate has been demonstrated over
nication simultaneously. However, the modulation bandwidth 5 m underwater transmission [66].
is the major limitation that affects the achievable data rate. On The general architecture of UOWC system is similar to that of
the other hand, near-infrared laser-based OWCs have the key the indoor OWC system shown in Fig. 2(b), with the free-space
advantage of broader bandwidth, thus allowing high data rates channel replaced by the underwater channel. Similar with the in-
with simple modulation format and minimal signal processing. door free-space channel, the underwater channel is also affected
Another advantage of near-infrared laser-based OWCs is invis- by background light, where sunlight introduces strong noise at
ibility, where high-speed data transmissions can be achieved in the receiver side, particularly when the receiver is facing upward.
the background. This advantage is particular attractive for the In addition, compared with the indoor free-space channel, the
uplink to avoid employing a visible and illuminated LED at the underwater channel introduces higher signal attenuation. The
user side, and it is explored in LiFi designs and implementations loss in the underwater channel is typically characterized by the
[60]. However, due to the coherent nature of the light wave, the attenuation coefficient c(λ):
transmission power is highly limited, resulting in the typical
c(λ) = a(λ) + b(λ) (1)
usage of beam steering. Whilst the energy efficiency is better,
the transmitter side also becomes more complicated, including where a(λ) and b(λ) are the absorption and scattering coeffi-
both the physical structure and the control perspectives. cients, respectively, which depend on the water type (e.g, pure
sea water, clear ocean water, coastal ocean water, or turbid harbor
water). The typical value of c(λ) can range from 0.056 m−1 in
B. Underwater OWC Systems pure sea water to 2.17 m−1 in turbid harbor water. Therefore,
In addition to in-building and indoor scenarios, the OWC the type of water has significant impact on the performance of
technology has also been explored to realize high-speed under- UOWC systems. Furthermore, the UOWC system performance
water communications [28]. This is mainly driven by the recent is also affected by water turbulence, which can lead to signal
trend of ocean exploration for natural resources, connecting scintillation that further reduces the received signal power.
underwater sensor networks and unmanned underwater vehicles, In UOWC systems, typically the LOS link is used to achieve
maritime construction works, and defence applications such as high data rate, including both point-to-point LOS links and the
providing wireless connectivity to submarines. Conventionally, diffusive LOS links, as shown in Fig. 5(a) and (b). In addition,
underwater wireless communications are realized using acoustic the NLOS link has also been studied in UOWC systems by ex-
waves. Due to the low signal propagation loss, long communi- ploring the signal reflection at the water-air interface to provide
cation distance exceeding a few kilometers can be achieved. enhanced robustness against possible channel blocking [67].
However, due to the low carrier frequency, the supported data Compared with indoor OWC systems, where signal reflections
rate is highly limited. On the other hand, the widely used are typically diffusive, the signal reflection in UOWC systems is
microwave frequency is unsuitable for underwater applications mostly specular, and hence, the reflected signal strength can be
due to the high propagation loss. stronger. In addition, due to the larger refractive index of water
To achieve higher speed underwater wireless communica- compared with air, the total internal reflection condition can
tions, the optical frequency has been explored. In particular, be satisfied by adjusting the transmitter and receiver locations.
there is an underwater transmission window between about Therefore, the reflected signal strength can be further enhanced
450 nm and 550 nm wavelength range. Hence, underwater to support high data rate transmissions. However, on the other
OWC systems have been widely studied recently. Compared hand, surface waves can significantly affect the specular re-
with the acoustic wave based solution, the underwater OWC flection at the water-air interface, and its impact on the NLOS
signal still has higher propagation loss, and hence, normally it channel needs to be further investigated.
targets at relatively short range communications with distances In addition to the NLOS-based UOWC systems, surface
up to tens of meters. More importantly, the data rate supported waves also have significant impact on the air-to-water type of
UOWCs due to the signal propagation through the air-water loss, UOWC link range is limited. Therefore, the cooperative
interface. This practical limiting factor starts to attract research transmission using relays to extend the transmission distance
attentions recently, where its impact on the signal detection has provides a promising solution. The cooperative transmission in
been characterized [68]. Results have shown that the surface UOWC systems can be divided into two phases: (1) the data is
wave leads to larger signal variation and an anti-error and erasure broadcast to both the relay node and the destination receiver; and
coding scheme has been proposed to mitigate the adverse effect. (2) the relay node forward the received data to the receiver to
A fast-tracking system has also been proposed to mitigate the assist the wireless communication. Both amplify-and-forward
beam deflection caused by the surface wave, and the packet loss (AF) and decode-and-forward (DF) types of relay can be used.
ratio has been significantly reduced by about 90% [69]. Despite The DF relay requires 3R signal regeneration, and hence, the AF
these recent progresses, the study on the impact of surface method is simpler and more power efficient. With the relaying-
wave in UOWC systems is still in an early stage, and more based dual-hop system using DF, analysis has shown that in a
effective mitigation techniques need to be further investigated. point-to-point narrow beam underwater communication system,
One promising solution is MIMO with spatial diversity, where the performance can be improved by up to 39 dB, which in turn
spatial redundancy can be explored. can significantly extend the communication reach [73]. This
Similar with indoor OWC systems, both LEDs and lasers improvement is enabled by dividing the channel into several
have been explored as light sources in UOWC systems. Con- shorter parts with each relay node serving as an intermediate
sidering the attenuation in UOWC channel, LED and laser transceiver, where the signal is detected, detected, processed and
transmitters normally operate in the visible range. LEDs are re-generated by a new light source (i.e., 3R signal regeneration).
more suitable for the diffusive type of UOWC link due to the Hence, a much larger energy per bit can be achieved at the
inherent signal divergence, which reduces the requirement on receiver, improving the BER and transmission distance.
transceiver alignment, and lasers are inherently a good option Based on the relay-assisted cooperative transmission, the
for the point-to-point type of link with ultra-high data rate. Using allocation of resources have been further investigated in UOWC
a micro LED array operated at 450 nm wavelength with a wider systems to achieve full system potential. For example, the
modulation bandwidth than conventional LEDs, up to 4.92 Gb/s relay selection has been considered as the contextual bandit
has been demonstrated over 1.5 m underwater wireless link [48]. problem incorporating environmental information to maximize
On the other hand, using injection-locked laser diodes together the UOWC channel performance [74]. The power allocation
with WDM (RGB wavelengths), up to 150 Gb/s data rate has has been further incorporated to improve the performance of
been achieved over 5 m ocean water [66]. These demonstration relay-assisted cooperative UOWC systems based on both LOS
has relatively short link lengths, and when the channel distance and NLOS links [75]. With the relay assistance, high-speed
increases, the data rate reduces due to the higher loss. Up to communications can be achieved over hundreds of meters to
500 Mb/s data transmission has been achieved over 100 m provide the highly desirable underwater wireless connectivity.
UOWC channel using a 520 nm laser with the simple on-off-
keying (OOK) modulation [70], and by using more spectral
efficient DMT modulation, up to 5 Gb/s has been achieved over C. Optical Wireless Interconnects in Data-Centers
50 m underwater link based on a 450 nm laser transmitter [71]. In addition to in-building/indoor and underwater environ-
Therefore, high-speed wireless connections can still be achieved ments, the short-range OWC technology has been further consid-
over a relatively long distance in the underwater environment. ered to realize high-speed wireless interconnects in data-center
It also worth noting that most long distance demonstrations networks in the past decade. With the explosive growth of data
consider the pure sea water channel, which has small c(λ). and the wide deployment of cloud-based services, the data-
When considering the turbid harbor water channel with much center networks require high-speed, efficient and reconfigurable
larger c(λ), the achievable communication distance will be much interconnect links. Conventional electrical-based interconnects
shorter. struggle to meet these demands, since they have limited band-
Different types of receivers have been studied in UOWC width, high transmission loss, high power consumption and
systems, mainly including PIN, APD and photomultiplier tube also long transmission latency. Hence, optical interconnects
(PMT). PIN-based receivers are most widely used due to their have become a popular option, such as the optical fiber-based,
simple operation and low cost. Compared with PIN, APD re- the polymer waveguide-based and the silicon photonic inte-
quires higher operation voltage and is more expensive. However, grated circuit-based optical interconnects [76], [77], [78]. Us-
it can provide better receiver sensitivity to extend the communi- ing optical solutions, the data rate and capacity of data-center
cation range. In addition, PMT has also been utilized in UOWC interconnects have been boosted significantly. However, the
system demonstrations, taking advantage of its internal optical optical fiber, polymer waveguide and silicon integrated circuit
gain, low noise, and high sensitivity [72]. Nevertheless, very based interconnects are all wired interconnects connecting fixed
high operation voltage is required (in the order of hundreds of transmitters and receivers, and high-speed switches are needed
volts), limiting the practical application. to provide flexible connections, which typically require high-
In addition to the studies aiming to push the achievable data speed O-E and E-O conversions that are expensive and power
rate and transmission distance of the UOWC link, recently hungry.
the relay-based systems have attracted intensive interests in In addition, due to the bursty nature of traffic, the data-center
underwater environments. Due to the relatively high propagation interconnect loads are normally imbalanced, where parts of
selection block, the receiver optical assembly and the PD. At the
transmitter side, the data is first modulated onto the optical signal
generated by the light source, and the emitted optical beam then
passes through the optical assembly, which normally collimates
the beam to extend interconnect range. Before being emitted to
the free-space, a link selection block is used. By dynamically
changing the propagation direction of the optical beam towards
the desired destination receiver, reconfigurable interconnects
can be realized to satisfy the need in data-center networks to
offload bursty traffic. After the OWC channel, the modulated
optical signal is collected by the receiver optical assembly and
then detected by the PD.
In OWC-based interconnects, the link selection block pro-
vides the critical reconfiguration function to achieve the re-
quired connection flexibility. It has been realized using liquid
crystal on silicon (LCoS)-based SLMs, opto-VLSI processors,
MEMS-based steering mirrors, and recently, silicon integrated
OPAs [82], [83], [84], similar with the beam steering function in
near-infrared indoor OWC systems. To increase the interconnect
speed, two methods have been widely used: WDM and parallel
Fig. 6. (a) Architecture of OWC-based interconnects in data-center networks;
and (b) general operation principle. transmissions using arrayed light source and PD operated at
the same wavelength, such as VCSEL array and PIN PD array.
Compared with the WDM option, the parallel transmission
the interconnect network are under-utilized whilst other parts method does not require precise wavelength control, and hence,
face traffic congestion that leads to significant delays. To avoid it is more cost-effective.
excessive latency, redundant interconnects need to be planned The major limiting factor in OWC-based optical interconnects
and laid out at the initial design and implementation stage. is the natural divergence of light beam while propagating in
However, it increases the cost of data-center network, and it the free-space, particularly when VCSELs with a relatively
also leads to lower overall network efficiency due to the waste large divergence and simple collimation lenses are used due
of majority redundant interconnects. to the low-cost and low-complexity considerations. Due to the
To solve these issues, the wireless data-center network con- divergence of light beam, the optical signal footprint becomes
cept has been proposed and investigated, where wireless inter- larger after free-space propagation. This leads to two funda-
connect links are further added to the data-center network [79]. mental issues: first, inter-channel crosstalk is introduced if the
Due to the flexibility of wireless links, wireless interconnects can spacing between adjacent parallel channels is small; and second,
be deployed to connect the transceivers with high transmission substantial power loss is introduced due to the limited aperture of
capacity requirement to provide extra interconnect bandwidth. PD and the associated focusing optics element. Since the beam
Hence, redundant wired interconnects are no longer needed. For footprint increases with the free-space propagation distance, the
wireless data-center interconnects, one key requirement is large divergence of light beam sets the limit for possible intercon-
capacity, where comparable capacity as the wired interconnect nection range. Whilst increasing the aperture of each receiving
is highly desirable. Therefore, the high RF region has been element and the spacing between adjacent elements can relax
explored, such as the millimeter-wave and terahertz bands [80], this limitation, the compactness of OWC-based interconnect is
[81]. Due to the availability of broad unregulated bandwidth, sacrificed.
the optical frequency has also been investigated to provide To further increase the achievable data rate and communi-
high-speed wireless interconnects [22], [23]. Such OWC-based cation range in OWC-based interconnects, several techniques
interconnects are capable of providing over 100 Gb/s data rate have been proposed and studied. For example, 40 Gb/s data rate
and reconfigurable connections between flexible transmitters per channel has been demonstrated over 30 cm range using the
and receivers without the need of high-speed O-E and E-O higher-order 16-CAP modulation format [43], and by further
conversions. In addition, due to the smaller refractive index of incorporating space-time block coding (STBC), the intercon-
the free-space channel compared with optical fiber, OWC-based nection range has been extended to 50 cm while maintaining the
interconnects can also achieve lower transmission latency. data rate [23]. In addition, the accelerating beam has also been
The general architecture and working principle of OWC- explored in OWC-based interconnects [82]. The 2D Airy beam,
based optical interconnects is shown in Fig. 6. A dedicated which is a typical accelerating beam, can move along a curved
optical interconnect module is incorporated onto each electronic trajectory with invariant intensity profiles (i.e., diffraction-free)
card. Due to the fixed locations of these optical interconnect and is also robust to perturbations (i.e., self-healing). Hence,
modules in data-centers, LOS links can be guaranteed via proper better robustness to obstacles in the free-space channel than
initial design. The optical interconnect module mainly consists the conventional Gaussian beam is demonstrated. The 2D Airy
of the light source, the transmitter optical assembly, the link beam is generated and detected by SLM, and when combined
with WDM, over 100 Gb/s (4 channels) full-duplex (O-band

