See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/322217014

Artificial Intelligence (AI) Methods in Optical Networks: A Comprehensive

Survey

Article  in  Optical Switching and Networking · April 2018

DOI: 10.1016/j.osn.2017.12.006

CITATIONS READS

114 4,583

7 authors, including:

Javier Mata Ignacio de Miguel

Universidad de Valladolid Universidad de Valladolid
9 PUBLICATIONS   140 CITATIONS    128 PUBLICATIONS   995 CITATIONS   

SEE PROFILE SEE PROFILE

Ramón J. Durán Barroso N. Merayo

Universidad de Valladolid Universidad de Valladolid
105 PUBLICATIONS   848 CITATIONS    89 PUBLICATIONS   512 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IntelliMar View project

Application-Centric IP/optical Network Orchestration (ACINO) View project

All content following this page was uploaded by Sandeep Singh on 17 January 2018.

The user has requested enhancement of the downloaded file.

 Optical Switching and Networking 28 (2018) 43–57

Contents lists available at ScienceDirect

Optical Switching and Networking

journal homepage: www.elsevier.com/locate/osn

Artificial intelligence (AI) methods in optical networks: A comprehensive

survey
Javier Mata a,* , Ignacio de Miguel a , Ramón J. Durán a , Noemí Merayo a ,
Sandeep Kumar Singh b , Admela Jukan b , Mohit Chamania c
a
Universidad de Valladolid, ETSI de Telecomunicación, Campus Miguel Delibes, 47011, Valladolid, Spain
b Technische Universität Carolo-Wilhelmina zu Braunschweig, Germany
c
ADVA Optical Networking, Berlin, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Artificial intelligence (AI) is an extensive scientific discipline which enables computer systems to solve problems
Artificial intelligence by emulating complex biological processes such as learning, reasoning and self-correction. This paper presents a
Machine learning
comprehensive review of the application of AI techniques for improving performance of optical communication
Optical communications
systems and networks. The use of AI-based techniques is first studied in applications related to optical trans-
Optical networks
Optimization
mission, ranging from the characterization and operation of network components to performance monitoring,
Survey mitigation of nonlinearities, and quality of transmission estimation. Then, applications related to optical network
control and management are also reviewed, including topics like optical network planning and operation in both
transport and access networks. Finally, the paper also presents a summary of opportunities and challenges in
optical networking where AI is expected to play a key role in the near future.

1. Introduction affecting areas of transmission, switching and network management.

Optical communication networks and systems have not stayed on
Artificial intelligence (AI) entities and systems have the ability the sidelines, but have started to adopt this discipline towards AI-
to perform operations analogous to learning and decision making by based optical networking, from photonic devices to control and
imitating biological processes, with special emphasis on human cog- management.
nitive processes. AI applications such as virtual personal-assistants, The aim of this paper is to review some of the currently considered
smart vehicles, purchase prediction, speech recognition or smart home approaches to increase the performance of optical networks by the use
devices, are almost ubiquitous, and similar AI-based techniques are of AI mechanisms, providing a survey of the current research within
already changing our daily lives in ways that improve human produc- this area, as well as an overview of opportunities and challenges arising
tivity, safety or health, affecting even the way we entertain or commu- in this context.
nicate. The remainder of this paper is organized as follows. Section 2 pro-
For the most part, AI does not deliver completely autonomous sys- vides an introduction to the field of AI. Since that is a very broad
tems, but instead adds knowledge and reasoning to existing applica- area, we review those AI subfields –and their associated techniques–
tions, databases, and environments, to make them friendlier, smarter, which have had or are expected to have, in our opinion, a signifi-
and more sensitive to changes in their environments. Each small break- cant role in optical networking. Then, Sections 3 and 4 analyze the
through on AI research enables us to expand our skills to solve new role of AI techniques in optical communication systems and networks.
classes and scales of problems, thereby driving research and innovation We first survey the use of AI in optical transmission (Section 3), and
in almost every scientific discipline. then we focus on networking issues (Section 4). Finally, in Section 5
As an example, the improvement of the performance of telecom- we describe further opportunities and challenges, and we conclude in
munication networks by the application of AI-based techniques has Section 6.
become an area under extensive research over the past decades,

* Corresponding author.
E-mail addresses: [email protected] (J. Mata), [email protected] (I. de Migue).

https://doi.org/10.1016/j.osn.2017.12.006
Received 4 December 2017; Accepted 27 December 2017
Available online 2 January 2018
1573-4277/© 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

Fig. 1. AI subfields and techniques applied to optical networks.

2. An overview of AI and related techniques A stride forward towards making agents more intelligent comes by
incorporating the use of knowledge, reasoning and planning. In this case,
AI focuses on the study of intelligent or rational agents, i.e., entities intelligent agents keep a knowledge base (KB) where relevant knowl-
which perceive and act in an environment with the aim of achieving edge about the environment and about the impact of their actions is
their goals or maximizing a performance parameter. Moreover, they stored. That KB is used by the agents when devising plans of action on
can further improve their performance through learning [1]. how to succeed on decision-making, and can be updated in order to
In this section, we briefly go through some of the subfields of AI adapt to changing conditions. Along this line, holistic frameworks, like
that have been successfully employed in optical networking, stating the cognitive optical networks, which perceive, act, learn, adapt and opti-
motivation for their introduction, and providing some examples of their mize their performance, have been proposed by different researchers
use in the optical networking literature. Fig. 1 shows a diagram with AI [13–18].
subfields and techniques, and classifies the references reviewed in this Three noteworthy topics when it comes to incorporating intelligence
survey within those categories. to optical systems or networks are how to handle uncertainty, how to
The simplest type of networking scenarios that we can think of are tackle decision-making, and how to learn.
deterministic, observable, static and completely known. For these sce- For sure, in an optical network there are non-deterministic events
narios, search algorithms and optimization theory are key elements of taking place, and lack of full information about the environment is not
the AI area [1], and thus, they have been extensively used in optical a rare issue. Therefore, intelligent agents must be able to operate under
network design and control for a long time. Examples include the use uncertainty in a robust way. The laws of probability and, in particular,
of breadth-first-search algorithms for routing, and linear and mixed- Bayesian networks are useful tools to build those robust models (e.g. Ref.
integer linear programming formulations for network planning (e.g., [19]). Moreover, optical systems and networks are subject to constant
[2,3]). However, when some of the conditions previously mentioned changes. Hence, intelligent agents must include inference algorithms for
are relaxed, or when the network size prohibits the use of the for- temporal models to perform tasks like filtering, prediction or smooth-
mer techniques, those methods have been complemented or replaced ing, relying on techniques like hidden Markov models (HMM) [20] and
by local search algorithms and metaheuristics like simulated annealing, Kalman filters [21,22].
genetic algorithms, swarm optimization, and teaching-learning based A second key element is the use of decision-making algorithms.
optimization [4,5]. For instance, optical network planning [6,7] and The underlying principle for these algorithms is the maximization of
lightpath establishment [8,9] have benefited from those techniques. the expected utility, in which a utility function is defined in order to
In many cases, the optical network has a single point of intelligence assign a single number to express the desirability of a state and an
(like a centralized control node), i.e., a single agent. However, in other agent makes decisions with the aim of maximizing such a function (e.g.
cases, different intelligent agents are involved, so that the actions taken Ref. [23]). Realistic networking environments, however, must deal with
by an agent have an impact on the others. For those scenarios, game uncertainty and the utility of an agent usually depends on a sequence
theory may come into play, and proposals in the area of optical net- of decisions rather than on a single isolated one. Decision making in
working can be found, e.g., in Ref. [10] (in the context of hybrid radio- optical network agents can therefore be modelled as sequential deci-
frequency/free space optics networks), [11] (in the context of elastic sion problems in uncertain environments. These problems can be solved
optical networks, EONs), or [12] (a book completely devoted to the by Markov decision processes (MDPs) if the agent’s actions depend only
topic). on the current state of the agent, and not on its history. MDPs are

44
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

defined by a transition model, which specifies the probabilistic out- 3.1. Characterization and operation of transmitters
comes of actions, and by a reward function, which specifies the reward
in each state. The solution of an MDP is a policy that associates a deci- AI techniques facilitate statistical modeling of individual optical
sion with every state that the agent might reach. An optimal policy components by including the underlying physics. In all these cases
maximizes the utility of the state sequences encountered when it is where a deterministic approach results in an impractical computational
executed. The use of MDPs in optical networking has been shown in load, learning mechanisms are becoming a promising and accurate per-
Refs. [24–26]. formance improvement tool.
The third issue of paramount importance is learning. Learning With the advent of advanced modulation formats aiming to increase
enables an agent to improve its performance on future tasks due to the spectral efficiency, ranging from 16 quadrature amplitude modula-
acquired experience. The inclusion of learning is important for several tion (16 QAM) to 64 QAM and beyond, the need for robust carrier fre-
reasons. A learning-capable agent can adapt to changes in the environ- quency and phase synchronization becomes crucial. At this point, a pre-
ment and it is even able to adapt to unforeseen scenarios that could not cise characterization of amplitude and phase noise of lasers is essential.
be anticipated when the agent was designed. Moreover, in many cases, Conventional time-domain approaches perform coherent detection in
learning from existing data may be the only way to generate a work- combination with digital signal processing (DSP) to cope with this issue
ing model, or in other words, as stated in Ref. [1], sometimes human [60,61], but as higher order modulation formats are implemented, the
programmers (or engineers) have no idea on how to program a solu- accuracy of the phase noise estimation is compromised in the presence
tion themselves. Statistical learning and machine learning provide the of moderate measurement noise. Zibar et al. [46] present a framework
theory and tools to learn from existing data, which can be gathered of Bayesian filtering in combination with expectation maximization
in optical communications systems and networks thanks to monitoring (EM) to accurately characterize laser amplitude and phase noise that
techniques. outperforms conventional approaches. Results demonstrate an accurate
Although agents can handle uncertainty by using the methods estimation of the phase noise even in the presence of large measurement
of probability and decision theory, they must learn their probabilis- noise.
tic theories from experience. Thus, bayesian learning methods [27] Additional examples of the use of AI techniques in the optimization
formulate learning as a form of probabilistic inference, using the of transmitters and lasers include the work by Hragui et al. [47], who
observations to update a prior distribution over hypotheses; maxi- use simulated annealing to determine the optimal settings in terms of
mum a posteriori (MAP) learning [28] selects a single most likely flatness for optical comb sources for ultradense WDM passive optical
hypothesis given the data, and maximum-likelihood learning [29] sim- networks, and the work by Brunton et al. [48], who jointly use machine
ply selects the hypothesis that maximizes the likelihood of the learning, genetic algorithms and adaptive control techniques to provide
data. These techniques have been used in optical receivers, e.g., in a self-tuning mechanism for mode-locked fiber lasers.
Refs. [30–32].
Apart from the above mentioned techniques machine learning has 3.2. Operation of erbium-doped fiber amplifiers (EDFAs)
also been widely used. There are three main categories in machine
learning. In supervised learning [1] an agent observes some example EDFAs are another optical network component on which AI tech-
input-output pairs and learns a function that maps from input to out- niques have been extensively applied. EDFAs are one of the key ele-
put. Techniques include linear regression, logistic regression, decision ments of optical transport networks, capable of extending the reach
trees, artificial neural networks, nearest neighbor models and support of the transmitted optical signal by performing amplification of WDM
vector machines (SVM) to name just a few. Moreover, different mod- channels in the optical domain. Machine learning techniques offer effi-
els can be combined in ensemble learning, with the aim of improv- cient solutions to a wide range of challenges inherent to the operation
ing results. Supervised learning has been used, for instance, for opti- of EDFAs within optical fiber transmission.
cal performance monitoring [33], to estimate the quality of trans- Specifically, Huang et al. [49] define a regression problem with
mission (QoT) in optical networks [34–38] and for resource alloca- supervised machine learning (using a radial basis function) to statisti-
tion in data centers [39]. In unsupervised learning [1], an agent learns cally model the channel dependence of power excursions in multi-span
patterns from the input even though no explicit output is supplied. EDFA networks, learning from historical data. It provides the system
For instance, clustering and principal component analysis methods, with accurate recommendations on channel add/drop strategies to min-
which belong to this type of learning, have been used for optical per- imize the power disparity among channels. With the arrival of flex-grid
formance monitoring, modulation format recognition and impairment networks, in which dynamic defragmentation is often applied to reopti-
mitigation [40–42]. Finally, in reinforcement learning [43] an agent mize spectrum assignment to active connections in order to improve the
learns an optimal (or nearly optimal) policy from a series of reinforce- spectral efficiency, the previous study is extended in Ref. [50] to cope
ments (rewards) or punishments received from its interaction with the with the power excursion problem in dynamically changing spectral
environment. Some techniques include adaptive dynamic programming configurations. A ridge regression model is used to determine the mag-
and temporal-difference (TD) methods. Q-learning, a well-known tech- nitude of the impact of a given sub-channel, and a logistic regression
nique of the latter type, aims to find an optimal quality value (Q- is applied to specify whether the contribution will result in an increase
Value) of action-selection policy for any given (finite) Markov deci- or decrease in the discrepancy among post-EDFA powers. Additionally,
sion process [44]. For instance, Q-learning has been used for path and a novel method for autonomous adjustment of the operating point of
wavelength selection in the context of optical burst-switched (OBS) amplifiers in an EDFA cascade by using a multilayer perceptron neural
networks [45]. network is presented in Ref. [51]. The aim of this adjustment is to opti-
mize the performance of the link by minimizing both the noise figure
and the ripple of the frequency response of the transmission system
3. Applications of AI in optical transmission while ensuring predefined input and output power levels.

