EP4670332A1

EP4670332A1 - Control-level-initiated deployment of a machine learning model via user level for wireless networks

Info

Publication number: EP4670332A1
Application number: EP24706813.3A
Authority: EP
Inventors: Jerediah FEVOLD; Sakira HASSAN; Sari Kaarina Nielsen
Original assignee: Nokia Technologies Oy
Current assignee: Nokia Technologies Oy
Priority date: 2023-02-23
Filing date: 2024-02-16
Publication date: 2025-12-31
Also published as: WO2024176074A1; JP2026509756A; KR20250151526A; MX2025009686A; CN120752897A

Abstract

A method includes receiving, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and performing, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

Description

CONTROL PLANE INITIATED DELIVERY OF MACHINE LEARNING MODEL VIA USER PLANE FOR WIRELESS NETWORKS

TECHNICAL FIELD

[0001] This description relates to wireless communications.

BACKGROUND

[0002] A communication system may be a facility that enables communication between two or more nodes or devices, such as fixed or mobile communication devices. Signals can be carried on wired or wireless carriers.

[0003] An example of a cellular communication system is an architecture that is being standardized by the 3rd Generation Partnership Project (3GPP). A recent development in this field is often referred to as 4G or the long-term evolution (LTE) of 3G, or the Universal Mobile Telecommunications System (UMTS) radio-access technology. E-UTRA (evolved UMTS Terrestrial Radio Access) is the air interface of 3GPP's Long Term Evolution (LTE) upgrade path for mobile networks. In LTE, base stations or access points (APs), which are referred to as enhanced Node B (eNBs), provide wireless access within a coverage area or cell. In LTE, mobile devices, or mobile stations are referred to as user equipments (UE). LTE has included a number of improvements or developments. Aspects of LTE are also continuing to improve.

[0004] 5G New Radio (NR) development is part of a continued mobile broadband evolution process to meet the requirements of 5G, similar to earlier evolution of 3G and 4G wireless networks. In addition, 5G is also targeted at the new emerging use cases in addition to mobile broadband. A goal of 5G is to provide significant improvement in wireless performance, which may include new levels of data rate, latency, reliability, and security. 5G NR may also scale to efficiently connect the massive Internet of Things (loT) and may offer new types of mission-critical services. For example, ultra-reliable and low-latency communications (URLLC) devices may require high reliability and very low latency. 6G and other wireless networks are also being developed, or will be developed in the near future.

SUMMARY

[0005] A method may include receiving, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and performing, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0006] An apparatus may include: at least one processor; and at least one memory including computer program code; the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus at least to: receive, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and perform, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0007] A non-transitory computer-readable storage medium may include instructions stored thereon that, when executed by at least one processor, are configured to cause a computing system to: receive, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and perform, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0008] An apparatus may include: means for receiving, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and means for performing, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0009] The details of one or more examples of embodiments are set forth in the accompanying drawings and the description below. Other features will be apparent from the description and drawings, and from the claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] FIG. 1 is a block diagram of a wireless network according to an example embodiment.

[0011] FIG. 2 is a flow chart illustrating operation of a user device (or UE) according to an example embodiment.

[0012] FIG. 3 is a diagram illustrating a machine learning (ML) model download by a user equipment (UE) or user device according to an example embodiment.

[0013] FIG. 4 is a diagram illustrating a machine learning (ML) model upload by a user equipment (UE) or user device according to an example embodiment.

[0014] FIG. 5 is a diagram illustrating a ML model download procedure controlled by a LMF according to an example embodiment.

[0015] FIG. 6 is a diagram illustrating a ML model download procedure controlled by a LMF according to an example embodiment.

[0016] FIG. 7 is a block diagram of a wireless station or node (e.g., user node, user device, or UE, network node, relay node, gNB, or other node).

DETAILED DESCRIPTION

[0017] FIG. 1 is a block diagram of a wireless network 130 according to an example embodiment. In the wireless network 130 of FIG. 1, user devices 131, 132, 133 and 135, which may also be referred to as mobile stations (MSs) or user equipment (UEs), may be connected (and in communication) with a base station (BS) 134, which may also be referred to as an access point (AP), an enhanced Node B (eNB), a gNB or a network node. The terms user device and user equipment (UE) may be used interchangeably. A BS may also include or may be referred to as a RAN (radio access network) node, and may include a portion of a BS or a portion of a RAN node, such as (e.g., such as a centralized unit (CU) and/or a distributed unit (DU) in the case of a split BS or split gNB). At least part of the functionalities of a BS (e.g., access point (AP), base station (BS) or (e)Node B (eNB), gNB, RAN node) may also be carried out by any node, server or host which may be operably coupled to a transceiver, such as a remote radio head. BS (or AP) 134 provides wireless coverage within a cell 136, including to user devices (or UEs) 131, 132, 133 and 135. Although only four user devices (or UEs) are shown as being connected or attached to BS 134, any number of user devices may be provided. BS 134 is also connected to a core network 150 via a SI interface 151. A location management function (LMF) is also connected to BS/gNB 134 and core network 150. This is merely one simple example of a wireless network, and others may be used. [0018] A base station (e.g., such as BS 134) is an example of a radio access network (RAN) node within a wireless network. A BS (or a RAN node) may be or may include (or may alternatively be referred to as), e.g., an access point (AP), a gNB, an eNB, or portion thereof (such as a /centralized unit (CU) and/or a distributed unit (DU) in the case of a split BS or split gNB), or other network node. A network node may refer to or may include a BS, AP, gNB, CU and/or DU, RAN node, as examples. Also, at least in some cases, a network node may also refer to or may include a core network (e.g., a core network node or core network entity, such as AMF (access and mobility function) or other core network entity), a LMF (location management function), or other network node.

[0019] According to an illustrative example, a BS node (e.g., BS, eNB, gNB, CU/DU, . . .) or a radio access network (RAN) may be part of a mobile telecommunication system. A RAN (radio access network) may include one or more BSs or RAN nodes that implement a radio access technology, e.g., to allow one or more UEs to have access to a network or core network. Thus, for example, the RAN (RAN nodes, such as BSs or gNBs) may reside between one or more user devices or UEs and a core network. According to an example embodiment, each RAN node (e.g., BS, eNB, gNB, CU/DU, . . .) or BS may provide one or more wireless communication services for one or more UEs or user devices, e.g., to allow the UEs to have wireless access to a network, via the RAN node. Each RAN node or BS may perform or provide wireless communication services, e.g., such as allowing UEs or user devices to establish a wireless connection to the RAN node, and sending data to and/or receiving data from one or more of the UEs. For example, after establishing a connection to a UE, a RAN node or network node (e.g., BS, eNB, gNB, CU/DU, . . .) may forward data to the UE that is received from a network or the core network, and/or forward data received from the UE to the network or core network. RAN nodes or network nodes (e.g., BS, eNB, gNB, CU/DU, . . .) may perform a wide variety of other wireless functions or services, e.g., such as broadcasting control information (e.g., such as system information or on-demand system information) to UEs, paging UEs when there is data to be delivered to the UE, assisting in handover of a UE between cells, scheduling of resources for uplink data transmission from the UE(s) and downlink data transmission to UE(s), sending control information to configure one or more UEs, and the like. These are a few examples of one or more functions that a RAN node or BS may perform.

[0020] A user device or user node (user terminal, user equipment (UE), mobile terminal, handheld wireless device, etc.) may refer to a portable computing device that includes wireless mobile communication devices operating either with or without a subscriber identification module (SIM), including, but not limited to, the following types of devices: a mobile station (MS), a mobile phone, a cell phone, a smartphone, a personal digital assistant (PDA), a handset, a device using a wireless modem (alarm or measurement device, etc.), a laptop and/or touch screen computer, a tablet, a phablet, a game console, a notebook, a vehicle, a sensor, and a multimedia device, as examples, or any other wireless device. It should be appreciated that a user device may also be (or may include) a nearly exclusive uplink only device, of which an example is a camera or video camera loading images or video clips to a network. Also, a user node may include a user equipment (UE), a user device, a user terminal, a mobile terminal, a mobile station, a mobile node, a subscriber device, a subscriber node, a subscriber terminal, or other user node. For example, a user node may be used for wireless communications with one or more network nodes (e.g., gNB, eNB, BS, AP, CU, DU, CU/DU) and/or with one or more other user nodes, regardless of the technology or radio access technology (RAT). In LTE (as an illustrative example), core network 150 may be referred to as Evolved Packet Core (EPC), which may include a mobility management entity (MME) which may handle or assist with mobility/handover of user devices between BSs, one or more gateways that may forward data and control signals between the BSs and packet data networks or the Internet, and other control functions or blocks. Other types of wireless networks, such as 5G (which may be referred to as New Radio (NR)) may also include a core network.

[0021] In addition, the techniques described herein may be applied to various types of user devices or data service types, or may apply to user devices that may have multiple applications running thereon that may be of different data service types. New Radio (5G) development may support a number of different applications or a number of different data service types, such as for example: machine type communications (MTC), enhanced machine type communication (eMTC), Internet of Things (loT), and/or narrowband loT user devices, enhanced mobile broadband (eMBB), and ultra-reliable and low-latency communications (URLLC). Many of these new 5G (NR) - related applications may require generally higher performance than previous wireless networks.

