Transmission of machine learning models via the user plane, initiated by the control plane, for wireless networks.

Abstract

The method includes: a user device receiving a control message from a network node for controlling the 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 the transfer of the ML model over the user plane; and the user device transferring the ML model between the user device and the storage location indicated by the storage information via the protocol.

Description

Technical Field

[0001] This description relates to wireless communication.

Background Art

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

[0003] An example of a cellular communication system is an architecture standardized by the 3rd Generation Partnership Project (3GPP). Recent developments in this field are often referred to as 4G, or long-term evolution (LTE) of 3G, or Universal Mobile Telecommunications System (UMTS) radio access technology. E-UTRA (evolved UMTS Terrestrial Radio Access) is the radio interface of the 3GPP's Long Term Evolution (LTE) upgrade path for mobile networks. In LTE, a base station or access point (AP) called an evolved Node B (eNB) provides wireless access within a coverage area or cell. In LTE, a mobile device or mobile station is called a user equipment (UE). LTE includes many improvements or developments. Also, improvements to aspects of LTE continue.

[0004] The development of 5G New Radio (NR), like the previous evolutions of 3G and 4G wireless networks, is part of a continuous mobile broadband evolution process to meet the requirements of 5G. Furthermore, 5G targets not only mobile broadband but also newly emerging use cases. The goal of 5G is to provide a significant improvement in wireless performance, which may include new levels of data rates, latency, reliability, and security. 5G NR may also be scaled to efficiently connect large-scale Internet of Things (IoT) and may provide 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 under development or planned for development in the near future. [Overview of the Initiative] [Means for solving the problem]

[0005] The method may include: a user device receiving a control message from a network node for controlling the 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 the transfer of the ML model over the user plane; and the user device transferring the ML model between the user device and the storage location indicated by the storage information via the protocol.

[0006] The device may include: at least one processor and; at least one memory containing computer program code, wherein the at least one memory and the computer program code, using at least one processor, cause the device to receive control messages from a network node by a user device to control the transfer of a machine learning (ML) model, the control messages 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 the transfer of the ML model over the user plane; and the device is configured to cause the user device to transfer the ML model between the user device and the storage location indicated by the storage information via the protocol.

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

[0008] The device may include: means for receiving control messages from a network node by a user device for controlling the transfer of a machine learning (ML) model, wherein the control message includes 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 the transfer of the ML model over the user plane; and means for the user device to transfer the ML model between the user device and the storage location indicated by the storage information via the protocol.

[0009] Details of one or more examples of several embodiments are described in the accompanying drawings and the following description. Other features will become apparent from this specification and drawings, as well as from the claims. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram of a wireless network according to an exemplary embodiment. [Figure 2] This is a flowchart illustrating the operation of a user device (or UE) according to an exemplary embodiment. [Figure 3] This figure illustrates the download of a machine learning (ML) model by a user device (UE) or user equipment according to an exemplary embodiment. [Figure 4] This figure shows an example embodiment of uploading a machine learning (ML) model using a user device (UE). [Figure 5] This figure shows the download procedure for an ML model controlled by an LMF according to an exemplary embodiment. [Figure 6] This figure shows the download procedure for an ML model controlled by an LMF according to an exemplary embodiment. [Figure 7]This is a block diagram of a wireless station or node (e.g., a user node, user device, or UE, network node, relay node, gNB, or other node). [Modes for carrying out the invention]

[0011] Figure 1 is a block diagram of a wireless network 130 according to an exemplary embodiment. In the wireless network 130 of Figure 1, user devices 131, 132, 133, and 135 (sometimes called mobile stations (MS) or user equipment (UE)) may be connected to (and communicate with) a base station (BS) 134 (sometimes called an access point (AP), enhanced Node B (eNB), gNB, or network node). The terms user device and user equipment (UE) may be used interchangeably. The BS may include, or be called, a RAN (Radio Access Network) node, and may include a portion of the BS or a portion of a RAN node (e.g., a centralized unit (CU) and / or distributed unit (DU) in the case of a split BS or split gNB). At least part of the functionality of a BS (e.g., an access point (AP), base station (BS), or (e)node B (eNB), gNB, or RAN node) may be performed by any node, server, or host that can be operationally coupled to a transceiver such as a remote radio head. The BS (or AP) 134 provides wireless coverage within cell 136, including user devices (or UEs) 131, 132, 133, and 135. Although shown as having only four user devices (or UEs) connected to or attached to the BS 134, any number of user devices may be provided. The BS 134 is also connected to the core network 150 via the S1 interface 151. A location management function (LMF) is also connected to the BS / gNB 134 and the core network 150. This is just one simple example of a wireless network, and others could be used.

[0012] A base station (e.g., BS134) is an example of a radio access network (RAN) node in a wireless network. A BS (or RAN node) may be, or may include (or be referred to as) an access point (AP), gNB, eNB, or a part thereof (such as a central unit (CU) and / or distributed unit (DU) in the case of a split BS or split gNB), or other network nodes. A network node may refer to, or may include, a BS, AP, gNB, CU and / or DU, or a RAN node, for example. Also, in at least some cases, a network node may refer to, or may include, a core network (e.g., a core network node or core network entity, such as an AMF (Access and Mobility Function) or other core network entities), an LMF (Location Management Function), or other network nodes.

[0013] In a descriptive example, a BS node (e.g., BS, eNB, gNB, CU / DU, etc.) or a radio access network (RAN) may be part of a mobile communication system. A RAN (radio access network) may include one or more BS or RAN nodes that implement radio access technology, for example, enabling one or more UEs to access the network or core network. Thus, for example, a RAN (RAN node such as a BS or gNB) may be located between one or more user devices or UEs and the core network. According to an exemplary embodiment, each RAN node (e.g., BS, eNB, gNB, CU / DU, etc.) or BS may provide one or more wireless communication services for one or more UEs or user devices, for example, enabling UEs to wirelessly access the network via the RAN node. Each RAN node or BS may perform or provide wireless communication services, for example, enabling UEs or user devices to establish wireless connections to the RAN node, and sending data to and / or receiving data from one or more UEs. For example, after establishing a connection to the UE, a RAN node or network node (e.g., BS, eNB, gNB, CU / DU, etc.) can forward data received from the network or core network to the UE, and / or forward data received from the UE to the network or core network. A RAN node or network node (e.g., BS, eNB, gNB, CU / DU, etc.) can perform a variety of other wireless functions or services, such as broadcasting control information (e.g., system information or on-demand system information, etc.) to the UE, paging the UE when there is data to be transmitted to the UE, assisting with the handover of UEs between cells, scheduling resources for uplink data transmission from and downlink data transmission to the UE, and sending out control information to configure one or more UEs. These are just some examples of one or more functions that a RAN node or BS can perform.

[0014] User devices or user nodes (such as user terminals, user equipment (UEs), mobile terminals, and handheld wireless devices) may refer to portable computing devices, including wireless mobile communication devices that operate with or without a subscriber identification module (SIM), and may include, but are not limited to, the following types of devices: for example, mobile stations (MS), cell phones, cellular phones, smartphones, personal digital assistants (PDAs), handsets, devices using wireless modems (such as alarm or measuring devices), laptops and / or touchscreen computers, tablets, phablets, game consoles, notebooks, vehicles, sensors, and multimedia devices, or any other wireless devices. User devices may also be (or may include) devices that are almost exclusively uplink-only, for example, cameras or video cameras that load images or video clips onto the network. User nodes may also include user equipment (UEs), user devices, user terminals, mobile terminals, mobile stations, mobile nodes, subscriber devices, subscriber nodes, subscriber terminals, or other user nodes. For example, a user node can be used for wireless communication with one or more network nodes (e.g., gNB, eNB, BS, AP, CU, DU, CU / DU) and / or one or more other user nodes, regardless of the technology or radio access technology (RAT). In LTE (as an illustrative example), the core network may be called the Evolved Packet Core (EPC), which may include a Mobility Management Entity (MME) that can handle or assist with the mobility / handover of user devices between BSs, one or more gateways that can transfer data and control signals between BSs and the packet data network or the internet, and other control functions or blocks. Other types of wireless networks, such as 5G (sometimes called New Radio (NR)), may also include a core network.

[0015] Furthermore, the techniques described herein can be applied to various types of user devices or data service types, or to user devices on which multiple applications, which may be of different data service types, may be running. The development of New Radio (5G) can support several different applications or several different data service types, such as: machine-type communications (MTC), enhanced machine-type communications (eMTC), the Internet of Things (IoT), and / or narrowband IoT user devices, enhanced mobile broadband (eMBB), and ultra-high reliability low latency communications (URLLC). Many of these new 5G (NR) related applications generally require higher performance than previous wireless networks.

[0016] The Internet of Things (IoT) refers to a growing group of things that can have internet or network connectivity, and these things can send and receive information with other network devices. For example, many sensor-type applications or devices may monitor physical state or status and, for example, send reports to a server or other network device when an event occurs. Machine-Type Communications (MTC, or Machine to Machine communication) can be characterized by fully automated data generation, exchange, processing, and operation between intelligent machines, with or without human intervention. Enhanced Mobile Broadband (eMBB) can support much higher data rates than those currently available with LTE.

