Self-Organizing Networks in the 6G Era:

State-of-the-Art, Opportunities, Challenges, and

Future Trends
Abdelaali Chaoub, Senior Member, IEEE, Aarne Mämmelä, Senior Member, IEEE,
Pedro Martinez-Julia, Member, IEEE, Ranganai Chaparadza, Muslim Elkotob, Member, IEEE,
Lyndon Ong, Member, IEEE, Dilip Krishnaswamy, Senior Member, IEEE
arXiv:2112.09769v1 [cs.NI] 17 Dec 2021

Abstract—Self-organizing networks (SONs) need to be en- such as auto-discovery, have matured under the umbrella of
dowed with self-coordination capabilities to manage the complex self-managed systems and networks. Within them we find the
relations between their internal components and to avoid their inception of the well-known self-organizing networks (SONs),
destructive interactions. Existing communication technologies
commonly implement responsive self-coordination mechanisms which brings all self-* features to the network.
that can be very slow in run-time situations. The sixth gen- Self-organization [1], [2] is a general term that covers any
eration (6G) networks, being in their early stages of research kind of autonomous restructuring of a system and is, therefore,
and standardization activities, open new opportunities to opt the highest and most general in the hierarchy of all technical
for a design-driven approach when developing self-coordination systems. Thus, the main objective of SONs in communication
capabilities. This can be achieved through the use of hybrid
weakly coupled SON designs. In this article, we review the history systems is to reduce capital expenditures (CAPEX) and opera-
of SONs including the inherent self-coordination feature. We tional expenditures (OPEX) through automating as many fea-
then delve into the concept of hybrid SONs (H-SONs), and we tures as possible, spanning multiple network components and
summarize the challenges, opportunities, and future trends for various engineering phases. SONs will therefore continue to
H-SON development. We provide a comprehensive collection of empower future communication systems that must offer certain
standardization activities and recommendations, discussing the
key contributions and potential work to continue the evolution fundamental properties, i.e., functionality, stability, scalability,
and push for a wide adoption of the H-SON paradigm. More performance, dependability, security, cost-effectiveness, and
importantly, we propose that H-SONs must be weakly coupled resilience [3]. For instance, features such as densification or
networks, i.e., the various feedback loops must be almost isolated support of multiple radio access technology call for smoothly
from each other to improve the stability and to avoid chaotic scalable SONs [4]. On the other hand, the SON architecture is
situations. We finally conclude the paper with the key hints about
the future landscape and the key drivers of 6G H-SONs. decomposed into a set of smaller functional units referred to
as SON functions (SONFs). Since 5G and beyond networks
Index Terms—6G; self-organization; self-coordination; conflict are characterized by high system dynamics, SONFs will be
avoidance and resolution; weakly coupled system; hybrid self-
organizing network (H-SON). increasingly triggered and, consequently, they may interact
either constructively or destructively. Thereby, advanced self-
coordination capabilities are needed to ensure a conflict-free
I. I NTRODUCTION operation. Surprisingly, this feature has attracted very little
The road to 6G networks has recently begun. It continues research and standardization efforts. Most self-coordination
with the evolution of the capabilities of the network, with some principles are usually lightly covered [5], left open for future
emphasis on the convergence of fixed and wireless networks. research attention [4] or not considered at the design phase of
A key problem is being able to manage and control networks, new communication systems [6].
which are increasing in complexity and heterogeneity, and at To the best of our knowledge, the paper Bayazeed et al. [6]
the same time reducing the need for manual control. This prob- is the first attempt to provide a detailed survey about the
lem is not totally new since it was already targeted during the self-coordination functionality in cellular networks. It starts
development of 5G and its predecessors. Over the years, the by providing an historical overview of SONs after the 2000s,
standards that enable the design of functions for implementing i.e., starting from 3G networks. Then, it introduces a high-
self-* features, such as self-organization, self-configuration, level comprehensive framework to categorize self-coordination
self-diagnosis, self-repair, self-healing, self-optimization, self- logics into protective, reactive, and proactive. According to [6],
awareness, self-adaptation, and other automated operations protective methods that anticipate conflicts since the design
stage are only valid for static situations whereas system dy-
Abdelaali Chaoub is with the National Institute of Posts and Telecom- namics call for proactive methods that can predict the potential
munications (INPT), Morocco (email: chaoub.abdelaali@gmail.com). Pedro conflicts at the execution time using artificial intelligence (AI)
Martinez-Julia is with the National Institute of Information and Commu-
nications Technology (NICT), Japan (email: pedro@nict.go.jp). Ranganai and machine learning (ML).
Chaparadza is with Capgemini Engineering & IPv6 Forum, Germany (email: The present paper complements the work of Fourati et
ran4chap@yahoo.com). Muslim Elkotob is with Vodafone, Germany (email: al. [2] and Bayazeed et al. [6] by providing a comprehensive
muslim.elkotob@vodafone.com). Lyndon Ong is with Ciena, USA (email:
lyong@ciena.com). Dilip Krishnaswamy is with Sterlite Tech, USA (email: overview of the self-organization history from the late 1940s.
dilip@ieee.org). We then emphasize the importance of hierarchy with different
time scales, decoupling of different hierarchy levels (i.e., time Internet that was originally called the Arpanet and is based on
scale separation) or between subsystems at the same hierarchy packet switching. The Transmission Control Protocol (TCP)
level (i.e., separation using orthogonality), multiple degrees of and the Internet Protocol (IP) formed the present TCP/IP
centralization, and negotiation between various system parts as model, and the name of the Arpanet was changed to the
promising design paradigms to build the components of SONs Internet. At the same time, the generalized Open Systems