In this section, we describe applications of AI techniques in the 3.3. Performance monitoring

physical layer of optical networks, i.e., in optical transmission-related
issues. AI techniques can help improve the configuration and operation A challenge in network control and management is to adapt to the
of network devices, optical performance monitoring, modulation format time-varying link performance parameters, such as optical signal to
recognition, fiber nonlinearities mitigation and quality of transmission noise ratio (OSNR), nonlinearity factors, chromatic dispersion (CD) and
(QoT) estimation. These applications are summarized in Table 1. polarization mode dispersion (PMD). This subsection analyzes the suit-

45
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

Table 1
Applications in optical transmission taking advantage of AI techniques.
Applications AI techniques Literature
Transmitters Bayesian filtering and expectation- [46]: characterizes laser amplitude and phase noise.
maximization
Simulated annealing [47]: determines the optimal settings for optical comb sources for ultra-
dense WDM passive optical networks.
Machine learning (pattern learning meth- [48]: self-tuning mechanism for mode-locked fiber lasers.
ods) and genetic algorithms
Optical amplification control Kernelized linear regression [49]: defines regression model to study power excursions in multi-span
EDFA networks.
Linear/logistic regression [50]: uses a ridge regression model to cope with the discrepancy among
post-EDFA channel powers.
Multilayer perceptron neural network [51]: autonomous adjustment of the operating point of amplifiers in an
EDFA cascade.
Linear impairments identification Kalman filter [21]: carrier phase tracking, polarization tracking, and estimation of the
first-order PMD.
Neural networks [33]: identifies CD, PMD and OSNR provided that bit-rate and modula-
tion format is known.
Principal component analysis [41]: monitors CD, PMD and OSNR.
OSNR monitoring Deep neural networks (DNN) [52]: uses DNN, trained with asynchronously sampled raw data, for
OSNR monitoring.
Neural networks [53]: uses neural networks based nonlinear regression for OSNR estima-
tion.
Modulation format recognition Principal component analysis [41]: identifies modulation formats/bit rates from a known set.
Support vector machines (SVM) [53]: classifies modulation formats using the variance of eye opening
width.
Clustering k- means [40]: identifies modulation formats based on the number of levels and
clusters in constellation diagram.
Receivers, nonlinearity mitigation Maximum a posteriori [30]: looks for phase estimates feasible to calculate in real-time.
Maximum-likelihood [31]: proposes various equalization schemes for high capacity WDM
interconnects.
[54]: Maximum-likelihood detection for phase-modulated systems with
linear and nonlinear phase noise.
Maximum-likelihood and maximum a pos- [32]: proposes various estimators to recover the phase in Offset-QAM-
teriori based filterbank multicarrier systems.
Bayesian filtering and expectation- [55]: proposes state-space models for cross-polarization mitigation, car-
maximization rier synchronization, symbol detection.
Nonlinear support vector machines [56]: SVM is applied to create decision boundaries to avoid errors
induced by nonlinear impairment.
K-nearest neighbors [57]: proposes an algorithm that learns the link properties and generates
the nonlinear decision boundaries for maximizing transmission distance
and improving nonlinear tolerance.
Clustering k-means [42]: proposes a technique to mitigate the effect of time-varying impair-
ments, e.g., phase noise.
Nonlinear support vector machines and [58]: uses Newton-method (N-SVM) to reduce inter-subcarrier nonlinear
Newton method crosstalk effects.
QoT estimation Case-Based Reasoning (CBR) [34]: presents a QoT estimator to decide whether a lightpath fulfils QoT
requirements or not.
CBR + learning/forgetting [35]: optimizes previous CBR approach for QoT estimation with learning
and forgetting techniques.
CBR + learning/forgetting [36]: experimental demonstration of the QoT estimator [35] in a WDM
80 Gb/s PDM-QPSK testbed.
Random forests classifier [37]: predicts the probability that the BER of a candidate lightpath will
not exceed a given threshold.
Linear regression [38]: uses BER information monitoring and a learning process (based
on linear regression) in order to estimate the BER of each new service
request.
Support vector machines [59]: proposes a fast and accurate lightpath QoT estimator based on SVM
to decide whether a lightpath fulfils
QoT requirements or not.

ability of the application of AI techniques in monitoring some of the for carrier phase tracking, polarization tracking, and estimation of the
aforementioned factors. first-order PMD. However, techniques applied in Ref. [33] and simi-
The estimation and acquisition of physical parameters of transmit- lar ones [63,64] require prior knowledge about the type of signal (bit-
ted optical signals allow network-diagnosis in order to take actions rate and modulation format), or additional cross-layer communication
(repairing damages, driving compensators/equalizers or rerouting traf- is required at the intermediate nodes to acquire this information from
fic around non-optimal links) against malfunctions [62]. As an example, the upper-layer protocols, which would result in a significant increase
Wu et al. [33] present an extensive study of the application of artifi- in node complexity. In this context, a novel technique for simultaneous
cial neural networks in optical performance monitoring (OPM), which linear impairments identification (OSNR, CD and PMD) that is inde-
includes the simultaneous identification of accumulated nonlinearity, pendent from bitrate and modulation format, provided that this infor-
OSNR, CD and PMD, from eye-diagram and eye-histogram parame- mation belongs to a known set, is proposed in Ref. [41]. The study is
ters, while Szafraniec et al. [21] propose Kalman filter as an estimator performed using principal component analysis-based pattern recogni-

46
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

tion on asynchronous delay-tap plots and it yields accurate results in linear tolerance improvements are demonstrated in a 16 QAM coher-
the simultaneous monitoring of linear impairments. ent transmission system. Following the same line of study, Torres et al.
Another recent work facing the limited scalability of the studies pre- [42] propose a non-symmetric demodulation technique for receivers
viously mentioned, which are based on the prior knowledge of a deter- equipped with DSP based on clustering (using k-means algorithm),
mined set of signals is presented in Ref. [52], where a deep neural which mitigates the effect of time-varying impairments such as imbal-
network (DNN), trained with raw data asynchronously sampled by a ance of in-phase and quadrature signals (IQ imbalance), bias drift and
coherent receiver is proposed for OSNR monitoring. Results show that phase noise. This machine learning-based demodulator is computation-
OSNR is accurately estimated. Yet, this DNN needs to be configured ally highly efficient and also transparent with respect to the nonlinear-
with at least 5 layers and needs to be trained with 400,000 samples ity source. Finally, a recent study [58] extends previous approaches by
to achieve accurate results, requiring long training time. Alternately, introducing these techniques in more advanced systems, with greater
Thrane et al. [53] propose an OSNR estimator and a modulation format spectral efficiency, such as coherent optical orthogonal frequency divi-
classifier for systems employing advanced modulation formats (up to sion multiplexing (CO-OFDM) systems. The proposed algorithm is a
64 QAM) and direct detection. The OSNR estimator employs a neural nonlinear equalizer SVM of reduced classifier complexity using the
network, while the modulation format classifier uses a support vector Newton-method (N-SVM). It achieves an effective handling of inter-
machine (SVM), both in order to learn a continuous mapping function subcarrier nonlinear crosstalk effects and an increase of the launched
between input features extracted from the power eye-diagram after the optical power with low computational load.
photodetector and the reference OSNR and modulation format, respec-
tively. Although accurate results are obtained for OSNR estimation and
3.5. Quality of transmission (QoT) estimation
modulation format classification, the study only considers white Gaus-
sian noise, while ignoring for the moment linear and nonlinear optical
Optical connection (or lightpath) QoT estimation prior to deploy-
fiber impairments.
ment is particularly relevant in impairment-aware optical network
design and operation. Azodolmolky et al. [67] presented a QoT estima-
3.4. Receivers and mitigation of nonlinearities
tor tool, the Q-Tool, which computes the associated Q-factors of a set
of lightpaths, given a reference topology, by combining analytical mod-
Currently, the information capacity of fiber optic systems is limited
els and numerical methods. These estimates are relatively accurate, but
by nonlinear effects of the optical fiber. Extensive research effort has
the necessary high computing time to perform the calculations makes
attempted to address mitigation of nonlinearities on the transmission
this tool impractical in scenarios where time constraints are important.
over optical fiber. Among these nonlinearities, nonlinear phase noise
Several approaches propose cognitive techniques to solve this draw-
(NLPN) is one of the prominent factors. So far this issue has been treated
back. As an example, Jiménez et al. [34] present a QoT estimator capa-
with electronic methods relying on the deterministic information of the
ble of exploiting previous experience and thus, provide with fast and
fixed fiber link, like maximum likelihood estimation [54], digital back
correct decisions on whether a lightpath fulfils QoT requirements or
propagation [65] and stochastic digital back propagation [66], which
not. It is based on case-based reasoning (CBR) [68], an artificial intel-
may be computationally too heavy for practical implementation.
ligence mechanism that offers solutions to new problems by retrieving
Currently, machine learning techniques are being incorporated to
the most similar cases faced in the past whether by reusing them or after
digital signal processing to mitigate nonlinearities in a more efficient
adapting them. Cases are retrieved from a knowledge base (KB), which
way, allowing more accurate symbol detection. As an example, a
can be static [34] or optimized with learning and forgetting techniques
cognitive digital receiver is proposed in Ref. [40], which, by means
[35]. Results for CBR relying on an optimized KB show an excellent
of clustering algorithms, is able to identify the incoming signal for-
rate of successful classification of lightpaths into high/low QoT cate-
mat, QPSK/8PSK/16QAM, without the need to receive a prior con-
gories and more important, up to four orders of magnitude faster than
trol message, thus opening the door to the autonomous modification
the Q-Tool mentioned above. Furthermore, this study is experimentally
of the modulation format. In addition, state-space models in combi-
demonstrated in a WDM 80 Gb/s PDM-QPSK testbed [36], where, even
nation with Bayesian filtering and expectation maximization are pre-
with a very small KB, very high rates of successful classifications of
sented in Ref. [55] with the aim of taking into account the underlying
lightpaths are achieved. One step further, and with the aim of further
physics of the channel and optical elements in the formulation of signal
reducing the prediction time, Mata et al. [59] propose the use of an
processing algorithms. As a result, an overall system improvement is
SVM classifier to predict if a lightpath fulfils QoT requirements or not.
achieved, including cross-polarization mitigation, carrier synchroniza-
This classifier proves to be not only significantly faster but also more
tion and optimal symbol detection. However, expectation maximization
accurate than the proposal in Ref. [35].
depends on the parameters of the transmission link and consequently it
Another proposal for QoT estimation is that of Barletta et al. [37],
is not applicable to dynamic optical networks.
who apply a machine learning-based classifier, specifically a random
Furthermore, Wang et al. [56] propose a machine learning algo-
forest, to predict the probability that the BER of a candidate lightpath
rithm to mitigate NLPN affecting M-ary phase-shift keying (M-PSK)
will not exceed a given threshold. Finally, Oda et al. [38] present the
based coherent optical transmission systems. Specifically, the algorithm
concept of “living network”, an optical network which keeps records of
introduced is a nonlinear SVM classifier able to generate nonlinear
its path-level performance, which takes advantage of BER information
decision boundaries that allows to bypass the errors induced by non-
monitoring and of a learning process (based on linear regression) in
linear impairments in the constellations of M-PSK signals, resulting in
order to estimate the BER of each new service request.
improvements both in the maximum transmission distance and launch
power dynamic range. Notwithstanding, SVM is basically a binary clas-
sifier, so to deal with higher order modulation formats, many SVMs 4. Applications of AI in optical networking
would be necessary.
Drawbacks derived from both previously mentioned studies are AI presents several opportunities for automating operations and
solved in Ref. [57], where a k-nearest neighbors-based detector is introducing intelligent decision making in network planning and
described and demonstrated. This algorithm only needs a small set of in dynamic control and management of network resources, includ-
labeled data in order to learn the link properties and generate the non- ing issues like connection establishment, self-configuration and self-
linear decision boundaries. Moreover, it performs a multi-class classifi- optimization, through prediction and estimation by utilizing present
cation and, therefore, it is capable of classifying multiple kinds of data network state and historical data. In this section, we review these appli-
simultaneously. In this way, maximum transmission distance and non- cations as well as use cases of AI in optical burst-switched networks

