Method for providing machine learning models to mobile devices

Abstract

A method for providing a machine learning model to a mobile terminal of a mobile communication system is described, comprising the steps of determining, by a mobile wireless communication network of the communication system, that a machine learning model to be used by the mobile terminal for a machine learning task should be distributed to the mobile terminal, and providing the machine learning model to the mobile terminal via user plane communication.

Description

[Technical Field]

[0001] The present disclosure relates to a method for providing machine learning models to a mobile terminal. [Background technology]

[0002] In communication systems such as 5G mobile communication systems, it is important to ensure that a certain quality of service can be maintained. To this end, various information, such as load, resource usage, available components, user mobility, component status, etc., may be monitored and taken into account when controlling the communication system, e.g., taking measures when overload is imminent to avoid degradation of service quality. This information can be collected and / or evaluated using machine learning models appropriately trained for the corresponding predictive tasks. For example, the use of artificial intelligence (AI) and machine learning (ML) techniques is a promising approach to improving the performance of next-generation radio access network (NG-RAN) air interfaces in several use cases, such as improved channel status information (CSI) feedback, beam management optimization, and positioning accuracy. However, this requires that, at least in some cases, the machine learning model be used on the terminal side, but due to the effort required for training, model training on the terminal side is difficult (due to constrained power and computational resources), and model training should typically be offloaded to the network, i.e., the model should be trained on the network side (e.g., a network function or an application function) and provided to the mobile terminal when needed. Therefore, an efficient technique for providing machine learning models to mobile terminals in a mobile communication system is desired. Summary of the Invention

[0003] A method for providing a machine learning model to a mobile terminal of a mobile communication system is provided, comprising the steps of determining, by a mobile wireless communication network of the communication system, that a machine learning model to be used by the mobile terminal for a machine learning task should be distributed to the mobile terminal, and providing the machine learning model to the mobile terminal via user plane communication. [Brief explanation of the drawings]

[0004] In the drawings, like reference numbers generally refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention. In the following description, various aspects are described with reference to the following drawings: [Figure 1] 1 illustrates a communication system according to an embodiment. [Figure 2] A flow diagram 200 showing the procedure for obtaining a machine learning (ML) model (on the network side) is shown. [Figure 3] 1 shows a flow diagram illustrating an embodiment in which an ML model is provided from a User Plane Function (UPF) to a mobile terminal via a Packet Data Unit (PDU) session between the mobile terminal and the UPF. [Figure 4] 1 shows a flow diagram illustrating an embodiment in which an ML model is provided to a mobile terminal from a Network Exposure Function (NEF) via a PDU session between the mobile terminal and the UPF and a User Plane (UP) tunnel between the UPF and the NEF. [Figure 5]1 shows a flow diagram illustrating an embodiment in which an ML model is provided from a Model Training (Network) Function (MTF) or a Model Repository (Network) Function (MRF) to a UPF via a user plane tunnel, and from there to a mobile terminal via a PDU session. [Figure 6] We present an application function (AF)-based approach for providing ML models to mobile devices. [Figure 7] A flow diagram illustrating an example procedure for an AF-based approach as shown in FIG. 6 is shown. [Figure 8] FIG. 1 is a flow diagram illustrating a method for providing a machine learning model to a mobile terminal of a mobile communication system according to an embodiment. [Figure 9] FIG. 1 is a flow diagram illustrating a method for obtaining a machine learning model from a mobile communication system according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0005] The following detailed description refers to the accompanying drawings, which show, by way of example, specific details and aspects of the present disclosure in which the invention may be practiced. Other aspects may be utilized, and structural, logical, and electrical changes may be made, without departing from the scope of the present invention. Various aspects of the present disclosure are not necessarily mutually exclusive, as some aspects of the present disclosure may be combined with one or more other aspects of the present disclosure to form new aspects.

[0006] Various examples corresponding to aspects of the present disclosure are described below.

