Ericsson AI in Networks

| 2025-01-07 | Ericsson Confidential | Page 1 of 44

Content

AI/ML in Radio Access Network

AI-Apps in Operations and Extended SW Support for RAN on ENI platform

AI/ML in Monitoring and Orchestration of Transport with ETAC solution

AI/ML in Network Optimization with Cognitive SW Solution and evolution to

EIAP
Content

AI/ML in Radio Access Network

AI-Apps in Operations and Extended SW Support for RAN on ENI platform

AI/ML in Monitoring and Orchestration of Transport with ETAC solution

AI/ML in Network Optimization with Cognitive SW Solution and evolution to

EIAP
 Industry Standardization and Alliance
Standardization
● “6G warmup” studies already started in 3GPP Rel-18/19/20
– Positioning, CSI compression and beam management
– Energy efficiency in higher layers

Ericsson is a founding member for AI-RAN Alliance

– Focus: Research, exploration, publication, data sharing, testing and verification, advocacy

| Yashar Nezami | Gheorghe Spiride | PA1 | 2024-03-02 | AI-RAN Alliance Communication Package | Ericsson Confidential | Page 5
| Yashar Nezami | Gheorghe Spiride | PA1 | 2024-03-02 | AI-RAN Alliance Communication Package | Ericsson Confidential | Page 4

| 2025-01-07 | Ericsson Confidential | Page 4 of 44

Networks AI Journey

● AI native Link Adaptation

AI & data AI native
● Ericsson Network Benchmarking ● AI powered traffic control (coverage
driven features
prediction)
● Gen-AI supported development develop and rApps
and testing ● AI powered MIMO Sleep Mode
● AI-powered rApps (BNEW, SAN, BCSS)

Networks ● Ericsson Product Information Assistant

AI journey (AIGen)

● Centralized AI in rApps ● Founding member AI in RAN Alliance

RAN-AI Ecosystem
● RAN Compute build for Architecture supported ● Leading collaborator with HCPs, HW
Advanced RAN AI algorithms manufacturers, and CSPs
● Strong contribution in 3GPP standardization

Intent based
Rule based feature AI enhanced feature AI native feature
autonomous RAN

| 2025-01-07 | Ericsson Confidential | Page 5 of 44

 AI for Centralized Automation and AI in RAN SW
Centralized automation with rApps
• Slower control loops for network design,
network level and cross domain Weeks Example use cases
optimization and assurance, life cycle
management, etc.
Network benchmarking and

rApps
• Open interfaces enable 3pp integration Days evolution, AI/ML life-cycle
and more flexibility for customized management, software
solutions upgrades, ….
Seconds
Intelligent Automation Platform
Orchestration, optimization,
programmability …
Distributed automation in RAN software
• Fast control loops to enhance scheduler, Handover decisions, QoS,
mobility, etc. for best possible dual connectivity control,
performance of cell/site level Milliseconds spectrum load balancing, ….
optimization use cases RAN
• Access to very granular real-time data in Beamforming, scheduling,
the node Microseconds CoMP, fast spectrum
management, …
• Part of RAN SW, codesigned and
validated with features
|| 2025-01-07
2025-01-07 || Ericsson
EricssonConfidential
Confidential || Page
Page66of
of44
44
 AI powered MIMO sleep mode (LTE)
Energy efficiency VP (available for now, activated on 20 sites)
RAN
MIMO sleep mode traffic aware power savings
AI algorithms applied on the baseband to forecast traffic patterns and
autonomously apply a reduced MIMO when traffic capacity demands are low.
At every 1 minute, the forecasted traffic utilization is being compared with
actual PRBs utilization and a decision to reduce MIMO or not is applied.

The feature can operate in two modes:

•MI_DETECTION - provides observability on potential energy savings gain
without reconfiguring MIMO, thus without an impact on network and traffic
characteristics. This mode performs AI training on traffic pattern in a cell.
•MI_SWITCH – performs continuous AI training on traffic pattern in a cell,
evaluates and applies optimal MIMO configuration.

Benefits & differentiators

• Lower OPEX
• Maintained user-experience
• Lower power consumption 14% 0
• No manual configuration/tuning of MIMO Sleep Mode Average saving at a Manual configurations
• Possible to combine with Cell sleep and the Low Energy Scheduler for even proof-of-concept cluster needed
better savings.

| 2025-01-07 | Ericsson Confidential | Page 7 of 44

 AI Powered Advanced Cell Supervision (LTE)
Self-Organizing Networks VP (available for now, 0% activation)
RAN
AI powered advance cell supervision improve performance
Locally executed and self-learning, on RAN Compute, Machine Learning
algorithms allows for continuous full network supervision that continuously
looks for anomaly in cell’s performance. Capable of instantaneous and predictive
detection and immediate recovery actions with minimal impact.