[0022] loT may refer to an ever-growing group of objects that may have Internet or network connectivity, so that these objects may send information to and receive information from other network devices. For example, many sensor type applications or devices may monitor a physical condition or a status, and may send a report to a server or other network device, e.g., when an event occurs. Machine Type Communications (MTC, or Machine to Machine communications) may, for example, be characterized by fully automatic data generation, exchange, processing and actuation among intelligent machines, with or without intervention of humans. Enhanced mobile broadband (eMBB) may support much higher data rates than currently available in LTE.

[0023] Ultra-reliable and low-latency communications (URLLC) is a new data service type, or new usage scenario, which may be supported for New Radio (5G) systems. This enables emerging new applications and services, such as industrial automations, autonomous driving, vehicular safety, e-health services, and so on. 3GPP targets in providing connectivity with reliability corresponding to block error rate (BLER) of 10"⁵ and up to 1 ms U-Plane (user/data plane) latency, by way of illustrative example. Thus, for example, URLLC user devices/UEs may require a significantly lower block error rate than other types of user devices/UEs as well as low latency (with or without requirement for simultaneous high reliability). Thus, for example, a URLLC UE (or URLLC application on a UE) may require much shorter latency, as compared to an eMBB UE (or an eMBB application running on a UE).

[0024] The techniques described herein may be applied to a wide variety of wireless technologies or wireless networks, such as 5G (New Radio (NR)), cmWave, and/or mmWave band networks, loT, MTC, eMTC, eMBB, URLLC, 6G, etc., or any other wireless network or wireless technology. These example networks, technologies or data service types are provided only as illustrative examples.

[0025] According to an example embodiment, a machine learning (ML) model may be used within a wireless network to perform (or assist with performing) one or more tasks or functionalities. In general, one or more nodes (e.g., BS, gNB, eNB, RAN node, user node, UE, user device, relay node, or other wireless node) within a wireless network may use or employ a ML model, e.g., such as, for example a neural network model (e.g., which may be referred to as a neural network, an artificial intelligence (Al) neural network, an Al neural network model, an Al model, an Al machine learning (Al ML) model or algorithm, a model, or other term) to perform, or assist in performing, one or more ML-enabled tasks or functionalities. A ML-enabled task may include tasks that may be performed (or assisted in performing) by a ML model, or a task for which a ML model has been trained to perform or assist in performing).

[0026] ML-based algorithms or ML models may be used to perform and/or assist with performing a variety of wireless-related functionalities, such as radio resource management (RRM) functions or functionalities or tasks to improve network performance, such as, e.g., in the UE for beam prediction (e.g., predicting a best beam or best beam pair based on measured reference signals), antenna panel or beam control, RRM (radio resource measurement) measurements and feedback (channel state information (CSI) feedback), CSI report compression, link monitoring, Transmit Power Control (TPC), etc. In some cases, the use of ML models may be used to improve performance of a wireless network in one or more aspects or as measured by one or more performance indicators or performance criteria.

[0027] ML Models may be or may include, for example, computational models used in machine learning made up of nodes organized in layers. The nodes are also referred to as artificial neurons, or simply neurons, and perform a function on provided input to produce some output value. A neural network or ML model may typically require a training period to learn the parameters, i.e., weights, used to map the input to a desired output. The mapping occurs via the function. Thus, the weights are weights for the mapping function of the neural network. Each neural network model or ML model may be trained for a particular task.

[0028] To provide the output given the input, the neural network model or ML model should be trained, which may involve learning the proper value for a large number of parameters (e.g., weights) for the mapping function. The parameters are also commonly referred to as weights as they are used to weight terms in the mapping function. This training may be an iterative process, with the values of the weights being tweaked over many (e.g., thousands) of rounds of training until arriving at the optimal, or most accurate, values (or weights). In the context of neural networks (neural network models) or ML models, the parameters may be initialized, often with random values, and a training optimizer iteratively updates the parameters (weights) of the neural network to minimize error in the mapping function. In other words, during each round, or step, of iterative training the network updates the values of the parameters so that the values of the parameters eventually converge on the optimal values.

[0029] Neural network models or ML models may be trained in either a supervised or unsupervised manner, as examples. In supervised learning, training examples are provided to the neural network model or other machine learning algorithm. A training example includes the inputs and a desired or previously observed output. Training examples are also referred to as labeled data because the input is labeled with the desired or observed output. In the case of a neural network, the network learns the values for the weights used in the mapping function that most often result in the desired output when given the training inputs. In unsupervised training, the neural network model learns to identify a structure or pattern in the provided input. In other words, the model identifies implicit relationships in the data. Unsupervised learning is used in many machine learning problems and typically requires a large set of unlabeled data. [0030] According to an example embodiment, the learning or training of a neural network model or ML model may be classified into (or may include) multiple categories (including supervised and unsupervised), depending on whether there is a learning “signal” or “feedback” available to a model. Thus, for example, within the field of machine learning, there may be two main types of learning or training of a model: supervised, and unsupervised. The main difference between the two types is that supervised learning is done using known or prior knowledge of what the output values for certain samples of data should be. Therefore, a goal of supervised learning may be to learn a function that, given a sample of data and desired outputs, best approximates the relationship between input and output observable in the data. Unsupervised learning, on the other hand, does not have labeled outputs, so its goal is to infer the natural structure present within a set of data points. ML model training may also include reinforcement learning as well.

[0031] Challenges exist as to techniques that should be used to communicate or transfer ML models to a UE, e.g., such as what messages or information should be provided to initiate a ML model transfer, and whether to use the control plane and/or user (or data) plane to initiate the ML model transfer and/or transferring of the ML model. In general, the control plane may include control messages communicated to provide control for various aspects or functions of a wireless network, e.g., such as control messages that may be communicated to coordinate or control of: connection establishment, UE handover or cell change, power control, configuring a UE to perform some function, etc. Data or user plan may generally include transmission of user data to or from a UE. Control plane may typically include control messages, such as radio resource control (RRC) messages transmitted to or received from a gNB, CU/DU or other RAN node. Other types of control messages, for example, may include LPP (LTE positioning protocol) control messages transmitted to or received from a location management function (LMF), or network access stratum (NAS) control message transmitted to or received from a core network (e.g., such as the access and mobility function (AMF)), for example. User plane may include the transmission of data to and/or from a UE, such using one or more user plane data transmission protocols (e.g., such as, for example, File Transfer Protocol (FTP), Hypertext Transfer Protocol (HTTP), Transmission Control Protocol (TCP), User Datagram Protocol (UDP), or other user plane protocols). Also, at least in some cases, packets or messages transmitted via the control plane may typically have a different priority and/or QoS (quality of service), as compared to transmission of data via the user plane. It is unclear how to use the user plane and/or control plane to allow UEs to download ML models (or otherwise to facilitate ML model transfer with respect to UEs). [0032] The user (or data plane) may typically have greater capacity to communicate larger files or larger chunks of data, as compared to control messages (or control plane messages). According to an example embodiment, the control plane (including one or more control messages, e.g., such as RRC messages, NAS messages and/or LPP messages or other control plane messages) may have very limited capacity to convey large files or significant information (e.g., such as a ML model or ML model portion), but control messages or control plane may advantageously be used, for example, to provide control and/or communication to coordinate and/or initiate and/or confirm transfer of a ML model. Likewise, the user (or data) plane can accommodate larger data transfers and thus, after a ML model transfer has been coordinated, controlled or initiated via the control plane (e.g., via one or more control messages). Thus, a user device or UE may use the user (or data) plane to perform ML model transfer (e.g., either download a ML model or upload a ML model).

[0033] Therefore, according to an example embodiment, a control message(s) (e.g., provided via a control plane) may be used to initiate, coordinate, confirm and/or otherwise control a transfer of a ML model to or from a UE, while a user (or data) plane may be used by the UE to perform the ML model transfer (either upload or download). According to an example embodiment, the ML models transferred may include, e.g., trained or untrained models, and may include a complete ML model, a portion of a (or partial) ML Model, and/or a delta or difference that indicates a change (or delta) of a (or another) ML model.

[0034] Capability Exchange

[0035] Capability exchange may be a bi-directional message flow initiated by a network node or network entity such as the gNB, the access and mobility function (AMF) in the core network, or the location management function (LMF). The purpose may be to synchronize with the UE to understand what features, specified in the 3GPP specifications, are supported by the UE. According to an example embodiment, the capability exchange procedure may be enhanced to allow a UE (and/or network node) to indicate a capability to transfer (e.g., upload and/or download) a ML model, and may also indicate one or more user (or data) plane protocols supported by the UE for downloading or uploading ML model.

[0036] Command and Response Messaging

[0037] According to an example embodiment, a command a response messaging structure may be used by the UE and a network node to initiate, control confirm and/or indicate failure of a ML model transfer via a control plane messages (e.g., RRC message or other control messages). For example, a basic protocol structure may use or implement messages in the form of Command, CommandResponse, CommandFailure, and/or CommandComplete. The Command is may typically be sent by the network (or network node) to the UE. CommandResponse may be sent by the UE, e.g., to request additional data or additional information. CommandComplete may be sent by the network to the UE for upload, or from UE to network for download, when the procedure (e.g., ML model transfer) completes successfully, and CommandFailure may be sent by the network node to the UE for Model transfer, or from UE to network for ML Model download, when the procedure (ML model transfer) fails, for example.