and achieve a conflict-free behavior by design as opposed Interconnection (OSI) model was developed.
to reactive and proactive methods that can be too slow in Ad hoc or unstructured networks are multi-hop SONs with-
dynamic situations. We particularly delve into the definition out any fixed infrastructure. The history of ad hoc networks
of the concept of weakly or loosely coupled systems referred started from packet radio networks although the term ”ad hoc
to as Hybrid SONs (H-SONs) in [2] and [6], which are gaining networks” was only later officially adopted by an IEEE 802.11
importance within the current network landscape. From a subcommittee after which the term became popular. The
standardization perspective, there has been insufficient interest interest in SONs rose in the 1980s and were originally defined
in self-coordination aspects and the focus has been on the to be distributed, thus having no centralized control. The
3rd Generation Partnership Project (3GPP), the International 3GPP Rel. 8 (2008) divided SONs into three groups, including
Telecommunications Union-Telecommunication Standardiza- Centralized SON (C-SON), Distributed SON (D-SON), and
tion Sector (ITU-T), and the European Telecommunications H-SON, which is a combination of C-SON and D-SON.
Standards Institute (ETSI) activities [5], thus overlooking 3GPP Rel. 11 (2011) proposed the term self-coordination to
many initiatives driven by other standardization development avoid and resolve conflicts in SONs [6]. The research on
organizations (SDOs). coordination of conflicts in hierarchical systems started in
Based on the above introduction, the contributions of this the 1960s with Mesarovic’s pioneering work. In hierarchical
article are three-fold. First, we concentrate on the potential systems, the conflicts were classified into inter- and intra-level
and the promises of H-SONs for an improved design-driven conflicts which correspond to vertical and horizontal coupling.
self-coordination. Second, we provide an in-depth landscape One of the methods suggested was interaction decoupling.
analysis of past and current standardization activities re- Independently of SONs, distributed computing was devel-
lated to SONs and the inherent self-coordination functionality oped in computer science towards autonomic computing that
within various SDOs to explore the harmonization possibil- is based on self-management. The term was later adopted
ities. Broader standardization-related insights are presented also in communication networks. ETSI defined the Generic
in [7]. Finally, we summarize the challenges and the future Autonomic Networking Architecture (GANA) reference model
technological trends for the development of SONs over the as the ETSI Technical Specification (TS) [9]. GANA describes
next decade. the concept of “ownership” of managed entities (MEs) (i.e.,
The remainder of this paper is organized as follows. In system resources) that relay their configuration to specific
Section II, we present the background and state-of-the art of functional entities called decision elements (DEs). They have a
SONs. Then, Section III discusses the opportunities, use cases, one-to-one relation to avoid that multiple DEs manipulate a DE
technical hurdles, and standardization landscape of SONs. at the same time, requiring DEs to negotiate with the DE that
The challenges, trends as well as directions of research and “owns” the ME. Among others, GANA defines multiple run-
standardization activities regarding SONs are identified in time techniques to synchronize policies among DEs, handling
Section IV. Finally, Section V concludes the paper. intents, and self-assessment of objective accomplishment.
The most important principles in system design include
II. BACKGROUND AND S TATE - OF - THE - ART feedback, optimization and decision making, hierarchy, and
The historical development of SONs is presented in Fig. 1. degree of centralization [8]. Feedback is based on the sense-
To obtain intelligent behavior, usually a feedback loop is decide-act loop, which in modern AI forms the basis of an
needed. Systems based on the feedback concept can be divided intelligent and rational agent, for example an SONF. The idea
hierarchically into automatic, autonomous, and self-organizing led to multi-agent systems (MASs), which form a hierarchy
systems, from bottom up [8]. A system may have regular of interacting agents, a good basis for self-organization. Even
inputs and control inputs. Regular inputs include any sensing before the AI theory was developed, the research on neural
information received from the environment, and control inputs networks, pattern recognition, and game theory was started and
are information given by an external controller to perform finally merged to the MAS theory. Thus they were precursors
needed actions. Automatic systems are systems that do not of current AI.
need manual intervention. They can be divided into control Optimization and decision making is a broad area with many
and adaptive systems. They need a control input in the form of different approaches. There are usually many conflicting ob-
a set-point value or reference signal, respectively. Autonomous jectives measuring the efficiency of the use of basic resources
systems are automatic systems that do not need any control such as energy, time, and bandwidth. Optimization problems
input during operation. Autonomous systems are learning can be convex or nonconvex in nature. Different types of
systems, which are able to change their behavior based on optima can exist for multiobjective optimization problems such
earlier experience that is saved into a memory. Self-organizing as a global or a local optimum. A typical type of optimum that
systems are autonomous systems that are able to restructure is targeted in such problems is called the Pareto optimum,
themselves. valid for commensurate objectives, but it is not in general
The general term self-organization was proposed by Ashby unique and therefore a final selection must be made based
already in 1947. In communications, the first SON was the on subjective grounds, usually on fairness. Modern joint or
 Feedback Control systems Adap ve systems Autonomous systems
Self-organiza on
Hierarchy OSI model
Arpanet Internet
Cross-layer design
Self-organizing
TCP/IP model networks
Packet switching
Packet radios Ad hoc networks
Coordina on Self-coordina on
Distributed SON Hybrid SON
ETSI GANA, O-RAN
Computer Distributed compu ng Autonomic compu ng
Ar ficial intelligence Mul -agent systems
Game theory
Pareto op mum Mul objec ve op miza on
Pa#ern recogni on
Neural networks