[0017] Ultra-high reliability low latency communication (URLLC) is a new data service type or new use scenario that can be supported for New Radio (5G) systems. This will enable the emergence of new applications and services such as industrial automation, autonomous driving, vehicle safety, and electronic medical services. 3GPP provides 10 explanatory examples. -5 The aim is to provide a reliable connection that accommodates a block error rate (BLER) of 1 millisecond and a maximum U-Plane (user / data plane) latency of 1 millisecond. Therefore, for example, a URLLC user device / UE may require a significantly lower block error rate and lower latency (which may or may not require high reliability simultaneously) than other types of user devices / UEs. Therefore, for example, a URLLC UE (or a URLLC application on the UE) may require much shorter latency compared to an eMBB UE (or an eMBB application running on the UE).

[0018] The techniques described herein may be applied to a variety of wireless technologies or wireless networks, including 5G (New Radio (NR)), cmWave and / or mmWave band networks, IoT, MTC, eMTC, eMBB, URLLC, 6G, or any other wireless network or wireless technology. These exemplary networks, technologies, or data service types are provided for illustrative purposes only.

[0019] According to exemplary embodiments, a machine learning (ML) model may be used within a wireless network to perform (or assist in performing) one or more tasks or functionalities. Generally, one or more nodes within a wireless network (e.g., BS, gNB, eNB, RAN node, user node, UE, user device, relay node, or other wireless node) may use or employ an ML model, such as a neural network model (e.g., a neural network, artificial intelligence (AI) neural network, AI neural network model, AI model, AI machine learning (AI ML) model or algorithm, model, or other terminology), to perform or assist in performing one or more ML-enabled tasks or functionalities. An ML-enabled task may include tasks that may be performed (or assist in performing) by an ML model, or tasks that an ML model has been trained to perform (or assist in performing).

[0020] ML-based algorithms or ML models can be used to perform and / or assist in performing various wireless-related functionalities, such as beam prediction in UEs (e.g., predicting the best beam or best beam pair based on a measured reference signal), antenna panel or beam control, RRM (Radio Resource Measurement) measurement and feedback (Channel State Information (CSI) feedback), CSI report compression, link monitoring, Transmit Power Control (TPC), and other radio resource management (RRM) functions or functions or tasks to improve network performance. In some cases, the use of ML models can be used to improve the performance of a wireless network when measured in one or more ways or by one or more performance indicators or performance criteria.

[0021] An ML model can be, for example, a computational model used in machine learning composed of nodes arranged in a layer-like fashion, or can include such a computational model. A node, also called an artificial neuron or simply a neuron, executes a function on the provided input to generate some output value. A neural network or ML model typically requires a training period to learn the parameters, i.e., weights, used to map the input to the desired output. The mapping is performed by a function. Thus, the weights are the weights regarding the mapping function of the neural network. Each neural network model or ML model may be trained for a specific task.

[0022] A neural network model or ML model should be trained to provide an output when an input is given, which may involve learning appropriate values for a number of parameters (e.g., weights) regarding the mapping function. Since the parameters are used to weight the terms of the mapping function, they are generally also called weights. This training can be an iterative process, and the values of the weights are fine-tuned over a number (e.g., thousands) of training rounds until they reach the optimal or most accurate values (or weights). In the context of a neural network (neural network model) or ML model, the parameters are often initialized with random values, and a training optimizer iteratively updates the parameters (weights) of the neural network to minimize the error of the mapping function. In other words, during each round or step of the iterative training, the network updates the values of the parameters, causing the values of the parameters to ultimately converge to the optimal values.

[0023] A neural network model or ML model can be trained, for example, either in a supervised or an unsupervised manner. In supervised learning, training examples are provided to a neural network model or other machine learning algorithm. The training examples include an input and a desired or previously observed output. The training examples are also called labeled data because the input is labeled with the desired or observed output. In the case of a neural network, the network learns values for the weights used in a mapping function that most frequently yields the desired output when a training input is given. In unsupervised training, a neural network model learns to identify structures or patterns within the provided input. In other words, the model identifies implicit relationships within the data. Unsupervised learning is used in many machine learning problems and typically requires a large set of unlabeled data.

[0024] According to an exemplary embodiment, the learning or training of a neural network model or ML model may be classified (or may include multiple categories) into multiple categories (including supervised and unsupervised) depending on whether there is a learning “signal” or “feedback” available to the model. Thus, for example, in the field of machine learning, there can be two main types of learning or training of a model: namely, supervised and unsupervised. The main difference between the two types is that supervised learning is performed using known or prior knowledge about what the output value for a particular data sample should be. Thus, the goal of supervised learning can be to learn a function that best approximates the input-output relationship observable in the data given a sample of the data and the desired output. On the other hand, unsupervised learning has no labeled output and thus its goal is to infer the natural structure that exists within a set of data points. ML model training can also include reinforcement learning.

[0025] There are challenges regarding the techniques to be used to communicate or transfer ML models to the UE, such as what messages or information should be provided to initiate ML model transfer, and whether a control plane and / or user (or data) plane should be used to initiate and / or transfer the ML model. Generally, the control plane may include control messages communicated to provide control over various aspects or functions of the wireless network, such as control messages that may be communicated for the coordination or control of: establishing connections, UE handover or cell changes, power control, configuring UEs to perform certain functions, etc. Data or user plans generally may include the transmission of user data to and from the UE. The control plane may typically include control messages such as radio resource control (RRC) messages sent to and from gNBs, CU / DUs, or other RAN nodes. Other types of control messages may include, for example, LPP (LTE Positioning Protocol) control messages sent to and from a Location Management Function (LMF), or Network Access Stratum (NAS) control messages sent to and from a core network (e.g., Access and Mobility Function (AMF)). The user plane may include the transmission of data to and from the UE using one or more user plane data transmission protocols (e.g., File Transfer Protocol (FTP), Hypertext Transfer Protocol (HTTP), Transmission Control Protocol (TCP), User Datagram Protocol (UDP), or other user plane protocols). Also, in at least some cases, packets or messages transmitted over the control plane may typically have different priorities and / or QoS (Quality of Service) compared to data transmitted over the user plane.It is unclear how the user plane and / or control plane should be used to enable the UE to download ML models (or otherwise facilitate the transfer of ML models to the UE).

[0026] The user (or data plane) may typically have greater capacity to communicate larger files or larger data chunks compared to control messages (or control plane messages). According to exemplary embodiments, the control plane (including one or more control messages, e.g., RRC messages, NAS messages, and / or LPP messages, or other control plane messages) may have very limited capacity to carry large files or critical information (e.g., ML models or parts of ML models), while control messages or the control plane may, advantageously, be used to provide control and / or communication to coordinate and / or initiate and / or confirm the transfer of an ML model. Similarly, the user (or data) plane can accommodate larger data transfers after the 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 can use the user (or data) plane to perform ML model transfers (e.g., downloading or uploading an ML model).

[0027] Accordingly, according to the exemplary embodiment, control messages (e.g., provided via a control plane) may be used to initiate, coordinate, confirm, and / or otherwise control the transfer of ML models to and from the UE, while the user (or data) plane may be used by the UE to perform the transfer (upload or download) of ML models. According to the exemplary embodiment, the transferred ML models may include, for example, trained or untrained models, and may include a complete ML model, a portion of one (or partial) ML model, and / or a delta or difference showing a change (or delta) of one (or another) ML model.

[0028] Ability exchange Capability exchange may also be a bidirectional message flow initiated by a network node or network entity such as a gNB, an Access and Mobility Function (AMF) within the core network, or a Location Management Function (LMF). The purpose may be to synchronize with the UE to understand which functions specified in the 3GPP specification are supported by the UE. According to an exemplary embodiment, the capability exchange procedure may be extended to allow the UE (and / or network node) to demonstrate its ability to transfer (e.g., upload and / or download) ML models, and may indicate one or more user (or data) plane protocols supported by the UE for downloading or uploading ML models.

[0029] Command and response messaging According to exemplary embodiments, command and response messaging structures may be used by UEs and network nodes to initiate, confirm control of, and / or indicate failure of ML model transfers via control plane messages (e.g., RRC messages 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. A Command may typically be sent by the network (or network node) to the UE. A CommandResponse may be sent by the UE, for example, to request additional data or additional information. For example, a CommandComplete may be sent by the network to the UE in the case of an upload, or from the UE to the network in the case of a download, when a procedure (e.g., ML model transfer) is successfully completed, and a CommandFailure may be sent by a network node to the UE in the case of a model transfer, or from the UE to the network in the case of an ML model download, when a procedure (ML model transfer) fails.

[0030] Transfer method According to exemplary embodiments, one or more user (or data) plane protocols may be used for ML model transfer. A non-exhaustive list of exemplary protocols includes, for example, HTTP, FTP, TCP, and UDP raw data transfer. These are various well-known basic protocols used for data transfer, for example, for download and upload. Each protocol has trade-offs to consider, such as transfer reliability and overhead, but these are left to implementation details. The important thing is that a data transfer protocol exists and that the UE and associated network entities need to communicate their ability to use the data transfer protocol. According to exemplary embodiments, a control message provided to the UE by a network node to initiate or request an ML model transfer may indicate (for example, one of these protocols) the protocol (e.g., user or data plane) that the UE should use to perform the ML model transfer.

