Page 7 - 3GPP_Highlights_Issue_5_WEB

While specific AI/ML models are not expected to be specified and Protocol aspects to study along with potential specification
will be left to implementation, enabling AI/ML for air interface will impact include, e.g., capability indication, configuration and
require specification impacts at various levels which we briefly control procedures (training/inference), and management of data
describe next. and AI/ML model, and collaboration level specific specification
impact per use case.
In addition to the air interface enabling aspects involving physical
layer and protocol areas, the interoperability and testability (e.g., Lively discussions are already taking place in areas such as AI/
requirements and testing frameworks) of such implementations ML model training, i.e., offline vs. online, along with their actual
is an important aspect that will also be considered. The possible meaning and feasibility. Also, the concept of model transfer
need and implications for AI/ML processing capabilities definition has shown to be a controversial topic requiring more aligned
will be also assessed during the study. understanding of the implications and consequences.
Physical layer aspects to study along with potential specification To conclude, this study item spanning the entire duration of Rel-
impact include, e.g., AI model lifecycle management (LCM), 18 is expected to provide a solid understanding of 3GPP’s role in
dataset construction for training, validation and test for the enabling an improved support of AI/ML for air interface problems
given use case, new signaling required to enable specific use which is expected to lead to normative projects in future Releases
cases, means for training and validation, assistance information, of 5G-Advanced. At the same time, the findings of this project
measurement, and feedback. are expected to be leveraged in future generations of wireless
systems that 3GPP will develop.

For details on the Study on Artificial Intelligence (AI)/Machine Learning (ML) for NR Air Interface (FS_NR_AIML_Air)
see: RP-221348 and 3GPP TR38.843.

5GS SUPPORT FOR AI/ML

By Tricci So, Rapporteur for the 3GPP WG SA2
Study on 5G System Support for AI/ML-based

Services (FS_AIMLsys), OPPO.

The Release 18 (Stage 2) study TR 23.700-80 - Study on 5G system (5GS) support for AI/ML-based services - is based on the
Stage 1 requirements specified in TS 22.261 (Clauses 6.40 and 7.10) for 5GS assistance to support Artificial Intelligence (AI) /
Machine Learning (ML) model distribution, transfer, training for various applications, including video & speech recognition,
robot control and automotive.

The scope of the study is to enable AI/ML service providers to leverage the 5GS as the intelligent transmission platform to
assist data transfer during application layer AI/ML operation. It looks at the following possible 5GC extensions:

• Monitoring 5G network resource utilization relevant to the UE.
• Extending 5GC information exposure on the UE/network conditions and performance prediction (e.g. location, QoS, load,

congestion, etc.) to the application.
• Enhancements of external parameter provisioning to the 5G Core (e.g. expected UE positioning, expected UE mobility,

etc.) to assist Application AI/ML operation.
• Possible QoS, Policy enhancements to support Application AI/ML operational traffic while supporting regular (non

Application-AI/ML) 5GS user traffic.
• Assistance from 5G Core to the AF and the UE to coordinate and manage the Federated Learning (FL) operation (i.e. FL

members selection, group performance monitoring, adequate network resources allocation and guarantee) between the
application clients running on the UEs and the Application Servers.
3GPP TR 23.700-80 (2022-09) identifies the key issues faced and the potential solutions available for AI/ML-based services
over the 5G system.

|
Issue 05 - No v ember 2 0 22 07