White Paper:

Intelligent Networking,
AI and Machine Learning
A Telecommunications Operator’s Perspective
Authored by the Members of the Linux Foundation End User Advisory Group
(listed alphabetically by company name)

Lingli Deng (CMCC)

Kaixi Liu (CMCC)
Yuhan Zhang (CMCC)
Massimo Banzi (Telecom Italia)
Steve Casey (Verizon)
Beth Cohen (Verizon)

Please direct any questions to [email protected].

Table of Contents
1 Key Takeaways......................................................................................................3
2 Overview................................................................................................................4
2.1 LFN EUAG: Role and Mission.....................................................................5
3 Problem Statement..............................................................................................6
4 What is Driving Intelligent Networking Adoption?...........................................8
5 Purpose..................................................................................................................9
6 Assumptions....................................................................................................... 10
7 Defining Intelligent Networking...................................................................... 11
8 Survey Results.................................................................................................... 12
8.1 Methodology............................................................................................. 12
8.2 Survey Analysis......................................................................................... 13
9 Survey Respondents’ Challenges.................................................................... 15
10 Survey Implications........................................................................................... 17
11 Recommended Actions..................................................................................... 17
11.1 Build a Common Understanding of AI Platforms................................ 18
11.2 Encourage Development of AI Skillsets................................................. 18
11.3 Promote Industry Collaboration through Open Source Projects...... 19
11.4 Promote Data Model Sharing ................................................................ 20
11.5 Establish a Unified Testing and Certification Program........................ 21
12 Conclusion.......................................................................................................... 22
13 Glossary.............................................................................................................. 23
14 Detailed Definition............................................................................................ 24
15 References.......................................................................................................... 25

White Paper: Intelligent Networking, AI and Machine Learning 2

1 Key Takeaways
• On the whole, the maturity of intelligent networking in the Telecom Industry
is still relatively low

• Some intelligent networking has been deployed by Telecoms, but more

research and development is needed to establish industry wide best
practices

• Currently the industry is more interested in using artificial intelligence

and machine learning to address operational, maintenance, and service
assurance issues than network optimization

• The Telecom industry needs to develop common AI platforms and intelligent

networking frameworks and methodologies to support the delivery of new
services quickly and efficiently

• A shared understanding of intelligent networking will help to support

interoperability

• Creating commonly understood sources of reliable data is difficult, both

within a company across business units and across the Telecom industry

• The open source community can play a key role in furthering the
development of these frameworks and best practices. Some projects that
look promising include Anuket Thoth, O-RAN, 3GPP SA5 and ITU-T standards
organizations, as well as the ONAP, O-RAN, and TIP open source projects

White Paper: Intelligent Networking, AI and Machine Learning 3

2 Overview
Telecoms need to be able to incorporate new technologies and next-generation
connectivity such as 5G to customers and end users. To achieve these ambitious
goals, they need to optimize their networks – make them more intelligent if you
will. Some of the tools needed include artificial intelligence (AI), machine learning
(ML) and artificial intelligence operations (AIOps). This document will explore what
intelligent networking means to telecoms, vendors and customers, and how AI
and ML technologies and tools can be used, the cultural shifts the industry needs
to make it a success, and what to bear in mind when deploying machine learning
across a telecom network.

The LF Networking (LFN) End User Advisor Group (EUAG) is publishing this
document to identify and highlight the latest thinking and recommendations for
building and supporting intelligent networking and the tools needed to achieve it. It
will touch on the state of automation and adoption of intelligent networking tools by
telecom operators. This is a new area for many in the telecom industry, so the focus
will be on the requirements, tools and approaches that have been deployed, and
some potential future scenarios for intelligent networking and AI/ML tools. Some of
the topics covered will include:

• Fundamental issues and challenges that might be solved by applying

Intelligent networking technology

• Establish a definition of intelligent networking

• The promotion of intelligent networking transformations

• Results from a February 2021 survey of telecom operators and vendors in

the industry ecosystem about the current state of industry adoption

• How intelligent networking might be incorporated into networks and

processes to improve operations and ensure that the solutions work as
expected in production network environments

• Recommended approaches and the potential for open source projects to

contribute to the next generation of intelligent networking tools

White Paper: Intelligent Networking, AI and Machine Learning 4

2.1 LFN EUAG: Role and Mission
The LF Networking (LFN) End User Advisor Group (EUAG)’s mission is to share views,
challenges, and best practices among organizations in the telecommunications
industry; particularly highlighting areas of opportunity for open source developer
communities. The group is comprised of individuals from various organizations
from the industry including telecommunications carriers, cable operators,
network providers, and compute or storage service providers. As the voice of
the operator end user community, it represents the operators’ perspective for
telecommunications related open source projects, and their adoption in the
industry.

