Supporting creation and / or training of a learning model for a user equipment

Abstract

Various aspects of the present disclosure relate to supporting creation and / or training of a learning model for a user equipment. A core network entity may receive, from a radio access network entity, data of a network-side model associated with the RAN entity or a storage location of the data. The data may include one or more of: a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model. The core network entity may output the received data or the received storage location, for supporting creation and / or training of a learning model for a UE.

Description

SUPPORTING CREATION AND / OR TRAINING OF A LEARNING MODEL FOR A USER EQUIPMENTTECHNICAL FIELD

[0001] The present disclosure relates to wireless communications, and more specifically to supporting creation and / or training of a learning model for a user equipment (UE).BACKGROUND

[0002] A wireless communications system may include one or multiple network communication devices, such as base stations, which may support wireless communications for one or multiple user communication devices, which may be otherwise known as UE, or other suitable terminology. The wireless communications system may support wireless communications with one or multiple user communication devices by utilizing resources of the wireless communications system (e.g., time resources (e.g., symbols, slots, subframes, frames, or the like) or frequency resources (e.g., subcarriers, carriers, or the like). Additionally, the wireless communications system may support wireless communications across various radio access technologies including third generation (3G) radio access technology, fourth generation (4G) radio access technology, fifth generation (5G) radio access technology, among other suitable radio access technologies beyond 5G (e.g., sixth generation (6G)).

[0003] Artificial intelligence (AI) / machine learning (ML) can be applied in wireless communications. For some use cases, the network and the UE may collaborate to perform AI / ML operations.

[0004] 3GPP Technical Report (TR) 38.843 V18.0.0 (2023-12), which relates to a study on AI / ML for new radio (NR) air interface, discusses AI / ML model life cycle management (clause 4.2).SUMMARY

[0005] An article “a” before an element is unrestricted and understood to refer to “at least one” of those elements or “one or more” of those elements. The terms “a,” “at least one,” “one or more,” and “at least one of one or more” may be interchangeable. As used herein, including in the claims, “or” as used in a list of items (e.g., a list of items prefaced by a phrase such as “at least one of’ or “one or more of’ or “one or both of’) indicates an inclusive list such that, for example, a list of at least one of A, B, or C means A or B or C or AB or AC or BC or ABC (i.e., A and B and C). Also, as used herein, the phrase “based on” shall not be construed as a reference to a closed set of conditions. For example, an example step that is described as “based on condition A” may be based on both a condition A and a condition B without departing from the scope of the present disclosure. In other words, as used herein, the phrase “based on” shall be construed in the same manner as the phrase “based at least in part on”. Further, as used herein, including in the claims, a “set” may include one or more elements.

[0006] Implementations of the methods, systems, and apparatuses described herein support creation and / or training of a learning model for a UE.

[0007] In a first aspect, some implementations of the methods, systems, and apparatuses described herein may be performed or implemented by a first core network (CN) entity. The first CN entity may have at least one memory, and at least one processor coupled with the at least one memory and configured to cause the first CN entity to perform various steps, operations, etc.

[0008] Some implementations of the methods, systems, and apparatuses described herein may include, in the first aspect: receiving, from a radio access network (RAN) entity, data of a network-side model associated with the RAN entity or a storage location of the data; and outputting the received data or the received storage location, for supporting one or more of creating or training a learning model for a user equipment (UE). The data of the network-side model may include one or more of: a set of parameters of the networkside model, a set of training data of the network-side model, or a set of inference data of the network-side model. The learning model is an artificial intelligent (AI) / machine learning (ML) model and can be used as a UE-side model of the UE. The received data or the received storage location may be output to a second CN entity, which may in turn obtainand provide the data to an entity associated with the UE (e.g., a UE-side server) to create and / or train the learning model. In some implementations, the learning model (UE-side model) and the network-side model may belong to a two-sided model, in particular a two- sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the network-side model may be a decoder model. The encoder model and the decoder model may support encoding (UE side) and reconstruction (network-side) of channel state information (CSI).

[0009] Some implementations of the methods, systems, and apparatuses described herein may include, in the first aspect: prior to receiving the data of the network-side model from the RAN entity: receiving a first request to obtain the data; and outputting, to the RAN entity, a second request to provide the data. The RAN entity may be associated with the UE. The second request may include information contained in the first request. For example, the first request may be received from a second CN entity.

[0010] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the information in the first request may include an identifier of the network-side model.

[0011] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the information in the first request may include information indicating one or more required data types of the data. For example, the information may indicate that one or more of the following data types are required (i.e., need to be obtained from the RAN entity): a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model.

[0012] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the information in the first request may include one or more of: a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE.

[0013] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the information in the first request may include an area of interest associated with at least one RAN entity.

[0014] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the at least one processor may be configured to cause the first CN entity to: output the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity. The existing association may be configured to convey signaling to the UE. For example, the existing association may include at least one communication link (e.g., N2 link).

[0015] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the information in the first request may include the data request for the RAN entity. Some of these implementations of the methods, systems, and apparatuses may include obtaining an identity of the RAN entity, establishing an association between the first CN entity and the RAN entity, and outputting the second request to the RAN entity based at least in part on the established association. For example, the association may include at least one communication link, which may be used for non- UE associated signaling. The RAN entity may be the RAN entity serving the UE.

[0016] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the information in the first request may include the identifier of the UE. Some of these implementations of the methods, systems, and apparatuses may include determining presence of non-access stratum (NAS) connection between the UE and the first CN entity, or paging the UE to establish a NAS connection between the UE and the first CN entity; and outputting the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity. The existing association may be configured to convey signaling to the UE. For example, the existing association may include at least one communication link (e.g., N2 link).

[0017] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the information in the first request may include the data request for the RAN entity, the identifier of the UE, and the indication to not obtain any data from the UE. Some of these implementations of the methods, systems, and apparatuses may include outputting the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity. The existing association maybe configured to convey signaling to the UE. For example, the existing association may include at least one communication link (e.g., N2 link).

[0018] In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the second request may further include information for the UE. The information for the UE may include an indication to establish a NAS connection for radio measurement collection.

[0019] In some implementations described herein, in the first aspect: the first CN entity may be configured to implement an application management function (AMF). In some implementations of the methods, systems, and apparatuses described herein, in the first aspect: the second CN entity may be configured to implement a data collection network function.

[0020] In a second aspect, some implementations of the methods, systems, and apparatuses described herein may be performed or implemented by a second CN entity. The second CN entity may have at least one memory, and at least one processor coupled with the at least one memory and configured to cause the second CN entity to perform various steps, operations, etc.

[0021] Some implementations of the methods, systems, and apparatuses described herein may include, in the second aspect: receiving a first request associated with a learning model for a UE; obtaining, based at least in part on the first request, data of a network-side model; and outputting the obtained data, for supporting one or more of creating or training of the learning model. The data of the network-side model may include one or more of: a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model. The learning model is an AI / ML model and can be used as a UE-side model of the UE. For example, the first request may be received from an entity associated with the UE (e.g., UE-side server). For example, the obtained data may be output to the entity associated with the UE (e.g., UE-side server). The entity associated with the UE (e.g., UE-side server) may create and / or train the learning model based at least in part on the data. In some implementations, the learning model (UE-side model) and the network-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model)may be an encoder model and the network-side model may be a decoder model. The encoder model and the decoder model may support encoding (UE side) and reconstruction (network-side) of CSI.

[0022] In some implementations of the methods, systems, and apparatuses described herein, in the second aspect: the first request may include one or more of: an identifier of the network-side model, information indicating one or more required data types of the data, an identifier of the UE, or an area of interest associated with at least one RAN entity for obtaining the data. For example, the information may indicate that one or more of the following data types are required (i.e., need to be obtained from the RAN entity): a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model.

[0023] In some implementations of the methods, systems, and apparatuses described herein, in the second aspect: the first request may include the identifier of the UE. Some of these implementations of the methods, systems, and apparatuses may include determining, based at least in part on the first request, a first CN entity or a RAN entity associated with (e.g., serving) the UE; and outputting, to the first CN entity, a second request to obtain the data from the RAN entity associated with the UE.

[0024] In some implementations of the methods, systems, and apparatuses described herein, in the second aspect: the first request may include the area of interest. Some of these implementations of the methods, systems, and apparatuses may include determining, based at least in part on the first request, a first CN entity associated with (e.g., serving) the area of interest; determining, based at least in part on the first CN entity, at least one UE associated with (e.g., being served by) the first CN entity, wherein the at least one UE includes the UE; and outputting, to the first CN entity, a second request to obtain the data from the RAN entity associated with the UE.

[0025] In some implementations of the methods, systems, and apparatuses described herein, in the second aspect: the second request may include one or more of: a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE.

[0026] Some implementations of the methods, systems, and apparatuses described herein may include, in the second aspect: obtaining the data from one or more of: the at least one memory of the second CN entity or at least one memory of another CN entity.

[0027] In some implementations described herein, in the second aspect: the second CN entity may be configured to implement a data collection network function. In some implementations described herein, in the second aspect: the first CN entity may be configured to implement an AMF. In some implementations described herein, in the second aspect: the another CN entity may be configured to implement a data storage network function.

[0028] In a third aspect, some implementations of the methods, systems, and apparatuses described herein may be performed or implemented by a RAN entity. The RAN entity may have at least one memory, and at least one processor coupled with the at least one memory and configured to cause the RAN entity to perform various steps, operations, etc.

[0029] Some implementations of the methods, systems, and apparatuses described herein may include, in the third aspect: receiving a request to provide data of a networkside model associated with the RAN entity, wherein the data comprises one or more of a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model; and outputting, to a first CN entity, the data or a storage location of the data, for supporting one or more of creating or training a learning model for a UE. The learning model is an AI / ML model and can be used as a UE- side model of the UE. In some implementations, the learning model (UE-side model) and the network-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the network-side model may be a decoder model. The encoder model and the decoder model may support encoding (UE side) and reconstruction (network-side) of CSI.

[0030] Some implementations of the methods, systems, and apparatuses described herein may include, in the third aspect: the at least one processor may be configured to cause the RAN entity to receive the request from the UE.

[0031] Some implementations of the methods, systems, and apparatuses described herein may include, in the third aspect: the at least one processor may be configured to cause the RAN entity to receive the request from the first CN entity.

[0032] In some implementations of the methods, systems, and apparatuses described herein, in the third aspect: the request may include one or more of: an identifier of the network-side model, or information indicating one or more required data types of the data. For example, the information may indicate that one or more of the following data types are required (i.e., need to be obtained from the RAN entity): a set of parameters of the networkside model, a set of training data of the network-side model, or a set of inference data of the network-side model.

[0033] In some implementations of the methods, systems, and apparatuses described herein, in the third aspect: the request may include one or more of a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE.

[0034] Some implementations of the methods, systems, and apparatuses described herein may include, in the third aspect: receiving, from the UE, radio measurement obtained by the UE; and outputting, to the first CN entity, the radio measurement along with the data or the storage location of the data.

[0035] In some implementations described herein, in the third aspect: the first CN entity may be configured to implement an AMF.

[0036] In a fourth aspect, some implementations of the methods, systems, and apparatuses described herein may be performed or implemented by a UE. The UE may have at least one memory, and at least one processor coupled with the at least one memory and configured to cause the UE to perform various steps, operations, etc.

