hv photonics Review Machine Learning Techniques in Radio-over-Fiber Systems and Networks Jiayuan He 1,2, Jeonghun Lee 2, Sithamparanathan Kandeepan 2 and Ke Wang 2,* 1 School of Computing and Information Systems, The University of Melbourne, Melbourne, VIC 3010, Australia; [email protected] 2 School of Engineering, RMIT University, Melbourne, VIC 3000, Australia; [email protected] (J.L.); [email protected] (S.K.) * Correspondence: [email protected] Received: 30 September 2020; Accepted: 30 October 2020; Published: 7 November 2020 Abstract: The radio-over-fiber (RoF) technology has been widely studied during the past decades to extend the wireless communication coverage by leveraging the low-loss and broad bandwidth advantages of the optical fiber. With the increasing need for wireless communications, using millimeter-waves (mm-wave) in wireless communications has become the recent trend and many attempts have been made to build high-throughput and robust mm-wave RoF systems during the past a few years. Whilst the RoF technology provides many benefits, it suffers from several fundamental limitations due to the analog optical link, including the fiber chromatic dispersion and nonlinear impairments. Various approaches have been proposed to address these limitations. In particular, machine learning (ML) algorithms have attracted intensive research attention as a promising candidate for handling the complicated physical layer impairments in RoF systems, especially the nonlinearity during signal modulation, transmission and detection. In this paper, we review recent advancements in ML techniques for RoF systems, especially those which utilize ML models as physical layer signal processors to mitigate various types of impairments and to improve the system performance. In addition, ML algorithms have also been widely adopted for highly efficient RoF network management and resource allocation, such as the dynamic bandwidth allocation and network fault detection. In this paper, we also review the recent works in these research domains. Finally, several key open questions that need to be addressed in the future and possible solutions of these questions are also discussed. Keywords: radio-over-fiber; fiber-wireless; optical wireless integration; neural networks; artificial intelligence; machine learning 1. Introduction With the wide availability of high-performance and portable personal electronic devices, such as smart phones and tablets, our demand for ubiquitous wireless communications has grown explosively during the past decades [1,2]. In addition, the need of high-speed wireless communications has also increased substantially, driven by broadband and bandwidth-intensive applications, such as ultra-high-definition video-on-demand, virtual reality (VR) and augmented reality (AR). To meet these requirements, we have seen rapid development and deployment of wireless communication technologies, such as the widely adoption of small cell and massive multi-input multi-output (MIMO) [3]. In addition, due to the congestion of lower radio frequency (RF) band, higher RF spectral region has been explored, such as the millimeter-wave (mm-wave) region and the terahertz region [4,5]. Given that the terahertz communications are still in the early stage of investigation, the mm-wave frequency is widely studied for wireless communications. Compared with the lower RF band, much broader bandwidth is Photonics 2020, 7, 105; doi:10.3390/photonics7040105 www.mdpi.com/journal/photonics Photonics 2020, 7, 105 2 of 31 Photonics 2020, 7, x FOR PEER REVIEW 2 of 31 withavailable the lower in the RF mm-wave band, much region broader and hence, bandwidth higher speed is available exceeding in the tens mm-wave of gigabit-per-second region and hence, (Gbps) highercan be speed achieved. exceeding tens of gigabit-per-second (Gbps) can be achieved. However,However, the the use use of of high high RF RF (e.g., (e.g., mm-wave mm-wave frequency) frequency) also also induces induces new new challenges. challenges. Compared Compared withwith the the lower lower RF RF band, band, high high RF RFband band suffers suff ersfrom from much much higher higher free-space free-space propagation propagation loss. One loss. promisingOne promising way to way resolve to resolve this problem this problem is the is radi theo-over-fiber radio-over-fiber (RoF) (RoF) technology, technology, which which combines combines the bestthe bestof two of twoworlds: worlds: the theoptical optical fiber fiber and and the the wirele wirelessss communications communications [6,7]. [6,7 ].In In this this type type of of systems, systems, asas shown shown in in Figure Figure 11,, opticaloptical fibersfibers areare usedused toto distributedistribute wirelesswireless signalssignals andand