This Technical Report will study, based on requirements as specified in clauses 6.40 and 7.10 of TS 22.261, 5GS assistance to support Artificial Intelligence (AI) / Machine Learning (ML) model distribution, transfer, training for various applications, e.g. video/speech recognition, robot control, automotive, etc.
The scope of this study is on how the AI/ML service providers could leverage 5GS as the platform to provide the intelligent transmission support for application layer AI/ML operation based on the following objectives.
Study the possible architectural and functional extensions to support the Application layer AI/ML operations defined in TS 22.261, more specifically:
Support monitoring of network resource utilization in the 5G system relevant to the UE in order to support Application AI/ML operation.
Whether and how to extend 5GS information exposure for 5GC NF(s) to expose UE and/or network conditions and performance prediction (e.g. location, QoS, load, congestion, etc.) to the UE and/or to the authorized 3rd party to assist the Application AI/ML operation.
Enhancements of external parameter provisioning to 5GC (e.g. expected UE activity behaviours, expected UE mobility, etc.) based on Application AI/ML operation.
Investigate the enhancements of other 5GC features that could be used to assist the Application AI/ML operations as described in clause 6.40 of TS 22.261.
Study possible QoS, Policy enhancements to support Application AI/ML operational traffic while supporting regular (non Application-AI/ML) 5GS user traffic.
Study whether and how 5GS provides assistance to AF and the UE for the AF and UE to manage the Federated Learning (FL) operation and model distribution/redistribution (i.e. FL members selection, group performance monitoring, adequate network resources allocation and guarantee) to facilitate collaborative Application AI/ML based Federated Learning operation between the application clients running on the UEs and the Application Servers.
The following documents contain provisions which, through reference in this text, constitute provisions of the present document.
References are either specific (identified by date of publication, edition number, version number, etc.) or non-specific.
For a specific reference, subsequent revisions do not apply.
For a non-specific reference, the latest version applies. In the case of a reference to a 3GPP document (including a GSM document), a non-specific reference implicitly refers to the latest version of that document in the same Release as the present document.
For the purposes of the present document, the terms given in TR 21.905 and the following apply. A term defined in the present document takes precedence over the definition of the same term, if any, in TR 21.905.
For the purposes of the present document, the abbreviations given in TR 21.905 and the following apply. An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in TR 21.905.
The Application AI/ML operation logic is controlled by an Application Function (AF).
Any AF request to the 5G System in the context of 5GS assistance to Application AI/ML operation should be authorized by the 5GC.
The 5G System should ensure the performance KPIs specified in clause 7.10 of TS 22.261 to support the Application AI/ML traffic.
To support AI/ML based services/applications via 5GS, the following architectural assumptions are made in the present study:
In Rel-18, an Application AI/ML operation is conducted within a single slice, i.e. all UEs which are involved in a given Application AI/ML operation are served by the same S-NSSAI and the Application Function (AF) belongs to this S-NSSAI.
In Rel-18, roaming is not supported, i.e. inter PLMN coordination aspects will not be studied.
5GC can differentiate Application Layer AI/ML traffic by using existing mechanisms defined in TS 23.501.