47
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

(OBS), in passive optical networks (PONs) and intra-datacenter net- harms the channels’ utilization.
works. These applications are summarized in Table 2.
4.2. Connection establishment
4.1. Optical network planning
Metaheuristics like simulated annealing and evolutionary methods
As described in Section 2, search algorithms and optimization theory like genetic algorithms or particle swarm optimization, are effective
have been widely used for optical network planning and dimensioning in solving hard optimization problems because they are less likely to
(e.g., [2,3]), usually complemented or extended with local search algo- become trapped in local optima. Therefore, these methods are useful
rithms and metaheuristics like simulated annealing, swarm optimiza- to solve the optical connection (lightpath) establishment problem in
tion and genetic algorithms [5]. optical networks. In WDM networks, this involves searching a combi-
Optical network planning involves tasks like designing the physical nation of route and available wavelength, and is so called the routing
topology of the network and ensuring survivability while minimizing and wavelength assignment (RWA) problem. In elastic optical networks
costs. Morais et al. [69] propose the use of genetic algorithms (GAs) (EONs), it involves searching for a route and a portion of available spec-
to address those issues in an opaque optical transport network, and de trum and even a modulation format, i.e., solving the routing and spec-
Miguel et al. [72] also rely on a GA for dimensioning dynamic WDM ring trum allocation (RSA) or the routing, modulation level and spectrum
networks. A related optimization problem, like minimizing the number allocation (RMLSA) problems.
of all-optical regenerators, is tackled by Martinelli et al. [6] with a GA, A multi-objective GA for solving the impairment-aware static RWA
which also jointly solves the routing and wavelength assignment (RWA) problem is presented in Ref. [8], and Rubio-Largo et al. [76] present a
problem while ensuring the QoT for the lightpaths to be established. comparative study among three multiobjective evolutionary algorithms
The problem of placing regenerators in optical networks subject to fault (MOEAs) based on swarm intelligence to solve the RWA problem in
tolerance constraints has also been approached by means of ant colony real-world optical networks: artificial bee colony algorithm, gravita-
optimization (ACO) techniques [7]. This proposal guarantees surviv- tional search algorithm and firefly algorithm, concluding that swarm
ability of the underlying network whilst also minimizing the number of intelligence is very suitable for this task.
regenerators required. High quality solutions are provided, with reason- Wang et al. [9] include considerations of mixed line rate, physical
able runtimes. Furthermore, a GA aiming to jointly optimize the selec- impairments and traffic grooming functionality to solve the RWA prob-
tion of nodes performing 3R regeneration, code rate, and routing and lem by means of an ACO algorithm. Different configurations of this
spectrum allocation for lightpaths to be established in flex-grid code- distributed solution are compared to each other and also with a central-
rate adaptive optical networks is proposed in Ref. [70]. The addition ized grooming adaptive shortest path algorithm. Although the central-
of the code rate in the classical resource and allocation problem entails ized solution shows better efficiency in terms of blocking probability,
establishing a trade-off between minimizing the number of regeneration ACO shows great robustness and adaptability to varying network and
nodes and minimizing the need of spectral resources. Results indicate traffic conditions. Additionally, in Ref. [83], a cognitive approach (case-
that, in general, with just few nodes selected for regeneration, it is pos- based reasoning) is introduced into the traditional RWA algorithm for
sible to ensure QoT and exploit the advantages of code-rate adaptive- dynamic WRONs with the aim of reducing computing complexity. Simu-
ness. In the context of flex-grid network planning, it is worthy to note lation results indicate that taking advantage of similar past experiences
the work by Velasco et al. [71]. They propose a novel network archi- or cases stored in a knowledge base (KB) can reduce computational
tecture consisting of a set of IP/MPLS areas performing routing and time by 25% over classical RWA algorithms, while maintaining or even
flow aggregation, which are interconnected through a flex-grid optical improving performance. In addition, Kyriakopoulos et al. [81] propose
core. In order to obtain near-optimal solutions for this architecture for a heuristic method based on ACO to reduce network energy footprint
real-sized network and traffic instances, they employ GAs (in particular, by exploiting the basic principles of swarm intelligence for finding the
biased random-key GAs). Under these circumstances, simulation results most energy-efficient routes from source to the destination node per
reveal that extending the core toward the edges results in significant traffic request. A different ACO-based proposal [82], which introduces
savings in both the core and IP/MPLS networks. a heuristic on the way ants choose a request from demand space (those
A different approach [73], presents a particle swarm optimization that can be served with shorter route first), outperforms both regular
(PSO) algorithm in order to solve the problem of resource allocation ACO and shortest-path and most-used algorithms. Additionally, Araújo
based on the signal-to-noise plus interference ratio optimization in a et al. [89] present a mechanism to estimate the blocking probability
hybrid wavelength division multiplexing/optical code division multi- when establishing lightpaths in an optical network. It consists of an
plexing network (WDM/OCDM) under Quality of Service (QoS) restric- artificial neural network which uses as inputs topological properties
tions and the energy efficiency constraint problem. The PSO strategy and general physical layer characteristics (on which a principal compo-
allows the regulation of the transmitted power in order to maximize nent analysis is previously carried out). Results show a speed-up greater
the energy efficiency. Results show interesting trade-offs between per- than 7500 times than that of a discrete event simulator, and accurate
formance and complexity. Following the same trend [74], presents an estimates are obtained except when very small blocking probabilities
alternative algorithm, a heuristic ACO scheme, for allocation of trans- are evaluated.
mitted power with increasing energy efficiency applicable to optical The RSA problem in EONs is NP-hard [123], and has also been
WDM/OCDM transport networks. In addition, an analytical disciplined considered an appropriate candidate to be solved by metaheuristics.
convex optimization approach, taking into account the performance In fact, in the seminal paper by Christodoulopoulos et al. [84] on elastic
and complexity metrics, is proposed as comparison. Simulation results bandwidth allocation, simulated annealing is used to solve the RMLSA
demonstrate that the ACO scheme proves to be useful in order to obtain problem. Simulated annealing has been also used by Perelló et al. [85]
spectral-efficient and energy efficient systems suitable for WDM/OCDM in elastic networks with space division multiplexing (SDM), where not
networks, with promising performance–complexity trade-offs in com- only route, modulation format, and spectrum have to be assigned but
parison with the analytical approach. also a fiber core, finally solving the route, modulation format, core, and
Another example of the use of AI in resource allocation is the work spectrum assignment (RMCSA).
by Petridou et al. [75]. They propose a message scheduling algorithm, In the context of evolutionary methods, the RMLSA problem has
based on the k-means clustering algorithm, which addresses both mes- been solved by means of a GA where two populations evolve in par-
sage sequencing and channel assignment for a WDM star network. allel and use a migration operation to exchange individuals between
Based on the produced clusters, the scheduling algorithm manages to them [77]. Furthermore, the RSA problem has also been solved in Ref.
avoid scheduling consecutive messages to the same destination which [78] with GAs for multicast flows considering different selection and

48
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

Table 2
Applications in optical networking taking advantage of AI techniques.
Applications AI techniques Literature
Survivable optical networks Genetic algorithms [69]: designs the physical topology of the network ensuring survivabil-
ity.
Ant colony optimization [7]: guarantees survivability of the underlying network whilst minimiz-
ing the number of regenerators.
Regenerator placement Genetic algorithms [6]: minimizes the number of all-optical regenerators in order to mini-
mize network cost.
[70]: optimizes the selection of regenerators, code rate, and routing and
spectrum allocation in flex-grid code-rate adaptive optical networks.
Ant colony optimization [7]: guarantees survivability of the underlying network whilst minimiz-
ing the number of regenerators.
Resource allocation Genetic algorithms [71]: proposes a novel network architecture interconnecting a set of
IP/MPLS areas, performing routing and flow aggregation, through a flex-
grid optical core.
[72]: proposes a method for joint routing and dimensioning of dynamic
WDM ring networks.
Particle swarm optimization [73]: solves the problem of resource allocation under Quality of Service
(QoS) restrictions and the energy, efficiency constraint problem.
Ant colony optimization [74]: allocation of transmitted power for energy efficient optical
WDM/OCDM networks.
K-means clustering [75]: proposes a message scheduling algorithm that addresses both the
message sequencing and channel assignment issues for a WDM star net-
work.
Markov decision processes [26]: models the resource allocation problem as a MDP to optimize an
objective arbitrarily defined by the network operator.
Connection establishment Swarm intelligence [76]: proposes multiobjective algorithms based on swarm intelligence to
solve the RWA problem.
Genetic algorithms [6]: solves the RWA problem while also ensuring QoT of lightpaths to
be established.
[8]: solves the Impairment Aware static RWA problem.
[77]: solves the RMLSA problem.
[78]: considers multicast flows for RSA using different selection and
crossover strategies.
[79]: solves the RSA problem in flex-grid networking producing useful
insights into network design.
[80]: solves the RSA problem with joint anycast and unicast demands.
Ant colony optimization [9]: RWA solution with great robustness and adaptability to varying net-
work and traffic conditions.
[81]: reduces network’s energy footprint by finding the most energy-
efficient routes.
[82]: introduces a heuristic on the way ants choose a request from
demand space in order to find the shortest path.
Case-based reasoning [83]: reduces computing complexity when solving the traditional RWA
problem in dynamic WRONs.
Simulated annealing [84]: solves the RMLSA problem in elastic networks.
[85]: solves the RMCSA problem in elastic networks with space division
multiplexing (SDM).
Tabu search [86,87]: solve RSA problem providing also dedicated path protection.
Backpropagation neural network [88]: predicts the arrival time and holding time of future connections by
considering past experiences.
Q-learning [45]: solves path and wavelength selection in OBS networks.
Game theory [11]: solves RSA problem by properly balancing the spectrum utilization
and security-level of the domain in multidomain EONs.
Neural networks and principal component [89]: estimates the connection blocking probability.
analysis
Kalman filters [22]: reduces blocking ratio by estimating the wavelength occupancy
prior to the RWA decision.
Markov decision processes [24]: derives the optimal lightpath establishment policy for each service
class by means of a MDP.
[25]: proposes the use of an efficient dynamic-preemption call admission
control scheme based on the optimal policy derived from a MDP.
Network reconfiguration Genetic algorithms and ant colony opti- [90]: survivable mapping of a given WDM virtual topology.
mization
Genetic algorithms [91]: designs virtual topologies while reducing energy consumption and
network congestion.
[92]: techno-economic study of the introduction of cognitive techniques
in virtual topology design.
[93]: addresses reliable multicast Virtual Network mapping for OFDM
based EONs.
Genetic algorithms and cognition [94]: produces estimations that can help to anticipate changes in the
traffic and proactively reconfigure the virtual network topology.
Neural networks [95]: performs reconfigurations based on the traffic volume and direc-
tion predicted by a neural network.

(continued on next page)

49
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

Table 2 (continued)
Applications AI techniques Literature
Failure/fault detection Bayesian networks, clustering [96]: identifies or locates failures in the virtual network topology to
improve quality of service.
Cognition-based methods [97]: detects failures in centralized SDN-based networks by periodically
exchanging the messages
between controller and switches.
Bayesian inference/networks [19,98]: propose probabilistic modeling and machine learning for fault
diagnosis in optical access networks.
Software Defined Networking Cognition-based methods [99]: correct mapping of topologies in considerably low total times.
[100]: introduces a transport SDN controller that facilitates optical net-
work virtualization and
autonomic operation.
[101]: proposes a new inter-networking paradigm based on broker
agents with cognitive intelligence.
Neural networks [102]: maximizes link capacity after predicting link performance in cor-
relation with the OSNR.
Reduction/estimation of burst loss Learning automata [103]: achieves self-awareness, self-protection and self-optimization in
OBS networks.
Q-learning [45]: solves path and wavelength selection in OBS networks.
[104]: exploits the feedback loop to control the retransmission rate of
bursts that are lost.
[105]: introduces a low-complexity solution to resolve contention in OBS
networks
Hidden Markov model and expectation- [23]: proposes variations of the TCP protocols to enhance the perfor-
maximization mance of OBS networks.
Bayesian networks [106]: decreases the burst loss ratio (BLR) in a OBS network.
Feed-forward neural network and [107]: proposes deflection routing protocols that achieve smaller burst-
Q-learning loss probabilities
than previous approaches while deflecting bursts less frequently.
Extreme learning machine [108]: estimates burst loss probability.
Ant colony optimization [109]: proposes an ACO approach to reduce burst loss ratio, enhancing
at the same time the delay.
Statistical solutions for prediction Hidden Markov model (HMM) [20]: uses HMM based traffic prediction along with QoS aware light path
establishment in WDM networks.
Bayesian methods and game theory [10]: utilizes a Bayesian game-theoretic model to guarantee cooperative-
ness in RF/FSO networks.
Intelligent ROADM Linear regression [38]: autonomously keeps record of path-level performance.
Splitter placement in PONs Genetic algorithms [110–112]: these papers optimize the location of splitter to achieve var-
ious objectives in PON.
QoS guarantees and dynamic Genetic algorithms and neural networks [113–117]: use genetic algorithms or neural networks to assure QoS in
bandwidth allocation in PONs PON.
Bayesian estimation [118]: proposes estimation and prediction-based Just-In-Time dynamic
bandwidth allocation algorithm.
Placement of ONUs Teaching learning-based optimization [119]: reduces the required Optical Network Units (ONUs) that assure
connectivity among wireless routers and ONUs in a Fiber-Wireless net-
work.
Genetic algorithms [120]: another non-deterministic approach for placement of ONUs.
Cognitive optical networks Cognition-based methods [13–18]: these papers propose cognitive optical network architectures.
Intra-Datacenters Multilayer perceptron [39]: allocates resources (optical circuits/electrical switches) to flows
according to their requirements.
Neural networks [121]: presents a flow classifier at the edge of the network combined
with an SDN centralized controller.
Markov decision processes [122]: makes scheduling decisions in all-optical data center networks
guaranteeing throughput optimality
under a zero reconfiguration delay.