1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 2030

Fig. 1: Development of self-organizing networks.

multiobjective optimization started already in the 1950s. Now that combines centralized and distributed control, as in the
multiobjective optimization is an essential part of the MAS H-SON. A schematic view of this architecture is provided in
theory. In practice the basic resources are incommensurate. Fig. 2, showing essentially a hierarchical weakly controlled set
In this case one can use evolutionary methods, and the result of almost autonomous agents. There is only a weak vertical
depends on the availability of the resources and their need and horizontal coupling or interaction between agents to im-
according to the law of supply and demand. In this way the prove stability so that the feedback loops are weakly coupled
relative costs of the resources are defined when there is no or almost isolated from each other. Using different degrees of
objective basis for the definition of the costs. coupling, the hybrid form can implement all other degrees of
The idea of hierarchy is to decompose a complex problem centralization as special cases. Self-organizing systems should
into a hierarchical set of simpler problems if there are different have at least a weak centralized control to improve optimality,
time scales in the changes in the system. This is also the basis fairness, and stability so that the global behavior is predictable.
for the TCP/IP and OSI models. The hierarchy also supports Accordingly, the higher in the hierarchy we are, the more
the stability of the system. Modern research on hierarchy and intelligence and complexity we have.
modularity started in the 1960s with the work by Simon and
Mesarovic. It was noticed that biological systems are hierarchi- III. O PPORTUNITIES , U SE C ASES , T ECHNICAL H URDLES ,
cal and different subsystems or modules at the same hierarchy AND S TANDARDIZATION L ANDSCAPE
level are only weakly coupled. The hierarchies were classified In this section we discuss some opportunities that may bring
into nested, multilayer, and dominance hierarchies. The layer new possibilities, emerging use cases along with some biggest
and dominance hierarchies are common in communication impediments identified in SONs, as well as the standardization
networks. In a hierarchy, the basic idea is to use low speed, issues and initiatives derived from them.
broad range, and low resolution at the higher levels such as
in the network layer and high speed, narrow range, and high
resolution at the lower levels such as in the physical layer. The A. Opportunities
speed should correspond to the rate of change of the relevant Rethinking system design is needed to resolve the cur-
network parameters at each level. There is usually orders of rently weak coordination among SONFs. This will leverage
magnitude difference in the speed at different levels. Range the benefits of distributed systems and a pervasive approach
and resolution are defined in the amplitude, time, frequency, through using advanced multi-tenant resource sharing. By
and space domains. In a hierarchical system, a form of joint coordinating SONFs, different performance metrics will be
optimization is cross-layer optimization, which can be loosely optimized, especially in dynamic situations.
or tightly coupled. In the latter case the performance may be In addition, some SONFs mostly focus on short term
better, but the complexity is larger, and there may be stability optimization as they mainly operate at the lower layers. With
problems. the aid of pervasive intelligence, SONFs will be able to process
According to the conventional degrees of centralization, both external historical information gathered from previous
systems are classified as centralized, decentralized, and dis- experiences and internal contextual data to build solid long-
tributed. Centralized systems are based on forced cooperation term optimization and prediction, and thus will truly operate
of agents in a hierarchy, decentralized systems are based on the network in a proactive and autonomous fashion.
competition of autonomous agents without any hierarchy, and On the other hand, a key opportunity is to work on the
distributed systems are an intermediate form where the agents combined hierarchy and degrees of centralization to improve
cooperate at least with their closest neighbors by exchanging self-coordination. Hierarchy is used in SONs to manage the
information. An alternative intermediate form is a hybrid form inherent complexity whereas different degrees of centralization
 Goals/KPIs Range and
intelligence

Level-2 SONF
Slower
Decide
actions
Sense
Sense Sense Act Time scale
Act Act separation

Orthogonality Orthogonality
(time, frequency, (time, frequency, Faster
Level-1 SONF space,...) Level-1 SONF space,...) Level-1 SONF actions
Decide
Sense
Sense Sense Act
Act Act
System
Process1 Process2 Process3 resources Speed and
resolution

Fig. 2: Improved hierarchy using weakly coupled system design.