[0037] Some implementations of the methods, systems, and apparatuses described herein may include, in the fourth aspect: receiving, from a RAN entity, information associated with a network-side model; determining that one or more of creating or training a learning model for the UE is required; and outputting, to the RAN entity, a request to provide data of the network-side model, for supporting one or more of creating or training the learning model. The data of the network-side model includes one or more of a set ofparameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model. The learning model is an AI / ML model and can be used as a UE-side model of the UE. In some implementations, the learning model (UE-side model) and the network-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the network-side model may be a decoder model. The encoder model and the decoder model may support encoding (UE side) and reconstruction (NW-side) of CSI. In some implementations, the request may include a request to provide the data of the network-side model to a first core network entity.

[0038] In some implementations of the methods, systems, and apparatuses described herein, in the fourth aspect: the request may include a data request for the RAN entity.

[0039] In some implementations of the methods, systems, and apparatuses described herein, in the fourth aspect: the request may include an indication that the UE has no UE- side model that corresponds to or is compatible with the network-side model.

[0040] In some implementations of the methods, systems, and apparatuses described herein, in the fourth aspect: the request may include information indicating one or more required data types of the data.

[0041] In some implementations of the methods, systems, and apparatuses described herein, in the fourth aspect: the request may include an identifier of the network-side model.

[0042] Some implementations of the methods, systems, and apparatuses described herein may include, in the fourth aspect: receiving the created and / or trained learning model. For example, the created and / or trained learning model may be received from an entity associated with the UE (e.g., a UE-side server), which has created and / or trained the learning model based at least in part on the data of the network-side model.BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 illustrates an example of a wireless communications system that can support creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure.

[0044] Figure 2 illustrates an example wireless communications system with a UE-side model and a network (NW)-side model in accordance with aspects of the present disclosure.

[0045] Figure 3 illustrates an example wireless communications system with a two- sided AI / ML model in accordance with aspects of the present disclosure.

[0046] Figure 4 illustrates an example system for training a two-sided AI / ML model in accordance with aspects of the present disclosure.

[0047] Figure 5 illustrates example operations for creating and / or training a UE-side model of a two-sided AI / ML model based on the example of Figure 4 in accordance with aspects of the present disclosure.

[0048] Figure 6 illustrates an example of a wireless communications system that can support creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure.

[0049] Figure 7 illustrates example operations for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure.

[0050] Figure 8 illustrates example operations for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure.

[0051] Figure 9 illustrates example operations for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure.

[0052] Figure 10 illustrates example operations for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure.

[0053] Figure 11 illustrates an example of a UE in accordance with aspects of the present disclosure.

[0054] Figure 12 illustrates an example of a processor in accordance with aspects of the present disclosure.

[0055] Figure 13 illustrates an example of an NE in accordance with aspects of the present disclosure.

[0056] Figure 14 illustrates a flowchart of a method that can be performed by a UE in accordance with aspects of the present disclosure.

[0057] Figure 15 illustrates a flowchart of a method that can be performed by a NE in accordance with aspects of the present disclosure.

[0058] Figure 16 illustrates a flowchart of a method that can be performed by a first CN entity in accordance with aspects of the present disclosure.

[0059] Figure 17 illustrates a flowchart of a method that can be performed by a second CN entity in accordance with aspects of the present disclosure.DETAILED DESCRIPTION

[0060] To create or train a learning model for a UE, a UE may obtain data from a RAN node (e.g., next-generation NodeB (gNB)) via a radio interface (e.g., Uu interface), and then provide the obtained data to a UE-side server via internet protocol (IP) connectivity to create and / or train the learning model (the created and / or trained model may be used as a UE-side model). The data from the RAN node may include RAN-related data, such as data associated with a network (NW)-side model, which can be used for creating and / or training the learning model. This approach involves the transfer of the data from the RAN node to the UE via the radio, and may not be ideal in some cases. For example, in some cases, this approach may be resource inefficient in that it may consume control plane radio bandwidth for a user plane operation. For example, in some cases, for multiple UEs, this approach may have poor scalability as the same data (e.g., RAN related data) may be transferred from the RAN node to each of the UEs via the radio.

[0061] Embodiments disclosed herein provide various solutions to support the creation and / or training of a learning model for a UE. Embodiments disclosed herein may enable transfer of data for creating and / or training the learning model, in particular data of a NW-side model, via the CN (which may interact with the UE-side server and optionally the RAN node). This may improve resource efficiency and / or scalability. In some embodiments, the CN may identify the RAN node serving the UE (that needs / requests the learning model) and may trigger the RAN node to provide the data for creating and / or training the learning model (e.g., model parameters, training and / or inference data of the NW-side model) to the UE-side server via the CN. In some embodiments, the CN may determine that the data for creating and / or training the learning model is stored in the CN, and may retrieve the stored data and provide it to the UE-side server.

[0062] In some embodiments, a first CN entity may: (i) receive, from a RAN entity, data of a NW-side model associated with the RAN entity or a storage location of the data, the data may include one or more of: a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model; and (ii) output the received data or the received storage location, for supporting creation and / or training of a learning model for a UE. In some embodiments, a second CN entity may: (i) receive a first request associated with (creating and / or training) a learning model for a UE; (ii) obtain, based at least in part on the first request, data of a NW-side model, which may include one or more of: a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model; and (iii) output the obtained data, for supporting creation and / or training of the learning model. In some embodiments, a RAN entity may: (i) receive a request to provide data of a NW-side model associated with the RAN entity, the data may include one or more of: a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model; and (ii) output, to a first CN entity, the data or a storage location of the data, for supporting creation and / or training of a learning model for a UE. In some embodiments, a UE may: (i) receive, from a RAN entity, information associated with a NW-side model; (ii) determine that creation and / or training of a learning model for a UE is required; and (iii) output, to the RAN entity, a request to provide data of the NW-side model, for supporting one or more of creating or training the learning model. The data of the NW-side model may include one or more of a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model.

[0063] Some embodiments may provide a wireless communications system that include two or more of: the first CN entity, the second CN entity, the UE, and the RAN entity. Some embodiments may provide a CN for wireless communications, which include the first CN entity and the second CN entity.

[0064] Aspects of the present disclosure are described in the context of a wireless communications system. The wireless communications system is configured to support creation and / or training of a learning model for a UE.

[0065] Figure 1 illustrates an example of a wireless communications system 100 in accordance with aspects of the present disclosure. The wireless communications system 100 may support creation and / or training of a learning model for a UE. The wireless communications system 100 may include one or more NE 102, one or more UE 104, and a CN 106. The wireless communications system 100 may support various radio access technologies. In some implementations, the wireless communications system 100 may be a 4G network, such as a long-term evolution (LTE) network or an LTE- Advanced (LTE-A) network. In some other implementations, the wireless communications system 100 may be a new radio (NR) network, such as a 5G network, a 5G-Advanced (5G-A) network, or a 5G ultrawideband (5G-UWB) network. In other implementations, the wireless communications system 100 may be a combination of a 4G network and a 5G network, or other suitable radio access technology including Institute of Electrical and Electronics Engineers (IEEE) 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20. The wireless communications system 100 may support radio access technologies beyond 5G, for example, 6G.Additionally, the wireless communications system 100 may support technologies, such as time division multiple access (TDMA), frequency division multiple access (FDMA), or code division multiple access (CDMA), etc.

[0066] The one or more NE 102 may be dispersed throughout a geographic region to form the wireless communications system 100. One or more of the NE 102 described herein may be or include or may be referred to as a RAN entity, RAN node, a network node, a base station, a network element, a NodeB, an eNodeB (eNB), a next-generation NodeB (gNB), or other suitable terminology. An NE 102 and a UE 104 may communicate via a communication link, which may be a wireless or wired connection. For example, anNE 102 and a UE 104 may perform wireless communications (e.g., receive signaling, transmit signaling) over a Uu interface.

[0067] An NE 102 may provide a geographic coverage area for which the NE 102 may support services for one or more UEs 104 within the geographic coverage area. For example, an NE 102 and a UE 104 may support wireless communications of signals related to services (e.g., voice, video, packet data, messaging, broadcast, etc.) according to one or multiple radio access technologies. In some implementations, an NE 102 may be moveable, for example, a satellite associated with a non-terrestrial network (NTN). In some implementations, different geographic coverage areas associated with the same or different radio access technologies may overlap, but the different geographic coverage areas may be associated with different NE 102.

[0068] The one or more UE 104 may be dispersed throughout a geographic region of the wireless communications system 100. A UE 104 may include or may be referred to as a remote unit, a mobile device, a wireless device, a remote device, a subscriber device, a transmitter device, a receiver device, or some other suitable terminology. In some implementations, the UE 104 may be referred to as a unit, a station, a terminal, or a client, among other examples. Additionally, or alternatively, the UE 104 may be referred to as an Internet-of-Things (loT) device, an Internet-of-Everything (loE) device, or machine-type communication (MTC) device, among other examples.

[0069] A UE 104 may be able to support wireless communications directly with otherUEs 104 over a communication link. For example, a UE 104 may support wireless communications directly with another UE 104 over a device-to-device (D2D) communication link. In some implementations, such as vehicle-to-vehicle (V2V) deployments, vehicle-to-everything (V2X) deployments, or cellular-V2X deployments, the communication link may be referred to as a sidelink. For example, a UE 104 may support wireless communications directly with another UE 104 over a PC5 interface.

[0070] An NE 102 may support communications with the CN 106, or with another NE 102, or both. For example, an NE 102 may interface with other NE 102 or the CN 106 through one or more backhaul links (e.g., SI, N2, N2, or network interface). In some implementations, the NE 102 may communicate with each other directly. In some otherimplementations, the NE 102 may communicate with each other or indirectly (e.g., via the CN 106). In some implementations, one or more NE 102 may include subcomponents, such as an access network entity, which may be an example of an access node controller (ANC). An ANC may communicate with the one or more UEs 104 through one or more other access network transmission entities, which may be referred to as a radio heads, smart radio heads, or transmission-reception points (TRPs).

[0071] The CN 106 may include one or more core network entities arranged to implement one or more core network functions. The CN 106 may support user authentication, access authorization, tracking, connectivity, and other access, routing, or mobility functions. The CN 106 may be an evolved packet core (EPC), or a 5G core (5GC), or a beyond-5G core (e.g., 6GC), which may include a control plane entity that manages access and mobility (e.g., a mobility management entity (MME), an access and mobility management functions (AMF)) and a user plane entity that routes packets or interconnects to external networks (e.g., a serving gateway (S-GW), a Packet Data Network (PDN) gateway (P-GW), or a user plane function (UPF)). In some implementations, the control plane entity may manage non-access stratum (NAS) functions, such as mobility, authentication, and bearer management (e.g., data bearers, signal bearers, etc.) for the one or more UEs 104 served by the one or more NE 102 associated with the CN 106. The CN 106 may include an application function (AF) for creating and / or training learning models. The CN 106 may further include a data collection network function. The CN 106 may further include a data storage network function.

[0072] The CN 106 may communicate with a packet data network over one or more backhaul links (e.g., via an SI, N2, N2, or another network interface). The packet data network may include an application server. In some implementations, one or more UEs 104 may communicate with the application server. A UE 104 may establish a session (e.g., a protocol data unit (PDU) session, or the like) with the CN 106 via an NE 102. The CN 106 may route traffic (e.g., control information, data, and the like) between the UE 104 and the application server using the established session (e.g., the established PDU session). The PDU session may be an example of a logical connection between the UE 104 and the CN 106 (e.g., one or more network functions of the CN 106).