White Paper: Intelligent Networking, AI and Machine Learning 5

3 Problem Statement
Telecom operators are first and foremost technologists, but the reality is that there is
constant pressure to increase the efficiency and capacity of operators’ infrastructures
to delivery more services to customers for lower operational costs – to make the
business work more efficiently. The software industry, leveraging virtualization,
cloud native approaches, agile methodologies, and test-driven development has long
been able to build applications and infrastructure flexible enough to be seamlessly
modified multiple times a day.

Could and should the Telecommunications industry, with its stringent requirements
for high availability, and its distributed service delivery models, adopt these
methodologies for its own infrastructure and systems? As software defined
networking (SDN) becomes more robust, operators have found that it is not enough
to just convert everything to software, step back and expect it to all work.

Based on the information submitted by participating operator and vendor

organizations, the level of sophistication about intelligent networks is still relatively
low, with little cross-departmental and cross community sharing or tools. This leads to
further silos, fragmentation of development and research activities, and less efficiency
across the industry.

It has been said that data is king and that is certainly true for any kind of AI tools.
Machine learning needs lots of data, the more data to analyze; the more effective the
results will be. As an example of this phenomenon, the problem of how to translate
texts effectively and efficiently was long thought to be unsolvable. Then Google
applied copious data sets in a brute force method that worked. While it is far from
perfect, certainly not at the level of Star Trek’s Universal Translator, Google Translate
is a reasonably workable method of text language translation.

One of the most pressing problems is the lack of an understanding of the data itself
and how it needs to be organized and processed to successfully apply machine
learning to improving network efficiencies.

White Paper: Intelligent Networking, AI and Machine Learning 6

 Figure 1: Some data types used for intelligent networking

There is a need for intelligent networking tools both in support of internal as well as
customer facing processes. The amount of data needed to track network workflows
to effectively create AI tools that will be able to do real-time predictive modeling. In
summary, not only is there a very long list of components that need to be part of the
Intelligent network, but capturing the right data and at the level of granularity that is
needed to produce results is still very much more an art than a science.

White Paper: Intelligent Networking, AI and Machine Learning 7

4 What is Driving Intelligent
Networking Adoption?
• Market and Business Expansion: For many telecoms, the consumer market
has long been saturated, so there has been a strong push to look for new
market opportunities. Technology changes have opened growth potential for
digital transformation, which puts additional demands on network services.

• Network Technology Evolution: The industry’s embrace of 5G,

network slicing, Infrastructure and Network Functions Virtualization (NFV)
and other cutting edge technologies, has put pressure on companies to
develop new features such as flexible resource allocation and dynamic
scheduling, as well as the need to support disaggregated network,
software and hardware models.

• Optimization for Operations Management: As the complexity of

networks and network interoperability increase, it becomes more difficult
to support traditional network operation and maintenance management
models. AIOps models could be a way to improve efficiency by aiding in
process coordination, information screening and automation of operational
management workflows. Some of the possible applications are service
subscription, fault monitoring, and quality optimization.[1]

White Paper: Intelligent Networking, AI and Machine Learning 8

5 Purpose
Intelligent networking as a tool in the telecom industry is becoming more urgent as
business and technical pressures increase. In order to gain a better understanding
of the state of the use of AI and intelligent networking in the industry, a survey of
65 telecom operators and vendors was conducted [2]. From the survey results, the
EUAG is recommending some ways to improve the rate of adoption. These include,
using industry standards bodies to create reference models, validate technologies,
formulate industry standards and promote related Open Source activities.

White Paper: Intelligent Networking, AI and Machine Learning 9

6 Assumptions
• Intelligent Networking is something of common interest across the industry

• Care should be taken about how to share data and results across competing
companies.

• Creating a shared understanding is beneficial to the entire industry – it is not

a competitive advantage to have access to better AI tools.

• The open source communities and standards bodies are in the best position
within the industry to develop some of the tools

White Paper: Intelligent Networking, AI and Machine Learning 10

7 Defining Intelligent Networking
Starting from a general definition of AI, according to the European Commission’s
High-Level Expert Group on AI, AI systems are software (and possibly also hardware)
systems designed by humans that, given a complex goal, act in the physical or
digital dimension by gaining an understanding of the environment through data
acquisition, interpreting the collected structured or unstructured data, applying
reasoning to that information, or processing the information, derived from this data
and finally, deciding the best course of action to take to achieve the stated goal.
AI systems can either use symbolic rules or a numeric model, and they can also
adapt their behavior by analyzing how the environment is affected by their previous
actions [3].