and C-band) optical interconnects have been achieved. Larger
scale OWC interconnect has also been realized, where up to
16 Gb/s error-free transmission has been achieved using PAM-4
modulation in a 16 × 16 data-center network [22]. Therefore,
the short-range OWC technology provides a promising solution
to achieve high speed and flexible interconnects in data-centers.
D. OWC-Based V2V and V2I Communications

Vehicular communications, including the wireless communi-
cations between vehicles (i.e., V2V) and between vehicle and
the infrastructure (i.e., V2I, e.g., vehicle to smart traffic light Fig. 7. Principle of optical camera communications. (a) Global shutter
scheme; and (b) rolling shutter scheme.
at road intersections), are the physical platform to enable intel-
ligent transportation network (ITS), where connected vehicles
and autonomous driving applications have attracted intensive
interests [85]. In addition to transforming the way we drive the rolling shutter feature of modern CMOS-based cameras has
everyday, they are also widely envisioned to improve the road been explored [33]. In rolling shutter-based images, the pixels
safety and mitigate traffic congestion. are exposed row-by-row instead of being exposed all at once. If
Various wireless communication technologies have been stud- the transmitter (i.e., LED in vehicle lights) is turned on when a
ied to support V2V and V2I links, such as the 5G cellular row is being exposed, a bright line appears in the image; and if
networks [86]. However, due to the congested microwave spec- the transmitter is turned off, a dark line is shown in the image, as
trum, RF-based V2V and V2I links experience long latency, high shown in Fig. 7(b). Hence, each row in the image is equivalent
signal interference and low throughput, particularly in crowded to a time domain sampling point, and the effective sampling rate
traffic scenarios. Therefore, the broadband spectrum available in now equals to the product of frame rate and the number of rows
the optical band and the OWC technology has been intensively in the rolling shutter scheme. Therefore, the data rate supported
studied recently to provide high throughput, low latency and low in the rolling shutter-based vehicular OWC can be increased
interference V2V and V2I communications [25], [26], [27]. In significantly [88].
addition, considering the wide availability of LEDs in vehicles, In both vehicular OWC and VLC systems, LEDs are used as
such as the LEDs in the headlight, tail light or daytime driving the light source. However, unlike only capturing the intensity
light, OWC-based vehicular communications using the visible information with the PIN PD in VLC systems, the camera-
light wavelength are highly attractive. Similar with the indoor based receiver in vehicular communications captures images,
and underwater scenarios, the LED-based OWC in vehicular and hence, the first step in signal detection is deciding the ROI,
communications also faces the limited transmitter modulation i.e., the part of image that captures the LED transmitter. One
bandwidth challenge. Hence, advanced modulation formats have of the most widely used methods to decide the ROI is using
been studied to boost the V2V and V2I data rates [87]. LEDs arranged in special shapes at the transmitter, such as
At the receiver side, conventional PDs, such as PIN PDs, the circular shape of each LED together with the rectangular
have been deployed to detect the signal transmitted by LED shape of the LED array in the vehicle light panel [89], where
transmitters [25]. Recently, considering the widely availability results show 26% and 18% improvements on the detection
of vehicular cameras, such as the rear parking camera, safety accuracy and BER, respectively. Another promising solution is
camera and dash camera, the use of camera-based receivers the ROI-signalling technique, where both low-rate and high-rate
in vehicular communications, which is also referred as OCC, data streams are transmitted simultaneously with the low-rate
is becoming more attractive [27]. In this type of systems, the data stream assisting the detection and tracking of the ROI of
camera captures a video of the LED transmitter, which consists LEDs [88]. In addition, deep learning has also been proposed to
of a series of images (i.e., frames) taken at consecutive time accurately detect and recognize LEDs in the transmitter, such as
instances. If the LED is intensity modulated, the brightness of the LED segmentation recognition method combining ML and
the image region capturing the LED will change accordingly, conventional image processing [90].
and the outputs at different time instances can be considered Compared with VLC systems, one major challenge faced by
as time domain samples of the data transmitted by the LED, as vehicular OWC systems is the much more noisy environment,
shown in Fig. 7(a). Hence, the transmitted data can be recovered where various types of ambient background lights exist and can
and vehicular communications can be realized. be captured by the camera receiver. These include the sunlight in
Whilst the camera-based receiver can achieve OWC-based the outdoor environment and the headlight or taillight from other
data transmission using the method described above, which surrounding vehicles, which can degrade the SNR substantially.
is referred as the global shutter scheme, it is fundamentally To overcome this limitation, one most widely used and simple
limited by relatively low frame rate. Since Nyquist sampling solution is limiting the receiver FOV, which can effectively
needs to be satisfied to recover the data transmitted by LED, reduce the background lights incident from the sides [91].
the communication speed is highly limited. To solve this issue, However, this method also reduces the communication angular
range between vehicles that leads to frequent communication OWC systems have a key limitation of background light induced
interruptions. Another method to overcome the impact of am- extra noise, which particularly limits the system performance in
bient light is exploring novel modulation schemes, such as the outdoor vehicular communications.
color shift keying (CSK) method where data is modulated onto Short-range OWC technologies and systems have attracted
the color of LED transmitters [92]. Analysis has shown that intensive interests and we have witnessed significant and rapid
better SNR performance can be achieved in outdoor vehicular developments. For example, in in-building and indoor OWCs,
communication scenarios. the data rate has been pushed to over Tb/s scale using ten WDM
Another key challenge faced by camera-based vehicular com- near-infrared channels with a transmission distance over 3.5 m
munications is the trade-off between data rate and distance. and a coverage area of about 2.54 m2 (1.8 m diameter) [6]; in
The increased distance between vehicles leads to two major underwater OWCs, the exploration of water surface reflection
consequences. First, the received signal power decreases due has enabled unique NLOS link-based connections overcoming
to the divergence of light beam emitted by LEDs, which leads to channel blocking limitation, where results show using a laser
decreased SNR. Secondly, the increased distance also results in transmitter a data rate up to 100 Mbps over 6 m harbor water
the area in the image occupied by the LED transmitter becoming is possible [67]; in wireless data-center networks, diffraction
smaller. Hence, if rolling shutter is used, the number of pixel free Airy beam has enabled robust >100 Gb/s full-duplex
rows that provides time domain samples of the transmitter high-speed interconnects with minimal crosstalk [82]; and in
becomes smaller, reducing the communication data rate. To vehicular communication, the use of rolling shutter mechanism,
solve this issue, several methods have been proposed, such as ROI-signalling, and deep learning has realized robust and high
limiting the divergence of LED transmitters with collimation throughput V2V and V2I links [88], [90].
optics, and using multiple LEDs in the transmitter to increase In addition to the short-range OWCs introduced above, OWC
the transmission power and to enlarge the size of image capturing has also been widely considered for indoor localization. The
the transmitter at the camera receiver [33]. indoor localization function is highly demanded in near-infrared
In addition to the OWC-based vehicular communications OWC systems employing relatively narrow steerable beams to
relying on the LOS channel, the NLOS channel also has been maintain wireless connectivity. Whilst it can be realized using
explored, particularly taking advantage of the reflections from conventional RF technologies (e.g., WiFi), the accuracy is nor-
the surfaces of surrounding vehicles [93]. Results show that mally limited by signal shadowing, reflections and interference.
practical communication distances can be achieved using the The OWC-based solution has been proposed to achieve more
reflection-based NLOS channel from white, orange and black accurate positioning and at the same time avoid requiring an
vehicles. In particular, the white vehicle reflections can enable extra system for localization. A number of solutions have been
communication distances over 17.5 m, providing a promising demonstrated, exploring the received signal strength (RSS) [96],
solution to V2V and V2I communications beyond LOS. To time-of-arrival (ToA) [97] and angle-of-arrival (AoA) [98] in-
overcome the LOS limit and to increase the communication formation. OWC localization with camera receivers has also
range, the relay-assisted OWC vehicular communications has been studied [99]. With these techniques, centimeter-scale or
also been proposed [94], where intermediate vehicles between even better accuracy has been achieved [100] and simultaneous
the transmitter and the receiver are used as relay nodes. Both AF position and orientation estimation has also been realized [101].
and DF relaying schemes are studied, and results show that the
more complicated DF scheme can improve the communication III. ADVANCED TECHNIQUES IN SHORT-RANGE OWCS
distance by up to 150% compared with AF [95], providing
As discussed in the Section II, short-range OWCs have been
a potential solution for long-distance OWC-based vehicular
widely studied in various scenarios. There has been significant
communications.
progress recently, where a large number of advanced techniques
have been proposed and demonstrated. In this section, we discuss
E. Summary of Short-Range OWCs a few selected advanced techniques that have attracted intensive
interests in short-range OWCs, focusing on the spatial domain
In this section, we have briefly introduced four types of
MIMO techniques, and the ML and neural network techniques.
short-range OWC systems, ranging from indoor to outdoor and
underwater scenarios. Whilst the channel properties, require-
A. Spatial Domain MIMO in Short-Range OWCs
ments and limitations are different, they all rely on the trans-
mission of data modulated optical waves through the wireless To boost the performance of short-range OWC systems,
channel, and hence, the key advantages are the availability of the symbol/time domain and the frequency domain have been
broad unregulated bandwidth resource, the immunity to con- widely studied, such as advanced modulation formats and the
ventional electromagnetic interference, the typical LOS signal WDM technology to increase the communication data rate. In
transmission providing inherent physical layer security, and the addition, the spatial domain has also been explored, leveraging
reduced transmission latency than optical fiber systems due to the advantages of having multiple OWC transceivers. Both
the smaller refractive index of free-space channel. Both LEDs MIMO with multiple transmitters and multiple receivers, and
and lasers have been used as transmitters, and both photodiodes multi-input single-output (MISO) with multiple transmitters
and cameras have served as receivers. In addition, due to the and single receiver, have been studied. In particular, the spatial
ubiquitous existence of ambient light, all of these short-range diversity, spatial multiplexing and spatial modulation principles
single transmitter system or the OWC system with spatial diver-

sity. For example, spatial multiplexing can be applied in camera-
based OWC systems, such as vehicular communications, by
exploring the parallel signal detection capability enabled by the
large number of pixels in camera. Through spatial multiplexed
parallel data transmission, the capacity limited by camera frame
rate can be improved, where beyond 100 kb/s data transmission
has been achieved using a camera receiver with a low frame rate
of only 60 fps [103].
In both spatial diversity and spatial multiplexing schemes,
normally all available OWC transmitters are activated simultane-
ously. Another possible method to utilize the spatial transmitters
is activating them selectively, which is the spatial modulation
scheme as shown in Fig. 8(c). Here, multiple transmitters are
available and each is allocated a unique transmitter index. The
data to be transmitted is split into two parts. One part is trans-
mitted via normal symbol domain modulations (such as OOK or
multi-level PAM), and the other part is transmitted in the spatial
domain via the index of activated transmitter. One example is
shown in the figure, where four transmitters are available and
one transmitter is activated at any time. Hence, according to
the spatial mapping, 2 bits-per-symbol can be transmitted in
the spatial domain. After free-space transmission, the receiver
detects the data transmitted in both the symbol domain (i.e.,
modulated symbol) and the spatial domain (i.e., the index of
activated transmitter). Therefore, the OWC system data rate can
be improved by transmitting extra data in the spatial domain. To
Fig. 8. Principle of spatial domain MIMO techniques. (a) Spatial diversity; further increase the data rate, more than one transmitters can be
(b) spatial multiplexing; and (c) spatial modulation. selectively activated, which allows more data to be transmitted
in the spatial domain, and this is referred as generalized spatial
modulation [104].
In short-range OWC systems with multiple transmitters, as-
have been investigated, aiming to improve both robustness and sume the data transmitted by the ith transmitter at time t is
capacity of short-range OWCs. xi (t), then the detected signal yj (t) by the jth receiver can be
1) General Principles: the basic principles of spatial diver- expressed as:
sity, spatial multiplexing and spatial modulation in short-range n