Benefits & differentiators

• Instantaneous detection and prescription of cells with degraded KPI
• Improve In-Service Performance (ISP) >95 % of sleeping Improved ISP
• Immediate recovery actions with minimal impacting actions when suspicious
KPI performance is anticipated
cells detected*

*In FFI

| 2025-01-07 | Ericsson Confidential | Page 8 of 44

 Machine Intelligence Enabled Mobility (LTE)
Self-Organizing Networks VP (available for now, active on 24% sites)
RAN
MI Enabled Mobility shortens handover
With many bands the measurement time for inter-frequency handover takes
time where early search thresholds secure KPIs at expense of low coverage.
By introducing AI models created for search zone and identifying candidate
bands for handover we target handovers to band with coverage.

Benefits & differentiators

• Shorter mobility action with 60% shorter handover than legacy with 250ms

60% 5%
average gain
• Drop KPI improvements of up to 5% in TDD deployment
• AI model created and trained on-line in the RBS
Faster handover Drop KPI improvement
• Zero Touch – no configuration needed
than legacy in TDD deployment

| 2025-01-07 | Ericsson Confidential | Page 9 of 44

 Uplink Traffic Triggered Mobility (LTE)
Differentiated Mobile Broadband VP (available for now, 0% activation)
RAN
Uplink traffic triggered mobility throughout boost
Uplink capacity
Increase total spectrum efficiency during continuous uplink traffic. By shifting
users in a split second, from TDD to FDD cells (for example), uplink heavy users
will enjoy
better coverage, as well as better uplink speed. TDD Cell

FDD Cell

Large uplink taffic

TDD cell capacity: 20MHz 25% uplink 75% downlink

FDD cell capacity: 20MHz 100% uplink

Benefits & differentiators

• Up to 4x throughput boost and better coverage
• AI model tracking traffic type on millisecond level 4x
• Inter-frequency mobility to offer best performance for every use case Users with uplink traffic Downlink users in TDD
get a 4x throughput cell will enjoy better
boost and better throughout as the cells
coverage are less loaded

| 2025-01-07 | Ericsson Confidential | Page 10 of 44

 AI Powered Link Adaptation (LTE)
Spectral efficiency VP (included in new LTE SW scope)
RAN
AI powered link adaptation improved spectrum efficiency
Introduces neural network in Ericsson unique EMCA architecture to enhance link
properties giving an improved spectral efficiency and increased throughput.
Current link adaptation are optimized for low loaded systems. By utilizing
information from adjacent Cells, we enhance link adaptation for medium to high
loaded systems with significant improvements in the spectrum efficiency.

Benefits & differentiators

• Unique functionality in the market
• Improve user experience, increased cell edge DL throughput up to 50%
• Improve up to 15% spectral efficiency in areas with high cell overlap and
15% 50%
medium to high loaded Improved spectrum Increased cell edge
efficiency DL throughput

| 2025-01-07 | Ericsson Confidential | Page 11 of 44

 Content

AI/ML in Radio Access Network

AI-Apps in Operations and Extended SW Support for RAN on ENI platform

AI/ML in Monitoring and Orchestration of Transport with ETAC solution

AI/ML in Network Optimization with Cognitive SW Solution and evolution to

EIAP

| 2025-01-07 | Ericsson Confidential | Page 12 of 44

 Ericsson Network Intelligence
Ericsson Network Intelligence is our predictive and preemptive support service that helps detect and
resolve potential issues in the network or acting on specific parameters to improve the performance
and quality of the network according to the objectives of the service provider. It is operational 24/7,
ensuring complete oversight of your E2E network.

Characteristics of Ericsson Network Intelligence

Predictive and Product-near Near real time Continuously

preemptive and root cause based and automatic improved
fixing possible & customizable

| 2025-01-07 | Ericsson Confidential | Page 13 of 44

 Ericsson Network Intelligence AI-App suite
Library of AI apps
Insights
AI apps which perform network wide or large-scale audit work to help prioritize
network operation activities and increase operational efficiency

Assurance
AI apps which check symptoms and fingerprints to alert potential system
failures ahead of alarms to avoid service outages and improve network quality

Self-healing
AI apps which detects network anomalies and perform
automatic recovery actions to preempt service outages

Performance
AI apps which alert on top offending cells in critical KPI's and
automate the fault isolation process to improve network
performance and increase operational efficiency

