Methods and apparatus for native 3GPP support of artificial intelligence and machine learning operations
Methods and apparatus for supporting artificial intelligence and machine learning operations in a 3GPP communication network are provided. A method may include receiving, by a WTRU, an indication of requirements and/or constraints associated with one or more WTRU applications. Based on the requirements and/or constraints associated with the one or more WTRU applications, the method may include determining one or more machine learning (ML) based analytics, data, and/or predictions to request from a network and/or internal WTRU ML modules. The method may then include transmitting, to the network and/or the internal WTRU ML modules, a request or subscription for the one or more ML based analytics, data, and/or predictions, and receiving a response or notification of the one or more ML based analytics, data, and/or predictions from the network and/or the internal WTRU ML modules.
1 . A Wireless Transmit/Receive Unit (WTRU), comprising:
a transceiver; and
a processor configured to:
receive an indication of any of a requirement and a constraint associated with one or more WTRU applications;
based on any of the requirement and the constraint associated with the one or more WTRU applications, determine one or more machine learning (ML)-based analytics to request or subscribe from a network element;
send, to the network element, a request or a subscription for the one or more ML-based analytics; and
receive a response indicating the one or more ML-based analytics from the network element; and
a protocol data unit (PDU) session modifier (PSM) entity configured to modify a PDU session associated with the one or more WTRU applications, wherein the PSM is configured to:
receive one or more application resource requirements from at least one of the one or more WTRU applications,
select at least one prediction module relevant to a determination of suitable PDU session parameters for a PDU session associated with the at least one of the one or more WTRU applications, and
activate the at least one selected prediction module to generate a prediction of the suitable PDU session parameters.
2 . The WTRU of claim 1 , wherein the request or the subscription is sent to the network element via any of non-access stratum (NAS) signaling or a user plane (UP) protocol.
3 . The WTRU of claim 1 , wherein any of the requirement and the constraint comprises any of:
quality of service (QOS) requirements;
a latency;
a bandwidth;
an availability; or
a reliability.
4 . The WTRU of claim 1 , wherein the one or more ML-based analytics are associated with any of:
an availability of computational resources;
a network capacity;
a signal strength;
a signal quality; or
a battery status.
5 . The WTRU of claim 1 , comprising:
a predictor engine coordinator (PEC) entity configured to:
generate the request or the subscription; and
send the generated request or the generated subscription to the network element via non-access stratum (NAS) signaling.
6 . The WTRU of claim 5 , wherein the PEC entity is configured to operate as part of a NAS-session management (NAS-SM) layer or to operate standalone on a layer above a NAS-mobility management (NAS-MM) layer.
7 . The WTRU of claim 1 , comprising one or more internal WTRU ML modules configured to generate the one or more ML-based analytics relating to operation of the WTRU, wherein the one or more internal WTRU ML modules comprise any of:
a ML module configured to generate a prediction of available bit rate at the WTRU;
a ML module configured to generate a prediction of a mobility condition of the WTRU; or
a ML module configured to generate a prediction of a usage rate of the processor.
8 . The WTRU of claim 1 , wherein the PSM is configured to modify the PDU session to satisfy a capacity requirement of an associated application or to improve usage of WTRU resources.
9 . A method implemented by a Wireless Transmit/Receive Unit (WTRU), the method comprising:
receiving an indication of any of a requirement and a constraint associated with one or more WTRU applications;
based on any of the requirement and the constraint associated with the one or more WTRU applications, determining one or more machine learning (ML)-based analytics to request or subscribe from a network element;
sending, to the network element, a request or a subscription for the one or more ML-based analytics;
receiving a response indicating the one or more ML-based analytics from the network element;
modifying, by a protocol data unit (PDU) session modifier (PSM) entity, a PDU session associated with the one or more WTRU applications;
receiving, by the PSM, one or more application resource requirements from at least one of the one or more WTRU applications;
selecting, by the PSM, at least one prediction module relevant to a determination of suitable PDU session parameters for a PDU session associated with the at least one of the one or more WTRU applications; and
activating the at least one selected prediction module to generate a prediction of the suitable PDU session parameters.
10 . The method of claim 9 , wherein the sending comprises sending the request or the subscription to the network element via any of non-access stratum (NAS) signaling or a user plane (UP) protocol.
11 . The method of claim 9 , wherein any of the requirement and the constraint comprises any of:
quality of service (QOS) requirements;
a latency;
a bandwidth;
an availability; or
a reliability.
12 . The method of claim 9 , wherein the one or more ML-based analytics are associated with any of:
an availability of computational resources;
a network capacity;
a signal strength;
a signal quality; or
a battery status.
13 . The method of claim 9 , comprising:
generating, by a predictor engine coordinator (PEC) entity, the request or the subscription; and
sending, by the PEC, the generated request or the generated subscription to the network element via non-access stratum (NAS) signaling.
14 . The method of claim 13 , wherein the PEC entity is configured to operate as part of a NAS-session management (NAS-SM) layer or to operate standalone on a layer above a NAS-mobility management (NAS-MM) layer.
15 . The method of claim 9 , comprising generating, by the one or more internal WTRU ML modules, the one or more ML-based analytics, data, or predictions relating to operation of the WTRU, wherein the one or more internal WTRU ML modules comprise any of:
a ML module configured to generate a prediction of available bit rate at the WTRU;
a ML module configured to generate a prediction of a mobility condition of the WTRU; or
a ML module configured to generate a prediction of a usage rate of a processing unit of the WTRU.
16 . The method of claim 9 , comprising modifying, by the PSM, the PDU session to satisfy a capacity requirement of an associated application or to improve usage of WTRU resources.
17 . A non-transitory computer readable medium comprising program instructions stored thereon, wherein the program instructions are configured to control an apparatus to:
receive an indication of any of a requirement and a constraint associated with one or more WTRU applications;
based on any of the requirement and the constraint associated with the one or more WTRU applications, determine one or more machine learning (ML)-based analytics to request or subscribe from a network element;
send, to the network element, a request or a subscription for the one or more ML-based analytics; and
receive a response indicating the one or more ML-based analytics from the network element;
modify, by a protocol data unit (PDU) session modifier (PSM) entity, a PDU session associated with the one or more WTRU applications;
receive, by the PSM, one or more application resource requirements from at least one of the one or more WTRU applications;
select, by the PSM, at least one prediction module relevant to a determination of suitable PDU session parameters for a PDU session associated with the at least one of the one or more WTRU applications; and
activate the at least one selected prediction module to generate a prediction of the suitable PDU session parameters.