[0073] In the wireless communications system 100, the NEs 102 and the UEs 104 may use resources of the wireless communications system 100 (e.g., time resources (e.g., symbols, slots, subframes, frames, or the like) or frequency resources (e.g., subcarriers, carriers)) to perform various operations (e.g., wireless communications). In some implementations, the NEs 102 and the UEs 104 may support different resource structures. For example, the NEs 102 and the UEs 104 may support different frame structures. In some implementations, such as in 4G, the NEs 102 and the UEs 104 may support a single frame structure. In some other implementations, such as in 5 G and among other suitable radio access technologies, the NEs 102 and the UEs 104 may support various frame structures (i.e., multiple frame structures). The NEs 102 and the UEs 104 may support various frame structures based on one or more numerologies.

[0074] One or more numerologies may be supported in the wireless communications system 100, and a numerology may include a subcarrier spacing and a cyclic prefix. A first numerology (e.g., / r=0) may be associated with a first subcarrier spacing (e.g., 15 kHz) and a normal cyclic prefix. In some implementations, the first numerology (e.g., / r=0) associated with the first subcarrier spacing (e.g., 15 kHz) may utilize one slot per subframe. A second numerology (e.g., / r=l) may be associated with a second subcarrier spacing (e.g., 30 kHz) and a normal cyclic prefix. A third numerology (e.g., / r=2) may be associated with a third subcarrier spacing (e.g., 60 kHz) and a normal cyclic prefix or an extended cyclic prefix. A fourth numerology (e.g., / r=3) may be associated with a fourth subcarrier spacing (e.g., 120 kHz) and a normal cyclic prefix. A fifth numerology (e.g., / r=4) may be associated with a fifth subcarrier spacing (e.g., 240 kHz) and a normal cyclic prefix.

[0075] A time interval of a resource (e.g., a communication resource) may be organized according to frames (also referred to as radio frames). Each frame may have a duration, for example, a 10 millisecond (ms) duration. In some implementations, each frame may include multiple subframes. For example, each frame may include 10 subframes, and each subframe may have a duration, for example, a 1 ms duration. In some implementations, each frame may have the same duration. In some implementations, each subframe of a frame may have the same duration.

[0076] Additionally or alternatively, a time interval of a resource (e.g., a communication resource) may be organized according to slots. For example, a subframe may include a number (e.g., quantity) of slots. The number of slots in each subframe may also depend on the one or more numerologies supported in the wireless communications system 100. For instance, the first, second, third, fourth, and fifth numerologies (i.e., / r=0, jU=l , / r=2, jU=3, / r=4) associated with respective subcarrier spacings of 15 kHz, 30 kHz, 60 kHz, 120 kHz, and 240 kHz may utilize a single slot per subframe, two slots per subframe, four slots per subframe, eight slots per subframe, and 16 slots per subframe, respectively. Each slot may include a number (e.g., quantity) of symbols (e.g., OFDM symbols). In some implementations, the number (e.g., quantity) of slots for a subframe may depend on a numerology. For a normal cyclic prefix, a slot may include 14 symbols. For an extended cyclic prefix (e.g., applicable for 60 kHz subcarrier spacing), a slot may include 12 symbols. The relationship between the number of symbols per slot, the number of slots per subframe, and the number of slots per frame for a normal cyclic prefix and an extended cyclic prefix may depend on a numerology. It should be understood that reference to a first numerology (e.g., / r=0) associated with a first subcarrier spacing (e.g., 15 kHz) may be used interchangeably between subframes and slots.

[0077] In the wireless communications system 100, an electromagnetic (EM) spectrum may be split, based on frequency or wavelength, into various classes, frequency bands, frequency channels, etc. By way of example, the wireless communications system 100 may support one or multiple operating frequency bands, such as frequency range designations FR1 (410 MHz - 7.125 GHz), FR2 (24.25 GHz - 52.6 GHz), FR3 (7.125 GHz - 24.25 GHz), FR4 (52.6 GHz - 114.25 GHz), FR4a or FR4-1 (52.6 GHz - 71 GHz), and FR5 (114.25 GHz - 300 GHz). In some implementations, the NEs 102 and the UEs 104 may perform wireless communications over one or more of the operating frequency bands. In some implementations, FR1 may be used by the NEs 102 and the UEs 104, among other equipment or devices for cellular communications traffic (e.g., control information, data). In some implementations, FR2 may be used by the NEs 102 and the UEs 104, among other equipment or devices for short-range, high data rate capabilities.

[0078] FR1 may be associated with one or multiple numerologies (e.g., at least three numerologies). For example, FR1 may be associated with a first numerology (e.g., / r=0), which includes 15 kHz subcarrier spacing; a second numerology (e.g., / r=l), which includes 30 kHz subcarrier spacing; and a third numerology (e.g., / r=2), which includes 60 kHz subcarrier spacing. FR2 may be associated with one or multiple numerologies (e.g., at least 2 numerologies). For example, FR2 may be associated with a third numerology (e.g., / r=2), which includes 60 kHz subcarrier spacing; and a fourth numerology (e.g., / r=3), which includes 120 kHz subcarrier spacing.

[0079] Figure 2 illustrates an example wireless communications system 200 with a UE- side model and a NW-side model in accordance with aspects of the present disclosure. In some implementations, the wireless communications system 200 may be implemented as part of the wireless communications system 100. The system 200 may include a UE 202 and a RAN node 204 (e.g., gNB) arranged to communicate via a radio interface 206 (e.g., Uu interface). The UE 202 may include a UE-side model, which may be otherwise known as UE-side AI / ML model or other suitable terminology. The UE-side model is a learning model that can be trained and can perform inference. The RAN node 204 may include a NW-side model, may be otherwise known as NW-side AI / ML model or other suitable terminology. The NW-side model may be a learning model that can be trained and can perform inference.

[0080] In some implementations, the UE 202 and the RAN node 204 may collaborate to perform AI / ML operations using one or both of the UE-side model and the NW-side model. In some implementations, the UE-side model and the NW-side model may perform AI / ML operations independent of each other. In some implementations, the UE 202 and the RAN node 204 may cooperate (e.g., communicate) to supporting training of one or both of the UE-side model and the NW-side model. In some implementations, the UE 202 and the RAN node 204 may cooperate (e.g., communicate) to facilitate performing inference using one or both of the UE-side model and the NW-side model. In some implementations, the UE-side model and the NW-side model may belong to a two-sided model in particular a two-sided AI / ML model. For example, the two-sided model may be collaboratively trained. For example, the UE 202 and the RAN node 204 may perform joint inference using thetwo-sided model (part of the inference performed by the UE 202 and part of the inference performed by the RAN node 204).

[0081] Figure 3 illustrates an example wireless communications system with a two- sided AI / ML model in accordance with aspects of the present disclosure. In some implementations, the wireless communications system 300 may be implemented as part of the wireless communications system 100. The system 300 may include a UE 302 and a gNB 304 arranged to communicate via a Uu interface. The UE 302 may include a UE-side AI / ML model that can be trained to perform inference. The gNB 304 may include a NW- side AI / ML model that can be trained to perform inference. The UE-side AI / ML model and the NW-side AI / ML model may belong to or define a two-sided AI / ML model. In some implementations, the two-sided AI / ML model may be arranged to perform channel state information (CSI) compression. For example, the UE 302 may use the UE-side AI / ML model to determine optimal CSI encoding parameters and the gNB 304 may use the NW- side AI / ML model to determine optimal CSI decoding parameters. In this example, the UE- side AI / ML model may be referred to as a CSI generation model, and the NW-side AI / ML model may be referred to as a CSI reconstruction model.

[0082] 3GPP TR 38.843 V18.0.0 (2023-12) has disclosed some options to define pairing information used to enable a UE to select CSI generation model(s) compatible with CSI reconstruction model(s) used by a gNB (clause 7.1.2).

[0083] Figure 4 illustrates an example system 400 for training a two-sided AI / ML model in accordance with aspects of the present disclosure. The system 400 may include a UE 402 and a gNB 404 arranged to communicate via a Uu interface. The system 400 may further include an entity 406 (e.g., NW-side server) arranged to communicate with the gNB 404 via IP connectivity, and an entity 408 (e.g., a UE-side server) arranged to communicate with the UE 402 via IP connectivity. The UE-side AI / ML model and the NW-side AI / ML model may belong to or define a two-sided AI / ML model. For example, the UE-side AI / ML model may be a CSI generation model, the NW-side AI / ML model may be a CSI reconstruction model, and the two-sided AI / ML model may be arranged to perform CSI compression.

[0084] To create and / or train the NW-side AI / ML model, the gNB 404 may provide related data to the entity 406 (e.g., NW-side server). The entity 406 can then create and / or train the NW-side AI / ML model based at least in part on the data provided by the gNB 404.

[0085] To create and / or train the UE-side AI / ML model (e.g., as part of a two-sided AI / ML model), the gNB 404 may provide data including data and / or model parameters of the NW-side AI / ML model to the UE 402 via the Uu interface. The UE 402 may then provide data, which includes the data received from the gNB 404, to the entity 408 (e.g., a UE-side server) via IP connectivity (user plane connection). The entity 408 can then create and / or train the UE-side AI / ML model based at least in part on the data provided by the UE 402.

[0086] Figure 5 illustrates example operations 500 for creating and / or training a UE- side model of a two-sided AI / ML model based on the example of Figure 4 in accordance with aspects of the present disclosure. The example operations 500 may be used for creating and / or training a UE-side AI / ML model for scenarios where a UE and a gNB may jointly use a two-sided AI / ML model (including the UE-side AI / ML model and a NW-side AI / ML model) for inference. The example operations 500 may be performed by a UE, a gNB, and a UE-side server. In the example operations 500, the data for creating and / or training the UE-side AI / ML model may not be transmitted via the CN.

[0087] At 502, the gNB may already have a trained NW-side AI / ML model for a AI / ML related use case (e.g., CSI).

[0088] At 504, the UE may provide its capabilities (including AI / ML capabilities) to the gNB via radio resource control (RRC) signaling.

[0089] At 506, the gNB, based on the received UE capabilities, may decide to use a two-sided AI / ML model for a use case (e.g., CSI compression).

[0090] At 508, the gNB may provide, to the UE via RRC signaling, information associated with or indication of the NW-side AI / ML model of the two-sided AI / ML model to be used for inference. In response to the received information, the UE may determine whether it has a corresponding or compatible UE-side AI / ML model for the two-sided AI / ML model. If the UE determines that it does not have a corresponding or compatibleUE-side AI / ML model, the UE may check with the UE-side server for an available AI / ML model which can be used as the UE-side AI / ML model.

[0091] At 510, the UE may transmit, to the UE-side server via IP connection, a request to check whether the UE-side server has an AI / ML model which can be used as the UE-side AI / ML model (i.e., whether the UE-side server has the UE-side model).

[0092] At 512, the UE-side server may provide, based on the received request, an acknowledgement message via IP connection. The UE-side server may check, based on the received request, whether it has an AI / ML model which can be used as the UE-side AI / ML model.

[0093] At 514, the UE-side server may determine that no AI / ML model that can be used as the UE-side AI / ML model is available. The UE-side server may decide to create and / or train an AI / ML model that can be used as the UE-side AI / ML model. The UE-side server may request for data for creating and / or training the AI / ML model.

[0094] At 516, the UE-side server may transmit a data request to the UE via IP connection. The data request may include a request for the UE to provide the data. The data request may include a RAN model ID associated with the NW-side AI / ML model and optionally other contents.