crossover strategies, and in Ref. [86] with tabu search techniques with by reducing as much as possible spectrum fragmentation. The latter,
the aim of providing dedicated path protection. Another coevolutive where source-destination pair bandwidth demands are located as far
approach is introduced in Ref. [80] to solve the RSA problem with joint apart from one another as possible across the optical spectrum solved,
anycast and unicast demands, outperforming previous proposals based employs a GA-based optimization in order to produce useful insights
on tabu search [87] and simulated annealing [124] approaches. into network design.
An alternative approach to solve the RSA problem in EONs is pre-
sented in Ref. [88], where a backpropagation neural network is pro-
posed to improve the RSA algorithm by predicting the arrival time 4.3. Network reconfiguration: virtual topologies
and holding time of future connections by considering past experi-
ences. Results confirm this approach outperforms RSA algorithms that The virtual topology is the set of optical connections (or lightpaths)
do not make use of historical information. Another work [79], applies established in a network. It does not have to be statically configured,
a Shannon entropy-based fragmentation metric to the RSA problem in but it can be dynamically reconfigured in order to better adapt to evolv-
flex-grid networking by utilizing two complementary approaches: mini- ing traffic demands with some objectives like reducing energy consump-
mum and maximum entropy. The former allows to increase the number tion, network congestion, end-to-end delay or blocking probability or
of demands that can be served before reaching critical blocking levels trying to ensure quality of transmission (QoT), etc. For that purpose,
two nature inspired heuristics, GA and ACO, are used in Ref. [90] to

50
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

obtain a survivable mapping of a given WDM virtual topology. Feasi- a dynamic routing algorithm to update the topology and restore the
ble solutions are obtained even for large topologies when integer linear optical path in a significantly short space of time.
programming methods cannot. Also, a multiobjective genetic algorithm Furthermore, Oliveira et al. [100] introduce a transport SDN
to design virtual topologies with the aim of reducing both the energy controller that facilitates optical network virtualization and auto-
consumption and the network congestion is presented by Fernández nomic/cognitive operation by means of two adaptive algorithms that
et al. [91]. The GA proposed there is enhanced with the capability of allow to reconfigure, on one hand, the transmission modulation for-
remembering solutions successfully used in the past, as well as connec- mat and spectrum utilization according to network conditions and on
tions with low QoT. The incorporation of those mechanisms leads to the other hand, the attenuation applied at the ROADMs to improve
improvements in performance. Furthermore, the introduction of cogni- the OSNR of the signals at the reception. Yan et al. [102] propose
tive techniques in virtual topology design also exhibits significant sav- and demonstrate in a field trial the planning of an SDN-based opti-
ings in terms of the total cost of ownership compared to conventional cal network utilizing neural network-based methods, which are able
methods. As a matter of fact, savings up to 20% and 25% in capital to predict link performance in correlation with the OSNR. By means
and operational expenditures, respectively, via a GA-based method, are of probabilistic-shaping bandwidth variable transmitters (BVTs), which
demonstrated in Ref. [92]. One step forward, the extension presented in are configured by the SDN controller based on these predictions, spec-
Ref. [94] uses monitored data to produce estimations that can help to tral efficiency can be adapted, maximizing the link capacity. Addition-
anticipate changes in the traffic and proactively reconfigure the virtual ally, a new inter-networking paradigm based on broker agents with cog-
network topology. nitive intelligence that compete to provide desirable inter-networking
An algorithm to identify/locate failures in the virtual network topol- services to autonomous systems through market-driven incentives, is
ogy that can lead to an unacceptable quality of service is proposed by proposed in Ref. [101].
Ruiz et al. [96]. They first perform the experimental characterization
of several causes of failure (which is done with the help of a cluster- 4.5. Applications in optical burst switching
ing algorithm), and then use those characterizations to train a Bayesian
network (BN). This trained BN is used to localize and identify the most Optical burst-switched (OBS) networks [2, Chap. 18] have also
probable cause of failure impacting a given service. taken advantage of artificial intelligence, and in particular, of machine
A virtual network topology reconfiguration approach is introduced learning techniques. Praveen et al. [103] propose a novel OBS archi-
in Ref. [95]. It performs reconfigurations based on the traffic volume tecture which takes advantages of learning automata to achieve self-
and direction predicted by an artificial neural network proposed for awareness, self-protection and self-optimization, consequently reduc-
every origin-destination pair. Periodically, collected monitoring data ing burst loss probability significantly. Work done in Ref. [103] has
are transformed into modelled data and the artificial neural networks been extended in different studies by using other machine learning
are used to predict the next-period traffic. Results show savings in both techniques, such as Q-learning, in order to solve the path and wave-
capital and operational expenditures. length selection problem [45], or by exploiting the feedback loop to
A different approach is followed in Ref. [93], where an efficient control the retransmission rate of bursts that are lost [104]. Moreover,
virtual network (VN) mapping for multicast services over both general variations of the TCP protocols to enhance the performance of OBS net-
IP networks and orthogonal frequency division multiplexing (OFDM)- works, including supervised and unsupervised learning techniques, are
based EONs, is presented. This proposal takes into consideration the also proposed in Ref. [23].
max-min fairness in terms of reliability among distinct VNs. In the IP Burst blocking or loss probability –the ratio of the number of lost
networks case and with the aim to globally optimize the reliability and bursts to the total number of transmitted bursts– is commonly used
fairness of all the multicast VN requests, a mixed integer linear pro- for the performance measurement of OBS network technologies. So far,
gramming (MILP) model to determine the upper bound on the reliabil- proposed techniques have proven to be too slow or not accurate enough
ity is presented, as well as a GA that addresses reliable multicast VN to estimate this parameter. New approaches based on machine learning
mapping. For OFDM based EONs, this solution is extended by consid- have outperformed previous studies, especially in terms of computation
ering the most efficient modulation format selection strategy, spectrum time. Leung et al. [108] present two models for burst loss ratio (BLR)
continuity, and conflict constraints. estimation employing neural networks based on the extreme learning
machine (ELM) framework. By using these models, estimates can be
4.4. Software defined networking obtained much faster than by means of simulations. Moreover, the accu-
racy of the BLR estimates outperforms that obtained with an existing
The software defined networking (SDN) paradigm [125], which analytical approach and are very close to the values obtained by sim-
decouples control and data planes, and enables programmability on the ulation. Three cognitive mechanisms –Bayesian networks, closed loop
former plane, has aroused the interest of both industry and research control and open loop control– to decrease BLR in an OBS network are
communities by allowing networks managers to manage, configure, introduced in Ref. [106]. Simulation results confirm that the applica-
automate and optimize network resources via software. In the context of tion of these methods in the admission process leads to a BLR reduc-
SDN over optical networks, a correct mapping of the underlying topol- tion in OBS networks. Similarly, in Ref. [105] a novel node degree
ogy at the control plane level is crucial. Following this requirement, dependent signalling algorithm in combination with Q-learning is pro-
a novel SDN-based cost-effective topology discovery method, allowing posed as a low-complexity deflection routing protocol with the aim
transparent optical networks to automatically learn physical adjacen- to resolve contention in OBS networks. This solution scales well for
cies between optical devices, is introduced in Ref. [99]. This is achieved large networks, since its complexity depends on the node degree rather
by means of a test-signal mechanism –by exchanging and verifying iden- than network size. Simulation results show that despite its lower com-
tifier information between discovery agents– and the OpenFlow proto- plexity, burst loss probability of the proposed algorithm is comparable
col, resulting in correct mapping of the topologies in low total times. to other existing reinforcement learning-based deflection routing algo-
In relation to this paradigm, fault tolerance is of paramount impor- rithms. The previous work is extended in Ref. [107], where a framework
tance when it comes to characterizing optical networks. In Ref. [97], that adds intelligence to deflection routing in buffer-less architectures
an efficient cognitive process for failure detection in centralized SDNs is is presented. In particular, by means of the combination of the node
proposed. Specifically, a network controller interacts with optical cross- degree-dependent signalling algorithm with a feed-forward neural net-
connects (OXCs) exchanging messages periodically to efficiently detect work (NN) and a feed-forward NN with episodic updates, both contain-
failures by using the link layer discovery protocol (LLDP). Fast and ing Q-learning-based decision-making modules, the proposed deflection
accurate communication of these link events to the controller allows routing protocols are proved to achieve smaller burst-loss probabili-

