TY - BOOK
T1 - 5G-PPP Technology Board
T2 - AI and ML – Enablers for Beyond 5G Networks
AU - Anastasopoulos, Markos
AU - Tzanakaki, Anna
AU - Srinivasan, Gokul Krishna
AU - Antevski, Kiril
AU - Baranda, Jordi
AU - De Schepper, Koen
AU - Casetti, Claudio Ettore
AU - Chiasserini, Carla Fabiana
AU - Garcia-Saavedra, Andres
AU - Guimarães , Carlos
AU - Kondepu, Koteswararao
AU - Li, Xi
AU - Magoula, Lina
AU - Malinverno, Marco
AU - Mangues-Bafalluy, Josep
AU - Martín-Pérez, Jorge
AU - Martínez, Ricardo
AU - Papagianni, Chrysa
AU - Valcarenghi, Luca
AU - Zeydan, Engin
AU - Alay, Özgü
AU - Aumayr, Erik
AU - Bosneag, Anne-Marie Cristina
AU - Caso, Giuseppe
AU - Jimeno, Elisa
AU - Diamanti, Maria
AU - Kakkavas, Grigorios
AU - Karyotis, Vasileios
AU - Papavassiliou, Symeon
AU - Stamou, Adamantia
AU - Bosneag, Anne-Marie Cristina
AU - Feghhi, Saman
AU - Mohamed, Ramy
AU - Xie, Min
AU - Zemouri, Sofiane
AU - Gramaglia, Marco
AU - Dreibholz, Thomas
AU - Elmokashfi, Ahmed
AU - Majumdar, Sayantini
AU - Roig, Joan Pujol
AU - Wang, Yue
AU - Poe, Wint Yi
AU - Lopez, Diego
AU - Mahmood, Kashif
AU - Behravesh, Rasoul
AU - Carrozzo, Gino
AU - Gil Pérez, Manuel
AU - Huertas Celdrán, Alberto
AU - Valero, José María Jorquera
AU - Lekidis, Alexios
AU - Martínez Pérez, Gregorio
AU - Sánchez, Pedro Miguel
AU - Subramanya, Tejas
AU - Martrat, Josep
AU - Trakadas, Panagiotis
AU - Giannopoulos, Anastasios
AU - Spantideas, Sotirios
AU - Yaqub, Edwin
AU - Desai, Rachana
AU - Klinkenberg, Ralf
AU - Boulogeorgos, Alexandros-Apostolos A.
AU - Alexiou, Angeliki
AU - He, Jiguang
AU - Katzouris, Nikos
AU - Lazarakis, Fotis
AU - Di Renzo, Marco
AU - Kokkoniemi, Joonas
AU - Bisson, Pascal
AU - Ayed, Dhouha
AU - Chollon, Geoffroy
AU - Skarmeta, Antonia
AU - Gürkan, Gür
AU - Benzaid, Chafika
AU - De Oca, Edgardo Montes
AU - Perales, Antonio Augustin Pastor
AU - Cosmas, John
AU - Meunier, Ben
AU - Shi, Lina
AU - Zhang, Xun
AU - Conti, Andrea
AU - Kennouche, Takai Eddine
AU - Morselli, Flavio
AU - Murphy, Chris
AU - Chergui, Hatim
AU - Devoti, Francesco
AU - Xu, Zhao
AU - Zanzi, Lanfranco
AU - Alcaraz Calero, Jose
AU - Wang, Qi
AU - Abdulkadir, MohammedA
AU - Katta, Saimanoj
AU - Costa-Requena, Jose
AU - Yesilkaya, Anil
AU - Videv, Stefan
AU - Prados-Garzon, Jonathan
AU - Chinchilla-Romero, Lorena
AU - Muñoz-Luengo, Pablo
AU - Ramos-Munoz, Juan J.