[0095] At 518, the UE, based on the request received from the UE-side server, may transmit a request to the gNB via RRC signaling. The request may include the RAN model ID, a request for data for creating and / or training the UE-side AI / ML model, etc. The data for creating and / or training the UE-side AI / ML model may include training data and parameters of the NW-side AI / ML model.

[0096] At 520, the gNB, based on the request received from the UE, may transmit a response to the UE via RRC signaling. The response may include the data requested (e.g., training data and parameters of the NW-side AI / ML model).

[0097] At 522, the UE, based on the response received from the gNB, may transmit the data for creating and / or training the UE-side AI / ML model (e.g., training data and parameters of the NW-side AI / ML model) to the UE-side server via IP connection.

[0098] At 524, the UE-side server may, based on the response and the data received from the UE, create and / or train an AI / ML model that can be used as the UE-side AI / ML model.

[0099] Figure 6 illustrates an example of a wireless communications system 600 that can support creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure. In some implementations, the wireless communications system 600 may be implemented as part of the wireless communications system 100. The system 600 illustrates an example architecture that can support the transfer of RAN node data (in particular data for creating and / or training a learning model for a UE, such as data of a NW-side model) to a UE-side server.

[0100] The system 600 may include a UE 602 and a RAN node (e.g., gNB) 604. The UE may include a UE-side model and the RAN node 604 may include a NW-side model. The UE-side model and the NW-side model may belong to or define a two-sided AI / ML model for CSI operation (the UE-side model for encoding and the NW-side model for decoding). The system 600 may further include a UE-side server for creating and / or training AI / ML models (which may be used as UE-side AI / ML models). In some implementations, the UE-side server 606 may be an application function (AF) of the CN. In some other implementations, the UE-side server 606 may not belong to the CN. The system 600 may further include a data client 605 for the RAN node 604. The data client 605 may be used to store model parameters and training and / or inference data of one or more AI / ML models (e.g., NW-side models). The system 600 may further include a CN with a data collection network function (NF) 608, an AMF 610, and an operations, administration, and maintenance (0AM) function.

[0101] In some implementations, the data collection NF 608 may be responsible for retrieving or obtaining the data for creating and / or training the learning model for the UE (e.g., data of the NW-side model) from the RAN node 604 directly (via a user plane SBI connection).

[0102] In some implementations, the data collection NF 608 may be responsible for retrieving or obtaining the data for creating and / or training the learning model for the UE(e.g., data of the NW-side model) from the RAN node 604 via the control plane via the AMF 610.

[0103] In some implementations, the RAN node 604 (or data client 605) may provide the data for creating and / or training the learning model for the UE (e.g., data of the NW- side model) to the 0 AM 612 via a direct interface, and the data collection NF 608 may retrieve or obtain the data from the 0AM 612.

[0104] In some implementations, the RAN node 604 may include the data client 605 responsible for storing data for creating and / or training the learning model for the UE (e.g., data of the NW-side model). In some implementations, the data client 605 may be in communication with the data collection NF 608. In some implementations, the data client 605 may be in communication with the 0AM 612.

[0105] In some implementations, the learning model for the UE may be used as a UE- side model of a two-sided AI / ML model, and the data collection NF 608 may retrieve or obtain data for creating and / or training the learning model for the UE (e.g., the data of the NW-side model) based at least in part on a request from the UE-side server 606. The request from the UE-side server 606 to the data collection NF 608 may be routed via a network exposure function (NEF).

[0106] In some implementations, the UE-side server 606 may retrieve or obtain the data directly from the RAN node 604 (or the data client 605) via an NEF. In some of these implementations, the data collection NF 608 may be omitted from the system 600).

[0107] It should be noted that the system 600 described herein describes at least one possible implementation, and that the system 600 may be modified and that other implementations are possible. For example, two or more of the core network functions may be combined. For example, the system 600 described herein include multiple routes or arrangements for the UE-side server 606 to obtain data from the RAN node 604 (or the data client 605) via the CN. However, the system 600 may be modified such that it may include only one or only some of these such routes or arrangements.

[0108] Figures 7 to 10 provide some example implementations for providing data for creating and / or training the learning model for the UE (e.g., data of the NW-side model) to an entity arranged to create and / or train the learning model.

[0109] Figure 7 illustrates example operations 700 for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure. In some implementations, the example operations 700 may be implemented based on at least part of the system 600. The example operations 700 may be used for creating and / or training a UE-side AI / ML model for a two-sided AI / ML model. The example operations 700 may be performed by a UE, a RAN node, an AMF, a data collection NF, and a UE-side server. The example operations 700 may include the AMF finding the RAN node serving the UE and requesting related data for creating and / or training a learning model for the UE.

[0110] At 702, the RAN node may already have a trained NW-side AI / ML model for a AI / ML related use case (e.g., CSI). The RAN node may have a list of all available NW-side AI / ML models.

[0111] At 704, the UE may provide its capabilities (including AI / ML capabilities) to the RAN node. The RAN node, based at least in part on the received capabilities, may decide to use a two-sided AI / ML model for a use case (e.g., CSI compression). The RAN node may provide, to the UE, information associated with or indication of the NW-side AI / ML model of the two-sided AI / ML model to be used. In response to the received information, the UE may determine whether it has a corresponding or compatible UE-side AI / ML model for the two-sided AI / ML model. If the UE determines that it does not have a corresponding or compatible UE-side AI / ML model, the UE may check with the UE-side server for an available AI / ML model which can be used as the UE-side AI / ML model.

[0112] At 706, the UE may transmit a request to check whether the UE-side server has an AI / ML model which can be used as the UE-side AI / ML model (i.e., whether the UE-side server has the UE-side model). The request may include an identifier of the NW-side AI / ML model (e.g., the RAN model ID) and / or other relevant information associated with (e.g., used by) the RAN node. The request may be transmitted via an IP connection. In some implementations, the request may be transmitted to the data collection NF (706a), which may in turn relay the request to the UE-side server (706b). In some implementations,the request may be transmitted to the UE-side server directly (without involving the data collection NF).

[0113] At 708, the UE-side server may determine, based at least in part on the received request, whether it has an AI / ML model which can be used as the UE-side AI / ML model.

[0114] At 710, the UE-side server may determine that no AI / ML model which can be used as the UE-side AI / ML model is available. The UE-side server may decide to create and / or train an AI / ML model that can be used as the UE-side AI / ML model. The UE-side server may request for data for creating and / or training the AI / ML model. To this end, in some implementations, the UE-side server may request for RAN data such as data of the NW-side AI / ML model (optionally among other things).

[0115] At 712, the UE-side server may transmit a data request to the data collection NF. The request may be transmitted via an NEF. The request may include information on the data requested. The request may include an identifier / address of the UE (e.g., globally unique permanent subscriber identifier (GPSI) or subscription permanent identifier (SUPI)) that is served by the RAN node containing the related data. The request may include an indication of the type of data to collect. The request may include an identifier of the NW- side AI / ML model. The request may include an indication to provide, from the RAN node serving the UE, the data set used to train the NW-side AI / ML model.

[0116] At 714, the data collection NF may determine the AMF and / or the RAN node serving the UE. For example, the data collection NF may determine the AMF serving the UE by interfacing with the unified data management (UDM) function.

[0117] At 716, the data collection NF may transmit or forward the data request to the AMF. This may be performed in various ways. In a first example, the data collection NF may invoke an Namf_Communication_NlN2MessageTransfer including, in a container, a data request message for the RAN node, the address of the UE (e.g. SUPI), and an indication not to retrieve any data from the UE. In a second example, the data collection NF may request the AMF via an Namf EventTrigger SBI operation to provide the identity of the RAN or RAN node serving the UE. The data collection NF may then invoke an Namf_Communication_NonUEN2MessageTransfer SBI message to the AMF, and themessage may include a container with a data request message for the RAN node. In a third example, the data collection NF may invoke a Namf SBI service specifically used to retrieve data from the RAN node and the request may include an identifier of the UE (UE ID). Other examples are possible.

[0118] At 718, the AMF transmits the data request to the RAN node. In some implementations, the AMF may include the container for requesting data from the RAN node in an N2 message towards the RAN node, to request the RAN node provide data of the NW-side AI / ML model (parameters, training data, and / or inference data of the NW-side AI / ML model). The request may include an identifier of the NW-side AI / ML model. In respect of the first example in 716, in 718, the AMF may transmit the N2 message to the RAN node based at least in part on the existing N2 link for the NAS connection between the UE and the AMF. The existing N2 link may be configured to convey signaling to the UE. By using the existing N2 link, there is no need to establish a new N2 link for 718. The existing N2 link may be the same next generation application protocol (NGAP) user equipment - transport network layer association (UE-TNLA)-binding (i.e., may use the same transport network layer (TNL) association and same NGAP association for the UE) when the UE is in connection management (CM) connected state. Additional details pertaining to this feature can be found, for example, in 3 GPP Technical Specification (TS) 23.501 V19.2.1 (2025-01) clause 5.21.1. The N2 message may not include any N1 message for the UE. In respect of the second example in 716, in 718, the AMF may transmit the N2 message to the RAN node using a new N2 link independent of any NAS connection between the UE and the AMF. The new N2 link may be a TNL association used for non- UE associated signaling. Additional details pertaining to this feature can be found, for example, in 3GPP TS 23.501 V19.2.1 (2025-01) clause 5.21.1.1. In respect of the third example in 716, in 718, in some implementations, the AMF may determine the RAN node serving the UE by identifying where there is an existing NAS connection between the UE and the AMF and, if it exists, using the existing N2 link to retrieve data from the RAN node. The existing N2 link may be configured to convey signaling to the UE. By using the existing N2 link, there is no need to establish a new N2 link for 718. In respect of the third example in 716, in 718, in some implementations, the AMF may determine the RAN node serving the UE by paging the UE and, once a NAS connection is established between theUE and the AMF, the AMF may use the existing N2 link to retrieve data from the RAN node. The existing N2 link may be configured to convey signaling to the UE. By using the existing N2 link, there is no need to establish a new N2 link for 718. In some implementations, the data request transmitted from the AMF to the RAN node may include an N1 message. The N1 message may contain a NAS message towards the UE, which may include an indication to establish NAS signaling for radio measurement collection. In some implementations, the NAS message may be a UE configuration update message indicating to the UE to establish a NAS connection for radio measurements and / or AI / ML model transfer.

[0119] At 720, the RAN node may forward the N1 message to the UE via RRC signaling. This may be performed only for the cases where the data request from the AMF to the RAN node includes the N1 message.

[0120] At 722, the UE provides a N1 message response to the RAN node via RRC signaling. The response may include radio measurements obtained by the UE and optionally other suitable data.

[0121] At 724, the RAN node may, based at least in part on the data request from the AMF, provide the requested data of the NW-side AI / ML model (parameters, training data, and / or inference data of the NW-side AI / ML model). Alternatively, the RAN node may, based at least in part on the data request from the AMF, provide information associated with the storage location of the data of the NW-side AI / ML model (e.g., the storage location of the data of the of the NW-side AI / ML model in the CN such as any of the 0AM, the data storage network function, etc.).

[0122] At 726, the RAN node may transmit a N2 message including the data (or its storage location) and the N1 message response provided by the UE (if any) to the AMF.

[0123] The AMF, based at least in part on the received N2 message, extracts or otherwise obtains the data (or its storage location), and at 728, transmits it to the data collection NF.