[0007] Example 1 is a method for providing a machine learning model to a mobile terminal of a mobile communication system, the method comprising: determining, by a mobile wireless communication network of the communication system, that a machine learning model to be used by the mobile terminal for a machine learning task should be distributed to the mobile terminal; and providing the machine learning model to the mobile terminal via user plane communications.

[0008] Example 2 is the method of example 1, including determining that the machine learning model should be distributed to the mobile terminal based on one or more of temporal and / or spatial variations in operating conditions, performance degradation of the machine learning model, configuration of the mobile terminal, and operator policies of the mobile communications network.

[0009] Example 3 is the method of example 1, including, in response to receiving, by the mobile wireless communications network, a request for the machine learning model from the mobile terminal, determining that the machine learning model should be delivered to the mobile terminal, and, in response to the request, providing the machine learning model to the mobile terminal.

[0010] Example 4 is the method of example 3, including receiving the request by a user plane function of the communication network.

[0011] Example 5 is the method of example 4, including the user plane function discovering a storage location of the machine learning model, retrieving the machine learning model from the discovered storage location, and providing the machine learning model to the mobile terminal.

[0012] Example 6 is the method of any one of Examples 3-5, including providing the machine learning model to the mobile terminal via the packet data unit session.

[0013] Example 7 is the method of any one of Examples 3-5, including receiving a request via a packet data unit session and providing the machine learning model to the mobile terminal via the packet data unit session or another packet data unit session.

[0014] Example 8 is the method of any one of Examples 3-7, including receiving the request by a core network component of the wireless communication network via a user plane tunnel.

[0015] Example 9 is the method of any one of Examples 3-8, including receiving the request at a user plane function of the wireless communication network and forwarding the request to a network publishing function of the wireless communication network via a user plane tunnel.

[0016] Example 10 is the method of example 9, including the steps of the network publishing function discovering a storage location of the machine learning model, obtaining the machine learning model from the discovered storage location, and providing the machine learning model to the mobile terminal via the user plane function.

[0017] Example 11 is the method of example 3, including receiving a request at an application function of a wireless communication network.

[0018] Example 12 is the method of example 11, including the application function discovering a storage location of the machine learning model, retrieving the machine learning model from the discovered storage location, and providing the machine learning model to the mobile terminal.

[0019] Example 13 is the method of any one of Examples 3 to 12, including receiving a request via the quality of service flow and providing the machine learning model to the mobile terminal in response to the request via the quality of service flow or another quality of service flow.

[0020] Example 14 is the method of any one of Examples 1-13, including retrieving the machine learning model from a storage device that stores the machine learning model via control plane communication or via a user plane tunnel.

[0021] Example 15 is the method of any one of Examples 1-14, wherein the machine learning model is used by the mobile terminal for prediction tasks and / or for compression and / or encoding and decompression and / or decoding tasks.

[0022] Example 16 is a communication network configured to perform the method of any one of Examples 1-15.

[0023] Example 17 is a method for obtaining a machine learning model from a mobile communication system, comprising: sending a request from the mobile terminal to a mobile wireless communication network of the communication system for a machine learning model to be used by the mobile terminal for a machine learning task; and receiving the machine learning model in response to the request via user plane communications.

[0024] Example 18 is the method of example 17, including determining, by the mobile terminal, that the machine learning model should be requested based on one or more of temporal and / or spatial variations in operating conditions, performance degradation of the machine learning model, configuration of the mobile terminal, and operator policies of the mobile communications network.

[0025] Example 19 is a mobile terminal configured to perform the method of example 17 or 18.

[0026] It should be noted that one or more features of any of the above examples may be combined with any one of the other examples. In particular, embodiments described in the context of a device are equally valid for a method.

[0027] According to further embodiments, there is provided a computer program and computer readable medium comprising instructions which, when executed by a computer, cause the computer to perform the method of any one of the above examples.

[0028] Various examples are described in more detail below.

[0029] FIG. 1 shows a communication system according to an embodiment.