[0038] Transfer Aspects

[0039] According to an example embodiment, one or more user (or data) plane protocols may be used for ML model transfers, as a non-exhaustive list of example protocols, e.g., such as HTTP, FTP, TCP, and UDP raw data transfer. Each of these are different well-known fundamental protocols used for data transfer, e.g., for download and upload. They each have trade-offs to consider, such as reliability of transfer and overhead, but those are left to implementation detail. The key is that there will be a data transfer protocol, and the UE and relevant network entity will need to communicate the capability to use a data transfer protocol. According to an example embodiment, a control message provided by the network node to the UE to initiate or request ML model transfer may indicate a (e.g., a user or data plane) protocol (e.g., indicating one of these protocols) that should be used by the UE to perform ML model transfer.

[0040] Control Plane and User Plane

[0041] According to an example embodiment, the control plane may, for example, include or may provide a signalling radio bearer (SRB), and may be used to send and receive, typically structured, messages to initiate, request, configure, coordinate, confirm and/or control the UE or user device for network connection procedures, to make and report measurements (e.g., to send a CSI report), and to establish data connectivity, and to perform other functions. Generally, the control plane is highly reliable, and used for small amounts of control data. The user plane, through a data radio bearer (DRB), supports a large amount of data, and is generally used for application layer traffic like the data transfer protocols discussed above. According to an example embodiment, as described in greater detail below, various embodiments described herein may use both the control plane (e.g., one or more control messages to initiate/request a ML model transfer, and/or to indicate a complete/successful ML Model transfer, and/or to indicate a failure of the ML model transfer, for example) and a user (or data) plane to perform the ML model transfer (e.g., either upload or download of the ML model). Also, user (or data) plane may use the UPF (user plane function). [0042] FIG. 2 is a flow chart illustrating operation of a user device (or UE) according to an example embodiment. Operation 210 includes receiving, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane. And, operation 220 includes performing, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information. The ML model transfer may be performed by the UE via user (or data) plane using the protocol indicated by the protocol information included in the control message.

[0043] With respect to the method of FIG. 2, the storage information, indicating a storage location for the ML model, may include: a network address of a host node; and at least one of the following: a path on the host node for the ML model; or a file name.

[0044] With respect to the method of FIG. 2, the control message may further include a ML model metadata container including at least one of the following ML model metadata: a ML model identifier that uniquely identifies the ML model in the user device or in a network; a functionality identifier that identifies a functionality for which the ML model is to be used to perform; and/or an area indication that identifies an area or one or more cells for which the ML model is valid for or to be used for.

[0045] With respect to the method of FIG. 2, the ML model may include at least one of: a complete ML model; a portion of a ML model; or a delta or change of a ML model indicating a change or difference of a ML model, e.g., as compared to another ML model that the UE may already have.

[0046] With respect to the method of FIG. 2, the network node may include at least one of the following: a gNB ; an eNB ; a base station or access point; a centralized unit (CU) and/or distributed unit (DU); a radio access network (RAN) node; an access and mobility function (AMF); a core network or core network node; or a location management function (LMF).

[0047] With respect to the method of FIG. 2, the control message may include at least one of the following: a first radio resource control (RRC) control message received from a gNB or a RAN (radio access network) node; a first LPP (LTE positioning protocol) control message received from a location management function (LMF); or a first NAS (network access stratum) control message received from the core network. [0048] With respect to the method of FIG. 2, the method may further include the UE transmitting at least one of the following messages regarding the transfer of the ML model: a second radio resource control (RRC) control message transmitted to the gNB or the RAN (radio access network) node, in response to the first RRC control message; a second LPP (LTE positioning protocol) control message transmitted to the location management function (LMF) in response to the first LPP control message; or a second NAS (network access stratum) control message transmitted to the core network in response to the first NAS control message.

[0049] With respect to the method of FIG. 2, the protocol to be used for transfer of the ML model via the user plane between the user device and the storage location indicated by the storage information may include (by way of example) at least one of: File Transfer Protocol (FTP); Hypertext Transfer Protocol (HTTP); Transmission Control Protocol (TCP); or User Datagram Protocol (UDP). Other protocols may be used to transfer the ML model.

[0050] With respect to the method of FIG. 2, the method may further include receiving, by the user device from the network node, a capabilities request; transmitting, by the user device to the network node, a capabilities response indicating that the user device has a capability to perform transfer of ML models.

[0051] With respect to the method of FIG. 2, the command may include a download command that instructs the user device to download to the user device the ML model from the storage location for the ML model indicated by the storage information; the method further including: downloading, by the user device from the storage location for the ML model, the ML model.

[0052] With respect to the method of FIG. 2, the control message may include a ML model metadata container including at least a functionality identifier that identifies a functionality for which the ML model is to be used to perform, wherein the method further including: performing, by the user device, the functionality specified by the functionality identifier using the downloaded ML model.

[0053] With respect to the method of FIG. 2, the storage information indicates a storage location from which the ML model may be downloaded, including the network address of the host node that is storing the ML model, and at least one of a path on the host node for the ML model or a file name associated with the ML model.

[0054] With respect to the method of FIG. 2, the command may include a download command instructing the user device (e.g., UE) to download the ML model from the storage location for the ML model; the method may further include providing, by the user device to the network node either: a download complete indication that indicates a download of the ML model was completed by the user device, or a download failure indication that indicates a download of the ML model failed.

[0055] With respect to the method of FIG. 2, the command may include a download command instructing the user device to download the ML model from the storage location for the ML model; and, wherein the control message further indicates a size of the ML model; the method further including: determining, by the user device, whether there are sufficient storage resources at the user device to store and/or use the ML model; and, transmitting, by the user device, to the network node, a control message including a failure indication indicating that a download of the ML model has failed including providing a failure reason indicating insufficient storage resources, if there are insufficient storage resources at the user devices to store and/or use the ML model.

[0056] With respect to the method of FIG. 2, the command may include a download command instructing the user device to download the ML model from the storage location for the ML model; wherein the control message may include a first checksum or hash for the ML model that may be used by the user device to verify integrity of the ML model that may be downloaded by the user device. Also, for example, the performing the transfer of the ML model may include: downloading, by the user device from the storage location for the ML model, the ML model. The method may further include: calculating a second checksum or hash for the downloaded ML model; comparing the second checksum or hash to the first checksum or hash to determine whether there is a match; determining whether an integrity of the downloaded ML model is verified or not based on whether the second checksum or hash matches the first checksum or hash; transmitting, by the user device to the network node, either: a model download complete indication confirming that download of the ML model has been completed if the integrity of the ML model was verified; or a model download failure indication indicating that a download of the ML model has failed including providing a failure reason indicating verification failure, if the integrity of the ML model was not verified.

[0057] With respect to the method of FIG. 2, the command may include an upload command that instructs the user device to upload the ML model to the storage location for the ML model.

[0058] With respect to the method of FIG. 2, the storage information may indicate a storage location to which the ML model may be uploaded by the user device, including the network address of the host node to store the ML model, and at least one of a path on the host node for the ML model and a file name associated with the ML model. [0059] With respect to the method of FIG. 2, the command may include an upload command that instructs the user device to upload the ML model to the storage location for the ML model; the method further including: uploading, by the user device, via the protocol indicated by the protocol information included in the control message, the ML model to the storage location.

[0060] With respect to the method of FIG. 2, the method may further include calculating, by the user device, a checksum or hash for the uploaded ML model; and transmitting, by the user device to the network node, the checksum or hash for the uploaded ML model to allow the network node to determine an integrity of the uploaded ML model.

[0061] With respect to the method of FIG. 2, the method may further include receiving, by the user device from the network node, either: an upload complete indication that indicates the upload of the ML model was completed by the user device and successfully received at the storage location, or an upload failure indication that indicates that the upload of the ML model failed.

[0062] With respect to the method of FIG. 2, the method may further include performing the following if the user device receives an upload failure indication for the ML model: re-uploading, by the user device to the storage location via the protocol indicated by the protocol information, the ML model; transmitting, or retransmitting, by the user device to the network node the checksum or hash for the re-uploaded ML model to allow the network node to determine an integrity of the re-uploaded ML model.

[0063] Various techniques are described that may provide or allow for ML model delivery or transfer between a UE and a network, such as the 3GPP wireless network, which may be or may include, for example, a RAN node or a gNodeB/gNB (e.g., which may be or may include a split arrangement of a centralized unit (CU) and/or distributed unit (DU)), a LMF, or a core network function (CNF), or other network node. It may be desirable or advantageous, for example, that the 3GPP network (or network node) is able to manage or control the ML model lifecycle, one aspect of which may be or include ML model delivery/transfer, which is the download or upload of a ML model (which may be or may include part of a ML model or a delta or difference of an ML model that indicates changes or difference with respect to a ML model). The various example embodiments and techniques described herein may take advantage of both the control plane (e.g., using control message(s) for requesting, commanding, initiating, controlling, managing, confirming, . . .) the ML model transfer, and the user (or data) plane for performing the ML model transfer (e.g., either upload by UE or download to the UE). [0064] As used herein, the terms delivery and transfer may mean the same thing and/or may be used interchangeably, e.g., which may refer to, or may include or mean, e.g., a communication, delivery, transfer, or transport of a ML model (e.g., either an upload of the ML Model from the UE to a storage location, or a download of the ML model to the UE from the storage location).