OWC systems are shown in Fig. 8. We focus on the transmitter yj (t) = Rj · hij (t − τij ) · xi (t − τij ) + nj (t) (2)
side, where multiple OWC transmitters are employed, and these i=1
principles can be applied to both MIMO and MISO configura-
tions. In the OWC system with spatial diversity, the same data is where hij (t) is the channel response from the ith transmitter to
transmitted by multiple transmitters, and hence, spatial redun- the jth receiver, τij is the corresponding channel propagation
dancy is achieved. Since the possibility of all OWC channels delay, nj (t) is the additive noise of the jth receiver, and Rj is
being blocked at the same time is very low, the robustness of the PD responsivity. Here we only consider direct LOS channels
OWC system can be enhanced significantly. This is particularly and ignore multi-path signals to better focus the subsequent
important for in-building and indoor OWC systems that rely on discussions on channel delay and similarity.
high-speed LOS links, where the end user movement may lead to 2) Channel Delay in Short-Range OWCs With Spatial Diver-
some links being blocked. It also worth mentioning that in spatial sity: When spatial diversity is applied, the same data is trans-
diversity, all OWC transmitters may transmit identical signals, mitted by all transmitters. Whilst redundancy is achieved, due to
such as the repetition coding (RC) scheme, or transmit different different propagation lengths of multiple OWC channels, optical
signals with repeated information in adjacent time slots (i.e., signals carrying the same data arrive at the receiver at different
coded spatial-temporally), such as the space-time block coding time instances. Taking an OWC system with two transmitters as
(STBC) [102]. an example, the received signal y(t) can be described as:
Another MIMO principle that has been widely studied in y(t) = R · [h1 · x(t − τ1 ) + h2 · x(t − τ2 )] + n(t) (3)
OWC systems is spatial multiplexing. As shown in Fig. 8(b),
spatial multiplexing uses multiple transmitters to send different where we assume the OWC channels are stable during this data
data at the same time instance, i.e., parallel data transmission. transmission period. It can be seen from (3) that the channel
Therefore, the overall data rate is improved compared with the delay leads to inter-symbol-interference (ISI), which degrades
the system performance. To suppress the impact of channel by this type of spatial multiplexing is that as expressed by (2),
delay, one simple method is synchronizing the OWC channels multiple signals from different transmitters need to be detected
through channel state feedback and adding the relative delay simultaneously, which requires low channel correlation for cor-
to transmitted signals. However, frequent channel feedback is rect detection. However, in short-range OWC scenarios, due to
required, as in high-speed short-range OWC systems even a the short free-space transmission distance, the channels are typi-
small path length difference can lead to a large channel delay cally highly similar and highly correlated, particularly in indoor
and severe ISI. To solve this issue, the channel equalization environments. To solve this issue, the adaptive power allocation
together with added cyclic prefix (CP) or zero padding (ZP) has method has been proposed, where the power imbalance between
been demonstrated [105]. However, the effective system data transmitters reduces the channel correlation and improves the
rate is reduced due to the extra CP or ZP, particularly for long spatial gain [112].
channel delays. The code division multiple access (CDMA) is 4) Short-Range OWC Systems With Spatial Modulation: in
another option [106]. Whilst orthogonal codes can suppress the spatial multiplexing, the data rate is improved through paral-
ISI amongst multiple channels, typically long codes with high lel data transmission, which increases the overall transmission
complexity are needed. power that is particularly problematic for short-range OWC
Another recently proposed method to tackle the channel delay systems with laser transmitters due to the safety concern. On
issue in short-range OWC systems with spatial diversity is the other hand, the spatial modulation principle transmits extra
exploring orthogonal waveform filters [107], where a dedicated data via the index of selectively activated transmitter, and hence,
filter is allocated to each transmitter. Due to the orthogonality it can improve the data rate without increasing the transmission
of these filters, by applying matched filters at receiver, signals power. Therefore, spatial modulation has also attracted intensive
from different transmitters can be decoupled and hence, the interests in short-range OWCs.
ISI can be suppressed without sacrificing the effective data rate As shown in Fig. 8(c), data can be transmitted via both the
or requiring long complicated codes. The proposed orthogonal spatial domain and the symbol domain. The simplest way to
waveform filters-based method has been experimentally veri- implement spatial modulation in short-range OWC systems is
fied in a 2.5 Gb/s near-infrared indoor OWC system with RC, only transmitting data in the spatial domain, i.e., the transmitter
and results have shown that similar BER performance can be remains on for the whole period when it is activated. This scheme
achieved for over 8 symbol periods channel delay. In addition is also referred as space shift keying (SSK). To use the spatial-
to this method, the use of ML has also been studied to solve temporal resources more efficiently, modulated symbols can be
the channel delay issue [108], [109], which will be discussed in transmitted when a transmitter is turned on selectively, such as
detail in Section III.B. It is also worth noting that all these studies sending out PAM symbols to convey extra information. Another
focus on the single-user scenario. Overcoming the channel delay more efficient way to use the spatial-temporal resources is still
in OWC spatial diversity systems with multiple users is still only transmitting data via the spatial domain, but instead of
challenging, and novel approaches and detailed investigations activating one transmitter at any time, multiple transmitters are
are required. activated simultaneously. Therefore, more complicated spatial
3) Short-Range OWC Systems With Spatial Multiplexing: constellations can be sent to increase the data rate, and this
the spatial multiplexing has been a hot topic in short-range scheme is referred as the generalized space shift keying (GSSK).
OWC systems to boost the transmission data rate. In particular, Further exploring this possibility, multiple transmitters can be
spatial multiplexing is a natural match with imaging receivers activated simultaneously to transmit modulated symbols, result-
and camera-based receivers. In the imaging receiver, the imaging in the generalized spatial modulation scheme that increases
ing lens maps transmitters to the focal plane. Due to different the OWC data rate via both spatial and symbol domains.
locations of these transmitters, the corresponding images locate In spatial modulation, the modulated symbols and the indices
differently on the focal plane. Hence, by using a PD array, mul- of selectively activated transmitters are detected simultaneously.
tiple signals can be collected and detected by different elements Hence, the detection errors in both domains are coupled and
in the array, and parallel data transmission can be realized. Using affect each other. Due to this, low channel correlation is required
an imaging receiver with a 12 × 24 PD array in an indoor OWC so that the spatial domain information can be detected correctly,
system, signals with spatial multiplexing have been separately which is the prerequisite for correct detection in the symbol
spatially and over 8 Gb/s data rate has been realized with the domain. However, in short-range applications, particularly in
simple OOK modulation [110]. The camera-based receiver is indoor environments, the multiple OWC channels are typically
another natural option to detect signals with spatial multiplexing highly similar. Therefore, appropriate design of the transmitter
in short-range OWC systems. Due to the availability of large and/or receiver is required to obtain effective gain using the
number of pixels, the camera-based receiver can be considered spatial modulation principle.
as a high-dimensional PD array. By implementing multiple ROI Several possible solutions have been proposed and studied,
detection, spatial multiplexed signals from different transmitters such as exploring the power allocation amongst multiple trans-
can be separated and detected to increase the effective data mitters to reduce the similarity in channel gains [113], using
rate [111]. transmitters with different orientations or PDs with angle di-
In addition to the imaging and camera receivers, spatial versity to mitigate the channel similarity [114], and optimizing
multiplexing has also been widely studied in short-range OWC the constellation design considering both spatial and symbol
systems with conventional PIN PDs. The major challenge faced domains to maximize the Euclidean distance between received
constellation points [115]. With these methods, the high channel not necessarily satisfied in short-range OWCs, particularly in
correlation issue in short-range OWC systems has been relieved stable indoor environments. A few recent works aiming to solve
and the communication performance has been substantially these limitations have been briefly reviewed. We would like
improved. Another recently proposed method to solve the chan- to highlight that in addition to the limited number of works
nel similarity issue in short-range OWC systems with spatial mentioned above, there have been a large number of recent
modulation applies unique waveform filters to multiple OWC relevant studies focusing on various challenging aspects and
transmitters [116]. One dedicated waveform filter is allocated to pushing the system capability [118], [119]. In particular, the
each OWC transmitter to enhance the channel uniqueness, and massive MIMO has been investigated, leveraging the typical
these waveform filters are designed to be orthogonal. Therefore, large number of LEDs in the transmitter array [120]. Due to
by using matched filters at the receiver side, the index of activated the large number of signal beams, massive MIMO can enable
OWC transmitter can be detected by monitoring the outputs beam division multi-access to serve a large number of users,
of matched filters since only the matched filter corresponding where high spectral efficiency and access to hundreds of users
to the activated transmitter has significant non-zero output. are illustrated. The collaborative efforts have resulted in the rapid
This transmitter index detection process also separates symbol advancement in short-range OWCs.
detection from spatial detection, and hence, the BER perfor-
mance can be significantly enhanced. The filter enhanced spatial
modulation principle is experimentally demonstrated. With two B. ML and Neural Networks in Short-Range OWC Systems
near-infrared OWC transmitters and PAM-4 symbol modulation, In the past a few years, we have witnessed an explosive
up to 7.5 Gb/s data rate has been achieved, which provides a increase of the interest in the ML technique, particularly deep
50% data rate improvement compared with the system without learning and neural network algorithms [121]. As a data-driven
spatial modulation. This principle has also been extended to technique, the ML algorithm learns complicated effects and de-
generalized spatial modulation, where multiple transmitters with pendencies based on training data and it has been widely applied
dedicated waveform filters are activated simultaneously [117]. in various fields, such as computer vision, image processing,
Using three near-infrared laser transmitters and 16-CAP symbol natural language processing, and recommender systems. In ad-
modulation, up to 15 Gb/s data rate is achieved with 3 bits-per- dition, ML algorithms have also been investigated in optical
symbol transmitted via the spatial domain. Results also show communications to capture and handle various impairments
that compared with the indoor OWC system with conventional during optical signal generation, modulation, transmission, am-
spatial modulation, the BER performance is improved by over plification, switching and detection [122].
two-orders-of-magnitude. In general, ML algorithms, which can be categorized into
5) Summary: In the previous sub-sections, we have briefly unsupervised learning, supervised learning and reinforcement
introduced the basic principles of spatial domain MIMO tech- learning types, have been applied in both the physical layer
niques to boost the performance of short-range OWC sys- and the network layer in optical communications [122], [123].
tems.MIMO techniques explore the multipath transmission us- In supervised learning, the training dataset with known desired
ing multiple transmissions and/or receptions, and in short-range output is used and the algorithm learns a mapping from the input
OWCs the spatial diversity, spatial multiplexing and spatial to the output; in unsupervised learning, the training dataset only
modulation MIMO schemes have been widely studied. Spatial has unlabeled inputs without knowing desired outputs and the
diversity explores the redundancy provided by multiple OWC ML model tries to cluster the inputs into groups; and in rein-
links and it has been widely used to improve system robustness, forcement learning, the algorithm takes an action and receives
particularly against possible LOS link blockages. However, feedback from the environment that indicates the degree of
since typically the same data is transmitted over multiple chan- satisfying the goals of the learner to sequentially learn a problem
nels, the data rate is not improved and the channel delay can solving strategy. In the network layer, ML has been studied for
lead to substantial performance degradation. On the other hand, many applications, such as modulation format recognition, fault
spatial multiplexing and spatial modulation explore the spatial detection and positioning, dynamic bandwidth allocation and
resources to increase the communication data rate. In spatial latency reduction, where the supervised learning and reinforce-
multiplexing, different data streams are transmitted in parallel ment learning methods are typically used. Taking the dynamic
by all available OWC transmitters, and hence, the effective bandwidth allocation as an example, the ML algorithm can learn
data rate can be enhanced significantly. However, the signal key characteristics of the network traffic and then predict possi-
detection is subject to high level of interference and requires ble future traffic pattern. Instead of defining and extracting traffic
knowing accurate channel state information (CSI), leading to characteristics (e.g., the arrival rate distribution, the network
lower quality of transmission. In spatial modulation, the OWC load distribution, and the traffic recurrent pattern) manually,
transmitters are selectively activated to convey extra data via which is difficult due to highly complicated network traffic,
the index (indices) of OWC transmitters. Due to the selected the ML algorithm automatically extracts traffic profiles from
access of the wireless channel, the interference is better than historical traffic and makes accurate predictions, which enable
spatial multiplexing. However, the data rate improvement is also dynamic resource allocation in a proactive manner to optimize
smaller whilst knowing accurate CSI is still required. In addition, the network performance. In addition to the network layer, ML
both spatial multiplexing and spatial modulation requires low has also been widely investigated for the physical layer signal
channel correlation among multiple OWC channels, which is processing by capturing the impacts of various impairments and
TABLE I
ML AND NEURAL NETWORK IN SHORT-RANGE OWCS
suppress them accordingly. Typically the physical layer pro-