[0030] In this example, the communication system includes a radio access network (RAN) 101, a core network 103, and a transport (communication) network 102 connecting the RAN 101 to the core network 103. Subscriber terminals (denoted as UEs according to 3GPP) 104 are connected to distributed units 105 of the RAN 101 which are connected to a centralized unit 106 (for implementing a base station of the RAN 101). Application functions 107 are connected to the core network 103. The core network 103, for example a 5G core network (5GC), includes various network functions (NFs), such as an Access and Mobility Management Function (AMF) 108, a Session Management Function (SMF) 109, a Network Exposure Function (NEF) 110, a User Plane Function (UPF) 111, a Network Data Analytics Function (NWDAF) 112, and a Policy Control Function (PCF) 113.

[0031] The core network 103 is coupled to an Operation, Administration and Maintenance (OAM) system 114 .

[0032] In a communication system such as that shown in FIG. 1, machine learning (ML) models may be used to perform predictive tasks for various purposes, such as optimizing the performance of the radio link between the subscriber terminal 104 and the RAN 101, such as: Improved channel state information (CSI) feedback, e.g., reduced overhead and improved accuracy of predictions Spatial-frequency domain CSI compression using two-sided AI models Time-domain CSI prediction using UE-side model Beam management, e.g., beam prediction in the time and / or spatial domain to reduce overhead and latency and improve beam selection accuracy Spatial domain downlink beam prediction for a first set of beams A based on measurements of a second set of beams Temporal downlink beam prediction for a first set of beams based on historical measurements of a second set of beams Improved positioning accuracy for different scenarios, including those with heavy non-line of sight (NLOS) conditions ○Direct AI / ML positioning ○AI / ML-assisted positioning Furthermore, AI / ML technologies, e.g. Network energy savings, e.g., ML models are used with Input: UE mobility / trajectory, current energy efficiency, UE measurement reports, etc. Output: predicted energy efficiency, handover strategy, etc. Load balancing, e.g., ML models are used using Input: UE trajectory, UE traffic, RAN resource status, etc. Output: Predicted self-resource status information, etc. Mobility optimization, e.g., ML models are used with Input: UE location information, radio measurements, UE handover, etc. Output: Predicted handovers, UE traffic predictions, etc. However, it can be used, among other things, to improve RAN performance, for example.

[0033] For one of the above purposes, the corresponding ML model may be trained by an entity, for example, a RAN entity, in particular, a base station (denoted as gNB in ​​5G), an OAM 114, a component of the core network 103 (e.g., a network function), or a third party (e.g., corresponding to the AF 107 or connected to the core network 103 via the AF 107). This training entity is hereinafter referred to as a Model Training NF (MTF). The ML model can be trained using training data collected from other NFs, UEs, OAMs, etc. The MTF can be, for example, an NWDAF 112 including a Model Training Logical Function (MTLF) with an extension supporting training models for RAN use cases (e.g., CSI prediction models) that provide models to NFs other than Analytical Logical Functions (AnLFs), such as the AMF, UPF, and SMF. The Model Training Entity or MTF collects training data and trains the respective ML model. The trained model (i.e., the specification of the trained model in terms of, for example, neural network weights) is then stored in a storage, hereafter referred to as a Model Repository NF (MRF), which can store the trained model and provide the model to other components and entities (e.g., NFs). The MRF is an Analytics Data Repository Function (ADRF) that has the ability to provide the stored model to NFs other than AnLF, such as AMF, UPF, and SMF.

[0034] The trained model may then be used at multiple inference locations, which may include, among other things, one or more mobile terminals. Thus, according to various embodiments, techniques are provided for distributing ML models from the network side, e.g., the RAN, 5GC, OAM, or a third party or AF, to the UE (via the RAN).

[0035] According to various embodiments, model distribution is enabled by extending the interfaces in the UE and 5GC NF for model distribution in the user plane (UP). It should be noted that training locations inside and outside the core network 103 (e.g., 5GC) are supported. The ML model may be a RAN model, i.e., an ML model trained specifically for a RAN task, i.e., to perform predictions related to RAN operation.

[0036] In the embodiments described below, it is assumed that there is a registration and discovery procedure whereby an MTF registers its supported models along with associated ML model filter information with a network repository function (NRF), and other NFs discover and select MTFs via querying the NRF for specific models and corresponding filters.