[0124] At 730, the data collection NF may forward the data to the UE-side server. Optionally, the data collection NF may store the data (e.g., the parameters, training data,and / or inference data of the NW-side AI / ML model) in a repository of the CN (e.g., application data repository function (ADRF)).

[0125] At 732, the UE-side server may create and / or train a corresponding or compatible UE-side AI / ML model based at least in part on the data received from the data collection NF (e.g., the parameters, training data, and / or inference data of the NW-side AI / ML model).

[0126] It should be noted that the example operations 700 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0127] Figure 8 illustrates example operations 800 for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure. In some implementations, the example operations 800 may be implemented based on at least part of the system 600. The example operations 800 may be used for creating and / or training a UE-side AI / ML model for a two-sided AI / ML model. The example operations 800 may be performed by a UE, a RAN node, an AMF, a data collection NF, and a UE-side server. The example operations 800 may relate to collecting data for creating and / or training a learning model for the UE from all RAN nodes in an area of interest.

[0128] At 802, the UE-side server determines a need to create and / or train an AI / ML model for use as a UE-side AI / ML model of a two-sided model. The UE-side server may determine a need to obtain data (e.g., RAN data such as data of NW-side AI / ML models) for creating and / or training the UE-side AI / ML model.

[0129] At 804, the UE-side server transmits a data request to the data collection NF. The request may be transmitted via an NEF. The data request may include information on the data requested. The request may include an area of interest associated with the RAN or RAN nodes from where the data needs to be collected. The data may include data associated with NW-side AI / ML model(s), such as parameters, training / inference data of NW-side AI / ML model(s). The UE-side server may include the area of interest when it is aware that there is no corresponding UE-side AI / ML model for a NW-side ML model in anarea (e.g. a UE may, based at least in part on error events or other events, indicate to a UE- side server that no UE-side AI / ML model is available).

[0130] At 806, the data collection NF determines the AMF(s) serving the area of interest. The data collection NF may retrieve AMF address(es) from the NRF, and may then obtain a list of UEs served by each of the AMF(s).

[0131] At 808, 714 to 732 in the example operations 700 may be performed to create and / or train a corresponding or compatible UE-side AI / ML model.

[0132] It should be noted that the example operations 800 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0133] Figure 9 illustrates example operations 900 for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure. In some implementations, the example operations 900 may be implemented based on at least part of the system 600. The example operations 900 may be used for creating and / or training a UE-side AI / ML model for a two-sided AI / ML model. The example operations 900 may be performed by a UE, a RAN node, an AMF, and a data collection NF. The example operations 900 may relate to the RAN node adding related data over N2 signaling via an existing NAS signaling between the UE and the AMF.

[0134] At 902, the RAN node may already have a trained NW-side AI / ML model for a AI / ML related use case (e.g., CSI).

[0135] At 904, the UE may provide its capabilities (including AI / ML capabilities) to the RAN node via RRC signaling.

[0136] At 906, the RAN node, based at least in part on the received UE capabilities, may decide to use a two-sided AI / ML model for a use case (e.g., CSI compression).

[0137] At 908, the RAN node may provide, to the UE via RRC signaling, information associated with or indication of the NW-side AI / ML model of the two-sided AI / ML model to be used (e.g., for inference).

[0138] In response to the received information, the UE may determine whether it has a corresponding or compatible UE-side AI / ML model for the two-sided AI / ML model. If the UE determines that it does not have a corresponding or compatible UE-side AI / ML model, the UE may check with the UE-side server for an available AI / ML model which can be used as the UE-side AI / ML model. If the UE cannot find or obtain the UE-side AI / ML model, it may determine that the creation and / or training of the UE-side AI / ML model is needed, and it may trigger a NAS message (e.g. a registration update NAS request) and include an indication that the RAN node needs to provide the data for creating and / or training the UE-side AI / ML model. The data may include RAN data such as data of the NW-side AI / ML model, including parameters, training data, and / or inference data of the NW-side AI / ML model. The trigger may be based at least in part on an indication by a UE- side server.

[0139] At 910, the UE transmits, to the RAN node via RRC signaling, a request for the RAN node to provide the related data. The request may include information on the data requested. The message may include a RAN model ID associated with the NW-side AI / ML model. The message may include a N1 NAS message. In some implementations, the message may include a flag for the RAN or RAN node to provide the related data. In some implementations, the message may include an indication that the UE has no UE-side AI / ML model corresponding to the NW-side AI / ML model selected.

[0140] At 912, the RAN node may, based at least in part on the message received from the UE, provide the related data (e.g., parameters, training data, and / or inference data of the NW-side AI / ML model) or a storage location of the related data. The storage location of the related data may be in the CN such as any of the 0AM, the data storage network function, etc. The RAN node may add the related data or the storage location to a N2 message.

[0141] At 914, the RAN node may transmit a N2 message to the AMF. The N2 message may include the related data (e.g., parameters, training data, and / or inference data of the NW-side AI / ML model) or a storage location of the related data. The N2 message may also include the N1 NAS message of the UE.

[0142] At 916, the AMF may extract or otherwise obtain the related data or its storage location from the N2 message. The AMF may further store the data or the storage locationlocally. The stored data or storage location may be used to handle future requests from the UE-side server.

[0143] At 918, the AMF may provide the data or storage location to the data collection NF. The data collection NF may store the data or the storage location at a local database (e.g., data storage network function). The data collection NF may then provide the data to the UE-side server to create and / or train a corresponding or compatible UE-side AI / ML model.

[0144] It should be noted that the example operations 900 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0145] Figure 10 illustrates example operations 1000 for supporting creation and / or training of a learning model for a UE in accordance with aspects of the present disclosure. In some implementations, the example operations 1000 may be implemented based on at least part of the system 600. The example operations 1000 may be used for creating and / or training a UE-side AI / ML model for a two-sided AI / ML model. The example operations 1000 may be performed by a UE, a RAN node, a data collection NF, and a UE-side server. The example operations 1000 may relate to the UE-side server retrieving data for creating and / or training a learning model for the UE from the CN.

[0146] At 1002, the RAN node may already have a trained NW-side AI / ML model for a AI / ML related use case (e.g., CSI). The RAN node may have a list of all available NW- side AI / ML models.

[0147] At 1004, the UE may provide its capabilities (including AI / ML capabilities) to the RAN node. The RAN node, based at least in part on the received capabilities, may decide to use a two-sided AI / ML model for a use case (e.g., CSI compression). The RAN node may provide, to the UE, information associated with or indication of the NW-side AI / ML model of the two-sided AI / ML model to be used. For example, the RAN node may communicate to the UE the UE-side AI / ML model(s) paired with the NW-side AI / ML model at the RAN node, e.g., using a RAN model ID. In response to the received information, the UE may determine whether it has a corresponding or compatible UE-sideAI / ML model (corresponding to the NW-side AI / ML model) for the two-sided AI / ML model. If the UE determines that it does not have a corresponding or compatible UE-side AI / ML model, the UE may check with the UE-side server for an available AI / ML model which can be used as the UE-side AI / ML model.

[0148] At 1006, the UE may transmit a request to check whether the UE-side server has an AI / ML model which can be used as the UE-side AI / ML model (i.e., whether the UE-side server has the UE-side model). The request may include an identifier of the NW-side AI / ML model (RAN model ID) and / or other relevant information associated with (e.g., used by) the RAN node. The request may be transmitted via an IP connection. In some implementations, the request may be transmitted to the data collection NF (1006a), which may in turn relay the request to the UE-side server (1006b). In some implementations, the request may be transmitted to the UE-side server directly (without involving the data collection NF).

[0149] At 1008, the UE-side server may determine, based at least in part on the received request, whether it has an AI / ML model which can be used as the UE-side AI / ML model.

[0150] At 1010, the UE-side server may determine that no AI / ML model which can be used as the UE-side AI / ML model is available. The UE-side server may decide to create and / or train an AI / ML model that can be used as the UE-side AI / ML model. The UE-side server may request for data for creating and / or training the AI / ML model. The data for creating and / or training the AI / ML model may include RAN data such as the parameters, training data, and / or inference data of the NW-side AI / ML model.

[0151] At 1012, the UE-side server may transmit a data request to the data collection NF. The request may be transmitted via an NEF. The request may include information on the data requested. The request may include an indication of the type of data to collect. The request may include an identifier of the NW-side AI / ML model. The request may include the RAN model ID. The data collection NF may determine the data that needs to be transferred to the UE side server based at least in part on the RAN model ID.

[0152] At 1014, the data collection NF may determine, based at least in part on the data request, that the data requested is available in the CN (e.g., stored in the data collection NF, a data storage NF, or like core network function / repository). The data collection NF may find the related data based at least in part on the RAN model ID. In respect of 1014, the data corresponding to the RAN model ID (the data for creating and / or training the AI / ML model) may already be stored in the CN (e.g., data collection NF or storage NF). This may be the case after the NW-side has completed the training of the NW-side AI / ML model and has obtained / developed the data (e.g., model parameter, training data, inference data, etc.) of the NW-side AI / ML model, and the obtained / developed data may be stored at the designated location. In some implementations, the NW-side may request for an ID when storing the data and may use the ID as the RAN model ID. In some implementations, the NW-side may receive an ID from the node storing the data and may use the ID as the RAN model ID.

[0153] At 1016, the data collection NF may transmit the obtained data to the UE-side server.

[0154] At 1018, the UE-side server may create and / or train a corresponding or compatible UE-side AI / ML model based at least in part on the data received from the data collection NF (e.g., the parameters, training data, and / or inference data of the NW-side AI / ML model).

[0155] It should be noted that the example operations 1000 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0156] Figure 11 illustrates an example of a UE 1100 in accordance with aspects of the present disclosure. The UE 1100 may include a processor 1102, a memory 1104, a controller 1106, and a transceiver 1108. The processor 1102, the memory 1104, the controller 1106, or the transceiver 1108, or various combinations thereof or various components thereof may be examples of means for performing various aspects of the present disclosure as described herein. These components may be coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or moreinterfaces. The UE 1100 may be configured to support the creation and / or training of a learning model for a UE (e.g., the UE 1100).

[0157] The processor 1102, the memory 1104, the controller 1106, or the transceiver1108, or various combinations or components thereof may be implemented in hardware (e.g., circuitry). The hardware may include a processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or other programmable logic device, or any combination thereof configured as or otherwise supporting a means for performing the functions described in the present disclosure.

[0158] The processor 1102 may include an intelligent hardware device (e.g., a general- purpose processor, a DSP, a CPU, an ASIC, an FPGA, or any combination thereof). In some implementations, the processor 1102 may be configured to operate the memory 1104. In some other implementations, the memory 1104 may be integrated into the processor 1102. The processor 1102 may be configured to execute computer-readable instructions stored in the memory 1104 to cause the UE 1100 to perform various functions of the present disclosure.

[0159] The memory 1104 may include volatile or non-volatile memory. The memory 1104 may store computer-readable, computer-executable code including instructions when executed by the processor 1102 cause the UE 1100 to perform various functions described herein. The code may be stored in a non-transitory computer-readable medium such the memory 1104 or another type of memory. Computer-readable media includes both non- transitory computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A non-transitory storage medium may be any available medium that may be accessed by a general-purpose or special-purpose computer.

[0160] In some implementations, the processor 1102 and the memory 1104 coupled with the processor 1102 may be configured to cause the UE 1100 to perform one or more of the functions described herein (e.g., executing, by the processor 1102, instructions stored in the memory 1104). For example, the processor 1102 may support wireless communications at the UE 1100 in accordance with examples as disclosed herein.