cessing relies on unsupervised learning and supervised learning
techniques. Compared with conventional digital signal process-
ing (DSP) algorithms, ML algorithms can learn the effects of
multiple impairments simultaneously and are particularly good
at handling nonlinearities. Therefore, better data transmission
performance has been demonstrated [124]. In this section, we
will introduce and review some recent works on ML assisted
short-range OWC systems, which are also summarized in Table
1.
1) ML-Based Signal Processing and Compensation in Short-
Range OWC Systems: as briefly discussed above, two key Fig. 9. (a) UOWC system with deep neural network; and (b) architecture of
the Gaussian kernal-aided deep neural network. Redrawn from [125].
strengths of ML techniques are the capabilities in handling
multiple impairments simultaneously considering the cross-
interactions, and learning and suppressing nonlinear effects hidden layers learn the impact of nonlinearities on the received
in optical communication systems, which are challenging for signal, and the output layer generates the final symbol decision
conventional DSPs. These capabilities have been explored in using a softmax classifier. In addition, a Gaussian kernal layer
short-range OWC systems recently. Focusing on the nonlinear is added between the input layer and the first hidden layer,
effect compensation strength, a LED-based UOWC system with which is used to better model the ISI. The cross-entropy loss
a ML-based signal processor has been proposed, as shown in function is adopted for the training process, which relies on the
Fig. 9(a) [125]. Due to the strong nonlinear response of the gradient descent back propagation. The UOWC system with the
LED, the nonlinear amplification of the driver circuit and the neural network-based signal processor has been experimentally
nonlinear response of the PD, the LED typically can only be demonstrated. Results show that the use of Gaussian kernal
driven by signals within a small range and the communication can reduce the training iterations by over 45% and outperform
performance is limited. The ML-based signal processor is used at the fully-connected neural network without the Gaussian kernal
the receiver to solve this issue after channel equalization, where by about 25%. The LED transmitter data modulation range is
a Gaussian kernal-aided deep neural network is employed. The extended from 0.2 V to 0.35 V, and the BER performance is
architecture of the neural network is shown in Fig. 9(b), which is improved by more than 1.5 dB compared with the Volterra
based on a fully-connected neural network. The input layer takes nonlinear equalizer, showing the capability of ML algorithm to
the received signal after O-E conversion and equalization, the suppress the nonlinear effects in the short-range OWC system.
Fig. 10. Architecture of two-connected neural networks for signal processing

and detection in UOWC systems with SPAD. Redrawn from [126].
The fully-connected neural network has also been studied to

combat the nonlinearity of single photon avalanche photodiode
(SPAD) in an UOWC system [126]. SPAD has better sensitivity
compared with other receivers to extend the communication
range, whilst its large nonlinearity limits the signal detection.
This issue is solved by using the deep learning aided signal
detection. Different from the previous work, here two connected
neural networks are integrated as the signal processing block,
as shown in Fig. 10. Both neural networks are based on the
fully-connected architecture, with the first one as the channel
compensation block and the second one as the demodulator.
Hence, both equalization and data detection are realized simulta- Fig. 11. (a) RNN-based signal processor in the indoor OWC system with
neously and the channel information is no longer needed. Results spatial diversity; (b) architecture of the LSTM-RNN; and (c) structure of the
show that the BER performance can be improved compared with LSTM cell. Redrawn from [108].
the logarithmic likelihood ration (LLR)-based soft-decision.
Compared with the use of one neural network together with
LLR soft-decision, the underwater communication range can be t, a received data rt is passed to the LSTM cell, and it updates
extended by up to 3 m in clear ocean water. the cell state and the hidden state via four gates, i.e., a forget
Whilst the fully-connected neural network has shown the gate ft that controls which information in the hidden state of
capability in short-range OWC systems, it has a large number of the previous time step needs to be removed, an input gate it
connections between neurons, which leads to the large amount that decides what information needs to be updated in the current
of learnable parameters and the need for large training datasets. cell state, a candidate gate ct that controls the amount of update
To solve it, the long-short term memory (LSTM)-based recurrent needed, and an output gate ot that decides outputs of current
neural network (RNN) has been studied. It has been studied to time step. The tanh nonlinear activation function is also used
suppress the ISI and nonlinear responses in an indoor OWC in the LSTM cell. Hence, the processing inside the LSTM cell
system with multiple transmitters and spatial diversity [108], can be considered as linear projections wrapped in nonlinear
where different path lengths from transmitters to the receiver functions. Through this way, various impairments in the indoor
lead to severe ISI and limit the performance. Compared with the OWC system with spatial diversity are learnt and processed.
fully-connected neural network, RNN processes the received Experimental results show that the LSTM-RNN-based scheme
signal in a recurrent manner, and hence, it is naturally suitable is capable of simultaneous signal processing and decision in the
to process periodic data received. indoor OWC system with two near-infrared transmitters, and
The principle of LSTM-RNN in indoor OWC systems is channel delay with lengths up to 0.6 symbol period is suppressed.
shown in Fig. 11 [108], which serves as a one-stop signal It outperforms the fully-connected neural network by >0.6 dB,
processor and detector. Here, the received data is divided into together with a small training dataset and a fast training process.
symbol sequences, where a symbol sequence is a number of sym- LSTM-RNN has also been further extended to the attention
bols r1 , r2 , . . ., rl that are processed and decided together. The augmented scheme (ALSTM-RNN) by adding an attention layer
channel impairments are learnt from these symbol sequences before the output layer [109]. The attention layer can maintain
passed from the input layer to the bidirectional layer, where information at all time steps, and hence, it has shown enhanced
LSTM cells are used. Afterwards, the information learnt in the capability in capturing impairments that lead to long-term effects
forward and backward directions are combined in the output amongst the received data in short-range OWC systems, further
layer and the final symbol decision y1 , y2 , . . ., yl is generated improving the BER performance. In the same indoor OWC
after the softmax function. In the LSTM-RNN signal processor, system with two transmitters, up to 5.5 symbol periods channel
channel impairments are captured and suppressed by LSTM delays are suppressed, and the BER improvement compared
cells. The basic structure of LSTM cell is shown in Fig. 11(c), with the LSTM-RNN scheme is >60%, which increases with
which keeps both a cell state and a hidden state. At a time step the length of channel delay.
square error (MSE) and the autocorrelation of learnt mapping

symbols are incorporated in the loss function. The MSE is widely
used in ML algorithms to optimize the difference between
the processed data and ground-truth. On the other hand, the
Fig. 12. Autoencoder-based VLC system. Redrawn from [128]. autocorrelation captures the unique peak-to-average power ratio
(PAPR) issue that exists in the LED-based system with OFDM
modulation. Hence, both the data-driven and the model-driven
In addition to the fully-connected neural network, LSTM- aspects are considered, and system impairments can be better
RNN and ALSTM-RNN, other types of ML architectures have mitigated. Results show that compared with Volterra-based DSP
also been explored, such as the convolutional neural network and purely data-driven autoencoder scheme, the PAPR can be
(CNN), the deep belief network (DBN), and the adaptive boost- reduced by >5 dB, the BER can be improved by over one
ing (AdaBoost) [127]. These neural networks have been used order-of-magnitude, and the training process can be accelerated.
for the physical layer signal demodulation in a VLC system with An autoencoder-based ML scheme is also studied in [131],
LED transmitter. The CNN scheme converts received signals to where the binary signal design in a VLC system has been
images and uses image classification to achieve demodulation; considered. Similarly, the loss function is designed to explicitly
the DBN scheme uses three restricted Boltzmann machines to incorporate the dimming constraint to achieve better system
capture and process modulation features; and the AdaBoost performance.
uses a strong classifier constructed from k-nearest neighbor 3) FPGA-Based Neural Network Hardware Accelerator in
(kNN)-based weak classifiers. These ML schemes have been Short-Range OWCs: in the neural network-based short-range
studied under different modulation formats that are widely used OWC works discussed above, the neural network physical layer
in VLC systems, including OOK, pulse position modulation signal processors are implemented offline using high-end CPUs
(PPM), and various orders of quadrature amplitude modulation or GPUs. Whilst they have exceptional computation capabil-
(QAM). Results show that the AdaBoost scheme achieves the ity and software programming flexibility, high-end CPUs and
best BER performance. GPUs typically have high cost and high power consumption.
2) ML Autoencoder in Short-Range OWCs: in addition to This is particularly problematic for short-range OWC systems,
signal processing and detection at the receiver side, ML tech- as the neural network needs to be implemented mostly at
niques have also been explored as the autoencoder in short-range the end user side, which are highly sensitive to cost and are
OWC systems, where the transmitter, OWC channel and receiver power-constrained. In addition, the latency caused by neural
are considered as a single ML network that is trained as an network signal processing is not considered, whilst it is critical
autoencoder [128]. Autoencoder is a ML algorithm that typically for the communication type of applications.
leverages neural network for representation learning. It aims to To solve this issue, neural network hardware accelerators
copy its input to its output by encoding the input to a lower have been proposed and investigated [132], [133], [134].
dimensional latent representation and decoding it back with min- Both application-specific integrated circuit (ASIC) and field-
imal reconstruction error. Hence, in OWC systems the autoen- programmable gate arrays (FPGA) have been studied to realize
coder encodes the data to be transmitted d (i.e., the transmitter neural network hardware accelerators. ASICs have high com-
side) into a representation that can be recovered to the original putation speed as well as low power consumption. However,
input at the output side (i.e., minimize the difference between dˆ they are normally designed for only one or few selected tasks,
and d), as shown in Fig. 12. Hence, the whole communication and hence, they are fabricated in low column leading to high
system is trained in an end-to-end manner without the need of overall cost and have limited application flexibility. On the other
rigorous mathematical models, and the autoencoder can jointly hand, FPGAs are highly reconfigurable. Therefore, they can
optimize the overall data recovery performance. The autoen- support the implementation of reconfigurable neural network
coder scheme has been applied to a RGB-based multi-colored architectures and algorithms for various applications flexibly.
VLC system with dimming control, focusing on overcoming FPGAs also have the advantages of parallel computing capabil-
the peak intensity and dimming constraints. Efficient coding ity and low power consumption. Consequently, the FPGA-based
rules can be learnt to optimize symbol constellation design, sig- hardware accelerator is considered as a promising solution to
nificantly improving the symbol-error-rate (SER) performance practical neural network implementations.
compared with conventional constellations. Results show that In the previous FPGA-based neural network hardware accel-
the SNR required to achieve the same SER can be reduced by erator studies, the focus is on processing large and complicated
>5 dB. datasets with high resource usages, whilst the processing latency
The autoencoder-based ML, which is an unsupervised learn- and power consumption are not critical issues [132]. However,
ing algorithm, has also been studied to mitigate the nonlinearity in short-range OWC systems, the dataset is typically small and
in VLC systems. Instead of fully relying on data-driven, the simple, whilst the key requirements include low processing
communication domain knowledge is incorporated in the model- delay for time-critical and real-time communications, low power
driven autoencoder in the design of ML network architecture and consumption for portable end users, and low hardware resource
the training loss function [129]. Separate fully-connected neural usage for low cost (e.g., compact FPGAs).
networks are used at both the encoder (i.e., transmitter) and the To achieve these targets, a FPGA-based RNN hardware ac-
decoder (i.e., receiver) sides. To train the network, both the mean celerator has been proposed and demonstrated in short-range
system or network conditions, and it needs to be re-trained