51
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

ties than previous approaches while deflecting bursts less frequently. In requirements. Thus, Jiménez et al. [115] develop an automatic tun-
addition, the solution requires less memory and CPU resources, which ing technique based on GA to tune a proportional controller to pro-
are more significant as the size of the network grows. Additionally, vide delay guarantees. In addition, they integrate a dynamic admission
Coulibaly et al. [109] propose an ACO approach to reduce burst loss control module to transmit or drop packets in order to achieve a bet-
ratio, enhancing at the same time the delay, by means of an adap- ter delay control. Besides, same authors in Ref. [117] propose a PID
tive and quality of service (QoS)-aware route, wavelength and timeslot controller integrated with an ANN to efficiently ensure QoS bandwidth
assignment algorithm. requirements in EPONs. Finally, Dias et al. [118] propose a dynamic
bandwidth allocation algorithm which uses Bayesian estimation to esti-
4.6. Applications in passive optical networks (PONs) mate the average interarrival time of packets at the ONUs, with the aim
of minimizing the queuing delay introduced by the sleep- and doze-
Network planning of passive optical networks (PONs) [2, Chap. 5] mode operations.
has also taken advantage of AI techniques. Villalba et al. [110] propose
a GA for topology searching and splitter placement in these networks.
4.7. Applications in intra-datacenter networking
A graph representation scheme, associated with a street map, is used
as a reference, and the formulation minimizes the amount of optical
Intra-datacenter (DC) networks are also embracing machine learn-
cabling and number of splitters and power budget. Besides, Kokangul
ing techniques in order to improve performance. For instance, in
[111] proposes a GA and mathematical modeling techniques to opti-
hybrid-switching-based DCs, where an electrical packet-switched and
mize the position of the primary and secondary nodes (i.e., the points
an optical circuit-switched network live together, machine learning-
where the signal is split the first and the second time, respectively),
based flow classification may be a decisive solution to improve speed
their split levels, the association of customers to secondary nodes,
and accuracy, besides improving adaptability to traffic dynamics. As
and the association of secondary nodes to primary nodes. This place-
an example, Rastegarfar et al. [39] have proposed a multi-layer per-
ment is done under constraints such as attenuations and characteristics
ceptron to allocate resources to TCP flows according to their require-
of the optical devices. Liu et al. [112] propose a topology optimiza-
ments (e.g., allocating optical circuits to bulk data transfer and mapping
tion model in long-reach passive optical networks (LR-PON) using a
short mice flows to electrical switches), and have obtained very signif-
nested genetic algorithm (NGA). The outer-loop of the NGA algorithm
icant improvements in network throughput due to both optical chan-
deals with the location of splitters and the inner-loop of the algorithm
nel bandwidth consolidation and adaptive flow classification. Further-
builds the spanning tree. Bhatt et al. [119] propose a teaching-learning
more, a neural network flow classifier at the edge of the network, com-
based optimization (TLBO) algorithm to reduce the required optical net-
bined with an SDN centralized controller able to take advantage of this
work units (ONUs) that assure connectivity among wireless routers and
classification outcome along with its global view of the resources has
ONUs in a Fiber-Wireless network. Results of the simulations carried
been proposed in Ref. [121]. Finally, Wang and Javidi [122] have also
out for different grid sizes of the geographical area and variable wire-
recently considered the end-to-end scheduling problem in all-optical
less routers confirm that the implemented scheme requires less ONUs,
data centers networks, whose main challenges are the bufferless nature
reaching a globally optimum solution that previous random and deter-
and the nonzero reconfiguration delay of optical switches, making it
ministic approaches failed to provide. Along the same line, other non-
necessary to use a centralized controller that can efficiently schedule
deterministic approaches for placement of ONUs also provide desired
the end-to-end transmissions. The proposed method employs an adap-
levels of performance, e.g., using genetic algorithms [120].
tive Markov scheduling policy, which makes decisions every time slot,
Other research works focus on the diagnosis and the self-diagnosis
and determines both the schedule and the time to reconfigure the sched-
of PONs. As an example, Gosselin et al. [19,98] propose probabilistic
ule based on the most recent queue length information.
modeling and machine learning for fault diagnosis in optical access
networks. The proposal is based on a Bayesian network which encodes
expert knowledge. Indeed, they develop a Bayesian inference engine, 5. New opportunities and challenges for the use of AI in optical
named probabilistic tool for GPON-FTTH access network self-diagnosis networks
(PANDA), to efficiently allow fault diagnosis in GPONs. Sarigiannidis
et al. [126] propose a 10-gigabit-capable PON (XG-PON) together with In this section, we describe a number of new opportunities and chal-
multiple long term evolution (LTE) radio access networks in the fron- lenges that we envision in the area of optical systems and networking.
thaul. The mechanism receives traffic-aware knowledge from the SDN We envision increasingly challenging roles of the use of AI in the phys-
controllers and it modifies the uplink-downlink configuration in the LTE ical layer, where it will continue being a useful tool not only in the
radio communication. This strategy allows to calculate an optimal con- framework of emerging optical transmission technologies but also in
figuration based on the traffic dynamics in the global network, allowing helping increasing security by means of attack or intrusion detection
for an improvement in the packet latency and jitter. and localization. We also describe the relevant role of AI in the automa-
Additional research is focused on assuring quality of service (QoS) tion of network management operations, and in the support of emerging
requirements and media access control (MAC) issues in PONs. Some networking paradigms.
approaches apply GAs to deal with these issues. As an example, Huang
[113] proposes to balance asymmetric traffic load between ONUs in
PON architectures using a GA, which also decreases congestion in 5.1. Optical transmission systems, and attack and intrusion detection
ONUs. Besides, Hwang et al. [116] formulate a genetic expression pro-
gramming (GEP) algorithm for the QoS traffic prediction integrated Many control decisions in a network are made based on accurate
with the limited packet transmission strategy to deal with the queue measurement or estimation of physical parameters. Thus, in line with
variation and as a consequence with the reduction of the delay of high the advances reported in Section 3, we expect AI will continue playing
priority traffic. In addition, Moradpoor et al. [114] propose a dynamic an important role in supporting emerging transmission technologies like
excess bandwidth allocation algorithm for an integrated hybrid PON space division multiplexing, multimode/multicore fibers and advanced
with wireless technology employing a GA. The algorithm is able to pro- modulation formats and constellation shaping. We also expect further
vide optimal/near-optimal solutions for the excess bandwidth assign- progresses in the use of AI techniques for QoT estimation and perfor-
ment in this converged network scenario. mance monitoring. However, a key area closely related to monitoring
Other approaches implement proportional-integral-derivative (PID) where we see a significant opportunity for AI techniques, but have not
controllers combined with GA or ANN techniques to comply with QoS yet emerged (to the best of our knowledge) in the optical arena, is that

52
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

of attack and intrusion detection. control loop.

Increasing Internet proliferation has driven businesses and other The widespread deployment of SDN, coupled with a telemetry-
institutions to provide essential services over the Internet infrastruc- driven feedback loop, would reduce the barrier for automating network
ture. As a result, these entities have to contend with higher risk and control operations. AI-based techniques presented in Section 4.2 can
potential costs associated with data breaches, which is currently esti- then be used to optimize RWA/RMLSA/RMCSA problems, performing
mated to be as high as $3.6 million [127] on average. The optical net- in-operation planning to optimize resources based on existing optical
work layer is also susceptible to attacks by malicious entities, which can service demands. As AI-based techniques in the area of network con-
be broadly classified as disruption and eavesdropping attacks. Disruption trol and management mature, we can also expect advanced applications
attacks attempt to exploit physical characteristics of optical transmis- such as preemptive relocation of services away from a predicted failure
sion to degrade the operation of existing services and/or block incom- in the network, a topic in which first steps have already been taken
ing service requests, while eavesdropping attacks are targeted towards [141–144]. As outlined in Section 4, network virtualization and recon-
gaining unauthorized access to the data transmitted over the optical figuration could also take advantage of AI-based techniques by utilizing
network. A variety of attack vectors for both these attack categories traffic prediction and classification, with or without SDN controller in
have been outlined in Refs. [128–130]. the loop. Nevertheless, the introduction of flex-grid devices, such as
Addressing attacks on optical infrastructure consists of mechanisms bandwidth-variable sliceable transponders and WSSs, and space divi-
to detect and localize an attack on the optical infrastructure, and the sion multiplexing techniques based on multi-core or few-/multi-mode
mitigation mechanisms in response to a detected attack. Statistical fibers have further increased the computation complexity of AI-enabled
approaches using optical measurements to identify and localize attacks RWA/RMLSA/RMCSA optimization solutions [135]. Nevertheless, AI-
in optical networks have been proposed in Refs. [131,132] and can be based techniques, especially deep learning, will play a major role in
employed to detect a range of attacks on the optical network infrastruc- optical network planning and reconfiguration, since optical devices are
ture. However, the computation complexity involved in applying the expected to imbibe more autonomous programmability features in the
proposed techniques in a large-scale network is non-trivial. Therefore, next decade.
AI techniques could be applied in this context to identify and localize
optical attacks. 5.3. Efficient joint operation of networks and computing resources
On successful identification of attacks, mitigation techniques, espe-
cially in the context of optical telecommunication networks involves all- Emerging paradigms like the Internet of things (IoT), Industry 4.0 or
optical encryption or re-routing of traffic away from attack locations. the tactile Internet [145], impose stringent requirements on networks,
The application of AI for network optimization and RWA problems has such as low latency, and high bandwidth, availability and security, thus
already been discussed in Sections 4.1–4.3, and can be employed to posing a significant challenge. The combination of 5G mobile communi-
move vulnerable optical traffic away form attack locations. cations systems with high-speed fault-tolerant fiber backhaul infrastruc-
tures will be key enabling technologies for these networks [145]. End-
5.2. Automating network management operations to-end latency for some applications can be limited to few milliseconds
(e.g., 1 ms for tactile internet). Thus, the distance between the edge and
In the area of optical network operation, heterogeneous (multi- computing resources must be limited to some tens of kilometers [146],
technology and multi-vendor) network devices make operation, admin- and a decentralized service platform architecture based on Mobile Edge
istration and maintenance of optical networks a complex and challeng- Computing (MEC) or Fog Computing (FC) is required. However, the
ing process. This is due to the fact that network state information, e.g., integration of various computing paradigms (MEC, Fog and cloud)
topology, congestion, failure discovery, etc., collected from different involves the development of integrated resource management, task allo-
devices have different and limited state information, which pose a big cation and failure handling techniques, to name just a few. Therefore,
challenge in data collection, processing and decision making. Thus, pro- the joint allocation of computing and networking resources (also includ-
cess automation has been identified as a key enabler for driving down ing inter-datacenter networking) is receiving increasing attention.
operational costs, and also to help move human intelligence away from AI is expected to play a key role to facilitate efficient joint oper-
repetitive tasks [133]. AI-based techniques are seen as a key enabler ation of network and computing devices, performing tasks like virtual
for automating network management operations [134,135], and cog- network function (VNF) distribution, task allocation, predictive caching
nitive optical networks were in fact proposed with that aim [13–18] and interpolation/extrapolation of human actions, and thus enhancing
but many open issues remain. The primary pre-requisite for automat- performance and providing better support for IoT and tactile Internet
ing network operations is the capability to efficiently collect and ana- applications. For instance, along this line, a recent work by Wong et al.
lyze telemetry information from the network. SDN and related stan- [146] has proposed a novel tactile Internet capable PON and a dynamic
dards have introduced streaming telemetry which can be used to effi- wavelength and bandwidth allocation method, which incorporates a
ciently collect information from the network all the way to the network mechanism to predict the traffic load to vary the number of active
edge. The scale of information collected in a large telco [136] poses wavelength channels in the network, and prioritize the transmission of
a challenge for processing and identifying anomalies such as network tactile Internet traffic (vs. other traffic) to comply with delay require-
mis-configurations, alarm-correlation, failure localization and predic- ments. Due to the huge expansion of IoT applications and services, we
tion. AI-based techniques can be efficiently used to target these prob- envision more advances to come along this line in the next years.
lems at scale, especially when combined with Big Data frameworks
like the Apache Hadoop [137] or Spark [138] ecosystems. Another 5.4. Applications in on-chip networks
key enabler for automating network management involves the use of
declarative intents which provide a syntax for users to describe their Finally, another promising field of application in our opinion is
requirements to network orchestration platforms [139,140]. Intents are that of on-chip networks. Today, major tech companies like Google,
widely employed in natural language processing (NLP) and AI-based Microsoft, Intel, etc. are pushing chip manufacturing to enable AI com-
techniques from these domains can be used to integrate user interac- puting on a single chip. On-chip networks provide communication sub-
tion platforms such as chatbots, voice command devices, etc., to inter- strates for the constantly increasing number of cores on a single chip.
pret user requests into intents for network orchestration platforms. As Generally, on-chip networks are designed to enable efficient comput-
discussed in Ref. [135] AI-based techniques can also be used to learn ing in less time while consuming less power. Though it is early days
and automate the process orchestration workflow for incoming intent in chip manufacturing to bring AI-enabled on-chip networks in market,
requests, while retaining the possibility of human intervention in the recently Google has come up with its own AI chip, known as Tensor Pro-

53
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

cessing Unit (TPU), which allows TensorFlow (a deep neural network ing a crucial role in supporting technologies like space division multi-
software) to run. Similarly, Intel announced an experimental chip called plexing, multimode/multicore fibers, advanced modulation formats and
“Loihi”, which is designed based on neuromorphic technology that uses constellation shaping, etc. Notwithstanding, the role that AI can play
neurons instead of logic gates. AI algorithms are known for time and in attack and intrusion detection in optical networks seems of special
computation complexity, thus requiring multiple central and graphical relevance and has not yet been significantly explored to the best of our
processing units, therefore AI-enabled optical network on-chip (ONoC) knowledge. Furthermore, the automation of network management oper-
is an alternative to the electronic NoC which will further reduce the ations, especially in the current context of networks becoming increas-
power consumption and computation time. For instance, Gu et al. [147] ingly heterogeneous, and the efficient joint operation of networks and
have recently proposed an optical network on-chip architecture by com- MEC/fog/cloud computing resources, together with on-chip networking
bining TDM and WDM technologies. This novel architecture allows to are, in our opinion, fields where AI can be decisive.
solve the blocking problem faced by previously proposed optical circuit-
switching (OCS) based ones [148]. Specifically, as an aspect of special Acknowledgment
interest for the subject of this article, the number of wavelength groups
and time slots is optimized by using a genetic algorithm, which helps This work has been partially supported by the Spanish Ministry
TDM-WDM-based ONoC to outperform results from equivalent propos- of Economy and Competitiveness (TEC2014-53071-C3-2-P, TEC2015-
als based on OCS-mesh. 71932-REDT).