Energy Efficiency
AI apps which save power and improve energy efficiency
while protecting and improving network performance
| 2025-01-07 | Ericsson Confidential | Page 14 of 44
 Ericsson Network Intelligence (ENI)
Energy Automation with AI/ML
What: Energy Automation with AI/ML focuses on How: The solution uses ML prediction model,
automating energy-saving features to optimize regularly analyzing data in order to make a
power consumption, improve energy efficiency, decision: should it close a component down,
and enhance network performance. should it activate a component, or should it do
● The ENI system connected with 5G and 4G nothing? That decision is predicated on predictions
nodes, scope consists of 3 key use cases: based on the live network data it is trained on, and
also active monitoring of activity in neighboring
1. 5G Dynamic Deep Sleep Activation cells
2. Radio Power Efficiency Map – Site profiling, to autonomously determine the activation
and deactivation of power-saving features period.
3. Intelligent RAN Power Saving for 4G (Cell
Sleep Mode) – Ensures optimal energy efficiency without compromising
network performance.
– Uses Deep Sleep feature in supported radios

| 2025-01-07 | Ericsson Confidential | Page 15 of 44

 Intelligent RAN Power Saving for 4G/5G trial in Europe
• ENI applies deep learning ML technique to model the cells in the network
based on the traffic behavior such as data volume and RRC users and
predict power saving features activations and deactivations (4G and 5G)
• The use case will optimize energy saving without impact on network
performance due to traffic prediction and KPI monitoring

Intelligent RAN Power Saving 5G with Deep Sleep Using Dynamic AI/ML models Use Case
Energy Saving %Performance Impact
Strategy
Radio Units Consumption ENDC Users 5G DL Throughput 5G DL Volume
Conservative 22.10% -3% -17% -13%
Balanced 23.50% -3% -10% -16%
Aggressive 32.80% -6% -20% -20%

Intelligent RAN Power Saving 4G (Cell Sleep Mode) UC optimizing the energy performance from the baseline.
Energy Saving %Performance Impact
Strategy DL User THP DL User THP
Radio Units Consumption Samples of DL User THP > 40Mbps
[per cell per hour] [Aggregated on all cells]
Conservative 10.97% -5% Reduced from 36% to 33% -0.5%
Balanced 18.62% -10% Reduced from 36% to 29% No degradation observed

| 2025-01-07 | Ericsson Confidential | Page 16 of 44

 Radio Power Efficiency Map
RAN ENM

Radio Power Efficiency Map Radio channel

check
Radio sites Map created via network data acquisition reveals the RRC user check
inefficiency using a color ranking based on traffic exchanged per Recommendations
Clustering and Sleeping cell
energy unit. ranking sites per check shared with O&M
energy efficiency CSP team and/or
Linear regression ML clustering algorithm is used to cluster and rank ratio (and ML HW local CS team, plus
value selection temperature custom report
the cells based on their energy efficiency performance grouping) check

Preloaded Fault Isolation algorithms will be used to isolate the fault Traffic KPIs
analisys
causing energy inefficiencies 12+Fault isolation steps

Benefits & differentiators

• Visualize the overall network energy efficiency with site
granularity
Reduce Energy
• Understand the root cause of inefficiencies
• Provide recommendations for resolution
RCA Waste
of inefficiencies with
• Reduce energy waste by addressing possible irregularities due to recommended resolution
Over dimensioning, Optimization problems, HW issues, SW issues,
and Configuration issues

| 2025-01-07 | Ericsson Confidential | Page 17 of 44

 Other ENI ML use cases
RRU HW fault alarm fault
SFP Failure Prediction VoLTE QoS anomaly detection
analysis

• ENI monitor and predict when SFP • ENI monitor HW fault alarms, • ENI apply unstructured machine
TX bias limit will be reached causing collect the product information of learning techniques to detect any
the connection CPRI link to go down the impacted radio units and abnormalities in vSBG, vMTAS and
perform automatic vCSCF
• ENI will recommend site visit before
TX bias reaches certain threshold so radio restart • ENI will provide early detection of
service outage can be avoided • This use case will help to reduce site underlying network issues affecting
visits, less HW replacements and VoLTE quality of service E2E
• The use case will enable preemptive
maintenance by efficient planning reduced manhours spent trouble • The use case will increase
of site maintenance and prevent shooting operational efficiency with
service outage automation of KPI troubleshooting
procedures and improve end user
experience

>95% 30% Improved voice quality and service

reduction of unnecessary RRU availability
of SFP link failures prevented replacements

| 2025-01-07 | Ericsson Confidential | Page 18 of 44

 Increase network stability with CSP in Latin America
The solution: The results:

Tailor-made solution addressing CSP challenges reduction of unnecessary

30%
related to network stability and high OPEX. Some of Remote Radio Units (RRU)
the implemented AI apps: replacements with automatic
Sleeping cell Low RSSI with zero traffic recovery
alert