regularly. Furthermore, the high power consumption is also a
major limitation of using ML techniques in OWCs, particularly
for short-range applications with highly limited power supplies.
Whilst this section introduces some recent progresses, it should
be noted that there are many more relevant studies are not
introduced. For example, ML algorithms have been applied in
vehicular OWC systems to provide more accurate and compre-
hensive channel modelling considering additional vehicle mo-
bility and environmental effects [135], in indoor OWC systems
Fig. 13. FPGA-based RNN hardware accelerator for indoor OWC systems.
Redrawn from [130].
to provide accurate positioning in addition to high-speed data
transmission [136], and in VLC systems to significantly enhance
the security against eavesdropping [137]. In addition, recently
the federated learning (FL), which is type of distributed ML
OWC systems [130], with the general architecture shown in technique, has also been proposed and studied in VLC systems
Fig. 13. The hardware accelerator mainly consists of the off-chip to overcome the privacy and communication overhead concerns
memory that stores the dataset for processing (i.e., the received associated with the centralized ML algorithms [138]. Whilst
OWC data after detection and sampling), the internal on-chip still at an early stage, incorporating FL is a promising solution
memory that keeps RNN parameters (e.g., weights and bias), to provide ubiquitous and reliable connections in VLC systems.
the micro-controller that controls the overall data flow, the AXI
stream bus that provides the data movement path, the clock
generator that generates the clock signal for parallel processing,
C. Other Recent Advanced Techniques
and RNN accelerator core that implements the IP (i.e., RNN
architecture) for indoor OWC systems. In the FPGA-based In addition to the spatial domain and ML techniques briefly
RNN hardware accelerator, in order to use the available hard- introduced above, a large number of other advanced techniques
ware resources more efficiently and to reduce the processing have been proposed and studied in short-range OWCs during the
latency, parallel RNN cells are implemented and novel additive past a few years. One of them is the exploration of NLOS-based
data insertion and tripe-phase clocking optimizations are been indoor OWC systems to avoid communication interruptions
used. The RNN hardware accelerator has been experimentally caused by channel blockage and improve robustness. In [139],
demonstrated based on an indoor near-infrared OWC system a beam reconfigurable NLOS OWC system based on the diffuse
with spatial diversity (two transmitters), and vanilla RNN is reflection focusing principle is proposed, where the wavefront of
implemented using the Xilinx VC709 FPGA board. Results the light incident onto a diffuse reflection surface (e.g., wall) is
show that up to 5.5 symbol periods delay can be handled with spatially shaped at the transmitter so that the amount of diffusely
similar BER results compared with the GPU-based neural net- reflected light reaching the receive is enhanced. A remarkable
work implementation, whilst the latency is reduced by about 30 Gbit/s data rate has been achieved with an over 17 dB power
61% and the power consumption is reduced by over 58%. In gain. In addition, the concept of intelligent reflection surface
addition to the RNN, FPGA-based hardware accelerator has also (IRS) is also promising in enabling short-range OWC systems
been demonstrated for CNN and binarized CNN (BCNN) [134]. with NLOS links, where reconfigurable reflecting elements are
Whilst the millimeter-wave radio-over-fiber systems are consid- deployed to change the wireless propagation channel [140]. With
ered, the hardware accelerator can also be adapted to short-range IRS, the LOS channel blockage issue can be solved by using the
OWC systems, taking the reconfigurability advantage of the NLOS channel instead, which consists of two non-blocked LOS
FPGA platform. channels from the transmitter to the IRS and from the IRS to the
4) Summary: In this section, we have briefly introduced the receiver, respectively. The IRS can be dynamically reconfigured
trending topic of applying ML techniques in short-range OWC to guide the incoming light wave to different receiver locations.
systems. Here we have mainly focused on the application in The short-range OWC technology recently has also been
the physical layer to combat various types of impairments and studied in device-to-device (D2D) communications, particularly
improve the signal reception performance. The hardware accel- for Internet-of-Things (IoT) devices to directly exchange data
erator to implement such ML algorithms in practical systems has between IoT devices to improve resource usage, energy effi-
also been discussed. The major strengths of ML in short-range ciency and latency. Leveraging LiFi and the wide availability of
OWCs include the capability of handling multiple cross-coupled LED-based lighting infrastructure in industrial environments,
impairments simultaneously, extracting complicated historical OWC-based industrial IoT network with D2D connections has
information for accurate predictions, and solving various types been investigated [141]. Taking advantage of the maturing VC-
of nonlinearity existing in the system or network, which are chal- SEL technology, VCSEL-based D2D communications have also
lenging for conventional signal processing techniques. However, been studied [142]. In particular, due to the broad modulation
as a data-driven technique, large training datasets are required, bandwidth, up to 7.2 Gb/s D2D data rate has been demonstrated
resulting in high computation requirement and long training using a 940 nm VCSEL array together with OFDM, bit loading
time. Such training datasets may be difficult to obtain as well and adaptive power allocation. In addition, hybrid RF-OWC
in reality, such as leading to large communication overhead. In communications that combine the robustness against blockage in
addition, the ML algorithm is typically trained under particular RF and the broad bandwidth in OWC systems have also attracted
substantial interest, providing a promising solution for future where it can provide a promising solution to the wireless ICT
D2D connections [143]. infrastructure in industrial environments.
To further improve the capacity, novel modulation techniques In addition to the rapid technological development in short-
have been studied in short-range OWCs as well, such as multi- range OWCs, standardization activities have also progressed.
band CAP (mCAP). Whilst CAP with good spectral efficiency IEEE released the standard Short-Range Wireless Optical Com-
can be implemented using intensity modulation and direct de- munication Using Visible Light, IEEE 802.15.7, in 2011, which
tection with low complexity, it is sensitive to the attenuation includes three different physical (PHY) implementation types
at high frequency, which is particularly problematic in VLC supporting data rates up to 96 Mb/s [148]. Later the revised
systems. mCAP is capable of solving this problem by using version further extends the communication spectrum to the
multiple subcarriers, where each has a relatively small band- whole light spectrum including ultra-violet (UV), visible and
width. Bit-loading can be further applied to these subcarriers to infrared wavelengths. Additional PHY types are also included
further improve the system capacity. Using LEDs with 4 MHz for image sensor-based OWCs [149]. However, the PD-based
bandwidth and 20 subcarriers, up to 249 Mb/s data rate has high-speed OWC, which captures recent progress in LiFi and
been achieved [144]. To achieve better data rate, non-orthogonal other short-range OWC technologies, is not included in the
m-CAP has been further proposed by allowing partial overlap revised version of IEEE 802.15.7. Hence, new standardization is
between sub-bands [145]. Up to 43% better spectral efficiency needed to promote short-range OWC technologies into practical
than orthogonal mCAP has been demonstrated in a visible light applications, and also to ensure the interoperability among man-
UOWC system using a neural network based receiver. ufacturers. Whilst more efforts are required, the IEEE 802.11bb
Exploring new light sources for short-range OWCs is also a is promising, which considers the integration of LiFi in IEEE
recently trending topic, such as the organic LED (OLED), which 802.11 local area networks and supports up to 5 Gbit/s data rate
has better flexibility, easier integration and wider coverage angle. [150].
However, the modulation bandwidth of OLED is typically lower
than conventional semiconductor LED. To overcome this issue, REFERENCES
advanced modulation, MIMO and equalization techniques have
[1] M. A. Khalighi and M. Uysal, “Survey on free space optical communi-
been studied, pushing the data rate up to ten of Mb/s [146]. In cation: A communication theory perspective,” IEEE Commun. Surveys
addition, whilst the small modulation bandwidth is a limitation Tuts., vol. 16, no. 4, pp. 2231–2258, Oct.–Dec. 2014.
for high-speed applications, it is not a major issue in many IoT [2] T. Koonen, “Indoor optical wireless systems: Technology, trends, and
applications,” J. Lightw. Technol., vol. 36, no. 8, pp. 1459–1467,
scenarios for sensor connections, where printed OLEDs have Apr. 2018.
been studied [147]. [3] S. Buzzi, C. D’Andrea, D. Li, and S. Feng, “MIMO-UFMC transceiver
schemes for millimeter-wave wireless communications,” IEEE Trans.
Commun., vol. 67, no. 5, pp. 3323–3336, 2019.
IV. CONCLUSION [4] X. Wang et al., “Millimeter wave communication: A comprehensive
survey,” IEEE Commun. Surveys Tuts., vol. 20, no. 3, pp. 1616–1653,
In this paper we have provided an overview of short-range Jul.–Sep. 2018.
[5] H.-J. Song, “Terahertz wireless communications: Recent developments
OWC principles and systems and the recent evolution of this including a prototype system for short-range data downloading,” IEEE
rapidly developing research field. The short-range OWC relies Microw. Mag., vol. 22, no. 5, pp. 88–99, May 2021.
on the wireless transmission of data-carrying optical wave, and [6] Y. Hong et al., “Beyond terabit/s WDM optical wireless transmission
using wavelength-transparent beam tracking and steering,” in Proc. Opt.