[0037] FIG. 2 shows a flow diagram 200 illustrating the ML model acquisition procedure (on the network side).

[0038] The flow includes a consumer NF 201 (i.e., a network function (NF) that acts as a consumer, i.e., obtains a trained ML model), such as an AMF, SMF, UPF, or NEF, an NRF 202, an MTF 203, and an MRF 204.

[0039] At 205, the MTF 203 registers with the NRF 202, informing the NRF 202, among other things, that the MTF 203 has ML model training and serving capabilities.

[0040] At 206, the consumer NF 201 sends a discovery request to the NRF 202, and at 207 the NRF 202 responds to the NRF 220 with a discovery request response informing the consumer NF 201 about the MTF 203.

[0041] At 208, the consumer NF 201 requests (by get or subscribe) an ML model from the MTF. The request may include various parameters related to the requested model. In particular, if the ML model is to be used on the terminal (UE) side, the request may include information about the UE type, UE vendor, and UE location.

[0042] If the requested ML model is stored in the MTF 203, then at 209 the MTF 203 provides the requested model to the consumer NF 201.

[0043] If the requested ML model is not stored in the MTF 203 but is stored in the MRF 204, then at 210 the MTF 203 informs the consumer NF 201 of the identity of the MRF 204, at 211 the consumer NF 201 requests the ML model from the MRF, and at 212 the MRF provides the requested ML model.

[0044] Both the terminal and the network should be capable of distributing ML models from the network to the terminal, i.e., the terminal should be able to request and accept models from the network. The network should also be able to train and distribute models to the terminal. The mobile terminal should be able to know its capabilities and characteristics regarding ML models, e.g., UE model provision capability (i.e., the UE has the capability to obtain an AIML model from the network (e.g., a 5G communication system (5GS))) ● A list of supported ML models (e.g., ML models for CSI) User device information (e.g., UE type, vendor, hardware, etc.) The network may be notified about

[0045] The mobile terminal may provide information about its capabilities and characteristics, for example, as part of a UE capability transfer (i.e., with a UECapabilityInformation message to the network in response to a UECapabilityEnquiry message from the network) or in its registration request (sent to the RAN to register with the network).

[0046] The network may perform network ML model provisioning capability advertisement, i.e., may inform the UE of its capabilities for ML model provisioning, for example through system information in a system information block (SIB), for example through a SystemInformation message (for example in response to a SystemInformationRequest). The network may provide, for example, the following system information: Network model provision capability (i.e., the network has the capability to provide ML models to the UE) A list of supported ML models (e.g., ML models for CSI)

[0047] In the following, an embodiment will be described in which an ML model is provided to a mobile terminal from the network side via user plane communication.

[0048] FIG. 3 shows a flow diagram 300 illustrating an embodiment in which an ML model is provided to a mobile terminal 301 from a UPF 302 via a PDU session 303 between the mobile terminal 301 and the UPF 302 .

[0049] The UE 301 and the UPF 302 each include a respective ML agent (MLA) 304. The ML agent 304 is an entity similar to a Performance Measurement Functionality (PMF). The PMF may be used by a mobile terminal to obtain access performance measurements on the user plane. Similarly, the MLA 304 enables the exchange of data, i.e., ML models, between the UE and the UPF in the user plane.

[0050] Additionally, an NRF 305 and an MTF or MRF 306 are included in the flow.

[0051] At 307, the UE 301 establishes a PDU session 303 with the UPF 302.

[0052] At 308, the mobile terminal 301 requests an ML model from the UPF 302. The request may include ML model provision content according to 3GPP, e.g. List of analysis IDs: Identifies the analysis in which the ML model will be used Use case context: Provides the context for the use of analytics to select the most relevant ML model. ●ML model interoperability information ●ML model target period The accuracy level you are interested in ●Time when the model is needed It also contains ML model information.

[0053] Furthermore, according to various embodiments, the ML model information is extended with an indication that the requested model should support that the UE is the ultimate model consumer (i.e., the ML model may be used on the terminal side).