Hardware temperature alert PLMN service unavailable

25%
reduction of site visits due to
RRU Hardware fault Power disturbance fault Remote Radio Units (RRU)
automatic recovery isolation automatic recovery

500
zero touch fixes per quarter
improving network stability and
operational efficiency

Continuous improvements with new customized AI-apps, focusing on early detection and automation
| 2025-01-07 | Ericsson Confidential | Page 19 of 44
 Maximize end user experience at international events with
CSP in Asia
❑ 50 AI apps working together as Event dashboard
to support network monitoring in real-time
one solution
“0”
❑ Fully embedded solution with CSP Critical incident
operations processes

Site activities automation

❑ Network wide preemptive audit for to support easy operation
risk cleanup and 30 min network
health checks 35%
Efficiency
improvement
❑ Realtime monitoring with 1 min
granularity for Event Support Intelligent AI apps
Dashboard to support anomaly detection and fast fault isolation

❑ Fast execution with 5 min from

Customer
data collection to fault isolation to Satisfaction
preempt incidents
| 2025-01-07 | Ericsson Confidential | Page 20 of 44
 The value of human guided AI/ML with CSP in Middle East
AI apps: CPRI fault prevention and Radio unit fault classifier

Reported Customer success

45% of issues detected by • 91 critical and major issues
Ericsson Network was detected and mitigated
45% Intelligence was not during the first two months of
detected by Customer the service
AI/ML tool • 100% of the issues reported
by Ericsson Network
Intelligence are valid cases
• Ericsson Network Intelligence
detected almost twice as
many issues by using human
guided AI/ML compared to
55% of the issues detected
self learning AI/ML
had a critical impact on the
55% • New AI-app requirements
node and 45% had a major
was developed and
impact
implemented in a short time

|| 2025-01-07
2025-01-07 || Ericsson
EricssonConfidential
Confidential || Page
Page21
21of
of44
44
 Proven Benefits per AI-App
• Heartbeat Failure Alarm for Tier1 Asian Operator – over 200 hours saved yearly in operations by automating of
weekly fault isolation report (4 hours x 52 weeks)
• Proactive Antenna performance improvement. Tier 1 operator in Asia reports approximately 10 cases per month
where they save 4 hours in every case in reduced troubleshooting, approx. 40 hours per month. In case the site
visit is required the cost gets up compared to remote troubleshooting.
• LTE Sleeping cells use case is implemented widely. Tier 1 Latin America have reported 1282 trouble tickets
avoided and 92 site visits avoided in 12 months.
• 5G Sleeping cells is deployed in several operators and operators report only 5min time needed from data collection
to automatic resolution.
• PTP Sync Frequency is implemented in Tier 1 Latin America operator in 2022. It was developed together with one
operator in co-creation mode and is available for all operators globally.
• Radio power efficiency map is deployed in Far EasTone in Taiwan, and it is officially communicated.
https://www.ericsson.com/en/cases/2023/ericsson-far-eastone.

| 2025-01-07 | Ericsson Confidential | Page 22 of 44

 Content

AI/ML in Radio Access Network

AI-Apps in Operations and Extended SW Support for RAN on ENI platform

AI/ML in Monitoring and Orchestration of Transport with ETAC solution

AI/ML in Network Optimization with Cognitive SW Solution and evolution to

EIAP

| 2025-01-07 | Ericsson Confidential | Page 23 of 44

 Ericsson Transport Automation Controller
Overview

An intelligent, cloud-native transport controller that uses AI and ML to analyze and automate Microwave, IP
and optical networks.

Core capabilities

Observability and analytics Automation and control

Core characteristics

Analytics Endorsing
AI-driven intelligence Flexible and scalable
and automation standardization
and insights architecture
made easy and openness

| 2025-01-07 | Ericsson Confidential | Page 24 of 44

AI based Microwave analytics on ETAC
ETAC
Microwave analytics uses AI to identify network issues. The interferer Root
cause can also be identified
• Fading event detection
• Interference detection
• Antenna misalignment detection
• Utilization monitoring and trend visualization
• Availability analysis

Future ML-use cases (ML-model development on-going) - Traffic prediction, SFP

monitoring, Cable monitoring for microwave

Benefits & differentiators

• Improve In-Service Performance (ISP)
Root cause
>95%
• Root cause for traffic impact is known immediately without manual analysis
• Fast decision for corrective action leads to fewer site visits identified
• An AI that runs in the customer premises are easy to deploy also where data
collection can be sensitive for degraded Detection accuracy
KPIs
 Microwave Energy Optimization