[0031] Control plane and user plane According to exemplary embodiments, the control plane may include, or provide, a signaling radio bearer (SRB), and can be used to typically send and receive structured messages to initiate, request, configure, adjust, verify, and / or control a UE or user device for network connectivity procedures, perform and report measurements (e.g., send CSI reports), establish data connectivity, and perform other functions. Generally, the control plane is used for reliable, small amounts of control data. The user plane, through a data radio bearer (DRB), supports large amounts of data and is generally used for application layer traffic such as the data transfer protocols discussed above. According to exemplary embodiments, as will be described in more detail below, various embodiments described herein can use both a control plane (e.g., one or more control messages to initiate / request an ML model transfer and / or indicate completion / success of an ML model transfer and / or failure of an ML model transfer) and a user (or data) plane to perform an ML model transfer (e.g., upload or download an ML model). Additionally, the user (or data) plane may use UPF (User Plane Function).

[0032] Figure 2 is a flowchart illustrating the operation of a user device (or UE) according to an exemplary embodiment. Operation 210 includes the user device receiving a control message from a network node to control the 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 the storage location for the ML model, and protocol information indicating the protocol to be used by the user device for the transfer of the ML model over the user plane. Furthermore, operation 220 includes the user device transferring the ML model between the user device and the storage location indicated by the storage information via the protocol. The ML model transfer may be performed by the UE over the user (or data) plane using the protocol indicated by the protocol information contained in the control message.

[0033] Regarding the method in Figure 2, the storage information indicating the storage location for the ML model may include: the network address of the host node and at least one of the following: the path on the host node related to the ML model; or the file name.

[0034] With respect to the method in Figure 2, the control message may further include an ML model metadata container containing at least one of the following ML model metadata: an ML model identifier that uniquely identifies the ML model within the user device or network; a functionality identifier that identifies the functionality that the ML model should perform; and / or an area indicator that identifies the area or one or more cells in which the ML model is valid or can be used.

[0035] With respect to the method in Figure 2, the ML model may include at least one of the following: a complete ML model; a part of an ML model; or a delta or modification of an ML model showing changes or differences in an ML model compared to another ML model that the UE may already have, for example.

[0036] With respect to the method in Figure 2, a network node may include at least one of the following: gNB; eNB; base station or access point; centralized unit (CU) and / or distributed unit (DU); radio access network (RAN) node; access and mobility function (AMF); core network or core network node; or location management function (LMF).

[0037] With respect to the method in Figure 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 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.

[0038] With respect to the method in Figure 2, this method may further include the step of the UE sending at least one of the following messages relating to the forwarding of the ML model: a second RRC control message sent to a gNB or RAN (Radio Access Network) node in response to a first Radio Resource Control (RRC) control message; a second LPP control message sent to a Location Management Function (LMF) in response to a first LPP (LTE Positioning Protocol) control message; or a second NAS control message sent to the core network in response to a first NAS (Network Access Stratum) control message.

[0039] Regarding the method in Figure 2, the protocol to be used for transferring ML models over the user plane between the user device and the storage location indicated by the storage information may include at least one of the following (for example): File Transfer Protocol (FTP); Hypertext Transfer Protocol (HTTP); Transmission Control Protocol (TCP); or User Datagram Protocol (UDP). Other protocols may be used to transfer ML models.

[0040] With respect to the method shown in Figure 2, this method may further include the user device receiving a capability request from a network node and the user device sending a capability response to the network node indicating that the user device has the capability to transfer an ML model.

[0041] With respect to the method in Figure 2, the command may include 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 to the user device; this method further includes: the user device downloading the ML model from the storage location for the ML model.

[0042] With respect to the method in Figure 2, the control message may include an ML model metadata container that includes at least a functionality identifier that identifies the functionality to be performed using the ML model, and this method further includes: the user device using the downloaded ML model to perform the functionality specified by the functionality identifier.

[0043] Regarding the method in Figure 2, the storage information indicates a storage location from which the ML model can be downloaded, including the network address of the host node storing the ML model and at least one of the path on the host node associated with the ML model or the file name associated with the ML model.

[0044] With respect to the method in Figure 2, the command may include a download command that instructs a user device (e.g., UE) to download the ML model from a storage location for the ML model, and this method may further include a step by which the user device provides a network node with either: a download complete indicator indicating that the user device has finished downloading the ML model, or a download failure indicator indicating that the download of the ML model has failed.

[0045] With respect to the method in Figure 2, the command may include a download command that instructs the user device to download the ML model from a storage location for the ML model, and the control message may further indicate the size of the ML model; the method further includes: the user device determining whether it has sufficient storage resources to store and / or use the ML model; and the user device sending a control message to the network node indicating that the download of the ML model failed, including providing a reason for failure indicating insufficient storage resources if the user device does not have sufficient storage resources to store and / or use the ML model.

[0046] With respect to the method in Figure 2, the command may include a download command that instructs the user device to download the ML model from a storage location for the ML model, and the control message may include a first checksum or hash of the ML model that can be used by the user device to verify the integrity of the ML model that can be downloaded by the user device. Also, for example, transferring an ML model may include: the user device downloading the ML model from a storage location for the ML model. This method may further include: calculating a second checksum or hash of the downloaded ML model; comparing the second checksum or hash with the first checksum or hash to determine if they match; determining whether the integrity of the downloaded ML model is verified based on whether the second checksum or hash matches the first checksum or hash; and the user device sending one of the following to the network node: a model download complete indication confirming that the download of the ML model is complete if the integrity of the ML model is verified; or a model download failure indication indicating that the download of the ML model has failed, including providing a reason for failure indicating verification failure if the integrity of the ML model is not verified.

[0047] Regarding the method in Figure 2, the command may include an upload command that instructs the user device to upload the ML model to a storage location for the ML model.

[0048] Regarding the method in Figure 2, the storage information may indicate a storage location where the ML model can be uploaded by a user device, including the network address of the host node for storing the ML model, and at least one of the path on the host node related to the ML model and the file name associated with the ML model.

[0049] With respect to the method in Figure 2, the command may include an upload command that instructs the user device to upload the ML model to a storage location for the ML model; this method further includes: the user device uploading the ML model to the storage location via a protocol indicated by the protocol information contained in the control message.

[0050] With respect to the method in Figure 2, this method may further include: the user device calculating a checksum or hash of the uploaded ML model; and the user device sending the checksum or hash of the uploaded ML model to a network node so that the network node can determine the integrity of the uploaded ML model.

[0051] With respect to the method in Figure 2, this method may further include the user device receiving one of the following from the network node: an upload complete indicator indicating that the upload of the ML model has been completed by the user device and successfully received at the storage location, or an upload failure indicator indicating that the upload of the ML model has failed.

[0052] With respect to the method in Figure 2, this method may further include the steps of: if the user device receives an upload failure indication for an ML model, the user device re-uploads the ML model to the storage location via the protocol indicated by the protocol information; and the user device sends or re-transmits a checksum or hash of the re-uploaded ML model to the network node so that the network node can determine the integrity of the re-uploaded ML model.

[0053] Various techniques are described that can provide or enable ML model transmission or forwarding between a UE and a network such as a 3GPP wireless network, where the network may be, or include, a RAN node or gNodeB / gNB (e.g., a split arrangement of centralized units (CUs) and / or distributed units (DUs)), an LMF, or a core network function (CNF), or other network node. For example, it may be desirable or advantageous for a 3GPP network (or network node) to be able to manage or control the ML model lifecycle, one aspect of which may be, or include, ML model transmission / forwarding, which may be, or include, a download or upload of an ML model (which may be, or include, a portion of an ML model, or a delta or difference of an ML model indicating a change or difference to the ML model). Various exemplary embodiments and techniques described herein may utilize both a control plane (e.g., using control messages to request, command, initiate, control, manage, confirm, etc., ML model forwarding) and a user (or data) plane for performing ML model forwarding (e.g., uploading by or downloading to the UE).

[0054] As used herein, the terms transmission and transfer may mean the same thing and / or be used interchangeably, for example, to refer to, include, or mean the communication, transmission, transfer, or transport of an ML model (e.g., uploading an ML model from a UE to a storage location, or downloading an ML model from a storage location to a UE).

[0055] Therefore, for example, ML model propagation / transfer can be initiated by using both the control plane and the user (or data) plane. For example, a network node may initiate, request, or command an UE to perform an ML model transfer by sending a control message to the UE via the control plane, which may include commands for the UE to upload or download an ML model (e.g., an upload command or a download command). A control message sent by a network node to the UE to initiate, control, or trigger an ML model transfer may also include storage information indicating the storage location for the ML model (e.g., where the ML model can be downloaded from or where the ML model should be uploaded), and protocol information indicating the protocol to be used by the UE to transfer the ML model via the user (or data) plane. Storage information may indicate the storage location for the ML model and may indicate or include, for example: the network address of the host node (e.g., the IP address or other network address of the server or node in the network), the path on the host node for the ML model (e.g., indicating the location on or within the host node where the ML model is stored for download or should be stored for upload), and / or at least one of the filenames associated with the ML model.

[0056] Furthermore, for example, a control message may include an ML model metadata container, which may contain ML model metadata including one or more of the following: an ML model identifier (ML model ID) that uniquely identifies the ML model within the UE or network; a functionality identifier that identifies the functionality that the ML model is supposed to perform (e.g., a functionality ID indicating a function that the UE should perform using the ML mode, such as power control, CSI report compression, or beam selection); and / or an area indication that identifies the area or one or more cells in which the ML model is valid or can be used (e.g., PCI (physical cell identifier), TAC (tracking area code that identifies an area or a tracking area that may include multiple PCIs or cells), or a geofence indication that identifies a geographic area).