Clearly, that definition is far too broad to apply to intelligent networking. A

better approach is to establish a common understanding of what it is within the
telecommunications context. However, it is quickly apparent with a web search
that AI, ML, and AIOps mean different things to different groups even within the
networking and telecommunications industry. While AI and machine learning is
often understood to be how automation can be used to optimize the network,
making sure the networks, connections, and packet flows remain optimized as
traffic naturally shifts over time. That optimization can happen within a telecom’s
own infrastructure or it might be tools that customers apply to their own networks.

This can also be more broadly applied to operational aspects of the telecom
infrastructure, which implies the use of AI/ML outside of the network itself, often
referred to as AIOps. According to Gartner, AIOps combines big data and machine
learning to automate IT operations processes, including event correlation, anomaly
detection and causality determination [4].

For the purposes of this document, intelligent networking is a network empowered

by AI technologies and systematic integration of AI and communication network
on hardware, software, systems, and processes. This includes but is not limited
to transforming network software and hardware into an AI system, as well as the
operational intelligence represented by the AIOps components [5].

The relation between intelligent networking and AIOps could be further explained
this way. The ultimate goal of intelligent networking is to turn a network composed
of software and hardware into an AI system, which covers but is not only subject to
AIOps.

White Paper: Intelligent Networking, AI and Machine Learning 11

8 Survey Results
8.1 Methodology
The LFN EUAG group developed the survey with the intention of gathering data
about the current state of the use of intelligent networking within the telecom
industry. The survey closed in February 2021 with a total of 65 respondents from
within the LFN community, made up of a broad spectrum of telecom operators
and vendors.

The criteria used for evaluating the findings and establishing the assumptions
as part of that evaluation were:

• Overall industry type and size of respondents: Respondent profiles

• Resource capacity: Level of interest in developing intelligent networking tools

Figure 2: LFN Survey Methodology

As shown in Figure 2, the survey was organized into the following areas of interest:
status of application requirements, research and development strategy, ecological
strategy, and challenges. A comprehensive analysis of the data used a combination
of classification, correlation and benchmarking techniques.

White Paper: Intelligent Networking, AI and Machine Learning 12

8.2 Survey Analysis
Overall progress in adopting intelligent networking

Based on the survey results, there is much that needs to be done before operators
and vendors achieve a fully intelligent network. Nearly half of the operator
respondents are still in the early stages of assisted operational management – one
of the first steps in making a network more intelligent. Respondents indicating they
have a conditional autonomous network or higher only account for about a quarter
of respondents.

Figure 3: Industry Progress

On the other hand, over 80% of vendors claim to have already applied AI technology
in some form in their products. Further analysis notes that 36% are only piloting
their solutions, so the claim that vendors have AI solutions seems to be more
aspirational than real at this point. Also interesting to note is that only 16% of
vendors have a long-term plan for deployment of AI into their architecture or
products. From these results, it can be inferred that many vendors are just starting
to dip their toes into the intelligent networking waters, so to speak.

Development strategy

The responses indicate that companies at more advanced stages of their AI/ML
adoption are more likely to adopt the strategy of building a unified AI platform that
can be used for all intelligent network application scenarios, rather than a piecemeal
approach to building their systems.

White Paper: Intelligent Networking, AI and Machine Learning 13

 Figure 4: Development Strategy Analysis

Looking at the more detailed responses specific to what respondents are looking
for, 60% said that basic AI algorithms, algorithm frameworks, and training
capabilities were considered most desirable. However, more than half of the
respondents said that they needed to focus on data access, which includes the
availability of raw data, training data and subject data. Less than half attached any
importance to providing reusable common business capabilities and applications
based on network intelligence, organized research and development, deployment
of common intelligent networking capabilities, and trusted AI capabilities. This
might be more related to the fact that few of the respondents are researchers, while
most are architects and network engineers, who are going to be more interested in
practical solutions that can be applied directly to solving their immediate business
problems.

Application scenarios

The top three AI application use cases the industry was most interested in are
operations and maintenance, service assurance and network optimization. This is
not surprising, since these use cases have direct application to serving customers
and front-line operations.