[0054] At 309, the UPF 302 performs MTF discovery and selection, followed by ML model acquisition at 310 (e.g., as described with reference to FIG. 2). ML model acquisition from the MTF or MRF 306 (depending on where the model is stored) is performed via the control plane.

[0055] At 311, the UPF 302 provides the mobile terminal 301 with the requested ML model (i.e., the trained model's specifications, e.g., in terms of neural network weights).

[0056] The communication between the mobile terminal 301 and the UPF 302 at 308 and 311 is carried out via a PDU session 302 and uses an MLA 304 that allows data exchange between the mobile terminal and the UPF in the user plane.

[0057] FIG. 4 shows a flow diagram 400 illustrating an embodiment in which an ML model is provided from an NEF 403 to a mobile terminal 401 via a PDU session 404 between the mobile terminal 401 and a UPF 402 and a UP tunnel 405 between the UPF 402 and a NEF 403.

[0058] The UE 401 may include an ML Agent (MLA) 406 that enables communication with the NEF 405 over the user plane.

[0059] Additionally, an NRF 407 and an MTF or MRF 408 are included in the flow.

[0060] At 409, the UE 301 establishes a PDU session 404 and an UP tunnel 405 with the UPF 302.

[0061] At 410, the mobile terminal 401 requests an ML model from the NEF 403. The request includes ML model information, which may be, for example, an ML model provisioning content according to 3GPP as described with reference to FIG.

[0062] Furthermore, according to various embodiments, the ML model information is extended with an indication that the requested model should support that the UE is the ultimate model consumer (i.e., the ML model may be used on the terminal side).

[0063] At 411, the NEF 403 performs MTF discovery and selection, followed by ML model acquisition at 412 (e.g., as described with reference to FIG. 2). ML model discovery and selection, as well as ML model acquisition from the MTF or MRF 406 (depending on where the model is stored), are performed via the control plane.

[0064] At 413, the NEF 403 provides the mobile terminal 401 with the requested ML model (i.e., the trained model's specifications, e.g., in terms of neural network weights).

[0065] The communication between the mobile terminal 401 and the NEF 402 at 410 may be based, for example, on a technique that allows the UE to directly call the NEF APIs via the user plane tunnel 405 (e.g., according to Resource owner-aware Northbound API Access (RNAA)).

[0066] FIG. 5 shows a flow diagram 500 illustrating an embodiment in which an ML model is provided from an MTF 508 or MRF 509 to a UPF 507 via a user plane tunnel 519 and from there to a mobile terminal 501 via a PDU session 520.

[0067] Furthermore, the RAN 502, AMF 503, SMF 504, PCF 505 and NRF 506 are included in the flow.

[0068] At 510, the mobile terminal 501 sends a request to the AMF 503 (via the RAN 502) to establish a PDU session 519. In response, at 511, the AMF 503 requests the SMF 504 to create a session management context. Then, at 512, the SMF 504 performs authentication and policy decisions for the PDU session 519 with the PCF 505.

[0069] At 513, the SMF 504 performs MTF discovery and selection with the NRF 506 (e.g., as described with reference to FIG. 2).

[0070] Then, at 514, the SMF 504 sends an N4 session establishment / modification request to the UPF 507, which indicates the discovered MTF 508 at 515.

[0071] At 515, a PDU session 519 is established (according to normal 3GPP procedures).

[0072] At 516, the UPF establishes a UP tunnel 520 with the MTF 508. This is performed, for example, according to an extension of the user plane tunnel establishment over the N4 interface according to 3GPP.

[0073] At 517, the mobile terminal 501 requests an ML model from the MTF 508 via the PDU session 519 and the UP tunnel 520, and at 518, the MTF 508 provides the requested ML model to the mobile terminal 501 via the PDU session 519 and the UP tunnel 520.

[0074] If the model is stored in the MRF 509 instead of the MTF 508, the MRF 509 replaces the MTF 508 in the above procedure.

[0075] FIG. 6 illustrates an AF-based approach for providing a mobile terminal 601 with an ML model.