Functionalities Benefits

▪ Centralized Radio Deep-Sleep (RDS) ▪ Save OPEX

based on AI/ML-traffic prediction ▪ Build a sustainable network
▪ Traffic Aware Power Save (TAPS) ▪ Take energy optimization measures
configuration that do not affect network performance
▪ Monitoring of energy optimization
features (RDS, TAPS)

| 2025-01-07 | Ericsson Confidential | Page 26 of 44

 Machine Learning in ETAC
• Current ETAC ML mode of operation: Model development and training on Ericsson site, model inference on customer site
• In the future ETAC is planned to integrate Ericsson MLOps components for enhanced MLOps capabilities at customer site

Model development by Ericsson Model Model training, monitoring and retraining by

R&D at Ericsson site inference ETAC at customer site
by ETAC at
customer
site

Current AI/ML mode of operation Future option for more complex AI/ML use cases
| 2025-01-07 | Ericsson Confidential | Page 27 of 44
Ericsson Transport Automation Controller
AI/ML in Transport Operations

✓ ETAC is the SW intended to monitor, analyze, and predict performance of MW links.

✓ ETAC is based on 5-10 use cases (rain, interference, wind, bending, energy
optimization, etc), provides recommendations, and enforces improvements in
configurations, reducing MNO OPEX
✓ It is a cloud-native transport controller that uses AI/ML to analyze and automate
the operations of Microwave networks.
✓ In future releases ETAC will also cover the IP Routers and Optical Networks
(Singapore’s leading Mobile Network Operator M1 and Ericsson deploys world’s
first live combined IP and microwave networks with advanced Artificial Intelligence
and Machine Learning capabilities of Ericsson Transport Automation Controller).
Content

AI/ML in Radio Access Network

AI-Apps in Operations and Extended SW Support for RAN on ENI platform

AI/ML in Monitoring and Orchestration of Transport with ETAC solution

AI/ML in Network Optimization with Cognitive SW Solution and evolution to

EIAP
Transforming and innovating new functionality
in the rApp examples
Energy Management Network Evolution Network Deployment Network Optimization
Capacity Planning Configuration Consistency Energy Efficiency
Performance Diagnostics
Network Slicing Performance Diagnostics

Energy Efficiency Capacity Planning Network Healing Cloud RAN LCM Performance Optimizers
Interference Management Cell Shaping
Optimization Mobility Optimization Performance Optimizers Incident Management Foundation rApps
Energy Management
Cloud RAN Observability Enabler rApps
Assurance Cell Shaping – AAS Cell Shaping – RET Geolocated diagnostics

Intelligent services C-SON Cognitive Software

Ericsson rApps CSP rApps 3PP rApps

Intelligent
Configuration Monitoring services
Ericsson Intelligent Automation Platform SDK

3rd party
ENM
EMS 3rd Party
EMS

3rd party Integrated Cloud Open 3rd Party

RAN
RAN RAN RAN RAN RAN

2021-2028 (Cognitive SW) 2025 - onwards

Capacity Management

Capacity planning
Cognitive SW

Capacity planning to maximize return of investment

• AI-powered KPI performance prediction, cell-by-cell for the
whole network
• Identify network bottlenecks to prioritize capex investment
• Predict optimal action plan & performance
• Evaluate what-if scenarios using ML-model
• Used for carrier upgrades, new sectors & sites recommendations
• Multi-vendor & multi-technology support, key for Re-farming

Benefits & differentiators

• Optimized RoI
– Expansions prioritized over time & in order of performance impact
20-40%1 83%2
• Improved accuracy Capex savings less Increased operational
– Field-proven hit rate to identify cells with capacity issues in advance carrier expansions efficiency increased
• Subscriber experience driven compared to operational efficiency
– AI model learns from the existing network to be able to predict future traditional approach on dimensioning tasks
SKPIs on cell level (Video resolution, Web page load time…)
Based on real customer outcomes: 1. Asian Operator, 2. South American Operator
 RF Shaping

Cell Shaping – Power Optimizer

Cognitive SW
Tx power optimization for energy & capacity savings
action At
Downlink power optimization solution uses reinforcement learning technology
to identify if cell TX power can be reduced without compromising coverage or
performance. Equally the solution identifies cells where power increase is Network final settings Live
Power
required for performance improvement. Power optimization saves energy and Digital network
allows maximizing radio capacity in markets with strict RF Optimizer
Twin partial copy
emissions regulations.
Rt+1 St+1
reward Rt
state St