maintain SONs scalability. A hybrid form of hierarchy, i.e., O-RAN design envisions the use of the H-SON concept
H-SON (Fig. 2) is envisioned to avoid the weaknesses of where there is interaction between SONFs at different posi-
pure centralized and distributed paradigms. Such hierarchical tions in the network. The definition of the A1 interface, in
networks can be created statically in a structured manner or particular, allows a higher-level SON in the Non-RT RIC to
dynamically in an unstructured manner when needed. Struc- access information from the Near-RT RIC and to exert policy
tured network partitions can provide better support for AI/ML controls over the Near-RT RIC. Support of integrated SONFs
within and across network partitions. has been adopted as one of the use cases of interest within
The intrinsic priorities of hierarchical systems are a very O-RAN. The key high-level O-RAN functions are depicted
powerful mechanism to resolve conflicts in SONs. Sens- in Fig. 3 allowing for a multitude of more specialized use
ing information propagates upwards and control information cases ranging from RAN sharing to unmanned aerial vehicles
downwards in the hierarchy so that higher levels will detect the (UAVs) radio resource allocation using a weakly coupled
conflict and act by preempting the lower levels where decisions design.
are usually made because they are faster. This hierarchy will
prevent a conflict from lasting for a long time causing system
C. Technical hurdles
performance degradation. However, hierarchical systems may
need a delay to detect, analyze, and trigger the best coun- Optimization and decision making in SONs are facing sev-
termeasure commensurate with the number of the hierarchy eral hurdles. The reason is that the complexity of optimization
levels. problems depends exponentially on the size of the system to be
optimized. Therefore in practice, an exhaustive search cannot
be used and various heuristic methods have been developed. In
B. Use cases general, optimization methods can be divided into top-down
The open radio access network (O-RAN) architecture has deductive methods using parametric models and bottom-up
been deliberately designed with flexibility to deploy SON inductive methods based on pattern recognition. The method
functions at the core, metro or edge depending on the control based on parametric models is usually simple and it converges
loop latency requirements, using the non real-time (Non-RT) fast since only the parameters of the model must be identified,
RAN Interface Controller (RIC), near real-time (Near-RT) RIC but it may fail if there is high uncertainty about what kind
and potential use of edge platforms hosting virtual network of models should be used. In pattern recognition, also called
functions (VNFs) as hosts for AI/ML applications as well structural identification, the model is formed from scratch,
(Fig. 3). for example using neural networks or evolutionary methods
Development of third-party SON software realized as rApps (e.g., genetic algorithms and game theoretical principles). Such
and xApps will be made easier by the definition of reference methods are in general complex and they converge slowly, but
designs for the supporting platforms’ hardware (e.g., commer- they may work even in uncertain environments if the changes
cial off-the-shelf (COTS) equipment) and software and through are not too fast. A combination of the two approaches is known
the definition of interface standards for the sensing input and as hybrid intelligent systems.
action output of SONFs at different levels (e.g., O1 and E2 On the other hand, an additional technical impediment is
interfaces). to find mechanisms that are optimized to enforce changes
 Control O-RAN Service Management and
loops Orchestration Framework
Non-RT RIC rApps Micro Micro
serviceserviceMicro
AI/ML models Governance service
training/inference and policies Non-RT RIC
O-RAN A1
>1s
Data analytics Interface
Micro Micro
serviceserviceMicro
service
Near-RT RIC xApps Near-RT RIC
O-RAN O1
AI/ML models O-RAN E2 Interface
training/inference Resource control 10 ms - 1 s Interface

>1s
Decisions O-CU
enforcement
3GPP F1
Interface
gNB
O-DU
Decisions Resource
execution management O-RAN Open
< 10 ms Fronthaul Interface

Data provisioning
to higher levels O-RU

Fig. 3: Open RAN use case control loops hierarchy.

required by adapting SONs as soon as they can be enforced, as defined by ETSI in TS 128 313, in order to standardize
considering the next steps in optimization to avoid unneeded the mechanisms needed to build a SON, either distributed or
changes but do not wait until all changes are definite. We centralized. More specifically, they tend to cover optimization
must balance the cost of a suboptimal network and the cost issues and configuration management. However, there are open
of making the required changes to optimize it. Incremental questions regarding the cross interaction among ETSI, ITU-
improvements towards the objective need to be made. This is T, and IETF views, such as the interface compatibility and
still an open research topic. adaptation, the function definition for consistency, and the
consolidation of architecture components for allowing multi-
vendor deployments. These are required in the near future to
D. Standardization landscape
define a stable framework for SONs [7].
SONs are a key target of SDOs [5], [10]. They are included SDOs are leveraging the power of digital twins to imple-
in the key framework of the ITU-T (Rec. ITU-T Y.3324) as a ment the capabilities required by SONs. They heavily rely
target for Long Term Evolution (LTE) and beyond, thus SON on telemetry. Thus, the procedures, tools, and protocols for
becomes a source of standardization efforts for 4G successors. telemetry collection, transmission, and processing are being
Some features of SONs have been envisioned for the overall worked on for standardization by SDOs. For instance, the IETF
standardization landscape for network management. For in- has begun the initial steps for this task by the adoption of
stance, as argued in [11], a solution for a network to self-adapt the Network Telemetry Framework (NTF) as the foundation
to changes in its environment is aligned with ETSI Network of a future set of standards for this purpose. In addition,
Function Virtualization (NFV) Management & Orchestration standardizing the mechanisms required for leveraging the
(MANO) and assumed by the Internet Engineering Task Force information provided by a digital twin of a SON is an open
(IETF) as part of the Network Management Research Group issue for SDOs. A key is to extend the current abilities of
(NMRG) and the Anima WG. In this sense, standardizing self- SONs by eliminating some elements or actors from distributed
organization and intelligent reasoning procedures becomes a coordination in response to models obtained by “simulating”
key for the correct evolution of the network. It is agreed that possible scenarios, as the main purpose of digital twins.
all components, hardware and software must be assessed, both Table I summarizes current and future SDO activities [7].
in terms of their SON capabilities (i.e., auto-*, self-*), as well Obviously, most of those ongoing standardization initiatives
as their overall adequacy and quality, thus the tenant of the are either inspired by or converge toward the reference GANA
network can have a high degree of certainty that the network model, while the latter adheres to the H-SON paradigm
will work as desired and expected. guidelines.
In addition, the evolution within ETSI for designing SONs
is taking part as Zero Touch Network and Service Manage- IV. C HALLENGES , T RENDS , AND R ESEARCH /S TANDARD
ment (ZSM), Experiential Network Intelligence (ENI) [5], D IRECTIONS
and GANA [9]. They provide underlying mechanisms to In this section, we enumerate the fundamental challenges
build SONs. However, there are many topics to work on, to be addressed for providing conflict-free SONs, and propose
 TABLE I: Examples of current and future standardization activities (not an exhaustive list).