[0076] Included in the flow are an ML Management AF 602, an NRF 603, a Model Training AF 604, a Proprietary ML AF 605, an NEF 606, an External ML Repository 607, an MTF 608, and an MRF 609 (or at least some of these, depending on where the ML models are trained and stored).

[0077] The mobile terminal 601 includes an MLA 610 and an AnLF 611. The MLA 610 is an application in the mobile terminal 601 that communicates with a (proprietary) ML AF 605. The ML 605 requests an ML model by calling the API of the NEF 606.

[0078] The ML management AF 602 acts as a gateway to obtain ML models from ML model providers (i.e., for example, the model training AF 604, external repository 607, OAM or MTF 608, or MRF 609) and provide them to the mobile terminal 601. The model training AF 604 can train models based on available data (sample datasets, simulations, etc.). Sources of ML models (e.g., the MTF 608) may be discovered using the NRF 603.

[0079] FIG. 7 shows a flow diagram 700 illustrating an example procedure for an AF-based approach such as that shown in FIG.

[0080] Included in the flow are a mobile terminal 701, a RAN 702, an AMF 703, an SMF 704, a PCF 705, a UPF 706, a NEF 707, an ML Management AF (MMF) 708, an ML Training AF (MTAF) 709, an MTF 410, an MRF 411, and an external ML model repository 412 (or at least some of these, depending on where the ML models are trained and stored).

[0081] At 713, the mobile terminal 701 sends a PDU session modify NAS message to the AMF 703 to initiate a QoS flow in the PDU session (assumed to have been previously established).

[0082] At 714, the AMF 703 responds by sending a request to the SMF 704 to update the SM context for the PDU session.

[0083] Alternatively, at 715, the ML management AF 708 may initiate the establishment of the QoS flows (in the PDU session) by sending a corresponding request to the NEF 707. In this case, at 716, the NEF 709 sends a policy / permission creation message to the PCF 705 to trigger the establishment of the QoS flows.

[0084] At 717, the QoS flow is established.

[0085] At 718, the mobile terminal 701 requests an ML model from the ML management AF 708. The request includes ML model information, which may be, for example, an ML model provisioning content according to 3GPP as described above with reference to Figure 3. Furthermore, according to various embodiments, the ML model information is extended with an indication that the requested model should support that the UE is the ultimate model consumer (i.e., that the ML model may be used on the terminal side).

[0086] At 719, the ML Manager AF 708 identifies the storage location of the requested ML. Model.

[0087] The ML management AF 708 performs the following steps according to the storage location of the identified requested ML model: ● From model training AF709 in 720, ● Directly and indirectly from MTF 710 or MRF 711 in 721, or ●From external repository 712 in 722 Get the requested ML model.

[0088] At 723, the ML management AF 708 provides the requested ML model to the mobile terminal 701.

[0089] It should be noted that all the request messages (to request an ML model) and response messages (to provide an ML model) in the above examples can be implemented as pairs of subscribe and notify messages. This enables network-triggered ML model delivery: a mobile terminal subscribes to a particular model, and the network then provides customized / updated versions of the model, taking into account, for example, operator policies, ML model performance monitoring, mobile terminal mobility / handover, etc.

[0090] It should further be noted that the various approaches for serving ML models are not exclusive, and multiple approaches may be implemented and used according to operator policies, mobile terminal configuration, etc.

[0091] In summary, in accordance with various embodiments, a method is provided as shown in FIGS.

[0092] FIG. 8 is a flow diagram 800 illustrating a method for providing machine learning models to a mobile terminal of a mobile communication system according to an embodiment.

[0093] In 801, a mobile wireless communication network of a communication system a determines that a machine learning model to be used by a mobile terminal for a machine learning task (e.g., inference, e.g., a prediction task, e.g., a prediction task for controlling radio access network operation or data compression or encoding (e.g., as an autoencoder)) should be distributed to the mobile terminal. For example, this is determined in response to receiving a request for a machine learning model from the mobile terminal. The request may also be a subscription (i.e., a subscribe message) to a corresponding model providing service.