Autonomous network elements, energy efficiency, and application optimization

were generally of less interest, mostly due to a perceived lack of feasibility and lack
of operators’ control over the algorithms and immaturity of the technology.

White Paper: Intelligent Networking, AI and Machine Learning 14

Ecology strategy

Questions related to the ecology strategy showed that there is not yet a consensus
on what open source projects, standards groups, and certifications (which do
not currently exist) should take priority. However, higher priority was given to
network service and AI algorithms certification efforts. Lower priority was given to
certification of network elements and applications residing above the network layer.

Figure 5: Ecology Strategy Analysis

White Paper: Intelligent Networking, AI and Machine Learning 15

9 Survey Respondents’
Challenges
From the survey results, clearly there is much work to be done. AI and machine
learning is still fairly new in the industry. Many telecoms have done optimization by
hand over the years – looking at historic data and tweaking networks – but what has
changed recently is the ability to do it more in real time. However, the technology
and its adoption are still evolving.

The operators, as the consumers of network intelligent technology, identified a

lack of a truly controllable and experimental network environments, no shared
or open network data sets, scarce AI talent, and technical resources. These
represent long-term common challenges faced by operators, wherever they are
on the intelligent networking journey. According to the survey results, the lack of
quantitative indicators for intelligent network applications and third party testing
and certification services effects, is a problem that operators must solve. Building
a test and certification effect evaluation system oriented to intelligent networking
is a common industry demand. From the operator’s perspective, the more
fruitful targets for automation and AI are operations and network management
components.

From the vendor perspective, network element standardization and lack of research
environments are the common challenges. A lack of trust in the ability for intelligent
network control is the main challenge for vendors to continue long-term intelligent
network planning and development of network intelligent tools and systems. A lack
of unified and trusted data normalization, shared and open network data sets are
also challenges for vendors to their commitment to develop intelligent networking
tools.

Yes, we have come a long way in building the intelligent network, but don’t be
fooled, the Real-Time enterprise requires significant integration work to pull
together all the components needed to support it. It is not just a matter of dropping
in some new equipment; it requires fundamental changes to the very business
delivery model itself. There are opportunities to inject intelligence at all these
different points, from the customer experience, automatic portals that anticipate
customers’ needs and interests at the front end. The very nature of how networks
are supported has to change. And not just one component, but all the piece parts
changing in a coordinated manner.

White Paper: Intelligent Networking, AI and Machine Learning 16

10 Survey Implications
• In general, the telecom industry is still in the early stages of transformation,
with relatively low levels of intelligent networking deployed

• As part of the overall industry intelligent transformation strategy, operators

and vendors should coordinate intelligent network applications through a
unified understanding of the AI platforms

• For now, operators are most interested in AI applications for operations and
maintenance, service assurance, and network optimization

• Many companies are involved in open source and standards bodies such as
O-RAN, 3GPP SA5 ITU-T, ONAP, and TIP

• Certifications and standards for intelligent network services and AI

algorithms are needed

• Operators need to prepare for the long-term challenges of introducing

intelligent networking. These include, the development of intelligent network
element and data standardization, open research environments, and the
establishment of shared data sets

White Paper: Intelligent Networking, AI and Machine Learning 17

11 Recommended Actions
11.1 Build a Common Understanding
of AI Platforms
The industry has just started down the intelligent network transformation journey.
To achieve the ultimate goal of fully autonomous networks, the recommendation
is that operators avoid introducing AI capabilities in specific, scattered use cases.
Rather, it makes more sense to build an enterprise-level unified AI platform
or framework for all intelligent network application use cases, with the goal to
provide intelligent application for all the required infrastructure services including
centralized computing power, algorithm frameworks, and reusable AI capabilities,
therefore creating one-stop management of operational and intelligent network
applications.

While it is hard to draw conclusions directly from the survey, it does make sense
for the telecom community to unite with the common goal of building a testing and
certification platform for intelligent network solution across the industry; if for no
other reason but to provide effective evaluation and testing methods for validating
intelligent applications and models. It is also critical to provide an open certification
laboratory, as well as a unified testing environment and general evaluation DevOps
pipeline for both in-house and outsourced solutions. The current candidates for this
work are some of the standard bodies already engaged in this work, such as O-RAN,
3GPP, and ITU-T. Nearly half of the survey respondents are already participating in
ONAP or O-RAN, so these projects seem like a natural place to start.

In conclusion, the industry is currently in need of both specification and open

certification of intelligent network service and network AI algorithms.