Benefits & differentiators

• Optimize TX power individually for each cell
• Continuous power optimization automatically maintains the optimum settings
as the network evolves and traffic distributions change
57% 3%
Average DL power Network level increase
• 57% average power reduction on coverage layer while maintaining traffic
volume & improving performance1 reduction on coverage in DL & UL throughput
layer while traffic & 5% reduction in DCR1
volume unchanged1
Based on real network outcomes when used in conjunction with RET Optimizer: 1. Service provider
in Europe
 RF Shaping

Cell Shaping – RET & AAS Optimizers

Cognitive SW rApp
AI powered automatic antenna tilt optimization
Reinforcement learning enables continuous tilt optimization that learns how the RET/AAS
performance of each cell reacts to antenna tilt changes. The ability to optimization
automatically adapt to the characteristics of each cell and the surrounding
network leads to an optimized radio environment and traffic distribution state reward action
significantly improving the end user experience. St Rt At
• Remote Electrical Tilt (RET) optimization is used to improve coverage, minimize
inter-cell interference and distribute traffic for improved network utilization. Rt+1
• With planned support for the NR Flexible Cell Shaping RAN feature, both the St+1 Network
digital tilt (vertical plane) and the beamwidth (horizontal plane) are optimized
for Active Antenna Systems (AAS).

Benefits & differentiators

• Reinforcement learning adapts the optimization based on the characteristics
of each cell and it’s influencing area
• 30%/5% throughput improvement (UL/DL) and 5% DCR reduction while
30% 5%
carrying more traffic1 Network level increase Network level increase
• Continuous optimization automatically maintains the optimum settings
as the network evolves and traffic distributions change
in uplink throughput1 in downlink throughput
& reduction in DCR1

Based on real customer outcomes: 1. Service provider in Europe

Spectrum Utilization

UL Interference Optimizer
Cognitive SW rApp
AI recommender for UL interference optimization
Conflict resolution logic
Highly loaded cells are a frequent source of uplink interference causing user
experience degradation.
UL Interference Optimizer uses AI to learn from different configurations,
environments, neighbors etc. When a cell impacted by UL interference is
detected the optimizer automatically recommends optimum UL power control
parameters to resolve the issue.
Performance
prediction pZeroPUCCH
pZeroPUSCH
Multiple pattern & combination analysis alpha

Benefits & differentiators

• Automatic optimization of pZeroNominalPusch & alpha parameters in the cell
area to maximize UL SINR
• Machine learning powered optimization learns from the network to be able to
26% 7%
recommend the optimum settings UL SINR improvement Capacity gain
• Continuous optimization reacts to new problems whenever they occur
(UL spectral
efficiency improvement)
Traffic & Mobility Management

Inter-frequency Mobility Optimizer

Cognitive SW rApp
AI optimizer for handover optimization
Robust and correctly timed handovers directed at the most suitable target cells
are the key for consistent experience of moving subscribers. The Inter-Frequency
Mobility Optimizer uses AI to learn from different configurations, environments,
neighbors etc. When a cell is performing worse than comparable mobility
scenarios, the optimizer automatically proposes optimum one-shot handover
parameters updates, with focus on improving poor performance at cell edge.

Benefits & differentiators

• Automatic optimization of inter-frequency mobility (A2, A3, A5) event triggers

9% 6%
across different frequency layers
• Online machine learning component ensures alignment between
recommended settings & customer specific strategies
• One-shot parameter updates (weekly-monthly) on a limited set of cells that
ERAB drop rate DL throughput
yield maximum performance impact as predicted by AI improvement improvement
 Analytics

Performance Diagnostics
Cognitive SW rApp
Performance Diagnostics for AI powered troubleshooting
• AI powered automatic analytics for the entire RAN every day. Automatic
detection of >50 different problem classes and >200 root causes enriched with
AI reasoning contextualize a cell issue in relation to their neighbors. The
optimization team resolves more issues faster, directly boosting network
performance and minimizing degradations on user experience.
• Multi-vendor & Multi-technology support.

Benefits & differentiators

• AI driven quantitative causal analysis, explainable AI & reasoning
• Unique functionality in the market
• 75% faster issue detection and classification1 compared to traditional methods 30% 15%
• Detection of 50% more performance impacting issues2 Increase in capacity per Better downlink speed
• 92-98% field verified accuracy2,3 in issue detection and classification optimization FTE1 in cells with issues4

Based on real customer outcomes: 1. European operator, 2. European operator,

3. Japanese operator, 4. Aggregated gains
 Analytics

Virtualized Drive Test

rApp offering renamed as Geolocated Diagnostics
Cognitive SW rApp
Virtual Drive Test for advanced optimization
• VDT provides call trace-based analytics of every subscriber transaction with
the RAN. Combined with geolocation this allows proactive analysis of every
failure, investigation of customer complaints and performance degradations
as well as insights on terminal performance, traffic maps etc.
• Multi-vendor & Multi-technology support.