6. Summary References

[1] S. Russell, P. Norvig, Artificial Intelligence: a Modern Approach, third ed.,

This paper has provided a comprehensive survey of the current
Prentice Hall Press, Upper Saddle River, NJ, USA, 2009.
research within the application of Artificial Intelligence (AI) techniques [2] B. Mukherjee, Optical WDM Networks, Springer-Verlag New York, Inc., Secaucus,
in optical networks, as well as an overview of opportunities and chal- NJ, USA, 2006.
lenges arising in this context. [3] J.M. Simmons, Optical Network Design and Planning, second ed., Springer
International Publishing, 2014.
In order to provide the reader with a clear and general vision of the [4] R.V. Rao, V.J. Savsani, D.P. Vakharia, Teaching-Learning-Based Optimization: an
numerous techniques and methods that make up this scientific disci- optimization method for continuous non-linear large scale problems, Inf. Sci. 183
pline, we have first described those AI subfields that have been success- (1) (2012) 1–15.
[5] K.-L. Du, M.N.S. Swamy, Search and Optimization by Metaheuristics, Springer
fully employed in optical networking: (1) search methods and optimiza- International Publishing, 2016.
tion theory, (2) game theory, (3) knowledge-based, reasoning and plan- [6] F. Martinelli, N. Andriolli, P. Castoldi, I. Cerutti, Genetic approach for optimizing
ning methods, (4) statistical models, (5) decision-making algorithms, the placement of all-optical regenerators in WSON, J. Opt. Commun. Netw. 6
(11) (2014) 1028–1037.
and (6) learning methods, and have classified relevant literature in the [7] M. Morgan, An ant colony approach to regenerator placement with fault
diagram shown in Fig. 1. tolerance in optical networks, in: 2015 7th International Workshop on Reliable
Later on, we have extensively reviewed the application of these tech- Networks Design and Modeling (RNDM), 2015, pp. 85–91.
[8] D. Monoyios, K. Vlachos, Multiobjective genetic algorithms for solving the
niques in order to improve the efficiency of both optical transmission impairment-aware routing and wavelength assignment problem, IEEE/OSA J.
(Table 1), and the design and control of optical networks (Table 2). Opt. Commun. Netw. 3 (1) (2011) 40–47.
Specifically, in relation to optical transmission, we have addressed the [9] X. Wang, M. Brandt-Pearce, S. Subramaniam, Distributed grooming, routing, and
wavelength assignment for dynamic optical networks using ant colony
suitability of AI techniques when dealing with the characterization and
optimization, IEEE/OSA J. Opt. Commun. Netw. 6 (6) (2014) 578–589.
operation of transmitters, EDFAs and receivers, as well as for perfor- [10] O. Awwad, A. Al-Fuqaha, B. Khan, D. Benhaddou, M. Guizani, A. Rayes,
mance monitoring, mitigation of nonlinearities, and for QoT estima- Bayesian-based game theoretic model to guarantee cooperativeness in hybrid
tion, which is particularly relevant in impairment-aware optical net- RF/FSO mesh networks, in: IEEE Global Telecommunications Conference, 2009.
GLOBECOM 2009, 2009, pp. 1–7.
work operation. In many cases, AI techniques have proven to be more [11] J. Zhu, B. Zhao, Z. Zhu, Leveraging game theory to achieve efficient attack-aware
efficient than classical approaches, which are typically limited by the service provisioning in EONs, J. Lightwave Technol. 35 (10) (2017) 1785–1796.
complexity, lack of adaptability, and/or scalability of the deterministic [12] L. Pavel, Game Theory for Control of Optical Networks, 2012. Birkhäuser Basel.
[13] I. de Miguel, R.J. Durán, T. Jiménez, N. Fernández, J.C. Aguado, R.M. Lorenzo,
or semi-analytical models on which they rely. AI techniques provide A. Caballero, I.T. Monroy, Y. Ye, A. Tymecki, et al., Cognitive dynamic optical
similar advantages in issues related to the control and design of opti- networks [invited], J. Opt. Commun. Netw. 5 (10) (2013) A107–A118.
cal networks. The thorough review of the state of the art in this matter [14] R. Borkowski, R.J. Duran, C. Kachris, D. Siracusa, A. Caballero, N. Fernandez, D.
Klonidis, A. Francescon, T. Jimenez, J.C. Aguado, I. de Miguel, E. Salvadori, I.
has included numerous applications in the following fields: optical net- Tomkos, R.M. Lorenzo, I.T. Monroy, Cognitive optical network testbed: EU
work planning, connection establishment (i.e., the pursuit of the opti- project CHRON, IEEE/OSA J. Opt. Commun. Netw. 7 (2) (2015) A344–A355.
mal solution for the RWA, RSA, RMLSA or RMCSA problems), network [15] G.S. Zervas, D. Simeonidou, Cognitive optical networks: need, requirements and
architecture, in: 2010 12th International Conference on Transparent Optical
reconfiguration, software-defined networking and applications in spe- Networks, 2010, pp. 1–4.
cific types of networks such as OBS networks, Passive Optical Networks [16] W. Wei, C. Wang, J. Yu, Cognitive optical networks: key drivers, enabling
and data centers. techniques, and adaptive bandwidth services, IEEE Commun. Mag. 50 (1) (2012)
106–113.
In general, AI techniques’ ability to find optimal or near-optimal
[17] S.J.B. Yoo, Multi-domain cognitive optical software defined networks with
solutions in highly complex scenarios (i.e., problems with very high market-driven brokers, in: 2014 The European Conference on Optical
dimensionality, fed with huge amounts of data) without the need to Communication (ECOC), 2014, pp. 1–3.
develop exhaustive analytical or semi-analytical models, makes them [18] V.W.S. Chan, E. Jang, Cognitive all-optical fiber network architecture, in: 2017
19th International Conference on Transparent Optical Networks (ICTON), 2017,
essential to meet the challenge posed by the increased complexity and pp. 1–4.
dynamism of current optical communication networks, which need to [19] S. Gosselin, J.L. Courant, S.R. Tembo, S. Vaton, Application of probabilistic
face, among other impairments, the physical limitations imposed by modeling and machine learning to the diagnosis of FTTH GPON networks, in:
2017 International Conference on Optical Network Design and Modeling
noise and nonlinear distortions. No less important is AI techniques’ abil- (ONDM), 2017, pp. 1–3.
ity to adapt to changing conditions, to learn from them and to propose [20] K. Chitra, M.R. Senkumar, Hidden Markov model based lightpath establishment
solutions to unexpected situations or cases, turning them into promis- technique for improving QoS in optical WDM networks, in: Second International
Conference on Current Trends in Engineering and Technology - ICCTET 2014,
ing candidates to, for example, reconfigure these dynamic networks in 2014, pp. 53–62.
short time scales in order to meet changing demand patterns. [21] B. Szafraniec, T.S. Marshall, B. Nebendahl, Performance monitoring and
Last but not least, we have described a number of opportunities measurement techniques for coherent optical systems, J. Lightwave Technol. 31
(4) (2013) 648–663.
and challenges that we envision in the area of optical networking. In
the context of optical transmission, we expect AI will continue play-

54
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

[22] M. Yannuzzi, X. Masip-Bruin, S. Sanchez-Lopez, E.M. Tordera, J. Sole-Pareta, J. [50] Y. Huang, P.B. Cho, P. Samadi, K. Bergman, Dynamic power pre-adjustments
Domingo-Pascual, Interdomain RWA based on stochastic estimation methods and with machine learning that mitigate EDFA excursions during defragmentation, in:
adaptive filtering for optical networks, in: IEEE Globecom 2006, 2006, pp. 1–6. Optical Fiber Communication Conference, Optical Society of America, 2017.
[23] A. Jayaraj, T. Venkatesh, C.S.R. Murthy, Loss classification in optical burst Th1J–2.
switching networks using machine learning techniques: improving the [51] E. d. A. Barboza, C.J. Bastos-Filho, J.F. Martins-Filho, U.C. de Moura, J.R. de
performance of TCP, IEEE J. Sel. Area. Commun. 26 (6) (2008) 45–54. Oliveira, Self-adaptive erbium-doped fiber amplifiers using machine learning, in:
[24] T. Tachibana, I. Koyanagi, MDP-based lightpath establishment for service Microwave & Optoelectronics Conference (IMOC), 2013 SBMO/IEEE MTT-s
differentiation in all-optical WDM networks with wavelength conversion International, IEEE, 2013, pp. 1–5.
capability, Photonic Netw. Commun. 20 (2) (2010) 183–192. [52] T. Tanimura, T. Hoshida, J.C. Rasmussen, M. Suzuki, H. Morikawa, OSNR
[25] C. Sue, Y. Hsu, P. Ho, Dynamic preemption call admission control scheme based monitoring by deep neural networks trained with asynchronously sampled data,
on Markov decision process in traffic groomed optical networks, IEEE/OSA J. in: 2016 21st OptoElectronics and Communications Conference (OECC) Held
Opt. Commun. Netw. 3 (4) (2011) 300–311. Jointly with 2016 International Conference on Photonics in Switching (PS), IEEE,
[26] R.R. Reyes, T. Bauschert, Adaptive and state-dependent online resource 2016, pp. 1–3.
allocation in dynamic optical networks, IEEE/OSA J. Opt. Commun. Netw. 9 (3) [53] J. Thrane, J. Wass, M. Piels, J.C. Diniz, R. Jones, D. Zibar, Machine learning
(2017) B64–B77. techniques for optical performance monitoring from directly detected PDM-QAM