11.2 Encourage Development of AI Skillsets

As intelligent networking becomes more prevalent in the industry, operational
personnel roles will change as many formerly manual processes are replaced by
automation and intelligence. This means that staff need to know how to manage
automated workflows. The operators will need to make changes to how operations
staff are trained and to develop the specific skill sets that are required for successful
operation support.

White Paper: Intelligent Networking, AI and Machine Learning 18

11.3 Promote Industry Collaboration through
Open Source Projects
In the process of deploying intelligent networks, the operations and maintenance
activities will gradually shift from people to autonomous network systems. The basis
for system decision-making will extend from explainable experiences to complex
AI algorithms and models. That is not just a simple technical issue, but requires
industry cooperation to establish fair and objective evaluation standards, open and
reusable test environments, and well-recognized certification services.

In terms of industrial ecology, the following standard organizations and open-source

communities are expected to be relevant:

• ONAP is a project building a reference implementation for an open source

autonomous network platform under the LF Networking umbrella. It
currently has few components that can be used to promote AI/ML activities,
but it is a project that has the support across the industry, so might serve as
a jumping off platform.

• Anuket is a project that is defining reference models and architectures in

support of cloud native Virtual Network Functions (VNF). It is under the LF
Networking umbrella. It recently started an intelligent networking project
called Thoth.

• ETSI ZSM ISG defines an automation framework for the whole process
of delivery, deployment, configuration, SLA, and optimization of cross-
domain E2E network operation and maintenance. Again, while there are few
elements in the platform today that can be used for AI/ML projects, it might
serve as a jumping off platform.

• O-RAN considers network autonomy as one of its most important

development directions. By introducing big data, artificial intelligence
and other methods, it helps realize efficient operation and maintenance
management in complex network environments and improve the utilization
of spectrum resources, as well as reduce network energy consumption.

• TM Forum has taken the initiative of organizing cross-organization

collaboration for autonomous networks and establishing a working group. It
is defining a level-based architecture for autonomous network, proposes the
concept of autonomous domains, and evolves resource management into
relatively independent autonomous domains.

White Paper: Intelligent Networking, AI and Machine Learning 19

While all this activity looks very promising, there is little or no coordination
or collaboration of projects or deliverables among the standards bodies or
participating organizations. For the most part, each of them only focuses on
a specific area of interest. The situation is further aggravated by the different
standards groups and open source projects having both overlapping and conflicting
objectives of the few standards that do exist, with no shared definition of how to
support Day 2 operational infrastructure changes. This has resulted in unnecessary
barriers for vendors, promoted technology silos and increased complexity across
the industry in general. As NFV applications and workloads mature, hardware and
software disaggregation will increase with more multi-vendor solutions, requiring a
more detailed understanding of the virtualized environments and more integration
testing of the intelligent networking systems and tools.

It makes sense to focus on building a reference implementation of intelligent

networking that can be used across the standards bodies and open source
organizations.

11.4 Promote Data Model Sharing

Data standardization and shared data sets and models have been long-term
challenges for the adoption of intelligent networking. A common understanding
of the data models themselves is a basic requirement. For example, how jitter is
defined can vary wildly across operators or even within a single operator. Even
something as seemingly simple as a basic AI algorithm framework is a much needed
capability to advance the industry.

Vendors and operators need to develop common AI models for data through a
mechanism for model and data sharing. An AI/ML and model sharing project [6]
would be a good way to foster industry collaboration and promote the sharing of
data and models through the joint construction of intelligent networking scenarios.
Some use cases under consideration include:

• Congestion Prediction & Mitigation, which focuses on how AI/ML may

be used to predict congestion and perform closed loop automation for
executing configuration changes to mitigate

• Sleeper Cell Detection, which is used to predict a cell going to “sleep” and
handover a critical UE (e.g. ambulance) to another cell

• Traffic Steering, which aims to improve Quality of Experience (QoE) by

steering UE traffic among multiple cells

White Paper: Intelligent Networking, AI and Machine Learning 20

 • Soft fault detection and resolution, which is used to detect “soft” faults that
are not often caught and hidden by the redundant systems

• Deterministic Predictive capacity planning, which is used to detect usage

patterns so that the network can be used more efficiently

11.5 Establish a Unified Testing and

Certification Program
From the feedback of the survey, the highest priority for testing and certification
service for intelligent network is effectiveness evaluation and testing system for
intelligent applications (test cases, data collection, and quantitative metrics).