SDO or
WG or Framework Standardization Activities Activities Type and Maturity Launch
Forum
Service overlay networks as the main abstraction
NGSON WG (P1903 standards) level for autonomics via embracing context aware- Regular track 2013
IEEE ness and self-organization capabilities.
Outlines standardization items and approach for en-
INGR SysOpt WG ( [7]) Plans for standardization 2019
hancing standards on autonomics in other SDOs/fora.
Defines a reduced-scope Autonomic Networking
(AN) (which is a reference model to describe node
Partly-mature standardization
behavior and self-management properties) with pro-
ANIMA (e.g., RFC 8993). track, inspired and aligned to ETSI 2014
IETF gressive introduction of autonomic functions (AFs).
GANA.
No implementation specifications for coordination
among AFs.
Architectural framework for network telemetry. Pro-
Extend network management be-
NTF tocols to gather accurate granular network data for 2018
yond conventional OAM.
full visibility.
GANA model and its instantiations onto various
TC NTECH/AFI WG and TC Regular Standardization track for
types of fixed, mobile and wireless networks. Run-
INT/AFI WG (e.g., ETSI TS 103 technical standards and detailed 2009
ning a 5G PoC program to implement some GANA
195-2 [9] and White Paper No. 16) specifications.
aspects.
ETSI
Defines an AI-based architecture to help external
systems improve their environmental awareness and
ENI ISG Pre-standardization track. 2017
adapt accordingly. Envisions the translation of input
data as well as output recommendations/commands.
Reuses existing standards and frameworks into a
holistic design to achieve E2E automation in multi-
ZSM ISG Pre-standardization track. 2017
vendor environments using AI-based data collection
and closed-loop control.
Rec. ITU-T Y.3324: defines the functional and archi-
tectural requirements of autonomic management and Regular standardization track. 2018
ITU SG13 control (AMC) for IMT-2020 networks.
Rec. ITU-T Y.3177: specifies a high-level architec-
ture of AI-based network automation for resource
Regular standardization track. 2021
and fault management for future networks including
IMT-2020.
FG-AN: support standardisation activities of au-
tonomous networks via building upon the existing Pre-standardization track. 2020
standards’ gaps.
Release 8 (e.g., TS 32.500) Basics of LTE-SON. Mature regular standard. 2008
Release 10 (e.g., TS 32.522) Self-coordination. Mature regular standard. 2011
3GPP
Release 16 (e.g., TR 28.861) Introduction to 5G NR-SON. Mature regular standard. 2018
Describes a high-level vision of architecture princi-
ples and requirements to guide other SDOs/Fora and
5G E2E architecture framework Requirements matured as inputs to
NGMN promote interoperablity. Its automation capabilities 2019
v3.0.8 standards.
of the network and services are based on the ETSI
GANA model.
Open source model-driven framework that brings
ONAP Istanbul Release orchestration and automation capabilities to physical Regular standardization track. 2017
and virtual network components.
Builds on ETSI GANA and ITU Recs. to define
BBF AIM AFs for fixed access and E2E converged fixed/mobile Regular standards, mature. 2018
networks.
Mapping of the ETSI GANA framework to the ODA
TMF ODA (e.g., IG1167 and IG1177) Standardized frameworks, mature. 2018
intelligence management model.
AFI = Autonomic network engineering for the self-managing Future Internet; AIM = Automated Intelligent Management; ANIMA = Autonomic Networking
Integrated Model and Approach; BBF = Broadband Forum; ENI = Experiential Networked Intelligence; FG-AN = Focus Group on Autonomous Networks;
IMT-2020 = International Mobile Telecommunications-2020; INGR = IEEE International Network Generations Roadmap; INT = Core Network and
Interoperability Testing; ISG = Industry Specification Group; NGSON = Next Generation Service Overlay Network; NR = New Radio; NTECH = Network
Technologies; NTF = Network Telemetry Framework; ODA = Open Digital Architecture; ONAP = Open Network Automation Platform, PoC = Proof Of
Concept; Rec. = Recommendation; RFC = Request for Comment; SG = Study Group; SysOpt = Systems Optimization; TC = Technical Committee; TMF
= TeleManagement Forum; TR = Technical Report; WG = Working Group; ZSM = Zero touch network & Service Management.