Benefits & differentiators

• AI powered Geolocation provides 4x location accuracy
• 100% visibility of all subscribers, terminals & traffic types, as opposed to
1 device along a limited drive test route
90% 70%
• 90% reduction in drive test1 activities & CO2 emissions. Reduction in drive Increase in customer
50% more cost efficient2
tests & related CO2 complaint handling
• GPS level location accuracy with Minimization of Drive
Tests (MDT) RAN feature
emissions1 capacity2

Based on real customer outcomes: 1. Operator in Southeast Asia, 2. Operator in Latin America
Energy Efficiency

RAN energy control/Intelligent RAN power saving

Intelligent Services rApp
RAN energy control/Intelligent RAN power saving
• Intelligently modelling the entire network’s cells based on the traffic behavior
such as data volume and RRC users. Using advanced ML techniques,
determining the power saving features activations and deactivations every
15 min autonomously across the network on a radio level granularity.
• Current features supported: Deep sleep and lock cell

Benefits & differentiators

25%
• 25% reduction of daily power consumption per radio with automatic and
intelligent activation and de-activation of power saving features such as
Radio Deep Sleep Zero impact
• Zero impact on network performance Reduction of daily on network
power consumption
for 5G*
performance

*Based on outcomes from Northeast Asia operator

Energy Efficiency

RAN energy cockpit rApp/radio power efficiency map

Intelligent Services rApp
RAN energy cockpit rApp/Radio Power Efficiency Map
• Measure energy efficiency per cell using energy consumption and traffic
metrics. Cluster and grade the sites reveals the inefficiency using a
color ranking.
• Then analyzing each cell against dimensioning, HW, SW and RF problems for
detailed understanding of the root cause of the inefficiency.

Benefits & differentiators

• Increased visibility of energy efficiency per site in the network
• Improved energy efficiency as well as network performance when addressing
the root cause of the inefficiency
Improve Improve
Network Energy
Performance Efficiency
Incident Management

Intelligent services – Insights

Intelligent Services rApp
Intelligent services – Insights
Intelligent services leverage AI technology to create and deliver customized use
case outcomes that automatically perform network wide or large-scale audit
work. The outcomes of the use cases are visible real-time via a web portal.
Examples of use cases are Node health check and SFP performance audit.

Benefits & differentiators

6 days
• Improved network KPI performance with improved handover success rate,
reduced drop call rate, improved throughput etc.
• Automates time consuming tasks, augmenting engineers so they can focus on
complex activities that require the power of the human brain From customer
Out-of-the-box
• Customized use cases to get insights to get deep understanding of what is requirement to use network insights
going on in the network which helps to prioritize network operations activities
case online
Healing of Cell Outage

Intelligent services – Assurance

Intelligent Services rApp
Intelligent services – Assurance
Intelligent services leverage AI technology to create and deliver customized use
case outcomes that automatically detects network anomalies and perform
automatic recovery actions. The outcomes of the use cases are visible real-time
via a web portal. Examples of use cases are sleeping cell and silent issue
detection and auto recovery.

Benefits & differentiators

5 min
• Increased operational efficiency with no human intervention and automate
troubleshooting activities
• Increased network quality by detecting and preempting issues before they
occur and provide automatic root cause remediation From data collection
Zero touch
• Near real-time data collection and automatic resolution to automatic resolution network stability
Healing of Network Services

Intelligent services – Self-Healing

Intelligent Services rApp
Intelligent services – Self-Healing
Intelligent services leverage AI technology to create and deliver customized use
case outcomes that automatically check symptoms and fingerprints to predict
and alert on potential system failures ahead of alarms. The outcomes of the use
cases are visible real-time via a web portal. Examples of use cases are SFP
failure prediction and Passive Hardware monitoring.

Benefits & differentiators

• Reduced network downtime by avoidance of service outages
• Reduced site visits and HW replacements by prediction of HW failures
• Reduced time spent trouble shooting by anticipating network issues and 35% Improved
provide recommendations according to the different root causes
Reduction of critical operational
incidents efficiency
Healing of Network Services

Intelligent services – Performance

Intelligent Services rApp
Intelligent services – Performance
Intelligent services leverage AI technology to create and deliver customized use
case outcomes that automatically alert on top offending cells in critical KPI's and
automate the fault isolation process. The outcomes of the use cases are visible
real-time via a web portal. Examples of use cases are KPI degradation fault
isolation and Hardware fault alarm fault isolation.