it has shown outstanding capabilities in realizing high data rate Fiber Commun. Conf., 2020, Paper W1G.4.
with simple modulation formats and low-cost direct detection [7] A. S. Hamza, J. S. Deogun, and D. R. Alexander, “Classification frame-
receivers, the immunity against electromagnetic interference, work for free space optical communication links and systems,” IEEE
Commun. Surveys Tuts., vol. 21, no. 2, pp. 1346–1382, Apr.–Jun. 2019.
and the enhanced physical layer security. Due to these advan- [8] A. Nirmalathas et al., “Indoor optical wireless access networks–recent
tages, the short-range OWC technology has been investigated progress,” J. Opt. Commun. Netw., vol. 13, no. 2, pp. A178–A186,
widely and here we have introduced the OWC-based in-building 2021.
[9] D. R. Dhatchayeny, S. Arya, and Y. H. Chung, “Infrared-based multiple-
and indoor personal and local area communications, underwater patient monitoring in indoor optical wireless healthcare systems,” IEEE
wireless links, flexible interconnects in data-centers, as well as Sensors J., vol. 19, no. 14, pp. 5594–5599, Jul. 2019.
V2V and V2I connections. We have focused on the basic work- [10] X. Zhao, H. Chen, and J. Sun, “On physical-layer security in multiuser
visible light communication systems with non-orthogonal multiple ac-
ing principles, system architectures, key function blocks, major cess,” IEEE Access, vol. 6, pp. 34004–34017, 2018.
limitations and potential solutions in this tutorial, to provide a [11] L. Bariah et al., “A prospective look: Key enabling technologies, appli-
basis for further studies. In addition, we have also introduced two cations and open research topics in 6G networks,” IEEE Access, vol. 8,
pp. 174792–174820, 2020.
major recent advanced techniques that have attracted intensive [12] S. Dang, O. Amin, B. Shihada, and M.-S. Alouini, “What should 6G be,”
interests to push the limit of short-range OWCs, including the Nature Electron., vol. 3, no. 1, pp. 20–29, 2020.
spatial domain diversity, multiplexing and spatial modulation [13] A. Biswas, “NASA’s deep space optical communications–an update,” in
Proc. Appl. Lasers Sens. Free Space Commun., 2019, Paper LTh1B.1.
schemes, and ML techniques. With these new techniques, the [14] H. Safi, A. Dargahi, J. Cheng, and M. Safari, “Analytical channel
performance and capability of short-range OWCs have been model and link design optimization for ground-to-hap free-space optical
further enhanced, which have been widely considered as a communications,” J. Lightw. Technol., vol. 38, no. 18, pp. 5036–5047,
2020.
new technological candidate for beyond-5G communications. [15] A. U. Chaudhry and H. Yanikomeroglu, “Free space optics for next-
Further considering the enhanced physical security, the short- generation satellite networks,” IEEE Consum. Electron. Mag., vol. 10,
range OWC technology is also highly attractive in Industry 4.0, no. 6, pp. 21–31, Nov. 2021.
[16] M. M. Tawfik, M. F.A. Sree, M. Abaza, and H. H. M. Ghouz, “Perfor- [40] Y. Tanaka, T. Komine, S. Haruyama, and M. Nakagawa, “Indoor visible
mance analysis and evaluation of inter-satellite optical wireless commu- light data transmission system utilizing white LED lights,” IEICE Trans.
nication system (IsOWC) from GEO to LEO at range 45000 km,” IEEE Commun., vol. 86, no. 8, pp. 2440–2454, 2003.
Photon. J., vol. 13, no. 4, pp. 1–6, Aug. 2021. [41] R. Bian, I. Tavakkolnia, and H. Haas, “15.73 Gb/s visible light commu-
[17] A. U. Chaudhry and H. Yanikomeroglu, “Laser intersatellite links in a nication with off-the-shelf LEDs,” J. Lightw. Technol., vol. 37, no. 10,
starlink constellation: A classification and analysis,” IEEE Veh. Technol. pp. 2418–2424, May 2019.
Mag., vol. 16, no. 2, pp. 48–56, Jun. 2021. [42] G. Cossu, W. Ali, R. Corsini, and E. Ciaramella, “Gigabit-class optical
[18] M. T. Dabiri et al., “UAV-assisted free space optical communication wireless communication system at indoor distances (1.5–4m),” Opt. Exp.,
system with amplify-and-forward relaying,” IEEE Trans. Veh. Technol., vol. 23, no. 12, pp. 15700–15705, 2015.
vol. 70, no. 9, pp. 8926–8936, Sep. 2021. [43] K. Wang, A. Nirmalathas, C. Lim, E. Skafidas, and K. Alameh, “Experi-
[19] A. Elzanaty and M.-S. Alouini, “Adaptive coded modulation for IM/DD mental demonstration of free-space based 120 Gb/s reconfigurable card-
free-space optical backhauling: A probabilistic shaping approach,” IEEE to-card optical interconnects,” Opt. Lett., vol. 39, no. 19, pp. 5717–5720,
Trans. Commun., vol. 68, no. 10, pp. 6388–6402, Oct. 2020. 2014.
[20] T. A. Nirmalathas et al., “Gigabit/s optical wireless access and indoor [44] V. Jungnickel, V. Pohl, S. Nonnig, and C. von Helmolt, “A physical
networks,” in Proc. Opt. Fiber Commun. Conf., 2020, Paper M2H.5. model of the wireless infrared communication channel,” IEEE J. Sel.
[21] F. Feng, P. Sangwongngam, G. Faulkner, and D. O’Brien, “Wide field- Areas Commun., vol. 20, no. 3, pp. 631–640, Apr. 2002.
of-view optical broadcasting for bi-directional indoor optical wireless [45] K. Wang, A. Nirmalathas, C. Lim, and E. Skafidas, “Impact of back-
communications employing PAM-4 modulation,” Opt. Lett., vol. 44, ground light induced shot noise in high-speed full-duplex indoor op-
no. 24, pp. 6009–6012, 2019. tical wireless communication systems,” Opt. Exp., vol. 19, no. 22,
[22] S. Zhang et al., “Feasibility study of optical wireless technology in pp. 21321–21332, 2011.
data center network,” IEEE Photon. Technol. Lett., vol. 33, no. 15, [46] Y. Wang, L. Tao, Y. Wang, and N. Chi, “High speed WDM VLC system
pp. 773–776, Aug. 2021. based on multi-band CAP64 with weighted pre-equalization and modified
[23] K. Wang, C. Lim, E. Wong, K. Alameh, S. Kandeepan, and E. Skafi- CMMA based post-equalization,” IEEE Commun. Lett., vol. 18, no. 10,
das, “High-speed reconfigurable free-space optical interconnects with pp. 1719–1722, Oct. 2014.
carrierless-amplitude-phase modulation and space-time-block code,” J. [47] G. Stepniak, M. Schüppert, and C.-A. Bunge, “Advanced modulation
Lightw. Technol., vol. 37, no. 2, pp. 627–633, Jan. 2019. formats in phosphorous LED VLC links and the impact of blue filtering,”
[24] W. Ali et al., “10 Gbit/s OWC system for intra-data centers links,” IEEE J. Lightw. Technol., vol. 33, no. 21, pp. 4413–4423, Nov. 2015.
Photon. Technol. Lett., vol. 31, no. 11, pp. 805–808, Jun. 2019. [48] E. Xie et al., “Over 10 Gbps VLC for long-distance applications using
[25] B. Béchadergue, C. Dominguez, A. Pesala, P. Chandra, G. Allegretto, and a GaN-based series-biased micro-LED array,” IEEE Photon. Technol.
S. Richer, “Vehicle-to-vehicle optical wireless communication with the Lett., vol. 32, no. 9, pp. 499–502, May 2020.
smart corner automotive headlamp,” in Proc. Global LIFI Congr., 2019, [49] L. Wang et al., “Green InGaN quantum dots breaking through efficiency
pp. 1–5. and bandwidth bottlenecks of micro-LEDs,” Laser Photon. Rev., vol. 15,
[26] Y. Kaymak, S. Fathi-Kazerooni, and R. Rojas-Cessa, “Indirect diffused no. 5, 2021, Art. no. 2000406.
light free-space optical communications for vehicular networks,” IEEE [50] G. Marsh and J. Kahn, “50-Mb/s diffuse infrared free-space link using on-
Commun. Lett., vol. 23, no. 5, pp. 814–817, May 2019. off keying with decision-feedback equalization,” IEEE Photon. Technol.
[27] M. K. Hasan, M. O. Ali, M. H. Rahman, M. Z. Chowdhury, and Y. Lett., vol. 6, no. 10, pp. 1268–1270, Oct. 1994.
M. Jang, “Optical camera communication in vehicular applications: A [51] M. T. Alresheedi and J. M. Elmirghani, “Performance evaluation of 5
review,” IEEE Trans. Intell. Transp. Syst., vol. 23, no. 7, pp. 6260–6281, Gbit/s and 10 Gbit/s mobile optical wireless systems employing beam
Jul. 2022. angle and power adaptation with diversity receivers,” IEEE J. Sel. Areas
[28] Z. Zeng, S. Fu, H. Zhang, Y. Dong, and J. Cheng, “A survey of underwater Commun., vol. 29, no. 6, pp. 1328–1340, Jun. 2011.
optical wireless communications,” IEEE Commun. Surveys Tuts., vol. 19, [52] X. Zhang et al., “40 Gb/s indoor optical wireless system enabled by a
no. 1, pp. 204–238, Jan.–Mar. 2017. cyclically arranged optical beamsteering receiver,” Opt. Lett., vol. 43,
[29] S. Teli, W. A. Cahyadi, and Y. H. Chung, “Optical camera commu- no. 4, pp. 723–726, 2018.
nication: Motion over camera,” IEEE Commun. Mag., vol. 55, no. 8, [53] K. Wang et al., “Experimental demonstration of indoor infrared optical
pp. 156–162, Aug. 2017. wireless communications with a silicon photonic integrated circuit,” J.
[30] P. H. Pathak, X. Feng, P. Hu, and P. Mohapatra, “Visible light com- Lightw. Technol., vol. 37, no. 2, pp. 619–626, Jan. 2019.
munication, networking, and sensing: A survey, potential and chal- [54] T. Koonen, J. Oh, K. Mekonnen, Z. Cao, and E. Tangdiongga, “Ultra-high
lenges,” IEEE Commun. Surveys Tuts., vol. 17, no. 4, pp. 2047–2077, capacity indoor optical wireless communication using 2d-steered pencil
Oct.–Dec. 2015. beams,” J. Lightw. Technol., vol. 34, no. 20, pp. 4802–4809, Oct. 2016.
[31] L. E. M. Matheus, A. B. Vieira, L. F. M. Vieira, M. A. M. Vieira, and [55] T. Koonen, F. Gomez-Agis, F. Huijskens, K. A. Mekonnen, Z. Cao, and
O. Gnawali, “Visible light communication: Concepts, applications and E. Tangdiongga, “High-capacity optical wireless communication using
challenges,” IEEE Commun. Surveys Tuts., vol. 21, no. 4, pp. 3204–3237, two-dimensional IR beam steering,” J. Lightw. Technol., vol. 36, no. 19,
Oct.–Dec. 2019. pp. 4486–4493, Oct. 2018.
[32] N. Chi, Y. Zhou, Y. Wei, and F. Hu, “Visible light communication in 6G: [56] K. Wang, “Quasi-passive indoor optical wireless communication sys-
Advances, challenges, and prospects,” IEEE Veh. Technol. Mag., vol. 15, tems,” IEEE Photon. Technol. Lett., vol. 32, no. 21, pp. 1373–1376,
no. 4, pp. 93–102, Dec. 2020. Nov. 2020.
[33] A.-M. Căilean and M. Dimian, “Current challenges for visible light com- [57] K. Wang, “Remote-powered infrared indoor optical wireless communi-
munications usage in vehicle applications: A survey,” IEEE Commun. cation systems,” IEEE Photon. Technol. Lett., vol. 34, no. 9, pp. 455–458,
Surveys Tuts., vol. 19, no. 4, pp. 2681–2703, Oct.–Dec. 2017. May 2022.
[34] F. Gfeller and U. Bapst, “Wireless in-house data communication via [58] H. Haas, “LiFi is a paradigm-shifting 5G technology,” Rev. Phys., vol. 3,