hence,hence, thethe broadbroad bandwidthbandwidth and low-losslow-loss advantages advantages of of optical optical fibers fibers are are utilized. utilized. The RoFThe technologyRoF technology has been has widely been widelyinvestigated investigated for different for applicationdifferent application scenarios and scenarios it has been and combined it has been with othercombined technologies with other such technologiesas wavelength-division-multiplexing such as wavelength-division-multiplexing (WDM) networks and (WDM)/or conventional networks passive and/or optical conventional networks passive(PONs) [optical8,9]. In networks addition, driven(PONs) by [8,9]. the 5G In and addition, beyond-5G driven wireless by the communications, 5G and beyond-5G in the pastwireless a few communications,years, the RoF technology in the past has a alsofew beenyears, studied the RoF for technology mobile fronthaul has also to supportbeen studied higher for speed mobile and fronthaulsystem throughput to support [ 10higher]. speed and system throughput [10]. FigureFigure 1. 1. Radio-over-fiberRadio-over-fiber (RoF) (RoF) system system architecture, architecture, including including RoF-based RoF-based backhaul, backhaul, RoF-based RoF-based fronthaul,fronthaul, and and RoF-based RoF-based fiber-wireless fiber-wireless converged converged ac accesscess networks (e.g., passive optical networks). Whilst the RoF technology has been widely studied, it su ers from several fundamental Whilst the RoF technology has been widely studied, it suffersff from several fundamental limitations,limitations, which which are are mainly mainly caused caused by by the the fact fact th thatat the the RoF RoF system system essentia essentiallylly utilizes utilizes the the analog analog opticaloptical link. link. For For example, example, fiber fiber chromatic chromatic dispersi dispersion,on, distortion distortion and and nonlinear nonlinear effects effects all all limit limit the the performanceperformance of of RoF RoF systems. systems. Several Several impairment impairment mi mitigationtigation principles principles and and techniques techniques have have been been proposedproposed and and demonstrated demonstrated to to improve improve the the performanc performancee of of RoF RoF systems, systems, such such as as the the optical optical single single sidebandsideband (OSSB) (OSSB) and and optical optical carrier carrier suppression suppression (OCS) (OCS) schemes schemes from from the the modulation modulation perspective perspective [7]. [7]. TheThe digitized-radio-over-fiber digitized-radio-over-fiber (DRoF)(DRoF) solutionsolution thatthat changeschanges the the fiber fiber transmission transmission link link from from analog analog to todigital digital has has also also been been investigated investigated [7]. Although[7]. Althou thegh DRoFthe DRoF scheme scheme can substantially can substantially improve improve the system the systemperformance, performance, high-speed high-speed and broadband and broadband analog-to-digital analog-to-digital converter converter (ADC) (ADC) and digital-to-analog and digital-to- analogconverters converters (DAC) (DAC) are required. are required. InIn addition addition to to these these solutions, solutions, various various electrical electrical domain domain dispersion dispersion and and nonlinearity nonlinearity mitigation mitigation algorithmsalgorithms have have been been studied studied as as well from the signal processing perspectiveperspective [[11].11]. Whilst Whilst these these algorithmsalgorithms havehave beenbeen shown shown to beto effbeective, effective, they typically they typically are designed are designed to handle variousto handle impairments various impairmentsseparately. separately. However, itHowever, is crucial it is to crucial mitigate to mitigate different different types of types impairments of impairments jointly jointly other other than thanseparately. separately. This This is because is because different different types types of impairments of impairments in the in same the same channel channel may may interact interact with with each eachother. other. In addition, In addition, these conventionalthese conventional signal processingsignal processing algorithms algorithms typically typically have limited have capability limited capabilityin suppressing in suppressing nonlinear nonlinear effects, whilst effects, the whilst analog the RoF analog systems RoF normally systems sunormallyffer from suffer substantial from substantialnonlinearity, nonlinearity, including both including optical fiber both nonlinearity optical fiber and nonlinearity the nonlinear and distortions the nonlinear caused bydistortions the signal causedmodulation by the (e.g., signal inter-modulation modulation