[0057] Control messages may also be exchanged or communicated between the UE and network nodes to manage and / or verify ML model transfers, such as confirming that an ML model transfer has completed (indicating success) or failed (indicating failure). For example, an ML model transfer failure may be due to (or caused by) insufficient storage resources, or due to a failure in integrity verification (e.g., a mismatch between the provided checksum or hash of the transferred ML model and the calculated checksum or hash of the transferred / received ML model).

[0058] Furthermore, control messages received by the UE, including upload or download commands, may indicate or include a storage location for the ML model. The storage location for the ML model may indicate any storage location in the network, such as a network node, server, cloud, core network, or RAN node. The storage location for the ML model data may indicate the storage location from which the ML model is uploaded (or should be uploaded) (in the case of an ML model upload command), or the storage location from which the UE can download the ML model (in the case of an ML model download command). The storage location for the ML model may indicate or include, for example, the network address of the host node (e.g., a RAN node, server, storage device or node in the cloud, core network entity, LMF, or any node, server, or storage device that stores or can store the ML model), the path (e.g., indicating the location of a file on the host node related to the ML model), and / or at least one of the file names related to (or associated with) the ML model. In other words, for example, the storage location of an ML model is not necessarily on the network node that sent the control message to initiate or request the transfer of the ML model by the UE.

[0059] Generally, for example, the storage location for an ML model may (or typically) be on a different node, device, or location than the network node that sent the control message to the UE to request or initiate the ML model transfer. This ML model storage location being (at least in some cases) typically different from and / or independent of the network node that requested the ML model transfer can improve network flexibility by allowing ML models to be stored in different locations (e.g., independent of the specific network node that may have requested the ML model transfer), and further, it can enable centralized storage of regionally applicable ML models without being limited to a specific location, a specific network node, or a specific protocol. However, in some cases, the storage location for an ML model may be on, or provided on, or the same as, the network node that initiated or requested the UE to perform the ML model transfer (sent the control message to the UE).

[0060] According to exemplary embodiments, a control plane, which may be, but is not limited to, RRC (e.g., RRC messages), NAS, or LPP, may be used to instruct or command the UE to download or upload an ML model (e.g., a complete ML model, a portion of an ML model, or an indication of the delta or difference of an ML model showing differences or changes to an ML model) via the user plane by indicating a download or upload protocol, e.g., but not limited to HTTP, FTP, TCP, or raw UDP data transfer, a host node address, the location of the ML model file (e.g., path and / or file name), a verification hash or checksum, and a container for the ML model metadata (e.g., an ML model metadata container). The user plane (e.g., often via UPF, using the indicated protocols) may be used to propagate / transfer ML model bytes. By using the user plane for propagating / transferring ML models, the drawbacks or concerns regarding segmentation or breach of control plane functionality in the control plane are mitigated or overcome. Furthermore, various exemplary embodiments, for example, can enable the synchronization of the ML model availability state of UEs within the network with minimal burden on the control plane, and allow for maximum utilization of the user plane's data capacity for ML model transfer.

[0061] Figure 3 shows a machine learning (ML) model downloaded by a user equipment (UE) or user device according to an exemplary embodiment. As shown in Figure 3, UE310 may communicate with gNB312, user plane functions (UPF)314 involved in the communication of user data, and access and mobility functions (AMF)316, the UPF and AMF being part of the core network. In step 1, UE can perform capability exchange with gNB312. For example, in response to receiving a capability request from gNB312, UE310 may send a capability response to gNB312 indicating that UE310 has the capability to transfer the ML model.

[0062] A network or network node (e.g., gNB312) may have an ML model that it wants to transfer or provide to the UE310 (e.g., a trained ML model, or a portion of an ML model, or a delta or difference related to the model, or one or more parameters that the UE can use to construct the ML model). Therefore, in step 2 of Figure 3, gNB312 sends a control message (e.g., an RRC message) to the UE310 to control the transfer of the ML model (and the UE310 also receives one from gNB312). In this example, a control message containing a download command is sent to the UE310 to control, instruct, or cause the UE310 to download the ML model from the storage location. For example, a control message sent to the UE310 may include a command for the UE310 to download the ML model, storage information indicating the storage location for the ML model, and protocol information indicating the protocol to be used by the user device for the transfer of the ML model over the user plane (e.g., over UPF314).

[0063] As shown in Figure 3, Information 2A includes information (all or part thereof) that may be included in the control message sent to the UE310 by the gNB312 in step 2. As previously mentioned, the control message may include a download command, protocol information, and storage information. As shown in Information 2A, the protocol information in this example may indicate the FTP protocol, which should be used by the UE310 to download the ML model from the storage location. The storage information in this example (shown in Information 2A) may include, for example, the network address of the host node (e.g., 10.10.10.10), the path on the host node to the ML model (e.g., in this case, the path to the location of the ML model on the host node) (e.g., the path shown in Information 2A: / models / ), and at least one of the file name (e.g., csiModel0 shown in Information 2A in this example). As shown in Information 2A of Figure 3, the control message may also include: a checksum or hash of the ML model (for example, in this example shown in Information 2A of Figure 2, md5_hash:ec55d3e698d289f2afd663725127bace). This may be used by the UE310 to verify or confirm the integrity of the downloaded ML model (for example, to confirm or verify that the downloaded ML model is complete, accurate, and error-free). The information contained within the control message may also include the size of the ML model (for example, in bytes).

[0064] Furthermore, as shown in Information 2A of Figure 3, control messages received by UE310 from gNB312 may include an ML model metadata container that may contain one or more of the following metadata: an ML model identifier that uniquely identifies the ML model within the user device or network (e.g., modelID:xxx shown in Information 2A); a functionality (or feature) identifier that identifies the functionality that the ML model should perform or assist in performing (e.g., functionID:xx shown in Information 2A); and / or an area indication that identifies the area or one or more cells in which the ML model is valid or can be used (e.g., validArea:{TAC(s),PCI(s),geofence} shown in Information 2A of Figure 3).

[0065] As shown in Figure 3, in 3A, UE310 can determine whether it has sufficient storage resources to store and / or use the ML model. If there are sufficient storage resources, UE310 does not send a reply or response to gNB312 at this point. However, if UE310 does not have sufficient storage resources, UE310 may send a control message to gNB312 that includes a failure indication. Thus, in step 3, if UE310 does not have sufficient storage resources to store and / or use the ML model, UE310 sends a control message to gNB312 that includes a failure indication (model download failed) indicating that the UE failed to download the ML model. In that control message sent to gNB312, UE310 may indicate a failure reason (e.g., failureReason: insufficient storage) indicating insufficient storage resources.

[0066] In step 4 of Figure 3, if the UE310 has sufficient storage resources to store and / or use ML mode, the procedure proceeds and the UE310 establishes a PDU (Protocol Data Unit) session with the UPF314 and / or AMF316.

[0067] In step 5 of Figure 3, the UE310 downloads the indicated ML model from the storage location indicated by the storage information (e.g., from the host node address, via the indicated path or file name) via the protocol indicated by the protocol information (in this example, the FTP protocol) (via the user plane function UPF314).

[0068] In step 6A of Figure 3, the UE310 can verify or confirm the integrity of the downloaded ML model. The received hash received in the control message of step 2 (in information 2A) may be the first checksum or hash. The UE310 may calculate a second checksum or hash for the downloaded ML model as the second checksum or hash. The UE310 can verify the integrity of the downloaded ML model by comparing these two checksums or hashes. For example, the UE310 may: 1) compare the calculated second checksum or hash with the received first checksum or hash to determine if they match, and 2) determine whether the integrity of the downloaded ML model has been verified based on whether the second checksum or hash matches the first checksum or hash. If the integrity of the downloaded ML model has been verified, or if it is correct, for example, error-free, these two checksums or hashes should match.

[0069] Depending on whether these two checksums or hashes match, UE310 does one of the following: If the integrity of the ML model is verified in step 6B (for example, if these two checksums or hash values ​​match), it sends a Model Download Complete indication (ModelDownloadComplete shown in step 6B) to gNB312, confirming that UE310 has finished downloading the ML model from the storage location; or, if the integrity of the ML model is not verified in step 6C (for example, if these two checksums or hash values ​​do not match, indicating that the downloaded ML model is corrupted, inaccurate, or contains errors), it sends a Model Download Failure indication (e.g., ModelDownloadFailure) to gNB312, including providing a reason for failure (e.g., MD5VerificationFailure shown in step 6C) indicating that the download of the ML model has failed.

[0070] Therefore, as shown in Figure 3, the gNB instructs the UE to download the ML model on the user plane (e.g., on the indicated protocol). The capability exchange procedure is used to determine whether the UE can transmit / transfer the ML model by download and which protocols the UE supports. The gNB issues a ModelDownloadCommand specifying one of the download protocols indicated as available to 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 the MD5 hash used by the UE to verify that the model download was successful (verify the integrity of the downloaded ML model). In addition, an ML model metadata container may also be sent to the UE310 (e.g., in a control message in step 2), which may contain information to identify the model for network control purposes. As an example, an effectiveness area may be indicated, which may assist in autonomous decisions regarding the activation, deactivation, or selection of the model. The included metadata may include any information specified by 3GPP, and the included parameters are not an exhaustive list.