[0161] In some implementations, the UE 1100 may be configured to receive, from a RAN entity, information associated with a NW-side model. The NW model is an AI / ML model.

[0162] In some implementations, the UE 1100 may be configured to determine that one or more of creating or training a learning model for the UE 1100 is required. The learning model is an AI / ML model and can be used as a UE-side model of the UE. For example, the learning model (UE-side model) and the NW-side model may belong to a two-sided model in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model, which can be used for conveying channel state information. In some implementations, the UE 1100 may be configured to determine that the UE does not include (e.g., store) any UE-side model corresponding to or compatible with the NW-side model (e.g., the UE does not have any UE-side model that, together with the NW-side model, can provide or form a required two- sided model). In some implementations, the UE 1100 may be configured to determine that the UE could not retrieve, e.g., from an entity associated with the UE such as a UE-side server, any model that can be used as the UE-side model corresponding to or compatible with the NW-side model.

[0163] In some implementations, the UE 1100 may be configured to output, to the RAN entity, a request to provide data of the NW-side model, for supporting the creation and / or training of the learning model for the UE. The data of the NW-side model may include one or more of a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model. The request may include a request to provide the data of the NW-side model to a first CN entity (e.g., a CN entity configured to implement AMF), which may in turn provide the data to an entity associated with the UE (e.g., a UE-side server) to create and / or train the learning model. The request may include one or more of: a data request for the RAN entity, an indication that the UE has no UE-side model that corresponds to or is compatible with the NW-side model, information indicating one or more required data types of the data, an identifier of the NW-side model, etc.

[0164] In some implementations, the UE 1100 may be configured to receive the created and / or trained learning model, e.g., from an entity associated with the UE (e.g., a UE-side server), which has created and / or trained the learning model based at least in part on the data of the NW-side model.

[0165] The controller 1106 may manage input and output signals for the UE 1100. The controller 1106 may also manage peripherals not integrated into the UE 1100. In some implementations, the controller 1106 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some implementations, the controller 1106 may be implemented as part of the processor 1102.

[0166] In some implementations, the UE 1100 may include at least one transceiver 1108. In some other implementations, the UE 1100 may have more than one transceiver 1108. The transceiver 1108 may represent a wireless transceiver. The transceiver 1108 may include one or more receiver chains 1110, one or more transmitter chains 1112, or a combination thereof.

[0167] A receiver chain 1110 may be configured to receive signals (e.g., control information, data, packets) over a wireless medium. For example, the receiver chain 1110 may include one or more antennas for receive the signal over the air or wireless medium. The receiver chain 1110 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. The receiver chain 1110 may include at least one demodulator configured to demodulate the receive signal and obtain the transmitted data by reversing the modulation technique applied during transmission of the signal. The receiver chain 1110 may include at least one decoder for decoding the processing the demodulated signal to receive the transmitted data.

[0168] A transmitter chain 1112 may be configured to generate and transmit signals (e.g., control information, data, packets). The transmitter chain 1112 may include at least one modulator for modulating data onto a carrier signal, preparing the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The transmitter chain 1112 may also include at least one poweramplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over the wireless medium. The transmitter chain 1112 may also include one or more antennas for transmitting the amplified signal into the air or wireless medium.

[0169] Figure 12 illustrates an example of a processor 1200 in accordance with aspects of the present disclosure. The processor 1200 may be an example of a processor configured to perform various operations in accordance with examples as described herein. The processor 1200 may include a controller 1202 configured to perform various operations in accordance with examples as described herein. The processor 1200 may optionally include at least one memory 1204, which may be, for example, an L1 / L2 / L3 cache. Additionally, or alternatively, the processor 1200 may optionally include one or more arithmetic-logic units (ALUs) 1206. One or more of these components may be in electronic communication or otherwise coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or more interfaces (e.g., buses). The processor 1200 may be configured to support the creation and / or training of a learning model for a UE.

[0170] The processor 1200 may be a processor chipset and include a protocol stack (e.g., a software stack) executed by the processor chipset to perform various operations (e.g., receiving, obtaining, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, reading) in accordance with examples as described herein. The processor chipset may include one or more cores, one or more caches (e.g., memory local to or included in the processor chipset (e.g., the processor 1200) or other memory (e.g., random access memory (RAM), read-only memory (ROM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), static RAM (SRAM), ferroelectric RAM (FeRAM), magnetic RAM (MRAM), resistive RAM (RRAM), flash memory, phase change memory (PCM), and others).

[0171] The controller 1202 may be configured to manage and coordinate various operations (e.g., signaling, receiving, obtaining, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, reading) of the processor 1200 to cause the processor 1200 to support various operations in accordance with examples as described herein. For example, the controller 1202 may operate as a control unit of the processor 1200, generating control signals that manage the operation of variouscomponents of the processor 1200. These control signals include enabling or disabling functional units, selecting data paths, initiating memory access, and coordinating timing of operations.

[0172] The controller 1202 may be configured to fetch (e.g., obtain, retrieve, receive) instructions from the memory 1204 and determine subsequent instruction(s) to be executed to cause the processor 1200 to support various operations in accordance with examples as described herein. The controller 1202 may be configured to track memory address of instructions associated with the memory 1204. The controller 1202 may be configured to decode instructions to determine the operation to be performed and the operands involved. For example, the controller 1202 may be configured to interpret the instruction and determine control signals to be output to other components of the processor 1200 to cause the processor 1200 to support various operations in accordance with examples as described herein. Additionally, or alternatively, the controller 1202 may be configured to manage flow of data within the processor 1200. The controller 1202 may be configured to control transfer of data between registers, arithmetic logic units (ALUs), and other functional units of the processor 1200.

[0173] The memory 1204 may include one or more caches (e.g., memory local to or included in the processor 1200 or other memory, such RAM, ROM, DRAM, SDRAM, SRAM, MRAM, flash memory, etc. In some implementations, the memory 1204 may reside within or on a processor chipset (e.g., local to the processor 1200). In some other implementations, the memory 1204 may reside external to the processor chipset (e.g., remote to the processor 1200).

[0174] The memory 1204 may store computer-readable, computer-executable code including instructions that, when executed by the processor 1200, cause the processor 1200 to perform various functions described herein. The code may be stored in a non-transitory computer-readable medium such as system memory or another type of memory. The controller 1202 and / or the processor 1200 may be configured to execute computer-readable instructions stored in the memory 1204 to cause the processor 1200 to perform various functions. For example, the processor 1200 and / or the controller 1202 may be coupled with or to the memory 1204, the processor 1200, the controller 1202, and the memory 1204 maybe configured to perform various functions described herein. In some examples, the processor 1200 may include multiple processors and the memory 1204 may include multiple memories. One or more of the multiple processors may be coupled with one or more of the multiple memories, which may, individually or collectively, be configured to perform various functions herein.

[0175] The one or more ALUs 1206 may be configured to support various operations in accordance with examples as described herein. In some implementations, the one or more ALUs 1206 may reside within or on a processor chipset (e.g., the processor 1200). In some other implementations, the one or more ALUs 1206 may reside external to the processor chipset (e.g., the processor 1200). One or more ALUs 1206 may perform one or more computations such as addition, subtraction, multiplication, and division on data. For example, one or more ALUs 1206 may receive input operands and an operation code, which determines an operation to be executed. One or more ALUs 1206 be configured with a variety of logical and arithmetic circuits, including adders, subtractors, shifters, and logic gates, to process and manipulate the data according to the operation. Additionally, or alternatively, the one or more ALUs 1206 may support logical operations such as AND, OR, exclusive-OR (XOR), not-OR (NOR), and not-AND (NAND), enabling the one or more ALUs 1206 to handle conditional operations, comparisons, and bitwise operations.

[0176] The processor 1200 may support wireless communications in accordance with examples as disclosed herein. The processor 1200 may be configured to or operable to support one or more means for supporting the creation and / or training of a learning model for a UE.

[0177] In some implementations, the processor 1200 may be configured to or operable to support a means for receiving, from a RAN entity, information associated with a NW- side model. The NW-side model is an AI / ML model.

[0178] In some implementations, the processor 1200 may be configured to or operable to support a means for determining that one or more of creating or training a learning model for a UE is required. The learning model is an AI / ML model and can be used as a UE-side model of the UE. For example, the learning model (UE-side model) and the NW-side model may belong to a two-sided model in particular a two-sided AI / ML model. Forexample, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model, which can be used for conveying channel state information. In some implementations, the processor 1200 may be configured to or operable to support a means for determining that the UE does not include (e.g., store) any UE-side model corresponding to or compatible with the NW-side model (e.g., the UE does not have any UE-side model that, together with the NW-side model, can provide or form a required two-sided model). In some implementations, the processor 1200 may be configured to or operable to support a means for determining that the UE could not retrieve, e.g., from an entity associated with the UE such as a UE-side server, any model that can be used as the UE-side model corresponding to or compatible with the NW-side model.

[0179] In some implementations, the processor 1200 may be configured to or operable to support a means for outputting, to the RAN entity, a request to provide data of the NW- side model, for supporting the creation and / or training of the learning model for the UE. The data of the NW-side model may include one or more of a set of parameters of the NW- side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model. The request may include a request to provide the data of the NW-side model to a first CN entity (e.g., a CN entity configured to implement AMF), which may in turn provide the data to an entity associated with the UE (e.g., a UE-side server) to create and / or train the learning model. The request may include one or more of: a data request for the RAN entity, an indication that the UE has no UE-side model that corresponds to or is compatible with the NW-side model, information indicating one or more required data types of the data, an identifier of the NW-side model, etc.

[0180] In some implementations, the processor 1200 may support a means for receiving the created and / or trained learning model, e.g., from an entity associated with the UE (e.g., a UE-side server), which has created and / or trained the learning model based at least in part on the data of the NW-side model.

[0181] In some implementations, the processor 1200 may be configured to or operable to support a means for receiving a request to provide data of a NW-side model associated with a RAN entity. The data of the NW-side model may include one or more of a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set ofinference data of the NW-side model. In some implementations, the request may be received from the UE. In some implementations, the request may be received from a CN entity (e.g., a CN entity configured to implement AMF). In some implementations, the request may include one or more of: an identifier of the NW-side model, or information indicating one or more required data types of the data. For example, the information may indicate that one or more of the following data types are required (i.e., need to be obtained from the RAN entity): a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model. Additionally or alternatively, the request may include one or more of: a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE.

[0182] In some implementations, the processor 1200 may be configured to or operable to support a means for outputting, to a CN entity (e.g., a CN entity configured to implement AMF), the data or a storage location of the data, for supporting creation and / or training of a learning model for a UE. The learning model is an AI / ML model and may be used as a UE- side model. The NW-side model is an AI / ML model. For example, the learning model (UE- side model) and the NW-side model may belong to a two-sided model, in particular a two- sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model, and they may support encoding (UE side) and reconstruction (NW-side) of CSI.

[0183] In some implementations, the processor 1200 may be configured to or operable to support a means for receiving, from the UE, radio measurement obtained by the UE. In some implementations, the processor 1200 may be configured to or operable to support a means for outputting, to the CN entity, the radio measurement along with the data or the storage location of the data.

