​​​Networks beyond IMT-2020 and machine learning: Requirements and architecture

(Continuation of Question Q20/13)

Motivation

​​The objective of this question is to study the requirements, architecture and use of technologies including artificial intelligence (AI)/machine learning (ML) to realize networks beyond IMT-2020, in order to address the anticipated needs of network and application services in the upcoming years.

Network requirements and architecture for IMT-2020 networks have been baselined and successful deployments have been reported since its inception in the early 2010s. The next generation of IMT networks (following IMT-2020 networks) is already under study in many countries. Considering that a new generation network is commercialized around every 10 years, the next generation of IMT networks is expected to be deployed around 2030. It is the right time to study the requirements and architecture of networks beyond IMT-2020.

IMT-2020 has gone through several major paradigm shifts in network technologies such as the adoption of network slicing and service-based architecture. However, there are still many aspects to improve in the current architecture. An evolutionary approach can be sought on the current generation of IMT networks to address some remaining issues. A network is not a simple packet delivery system anymore; it is becoming a neural system of our society. To meet the requirements and derive necessary architecture enhancements, consideration should be given to key aspects of networks beyond IMT-2020.

Integration of AI/ML applications is also regarded as one of the key architectural aspects to consider for networks beyond IMT-2020. The complexity coming from distributed architecture and heterogeneous nature of use cases makes it imperative to study the service requirements and overheads related to the AI/ML applications. A comprehensive study of the impact, KPIs and evaluation of AI/ML applications is a must for the design of network architecture. The study should also include test methodologies and deployment guidelines for AI/ML applications in the networks.

In summary, this question focuses on the study of the requirements, architecture and use of technologies including AI/ML to realize networks beyond IMT-2020.

The following major Recommendations, in force at the time of approval of this Question, fall under its responsibility:
    –           Y.3100, Y.3101, Y.3102 and Y.3104
    –           Y.3172, Y.3173 and Y.3174.​

Study items
Study items to be considered include, but are not limited to:​

• What are the key requirements and capabilities of networks beyond IMT-2020  including AI/ML based on the emerging service scenarios?
• What framework and architecture are required to realize networks beyond IMT-2020 including AI/ML based on the identified requirements and capabilities?
• What key technologies related to networks beyond IMT-2020 including AI/ML are required to realize the networks?
• How to incorporate network intelligence from AI/ML into networks beyond IMT-2020?
• How to build and/or guide the ecosystem on networks beyond IMT-2020 including AI/ML taking into account business models and use cases?
• How to utilize and guide the open source software activities related to networks beyond IMT-2020 and AI/ML to meet the requirements of the networks?​