[0071] If the UE310 does not have sufficient storage for the ML model, it issues or sends a ModelDownloadFailure with the reason insufficientStorage to the gNB312 to cancel the model propagation / transfer procedure. Otherwise, it triggers the establishment of a sufficient PDU session to access the model's host address, which has not yet been established by attempting the ML model download procedure. The UE310 initiates the ML model download procedure using the specified protocol. The download is either completed or declared as failed by the UE310. The UE310 calculates the MD5 hash (or other checksum) of the downloaded model and compares it to the one supplied in the ModelDownloadCommand to verify that the download was successful. If the verification is successful, the UE sends a ModelDownloadComplete to the gNB312. Similarly, if the download fails, for example, if the integrity verification of the ML model fails, the UE310 sends a ModelDownloadFailure indicating that the download failed, as well as a failureReason which may be insufficientStorage or MD5 verification failure, but is not limited to these.

[0072] The exemplary embodiment of ML model download shown in Figure 3 can, as an alternative, be implemented in the core network via NAS messaging by replacing the capability exchange between gNB312 and UE310 in step 1 with a NAS capability exchange between UE310 and AMF316. The content of ModelDownloadCommand in step 2 remains the same, and a PDU session can be established in step 3 as before. The remaining steps 4-5 remain the same except that the model download response (ModelDownloadComplete or ModelDownloadFailure) is addressed to AMF316 instead of gNB312 (sent to AMF316 by UE310).

[0073] Similarly, the exemplary embodiment of ML model download shown in Figure 3 can be implemented in the LMF using LPP protocol messaging by replacing the capability exchange between gNB312 and UE in step 1 with an LPP capability exchange between UE and LMF (not shown in Figure 3). The contents of ModelDownloadCommand in step 2 remain the same, and as before, a PDU session is established in step 3. The remaining steps 4-5 remain the same except that the model download response (ModelDownloadComplete or ModelDownloadFailure) is addressed to the LMF instead of gNB312 (or sent to the LMF by UE310).

[0074] In the exemplary embodiment, colocation of 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 the ML model download (colocation), or the storage location may be provided on a different node, in which case the host node (or storage location for ML mode) is different from the network node that requested the ML model download (the storage location is not in the same location as the gNB or network node that requested the ML model download). The UE should also be able to access the host node via UPF so that it can transfer the ML model to and from the storage location on the host node. Furthermore, a control entity (e.g., gNB, LMF, or AMF) that requests the UE to download or upload an ML model typically has access to the host node's internal or external storage or storage location, and therefore such a control entity can also access such ML models to verify, for example, the integrity of the uploaded model.

[0075] Figure 4 shows an example embodiment of uploading a machine learning (ML) model by a user equipment (UE) or user device. The upload example in Figure 4 is similar to the ML model download example in Figure 3, and the differences between these two figures will be explained. As shown in Figure 4, UE310 may communicate with gNB312, User Plane Function (UPF)314 involved in the communication of user data, and Access and Mobility Function (AMF)316, the UPF and AMF being part of the core network. In step 1, the UE can perform capability exchange with gNB312. For example, in response to receiving a capability request from gNB312, UE310 may send a capability response to gNB312 indicating that UE310 has the capability to transfer the ML model.

[0076] A network or network node (e.g., gNB312) may want the UE to provide or upload the ML model to a storage location. Therefore, in step 2 of Figure 4, gNB312 sends a control message (e.g., an RRC message) to the UE310 to control the transfer of the ML model (and the UE310 receives it from gNB312). In this example, a control message containing an upload command is sent to the UE310 to control, instruct, or cause the UE310 to upload the ML model to the storage location. For example, a control message sent to the UE310 may include a command for the UE310 to upload the ML model, storage information indicating the storage location where the ML model should be uploaded, and protocol information indicating the protocol to be used by the user device for uploading the ML model via the user plane (e.g., via UPF314).

[0077] As shown in Figure 4, the control message may include all or part of Information 2A, for example, an upload command, protocol information indicating the protocol to be used by the UE for the upload, and storage information. As shown in Information 2A in Figure 4, the protocol information may indicate the FTP protocol in this example, and this protocol should be used by the UE310 to upload the ML model to the storage location. The storage information in this example (shown in Information 2A) may include, for example, the network address of the host node (e.g., 10.10.10.10), the path on the host node related to the ML model (e.g., in this case, the path to where the ML model should be uploaded or stored on the host node) (e.g., the path shown in Information 2A: / models / ), and at least one of the file names that may indicate the name by which the ML model is stored or uploaded (e.g., csiModel0).

[0078] Furthermore, in Figure 4, the control message received from the gNB regarding this upload procedure does not include a checksum or hash of the ML model. This is because: the gNB312 or other network entity (not the UE) verifies the integrity of the uploaded ML model. Therefore, the control message containing the upload command in step 2 typically does not (and does not need to) include a checksum or hash value.

[0079] Furthermore, as shown in Information 2A of Figure 4, the control messages received by the UE310 from the gNB312 may include an ML model metadata container that may contain one or more of the following metadata: an ML model identifier that uniquely identifies the ML model within the user device or network; a functionality (or feature) identifier that identifies the functionality that the ML model should perform or assist in performing; and / or an area indicator that identifies the area or one or more cells in which the ML model is valid or can be used.

[0080] In step 3 of Figure 4, the UE310 establishes a PDU (Protocol Data Unit) session with the UPF314.

[0081] In step 4 of Figure 4, the UE310 uploads the indicated ML model to a storage location (for example, to the indicated path or filename on the host node) via the protocol indicated by the protocol information (in this example, the FTP protocol) (via the user plane function UPF314).

[0082] In step 5 of Figure 4, if the UE detects an error in the ML model upload, the UE310 can send a Model Upload response indicating upload failure in 5A, and the gNB can respond with Model Upload Failure in 5B.

[0083] In step 6 of Figure 4, if the UE310 does not detect any failures in the upload procedure, the UE310 sends a checksum or hash of the uploaded ML model to the gNB312, which can be used by network nodes, gNBs, or other network entities to verify the integrity of the uploaded ML model.

[0084] In step 7A of Figure 4, gNB312 (or another network entity) can verify the integrity of the uploaded ML model, for example by calculating a checksum or hash of the received ML model and comparing the calculated checksum or hash (calculated by gNB or the network entity based on the uploaded ML model) with the received checksum or hash (provided to gNB312 by UE310) to verify the integrity of the uploaded ML model. If the integrity of the ML model is verified, in step 7B of Figure 4, gNB312 sends a model upload complete indication to UE310. On the other hand, if the integrity verification of the uploaded ML model fails, in step 7C of Figure 4, gNB312 sends a model upload failure indication to UE310, which may indicate, for example, the reason for the verification failure. UE310 can then re-upload the ML model (for example, in response to the upload failure indication from gNB312), or UE310 can wait for another upload command from gNB312 before re-uploading the ML model to storage.

[0085] Figures 5 and 6 illustrate the download and upload procedures for ML models that may be controlled or requested by a Location Management Function (LMF) according to an exemplary embodiment, respectively. Figure 5 illustrates the download procedure for an ML model controlled by an LMF according to an exemplary embodiment. Figure 6 illustrates the download procedure for an ML model controlled by an LMF according to an exemplary embodiment. The message flow and operation for Figures 5 (download) and 6 (upload) are generally identical or very similar to those shown in Figures 3 and 4, respectively, except, for example, that the LMF is the controlling entity instead of the gNB, and the control messages are provided via LPP instead of RRC.

[0086] The LPP protocol is used by LMF entities to send and receive messages. The LPP protocol specifies the majority of its bidirectional messaging with two message types: RequestAssistanceData and ProvideAssistanceData. The UE requests data from the LMF, and the LMF provides data to the UE. RequestLocationInformation and ProvideLocationInformation messages may be used by the LMF to request location information from the UE, and by the UE to provide location information to the LMF. Furthermore, the LPP protocol supports the following message types to indicate errors or to halt a process: Abort allows for the cancellation of a procedure; Error allows for the sending of errors.

[0087] One exemplary technique for model delivery / transfer with respect to LMF may involve embedding ModelDownload and ModelUpload messages, or ModelDelivery messages, within a Request / Provide LocationInformation message (to / from LMF). To instruct the UE to download or upload an ML model, LMF may send a RequestLocationInformation along with the command to download or upload the ML model. Figure 5 shows a possible LMF adaptation using the ModelDelivery message option embedded in the LocationInformation message and using LPP Error for error delivery. The data contained in the ModelDownload and ModelUpload or ModelDelivery messages embedded within the LPP message may be the same as previously defined.

[0088] Similarly, ML model upload and download procedures can be adapted for use with NAS messages sent to and from the core network (e.g., AMF).

[0089] According to another exemplary embodiment, instead of having separate messages for the upload procedure and the download procedure, a single set of messages may be used, containing different content or commands within such messages to correspond to either the upload procedure or the download procedure.

[0090] The ModelDownload and ModelUpload protocol messaging may be merged into the ModelDelivery command, resulting in the following changes: ModelDownloadCommand and ModelUploadCommand are merged into ModelDeliveryCommand, which includes a direction field with supported values ​​"download" and "upload," and an optional MD5 hash field that can only be used for the "download" direction. ModelDownloadComplete and ModelUploadComplete are merged together with ModelDeliveryComplete.

[0091] ModelDownloadFailure and ModelUploadFailure will be merged into ModelDeliveryFailure. ModelUploadResponse will be changed to ModelDeliveryResponse.