[0184] Figure 13 illustrates an example of an NE 1300 in accordance with aspects of the present disclosure. The NE 1300 may include a processor 1302, a memory 1304, a controller 1306, and a transceiver 1308. The processor 1302, the memory 1304, the controller 1306, or the transceiver 1308, or various combinations thereof or various components thereof may be examples of means for performing various aspects of the present disclosure as described herein. These components may be coupled (e.g.,operatively, communicatively, functionally, electronically, electrically) via one or more interfaces. The NE 1300 may be configured to support the creation and / or training of a learning model for a UE.

[0185] The processor 1302, the memory 1304, the controller 1306, or the transceiver 1308, or various combinations or components thereof may be implemented in hardware (e.g., circuitry). The hardware may include a processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or other programmable logic device, or any combination thereof configured as or otherwise supporting a means for performing the functions described in the present disclosure.

[0186] The processor 1302 may include an intelligent hardware device (e.g., a general- purpose processor, a DSP, a CPU, an ASIC, an FPGA, or any combination thereof). In some implementations, the processor 1302 may be configured to operate the memory 1304. In some other implementations, the memory 1304 may be integrated into the processor 1302. The processor 1302 may be configured to execute computer-readable instructions stored in the memory 1304 to cause the NE 1300 to perform various functions of the present disclosure.

[0187] The memory 1304 may include volatile or non-volatile memory. The memory 1304 may store computer-readable, computer-executable code including instructions when executed by the processor 1302 cause the NE 1300 to perform various functions described herein. The code may be stored in a non-transitory computer-readable medium such the memory 1304 or another type of memory. Computer-readable media includes both non- transitory computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A non-transitory storage medium may be any available medium that may be accessed by a general-purpose or special-purpose computer.

[0188] In some implementations, the processor 1302 and the memory 1304 coupled with the processor 1302 may be configured to cause the NE 1300 to perform one or more of the functions described herein (e.g., executing, by the processor 1302, instructions stored in the memory 1304). For example, the processor 1302 may support wireless communicationsat the NE 1300 in accordance with examples as disclosed herein. The NE 1300 may be configured to support the creation and / or training of a learning model for a UE.

[0189] In some implementations, the NE 1300 may be configured to receive a request to provide data of a NW-side model associated with the RAN entity. The data of the NW- side model may include one or more of a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model. In some implementations, the request may be received from the UE. In some implementations, the request may be received from a CN entity (e.g., a CN entity configured to implement AMF). In some implementations, the request may include one or more of: an identifier of the NW-side model, or information indicating one or more required data types of the data. For example, the information may indicate that one or more of the following data types are required (i.e., need to be obtained from the RAN entity): a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW- side model. Additionally or alternatively, the request may include one or more of: a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE.

[0190] In some implementations, the NE 1300 may be configured to output, to a CN entity (e.g., a CN entity configured to implement AMF), the data or a storage location of the data, for supporting creation and / or training of a learning model for a UE. The learning model is an AI / ML model and may be used as a UE-side model. The NW-side model is an AI / ML model. For example, the learning model (UE-side model) and the NW-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model, and they may support encoding (UE side) and reconstruction (NW-side) of CSI.

[0191] In some implementations, the NE 1300 may be configured to receive, from the UE, radio measurement obtained by the UE. In some implementations, the NE 1300 may be configured to output, to the CN entity, the radio measurement along with the data or the storage location of the data.

[0192] The controller 1306 may manage input and output signals for the NE 1300. The controller 1306 may also manage peripherals not integrated into the NE 1300. In some implementations, the controller 1306 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some implementations, the controller 1306 may be implemented as part of the processor 1302.

[0193] In some implementations, the NE 1300 may include at least one transceiver 1308. In some other implementations, the NE 1300 may have more than one transceiver 1308. The transceiver 1308 may represent a wireless transceiver. The transceiver 1308 may include one or more receiver chains 1310, one or more transmitter chains 1312, or a combination thereof.

[0194] A receiver chain 1310 may be configured to receive signals (e.g., control information, data, packets) over a wireless medium. For example, the receiver chain 1310 may include one or more antennas for receive the signal over the air or wireless medium. The receiver chain 1310 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. The receiver chain 1310 may include at least one demodulator configured to demodulate the receive signal and obtain the transmitted data by reversing the modulation technique applied during transmission of the signal. The receiver chain 1310 may include at least one decoder for decoding the processing the demodulated signal to receive the transmitted data.

[0195] A transmitter chain 1312 may be configured to generate and transmit signals (e.g., control information, data, packets). The transmitter chain 1312 may include at least one modulator for modulating data onto a carrier signal, preparing the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The transmitter chain 1312 may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over the wireless medium. The transmitter chain 1312 may also include one or more antennas for transmitting the amplified signal into the air or wireless medium.

[0196] Figure 14 illustrates a flowchart of a method 1400 in accordance with aspects of the present disclosure. In some implementations, the operations of the method 1400 may be implemented by a UE as described herein. In some implementations, the UE may execute a set of instructions to control the function elements of the UE to perform the described functions. In some implementations, the operations of the method 1400 may be implemented by a processor as described herein.

[0197] At 1402, the method 1400 may include receiving, from a RAN entity, information associated with a NW-side model. The NW-side model is an AI / ML model. The operations of 1402 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1402 may be performed by a UE as described with reference to Figure 11 or a processor as described with reference to Figure 12.

[0198] At 1404, the method 1400 may include determining that creation and / or training of a learning model for the UE is required. The learning model is an AI / ML model and can be used as a UE-side model of the UE. For example, the learning model (UE-side model) and the NW-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model, and they may support encoding (UE side) and reconstruction (NW-side) of CSI. The determining in 1404 may be based on determining that the UE does not include (e.g., store) any UE-side model corresponding to or compatible with the NW-side model (e.g., the UE does not have any UE-side model that, together with the NW-side model, can provide or form a required two-sided model).Additionally or alternatively, the determining in 1404 may be based on determining that the UE could not retrieve, e.g., from an entity associated with the UE such as a UE-side server, any model that can be used as the UE-side model corresponding to or compatible with the NW-side model. The operations of 1404 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1404 may be performed by a UE as described with reference to Figure 11 or a processor as described with reference to Figure 12.

[0199] At 1406, the method 1400 may include outputting, to the RAN entity, a request to provide data of the NW-side model, for supporting the creation and / or training of the learning model for the UE. The data of the NW-side model may include one or more of a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model. The request may include a request to provide the data of the NW-side model to a first CN entity (e.g., a CN entity configured to implement AMF), which may in turn provide the data to an entity associated with the UE (e.g., a UE-side server) to create and / or train the learning model. The request may include one or more of: a data request for the RAN entity, an indication that the UE has no UE-side model that corresponds to or is compatible with the NW-side model, information indicating one or more required data types of the data, an identifier of the NW-side model, etc. The operations of 1406 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1406 may be performed a UE as described with reference to Figure 11 or a processor as described with reference to Figure 12.

[0200] It should be noted that the method 1400 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible. For example, in some implementations, the method 1400 may include, after 1406, receiving the created and / or trained learning model, e.g., from an entity associated with the UE (e.g., a UE-side server), which has created and / or trained the learning model based at least in part on the data of the NW-side model. The received learning model may then be used as the UE-side model for inference. In some implementations, aspects of these operations may be performed a UE as described with reference to Figure 11 or a processor as described with reference to Figure 12.

[0201] Figure 15 illustrates a flowchart of a method 1500 in accordance with aspects of the present disclosure. In some implementations, the operations of the method 1500 may be implemented by an NE (RAN entity) as described herein. In some implementations, the NE (RAN entity) may execute a set of instructions to control the function elements of the NE(RAN entity) to perform the described functions. In some implementations, the operations of the method 1500 may be implemented by a processor as described herein.

[0202] At 1502, the method 1500 may include receiving a request to provide data of a NW-side model associated with the RAN entity. The data of the NW-side model may include one or more of a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model. In some implementations, the request may be received from the UE. In some implementations, the request may be received from a CN entity (e.g., a CN entity configured to implement AMF). In some implementations, the request may include one or more of: an identifier of the NW-side model, or information indicating one or more required data types of the data. For example, the information may indicate that one or more of the following data types are required (i.e., need to be obtained from the RAN entity): a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW- side model. Additionally or alternatively, the request may include one or more of: a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE. The operations of 1502 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1502 may be performed by an NE as described with reference to Figure 13 or a processor as described with reference to Figure 12.

[0203] At 1504, the method 1500 may include outputting, based at least in part on the request and to a CN entity (e.g., a CN entity configured to implement AMF), the data or a storage location of the data, for supporting creation and / or training of a learning model for a UE. The learning model is an AI / ML model and may be used as a UE-side model. The NW-side model is an AI / ML model. For example, the learning model (UE-side model) and the NW-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model, and they may support encoding (UE side) and reconstruction (NW-side) of CSI. The operations of 1504 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of1504 may be performed by an NE as described with reference to Figure 13 or a processor as described with reference to Figure 12.

[0204] It should be noted that the method 1500 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible. For example, in some implementations, the method 1500 may include receiving, from the UE, radio measurement obtained by the UE, and outputting, to the CN entity, the radio measurement along with the data or the storage location of the data. In some implementations, aspects of these operations may be performed by an NE as described with reference to Figure 13 or a processor as described with reference to Figure 12.

[0205] Figure 16 illustrates a flowchart of a method 1600 in accordance with aspects of the present disclosure. In some implementations, the operations of the method 1600 may be implemented by a CN or a first CN entity as described herein. In some implementations, the first CN entity may be configured to implement AMF.

[0206] At 1602, the method 1600 may include receiving, from a RAN entity, data of a NW-side model associated with the RAN entity or a storage location of the data. The data of the NW-side model may include one or more of a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW- side model. The NW-side model is an AI / ML model.

[0207] At 1604, the method 1600 may include outputting the received data or the received storage location, for supporting creation and / or training of a learning model for a UE. The learning model is an artificial intelligent (AI) / machine learning (ML) model and can be used as a UE-side model of the UE. In some implementations, the received data or the received storage location may be output to a second CN entity (e.g., a CN entity configured to implement data collection network function), which may in turn obtain and provide the data to an entity associated with the UE (e.g., a UE-side server) to create and / or train the learning model. In some implementations, the learning model (UE-side model) and the NW-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model. The encoder model and the decoder model maysupport encoding (UE side) and reconstruction (NW-side) of channel state information (CSI).

[0208] In some implementations, prior to 1062, the method 1600 may include 1601 A and 1601B.

[0209] At 1601 A, the method 1600 may include receiving a first request to obtain the data of the NW-side model. The first request may be received from the second CN entity e.g., a CN entity configured to implement data collection network function). In some implementations, the information in the first request may include one or more of: an identifier of the NW-side model, the information in the first request may include information indicating one or more required data types of the data, an area of interest associated with at least one RAN entity, a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE. For example, the information indicating one or more required data types of the data may indicate that a set of parameters of the NW-side model is required. Additionally or alternatively, the information indicating one or more required data types of the data may indicate that a set of training data of the NW-side model is required. Additionally or alternatively, the information indicating one or more required data types of the data may indicate that a set of inference data of the NW- side model is required.

[0210] At 1601B, the method 1600 may include outputting, to the RAN entity associated with the UE, a second request to provide the data of the NW-side model. The RAN entity may be associated with the UE. The second request may include information contained in the first request. In some implementations, 160 IB may include outputting the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity. The existing association may be configured to convey signaling to the UE. For example, the existing association may include at least one communication link (e.g., N2 link).