Benefits & differentiators

• Improved overall network performance by quick detection of KPI degradations
• Increased operational efficiency with closed loop automation and automation
of KPI troubleshooting procedures 60% Improved
• Isolate negative network-impacting events and give recommendations to
Reduction in network operational
resolve product near root-causes
performance issues efficiency
Healing of Cell Outage

Automated Ducting Mitigation rApp

Exploration rApp
Mitigate Ducting effect
Ducting can be seen as an “extreme distance overshooting” in which Downlink
traffic (aggressor) interferes with, and degrades, uplink traffic (victim) at far
distances
Using information from RAN counters, weather forecast, geospatial data, Duct Interference
network configuration and by means of AI technique, Automated Ducting
Long Propagation distance (10s to 100+ miles)
Mitigation (ADM) rApp will be able to predict and detect ducting conditions and
apply mitigation actions automatically

Benefits & differentiators

• Improve uplink performance of gNodeBs impacted by remote interferers from
tropospheric over-propagation (ducting effect) Uplink Throughput Reduce temporary
• Open/Closed-loop solution with automated mitigation actions: Change UL improvement effect of “ducting”
Power Control parameters/Force handover to healthy neighbor cell (cell border)
• Visualizes Remote Interference and connections on a map
effect in TDD
• Predicts ducting conditions and applies mitigation actions to reduce the impact
Uplink layer
• Complements Ericsson 5G NR Interference Mitigation features
Upgrade batch planning rApp
within 3-years
RAN rApp

Minimize traffic and capacity loss, such as:

rApp
— Schedule optimal time for re-starting upgraded nodes
— Planning based on number of impacted UEs allowed

Minimize coverage loss, such as:

— Identify “coverage nodes” to avoid large coverage holes
— Detect shared radios
— Group them into the same batch together with the affected
basebands
— Release compatibility, neighboring nodes, configuration
aspects, …

Minimum to no traffic
Benefits & differentiators Minimum to no
and capacity
— Minimize end-user impact at SW upgrade coverage loss at
coverage loss at
upgrade
upgrade
 Cognitive SW Trial in Central Asia
Virtual Drive Tests
• Proven very high correlation between VDT and Operator Drive-tests.
• Top overshooting cells and cross-connections are identified
• 60% faster drive-test report generation
• 80% traditional drive-test reduction
Software modules: • 50% faster site acceptance, enabling 20% increase in rollout capacity
▪ Capacity Planning • 70% increase on complaint handling capacity, with 33% improved
▪ Virtualized Drive-Test efficiency in optimization
▪ Performance Diagnostics
Performance Diagnostics
• 54% of the cells have some performance issues. Top three reasons are
MIMO performance, IRAT and Cell load.
Carrier U/EARFCN Bandwidth #Sector
Technology Carrier Band
Name -DL (MHz) Carriers
• HO IRAT & MIMO are dominant classes –more than 70% of all issues.
2100 1
• PIM & External Interference is also quite significant issue.
L21 150 10 4011
• 50% more issues identified, 75% faster compared to traditional
LTE 1800 3 L18 1302 10 5538
methods
800 20 L8 6200 10 4664
• 30% increase on operational efficiency for optimization
2100 1 U21-1 10562 5 1085
WCDMA • 98% field-validated accuracy on issue detection & classification
2100 1 U21-2 10587 5 1072

Capacity Planning
• 20% CAPEX savings compared traditional planning methods
• 10 % additional CAPEX savings by load balancing actions
• 80% faster capacity planning execution
• >70% accuracy predicting bottlenecks 6 months in advance
Value of AI for Telco is multifaceted and field-proven
Examples from production deployments

Capacity Planning
Reducing CAPEX 10% CAPEX savings
Operator in Malaysia
CAPEX and ROI
Capacity Planning -
Faster investment planning 80% faster network capacity planning
> Thailand
60% better DL throughput with 50% higher DL data traffic PD & Optimizers -> Qatar
Higher user throughput 50% better UL throughput with 130% higher UL data traffic PD -> Operator in LatAm
20% higher 5G utilization
Performance Diagnostics
Proactive Network KPI improvements 50% more issues detected, 75% faster
European Operator
Customer experience
and service excellence Performance Diagnostics
Leading network quality benchmarks 40% less bad quality cells
China operator
Performance Diagnostics
Reducing customer complaints 67% less customer complaints
Indonesia
Virtualized Drivetest & PD
Faster complaint resolution >80% faster complaint resolution
Operator in Latin Am.
Performance Diagnostics
Reduce OPEX 80% reduction in time spent per analysis
Far Eastone
OPEX and productivity
Virtualized Drivetest in
Increase productivity 30% boost in productivity
Denmark

Virtualized Drivetest
Sustainability Greener operations 100 tons CO2 less per year (600K Km virtual drivetest)
Southeast Asia