[0092] Several further examples are provided.

[0093] Example 1. A method comprising: a user device receiving a control message from a network node for controlling the transfer of a machine learning (ML) model, wherein the control message includes 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 the transfer of the ML model over the user plane; and the user device transferring the ML model between the user device and the storage location indicated by the storage information via the protocol.

[0094] Example 2. The method according to Example 1, wherein the storage information indicating the storage location for the ML model includes the network address of the host node and at least one of the paths on the host node related to the ML model; or filename.

[0095] Example 3. The method according to Example 1 or 2, further comprising an ML model metadata container in which the control message further includes at least one of the following ML model metadata: an ML model identifier that uniquely identifies the ML model within a user device or network; a functionality identifier that identifies the functionality that the ML model should perform; and / or an area indicator that identifies an area or one or more cells in which the ML model is valid or can be used.

[0096] Example 4. The method of Example 1, wherein the ML model includes: a complete ML model; a portion of an ML model; or at least one of the deltas or modifications of an ML model that show a change or difference from the ML model.

[0097] Example 5. The method according to any one of Examples 1 to 4, wherein the network node includes at least one of: gNB; eNB; base station or access point; centralized unit (CU) and / or distributed unit (DU); radio access network (RAN) node; access and mobility function (AMF); core network or core network node; or location management function (LMF).

[0098] Example 6. The method according to any one of Examples 1 to 5, wherein the control message includes at least one of: a first radio resource control (RRC) control message received from a gNB or 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.

[0099] Example 7. The method as in Example 6, further comprising the UE sending at least one of the following messages relating to the transfer of an ML model: a second RRC control message sent to a gNB or RAN (Radio Access Network) node in response to a first Radio Resource Control (RRC) control message; a second LPP control message sent to a Location Management Function (LMF) in response to a first LPP (LTE Positioning Protocol) control message; or a second NAS control message sent to the core network in response to a first NAS (Network Access Stratum) control message.

[0100] Example 8. The method according to any one of Examples 1 to 7, wherein the protocol to be used for transferring ML models over the user plane between the user device and the storage location indicated by the storage information includes at least one of: File Transfer Protocol (FTP); Hypertext Transfer Protocol (HTTP); Transmission Control Protocol (TCP); or User Datagram Protocol (UDP).

[0101] Example 9. The method of any one of Examples 1 to 8, further comprising: the user device receiving a capability request from a network node; and the user device sending a capability response to the network node indicating that the user device has the capability to transfer an ML model.

[0102] Example 10. A download command that instructs a user device to download an ML model from a storage location for the ML model indicated by storage information to the user device; Method: The method of any of Examples 1 to 9, further comprising the user device downloading the ML model from a storage location for the ML model.

[0103] Example 11. The method of Example 10, wherein the control message includes an ML model metadata container that includes at least a functionality identifier that identifies the functionality to be performed using the ML model; and the method further includes the user device using the downloaded ML model to perform the functionality specified by the functionality identifier.

[0104] Example 12. The method according to Example 10 or 11, wherein the storage information indicates a storage location from which the ML model may be downloaded, the storage information including the network address of the host node storing the ML model and at least one of the path on the host node relating to the ML model or the file name associated with the ML model.

[0105] Example 13. A method comprising: a download command that instructs a user device to download an ML model from a storage location for the ML model; and a method that further comprises: the method of any of Examples 10 to 12, wherein the user device provides a network node with either a download complete indicator indicating that the user device has completed downloading the ML model, or a download failure indicator indicating that the download of the ML model has failed.

[0106] Example 14. The method of any one of Examples 10 to 13, further comprising: a download command that instructs a user device to download an ML model from a storage location for the ML model; a control message that further indicates the size of the ML model; and a method that: determines whether the user device has sufficient storage resources to store and / or use the ML model; and if the user device does not have sufficient storage resources to store and / or use the ML model, the user device sends a control message that includes a failure indication indicating that the download of the ML model failed, which includes providing a reason for failure indicating insufficient storage resources to the network node.

[0107] Example 15. The method of any one of Examples 9 to 14, wherein the command includes a download command that instructs the user device to download the ML model from a storage location for the ML model; and the control message includes a first checksum or hash of the ML model that can be used by the user device to verify the integrity of the ML model that can be downloaded by the user device.

[0108] Example 16. Transferring an ML model: The method of Example 15, further comprising: a user device downloading an ML model from a storage location for the ML model; the method: calculating a second checksum or hash of the downloaded ML model; comparing the second checksum or hash with a first checksum or hash to determine if they match; determining whether the integrity of the downloaded ML model is to be verified based on whether the second checksum or hash matches the first checksum or hash; and the user device sending to a network node either a model download complete indication, confirming that the download of the ML model is complete if the integrity of the ML model has been verified; or a model download failure indication, indicating that the download of the ML model has failed, including providing a reason for failure indicating verification failure if the integrity of the ML model has not been verified.

[0109] Example 17. The method of any of Examples 1 through 9, wherein the command includes an upload command that instructs the user device to upload the ML model to a storage location for the ML model.

[0110] Example 18. The method according to any one of Examples 1 to 9 and 17, wherein the storage information indicates a storage location where an ML model may be uploaded by a user device, the storage information including the network address of a host node for storing the ML model, and at least one of the path on the host node for the ML model and the file name associated with the ML model.

[0111] Example 19. The method according to any one of Examples 1 to 9, 17 and 18, wherein the command includes an upload command that instructs a user device to upload an ML model to a storage location for the ML model; and the method further includes the user device uploading the ML model to the storage location via a protocol indicated by protocol information contained in a control message.

[0112] Example 20. The method according to Example 19, further comprising: the user device calculating a checksum or hash of the uploaded ML model; and the user device sending the checksum or hash of the uploaded ML model to a network node so that the network node can determine the integrity of the uploaded ML model.

[0113] Example 21. The method of any one of Examples 1 to 9 and 17 to 20, further comprising receiving either an upload complete indication that the upload of the ML model from the network node by the user device has been completed by the user device and successfully received at the storage location, or an upload failure indication that the upload of the ML model has failed.

[0114] Example 22. The method of Example 21, further comprising: when a user device receives an upload failure indication for an ML model: the user device re-uploading the ML model to a storage location via the protocol indicated by the protocol information; and the user device sending or re-transmitting a checksum or hash of the re-uploaded ML model to a network node so that the network node can determine the integrity of the re-uploaded ML model.

[0115] Example 23. An apparatus comprising at least one processor and at least one memory containing computer program code, wherein the at least one memory and the computer program code are configured to cause the apparatus to perform at least one of the methods described in any of Examples 1 to 22, using at least one processor.

[0116] Example 24. A non-temporary computer-readable storage medium containing stored instructions, wherein the instructions are configured to cause a computing system to perform the actions described in any of Examples 1 to 22 when executed by at least one processor.

[0117] Example 25. An apparatus comprising means for performing the method described in any of Examples 1 to 22.

[0118] Example 26. A device comprising at least one processor and at least one memory containing computer program code, wherein the at least one memory and the computer program code are configured to cause the device to receive control messages from a network node by a user device for controlling the transfer of a machine learning (ML) model, the control messages comprising 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 the transfer of the ML model over the user plane; and the device is configured to cause the user device to transfer the ML model between the user device and the storage location indicated by the storage information via the protocol.

[0119] Example 27. A non-temporary computer-readable storage medium containing stored instructions, wherein when the instructions are executed by at least one processor, the computing system: causes a user device to receive a control message from a network node for controlling the transfer of a machine learning (ML) model, the control message comprising 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 the transfer of the ML model over the user plane; and is configured to cause the user device to transfer the ML model between the user device and the storage location indicated by the storage information via the protocol.

[0120] Example 28. A device comprising: means for a user device to receive a control message from a network node for controlling the transfer of a machine learning (ML) model, wherein the control message includes 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 the transfer of the ML model over the user plane; and means for the user device to transfer the ML model between the user device and the storage location indicated by the storage information via the protocol.

[0121] Figure 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 exemplary embodiment. The wireless station 1300 may include, for example, one or more (e.g., two as shown in Figure 7) RF (radio frequency) or wireless transceivers 1302A, 1302B, each wireless transceiver including a transmitter for transmitting signals and a receiver for receiving signals. The wireless station further includes a processor or control unit / entity (controller) 1304 for executing instructions or software and controlling the transmission and reception of signals, and a memory 1306 for storing data and / or instructions.

[0122] The processor 1304 may also make decisions or judgments, generate frames, packets, or messages for transmission, decode received frames or messages for further processing, and perform other tasks or functions described herein. For example, the processor 1304, which may be a baseband processor, may generate messages, packets, frames, or other signals and transmit them via the wireless transceiver 1302 (1302A or 1302B). The processor 1304 may control the transmission of signals or messages over a wireless network and may also control the reception of signals or messages over a wireless network (for example, after being down-converted by the wireless transceiver 1302). The processor 1304 may be programmable and may be capable of performing various tasks and functions described above, such as one or more of the tasks or methods described above, by executing software or other instructions stored in memory or other computer media. The processor 1304 may be, for example, a programmable processor that runs hardware, programmable logic, software, or firmware, and / or any combination thereof (or may include these). Using other terminology, the processor 1304 and the transceiver 1302 together can be considered, for example, a wireless transmitter / receiver system.