diffuse infrared radiation,” Proc. IEEE Proc. IRE, vol. 67, no. 11, pp. 26–31, 2018. [Online]. Available: https://www.sciencedirect.com/
pp. 1474–1486, 1979. science/article/pii/S2405428317300151
[35] T.-S. Chu and M. Gans, “High speed infrared local wireless communi- [59] H. Haas, L. Yin, Y. Wang, and C. Chen, “What is LiFi,” J. Lightw.
cation,” IEEE Commun. Mag., vol. 25, no. 8, pp. 4–10, Aug. 1987. Technol., vol. 34, no. 6, pp. 1533–1544, Mar. 2016.
[36] J. M. Kahn and J. R. Barry, “Wireless infrared communications,” Proc. [60] C. Chen, D. A. Basnayaka, A. A. Purwita, X. Wu, and H. Haas, “Wireless
IEEE, vol. 85, no. 2, pp. 265–298, 1997. infrared-based LiFi uplink transmission with link blockage and random
[37] V. Pohl, V. Jungnickel, and C. von Helmolt, “A channel model for wireless device orientation,” IEEE Trans. Commun., vol. 69, no. 2, pp. 1175–1188,
infrared communication,” in Proc. IEEE 11th Int. Symp. Pers. Indoor Feb. 2021.
Mobile Radio Commun., 2000, vol. 1, pp. 297–303. [61] Y. Wang, D. A. Basnayaka, X. Wu, and H. Haas, “Optimization of load
[38] J. Kahn, R. You, P. Djahani, A. Weisbin, B. K. Teik, and A. Tang, “Imaging balancing in hybrid LiFi/RF networks,” IEEE Trans. Commun., vol. 65,
diversity receivers for high-speed infrared wireless communication,” no. 4, pp. 1708–1720, Apr. 2017.
IEEE Commun. Mag., vol. 36, no. 12, pp. 88–94, Dec. 1998. [62] X. Wu, D. C. O’Brien, X. Deng, and J.-P. M. G. Linnartz, “Smart handover
[39] A. Al-Ghamdi and J. Elmirghani, “Line strip spot-diffusing transmitter for hybrid LiFi and WiFi networks,” IEEE Trans. Wireless Commun.,
configuration for optical wireless systems influenced by background vol. 19, no. 12, pp. 8211–8219, Dec. 2020.
noise and multipath dispersion,” IEEE Trans. Commun., vol. 52, no. 1, [63] V. Jungnickel et al., “LiFi for industrial wireless applications,” in Proc.
pp. 37–45, Jan. 2004. Opt. Fiber Commun. Conf. Exhib., 2020, pp. 1–3.
[64] N. Lorrière et al., “Photovoltaic solar cells for outdoor LiFi communica- [88] M. D. Thieu, T. L. Pham, T. Nguyen, and Y. M. Jang, “Optical-
tions,” J. Lightw. Technol., vol. 38, no. 15, pp. 3822–3831, Aug. 2020. RoI-signaling for vehicular communications,” IEEE Access, vol. 7,
[65] G. N. Arvanitakis et al., “Gb/s underwater wireless optical communica- pp. 69873–69891, 2019.
tions using series-connected GaN micro-LED arrays,” IEEE Photon. J., [89] T.-H. Do and M. Yoo, “A multi-feature LED bit detection algorithm
vol. 12, no. 2, pp. 1–10, 2020. in vehicular optical camera communication,” IEEE Access, vol. 7,
[66] X.-H. Huang et al., “Bidirectional white-lighting WDM VLC–UWOC pp. 95797–95811, 2019.
converged systems,” J. Lightw. Technol., vol. 39, no. 13, pp. 4351–4359, [90] X. Sun, W. Shi, Q. Cheng, W. Liu, Z. Wang, and J. Zhang, “An LED
Jul. 2021. detection and recognition method based on deep learning in vehicle
[67] B. Majlesein, A. Gholami, and Z. Ghassemlooy, “A complete model for optical camera communication,” IEEE Access, vol. 9, pp. 80897–80905,
underwater optical wireless communications system,” in Proc. 11th Int. 2021.
Symp. Commun. Syst. Netw. Digit. Signal Process., 2018, pp. 1–5. [91] A.-M. Căilean, B. Cagneau, L. Chassagne, M. Dimian, and V. Popa,
[68] C. Fu et al., “Anti error and erasure coding for water-to-air visible light “Novel receiver sensor for visible light communications in automo-
communication through wavy water surface with wave height up to 0.6 tive applications,” IEEE Sensors J., vol. 15, no. 8, pp. 4632–4639,
meters,” Opt. Exp., vol. 30, no. 11, pp. 18743–18761, 2022. Aug. 2015.
[69] Y. Di, Y. Shao, and L.-K. Chen, “Mitigation of wave-induced packet loss [92] S. Halawi, E. Yaacoub, S. Kassir, and Z. Dawy, “Performance analy-
for water-air optical wireless communication by a tracking system,” in sis of circular color shift keying in VLC systems with camera-based
Proc. Opt. Fiber Commun. Conf. Exhibit., 2021, pp. 1–3. receivers,” IEEE Trans. Commun., vol. 67, no. 6, pp. 4252–4266, Jun.
[70] J. Wang, C. Lu, S. Li, and Z. Xu, “100m/500 Mbps underwater optical 2019.
wireless communication using an NRZ-OOK modulated 520 nm laser [93] B. Turan, O. Narmanlioglu, O. N. Koc, E. Kar, S. Coleri, and M. Uysal,
diode,” Opt. Exp., vol. 27, no. 9, pp. 12171–12181, 2019. “Measurement based non-line-of-sight vehicular visible light commu-
[71] J. Du et al., “Experimental demonstration of 50-m/5-Gbps underwater op- nication channel characterization,” IEEE Trans. Veh. Technol., vol. 71,
tical wireless communication with low-complexity chaotic encryption,” no. 9, pp. 10110–10114, Sep. 2022.
Opt. Exp., vol. 29, no. 2, pp. 783–796, 2021. [94] O. Narmanlioglu, B. Turan, S. C. Ergen, and M. Uysal, “Cooper-
[72] J. Ning, G. Gao, J. Zhang, H. Peng, and Y. Guo, “Adaptive receiver control ative MIMO-OFDM based inter-vehicular visible light communica-
for reliable high-speed underwater wireless optical communication with tion using brake lights,” Comput. Commun., vol. 120, pp. 138–146,
photomultiplier tube receiver,” IEEE Photon. J., vol. 13, no. 4, pp. 1–7, 2018.
Aug. 2021. [95] E. Eso, Z. Ghassemlooy, S. Zvanovec, P. Pesek, and J. Sathian, “Vehicle-
[73] M. V. Jamali, A. Chizari, and J. A. Salehi, “Performance analysis of to-vehicle relay-assisted VLC with misalignment induced azimuth or
multi-hop underwater wireless optical communication systems,” IEEE elevation offset angles,” IEEE Photon. Technol. Lett., vol. 33, no. 16,
Photon. Technol. Lett., vol. 29, no. 5, pp. 462–465, Mar. 2017. pp. 908–911, Aug. 2021.
[74] X. Li, J. Liu, L. Yan, S. Han, and X. Guan, “Relay selection in underwater [96] K. Wang, T. Song, S. Kandeepan, H. Li, and K. Alameh, “Indoor opti-
acoustic cooperative networks: A contextual bandit approach,” IEEE cal wireless communication system with continuous and simultaneous
Commun. Lett., vol. 21, no. 2, pp. 382–385, Feb. 2017. positioning,” Opt. Exp., vol. 29, no. 3, pp. 4582–4595, 2021.
[75] F. Xing, H. Yin, X. Ji, and V. C. M. Leung, “Joint relay selection [97] T. Q. Wang, Y. A. Sekercioglu, A. Neild, and J. Armstrong, “Position
and power allocation for underwater cooperative optical wireless net- accuracy of time-of-arrival based ranging using visible light with appli-
works,” IEEE Trans. Wireless Commun., vol. 19, no. 1, pp. 251–264, cation in indoor localization systems,” J. Lightw. Technol., vol. 31, no. 20,
Jan. 2020. pp. 3302–3308, Oct. 2013.
[76] Q. Hu, M. Chagnon, K. Schuh, F. Buchali, and H. Bülow, “IM/DD beyond [98] M. H. Bergen, X. Jin, D. Guerrero, H. A. L. F. Chaves, N. V. Fredeen,
bandwidth limitation for data center optical interconnects,” J. Lightw. and J. F. Holzman, “Design and implementation of an optical receiver for
Technol., vol. 37, no. 19, pp. 4940–4946, Oct. 2019. angle-of-arrival-based positioning,” J. Lightw. Technol., vol. 35, no. 18,
[77] Y. Shi, L. Ma, Y. Zhuang, and Z. He, “Investigation on roughness-induced pp. 3877–3885, Sep. 2017.
scattering loss of small-core polymer waveguides for single-mode optical [99] B. Lin, Z. Ghassemlooy, C. Lin, X. Tang, Y. Li, and S. Zhang, “An
interconnect applications,” Opt. Exp., vol. 28, no. 26, pp. 38733–38744, indoor visible light positioning system based on optical camera com-
2020. munications,” IEEE Photon. Technol. Lett., vol. 29, no. 7, pp. 579–582,
[78] Y. Zhang et al., “Foundry-enabled scalable all-to-all optical interconnects Apr. 2017.
using silicon nitride arrayed waveguide router interposers and silicon [100] J.-Y. Sung, “Image frame rate limited sub-centimeter localization system
photonic transceivers,” IEEE J. Sel. Topics Quantum Electron., vol. 25, for hybrid radio-optical wireless communications,” J. Lightw. Technol.,
no. 5, pp. 1–9, Sep.–Oct. 2019. vol. 40, no. 13, pp. 4150–4160, Jul. 2022.
[79] Y. Cui, H. Wang, X. Cheng, and B. Chen, “Wireless data center network- [101] S. Shen, S. Li, and H. Steendam, “Hybrid position and orientation
ing,” IEEE Wireless Commun., vol. 18, no. 6, pp. 46–53, Dec. 2011. estimation for visible light systems in the presence of prior information
[80] S. A. Mamun, S. G. Umamaheswaran, A. Ganguly, M. Kwon, and A. on the orientation,” IEEE Trans. Wireless Commun., vol. 21, no. 8,
Kwasinski, “Performance evaluation of a power-efficient and robust 60 pp. 6271–6284, Aug. 2022.
Ghz wireless server-to-server datacenter network,” IEEE Trans. Green [102] T. Song et al., “Performance analysis of repetition-coding and space-time-
Commun. Netw., vol. 2, no. 4, pp. 1174–1185, Dec. 2018. block-coding as transmitter diversity schemes for indoor optical wireless
[81] C.-L. Cheng, S. Sangodoyin, and A. Zajić, “THz MIMO channel char- communications,” J. Lightw. Technol., vol. 37, no. 20, pp. 5170–5177,
acterization for wireless data center-like environment,” in Proc. IEEE Oct. 2019.
Int. Symp. Antennas Propag. USNC-URSI Radio Sci. Meeting, 2019, [103] L. Liu, R. Deng, J. Shi, J. He, and L.-K. Chen, “Beyond 100-kbit/s
pp. 2145–2146. transmission over rolling shutter camera-based VLC enabled by color
[82] X. Zhao, L. Zhu, S. Fu, C. Liu, M. Tang, and D. Liu, “Dual-band accelerat- and spatial multiplexing,” in Proc. Opt. Fiber Commun. Conf. Exhibit.,
ing beams enabled full duplex free-space optical interconnection,” IEEE 2020, pp. 1–3.
J. Sel. Topics Quantum Electron., vol. 27, no. 1, pp. 1–7, Jan./Feb. 2021. [104] C. R. Kumar and R. Jeyachitra, “Power efficient generalized spatial mod-
[83] M. Aljada, K. E. Alameh, Y.-T. Lee, and I.-S. Chung, “High-speed (2.5 ulation MIMO for indoor visible light communications,” IEEE Photon.
Gbps) reconfigurable inter-chip optical interconnects using opto-VLSI Technol. Lett., vol. 29, no. 11, pp. 921–924, Jun. 2017.
processors,” Opt. Exp., vol. 14, no. 15, pp. 6823–6836, 2006. [105] M. Li, X. Li, C. Yang, Q. Yang, and S. Yu, “Experimental demonstration
[84] Q. You, D. Chen, X. Xiao, and S. Yu, “10 Gb/s free space optical intercon- of indoor interfered visible light communication system employing suc-
nect with broadcasting capability enabled by a silicon integrated optical cessive interference cancellation,” Chin. Opt. Lett., vol. 15, no. 5, 2017,
phased array,” Chin. Opt. Lett., vol. 19, no. 12, 2021, Art. no. 120602. Art. no. 0 50604.
[85] S. Zeadally, M. A. Javed, and E. B. Hamida, “Vehicular communications [106] Y. Qiu, S. Chen, H.-H. Chen, and W. Meng, “Visible light communica-
for its: Standardization and challenges,” IEEE Commun. Standards Mag., tions based on CDMA technology,” IEEE Wireless Commun., vol. 25,