[0065] Thus, for example, the control plane and user (or data) plane may be used in conjunction to initiate the ML model delivery/transfer. For example, a network node may initiate, request or instruct the UE to perform a ML model transfer, e.g., by transmitting to the UE a control message via the control plane that may include a command (e.g., upload command or download command) for the UE to upload or download the ML model).

The control message transmitted by the network node to the UE to initiate, control or cause ML model transfer may also include storage information indicating a storage location for the ML model (e.g., where the ML model may be downloaded from, or where the ML model should be uploaded to), and protocol information indicating a protocol to be used by the UE for transfer of the ML model via the user (or data) plane. The storage information may indicate a storage location for the ML model, and, for example, may indicate or include: a network address of a host node (e.g., IP address or other network address of a server, or node within a network), and at least one of a path on the host node for the ML model (e.g., indicating a location on or within the host node where the ML model is stored for download or should be stored for upload), and/or a file name associated with the ML model.

[0066] Also, for example, the control message may also include a ML model metadata container, which may include ML model metadata, including one or more of the following ML model metadata: a ML model identifier (ML model ID) that uniquely identifies the ML model in the UE or in the network; a functionality identifier that identifies a functionality for which the ML model is to be used to perform (e.g., functionality ID indicating a function to be performed by the UE using the ML mode, e.g., such as power control, CSI report compression, beam selection, etc.); and/or an area indication (e.g., such as a PCI (physical cell identifier), TAC (tracking area code that identifies a tracking area that may include an area or multiple PCIs or cells), or a geofence indication that identifies a geographical area) that identifies an area or one or more cells for which the ML model is valid for or to be used for.

[0067] Control messages may also be exchanged or communicated between the UE and network node to manage and/or verify the ML model transfer, such as to confirm that ML model transfer was completed (indicating a successful ML model transfer) or has failed (indicating a failure of the ML model transfer). For example, an ML model transfer failure may be due to (or caused by), e.g., insufficient storage resources, or due to integrity verification failure (e.g., such as due to a non-match between a provided checksum or hash for the transferred ML model and a calculated checksum or hash for the transferred/received ML model).

[0068] Also, the control message received by the UE, that includes an upload command or a download command, may indicate or include a storage location for the ML model. The storage location for the ML model may indicate any storage location within the network, e.g., such as storage location on a network node, a server, the cloud, the core network, on a RAN node, etc. The storage location for the ML model data may indicate a storage location where the ML model will (or should) be uploaded to (for ML model upload command), or a storage location from which the UE may download the ML model from (for an ML model download command). The storage location for the ML model may indicate or may include, e.g., a network address of a host node (e.g., a RAN node, a server, a storage device or node in the cloud, a core network entity, LMF, or any node, server or storage device that is storing or may store the ML model), and at least one of a path (e.g., indicating a file location within host node for the ML Model) and/or a file name for (or associated with) the ML model. In other words, for example, the ML model storage location is not necessarily on the network node that transmitted the control message to initiate or request the ML model transfer by the UE.

[0069] In general, for example, the storage location for the ML model may (or may typically) be at a different node, device, or location than the network node that transmitted the control message to the UE to request or initiate ML model transfer. This ML model storage location being (at least in some cases) typically different from and/or independent from the network node that requested the ML model transfer may provide improved network flexibility, e.g., by allowing ML models to be stored at different locations (e.g., independent of a particular network node that may have requested the ML model transfer), and may, for example, allow for regionally applicable ML models to be stored centrally, without being limited to specific locations, specific network nodes or specific protocols, for example. Although, in some cases, the storage location for the ML model may be or may be provided on, or may be the same as, the network node that initiated or requested (transmitted the control message to the UE) the UE to perform the ML model transfer.

[0070] According to an example embodiment, the control plane, which could be, but is not limited to, RRC (e.g., such as a RRC message), NAS, or LPP, may be used to instruct or command the UE to download or upload a ML model (e.g., which may be a complete ML model, part of a ML model, or a delta or difference indication for a ML model indicating a difference or change in an ML model) via the user plane by indicating the download or upload protocol, such as, but not limited to, HTTP, FTP, TCP or raw UDP data transfer, the host node address, file location (e.g., path and/or file name) of the ML model, a verification hash or checksum, and a container (e.g., ML model metadata container) for ML model metadata. The user plane (e.g., which may be via UPF, and using the indicated protocol) may be used to deliver/transfer the ML model bytes. By using the user plane for the delivery/transfer of ML model, disadvantages or concerns about segmentation in the control plane or compromise of control plane functions are alleviated or overcome. Also, for example, the various example embodiments may allow for synchronization of the state of ML model availability of UEs in the network, while introducing little burden to the control plane, and allowing for the full use of the data capacity of the user plane for ML model transfer.

[0071] FIG. 3 is a diagram illustrating a machine learning (ML) model download by a user equipment (UE) or user device according to an example embodiment. As shown in FIG. 3, a UE 310 may be in communication with a gNB 312, a user plane function (UPF) 314 that is involved with communication of user data, and an access and mobility function (AMF) 316, where UPF and AMF are part of the core network. At step 1, the UE may perform a capabilities exchange with the gNB 312. For example, in response to receiving a capabilities request from gNB 312, the UE 310 may transmit to gNB 312 a capabilities response indicating that the UE 310 has a capability to perform transfer of ML models.

[0072] The network or network node (e.g., gNB 312) may have a ML model (e.g., a trained ML model, or portion of an ML model, or delta or difference for a model, or one or more parameters that the UE may use to configure a ML model) that it would like to transfer or provide to UE 310. Thus, at step 2 of FIG. 3, the gNB 312 transmits to UE 310 (and UE 310 receives from gNB 312) a control message (e.g., a RRC message) to control a transfer of a ML model. In this example, the control message, including a download command, is transmitted to the UE 310 to control, instruct, or cause the UE 310 to download a ML model from a storage location. For example, the control message transmitted to UE 310 may include a command for the UE 310 to download a ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane (e.g., via UPF 314).

[0073] As shown in FIG. 3, information 2 A includes information that (all or part of which) may be included in the control message that is transmitted by the gNB 312 to the UE 310 in step 2. As noted, the control message may include a download command, protocol information, and storage information. As shown in information 2A, the protocol information may indicate a protocol of FTP in this example, which should be used by the UE 310 to download the ML model from the storage location. The storage information in this example (shown in information 2A) may include, e.g., a network address (e.g., 10.10.10.10) of a host node, and at least one of a path (e.g., path:/models/ shown in information 2A) on the host node for the ML model (e.g., in this case path to the location of the ML model on the host node), and a file name (e.g., csiModelO in this example as shown in information 2A). As shown in the information 2A of FIG. 3, the control message may also include: a checksum or hash (e.g., md5_hash: ec55d3e698d289f2afd663725127bace in this example shown in information 2A of FIG. 2) for the ML model, which may be used by UE 310 to verify or confirm integrity of the downloaded ML model (e.g., to confirm or verify that the downloaded ML model is complete and accurate, without errors). Also, the information included within the control message may include a size (e.g., number of bytes) of the ML model.

[0074] Also, as shown in information 2A of FIG. 3, the control message received by UE 310 from gNB 312 may include a ML model metadata container that may include, e.g., one or more of the following metadata: a ML model identifier (e.g., modellD: xxx, shown in information 2A) that uniquely identifies the ML model in the user device or in a network, a functionality (or function) identifier (e.g., functionlD: xx, shown in information 2A) that identifies a functionality for which the ML model is to be used to perform or assist in performing; and/or an area indication (e.g., validArea: {TAC(s), PCI(s), geofence}, shown in information 2A of FIG. 3) that identifies an area or one or more cells for which the ML model is valid for or to be used for.

[0075] As shown in FIG. 3, at 3 A, the UE 310 may determine whether UE 310 has sufficient storage resources to store and/or use the ML model. If there are sufficient storage resources, then UE 310 does not send a reply or response to gNB 312 at this point. However, if there are insufficient storage resources at UE 310, the UE 310 may send a control message with failure indication to gNB 312. Thus, at step 3, if there are insufficient storage resources at UE 310 to store and/or use the ML mode, the UE 310 transmits to gNB 312 a control message including a failure indication (model download failure) indicating that a download by UE of the ML model has failed. UE 310 may indicate, in its control message transmitted to gNB 312, a failure reason indicating insufficient storage resources (e.g., failureReason: insufficient storage).

[0076] At step 4 of FIG. 3, if there are sufficient storage resources at the UE 310 to store and/or use the ML mode, then the procedure continues, and UE 310 establishes a PDU (protocol data unit) session with the UPF 314 and/or AMF 316. [0077] At step 5 of FIG. 3, the UE 310 downloads the indicated ML model from the storage location (e.g., from the host node address, via the indicated path or file name) indicated by the storage information via the protocol (via user plane function UPF 314) indicated by the protocol information (in this example indicating protocol of FTP).

[0078] At step 6 A of FIG. 3, the UE 310 may verify or confirm integrity of the downloaded ML model. The received hash, which was received within the control message of step 2 (within information 2A), may be a first checksum or hash. UE 310 may calculate a second checksum or hash for the downloaded ML model as a second checksum or hash. The UE 310 may compare these two checksums or hashes to verify integrity of the downloaded ML model. For example, UE 310 may: 1) compare the calculated second checksum or hash to the received first checksum or hash to determine whether there is a match, and 2) determine whether an integrity of the downloaded ML model is verified or not based on whether the second checksum or hash matches the first checksum or hash. These two checksums or hashes should match if the integrity of the downloaded ML model is verified or is correct, e.g., is without error.