[0123] Furthermore, referring to Figure 7, the controller (or processor) 1308 can execute software and instructions, provide overall control over the station 1300, provide control over other systems not shown in Figure 7, such as control of input / output devices (e.g., displays and keypads), and / or run software for one or more applications that may be provided on the wireless station 1300, such as email programs, audio / video applications, word processors, Voice over IP applications, or other applications or software.

[0124] Furthermore, a storage medium containing stored instructions may be provided, which, when executed by a controller or processor, cause processor 1304 or other controller or processor to perform one or more of the functions or tasks described above.

[0125] According to another exemplary embodiment, the RF or wireless transceiver 1302A / 1302B can receive signals or data, and / or transmit or send signals or data. The processor 1304 (and optionally the transceiver 1302A / 1302B) can control the RF or wireless transceiver 1302A or 1302B to receive, send, broadcast, or transmit signals or data.

[0126] Embodiments of the various techniques described herein may be implemented in digital electronic circuits, or in computer hardware, firmware, software, or combinations thereof. Some embodiments may be implemented as computer program products, i.e., computer programs tangibly embodied in information carriers, such as in machine-readable storage devices or propagated signals, for execution by or control of the operation of data processing devices, such as programmable processors, computers, or multiple computers. Some embodiments may also be provided on computer-readable media or computer-readable storage media, which may be non-temporary media. Embodiments of the various techniques may also include embodiments provided via temporary signals or media, and / or program and / or software embodiments that are downloadable via the Internet or other networks (wired and / or wireless networks). Furthermore, some embodiments may be provided via machine-type communications (MTC) and also via the Internet of Things (IoT).

[0127] Computer programs can be stored in source code format, object code format, or some intermediate format, and can be any entity or device capable of carrying the program, such as a carrier, distribution medium, or computer-readable medium. Such carriers include, for example, recording media, computer memory, read-only memory, photoelectric and / or electrical carrier signals, communication signals, and software distribution packages. Depending on the processing power required, computer programs may run on a single electronic digital computer or be distributed across multiple computers.

[0128] Furthermore, embodiments of the various techniques described herein may utilize cyber-physical systems (CPS) (systems that coordinate computational elements to control physical entities). CPS can enable the realization and utilization of a large number of interconnected ICT devices (sensors, actuators, processors, microcontrollers, etc.) embedded in physical objects in various locations. Mobile cyber-physical systems, in which the physical system in question possesses inherent mobility, are a subcategory of cyber-physical systems. Examples of mobile-physical systems include mobile robots and electronic devices transported by humans or animals. The increasing popularity of smartphones has led to growing interest in the field of mobile cyber-physical systems. Therefore, various embodiments of the techniques described herein may be provided by one or more of these techniques.

[0129] Computer programs, such as those described above, may be written in any form of programming language, including compiled or interpreted languages, and may be deployed in any form, including as standalone programs or as modules, components, subroutines, or other units or parts thereof, suitable for use in a computing environment. Computer programs may be deployed to run on a single computer, on a single site, or on multiple computers distributed across multiple sites and interconnected by a communication network.

[0130] The method steps may be performed by one or more programmable processors that execute a computer program or a portion of a computer program to perform a function by performing calculations on input data and generating an output. The method steps may also be performed by a dedicated logic circuit such as an FPGA (Field-Programmable Gate Array) or ASIC (Application-Specific Integrated Circuit), and the device may be implemented as such a dedicated logic circuit.

[0131] Processors suitable for executing computer programs include, for example, both general-purpose and dedicated microprocessors, and any one or more processors in any type of digital computer, chip, or chipset. Generally, a processor receives instructions and data from read-only memory, random-access memory, or both. The 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 may also include one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or may be operationally coupled to receive data from or transfer data to such mass storage devices, or both. Information carriers suitable for embodying computer program instructions and data include, for example, semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks, such as internal hard disks or removable disks; magneto-optical disks; and all forms of non-volatile memory, such as CD-ROM and DVD-ROM disks. Processors and memory may be complemented by or incorporated into dedicated logic circuits.

[0132] To provide user interaction, some embodiments may be implemented on a computer having a display device for displaying information to the user, such as a cathode ray tube (CRT) or liquid crystal display (LCD) monitor, and a user interface such as a keyboard and pointing device, such as a mouse or trackball, on which the user can provide input to the computer. Other types of devices may also be used to provide user interaction; for example, the feedback provided to the user may be any form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback; furthermore, input from the user may be received in any form, including acoustic, voice, or tactile input.

[0133] Some embodiments may be implemented in a computing system that includes, for example, a backend component as a data server, or a middleware component such as an application server, or a frontend component such as a client computer having a graphical user interface or a web browser on which a user can interact with the embodiment, or any combination of such backend, middleware, or frontend components. The components may be interconnected by any form or medium of digital data communication, such as a communication network. Examples of communication networks include local area networks (LANs) and wide area networks (WANs), such as the Internet.

[0134] While some features of the embodiments described above have been illustrated as stated herein, those skilled in the art will likely recall numerous modifications, substitutions, alterations, and equivalent forms. Therefore, it should be understood that the appended claims are intended to cover all such modifications and alterations that fall within the true spirit of the various embodiments.

Claims

1. The user device receives a control message from a network node to control the transfer of a machine learning (ML) model, wherein the control message includes a command for the user device to download or upload the ML model, storage information indicating the storage location for the ML model, and protocol information indicating the protocol to be used by the user device for the transfer of the ML model over the user plane. The user device transfers the ML model between the user device and the storage location indicated by the storage information via a protocol. Methods that include...

2. Storage information indicating the storage location for the ML model, The host node's network address and The path on the host node related to the ML model, or file name at least one of the and The method according to claim 1, including the method described in claim 1.

3. The control message contains the following ML model metadata, namely, An ML model identifier that uniquely identifies an ML model within a user device or network. A functionality identifier that identifies the functionality that should be performed using the ML model, or Area indication identifies an area or one or more cells where an ML model is valid or can be used. The method according to claim 1 or 2, further comprising an ML model metadata container containing at least one of the following.

4. The ML model, Complete ML model, A part of the ML model, or ML model delta or change indicating a change or difference in the ML model The method according to any one of claims 1 to 3, comprising at least one of the following.

5. Network nodes gNB, eNB, Base station or access point, Centralized Unit (CU) and / or Distributed Unit (DU), Wireless Access Network (RAN) node, Access and Mobility Functions (AMF), The core network or core network node, or Location management function (LMF) The method according to any one of claims 1 to 4, comprising at least one of the following.

6. The control message is A first radio resource control (RRC) control message received from a gNB or RAN (radio access network) node, A first LPP (LTE Positioning Protocol) control message received from the position management function (LMF), or The first NAS (Network Access Stratum) control message received from the core network. The method according to any one of claims 1 to 5, comprising at least one of the following.

7. The following message regarding the transfer of ML models, namely, A second RRC control message is transmitted to a gNB or RAN (Radio Access Network) node in response to a first Radio Resource Control (RRC) control message. A second LPP control message sent to the LMF (Local Positioning Function) in response to a first LPP (LTE Positioning Protocol) control message, or A second NAS control message is sent to the core network in response to the first NAS (Network Access Stratum) control message. The method according to claim 6, further comprising transmitting at least one of the following.

8. The protocol to be used for transferring ML models via the user plane between the user device and the storage location indicated by the storage information is: File Transfer Protocol (FTP) Hypertext Transfer Protocol (HTTP), Transmission Control Protocol (TCP), or User Datagram Protocol (UDP) The method according to any one of claims 1 to 7, comprising at least one of the following.

9. The user device receives capability requests from the network node. The user device sends a capability response to the network node indicating that the user device has the capability to transfer ML models. The method according to any one of claims 1 to 8, further comprising:

10. The command includes 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 is The user device downloads the ML model from its storage location. The method according to any one of claims 1 to 9, further comprising:

11. The control message includes an ML model metadata container which includes at least a functionality identifier that identifies the functionality that the ML model should perform, and the method is The user device uses the downloaded ML model to perform the functionality specified by the functionality identifier. The method according to claim 10, further comprising:

12. The method according to claim 10 or 11, wherein the storage information indicates a storage location from which an ML model is downloaded, the storage information including the network address of a host node storing the ML model and at least one of a path on the host node relating to the ML model or a file name associated with the ML model.

13. The command includes a download command that instructs the user device to download the ML model from the storage location for the ML model. The method further includes providing a network node with either a download completion indicator indicating that the user device has completed downloading the ML model, or a download failure indicator indicating that the download of the ML model has failed. The method according to any one of claims 10 to 12.

14. The command includes a download command that instructs the user device to download the ML model from the storage location for the ML model. The control message further indicates the size of the ML model. The method is The user device determines whether it has sufficient storage resources to store and / or use the ML model. If the user device does not have sufficient storage resources to store and / or use the ML model, the user device sends a control message to the network node that includes a failure indication indicating that the ML model download failed, including providing a reason for failure indicating insufficient storage resources. The method according to any one of claims 10 to 13, further comprising:

15. The command includes a download command that instructs the user device to download the ML model from the storage location for the ML model. The control message includes a first checksum or hash of the ML model used by the user device to verify the integrity of the ML model to be downloaded by the user device. The method according to any one of claims 9 to 14.