AU - Colman-Meixner, Carlos
AU - Zhou, Xueqing
AU - Yan, Shuangyi
AU - Aumayr, Erik
AU - Goodarzi, Meysam
AU - Cogalan, Tezcan
A2 - Mur, Daniel Camps
A2 - Gavras, Anastasius
A2 - Ghoraishi, Mir
A2 - Hrasnica, Halid
A2 - Kaloxylos, Alexandros
PY - 2021/5/21
Y1 - 2021/5/21
N2 - This white paper on AI and ML as enablers of beyond 5G (B5G) networks is based on contributions from 5G PPP projects that research, implement and validate 5G and B5G network systems. The white paper introduces the main relevant mechanisms in Artificial Intelligence (AI) and Machine Learning (ML), currently investigated and exploited for 5G and B5G networks. A family of neural networks is presented which are, generally speaking, non-linear statistical data modelling and decision-making tools. They are typically used to model complex relationships between input and output parameters of a system or to find patterns in data. Feed-forward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks belong to this family. Reinforcement learning is concerned about how intelligent agents must take actions in order to maximize a collective reward, e.g., to improve a property of the system. Deep reinforcement learning combines deep neural networks and has the benefit that is can operate on non-structured data. Hybrid solutions are presented such as combined analytical and machine learning modelling as well as expert knowledge aided machine learning. Finally, other specific methods are presented, such as generative adversarial networks and unsupervised learning and clustering. In the sequel the white paper elaborates on use case and optimisation problems that are being tackled with AI/ML, partitioned in three major areas namely, i) Network Planning, ii) Network Diagnostics/Insights, and iii) Network Optimisation and Control. In Network Planning, attention is given to AI/ML assisted approaches to guide planning solutions. As B5G networks become increasingly complex and multi-dimensional, parallel layers of connectivity are considered a trend towards disaggregated deployments in which a base station is distributed over a set of separate physical network elements which ends up in the growing number of services and network slices that need to be operated. This climbing complexity renders traditional approaches in network planning obsolete and calls for their replacement with automated methods that can use AI/ML to guide planning decisions. In this respect two solutions are discussed, first the network element placement problem is introduced which aims at improvements in the identification of optimum constellation of base stations each located to provide best network performance taking into account various parameters, e.g. coverage, user equipment (UE) density and mobility patterns (estimates), required hardware and cabling, and overall cost. The second problem considered in this regard is the dimensioning considerations for C-RAN clusters, in which employing ML-based algorithms to provide optimal allocation of baseband unit (BBU) functions (to the appropriate servers hosted by the central unit (CU)) to provide the expected gains is addressed. In Network Diagnostics, attention is given to the tools that can autonomously inspect the network state and trigger alarms when necessary. The contributions are divided into network characteristics forecasts solutions, precise user localizations methods, and security incident identification and forecast. The application of AI/ML methods in high-resolution synthesising and efficient forecasting of mobile traffic; QoE inference and QoS improvement by forecasting techniques; service level agreement (SLA) prediction in multi-tenant environments; and complex event recognition and forecasting are among network characteristics forecasts methods discussed. On high-precision user localization, AI-assisted sensor fusion and line-of-sight (LoS)/non-line-of-sight (NLoS) discrimination, and 5G localization based on soft information and sequential autoencoding are introduced. And finally, on forecasting security incidents, after a short introduction on modern attacks in mobile networks, ML-based network traffic inspection and real-time detection of distributed denial-of-service (DDoS) attacks are briefly examined. In regard to the Network Optimisation and Control, attention is given to the different network segments, including radio access, transport/fronthaul (FH)/backhaul (BH), virtualisation infrastructure, end-to-end 5G PPP Technology Board AI/ML for Networks 3 (E2E) network slicing, security, and application functions. Among application of AI/ML in radio access, the slicing in multi-tenant networks, radio resource provisioning and traffic steering, user association, demand-driven power allocation, joint MAC scheduling (across several gNBs), and propagation channel estimation and modelling are discussed. Moreover, these solutions are categorised (based on the application time-scale) into real-time, near-real-time, and non-real-time groups. On transport and FH/BH networks, AI/ML algorithms on triggering path computations, traffic management (using programmable switches), dynamic load balancing, efficient per-flow scheduling, and optimal FH/BH functional splitting are introduced. Moreover, federated learning across MEC and NFV orchestrators, resource allocation for service function chaining, and dynamic resource allocation in NFV infrastructure are among introduced AI/ML applications for virtualisation infrastructure. In the context of E2E slicing, several applications such as automated E2E service assurance, resource reservation (proactively in E2E slice) and resource allocation (jointly with slice-based demand