[27] Z. Ghahramani, Probabilistic machine learning and artificial intelligence, Nature signals, J. Lightwave Technol. 35 (4) (2017) 868–875.
521 (7553) (2015) 452. [54] A.P.T. Lau, J.M. Kahn, Signal design and detection in presence of nonlinear phase
[28] K.P. Murphy, Machine Learning: a Probabilistic Perspective, The MIT Press, noise, J. Lightwave Technol. 25 (10) (2007) 3008–3016.
2012. [55] D. Zibar, M. Piels, R. Jones, C.G. Schäeffer, Machine learning techniques in
[29] C.M. Bishop, Pattern Recognition and Machine Learning, Springer, 2006. optical communication, J. Lightwave Technol. 34 (6) (2016) 1442–1452.
[30] M.G. Taylor, Phase estimation methods for optical coherent detection using [56] D. Wang, M. Zhang, Z. Li, Y. Cui, J. Liu, Y. Yang, H. Wang, Nonlinear decision
digital signal processing, J. Lightwave Technol. 27 (7) (2009) 901–914. boundary created by a machine learning-based classifier to mitigate nonlinear
[31] F. Karinou, N. Stojanovic, C. Prodaniuc, M. Agustin, J. Kropp, N.N. Ledentsov, phase noise, in: Optical Communication (ECOC), 2015 European Conference,
Solutions for 100/400-Gb/s ethernet systems based on multimode photonic IEEE, 2015, pp. 1–3.
technologies, J. Lightwave Technol. 35 (15) (2017) 3214–3222. [57] D. Wang, M. Zhang, M. Fu, Z. Cai, Z. Li, Y. Cui, B. Luo, KNN-based detector for
[32] F. Rottenberg, T.H. Nguyen, S.P. Gorza, F. Horlin, J. Louveaux, ML and MAP coherent optical systems in presence of nonlinear phase noise, in: 2016 21st
phase noise estimators for optical fiber FBMC-OQAM systems, in: 2017 IEEE OptoElectronics and Communications Conference (OECC) Held Jointly with 2016
International Conference on Communications (ICC), 2017, pp. 1–6. International Conference on Photonics in Switching (PS), IEEE, 2016, pp. 1–3.
[33] X. Wu, J.A. Jargon, R.A. Skoog, L. Paraschis, A.E. Willner, Applications of [58] E. Giacoumidis, J. Wei, S. Mhatli, M.F. Stephens, N.J. Doran, A.D. Ellis, B.J.
artificial neural networks in optical performance monitoring, J. Lightwave Eggleton, Nonlinear inter-subcarrier intermixing reduction in coherent optical
Technol. 27 (16) (2009) 3580–3589. OFDM using fast machine learning equalization, in: Optical Fiber
[34] T. Jiménez, J.C. Aguado, I. de Miguel, R.J. Durán, N. Fernandez, M. Angelou, D. Communications Conference and Exhibition (OFC), 2017, IEEE, 2017, pp. 1–3.
Sánchez, N. Merayo, P. Fernández, N. Atallah, et al., A cognitive system for fast [59] J. Mata, I. de Miguel, R.J. Durán, J.C. Aguado, N. Merayo, L. Ruiz, P. Fernández,
quality of transmission estimation in core optical networks, in: Optical Fiber R.M. Lorenzo, E.J. Abril, A SVM approach for lightpath QoT estimation in optical
Communication Conference and Exposition (OFC/NFOEC), 2012 and the transport networks, in: 2017 IEEE International Conference on Big Data (IEEE
National Fiber Optic Engineers Conference, IEEE, 2012, pp. 1–3. BigData), 2017, pp. 4713–4715.
[35] T. Jiménez, J.C. Aguado, I. de Miguel, R.J. Durán, M. Angelou, N. Merayo, P. [60] T. Duthel, G. Clarici, C.R. Fludger, J.C. Geyer, C. Schulien, S. Wiese, Laser
Fernández, R.M. Lorenzo, I. Tomkos, E.J. Abril, A cognitive quality of linewidth estimation by means of coherent detection, IEEE Photon. Technol. Lett.
transmission estimator for core optical networks, J. Lightwave Technol. 31 (6) 21 (20) (2009) 1568–1570.
(2013) 942–951. [61] K. Kikuchi, Characterization of semiconductor-laser phase noise and estimation of
[36] A. Caballero, J.C. Aguado, R. Borkowski, S. Saldaña, T. Jiménez, I. de Miguel, V. bit-error rate performance with low-speed offline digital coherent receivers, Optic
Arlunno, R.J. Durán, D. Zibar, J.B. Jensen, et al., Experimental demonstration of Express 20 (5) (2012) 5291–5302.
a cognitive quality of transmission estimator for optical communication systems, [62] Z. Dong, F.N. Khan, Q. Sui, K. Zhong, C. Lu, A.P.T. Lau, Optical performance
Optic Express 20 (26) (2012) B64–B70. monitoring: a review of current and future technologies, J. Lightwave Technol.
[37] L. Barletta, A. Giusti, C. Rottondi, M. Tornatore, QoT estimation for unestablished 34 (2) (2016) 525–543.
lighpaths using machine learning, in: Optical Fiber Communications Conference [63] Z. Li, Z. Jian, L. Cheng, Y. Yang, C. Lu, A.P.T. Lau, C. Yu, H. Tam, P. Wai, Signed
and Exhibition (OFC), 2017, IEEE, 2017, pp. 1–3. chromatic dispersion monitoring of 100Gbit/s CS-RZ DQPSK signal by evaluating
[38] S. Oda, M. Miyabe, S. Yoshida, T. Katagiri, Y. Aoki, T. Hoshida, J.C. Rasmussen, the asymmetry ratio of delay tap sampling, Optic Express 18 (3) (2010)
M. Birk, K. Tse, A learning living network with open ROADMS, J. Lightwave 3149–3157.
Technol. 35 (8) (2017) 1350–1356. [64] F. Khan, A.P.T. Lau, Z. Li, C. Lu, P. Wai, OSNR monitoring for RZ-DQPSK systems
[39] H. Rastegarfar, M. Glick, N. Viljoen, M. Yang, J. Wissinger, L. Lacomb, N. using half-symbol delay-tap sampling technique, IEEE Photon. Technol. Lett. 22
Peyghambarian, TCP flow classification and bandwidth aggregation in optically (11) (2010) 823–825.
interconnected data center networks, IEEE/OSA J. Opt. Commun. Netw. 8 (10) [65] A. Napoli, Z. Maalej, V.A. Sleiffer, M. Kuschnerov, D. Rafique, E. Timmers, B.
(2016) 777–786. Spinnler, T. Rahman, L.D. Coelho, N. Hanik, Reduced complexity digital
[40] N.G. Gonzalez, D. Zibar, I.T. Monroy, Cognitive digital receiver for burst mode back-propagation methods for optical communication systems, J. Lightwave
phase modulated radio over fiber links, in: Optical Communication (ECOC), 2010 Technol. 32 (7) (2014) 1351–1362.
36th European Conference and Exhibition, IEEE, 2010, pp. 1–3. [66] N.V. Irukulapati, D. Marsella, P. Johannisson, E. Agrell, M. Secondini, H.
[41] M.C. Tan, F.N. Khan, W.H. Al-Arashi, Y. Zhou, A.P.T. Lau, Simultaneous optical Wymeersch, Stochastic digital backpropagation with residual memory
performance monitoring and modulation format/bit-rate identification using compensation, J. Lightwave Technol. 34 (2) (2016) 566–572.
principal component analysis, J. Opt. Commun. Netw. 6 (5) (2014) 441–448. [67] S. Azodolmolky, J. Perelló, M. Angelou, F. Agraz, L. Velasco, S. Spadaro, Y.
[42] J.J.G. Torres, A. Chiuchiarelli, V.A. Thomas, S.E. Ralph, A.M.C. Soto, N.G. Pointurier, A. Francescon, C.V. Saradhi, P. Kokkinos, et al., Experimental
González, Adaptive nonsymmetrical demodulation based on machine learning to demonstration of an impairment aware network planning and operation tool for
mitigate time-varying impairments, in: Avionics and Vehicle Fiber-optics and transparent/translucent optical networks, J. Lightwave Technol. 29 (4) (2011)
Photonics Conference (AVFOP), 2016 IEEE, IEEE, 2016, pp. 289–290. 439–448.
[43] L.P. Kaelbling, M.L. Littman, A.W. Moore, Reinforcement learning: a survey, J. [68] A. Aamodt, E. Plaza, Case-based reasoning: foundational issues, methodological
Artif. Intell. Res. 4 (1996) 237–285. variations, and system approaches, AI Commun. 7 (1) (1994) 39–59.
[44] C.J. Watkins, P. Dayan, Q-learning. Mach. learn. 8 (3–4) (1992) 279–292. [69] R.M. Morais, C. Pavan, A.N. Pinto, C. Requejo, Genetic algorithm for the
[45] Y.V. Kiran, T. Venkatesh, C.S.R. Murthy, A reinforcement learning framework for topological design of survivable optical transport networks, J. Opt. Commun.
path selection and wavelength selection in optical burst switched networks, IEEE Netw. 3 (1) (2011) 17–26.
J. Sel. Area. Commun. 25 (9) (2007) 18–26. [70] I. Cerutti, F. Martinelli, N. Sambo, F. Cugini, P. Castoldi, Trading regeneration
[46] D. Zibar, L.H.H. de Carvalho, M. Piels, A. Doberstein, J. Diniz, B. Nebendahl, C. and spectrum utilization in code-rate adaptive flexi-grid networks, J. Lightwave
Franciscangelis, J. Estaran, H. Haisch, N.G. Gonzalez, et al., Application of Technol. 32 (23) (2014) 4496–4503.
machine learning techniques for amplitude and phase noise characterization, J. [71] L. Velasco, P. Wright, A. Lord, G. Junyent, Saving CAPEX by extending
Lightwave Technol. 33 (7) (2015) 1333–1343. flexgrid-based core optical networks toward the edges [invited], IEEE/OSA J.
[47] A. Hraghi, M.E. Chaibi, M. Menif, D. Erasme, Demonstration of 16QAM-OFDM Opt. Commun. Netw. 5 (10) (2013) A171–A183.
UDWDM transmission using a tunable optical flat comb source, J. Lightwave [72] I. de Miguel, R. Vallejos, A. Beghelli, R.J. Durán, Genetic algorithm for joint
Technol. 35 (2) (2017) 238–245. routing and dimensioning of dynamic WDM networks, IEEE/OSA J. Opt.
[48] S.L. Brunton, X. Fu, J.N. Kutz, Self-tuning fiber lasers, IEEE J. Sel. Top. Quant. Commun. Netw. 1 (7) (2009) 608–621.
Electron. 20 (5) (2014) 464–471. [73] F.R. Durand, T. Abro, Energy-efficient power allocation for WDM/OCDM
[49] Y. Huang, W. Samoud, C.L. Gutterman, C. Ware, M. Lourdiane, G. Zussman, P. networks with particle swarm optimization, IEEE/OSA J. Opt. Commun. Netw. 5
Samadi, K. Bergman, A machine learning approach for dynamic optical channel (5) (2013) 512–523.
add/drop strategies that minimize EDFA power excursions, in: Proceedings of [74] M. de Paula Marques, F.R. Durand, T. Abro, WDM/OCDM energy-efficient
ECOC 2016; 42nd European Conference on Optical Communication, VDE, 2016, networks based on heuristic ant colony optimization, IEEE Syst. J. 10 (4) (2016)
pp. 1–3. 1482–1493.

