“Predictive Mobile Networks”

Slides:



Advertisements
Similar presentations
VSMC MIMO: A Spectral Efficient Scheme for Cooperative Relay in Cognitive Radio Networks 1.
Advertisements

LTE-A Carrier Aggregation
Overview.  UMTS (Universal Mobile Telecommunication System) the third generation mobile communication systems.
LTE pre-SI Test Suite Implementation in TTCN3: A case study 1 Pramod Shrivastava Intel Mobile Communication Nitin Bodas Larsen & Toubro Infotech Ranganath.
Multiple Criteria Optimisation for Base Station Antenna Arrays in Mobile Communication Systems By Ioannis Chasiotis PhD Student Institute for Communications.
College of Engineering Resource Management in Wireless Networks Anurag Arepally Major Adviser : Dr. Robert Akl Department of Computer Science and Engineering.
Self-Management for Unified Heterogeneous Radio Access Networks ISWCS 2015 Twelfth International Symposium on Wireless Communication Systems Brussels,
 The most intelligent device - “Human Brain”.  The machine that revolutionized the whole world – “computer”.  Inefficiencies of the computer has lead.
Interference Cancellation as a Mobile Enhancement to Improve Spectral Efficiency IEEE ComSoc Denver Chapter January 16, 2007.
Philipp Hasselbach Capacity Optimization for Self- organizing Networks: Analysis and Algorithms Philipp Hasselbach.
Millions of points of measurement Dense spatial and temporal data Need visual analytic tools as conventional analyses are too inefficient Visualization.
Session 2a, 10th June 2008 ICT-MobileSummit 2008 Copyright E3 project, BUPT Autonomic Joint Session Admission Control using Reinforcement Learning.
06/09/2005Master's Thesis Seminar - Jesse Kruus 1 Analyzing and Developing Base Load for WCDMA Base Station Automated Testing System Thesis written at.
Doc.: n-proposal-statistical-channel-error-model.ppt Submission Jan 2004 UCLA - STMicroelectronics, Inc.Slide 1 Proposal for Statistical.
1© Nokia Siemens Networks Confidential Realities of LTE Deployment Bill Payne Head of Innovation Team CTO Office.
All Rights Reserved, Copyright©2008, FUJITSU LIMITED. and FUJITSU LABORATORIES LIMITED. REV Technology Considerations for LTE-Advanced 3GPP TSG.
Cerion Optimization Services, Inc Dallas Parkway, Suite 404 Frisco, TX USA Optimiser TM Product Introduction This.
Architectures and Algorithms for Future Wireless Local Area Networks  1 Chapter Architectures and Algorithms for Future Wireless Local Area.
Business Id © NetHawk All rights reserved. Confidential April 2005NetHawk NetHawk Quality of Service products Markus Ahokangas, MSc Product.
Artificial Intelligence, Expert Systems, and Neural Networks Group 10 Cameron Kinard Leaundre Zeno Heath Carley Megan Wiedmaier.
INTRODUCTION:- The approaching 4G (fourth generation) mobile communication systems are projected to solve still-remaining problems of 3G (third generation)
Submission May 2016 H. H. LEESlide 1 IEEE Framework and Its Applicability to IMT-2020 Date: Authors:
5G. Overall Vision for 5G 5G will provide users with fiber-like access data rate and "zero" latency user experience be capable of connecting 100 billion.
ML in the Routers: Learn from and Act on Network Traffic Bing ietf95, April
Korean Intellectual Property Office – ICU seminar Ha, Jeongseok March 7, 2007 School of Engineering, Information and Communications University Wideband-CDMA.
Introduction to Machine Learning, its potential usage in network area,
MCA ECC PT1 Meeting
Optimisation of Radio Spectrum Usage
Guide of Genex Assistant for LTE
5 G.
Cost Effectively Deploying of Relay Stations (RS) in IEEE 802
LTE Long Term Evolution
Long Term Evolution (LTE) and System Architecture Evolution (SAE)
Success Stories.
Success Stories.
Proposal for Statistical Channel Error Model
Introduction Characteristics Advantages Limitations
Francisco da Silva, Senior Councillor, Huawei
OCC and LiFi based Light Communication for 5G Revolution
It’s All About Me From Big Data Models to Personalized Experience
11ax PAR Verification using UL MU-MIMO
1st Draft for Defining IoT (1)
System-Level simulation Inter-cell RRM Multi-cell RRM
LTE Long Term Evolution
Views for The LTE-Advanced Requirements
OCC and LiFi based Light Communication for 5G Revolution
LTE-A Relays and Repeaters
Evaluation Model for LTE-Advanced
Long Term Evolution (LTE)
Nortel Corporate Presentation
Emerging ICT needs – a Practitioners Perspective
AI emerging trend in QA Sanjeev Kumar Jha, Senior Consultant
New Adaptive Resource Allocation Scheme in LTE-Advanced
به نام خدا Big Data and a New Look at Communication Networks Babak Khalaj Sharif University of Technology Department of Electrical Engineering.
Mobile Synchronization Trends 4G to 4.5G to 5G
Speaker: I-LUN LEE ADVISOR: DR. HO-TING WU
Presented by Mohamad Haidar, Ph.D. May 13, 2009 Moncton, NB, Canada
Data collection methodology and NM paradigms
Dynamic Resource Scheduling Algorithm for Public Safety Network
OCC and LiFi based Light Communication for 5G Revolution
Integrating Deep Learning with Cyber Forensics
Francisco da Silva, Senior Councillor, Huawei
VHT in Below 6 GHz Frequency Bands
VHT in Below 6 GHz Frequency Bands
Christoph F. Eick: A Gentle Introduction to Machine Learning
LAA / Wi-Fi Coexistence evaluations with commercial hardware
Cellular and mobile communications (GSM, 3G and 4G (LTE))
The Intelligent Enterprise and SAP Business One
doc.: IEEE yy/xxxxr0 Date: September, 2019
Presentation transcript:

“Predictive Mobile Networks” Francisco Martín Pignatelli Group Head of Radio Product at Vodafone

Artificial Intelligence Today Classification Object Detection Tabby, tabby cat (57,03%) Tiger cat (14,6%) Washbasin, handbasin (9,6%) Egyptian cat (6,06%) Toilet seat (1,66%)

Artificial Intelligence in 4G and 5G A.I. is used in multiple industries: autonomous cars, weather prediction, voice, image, videos recognition… Machine Learning, is a subset of A.I that allows to identify patterns through the statistical data analysis Vodafone pioneers use of A.I. in Mobile Networks to evolve to a predictive network and automate complex tasks

Mobile Data growth has been folded in last 2 years Vodafone Group 225 TB 564 TB (1Q, 2016-2018)

AI will be essential in Networks Today 3-4 bands in use in 4G. In future, many more… And every band will use different technologies: 4G, 5G, FDD/TDD, 2x2 MIMO (2 transmitters), 4x4 MIMO (4 transmitters) or even Massive MIMO (32 or 64 Tx). Artificial Intelligence is necessary to solve the resources configuration, allocation and optimisation. 3.5 GHz (5G band) 2600 MHz TDD 2600 MHz 2100 MHz 1800 MHz 1400 MHz MaMIMO 900 MHz 800 MHz 700 MHz (5G)

How will it be implemented? Places where AI is located What’s AI in RAN?: The application of Machine Learning algorithms to big data to learn and automate actions. Machine Learning algorithms: Neural Network, Q-learning, Random Forest, etc Process time: eNB: milliseconds Servers: seconds/minutes Big data: eNB: all radio info Servers: counters, KPI, Traces (10 times less volume than eNB) Operational servers External Server Centralised SON Radio OSS Core OSS 2017 Field Trial & 2018 Commercial eNB Software Core Network NW elements 2017 PoC & 2018 Field Trial and commercial

What’s new? Problem areas: drop calls, interference, etc. LEARN: With every radio call info, the network learns: Problem areas: drop calls, interference, etc. Areas not managed efficiently: handover triggers, carrier aggregation combination, etc. PREDICT: The Network will estimate problems and inefficiencies when similar conditions to the learnt ones are happening. FIX per USER and per CELL before the issues happen, making changes such as fix the misconfiguration, change handover decisions and parameters… Drop predict

AI on the Radio from Evolution to Revolution 2017 SON Application 2018 eNB AI 2019 First 5G applications … 202x Self management features

World 1st Predictive Load pilot 1 hour prediction Problem: Traditional optimisation has 30-40mins delay due to counter cycle. New respect to traditional C-SON: AI predicts the traffic load 1 hour in advance after analysing several months counters. Trial Results: Load balanced before the load increases. +26% Accessibility +5.6% Voice Traffic +5.7% Data Traffic +5.8% Data Throughput Centralised SON

ML applied to VoLTE Optimisation New compared to traditional C- SON: ML algorithms (Dimension reduction, Q-learning and T-Test algorithms) analyse 400 counters and do a correlation of packet loss with Top 5 counters. Trial Result: Root cause analysis effort from 2.5 months of One Engineer to 4 hours Find the optimum parameter settings combination to reduce packet loss, improving VoLTE MOS quality. Overall result Top 10% worst cells Gain 11% Gain 33% Gain 22% Gain 81% Commercial C-SON July’18 Cenralised SONt

ML applied to cell edge user speed Problem: Cell edge scenarios are complex, too many parameters and many settings combinations. New respect to traditional C-SON: ML Algorithm (Q-learning) reduce configuration complexity of 6 categories / 100+ parameters. Trial Result: cell edge throughput improved. Power control Schedule GAP Coverage Resource Handover 6 parameter categories + 30% (0,86 to 1.1 Mbps) Overall + 49% (1,1 to 1.7 Mbps) Top 10% gain cells Centralised SON C-SON feature by July’18

AI @ eNB Radio 4G/5G Real NW Traces from London and Barcelona (Huawei). What’s new? AI algorithms learn per cell and per user what are the best frequencies to allocate, using Radio-frequency fingerprinting, without the UE to measure. Accuracy >80% when more than 10.000 samples for Handover; 76 to 99% accuracy for best Carrier Aggregation selection. PoC scope: Apply AI algorithms offline over real traces to estimate their performance gain and accuracy. Live data analysis IFHO from 2100 to 800 July’17 September’17 2Q-18 Traces Collection Evaluation of algo based on AI Features implementation: - IFHO, CA, Load Balancing, VoLTE and CSFB performance opt

First introduction of AI within eNB in 2H-2018 PoC result: Thousands of millions of inter-frequency handover in half time (1 second to 0.5 seconds). Adding automatically the right frequencies to do Carrier Aggregation, 10% higher average throughput. Load balancing, redirecting the terminals to the best cell in a faster way. A – 70 dBm B A – 95 dBm B – 102 dBm C – 100 dBm C Radio 4G/5G

Conclusions AI (Machine Learning) algorithms can learn in the network automatically, predicting the traffic, and anticipating the problems per Cell and per Device. Improvements in throughput, voice quality, OPEX troubleshooting, faster handover, more efficient carrier selection  and with precision achieved higher than 90% Vodafone will integrate the AI within the networks during 2018. These algorithms will be essential to manage, configure, and optimise the 4G and 5G network in the next years due to the traffic growth and therefore the deployment of new frequency bands.