[0079] Depending on whether these two checksums or hashes match or not, the UE 310 will either: at step 6B, transmit a model download complete indication (ModelDownloadComplete, shown in step 6B) to gNB 312 confirming that download of the ML model by UE 310 from the storage location has been completed, if the integrity of the ML model was verified (e.g., if these two checksums or hash values match); or at step 6C, transmit to gNB 312 a model download failure indication (e.g., ModelDownloadFailure) indicating that a download of the ML model has failed including providing a failure reason indicating verification failure (e.g., MD5VerificationFailure, shown in step 6C), if the integrity of the ML model was not verified (e.g., if these two checksums or hash values do not match, indicating the downloaded ML model is corrupt, inaccurate or has errors).

[0080] Thus, as shown in FIG. 3, the gNB commands the UE to download a ML model over the user plane (e.g., over an indicated protocol). The capability exchange procedure is used to determine whether the UE is capable of ML model delivery/transfer via download, and which protocols it supports. The gNB issues a ModelDownloadCommand specifying one of the download protocols indicated as available by the UE, the host (or host node) address of the ML model, the path to the ML model, the size of the ML model, and an MD5 hash used by the UE to verify that the model download was successful (verify integrity of the downloaded ML model). Additionally, a ML model metadata container may also be transmitted to the UE 310 (e.g., within the control message at step 2), which may include information to identify the model for network control purposes. As an example, a validity area is shown, which could aid in autonomous decisions about model activation, deactivation, or selection. The included metadata will be able to contain any information as specified by 3GPP, and the included parameters are not an exhaustive list.

[0081] If the UE 310 does not have sufficient storage for the ML model, it would issue or transmit a ModelDownloadFailure to gNB 312 with the reason insufficientStorage to cancel the model delivery/transfer procedure. Otherwise, a PDU session establishment sufficient for accessing the host address of the model is triggered, if not already established by attempting the ML model download procedure. The UE 310 initiates the ML model download procedure using the specified protocol. The download completes or is declared a failure by the UE 310. The UE 310 computes the MD5 hash (or other checksum) of the downloaded model and compares it to that supplied with the ModelDownloadCommand to verify the download was successful. The UE transmits ModelDownloadComplete to the gNB 312 if the verification was successful. Similarly, if the download failed, e.g., integrity verification for the ML model failed, UE 310 transmits ModelDownloadFailure to the gNB 312, indicating the download failed, as well as a failureReason, which could be, but is not limited to, insufficientStorage or MD5 validation failure.

[0082] The example embodiment shown in FIG. 3 for ML model download may alternatively be implemented in the core network through NAS messaging by replacing the capability exchange between the gNB 312 and UE 310 in step 1 with a NAS capability exchange between the UE 310 and the AMF 316. The contents of the ModelDownloadCommand in step 2 would remain the same, and just as before, a PDU session may be established in step 3. The remaining steps, 4-5, would remain the same, except that the model download response (ModelDownloadComplete or ModelDownloadFailure) would be directed at (transmitted by UE 310 to) the AMF 316 instead of the gNB 312.

[0083] Similarly, the example embodiment shown in FIG. 3 for ML model download may alternatively be implemented in the LMF, using LPP protocol messaging by replacing the capability exchange between the gNB 312 and UE in step 1 with a LPP capability exchange between the UE and the LMF (not shown in FIG. 3). The contents of the ModelDownloadCommand in step 2 would remain the same, and just as before, a PDU session would be established in step 3. The remaining steps, 4-5 would remain the same, except that the model download response (ModelDownloadComplete or ModelDownloadFailure) would be directed at (or transmitted by UE 310 to) the LMF instead of the gNB 312. [0084] In an example embodiment, co-location for the ML model storage location is not required. That is, the storage location (e.g., on the host node) may be on the same network node that initiated or requested ML model download (colocation), or the storage location may be provided on a different node, where the host node (or storage location for ML mode) is different from the network node that requested the ML model download (storage location is not co-located with gNB or network node that requested ML model download). Also, the UE should have access to the host node via UPF to be able to transfer the ML model to/from the storage location on the host node. Also, the controlling entity (e.g., the gNB, LMF, or AMF) that requests the UE to download or upload the ML model may typically have access to internal or external storage of the host node or storage location so that such controlling entity may also access such ML models, e.g., to verify integrity of an uploaded model, for example.

[0085] FIG. 4 is a diagram illustrating a machine learning (ML) model upload by a user equipment (UE) or user device according to an example embodiment. The upload example of FIG. 4 is similar to the ML model download example of FIG. 3, and the differences between these two figures will be described. As shown in FIG. 4, a UE 310 may be in communication with a gNB 312, a user plane function (UPF) 314 that is involved with communication of user data, and an access and mobility function (AMF) 316, where UPF and AMF are part of the core network. At step 1 , the UE may perform a capabilities exchange with the gNB 312. For example, in response to receiving a capabilities request from gNB 312, the UE 310 may transmit to gNB 312 a capabilities response indicating that the UE 310 has a capability to perform transfer of ML models.

[0086] The network or network node (e.g., gNB 312) may want the UE to provide or upload a ML model to a storage location. Thus, at step 2 of FIG. 4, the gNB 312 transmits to UE 310 (and UE 310 receives from gNB 312) a control message (e.g., a RRC message) to control a transfer of a ML model. In this example, the control message, including an upload command, is transmitted to the UE 310 to control, instruct, or cause the UE 310 to upload a ML model to a storage location. For example, the control message transmitted to UE 310 may include a command for the UE 310 to upload a ML model, storage information indicating a storage location to which the ML model should be uploaded, and protocol information indicating a protocol to be used by the user device for upload of the ML model via a user plane (e.g., via UPF 314).

[0087] As shown in FIG. 4, the control message may include all or part of the information 2 A, e.g., including an upload command, protocol information indicating a protocol to be used by the UE for upload, and storage information. As shown in information 2A of FIG. 4, the protocol information may indicate a protocol of FTP in this example, which should be used by the UE 310 to upload the ML model to the storage location. The storage information in this example (shown in information 2A) may include, e.g., a network address (e.g., 10.10.10.10) of a host node, and at least one of a path (e.g., path:/models/ shown in information 2A) on the host node for the ML model (e.g., in this case path to the location of where the ML model should be uploaded or stored on the host node), and a file name (e.g., csiModelO) that may indicate a name the ML model should be stored or uploaded as.

[0088] Also, in FIG. 4, for this upload procedure, the control message received from the gNB does not include a checksum or hash for the ML model. This is because: gNB 312 or other network entity (and not the UE) will verify integrity of the uploaded ML model. Therefore, the control message with upload command at step 2 does not typically include (and need not include) a checksum or hash value.

[0089] Also, as shown in information 2A of FIG. 4, the control message received by UE 310 from gNB 312 may include a ML model metadata container that may include, e.g., one or more of the following metadata: a ML model identifier that uniquely identifies the ML model in the user device or in a network, a functionality (or function) identifier that identifies a functionality for which the ML model is to be used to perform or assist in performing; and/or an area indication that identifies an area or one or more cells for which the ML model is valid for or to be used for.

[0090] At step 3 of FIG. 4, the UE 310 establishes a PDU (protocol data unit) session with a UPF 314.

[0091] At step 4 of FIG. 4, the UE 310 uploads the indicated ML model to the storage location (e.g., to the indicated path or file name on the host node via the protocol (via user plane function UPF 314) indicated by the protocol information (in this example indicating protocol of FTP).

[0092] At step 5 of FIG. 4, if an error in the ML model upload is detected by the UE, the UE 310 transmits a Model Upload response indicating upload failure at 5 A, and the gNB may respond with a model upload failure at 5B.

[0093] At step 6 of FIG. 4, if there is not failure detected by the UE 310 for the upload procedure, the UE 310 transmits to the gNB 312 a checksum or hash for the uploaded ML model, which may be used by the network node, gNB or other network entity to verify integrity of the uploaded ML model.

[0094] At step 7A of FIG. 4, the gNB 312 (or other network entity) may verify the integrity of the uploaded ML model, e.g., by calculating a checksum or hash for the received ML model, and comparing the calculated checksum or hash (calculated by gNB or network entity, based on uploaded ML model) to the received checksum or hash (provided by UE 310 to gNB 312), to verify integrity of the uploaded ML model. If integrity of the ML model is verified, then at step 7B of FIG. 4, the gNB 312 transmits to UE 310 a model upload complete indication. While if the integrity verification of the uploaded ML model has failed, at step 7C of FIG. 4, the gNB 312 transmits a model upload failure indication to UE 310, and may indicate a failure reason of verification failure, for example. The UE 310 may then re-upload the ML model (e.g., in response to the upload failure indication from gNB 312), or UE 310 may wait to receive another upload command from gNB 312 before re-uploading the ML model to the storage location.

[0095] FIGs. 5 and 6 are diagrams illustrating ML model download and upload procedures, respectively, that may be controlled or requested by a location management function (LMF) according to example embodiments. FIG. 5 is a diagram illustrating a ML model download procedure controlled by a LMF according to an example embodiment. FIG. 6 is a diagram illustrating a ML model download procedure controlled by a LMF according to an example embodiment. The message flow and operation for FIG. 5 (download) and FIG. 6 (upload) are generally the same or very similar to that shown in FIGs. 3 and 4, respectively, with LMF being the control entity instead of gNB, and control messages being provided via LPP instead of RRC, for example.