novel approaches along with promising research directions. A and open research questions is presented in Table II.
summary of those SON-specific challenges, potential solutions
A. Challenges Quantum technology. Self-coordination usually needs to
Nowadays, the development and standardization activities solve complex optimization problems, which can be resolved
of true SON solutions are still facing some key challenges. with quantum computing (QC). QC techniques exploit the
Notably, future communication systems are becoming increas- superposition of quantum states to concurrently explore dif-
ingly complex as a result of supporting new ecosystems with ferent possibilities to quickly arrive at an optimal solution.
huge information flow. Another challenge is that communi- In this perspective, QC will reduce computation cost (e.g.,
cation systems, as machines, do not have self-consciousness complex calculations) and the overall latency of the network
and they may act chaotically in unexpected situations. A clas- (e.g., overheads of various layers) especially in virtualized and
sical example is the semantics, particularly the translation of cloudified networks [13]. This can enhance both proactive
homonyms. This increases the risk of control loss. Advancing and reactive coordination algorithms in extremely dynamic
self-organization capabilities will also need to deal with the environments. However, quantum devices need to run under
interoperability issues. Despite the uptake of new communica- cryogenic temperatures or very high pressures. They are still
tion systems, some fundamental services will continue to rely in early development stages. On the other hand, quantum
on legacy technologies as the case of voice services. For this communication techniques use entanglement for quantum key
reason, forthcoming sophisticated SONFs need to be backward distribution to support secure information exchange in emerg-
compatible with the older ones. It is also noteworthy that self- ing networks. Such techniques can achieve secure optimiza-
organization is a process that spans various components and tions by enabling secure exchange of the inherent metrics and
functions in the end-to-end mobile architecture. This makes it decisions.
very difficult to design a holistic and turnkey SON solution. Federated and transfer learning. Federated learning can
Finally, the full potential of SONs will only be achieved simplify network automation by processing learning models
if they can continuously manage computationally expensive (e.g., on COTS hardware installed at the edge) instead of
optimizations because future systems must provide quasi- handling large training datasets to reduce the time and en-
instantaneous responsiveness. ergy costs of the proactive self-coordination. Further, transfer
learning will allow capitalizing on the experience gained in
avoiding or resolving previous SON conflicts to address new
B. Future Research Directions but similar ones [2].
The self-organization paradigm is expected to leverage the Extended reality (XR) based simulations. XR environ-
power of H-SON with its enabling technologies that will ments can simulate or emulate realistic scenarios in which
emerge within the next decade, as listed below. a network equipment operates under typical real-time condi-
Evolved H-SONs. Future communication systems call for tions. The resulting observations can be further improved using
a more seamless and adaptive H-SON to better balance the knowledge and experience gained from historical information
SONFs between the higher and lower levels. The conceptual of similar equipment together with contextual data provided
fixed levels of self-organization functionality can be made by AI techniques. This can enhance the predefined rules and
more flexible during the run-time phase depending on the policies implemented at the design stage to identify potential
application being considered. Departing from the predefined SONFs conflicts.
levels at the design stage (e.g., the four abstraction levels of Collective intelligence. Conventional AI techniques are
GANA [9]), real-time self-organization can scale up and down leveraged by each SONF to selfishly benefit its own actions,
adaptively on the transmission time interval (TTI) timescale and can cooperate with the remaining SONFs for global per-
based on service needs to enable an efficient reactive self- formance. Consequently, SONs will need to handle an inherent
coordination. More reference levels can be initially defined communication overhead. In the next decade, self-coordination
to include more use cases and enable cross-layer interactions. will leverage the power of the collective intelligence wherein
This tactical layers elasticity will reduce SON signaling, and local information (e.g., current and historical parameters and
can guarantee backward compatibility with legacy SONFs that observations) are the basis for each SONF to decide on its
have their own management levels even in run-time situations optimal actions that constructively impact the entire SON func-
and immediate demands. Below, we share some enabling tioning without the need for direct communications between
technologies that may deliver the flexibility needed in evolved the involved SONFs [14]. This can enable a truly proactive
H-SONs. coordination of various SONFs with minimum or no overhead
Virtualized, containerized, and multi-tenant architec- and loose coupling. For instance, the whole SON can be
tures. Virtualization, containerization and multi-tenancy ap- modelled based on a game-theoretic approach and then deep
proaches are expected to be largely adopted. Accordingly, the reinforcement learning techniques can be applied to converge
use of COTS equipment from any vendor will potentially to the optimal action for each SONF. This combination has
facilitate self-coordination capabilities because the network been successfully explored in [15] for the case of inter-cell
control parameters (NCPs) will be similar among various interference avoidance. However, research still needs to deal
network elements. This can also reduce the parameter types with the slow convergence speed of evolutionary methods and
to be controlled, and alleviate multi-vendor compatibility neural networks (e.g., via offline training), and minimize the
issues [12]. On the downside, different SONFs can access amount of data locally needed to achieve the global benefit.
the same database and consequently compromise each other’s Sustainability and energy efficiency. There is always an
security. interplay between network performance and the corresponding
 TABLE II: Future research directions for evolving SONs.

Challenge Potential Solutions Open Research Questions

Growing complexity (due to H-SON to weakly control Adoption of virtualized, container- Ensure full competition and logical isolation as well
more system elements, various sub-systems with ized and multi-tenant architectures as secure services (i.e., SONFs)
heterogeneous ecosystems, reduced intra-network Use of COTS equipment, open and
and increased data volumes). signaling Risk of obsolescence and need to proactively manage
portable services for multi-vendor
replacement strategies
interoperability
Federated learning and transfer
Heterogeneity of devices, subsystems, tasks and data
learning
Develop seamless fallback mechanisms to down-
Service redundancy: develop main
Support of backward H-SON to orchestrate grade the SONF to an older version or restore its
services on SONs and backup ver-
compatibility and coexistence network operations over previous configuration in case of outage situations
sions on legacy systems.
with legacy non-SONs or various technologies and to improve fault tolerance
partial SONs. standards Develop SON-capable gateways on
the top of legacy systems with
Heterogeneity and complexity of legacy systems
translation and normalization capa-
usually vendor-dependant.
bilities to endow legacy systems
with automation features
The risk of losing control and Define the human intervention scope and frontiers as
Human in the loop
situation awareness (because H-SON to keep weakly well as manage human weaknesses (e.g., illness)
of emergence). centralized control Advanced resilience capabilities
The duration of the temporary system unavailability
such as self-stabilization to self-
should be predictable and bounded.
restore a stable system state.
New standards for the H-SON concept to define more
GANA as a reference model, refined degrees of centralization depending on the
Harmonized H-SON many possible instantiations for strength of the intended coupling. Achieve an adap-
There is no one-size-fits-all
model for wired and customized implementations tive and seamless balance between the centralized
standard.
wireless networks and distributed SONFs
SDOs/Fora cooperation to unify A joint multi-SDO/fora development process should
and narrow taxonomies be adopted from the early standardization stages
Proactive coordination using AI
Need real-time computation offloading mechanisms
Computationally expensive H-SON to avoid getting and continuous learning
optimizations. stuck in local optima Quantum capabilities in terms of running under
Quantum technology for faster AI
cryogenic temperatures or very high pressures are
and computing
still an open problem
Sustainable energy provisioning Achieving a profound technology shift through large
(e.g., renewable sources) scale adoption of renewable-powered system units