[0211] In some implementations, the information in the first request may include the data request for the RAN entity, and the method 1600 may include obtaining an identity of the RAN entity, establishing an association between the first CN entity and the RAN entity, and, in 1601B, outputting the second request to the RAN entity based at least in part on theestablished association. For example, the association may include at least one communication link, which may be used for non-UE associated signaling. The RAN entity may be the RAN entity serving the UE. In some implementations, the information in the first request may include the identifier of the UE, and the method 1600 may include determining presence of NAS connection between the UE and the first CN entity, or paging the UE to establish a NAS connection between the UE and the first CN entity; and, in 1601B, outputting the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity. The existing association may be configured to convey signaling to the UE. For example, the existing association may include at least one communication link (e.g., N2 link). In some implementations, the information in the first request may include the data request for the RAN entity, the identifier of the UE, and the indication to not obtain any data from the UE, and the method 1600 may include, in 1601B, outputting the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity. The existing association may be configured to convey signaling to the UE. For example, the existing association may include at least one communication link (e.g., N2 link). In some implementations, the second request may further include information for the UE. The information for the UE may include an indication to establish a NAS connection for radio measurement collection.

[0212] It should be noted that the method 1600 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0213] Figure 17 illustrates a flowchart of a method 1700 in accordance with aspects of the present disclosure. In some implementations, the operations of the method 1700 may be implemented by a CN or a second CN entity as described herein. In some implementations, the second CN entity may be configured to implement a data collection network function.

[0214] At 1702, the method 1700 may include receiving a first request associated with a learning model for a UE. The learning model is an AI / ML model and can be used as a UE-side model of the UE. For example, the first request may be received from an entity associated with the UE (e.g., a UE-side server). In some implementations, the first requestmay include one or more of: an identifier of a NW-side model, information indicating one or more required data types of data required, an identifier of the UE, or an area of interest associated with at least one RAN entity for obtaining the data.

[0215] At 1704, the method 1700 may include obtaining, based at least in part on the first request, data of a NW-side model. The data of the NW-side model may include one or more of: a set of parameters of the NW-side model, a set of training data of the NW-side model, or a set of inference data of the NW-side model. In some implementations, the learning model (UE-side model) and the NW-side model may belong to a two-sided model, in particular a two-sided AI / ML model. For example, the learning model (UE-side model) may be an encoder model and the NW-side model may be a decoder model, and they may support encoding (UE side) and reconstruction (NW-side) of CSI. The first request may include one or more of: an identifier of the NW-side model, information indicating one or more required data types of the data of the NW-side model, an identifier of the UE, or an area of interest associated with at least one RAN entity for obtaining the data. For example, the information indicating one or more required data types of the data may indicate that a set of parameters of the NW-side model is required. Additionally or alternatively, the information indicating one or more required data types of the data may indicate that a set of training data of the NW-side model is required. Additionally or alternatively, the information indicating one or more required data types of the data may indicate that a set of inference data of the NW-side model is required.

[0216] In some implementations, the first request in 1702 may include the identifier of the UE, and the method 1700 includes, prior to or as part of 1704: determining, based at least in part on the first request, a first CN entity (e.g., a CN entity configured to implement AMF) or a RAN entity associated with (e.g., serving) the UE; and outputting, to the first CN entity, a second request to obtain the data from the RAN entity associated with the UE. The second request may include one or more of: a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE.

[0217] In some implementations, the first request in 1702 may include the area of interest, and the method 1700 includes, prior to or as part of 1704: determining, based at least in part on the first request, a first CN entity (e.g., a CN entity configured to implementAMF) associated with (e.g., serving) the area of interest; determining, based at least in part on the first CN entity, at least one UE (which includes the UE which the learning model is for) associated with (e.g., being served by) the first CN entity; and outputting, to the first CN entity, a second request to obtain the data from the RAN entity associated with the UE. The second request may include one or more of: a data request for the RAN entity, an identifier of the UE, or an indication to not obtain any data from the UE.

[0218] In some implementations, 1704 may include obtaining the data of the NW-side model from one or more of: the at least one memory of the second CN entity or at least one memory of another CN entity (e.g., a CN entity configured to implement a data storage network function). In some implementations, as the data of the NW-side model can be obtained locally from the CN, it may not be necessary to output any request to obtain the data of the NW-side model from the RAN entity associated with the UE.

[0219] At 1706, the method 1700 may include outputting the obtained data, for supporting creation and / or training of the learning model. In some implementations, the obtained data may be output to the entity associated with the UE (e.g., UE-side server), which may create and / or train the learning model based at least in part on the data of the NW-side model.

[0220] It should be noted that the method 1700 described herein describes at least one possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0221] Generally, embodiments disclosed herein relate to supporting creation and / or training of a learning model for a UE (UE-side AI / ML model). Some embodiments may relate to a CN determining a RAN node serving a UE for which a learning model (UE-side AI / ML model) needs to be created and / or trained. The CN may trigger the RAN node to provide NW-side AI / ML model data (model parameters, training data, inference data, etc., of the NW-side AI / ML model). The data may be stored at the CN and / or provided to the UE-side server. Some embodiments may relate to a CN (e.g., AMF) determining a RAN node serving a UE and requesting related RAN node data (NW-side AI / ML model data). Some embodiments may relate to collecting NW-side AI / ML model data (model parameters, training data, inference data, etc., of one or more NW-side AI / ML model(s))from all RAN nodes in an area of interest. Some embodiments may relate to a RAN node adding related RAN node data (NW-side AI / ML model data) over N2 signaling via an existing NAS signaling between the UE and the CN (e.g., the AMF).

[0222] The description herein is provided to enable a person having ordinary skill in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to a person having ordinary skill in the art, and the generic principles defined herein may be applied to other variations without departing from the scope of the disclosure. Thus, the disclosure is not limited to the examples and designs described herein but is to be accorded the broadest scope consistent with the principles and novel features disclosed herein.

[0223] For example, in some implementations, the UE-side model may belong to a two- sided model, in which case the UE-side model may be referred to as a UE-part of the two- sided model. For example, in some implementations, the UE-side model may not belong to a two-sided model. For example, in some implementations, the NW-side model may belong to a two-sided model, in which case the NW-side model may be referred to as a NW-part of the two-sided model. For example, in some implementations, the NW-side model may not belong to a two-sided model. For example, the UE-side model and NW-side model disclosed herein are not limited to models for CSI application.

Claims

CLAIMSWhat is claimed is:

1. A first core network (CN) entity for wireless communications, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the first CN entity to: receive, from a radio access network (RAN) entity, data of a network-side model associated with the RAN entity or a storage location of the data, wherein the data comprises one or more of a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model; and output the received data or the received storage location, for supporting one or more of creating or training a learning model for a user equipment (UE).

2. The first CN entity of claim 1, wherein the at least one processor is configured to cause the first CN entity to: prior to receiving the data of the network-side model from the RAN entity, receive a first request to obtain the data; and output, to the RAN entity, a second request to provide the data, wherein the RAN entity is associated with the UE, and the second request comprises information contained in the first request.

3. The first CN entity of claim 2, wherein the information in the first request comprises an identifier of the network-side model.

4. The first CN entity of claim 2 or 3, wherein the information in the first request comprises information indicating one or more required data types of the data.

5. The first CN entity of any one of claims 2 to 4, wherein the information in the first request comprises one or more of:a data request for the RAN entity; an identifier of the UE; an area of interest associated with at least one RAN entity; or an indication to not obtain any data from the UE.

6. The first CN entity of any one of claims 2 to 5, wherein the at least one processor is configured to cause the first CN entity to: output the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity; wherein the existing association is configured to convey signaling to the UE.

7. The first CN entity of claim 5, wherein the information in the first request comprises the data request for the RAN entity; and wherein the at least one processor is configured to cause the first CN entity to: obtain an identity of the RAN entity; establish an association between the first CN entity and the RAN entity; and output the second request to the RAN entity based at least in part on the established association.

8. The first CN entity of claim 5, wherein the information in the first request comprises the identifier of the UE; and wherein the at least one processor is configured to cause the first CN entity to: determine presence of non-access stratum (NAS) connection between the UE and the first CN entity, or page the UE to establish a NAS connection between the UE and the first CN entity; and output the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity;wherein the existing association is configured to convey signaling to the UE.

9. The first CN entity of claim 5, wherein the information in the first request comprises the data request for the RAN entity, the identifier of the UE, and the indication to not obtain any data from the UE; and wherein the at least one processor is configured to cause the first CN entity to: output the second request to the RAN entity based at least in part on an existing association between the first CN entity and the RAN entity; wherein the existing association is configured to convey signaling to the UE.

10. The first CN entity of any one of claims 2 to 9, wherein the second request further comprises information for the UE, the information for the UE comprises an indication to establish a NAS connection for radio measurement collection.

11. A radio access network (RAN) entity for wireless communications, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the RAN entity to: receive a request to provide data of a network-side model associated with the RAN entity, wherein the data comprises one or more of a set of parameters of the network-side model, a set of training data of the networkside model, or a set of inference data of the network-side model; and output, to a first core network (CN) entity, the data or a storage location of the data, for supporting one or more of creating or training a learning model for a user equipment (UE).

12. The RAN entity of claim 11, wherein the at least one processor is configured to cause the RAN entity to: receive the request from the UE; orreceive the request from the first CN entity.

13. The RAN entity of claim 11 or 12, wherein the request comprises one or more of: an identifier of the network-side model; information indicating one or more required data types of the data; a data request for the RAN entity; an identifier of the UE; or an indication to not obtain any data from the UE.

14. A second core network (CN) entity for wireless communications, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the second CN entity to: receive a first request associated with a learning model for a user equipment (UE); obtain, based at least in part on the first request, data of a networkside model, wherein the data comprises one or more of a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model; and output the obtained data, for supporting one or more of creating or training of the learning model.

15. The second CN entity of claim 14, wherein the first request comprises one or more of: an identifier of the network-side model; information indicating one or more required data types of the data; an identifier of the UE; or an area of interest associated with at least one radio access network (RAN) entity for obtaining the data.

16. The second CN entity of claim 15,wherein the first request comprises the identifier of the UE; and wherein the at least one processor is configured to cause the second CN entity to: determine, based at least in part on the first request, a first CN entity or a RAN entity associated with the UE; and output, to the first CN entity, a second request to obtain the data from the RAN entity associated with the UE.

17. The second CN entity of claim 15, wherein: the first request comprises the area of interest; and the at least one processor is configured to cause the second CN entity to: determine, based at least in part on the first request, a first CN entity associated with the area of interest; determine, based at least in part on the first CN entity, at least one UE associated with the first CN entity, wherein the at least one UE comprises the UE; and output, to the first CN entity, a second request to obtain the data from the RAN entity associated with the UE.

18. The second CN entity of claim 16 or 17, wherein the second request comprises one or more of: a data request for the RAN entity; an identifier of the UE; or an indication to not obtain any data from the UE.

19. The second CN entity of claim 14 or 15, wherein the at least one processor is configured to cause the second CN entity to: obtain the data from one or more of the at least one memory of the second CN entity or at least one memory of another CN entity.

20. A user equipment (UE) for wireless communications, comprising: at least one memory; andat least one processor coupled with the at least one memory and configured to cause the UE to: receive, from a radio access network (RAN) entity, information associated with a network-side model; determine that one or more of creating or training a learning model for the UE is required; and output, to the RAN entity, a request to provide data of a network-side model, for supporting one or more of creating or training the learning model, wherein the data comprises one or more of a set of parameters of the network-side model, a set of training data of the network-side model, or a set of inference data of the network-side model.

Images