[0096] The LPP protocol is used by the LMF entity to send and receive messages. The LPP protocol specifies the bulk of its bidirectional messaging in two message types: RequestAssistanceData and ProvideAssistanceData. The UE requests data from the LMF, and the LMF provides data to the UE. RequestLocationlnformation and ProvideLocationlnformation messages may be used by LMF to request location information from the UE, and for the UE provides location information to the LMF. Additionally, the LPP protocol supports the following message types to indicate errors or stop processes: Abort allows for the cancellation of a procedure. Error allows the transmission of errors.

[0097] One example approach for model delivery/transfer for the LMF may include embedding the ModelDownload and ModelUpload messages or the ModelDelivery messages inside of the Request/Provide Locationinformation messages (to/from the LMF). To command the UE to download or upload an ML model, the LMF may send RequestLocationlnformation with a command to download or upload an ML model. FIG. 5 shows a possible LMF adaptation using the ModelDelivery message option, embedded in the Locationinformation messages, and using LPP Error for error delivery. The data contained in the ModelDownload and ModelUpload or ModelDelivery messages embedded inside of the LPP messages may be the same as defined previously.

[0098] Similarly, the ML model upload and download procedures may be applied to be used with NAS messages sent to/from the core network (e.g., AMF).

[0099] According to another example embodiment, rather than having separate messages for upload procedure and download procedure, one set of message could be used, with different contents or commands within such messages to accommodate upload or download procedure.

[0100] The ModelDownload and ModelUpload protocol messaging could be merged into ModelDelivery commands, making the following changes. ModelDownloadCommand and ModelUploadCommand would be merged into ModelDeliveryCommand, which would contain a direction field, with supported values of “download” and “upload”, and an optional MD5 hash field, only to be used for a direction of “download”. ModelDownloadComplete and ModelUploadComplete would be merged with ModelDeliveryComplete.

[0101] ModelDownloadFailure and ModelUploadFailure would be merged into ModelDeliveryFailure. ModelUploadResponse would be changed to ModelDeliveryResponse.

[0102] Some further examples will be provided.

[0103] Example 1. A method comprising: receiving, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and performing, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0104] Example 2. The method of Example 1, wherein the storage information, indicating a storage location for the ML model, comprises: a network address of a host node; and at least one of the following: a path on the host node for the ML model; or a file name.

[0105] Example 3. The method of any of Examples 1-2, wherein the control message further comprises a ML model metadata container including at least one of the following ML model metadata: a ML model identifier that uniquely identifies the ML model in the user device or in a network; a functionality identifier that identifies a functionality for which the ML model is to be used to perform; and/or an area indication that identifies an area or one or more cells for which the ML model is valid for or to be used for. [0106] Example 4. The method of Example 1 , wherein the ML model comprises at least one of: a complete ML model; a portion of a ML model; or a delta or change of a ML model indicating a change or difference of a ML model.

[0107] Example 5. The method of any of Examples 1-4, wherein the network node comprises at least one of the following: a gNB; an eNB; a base station or access point; a centralized unit (CU) and/or distributed unit (DU); a radio access network (RAN) node; an access and mobility function (AMF); a core network or core network node; or a location management function (LMF).

[0108] Example 6. The method of any of Examples 1-5, wherein the control message comprises at least one of the following: a first radio resource control (RRC) control message received from a gNB or a RAN (radio access network) node; a first LPP (LTE positioning protocol) control message received from a location management function (LMF); or a first NAS (network access stratum) control message received from the core network.

[0109] Example 7. The method of Example 6, further comprising the UE transmitting at least one of the following messages regarding the transfer of the ML model: a second radio resource control (RRC) control message transmitted to the gNB or the RAN (radio access network) node, in response to the first RRC control message; a second LPP (LTE positioning protocol) control message transmitted to the location management function (LMF) in response to the first LPP control message; or a second NAS (network access stratum) control message transmitted to the core network in response to the first NAS control message.

[0110] Example 8. The method of any of Examples 1-7 wherein the protocol to be used for transfer of the ML model via the user plane between the user device and the storage location indicated by the storage information comprises at least one of: File Transfer Protocol (FTP); Hypertext Transfer Protocol (HTTP); Transmission Control Protocol (TCP); or User Datagram Protocol (UDP).

[0111] Example 9. The method of any of Examples 1-8, further comprising: receiving, by the user device from the network node, a capabilities request; and, transmitting, by the user device to the network node, a capabilities response indicating that the user device has a capability to perform transfer of ML models.

[0112] Example 10. The method of any of Examples 1-9: wherein the command comprises a download command that instructs the user device to download to the user device the ML model from the storage location for the ML model indicated by the storage information; the method further comprising: downloading, by the user device from the storage location for the ML model, the ML model. [0113] Example 11. The method of Example 10, wherein the control message includes a ML model metadata container including at least a functionality identifier that identifies a functionality for which the ML model is to be used to perform, wherein the method further comprises: performing, by the user device, the functionality specified by the functionality identifier using the downloaded ML model.

[0114] Example 12. The method of any of Examples 10-11, wherein the storage information indicates a storage location from which the ML model may be downloaded, including the network address of the host node that is storing the ML model, and at least one of a path on the host node for the ML model or a file name associated with the ML model.

[0115] Example 13. The method of any of Examples 10-12: wherein the command comprises a download command instructing the user device to download the ML model from the storage location for the ML model; the method further comprising providing, by the user device to the network node either: a download complete indication that indicates a download of the ML model was completed by the user device, or a download failure indication that indicates a download of the ML model failed.

[0116] Example 14. The method of any of Examples 10-13: wherein the command comprises a download command instructing the user device to download the ML model from the storage location for the ML model; wherein the control message further indicates a size of the ML model; the method further comprising: determining, by the user device, whether there are sufficient storage resources at the user device to store and/or use the ML model; and, transmitting, by the user device, to the network node, a control message including a failure indication indicating that a download of the ML model has failed including providing a failure reason indicating insufficient storage resources, if there are insufficient storage resources at the user devices to store and/or use the ML model.

[0117] Example 15. The method of any of Examples 9-14: wherein the command comprises a download command instructing the user device to download the ML model from the storage location for the ML model; wherein the control message comprises a first checksum or hash for the ML model that may be used by the user device to verify integrity of the ML model that may be downloaded by the user device.

[0118] Example 16. The method of Example 15: wherein the performing the transfer of the ML model comprises: downloading, by the user device from the storage location for the ML model, the ML model; the method further comprising: calculating a second checksum or hash for the downloaded ML model; comparing the second checksum or hash to the first checksum or hash to determine whether there is a match; determining whether an integrity of the downloaded ML model is verified or not based on whether the second checksum or hash matches the first checksum or hash; transmitting, by the user device to the network node, either: a model download complete indication confirming that download of the ML model has been completed if the integrity of the ML model was verified; or a model download failure indication indicating that a download of the ML model has failed including providing a failure reason indicating verification failure, if the integrity of the ML model was not verified.

[0119] Example 17. The method of any of Examples 1-9, wherein the command comprises an upload command that instructs the user device to upload the ML model to the storage location for the ML model.

[0120] Example 18. The method of any of Examples 1-9 and 17, wherein the storage information indicates a storage location to which the ML model may be uploaded by the user device, including the network address of the host node to store the ML model, and at least one of a path on the host node for the ML model and a file name associated with the ML model.

[0121] Example 19. The method of any of Examples 1-9, 17 and 18: wherein the command comprises an upload command that instructs the user device to upload the ML model to the storage location for the ML model; the method further comprising: uploading, by the user device, via the protocol indicated by the protocol information included in the control message, the ML model to the storage location.

[0122] Example 20. The method of Example 19, the method further comprising: calculating, by the user device, a checksum or hash for the uploaded ML model; and transmitting, by the user device to the network node, the checksum or hash for the uploaded ML model to allow the network node to determine an integrity of the uploaded ML model.

[0123] Example 21. The method of any of Examples 1-9 and 17-20, the method further comprising: receiving, by the user device from the network node, either: an upload complete indication that indicates the upload of the ML model was completed by the user device and successfully received at the storage location, or an upload failure indication that indicates that the upload of the ML model failed.

[0124] Example 22. The method of Example 21, further comprising performing the following if the user device receives an upload failure indication for the ML model: re-uploading, by the user device to the storage location via the protocol indicated by the protocol information, the ML model; and, transmitting, or retransmitting, by the user device to the network node the checksum or hash for the re-uploaded ML model to allow the network node to determine an integrity of the re-uploaded ML model. [0125] Example 23. An apparatus comprising: at least one processor; and at least one memory including computer program code; the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus at least to perform the method of any of Examples 1-22.

[0126] Example 24. A non-transitory computer-readable storage medium comprising instructions stored thereon that, when executed by at least one processor, are configured to cause a computing system to perform the method of any of Examples 1-22.

[0127] Example 25. An apparatus comprising means for performing the method of any of Examples 1-22.