energy efficiency [2]. One way to resolve the potential conflicts loops are almost isolated from each other to guarantee the
is to define priorities. Coordination between SONFs attempting stability of the network. Thus we discuss the potential and
to optimize performance and others attempting to optimize the promises of H-SONs for an improved self-coordination
energy usage will be an important area of future work given to avoid and resolve possible conflicts. We provided an in-
a goal of sustainable networks. depth landscape analysis of past and current standardization
Human factor. SONs aim at reducing human intervention activities related to SONs and the inherent self-coordination
in system optimization related tasks, but a human in the functionality within various standardization organizations to
self-coordination process is indispensable to prevent system explore the harmonization possibilities. Finally, we summa-
malfunction since machines cannot be endowed with self- rized the challenges and the future technological trends for
consciousness [6]. For a successful operational mode, human the development of SONs over the next decade. An important
intervention should be in the highest level loops, and ide- challenge is to adapt the network to the correct degree of
ally should not intervene at the lowest levels characterized centralization between centralized and distributed networks
by tighter constraints in terms of speed and responsiveness. depending on the situation in the network environment.
Moreover, and besides autonomous learning, human experts
can also contribute to enriching the coordination rules fed into ACKNOWLEDGMENT
the network.
This work was supported in part by the EU Horizon 2020
project DEDICAT 6G under Grant no. 101016499.
V. C ONCLUSION
In this paper we analyzed the problem of managing complex R EFERENCES
networks more efficiently. As it can be seen from history, a
[1] O. G. Aliu, A. Imran, M. A. Imran, and B. Evans, “A survey of self
potential solution is H-SON, which combines weak centralized organisation in future cellular networks,” IEEE Communications Surveys
control and distributed control so that the different feedback & Tutorials, vol. 15, no. 1, pp. 336–361, First Quarter 2013.
 [2] H. Fourati, R. Maaloul, L. Chaari, and M. Jmaiel, “Comprehensive Pedro Martinez-Julia [M] received the B.S. in Computer Science from the
survey on self-organizing cellular network approaches applied to 5G Open University of Catalonia, the M.S. in Advanced Information Technology
networks,” Computer Networks, vol. 199, pp. 1–24, 9 November 2021, and Telematics and the Ph.D. in Computer Science from the University
Art. no. 108435. of Murcia, Spain. He is currently a full-time researcher with the National
[3] S. Dobson, D. Hutchison, A. Mauthe, A. Schaeffer-Filho, P. Smith, Institute of Information and Communications Technology (NICT), Tokyo.
and J. P. G. Sterbenz, “Self-organization and resilience for networked He has been involved in EU-funded research projects since 2009, leading
systems: Design principles and open research issues,” Proceedings of several tasks/activities, and participating in IETF/IRTF for the standardization
the IEEE, vol. 107, no. 4, pp. 819–834, April 2019. of new network technologies. He has published over twenty papers in refereed
[4] T. Darwish, G. K. Kurt, H. Yanikomeroglu, G. Senarath, and P. Zhu, conferences and journals. His main expertise is in network architecture, control
“A vision of self-evolving network management for future intelligent and management, with particular interest in overlay networks and distributed
vertical HetNet,” IEEE Wireless Communications, vol. 28, no. 4, pp. systems and services. He is a member of ACM and IEEE.
96–105, August 2021.
[5] V. P. Kafle, T. Hirayama, T. Miyazawa, M. Jibiki, and H. Harai, “Net-
work control and management automation: Architecture standardization
perspective,” IEEE Communications Standards Magazine, vol. 5, no. 3,
pp. 106–114, September 2021.
[6] A. Bayazeed, K. Khorzom, and M. Aljnidi, “A survey of self-
coordination in self-organizing network,” Computer Networks, vol. 196, Ranganai Chaparadza Dr.-Ing./PhD, is a Senior Capgemini Consultant for
pp. 1–32, 4 September 2021, Art. no. 108222. Vodafone (and other Telecommunications Network Operators) and Solutions
[7] IEEE International Network Generations Roadmap (INGR), Systems Design Architect, Standardization of Autonomic/Autonomous Networking in
Optimization Working Group, “An IEEE 5G and beyond technology ETSI TC INT/AFI WG. IPv6 Forum Fellow representative in ETSI TC
roadmap: Systems optimization, 2021 edition,” 2021. INT/AFI WG. PhD in Telecomm & Computing Engineering from Technical
[8] A. Mämmelä, J. Riekki, A. Kotelba, and A. Anttonen, “Multidisciplinary University of Berlin (TUB)/Germany; MSc in Telecommunications Engineer-
and historical perspectives for developing intelligent and resource- ing from Warsaw University of Technology, Poland. Standardization Expert
efficient systems,” IEEE Access, vol. 6, pp. 17 464–17 499, 2018. (ETSI, BBF, ITU-T, NGMN, TMF, 3GPP, IEEE, IETF, etc); Innovation with
[9] ETSI TS 103 195-2 V1.1.1, “Autonomic network engineering for SDN, NFV and AMC (Autonomic Management & Control) using ETSI
the self-managing Future Internet (AFI); Generic Autonomic Network GANA-oriented standards; ETSI 5G PoC. IEEE INGR Future Networks SBB
Architecture; Part 2: An architectural reference model for autonomic WG & SystOpt WG & Testbeds WG.
networking, cognitive networking and self-management,” May 2018.
[10] X. Long, X. Gong, X. Que, W. Wang, B. Liu, S. Jiang, and N. Kong,
“Autonomic networking: Architecture design and standardization,” IEEE
Internet Computing, vol. 21, no. 5, pp. 48–53, September/October 2017.
[11] P. Martinez-Julia, V. P. Kafle, and H. Harai, “Exploiting external events
for resource adaptation in virtual computer and network systems,” IEEE
Transactions on Network and Service Management, vol. 15, no. 2, pp.
555–566, June 2018. Muslim Elkotob [M] Dr.-Ing./PhD, is a Principal Solutions Architect at
[12] L. Bonati, S. D’Oro, M. Polese, S. Basagni, and T. Melodia, “Intelli- Vodafone with a lead role and end-to-end responsibility in the Enterprise
gence and learning in O-RAN for data-driven NextG cellular networks,” Business Line. He works on driving innovation and standardizing architectures
IEEE Communications Magazine, 2020, to be published. in the areas of SDN/NFV, Autonomics, Slicing and Security in 5G and IoT.
[13] R. Ferrara, R. Bassoli, C. Deppe, F. H. Fitzek, and H. Boche, “The He is an IPv6-Forum Fellow and delegate with lead roles in various SDOs
computational and latency advantage of quantum communication net- including ETSI, TMForum, ITU-T and IEEE. Having a career background
works,” IEEE Communications Magazine, vol. 59, no. 6, pp. 132–137, with vendors, service providers and R&D, he has spent the last seven years
June 2021. strengthening Vodafone’s role in the enterprise Value Chain as a global player
[14] R. Li, Z. Zhao, X. Xu, F. Ni, and H. Zhang, “The collective advantage with a powerful infrastructure and autonomic IT services on top.
for advancing communications and intelligence,” IEEE Wireless Com-
munications, vol. 27, no. 4, pp. 96–102, August 2020.
[15] K.-K. Wong, G. Liu, W. Cun, W. Zhang, M. Zhao, and Z. Zheng, “Truly
distributed multicell multi-band multiuser MIMO by synergizing game
theory and deep learning,” IEEE Access, vol. 9, pp. 30 347–30 358, 2021.