16. Transferring ML models is possible. This includes downloading the ML model from the storage location for the ML model using the user device. The method is Calculate a second checksum or hash for the downloaded ML model, The second checksum or hash is compared with the first checksum or hash to determine if they match. The integrity of the downloaded ML model is verified based on whether the second checksum or hash matches the first checksum or hash. The user device communicates to the network node, If the integrity of the ML model is verified, a model download completion indicator will be displayed to confirm that the ML model download is complete, or If the integrity of the ML model could not be verified, a model download failure indication will be provided, including a reason for the verification failure. Sending either of the following The method according to claim 15, further comprising:

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

18. The method according to any one of claims 1 to 9 and 17, wherein the storage information indicates a storage location to which a user device should upload an ML model, the storage information including the network address of a host node for storing the ML model, and at least one of the path on the host node relating to the ML model and the file name associated with the ML model.

19. The command includes an upload command that instructs the user device to upload the ML model to a storage location for the ML model. The method is The user device uploads the ML model to the storage location via a protocol indicated by the protocol information contained in the control message. The method according to any one of claims 1 to 9, 17, and 18, further comprising:

20. The user device calculates a checksum or hash of the uploaded ML model, The user device sends a checksum or hash of the uploaded ML model to the network node, enabling the network node to determine the integrity of the uploaded ML model. The method according to claim 19, further comprising:

21. The user device receives either an upload complete indicator from the network node, indicating that the upload of the ML model has been completed by the user device and successfully received at the storage location, or an upload failure indicator, indicating that the upload of the ML model has failed. The method according to any one of claims 1 to 9 and 17 to 20, further comprising:

22. If the user device receives an upload failure indication regarding the ML model, The user device re-uploads the ML model to the storage location via the protocol indicated by the protocol information. The user device sends or resends a checksum or hash of the re-uploaded ML model to the network node, enabling the network node to determine the integrity of the re-uploaded ML model. The method according to claim 21, further comprising performing the following:

23. At least one processor, A device comprising at least one memory containing computer program code, At least one memory and computer program code are provided to the device using at least one processor, The network node receives a control message to control the transfer of a machine learning (ML) model, the control message includes a command for the device to download or upload the ML model, storage information indicating the storage location for the ML model, and protocol information indicating the protocol to be used by the device for transferring the ML model over the user plane. The protocol allows for the transfer of ML models between the device and the storage location indicated by the storage information. A device configured in such a way.

24. Storage information indicating the storage location for the ML model, The host node's network address and The path on the host node related to the ML model, or file name at least one of the and The apparatus according to claim 23, including the apparatus described in claim 23.

25. The control message contains the following ML model metadata, namely, An ML model identifier that uniquely identifies an ML model within a device or network. A functionality identifier that identifies the functionality that should be performed using the ML model, or Area indication identifies an area or one or more cells where an ML model is valid or can be used. The apparatus according to claim 23 or 24, further comprising an ML model metadata container containing at least one of the following.

26. The ML model, Complete ML model, A part of the ML model, or ML model delta or change indicating a change or difference in the ML model The apparatus according to any one of claims 23 to 25, comprising at least one of the following.

27. Network nodes gNB, eNB, Base station or access point, Centralized Unit (CU) and / or Distributed Unit (DU), Wireless Access Network (RAN) node, Access and Mobility Functions (AMF), The core network or core network node, or Location management function (LMF) The apparatus according to any one of claims 23 to 26, comprising at least one of the following.

28. The control message is A first radio resource control (RRC) control message received from a gNB or RAN (radio access network) node, A first LPP (LTE Positioning Protocol) control message received from the position management function (LMF), or The first NAS (Network Access Stratum) control message received from the core network. The apparatus according to any one of claims 23 to 27, comprising at least one of the following.

29. At least one memory and computer program code, using at least one processor, transmits the following message to the device regarding the transfer of an ML model, namely, A second RRC control message is transmitted to a gNB or RAN (Radio Access Network) node in response to a first Radio Resource Control (RRC) control message. A second LPP control message sent to the LMF (Local Positioning Function) in response to a first LPP (LTE Positioning Protocol) control message, or A second NAS control message is sent to the core network in response to the first NAS (Network Access Stratum) control message. The apparatus according to claim 28, configured to transmit at least one of the following.

30. The protocol to be used for transferring ML models via the user plane between the device and the storage location indicated by the storage information is: File Transfer Protocol (FTP) Hypertext Transfer Protocol (HTTP), Transmission Control Protocol (TCP), or User Datagram Protocol (UDP) The apparatus according to any one of claims 23 to 29, comprising at least one of the following.

31. At least one memory and computer program code are used in the device by at least one processor. The network node receives capability requests. The network node is instructed to send a capability response indicating that the device has the capability to transfer ML models. The apparatus according to any one of claims 23 to 30, configured as described above.

32. The command includes a download command that instructs the device to download the ML model from the storage location for the ML model indicated by the storage information, At least one memory and computer program code are used in the device by at least one processor. Download the ML model from its storage location. The apparatus according to any one of claims 23 to 31, configured as described above.

33. The control message includes an ML model metadata container which includes at least a functionality identifier that identifies the functionality to be performed using the ML model, and at least one memory and computer program code are provided to the device using at least one processor. The downloaded ML model is used to perform the functionality specified by the functionality identifier. The apparatus according to claim 32, configured as described above.

34. The apparatus according to claim 32 or 33, wherein the storage information indicates a storage location from which an ML model is downloaded, including the network address of a host node storing the ML model and at least one of a path on the host node relating to the ML model or a file name associated with the ML model.

35. The command includes a download command that instructs the device to download the ML model from the storage location for the ML model. At least one memory and computer program code are configured, using at least one processor, to cause the device to provide network nodes with either a download completion indicator indicating that the device has completed the download of the ML model, or a download failure indicator indicating that the download of the ML model has failed. The apparatus according to any one of claims 32 to 34.

36. The command includes a download command that instructs the device to download the ML model from the storage location for the ML model. The control message further indicates the size of the ML model. At least one memory and computer program code are used in the device by at least one processor. Determine whether the device has sufficient storage resources to store and / or use the ML model. The network node is instructed to send a control message that includes a failure indication indicating that the ML model download failed, including providing a reason for failure indicating insufficient storage resources if the device does not have enough storage resources to store and / or use the ML model. The apparatus according to any one of claims 32 to 35, configured as described above.

37. The command includes a download command that instructs the device to download the ML model from the storage location for the ML model. The control message includes a first checksum or hash of the ML model used by the device to verify the integrity of the ML model to be downloaded by the device. The apparatus according to any one of claims 31 to 36.

38. The transfer of an ML model involves at least one memory and computer program code, and the at least one memory and computer program code is transmitted to the device using at least one processor. Download the ML model from its storage location. It is configured in such a way, At least one memory and computer program code are used in the device by at least one processor. Calculate a second checksum or hash for the downloaded ML model. The second checksum or hash is compared with the first checksum or hash to determine if they match. The integrity of the downloaded ML model is verified based on whether the second checksum or hash matches the first checksum or hash. On the network node, If the integrity of the ML model is verified, a model download completion indicator will be displayed to confirm that the ML model download is complete, or If the integrity of the ML model could not be verified, a model download failure indication will be provided, including a reason for the verification failure. Send one of the following The apparatus according to claim 37, configured as described above.

39. The apparatus according to any one of claims 23 to 31, wherein the command includes an upload command that instructs the apparatus to upload an ML model to a storage location for the ML model.

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

41. The command includes an upload command that instructs the device to upload the ML model to a storage location for the ML model. At least one memory and computer program code are used in the device by at least one processor. The ML model is uploaded to the storage location via the protocol indicated by the protocol information contained in the control message. It is configured in such a way. The apparatus according to any one of claims 23 to 31, 39, and 40.

42. At least one memory and computer program code are used in the device by at least one processor. Calculate the checksum or hash of the uploaded ML model. Send a checksum or hash of the uploaded ML model to the network node so that the network node can determine the integrity of the uploaded ML model. The apparatus according to claim 41, configured as described above.

43. At least one memory and computer program code are used in the device by at least one processor. The network node should receive either an upload complete indicator, indicating that the ML model upload was completed by the device and successfully received at the storage location, or an upload failure indicator, indicating that the ML model upload failed. The apparatus according to any one of claims 23 to 31 and 39 to 42, configured as such.

44. At least one memory and computer program code, using at least one processor, transmits to the device an upload failure indication regarding an ML model when the device receives such an indication. Re-uploading the ML model to the storage location via the protocol indicated by the protocol information. Send or resend a checksum or hash of the re-uploaded ML model to the network node so that the network node can determine the integrity of the re-uploaded ML model. The apparatus according to claim 43, configured to perform the following.

45. A non-temporary computer-readable storage medium containing stored instructions, wherein when the instructions are executed by at least one processor, the computing system... The network node receives a control message to control the transfer of a machine learning (ML) model, the control message includes a command for the user device to download or upload the ML model, storage information indicating the storage location for the ML model, and protocol information indicating the protocol to be used by the user device for transferring the ML model over the user plane. The protocol facilitates the transfer of ML models between the user device and the storage location indicated by the storage information. A non-temporary, computer-readable storage medium configured in such a way.

46. Means for receiving control messages from a network node to control the transfer of a machine learning (ML) model, wherein the control message includes a command for the device to download or upload the ML model, storage information indicating the storage location for the ML model, and protocol information indicating the protocol to be used by the device for the transfer of the ML model via the user plane. Means for transferring ML models between a device and a storage location indicated by storage information via a protocol, A device equipped with the following features.

Images