prediction), slice isolation, and slice optimisation are presented. In regard to the network security, the application of AI/ML techniques in responding to the attack incidents are discussed for two cases, i.e. in moving target defence for network slice protection, and in self-protection against app-layer DDoS attacks. And finally, on the AI/ML applications in optimisation of application functions, the dash prefetching optimization and Q-learning applications in federated scenarios are presented.The white paper continues with the discussions on the application of AI/ML in the 5G and B5G network architectures. In this context the AI/ML based solutions pertaining to autonomous slice management, control and orchestration, cross-layer optimisation framework, anomaly detection, and management analytics, as well as aspects in AI/ML-as-a-service in network management and orchestration, and enablement of ML for the verticals' domain are presented. This is followed by topics on management of ML models and functions, namely the ML model lifecycle management, e.g., training, monitoring, evaluation, configuration and interface management of ML models. Furthermore, the white paper investigates the standardisation activities on the enablement of AI/ML in networks, including the definition of network data analytics function (NDAF) by 3GPP, the definition of an architecture that helps address challenges in network automation and optimization using AI and the categories of use cases where AI may benefit network operation and management by ETSI ENI, and finally the O-RAN definition of non-real-time and near-real-time RAN controllers to support ML-based management and intelligent RAN optimisation. Additionally, the white paper identifies the challenges in view of privacy and trust in AI/ML-based networks and potential solutions by introducing privacy preserving mechanisms and the zero-trust management approach are introduced. The availability of reliable data-sets as a crucial prerequisite to efficiency of AI/ML algorithms is discussed and the white paper concludes with a brief overview of AI/ML-based KPI validation and system troubleshooting. In summary the findings of this white paper conclude with the identification of several areas (research and development work) for further attention in order to enhance future network return-on-investment (ROI): (a) building standardized interfaces to access relevant and actionable data, (b) exploring ways of using AI to optimize customer experience, (c) running early trials with new customer segments to identify AI opportunities, (d) examining use of AI and automation for network operations, including planning and optimization, (e) ensuring early adoption of new solutions for AI and automation to facilitate introduction of new use cases, and (f) establish/launch an open repository for network data-sets that can be used for training and benchmarking algorithms by all.
AB - This white paper on AI and ML as enablers of beyond 5G (B5G) networks is based on contributions from 5G PPP projects that research, implement and validate 5G and B5G network systems. The white paper introduces the main relevant mechanisms in Artificial Intelligence (AI) and Machine Learning (ML), currently investigated and exploited for 5G and B5G networks. A family of neural networks is presented which are, generally speaking, non-linear statistical data modelling and decision-making tools. They are typically used to model complex relationships between input and output parameters of a system or to find patterns in data. Feed-forward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks belong to this family. Reinforcement learning is concerned about how intelligent agents must take actions in order to maximize a collective reward, e.g., to improve a property of the system. Deep reinforcement learning combines deep neural networks and has the benefit that is can operate on non-structured data. Hybrid solutions are presented such as combined analytical and machine learning modelling as well as expert knowledge aided machine learning. Finally, other specific methods are presented, such as generative adversarial networks and unsupervised learning and clustering. In the sequel the white paper elaborates on use case and optimisation problems that are being tackled with AI/ML, partitioned in three major areas namely, i) Network Planning, ii) Network Diagnostics/Insights, and iii) Network Optimisation and Control. In Network Planning, attention is given to AI/ML assisted approaches to guide planning solutions. As B5G networks become increasingly complex and multi-dimensional, parallel layers of connectivity are considered a trend towards disaggregated deployments in which a base station is distributed over a set of separate physical network elements which ends up in the growing number of services and network slices that need to be operated. This climbing complexity renders traditional approaches in network planning obsolete and calls for their replacement with automated methods that can use AI/ML to guide planning decisions. In this respect two solutions are discussed, first the network element placement problem is introduced which aims at improvements in the identification of optimum constellation of base stations each located to provide best network performance taking into account various parameters, e.g. coverage, user equipment (UE) density and mobility patterns (estimates), required hardware and cabling, and overall cost. The second problem considered in this regard is the dimensioning considerations for C-RAN clusters, in which employing ML-based algorithms to provide optimal allocation of baseband unit (BBU) functions (to the appropriate servers hosted by the central unit (CU)) to provide the expected gains is addressed. In Network Diagnostics, attention is given to the tools that can autonomously inspect the network state and trigger alarms when necessary. The contributions are divided into network characteristics forecasts solutions, precise user localizations methods, and security incident identification and forecast. The application of AI/ML methods in high-resolution synthesising and efficient forecasting of mobile traffic; QoE inference and QoS improvement by forecasting techniques; service level agreement (SLA) prediction in multi-tenant environments; and complex event recognition and forecasting are among network characteristics forecasts methods discussed. On high-precision user localization, AI-assisted sensor fusion and line-of-sight (LoS)/non-line-of-sight (NLoS) discrimination, and 5G localization based on soft information and sequential autoencoding are introduced. And finally, on forecasting security incidents, after a short introduction on modern attacks in mobile networks, ML-based network traffic inspection and real-time detection of distributed denial-of-service (DDoS) attacks are briefly examined. In regard to the Network Optimisation and Control, attention is given to the different network segments, including radio access, transport/fronthaul (FH)/backhaul (BH), virtualisation infrastructure, end-to-end 5G PPP Technology Board AI/ML for Networks 3 (E2E) network slicing, security, and application functions. Among application of AI/ML in radio access, the slicing in multi-tenant networks, radio resource provisioning and traffic steering, user association, demand-driven power allocation, joint MAC scheduling (across several gNBs), and propagation channel estimation and modelling are discussed. Moreover, these solutions are categorised (based on the application time-scale) into real-time, near-real-time, and non-real-time groups. On transport and FH/BH networks, AI/ML algorithms on triggering path computations, traffic management (using programmable switches), dynamic load balancing, efficient per-flow scheduling, and optimal FH/BH functional splitting are introduced. Moreover, federated learning across MEC and NFV orchestrators, resource allocation for service function chaining, and dynamic resource allocation in NFV infrastructure are among introduced AI/ML applications for virtualisation infrastructure. In the context of E2E slicing, several applications such as automated E2E service assurance, resource reservation (proactively in E2E slice) and resource allocation (jointly with slice-based demand prediction), slice isolation, and slice optimisation are presented. In regard to the network security, the application of AI/ML techniques in responding to the attack incidents are discussed for two cases, i.e. in moving target defence for network slice protection, and in self-protection against app-layer DDoS attacks. And finally, on the AI/ML applications in optimisation of application functions, the dash prefetching optimization and Q-learning applications in federated scenarios are presented.The white paper continues with the discussions on the application of AI/ML in the 5G and B5G network architectures. In this context the AI/ML based solutions pertaining to autonomous slice management, control and orchestration, cross-layer optimisation framework, anomaly detection, and management analytics, as well as aspects in AI/ML-as-a-service in network management and orchestration, and enablement of ML for the verticals' domain are presented. This is followed by topics on management of ML models and functions, namely the ML model lifecycle management, e.g., training, monitoring, evaluation, configuration and interface management of ML models. Furthermore, the white paper investigates the standardisation activities on the enablement of AI/ML in networks, including the definition of network data analytics function (NDAF) by 3GPP, the definition of an architecture that helps address challenges in network automation and optimization using AI and the categories of use cases where AI may benefit network operation and management by ETSI ENI, and finally the O-RAN definition of non-real-time and near-real-time RAN controllers to support ML-based management and intelligent RAN optimisation. Additionally, the white paper identifies the challenges in view of privacy and trust in AI/ML-based networks and potential solutions by introducing privacy preserving mechanisms and the zero-trust management approach are introduced. The availability of reliable data-sets as a crucial prerequisite to efficiency of AI/ML algorithms is discussed and the white paper concludes with a brief overview of AI/ML-based KPI validation and system troubleshooting. In summary the findings of this white paper conclude with the identification of several areas (research and development work) for further attention in order to enhance future network return-on-investment (ROI): (a) building standardized interfaces to access relevant and actionable data, (b) exploring ways of using AI to optimize customer experience, (c) running early trials with new customer segments to identify AI opportunities, (d) examining use of AI and automation for network operations, including planning and optimization, (e) ensuring early adoption of new solutions for AI and automation to facilitate introduction of new use cases, and (f) establish/launch an open repository for network data-sets that can be used for training and benchmarking algorithms by all.
M3 - Commissioned report
VL - 1.0
BT - 5G-PPP Technology Board
PB - European Commission
ER -