Therefore, the recommendation is to invest in building a testing and certification

program that could evaluate various intelligent applications with objective
performance metrics and evaluation approaches by scenarios, categories, and
levels. It would be used for the entire lifecycle management of AI application
research and development, deployment, expectation management, feasibility
analysis, closed loop upgrading, provisioning and cancellation, etc., to improve the
objectivity and effectiveness of related process decisions.

The Compliance and Verification Committee (CVC) is a community driven body

within LF Networking which defines policies and oversight for compliance and
verification programs. The CVC Anuket Assured program [7] is an open source,
community-led compliance and verification program to demonstrate the readiness
and availability of commercial cloud native and virtualized products and services
using Anuket and ONAP components. It offers testing of commercial products built
on requirements from ONAP, multiple standards bodies, and the LFN EUAG to
demonstrate the readiness and availability of commercial products based on the
Anuket and ONAP project work.

White Paper: Intelligent Networking, AI and Machine Learning 21

12 Conclusion
There is still a long way to go to achieve intelligent transformation for networks
with many challenges along the way. However, the industry must incorporate these
technologies into the infrastructure to meet the demands for telecommunications in
the twenty-first century.

• In general, the telecom industry is still in early stages of intelligent

transformation, with relatively low levels of autonomous, intelligent
networking deployed.

• As part of the overall industry intelligent transformation strategy, leading

operators and vendors should introduce and coordinate intelligent network
applications through a unified understanding of AI platforms. A consensus
could be reached that developing a unified platform is essential to reach the
goal of intelligent network autonomy.

• For now, operators are most interested in AI applications for operations

and maintenance, service assurance, and network optimization. Therefore,
operational personnel are encouraged to improve AI-related skills.

• Establishing a cross-industry shared data model poses considerable

challenges; however, data sharing amongst not only operators, but between
operators and vendors as well brings significant benefits.

White Paper: Intelligent Networking, AI and Machine Learning 22

13 Glossary
AI Artificial intelligence
AIOps Artificial Intelligence for Operations
CSP Communications Service Providers
CVC Compliance & Verification Committee
ETSI European Telecommunications Standards Institute
EUAG End User Advisory Group, a working group within the LFN
LFN Linux Foundation Networking
ML Machine Learning
M-SDO Multiple- Standards Organizations
ONAP Open Network Automation Platform
O-RAN Open-Radio Access Network
QoE Quality of Experience
SDO Standards Organizations
TM Forum TM Forum is a global industry association for service providers
and their suppliers in the telecommunications industry.
ZSM Zero-touch network and Service Management

White Paper: Intelligent Networking, AI and Machine Learning 23

14 Detailed Definitions
Communication network is structured by different layers. Network intelligence is
defined with the same logic:

• Terminal intelligence: the end-side intelligence embedded in various terminal

equipment. It performs 2 main functions: end-side data collection and
presentation of intelligent services to users. Data collection typically includes
measurement, signaling and sensing data concerned with end-side user
experience, network quality and environment awareness. This is also the
network intelligence closest to users.

• Network element intelligence: the distributed intelligence of network

elements and their related OMC, which collects real-time data on the
network side, provides closed-loop control capability for fundamental
network functions in the network element layer and network domain, such as
protocol processing, data transmission, route control, session management,
resource management, mobility management, etc., and executes decision
made in operation and service intelligence layer.

• Operation intelligence: centralized intelligence possessed by the network

management center of operation and maintenance. Based on the 5 core
capabilities of prediction, perception, diagnosis, decision-making and control,
it achieves whole lifecycle management of network planning, deployment,
maintenance, optimization and operation, supporting the service layer
intelligence.

• Service intelligence: as the interface between network and users, it supports

end-to-end network service intelligence. Using core technologies such as
intention, network slicing and SLA, service intelligence is able to meet the
diversified, and customized needs of users, realize the closed loop of end-to-
end service perception, assurance and optimization, and provides customers
with open capabilities to empower their own business.

White Paper: Intelligent Networking, AI and Machine Learning 24

15 References
[1] China Mobile’s Practice on Autonomous Network White Paper (2021)

[2] LFN EUAG: “Intelligent Network and AI for Survey for Telecom Operators,
Vendors and AI communities”

[3] European Commission

[4] Gartner

[5] Feng Junlan: “ITU- AI empowering industry potential online webinar. AI for 5G, 5G
for AI”

[6] EUAG AI/ML and model sharing project

[7] Anuket Assured Program

White Paper: Intelligent Networking, AI and Machine Learning 25