55
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

[75] S.G. Petridou, P.G. Sarigiannidis, G.I. Papadimitriou, A.S. Pomportsis, On the use [100] J. Oliveira, J. Oliveira, E. Magalhes, J. Janurio, M. Siqueira, R. Scaraficci, M.
of clustering algorithms for message scheduling in WDM star networks, J. Salvador, L. Mariote, N. Guerrero, L. Carvalho, F. v. Hooft, G. Santos, M. Garrich,
Lightwave Technol. 26 (17) (2008) 2999–3010. Toward terabit autonomic optical networks based on a software defined
[76] Á. Rubio-Largo, M.A. Vega-Rodriguez, J.A. Gomez-Pulido, J.M. Sanchez-Pérez, A adaptive/cognitive approach [invited], IEEE/OSA J. Opt. Commun. Netw. 7 (3)
comparative study on multiobjective swarm intelligence for the routing and (2015) A421–A431.
wavelength assignment problem, IEEE Trans. Syst. Man Cybern. C Appl. Rev. 42 [101] S.J.B. Yoo, Multi-domain cognitive optical software defined networks with
(6) (2012) 1644–1655. market-driven brokers, in: 2014 The European Conference on Optical
[77] L. Gong, X. Zhou, W. Lu, Z. Zhu, A two-population based evolutionary approach Communication (ECOC), 2014, pp. 1–3.
for optimizing routing, modulation and spectrum assignments (RMSA) in [102] S. Yan, N. Khan, A. Mavromatis, D. Gkounis, Q. Fan, F. Ntavou, K. Nikolovgenis,
O-OFDM networks, IEEE Commun. Lett. 16 (9) (2012) 1520–1523. F. Meng, E. Salas, C. Guo, C. Lu, A. Lau, R. Nejabati, D. Simeonidou, Field trial of
[78] P. Lechowicz, K. Walkowiak, Genetic algorithm for routing and spectrum machine-learning-assisted and SDN-based optical network planning with
allocation in elastic optical networks, in: Network Intelligence Conference network-scale monitoring database, in: 43rd European Conference on Optical
(ENIC), 2016 Third European, IEEE, 2016, pp. 273–280. Communication (ECOC 2017), 2017.
[79] P. Wright, M.C. Parker, A. Lord, Minimum- and maximum-entropy routing and [103] J. Praveen, B. Praveen, T. Venkatesh, Y.V. Kiran, C.S.R. Murthy, A first step
spectrum assignment for flexgrid elastic optical networking [invited], IEEE/OSA toward autonomic optical burst switched networks, IEEE J. Sel. Area. Commun.
J. Opt. Commun. Netw. 7 (1) (2015) A66–A72. 24 (12) (2006) 94–105.
[80] M. Przewoniczek, R. Gocie, K. Walkowiak, M. Klinkowski, Towards solving [104] T. Venkatesh, A. Sankar, A. Jayaraj, C.S.R. Murthy, A complete framework to
practical problems of large solution space using a novel pattern searching hybrid support controlled burst retransmission in optical burst switching networks, IEEE
evolutionary algorithm an elastic optical network optimization case study, Expert J. Sel. Area. Commun. 26 (3) (2008) 65–73.
Syst. Appl. 42 (21) (2015) 7781–7796. [105] S. Haeri, W.W.K. Thong, G. Chen, L. Trajkovi, A reinforcement learning-based
[81] C.A. Kyriakopoulos, G.I. Papadimitriou, P. Nicopolitidis, E. Varvarigos, algorithm for deflection routing in optical burst-switched networks, in: 2013
Energy-efficient lightpath establishment in backbone optical networks based on IEEE 14th International Conference on Information Reuse Integration (IRI), 2013,
ant colony optimization, J. Lightwave Technol. 34 (23) (2016) 5534–5541. pp. 474–481.
[82] Y. Sun, J. Yuan, M. Zhai, An optimized RWA method with services selection [106] H. Elbiaze, Cognitive mechanisms for providing QoS in OBS networks, in: 2011
based on ACO in optical network, in: 2016 15th International Conference on 13th International Conference on Transparent Optical Networks, 2011, pp. 1–4.
Optical Communications and Networks (ICOCN), 2016, pp. 1–3. [107] S. Haeri, L. Trajkovi, Intelligent deflection routing in buffer-less networks, IEEE
[83] Z. Chen, S. Wang, H. Zhang, Y. Liu, Y. Peng, Cognitive routing and wavelength Trans. Cybern. 45 (2) (2015) 316–327.
assignment algorithm for dynamic optical networks, in: 2014 12th International [108] H.C. Leung, C.S. Leung, E.W.M. Wong, S. Li, Extreme learning machine for
Conference on Optical Internet 2014 (COIN), 2014, pp. 1–3. estimating blocking probability of bufferless OBS/OPS networks, IEEE/OSA J.
[84] K. Christodoulopoulos, I. Tomkos, E. Varvarigos, Elastic bandwidth allocation in Opt. Commun. Netw. 9 (8) (2017) 682–692.
flexible OFDM-based optical networks, J. Lightwave Technol. 29 (9) (2011) [109] Y. Coulibaly, G. Rouskas, M.S. Abd Latiff, M.A. Razzaque, S. Mandala, QoS-aware
1354–1366. ant-based route, wavelength and timeslot assignment algorithm for optical burst
[85] J. Perelló, J.M. Gené, A. Pagès, J.A. Lazaro, S. Spadaro, Flex-grid/SDM backbone switchednetworks, Trans. Emerg. Telecommun. Technol. 26 (11) (2015)
network design with inter-core XT-limited transmission reach, IEEE/OSA J. Opt. 1265–1277.
Commun. Netw. 8 (8) (2016) 540–552. [110] T. V. y. Villalba, S.M. Rossi, M.P. Mokarzel, M.R. Salvador, H.M.A. Neto, A.C.
[86] K. Walkowiak, M. Klinkowski, B. Rabiega, R. Gocie, Routing and spectrum Cesar, M.A. Romero, M. d. L. Rocha, Design of passive optical networks using
allocation algorithms for elastic optical networks with dedicated path protection, genetic algorithm, in: 2009 SBMO/IEEE MTT-s International Microwave and
Opt. Switch. Netw. 13 (Supplement C) (2014) 63–75. Optoelectronics Conference (IMOC), 2009, pp. 682–686.
[87] R. Goścień, K. Walkowiak, M. Klinkowski, Tabu search algorithm for routing, [111] A.A.A. Kokangul, Optimization of passive optical network planning, Appl. Math.
modulation and spectrum allocation in elastic optical network with anycast and Model. 35 (7) (2011) 3345–3354.
unicast traffic, Comput. Network. 79 (Supplement C) (2015) 148–165. [112] X. Liu, R. Gu, Y. Ji, A nested genetic algorithm for topologyoptimization in
[88] W.B. Jia, Z.Q. Xu, Z. Ding, K. Wang, An efficient routing and spectrum LRPON, in: 2012 3rd IEEE International Conference on Network Infrastructure
assignment algorithm using prediction for elastic optical networks, in: 2016 and Digital Content, 2012, pp. 217–221.
International Conference on Information System and Artificial Intelligence (ISAI), [113] C.T. Huang, The study of balance traffic load with genetic algorithm for PON, in:
2016, pp. 89–93. 2008 International Conference on Intelligent Information Hiding and Multimedia
[89] D.R.B. de Araujo, C.J.A. Bastos-filho, J.F. Martins-filho, Methodology to obtain a Signal Processing, 2008, pp. 39–42.
fast and accurate estimator for blocking probability of optical networks, [114] N. Moradpoor, G. Parr, S. McClean, B. Scotney, A mathematical model for a
IEEE/OSA J. Opt. Commun. Netw. 7 (5) (2015) 380–391. GA-based dynamic excess bandwidth allocation algorithm for hybrid PON and
[90] F.C. Ergin, E. Kaldirim, A. Yayimli, A.S. Uyar, Ensuring resilience in optical WDM wireless technology integrations for next generation broadband access networks,
networks with nature-inspired heuristics, IEEE/OSA J. Opt. Commun. Netw. 2 (8) in: 2013 5th Computer Science and Electronic Engineering Conference (CEEC),
(2010) 642–652. 2013, pp. 23–28.
[91] N. Fernández, R.J. Durán, I. de Miguel, N. Merayo, J.C. Aguado, P. Fernández, T. [115] T. Jiménez, N. Merayo, A. Andrés, R. Durán, J. Aguado, I. de Miguel, P.
Jiménez, I. Rodríguez, D. Sánchez, R.M. Lorenzo, et al., Survivable and Fernández, R. Lorenzo, E. Abril, An auto-tuning PID control system based on
impairment-aware virtual topologies for reconfigurable optical networks: a genetic algorithms to provide delay guarantees in passive optical networks,
cognitive approach, in: 4th International Congress on Ultra Modern Expert Syst. Appl. 42 (23) (2015) 9211–9220.
Telecommunications and Control Systems and Workshops (ICUMT 2012), IEEE, [116] I.S. Hwang, J.Y. Lee, A.T. Liem, Genetic expression programming: a new
2012, pp. 793–799. approach for QoS traffic prediction in EPONs, in: 2012 Fourth International
[92] N. Fernández, R.J. Durán, E. Palkopoulou, I. de Miguel, I. Stiakogiannakis, N. Conference on Ubiquitous and Future Networks (ICUFN), 2012, pp. 249–254.
Merayo, I. Tomkos, R.M. Lorenzo, Techno-economic advantages of cognitive [117] N. Merayo, D. Juárez, J.C. Aguado, I. de Miguel, R.J. Durán, P. Fernández, R.M.
virtual topology design, in: 39th European Conference and Exhibition on Optical Lorenzo, E.J. Abril, PID controller based on a self-adaptive neural network to
Communication (ECOC 2013), IET, 2013, pp. 1–3. ensure QoS bandwidth requirements in passive optical networks, IEEE/OSA J.
[93] X. Gao, Z. Ye, J. Fan, W. Zhong, Y. Zhao, X. Cao, H. Yu, C. Qiao, Virtual network Opt. Commun. Netw. 9 (5) (2017) 433–445.
mapping for multicast services with max-min fairness of reliability, IEEE/OSA J. [118] M.P.I. Dias, B.S. Karunaratne, E. Wong, Bayesian estimation and prediction-based
Opt. Commun. Netw. 7 (9) (2015) 942–951. dynamic bandwidth allocation algorithm for sleep/doze-mode passive optical
[94] N. Fernández, R.J.D. Barroso, D. Siracusa, A. Francescon, I. de Miguel, E. networks, J. Lightwave Technol. 32 (14) (2014) 2560–2568.
Salvadori, J.C. Aguado, R.M. Lorenzo, Virtual topology reconfiguration in optical [119] U.R. Bhatt, N. Chouhan, R. Upadhyay, C. Agrawal, ONU placement in
networks by means of cognition: evaluation and experimental validation, J. Opt. fiber-wireless (FiWi) access networks using teacher phase of teaching learning
Commun. Netw. 7 (1) (2015) A162–A173. based optimization (TLBO) algorithm, in: 2017 3rd International Conference on
[95] F. Morales, M. Ruiz, L. Gifre, L.M. Contreras, V. López, L. Velasco, Virtual Computational Intelligence Communication Technology (CICT), 2017, pp. 1–4.
network topology adaptability based on data analytics for traffic prediction, J. [120] U.R. Bhatt, N. Chouhan, R. Upadhyay, Hybrid algorithm: a cost efficient solution
Opt. Commun. Netw. 9 (1) (2017) A35–A45. for ONU placement in Fiber-Wireless (FiWi) network, Opt. Fiber Technol. 22
[96] M. Ruiz, F. Fresi, A.P. Vela, G. Meloni, N. Sambo, F. Cugini, L. Poti, L. Velasco, P. (Supplement C) (2015) 76–83.
Castoldi, Service-triggered failure identification/localization through monitoring [121] M. Glick, H. Rastegarfar, Scheduling and control in hybrid data centers, in: 2017
of multiple parameters, in: Proceedings of ECOC 2016; 42nd European IEEE Photonics Society Summer Topical Meeting Series (SUM), 2017, pp.
Conference on Optical Communication, VDE, 2016, pp. 1–3. 115–116.
[97] X. Zhang, W. Hou, L. Guo, S. Wang, Y. Sun, X. Yang, Failure recovery solutions [122] C.H. Wang, T. Javidi, Adaptive policies for scheduling with reconfiguration
using cognitive mechanisms for software defined optical networks, in: 2016 15th delay: an end-to-end solution for all-optical data centers, IEEE/ACM Trans. Netw.
International Conference on Optical Communications and Networks (ICOCN), 25 (3) (2017) 1555–1568.
2016, pp. 1–3. [123] M. Klinkowski, K. Walkowiak, Routing and spectrum assignment in spectrum
[98] S.R. Tembo, S. Vaton, J.L. Courant, S. Gosselin, A tutorial on the EM algorithm sliced elastic optical path network, IEEE Commun. Lett. 15 (8) (2011) 884–886.
for Bayesian networks: application to self-diagnosis of GPON-FTTH networks, in: [124] M. Aibin, K. Walkowiak, Simulated Annealing algorithm for optimization of
2016 International Wireless Communications and Mobile Computing Conference elastic optical networks with unicast and anycast traffic, in: 2014 16th
(IWCMC), 2016, pp. 369–376. International Conference on Transparent Optical Networks (ICTON), 2014, pp.
[99] R. Montero, F. Agraz, A. Pagès, J. Perelló, S. Spadaro, Dynamic topology 1–4.
discovery in SDN-enabled transparent optical networks, in: 2017 International [125] W. Xia, Y. Wen, C.H. Foh, D. Niyato, H. Xie, A survey on software-defined
Conference on Optical Network Design and Modeling (ONDM), 2017, pp. 1–6. networking, Commun. Surv. Tutorials, IEEE 17 (1) (2015) 27–51.

56
J. Mata et al. Optical Switching and Networking 28 (2018) 43–57

[126] P. Sarigiannidis, A. Sarigiannidis, I. Moscholios, P. Zwierzykowski, “DIANA: A [137] Apache Hadoop. URL http://hadoop.apache.org.
Machine Learning Mechanism for Adjusting the TDD Uplink-Downlink [138] Apache Spark. URL https://spark.apache.org.
Configuration in XG-PON-LTE Systems”, Mobile Inform Syst., 2017 (2017), [139] Open Network Operating System (ONOS). URL http://onosproject.org/.
Article ID 8198017, 15 pages. [140] The Linux Foundation, OpenDaylight. URL https://www.opendaylight.org/.
[127] IBM, Cost of data breach study, (Accessed on 01.09. 2017). URL http://www-03. [141] D. Siracusa, F. Pederzolli, E. Salvadori, R.L. Cigno, I.T. Monroy, Proactive
ibm.com/security/data-breach/. restoration of slow-failures in optical networks, in: 2014 16th International
[128] N. Skorin-Kapov, M. Furdek, S. Zsigmond, L. Wosinska, Physical-layer security in Conference on Transparent Optical Networks (ICTON), 2014, pp. 1–4.
evolving optical networks, IEEE Commun. Mag. 54 (8) (2016) 110–117. [142] N. Fernández, R.J. Durán, I. de Miguel, J.C. Aguado, N. Merayo, R.M. Lorenzo, D.
[129] M.P. Fok, Z. Wang, Y. Deng, P.R. Prucnal, Optical layer security in fiber-optic Siracusa, A. Francescon, E. Salvadori, Demonstration of proactive restoration in
networks, IEEE Trans. Inf. Forensics Secur. 6 (3) (2011) 725–736. cognitive heterogeneous reconfigurable optical networks, in: 10th International
[130] M. Medard, D. Marquis, R.A. Barry, S.G. Finn, Security issues in all-optical Conference on Heterogeneous Networking for Quality, Reliability, Security and
networks, IEEE Netw. 11 (3) (1997) 42–48. Robustness, 2014, pp. 131–132.
[131] R. Rejeb, I. Pavlosoglou, M.S. Leeson, R.J. Green, Securing all-optical networks, [143] A.P. Vela, M. Ruiz, F. Fresi, N. Sambo, F. Cugini, G. Meloni, L. Pot, L. Velasco, P.
in: Proceedings of 2003 5th International Conference on Transparent Optical Castoldi, BER degradation detection and failure identification in elastic optical
Networks, 2003, vol. 1, 2003, pp. 87–90. networks, J. Lightwave Technol. 35 (21) (2017) 4595–4604.
[132] M. Medard, S.R. Chinn, P. Saengudomlert, Attack detection in all-optical [144] A.P. Vela, M. Ruiz, F. Cugini, L. Velasco, Combining a machine learning and
networks, in: Optical Fiber Communication Conference and Exhibit, 1998. OFC optimization for early pre-FEC BER degradation to meet committed QoS, in: 2017
’98, 1998, pp. 272–273. Technical Digest. 19th International Conference on Transparent Optical Networks (ICTON), 2017,
[133] S. Thomas, AI Key to Telcos’ Digital Transformation Survey, 2017. http://www. pp. 1–4.
lightreading.com/artificial-intelligence-machine-learning/ai-key-to-telcos- [145] M. Maier, M. Chowdhury, B.P. Rimal, D.P. Van, The Tactile Internet: vision,
digital-transformation-survey-/d/d-id/733482. recent progress, and open challenges, IEEE Commun. Mag. 54 (5) (2016)
[134] K. Kompella, The Self Driving Network: How to Realize it, 2017. https:// 138–145.
datatracker.ietf.org/meeting/98/materials/slides-98-nmrg-self-driving- [146] E. Wong, M.P.I. Dias, L. Ruan, Predictive resource allocation for tactile internet
networks/. capable passive optical LANs, J. Lightwave Technol. 35 (13) (2017) 2629–2641.
[135] A. Jukan, M. Chamania, Evolution towards smart optical networking: where [147] H. Gu, Z. Wang, B. Zhang, Y. Yang, K. Wang, Time-division-multiplexing-
artificial intelligence (AI) meets the world of photonics, in: Photonic Networks wavelength- division-multiplexing-based architecture for ONoC, IEEE/OSA J.
and Devices, Optical Society of America, 2017. NeM2B–4. Opt. Commun. Netw. 9 (5) (2017) 351–363.
[136] D. King, C. Rotsos, A. Aguado, L. Velasco, N. Georgalas, Network-based telemetry [148] G. Hendry, J. Chan, S. Kamil, L. Oliker, J. Shalf, L.P. Carloni, K. Bergman, Silicon
to facilitate the programmable management plane for optical transport nanophotonic network-on-chip using TDM arbitration, in: 2010 18th IEEE
infrastructure, in: 2016 18th International Conference on Transparent Optical Symposium on High Performance Interconnects, 2010, pp. 88–95.
Networks (ICTON), 2016. pp. 1–1.

57

View publication stats