vol. 4, no. 1, pp. 11–17, Mar. 2020. no. 2, pp. 178–185, Apr. 2018.
[86] S. A. A. Shah, E. Ahmed, M. Imran, and S. Zeadally, “5G for vehicular [107] T. Song et al., “Optical wireless communications adopting delay-tolerant
communications,” IEEE Commun. Mag., vol. 56, no. 1, pp. 111–117, repetition-coding with orthogonal-filters and on-demand equalization,”
Jan. 2018. J. Lightw. Technol., vol. 38, no. 16, pp. 4250–4259, Aug. 2020.
[87] Y. Liu, Z. Jiang, F. Wang, and N. Chi, “315 Mbps internet of vehicle [108] J. He, J. Lee, T. Song, H. Li, S. Kandeepan, and K. Wang, “Recurrent
communication system using car head lamp based on weighted pre- neural network (RNN) for delay-tolerant repetition-coded (RC) indoor
distortion,” in Proc. IEEE 18th Int. Conf. Commun. Technol., 2018, optical wireless communication systems,” Opt. Lett., vol. 44, no. 15,
pp. 509–512. pp. 3745–3748, 2019.
[109] J. He, J. Lee, T. Song, H. Li, S. Kandeepan, and K. Wang, “Delay-tolerant [131] H. Lee, I. Lee, T. Q. Quek, and S. H. Lee, “Binary signaling design for
indoor optical wireless communication systems based on attention- visible light communication: A deep learning framework,” Opt. Exp.,
augmented recurrent neural network,” J. Lightw. Technol., vol. 38, no. 17, vol. 26, no. 14, pp. 18131–18142, 2018.
pp. 4632–4640, Sep. 2020. [132] A. Shawahna, S. M. Sait, and A. El-Maleh, “FPGA-based accelerators of
[110] S. H. Younus, A. A. Al-Hameed, A. T. Hussein, M. T. Alresheedi, and deep learning networks for learning and classification: A review,” IEEE
J. M. Elmirghani, “Parallel data transmission in indoor visible light Access, vol. 7, pp. 7823–7859, 2019.
communication systems,” IEEE Access, vol. 7, pp. 1126–1138, 2018. [133] E. Nurvitadhi, J. Sim, D. Sheffield, A. Mishra, S. Krishnan, and D. Marr,
[111] T. Zinda, K. Ito, and W. Chujo, “Rolling-shutter-based optical camera “Accelerating recurrent neural networks in analytics servers: Compar-
communication using distributed LED array,” in Proc. 11th Int. Symp. ison of FPGA, CPU, GPU, and ASIC,” in Proc. 26th Int. Conf. Field
Commun. Syst. Netw. Dig. Signal Process., 2018, pp. 1–4. Programmable Log. Appl., 2016, pp. 1–4.
[112] T. Fath and H. Haas, “Performance comparison of MIMO techniques for [134] J. Lee, J. He, and K. Wang, “FPGA-based neural network accelerators
optical wireless communications in indoor environments,” IEEE Trans. for millimeter-wave radio-over-fiber systems,” Opt. Exp., vol. 28, no. 9,
Commun., vol. 61, no. 2, pp. 733–742, Feb. 2013. pp. 13384–13400, 2020.
[113] K.-H. Park, Y.-C. Ko, and M.-S. Alouini, “On the power and offset [135] B. Turan and S. Coleri, “Machine learning based channel modeling
allocation for rate adaptation of spatial multiplexing in optical wireless for vehicular visible light communication,” IEEE Trans. Veh. Technol.,
mimo channels,” IEEE Trans. Commun., vol. 61, no. 4, pp. 1535–1543, vol. 70, no. 10, pp. 9659–9672, Oct. 2021.
Apr. 2013. [136] A. H. A. Bakar, T. Glass, H. Y. Tee, F. Alam, and M. Legg, “Accu-
[114] J. Lian and M. Brandt-Pearce, “Multiuser mimo indoor visible light rate visible light positioning using multiple-photodiode receiver and
communication system using spatial multiplexing,” J. Lightw. Technol., machine learning,” IEEE Trans. Instrum. Meas., vol. 70, pp. 1–12,
vol. 35, no. 23, pp. 5024–5033, Dec. 2017. 2021.
[115] J.-Y. Wang, J.-B. Wang, Y. Wu, M. Lin, and M. Chen, “Constellation [137] L. Xiao, G. Sheng, S. Liu, H. Dai, M. Peng, and J. Song, “Deep rein-
optimization for spatial modulation based indoor optical wireless com- forcement learning-enabled secure visible light communication against
munications,” in Proc. IEEE GLOBECOM Glob. Commun. Conf., 2017, eavesdropping,” IEEE Trans. Commun., vol. 67, no. 10, pp. 6994–7005,
pp. 1–6. Oct. 2019.
[116] K. Wang, “Digital orthogonal-filters enhanced spatial modulation for [138] S. Naser et al., “Toward federated-learning-enabled visible light com-
high-speed indoor optical wireless communications,” J. Lightw. Technol., munication in 6G systems,” IEEE Wireless Commun., vol. 29, no. 1,
vol. 37, no. 24, pp. 5988–5995, Dec. 2019. pp. 48–56, Feb. 2022.
[117] K. Wang, “Indoor optical wireless communication system with filters- [139] Z. Cao, X. Zhang, G. Osnabrugge, J. Li, I. M. Vellekoop, and A.
enhanced generalized spatial modulation and carrierless amplitude and M. Koonen, “Reconfigurable beam system for non-line-of-sight free-
phase (CAP) modulation,” Opt. Lett., vol. 45, no. 18, pp. 4980–4983, space optical communication,” Light: Sci. Appl., vol. 8, no. 1, pp. 1–9,
Sep. 2020. 2019.
[118] K. L. Bober et al., “Distributed multiuser MIMO for LiFi in in- [140] S. Aboagye, T. M. Ngatched, O. A. Dobre, and A. R. Ndjiongue, “Intel-
dustrial wireless applications,” J. Lightw. Technol., vol. 39, no. 11, ligent reflecting surface-aided indoor visible light communication sys-
pp. 3420–3433, 2021. tems,” IEEE Commun. Lett., vol. 25, no. 12, pp. 3913–3917, Dec. 2021.
[119] P. W. Berenguer et al., “Optical wireless MIMO experiments in an [141] A. B. Ozyurt and W. O. Popoola, “LiFi-based D2D communica-
industrial environment,” IEEE J. Sel. Areas Commun., vol. 36, no. 1, tion in industrial IoT,” IEEE Syst. J., early access, doi: 10.1109/
pp. 185–193, Jan. 2018. JSYST.2022.3179183.
[120] C. Sun, X. Gao, J. Wang, Z. Ding, and X.-G. Xia, “Beam domain massive [142] Z. Wei et al., “Optical uplink, D2D and IoT links based on VCSEL
MIMO for optical wireless communications with transmit lens,” IEEE array: Analysis and demonstration,” J. Lightw. Technol., vol. 40, no. 15,
Trans. Commun., vol. 67, no. 3, pp. 2188–2202, Mar. 2019. pp. 5083–5096, Aug. 2022.
[121] Y. LeCun, Y. Bengio, and G. Hinton, “Deep Learning,” Nature, vol. 521, [143] S. Huang, G. Chuai, and W. Gao, “Coalitional games based resource
no. 7553, pp. 436–444, 2015. allocation for D2D uplink underlaying hybrid VLC-RF networks,” in
[122] F. Musumeci et al., “An overview on application of machine learning Proc. IEEE Wireless Commun. Netw. Conf., 2022, pp. 2316–2321.
techniques in optical networks,” IEEE Commun. Surveys Tuts., vol. 21, [144] K. Werfli et al., “Experimental demonstration of high-speed 4× 4 imaging
no. 2, pp. 1383–1408, Apr.–Jun. 2019. multi-CAP MIMO visible light communications,” J. Lightw. Technol.,
[123] J. He, J. Lee, S. Kandeepan, and K. Wang, “Machine learning techniques vol. 36, no. 10, pp. 1944–1951, 2018.
in radio-over-fiber systems and networks,” Photonics, vol. 7, no. 4, 2020, [145] J. Chen et al., “Neural network detection for bandwidth-limited non-
Art. no. 105. orthogonal multiband CAP UVLC system,” IEEE Photon. J., vol. 14,
[124] F. N. Khan, Q. Fan, C. Lu, and A. P. T. Lau, “An optical communication’s no. 2, pp. 1–9, Apr. 2022.
perspective on machine learning and its applications,” J. Lightw. Technol., [146] H. Chen, Z. Xu, Q. Gao, and S. Li, “A 51.6 Mb/s experimental VLC
vol. 37, no. 2, pp. 493–516, Jan. 2019. system using a monochromic organic LED,” IEEE Photon. J., vol. 10,
[125] N. Chi, Y. Zhao, M. Shi, P. Zou, and X. Lu, “Gaussian kernel-aided no. 2, pp. 1–12, Apr. 2018.
deep neural network equalizer utilized in underwater PAM8 visible light [147] A. Perera and M. Katz, “Feasibility study on the use of printed OLEDs
communication system,” Opt. Exp., vol. 26, no. 20, pp. 26700–26712, for wireless data and power transmission in light-based Internet of Things
2018. (LIoT),” in Proc. IEEE 16th Int. Conf. Ind. Inf. Syst., 2021, pp. 58–61.
[126] R. Jiang, C. Sun, L. Zhang, X. Tang, H. Wang, and A. Zhang, “Deep [148] S. Rajagopal, R. D. Roberts, and S.-K. Lim, “IEEE 802.15.7 visible
learning aided signal detection for spad-based underwater optical wire- light communication: Modulation schemes and dimming support,” IEEE
less communications,” IEEE Access, vol. 8, pp. 20363–20374, 2020. Commun. Mag., vol. 50, no. 3, pp. 72–82, Mar. 2012.
[127] S. Ma et al., “Signal demodulation with machine learning methods for [149] T. Nguyen, A. Islam, T. Yamazato, and Y. M. Jang, “Technical issues on
physical layer visible light communications: Prototype platform, open IEEE 802.15.7m image sensor communication standardization,” IEEE
dataset, and algorithms,” IEEE Access, vol. 7, pp. 30588–30598, 2019. Commun. Mag., vol. 56, no. 2, pp. 213–218, Feb. 2018.
[128] H. Lee, I. Lee, and S. H. Lee, “Deep learning based transceiver design for [150] M. Uysal, F. Miramirkhani, T. Baykas, and K. Qaraqe, “IEEE 802.11bb
multi-colored VLC systems,” Opt. Exp., vol. 26, no. 5, pp. 6222–6238, reference channel models for indoor environments,” Tech. Rep. 802.11-
2018. 18/1582r0, 2018.
[129] P. Miao, B. Zhu, C. Qi, Y. Jin, and C. Lin, “A model-driven
deep learning method for LED nonlinearity mitigation in OFDM-
based optical communications,” IEEE Access, vol. 7, pp. 71436–71446,
2019.
[130] J. Lee, T. Song, J. He, S. Kandeepan, and K. Wang, “Recurrent neural
network FPGA hardware accelerator for delay-tolerant indoor optical
wireless communications,” Opt. Exp., vol. 29, no. 16, pp. 26165–26182,
2021.

Evolution of Short-Range Optical Wireless Communications

Uploaded by

Copyright:

Available Formats

Evolution of Short-Range Optical Wireless Communications

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Evolution of Short-Range Optical Wireless Communications

Uploaded by

Copyright:

Available Formats

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 41, NO.

4, FEBRUARY 15, 2023 1019

Evolution of Short-Range Optical Wireless

Abstract—The optical wireless communication (OWC) technol- I. INTRODUCTION

Fig. 1. Application of optical wireless communications in beyond-5G.

VLC channel models, modulation techniques, medium access

To solve these limitations, advanced transmitter technologies

the free-space, the data-carrying light then propagates to the

vertical handover between the WiF signal with large coverage

with WDM, over 100 Gb/s (4 channels) full-duplex (O-band

D. OWC-Based V2V and V2I Communications

single transmitter system or the OWC system with spatial diver-

suppress them accordingly. Typically the physical layer pro-

Fig. 10. Architecture of two-connected neural networks for signal processing

The fully-connected neural network has also been studied to

square error (MSE) and the autocorrelation of learnt mapping

system or network conditions, and it needs to be re-trained

You might also like