[0128] Example 26. An apparatus comprising: at least one processor; and at least one memory including computer program code; the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus at least to: receive, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and perform, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0129] Example 27. A non-transitory computer-readable storage medium comprising instructions stored thereon that, when executed by at least one processor, are configured to cause a computing system to: receive, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and perform, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0130] Example 28. An apparatus comprising: means for receiving, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and means for performing, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

[0131] FIG. 7 is a block diagram of a wireless station or node (e.g., UE, user device, AP, BS, eNB, gNB, RAN node, network node, TRP, or other node) 1300 according to an example embodiment. The wireless station 1300 may include, for example, one or more (e.g., two as shown in FIG. 7) RF (radio frequency) or wireless transceivers 1302 A, 1302B, where each wireless transceiver includes a transmitter to transmit signals and a receiver to receive signals. The wireless station also includes a processor or control unit/entity (controller) 1304 to execute instructions or software and control transmission and receptions of signals, and a memory 1306 to store data and/or instructions.

[0132] Processor 1304 may also make decisions or determinations, generate frames, packets or messages for transmission, decode received frames or messages for further processing, and other tasks or functions described herein. Processor 1304, which may be a baseband processor, for example, may generate messages, packets, frames, or other signals for transmission via wireless transceiver 1302 (1302A or 1302B). Processor 1304 may control transmission of signals or messages over a wireless network, and may control the reception of signals or messages, etc., via a wireless network (e.g., after being down- converted by wireless transceiver 1302, for example). Processor 1304 may be programmable and capable of executing software or other instructions stored in memory or on other computer media to perform the various tasks and functions described above, such as one or more of the tasks or methods described above. Processor 1304 may be (or may include), for example, hardware, programmable logic, a programmable processor that executes software or firmware, and/or any combination of these. Using other terminology, processor 1304 and transceiver 1302 together may be considered as a wireless transmitter/receiver system, for example.

[0133] In addition, referring to FIG. 7, a controller (or processor) 1308 may execute software and instructions, and may provide overall control for the station 1300, and may provide control for other systems not shown in FIG. 7, such as controlling input/output devices (e.g., display, keypad), and/or may execute software for one or more applications that may be provided on wireless station 1300, such as, for example, an email program, audio/video applications, a word processor, a Voice over IP application, or other application or software. [0134] In addition, a storage medium may be provided that includes stored instructions, which when executed by a controller or processor may result in the processor 1304, or other controller or processor, performing one or more of the functions or tasks described above.

[0135] According to another example embodiment, RF or wireless transceiver(s) 1302A/1302B may receive signals or data and/or transmit or send signals or data. Processor 1304 (and possibly transceivers 1302A/1302B) may control the RF or wireless transceiver 1302 A or 1302B to receive, send, broadcast, or transmit signals or data.

[0136] Embodiments of the various techniques described herein may be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or in combinations of them. Embodiments may be implemented as a computer program product, i.e., a computer program tangibly embodied in an information carrier, e.g., in a machine-readable storage device or in a propagated signal, for execution by, or to control the operation of, a data processing apparatus, e.g., a programmable processor, a computer, or multiple computers. Embodiments may also be provided on a computer readable medium or computer readable storage medium, which may be a non-transitory medium. Embodiments of the various techniques may also include embodiments provided via transitory signals or media, and/or programs and/or software embodiments that are downloadable via the Internet or other network(s), either wired networks and/or wireless networks. In addition, embodiments may be provided via machine type communications (MTC), and also via an Internet of Things (IOT).

[0137] The computer program may be in source code form, object code form, or in some intermediate form, and it may be stored in some sort of carrier, distribution medium, or computer readable medium, which may be any entity or device capable of carrying the program. Such carriers include a record medium, computer memory, read-only memory, photoelectrical and/or electrical carrier signal, telecommunications signal, and software distribution package, for example. Depending on the processing power needed, the computer program may be executed in a single electronic digital computer, or it may be distributed amongst a number of computers.

[0138] Furthermore, embodiments of the various techniques described herein may use a cyber-physical system (CPS) (a system of collaborating computational elements controlling physical entities). CPS may enable the embodiment and exploitation of massive amounts of interconnected ICT devices (sensors, actuators, processors microcontrollers, . . .) embedded in physical objects at different locations. Mobile cyber physical systems, in which the physical system in question has inherent mobility, are a subcategory of cyber-physical systems. Examples of mobile physical systems include mobile robotics and electronics transported by humans or animals. The rise in popularity of smartphones has increased interest in the area of mobile cyber-physical systems. Therefore, various embodiments of techniques described herein may be provided via one or more of these technologies.

[0139] A computer program, such as the computer program(s) described above, can be written in any form of programming language, including compiled or interpreted languages, and can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit or part of it suitable for use in a computing environment. A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network.

[0140] Method steps may be performed by one or more programmable processors executing a computer program or computer program portions to perform functions by operating on input data and generating output. Method steps also may be performed by, and an apparatus may be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).

[0141] Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer, chip or chipset. Generally, a processor will receive instructions and data from a read-only memory or a random access memory or both. Elements of a computer may include at least one processor for executing instructions and one or more memory devices for storing instructions and data. Generally, a computer also may include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto-optical disks, or optical disks. Information carriers suitable for embodying computer program instructions and data include all forms of non-volatile memory, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The processor and the memory may be supplemented by, or incorporated in, special purpose logic circuitry.

[0142] To provide for interaction with a user, embodiments may be implemented on a computer having a display device, e.g., a cathode ray tube (CRT) or liquid crystal display (LCD) monitor, for displaying information to the user and a user interface, such as a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0143] Embodiments may be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an embodiment, or any combination of such back-end, middleware, or front-end components. Components may be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include a local area network (LAN) and a wide area network (WAN), e.g., the Internet.

[0144] While certain features of the described embodiments have been illustrated as described herein, many modifications, substitutions, changes, and equivalents will now occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the various embodiments.

Claims

WHAT IS CLAIMED IS:

1. A method comprising: receiving, by a user device from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and performing, by the user device via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

2. The method of claim 1, wherein the storage information, indicating a storage location for the ML model, comprises: a network address of a host node; and at least one of the following: a path on the host node for the ML model; or a file name.

3. The method of any of claims 1-2, wherein the control message further comprises a ML model metadata container including at least one of the following ML model metadata: a ML model identifier that uniquely identifies the ML model in the user device or in a network; a functionality identifier that identifies a functionality for which the ML model is to be used to perform; or an area indication that identifies an area or one or more cells for which the ML model is valid for or to be used for.

4. The method of any of claims 1-3, wherein the ML model comprises at least one of: a complete ML model; a portion of a ML model; or a delta or change of a ML model indicating a change or difference of a ML model.

5. The method of any of claims 1-4, wherein the network node comprises at least one of the following: a gNB; an eNB; a base station or access point; a centralized unit (CU) and/or distributed unit (DU); a radio access network (RAN) node; an access and mobility function (AMF); a core network or core network node; or a location management function (LMF).

6. The method of any of claims 1-5, wherein the control message comprises at least one of the following: a first radio resource control (RRC) control message received from a gNB or a RAN (radio access network) node; a first LPP (LTE positioning protocol) control message received from a location management function (LMF); or a first NAS (network access stratum) control message received from the core network.

7. The method of claim 6, further comprising transmitting at least one of the following messages regarding the transfer of the ML model: a second radio resource control (RRC) control message transmitted to the gNB or the RAN (radio access network) node, in response to the first RRC control message; a second LPP (LTE positioning protocol) control message transmitted to the location management function (LMF) in response to the first LPP control message; or a second NAS (network access stratum) control message transmitted to the core network in response to the first NAS control message.

8. The method of any of claims 1-7 wherein the protocol to be used for transfer of the ML model via the user plane between the user device and the storage location indicated by the storage information comprises at least one of:

File Transfer Protocol (FTP);

Hypertext Transfer Protocol (HTTP);

Transmission Control Protocol (TCP); or

User Datagram Protocol (UDP).

9. The method of any of claims 1-8, further comprising: receiving, by the user device from the network node, a capabilities request; transmitting, by the user device to the network node, a capabilities response indicating that the user device has a capability to perform transfer of ML models.

10. The method of any of claims 1-9, wherein the command comprises a download command that instructs the user device to download the ML model from the storage location for the ML model indicated by the storage information; the method further comprising: downloading, by the user device from the storage location for the ML model, the ML model.

11. The method of claim 10, wherein the control message includes a ML model metadata container including at least a functionality identifier that identifies a functionality for which the ML model is to be used to perform, wherein the method further comprises: performing, by the user device, the functionality specified by the functionality identifier using the downloaded ML model.

12. The method of any of claims 10-11, wherein the storage information indicates a storage location from which the ML model is downloaded, including the network address of the host node that is storing the ML model, and at least one of a path on the host node for the ML model or a file name associated with the ML model.

13. The method of any of claims 10-12, wherein the command comprises a download command instructing the user device to download the ML model from the storage location for the ML model; the method further comprising providing, by the user device to the network node, either a download complete indication that indicates a download of the ML model was completed by the user device, or a download failure indication that indicates a download of the ML model failed.

14. The method of any of claims 10-13, wherein the command comprises a download command instructing the user device to download the ML model from the storage location for the ML model; wherein the control message further indicates a size of the ML model; the method further comprising: determining, by the user device, whether there are sufficient storage resources at the user device to store and/or use the ML model; transmitting, by the user device, to the network node, a control message including a failure indication indicating that a download of the ML model has failed including providing a failure reason indicating insufficient storage resources, if there are insufficient storage resources at the user devices to store and/or use the ML model.