Lyndon Ong [M] is Principal, Advanced Architecture in the Office of the

CTO at Ciena Corporation. Dr. Ong is focused on control architecture and
5G/6G open RAN, and has headed numerous projects in the areas of optical
network control and SDN at Ciena, where he received the Ciena Technical
Fellow award in 2014. He is currently co-chair of the Open RAN Alliance (O-
Abdelaali Chaoub [SM] has been an Associate Professor in Telecommuni-
RAN) Orchestration and Cloudification WG and Project Leader for the ONF’s
cations, appointed at the Institut National des Postes et Télécommunications
(INPT) of Morocco since 2015. He obtained an engineering degree of Open Transport Configuration and Control (OTCC) Project. He previously
telecommunication from INPT in 2007 with the highest honors and received chaired the OIF Technical Committee and the IETF Sigtran WG, where he
participated in development of the SCTP transport protocol. Dr. Ong received
his Ph.D. degree in electrical engineering from Mohammed V-Agdal Uni-
his doctoral degree from Columbia University in 1991.
versity in 2013. His research interests are related to remote/rural connectivity
solutions for beyond 5G/6G networks, design and optimization of IoT-enabled
smart environments, dynamic spectrum access and cognitive radio networks,
adaptive multimedia streaming in wireless networks and mitigation of DoS
attacks in cellular networks. He has worked as a Senior VoIP solutions
consultant at Alcatel-Lucent (2007-2015).

Dilip Krishnaswamy [SM] received a PhD in electrical engineering from the

University of Illinois at Urbana-Champaign. He currently serves as a Senior
Principal Architect at Sterlite Access Solutions, and recently served as VP for
R&D related to new and emerging technologies at Jio Platforms. Previously
Aarne Mämmelä [SM] received the degree of D.Sc. (Tech.) (with honors) he has worked as a Platform Architect at Intel, as a Senior Staff Researcher
from the University of Oulu in 1996. He was with the University of Oulu in the Office of the Chief Scientist at Qualcomm Research, and as a Senior
from 1982 to 1993. In 1993 he joined the VTT Technical Research Centre Scientist at IBM Research. He is an inventor on 60+ granted US patents and
of Finland in Oulu. Since 1996 he has been a Research Professor of wireless has authored 70+ research publications. His current research interests include
communications. He visited the University of Kaiserslautern in Germany in 5G/6G networks, distributed optimization and machine learning, wireless
1990-1991 and the University of Canterbury in New Zealand in 1996-1997. networking, and quantum computing.
He has been an advisor to 10 doctoral students and published over 40 journal
papers and over 100 conference papers.