|
Work item:
|
Y.AI_HNO_ACPS
|
|
Subject/title:
|
Requirements and framework for AI/ML-based heterogeneous network optimization and automated communication primitive selection in future networks including IMT-2020 and beyond
|
|
Status:
|
Under study
|
|
Approval process:
|
AAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2027-Q4 (Medium priority)
|
|
Liaison:
|
ITU-T SG2, SG11, 3GPP SA, ETSI
|
|
Supporting members:
|
China Unicom, ICT CAS, BUPT, CICT, MIIT (China), China Mobile, China Telecommunications, ZTE, India (Republic of), TRAI,Mobile Communication Company of Iran (MCI), Iran University of science and Technology (IUST)
|
|
Summary:
|
In this Recommendation, the concept of heterogeneous network refers to a complex core network architecture characterized by the coexistence and collaborative operation of multiple structurally diverse core networks within a unified network environment. This environment may encompass core networks associated with IMT-Advanced, IMT-2020, as well as fixed network.
In the operation of heterogeneous networks, significant challenges arise in policy collaboration, capacity allocation, routing efficiency, and topology optimization due to the coexistence and interaction of multiple structurally diverse core networks. This Recommendation addresses an AI/ML-based network optimization approach for heterogeneous network to address above challenges by utilizing collaborative leaning capabilities (such as federated learning, knowledge distillation, feature fusion, model fusion, and policy fusion) and joint optimization capabilities (such as policy collaboration, capacity optimization, routing optimization, and topology optimization).
This Recommendation also aims to support AI/ML based dynamic communication primitive selection in future networks including IMT-2020 and beyond networks. The targeted AI/ML framework will learn from the base communication primitives and current network conditions to select a new set of optimized communication primitives, which will be applied to the underlay IMT-2020 and beyond networks. By applying AI/ML based communication primitive selection, the networks can reduce redundant processing and work more efficiently to achieve the user/service and system requirements.
Based on the architectural framework for machine learning in future networks including IMT-2020, as defined in [ITU-T Y.3172], and the framework for AI/ML-based network design optimization specified in [ITU-T Y.3142], this Recommendation aims to further address the requirements and framework for AI/ML-based heterogeneous network optimization and automated communication primitive selection.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2025-03-19 10:32:44
|
|
Last update:
2025-04-23 18:32:00
|
|