15. The method of any of claims 9-14, wherein the command comprises a download command instructing the user device to download the ML model from the storage location for the ML model; wherein the control message comprises a first checksum or hash for the ML model used by the user device to verify integrity of the ML model to be downloaded by the user device.

16. The method of claim 15, wherein the performing the transfer of the ML model comprises: downloading, by the user device from the storage location for the ML model, the ML model; the method further comprising: calculating a second checksum or hash for the downloaded ML model; comparing the second checksum or hash to the first checksum or hash to determine whether there is a match; determining whether an integrity of the downloaded ML model is verified or not based on whether the second checksum or hash matches the first checksum or hash; and transmitting, by the user device to the network node, either: a model download complete indication confirming that download of the ML model has been completed if the integrity of the ML model was verified; or a model download failure indication indicating that a download of the ML model has failed including providing a failure reason indicating verification failure, if the integrity of the ML model was not verified.

17. The method of any of claims 1-9, wherein the command comprises an upload command that instructs the user device to upload the ML model to the storage location for the ML model.

18. The method of any of claims 1-9 and 17, wherein the storage information indicates a storage location to which the ML model to be uploaded by the user device, including the network address of the host node to store the ML model, and at least one of a path on the host node for the ML model and a file name associated with the ML model.

19. The method of any of claims 1-9, 17 and 18, wherein the command comprises an upload command that instructs the user device to upload the ML model to the storage location for the ML model; the method further comprising: uploading, by the user device, via the protocol indicated by the protocol information included in the control message, the ML model to the storage location.

20. The method of claim 19, further comprising: calculating, by the user device, a checksum or hash for the uploaded ML model; and transmitting, by the user device to the network node, the checksum or hash for the uploaded ML model to allow the network node to determine an integrity of the uploaded ML model.

21. The method of any of claims 1-9 and 17-20, further comprising: receiving, by the user device from the network node, either: an upload complete indication that indicates the upload of the ML model was completed by the user device and successfully received at the storage location, or an upload failure indication that indicates that the upload of the ML model failed.

22. The method of claim 21, further comprising performing the following if the user device receives an upload failure indication for the ML model: re-uploading, by the user device to the storage location via the protocol indicated by the protocol information, the ML model; transmitting, or retransmitting, by the user device to the network node the checksum or hash for the re-uploaded ML model to allow the network node to determine an integrity of the re-uploaded ML model.

23. An apparatus comprising: at least one processor; and at least one memory including computer program code; the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus at least to: receive, from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the apparatus to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the apparatus for transfer of the ML model via a user plane; and perform, via the protocol, a transfer of the ML model between the apparatus and the storage location indicated by the storage information.

24. The apparatus of claim 23, wherein the storage information, indicating a storage location for the ML model, comprises: a network address of a host node; and at least one of the following: a path on the host node for the ML model; or a file name.

25. The apparatus of any of claims 23-24, wherein the control message further comprises a ML model metadata container including at least one of the following ML model metadata: a ML model identifier that uniquely identifies the ML model in the apparatus or in a network; a functionality identifier that identifies a functionality for which the ML model is to be used to perform; or an area indication that identifies an area or one or more cells for which the ML model is valid for or to be used for.

26. The apparatus of any of claims 23-25, wherein the ML model comprises at least one of: a complete ML model; a portion of a ML model; or a delta or change of a ML model indicating a change or difference of a ML model.

27. The apparatus of any of claims 23-26, wherein the network node comprises at least one of the following: a gNB; an eNB; a base station or access point; a centralized unit (CU) and/or distributed unit (DU); a radio access network (RAN) node; an access and mobility function (AMF); a core network or core network node; or a location management function (LMF).

28. The apparatus of any of claims 23-27, wherein the control message comprises at least one of the following: a first radio resource control (RRC) control message received from a gNB or a RAN (radio access network) node; a first LPP (LTE positioning protocol) control message received from a location management function (LMF); or a first NAS (network access stratum) control message received from the core network.

29. The apparatus of claim 28, wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: transmit at least one of the following messages regarding the transfer of the ML model: a second radio resource control (RRC) control message transmitted to the gNB or the RAN (radio access network) node, in response to the first RRC control message; a second LPP (LTE positioning protocol) control message transmitted to the location management function (LMF) in response to the first LPP control message; or a second NAS (network access stratum) control message transmitted to the core network in response to the first NAS control message.

30. The apparatus of any of claims 23-29, wherein the protocol to be used for transfer of the ML model via the user plane between the apparatus and the storage location indicated by the storage information comprises at least one of:

File Transfer Protocol (FTP);

Hypertext Transfer Protocol (HTTP);

Transmission Control Protocol (TCP); or User Datagram Protocol (UDP).

31. The apparatus of any of claims 23-30, wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: receive, from the network node, a capabilities request; transmit, to the network node, a capabilities response indicating that the apparatus has a capability to perform transfer of ML models.

32. The apparatus of any of claims 23-31, wherein the command comprises a download command that instructs the apparatus to download the ML model from the storage location for the ML model indicated by the storage information; wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: download, from the storage location for the ML model, the ML model.

33. The apparatus of claim 32, wherein the control message includes a ML model metadata container including at least a functionality identifier that identifies a functionality for which the ML model is to be used to perform, wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: perform the functionality specified by the functionality identifier using the downloaded ML model.

34. The apparatus of any of claims 32-33, wherein the storage information indicates a storage location from which the ML model is downloaded, including the network address of the host node that is storing the ML model, and at least one of a path on the host node for the ML model or a file name associated with the ML model.

35. The apparatus of any of claims 32-34, wherein the command comprises a download command instructing the apparatus to download the ML model from the storage location for the ML model; wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: provide, to the network node, either a download complete indication that indicates a download of the ML model was completed by the apparatus, or a download failure indication that indicates a download of the ML model failed.

36. The apparatus of any of claims 32-35, wherein the command comprises a download command instructing the apparatus to download the ML model from the storage location for the ML model; wherein the control message further indicates a size of the ML model; wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: determine whether there are sufficient storage resources at the apparatus to store and/or use the ML model; transmit, to the network node, a control message including a failure indication indicating that a download of the ML model has failed including providing a failure reason indicating insufficient storage resources, if there are insufficient storage resources at the apparatuss to store and/or use the ML model.

37. The apparatus of any of claims 31-36, wherein the command comprises a download command instructing the apparatus to download the ML model from the storage location for the ML model; wherein the control message comprises a first checksum or hash for the ML model used by the apparatus to verify integrity of the ML model to be downloaded by the apparatus.

38. The apparatus of claim 37, wherein the performing the transfer of the ML model comprises the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: download, from the storage location for the ML model, the ML model; wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: calculate a second checksum or hash for the downloaded ML model; compare the second checksum or hash to the first checksum or hash to determine whether there is a match; determine whether an integrity of the downloaded ML model is verified or not based on whether the second checksum or hash matches the first checksum or hash; and transmit, to the network node, either: a model download complete indication confirming that download of the ML model has been completed if the integrity of the ML model was verified; or a model download failure indication indicating that a download of the ML model has failed including providing a failure reason indicating verification failure, if the integrity of the ML model was not verified.

39. The apparatus of any of claims 23-31, wherein the command comprises an upload command that instructs the apparatus to upload the ML model to the storage location for the ML model.

40. The apparatus of any of claims 23-31 and 39, wherein the storage information indicates a storage location to which the ML model to be uploaded by the apparatus, including the network address of the host node to store the ML model, and at least one of a path on the host node for the ML model and a file name associated with the ML model.

41. The apparatus of any of claims 23-31, 39 and 40, wherein the command comprises an upload command that instructs the apparatus to upload the ML model to the storage location for the ML model; wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: upload, via the protocol indicated by the protocol information included in the control message, the ML model to the storage location.

42. The apparatus of claim 41, wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: calculate a checksum or hash for the uploaded ML model; and transmit, to the network node, the checksum or hash for the uploaded ML model to allow the network node to determine an integrity of the uploaded ML model.

43. The apparatus of any of claims 23-31 and 39-42, wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: receive, from the network node, either: an upload complete indication that indicates the upload of the ML model was completed by the apparatus and successfully received at the storage location, or an upload failure indication that indicates that the upload of the ML model failed.

44. The apparatus of claim 43, wherein the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: perform the following if apparatus receives an upload failure indication for the ML model: re-upload, to the storage location via the protocol indicated by the protocol information, the ML model; transmit, or retransmit, to the network node, the checksum or hash for the re-uploaded ML model to allow the network node to determine an integrity of the re-uploaded ML model.

45. A non-transitory computer-readable storage medium comprising instructions stored thereon that, when executed by at least one processor, are configured to cause a computing system to: receive, from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for a user device to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the user device for transfer of the ML model via a user plane; and perform, via the protocol, a transfer of the ML model between the user device and the storage location indicated by the storage information.

46. An apparatus comprising: means for receiving, from a network node, a control message to control a transfer of a machine learning (ML) model, the control message including a command for the apparatus to download or upload the ML model, storage information indicating a storage location for the ML model, and protocol information indicating a protocol to be used by the apparatus for transfer of the ML model via a user plane; and means for performing, via the protocol, a transfer of the ML model between the apparatus and the storage location indicated by the storage information.