[0094] At 802, the mobile wireless communication network provides the machine learning model to the mobile terminal via user plane communications (in response to determining that the machine learning model should be distributed to the mobile terminal).

[0095] According to various embodiments, there is provided a communications network configured to perform the method described with reference to FIG.

[0096] The communication network may correspond, for example, to the RAN 101, the transport network 102 and / or the core network 103 of the communication system of FIG.

[0097] FIG. 9 is a flow diagram 900 illustrating a method for obtaining a machine learning model from a mobile communication system according to an embodiment.

[0098] At 901, a mobile terminal transmits a request for a machine learning model to be used by the mobile terminal for an ML task to a mobile wireless communication network of a communication system.

[0099] At 902, the mobile terminal receives the requested machine learning model in response to the request via user plane communications.

[0100] The request may include an indication that a machine learning model is requested for a machine learning task at the mobile terminal.

[0101] According to various embodiments, there is provided a mobile terminal configured to perform the method described with reference to FIG.

[0102] The mobile terminal corresponds, for example, to one of the UEs 104 of the communication system of FIG.

[0103] In other words, according to various embodiments, the ML model is provided to the mobile terminal via the user plane, which allows for example to use the ML model on the terminal side, e.g., on the air interface, for communication optimization.

[0104] As noted above, many variations are possible that provide a flexible set of techniques.

[0105] Providing a machine learning model via user plane communication may be understood as transmitting the specifications of the machine learning model, i.e. the values ​​of the trained parameters of the machine learning model (such as neural network weights if the machine learning model is a neural network), via user plane communication (i.e. over the air interface) from the network side (i.e. the mobile radio communication network) to the mobile terminal, i.e. using communication channels, interfaces and / or reference points belonging to the user plane (of the mobile communication system).

[0106] The communication networks and mobile terminals may include and / or be implemented by data processing components such as one or more processors, memories, interfaces, receivers, transmitters, antennas, etc., and may be implemented, for example, by one or more circuits. A "circuit" may be understood as any kind of logic implementation entity, and may be a dedicated circuit or a processor that executes software stored in memory, firmware, or any combination thereof. Thus, a "circuit" may be a hardwired logic circuit or a programmable processor, for example, a programmable logic circuit such as a microprocessor. A "circuit" may also be a processor that executes software, for example, any kind of computer program. Any other kind of implementation of each of the above functions may also be understood as a "circuit."

[0107] While particular embodiments have been described, it should be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the embodiments of the present disclosure as defined by the appended claims. The scope is accordingly indicated by the appended claims, and all changes that come within the meaning and range of equivalents of the claims are therefore intended to be embraced.

Claims

1. 1. A method for providing a machine learning model to a mobile terminal of a mobile communication system, comprising: determining, by a mobile radio communication network of the communication system, in response to receiving a request for the machine learning model from the mobile terminal, including receiving a request at a user plane function of the radio communication network, by the mobile radio communication network, that a machine learning model to be used by the mobile terminal for a machine learning task should be delivered to the mobile terminal; the user plane function discovering a storage location of the machine learning model, retrieving the machine learning model from the discovered storage location, and providing the machine learning model to the mobile terminal; or forwarding the request to a network publishing function of the wireless communication network via a user plane tunnel, wherein the network publishing function discovers a storage location of the machine learning model, retrieves the machine learning model from the discovered storage location, and provides the machine learning model to the mobile terminal via the user plane function; providing the machine learning model to the mobile terminal via user plane communications in response to the request; A method comprising:

2. 2. The method of claim 1, comprising determining that the machine learning model should be distributed to the mobile terminal based on one or more of temporal and / or spatial variations in operating conditions, performance degradation of a machine learning model, a configuration of the mobile terminal, and an operator policy of a mobile communications network.

3. The method of claim 1 , comprising receiving the request by a user plane function of the communications network.

4. The method of claim 1 , comprising providing the machine learning model to the mobile terminal via a packet data unit session.

5. 2. The method of claim 1, comprising receiving the request over a packet data unit session and providing the machine learning model to the mobile terminal over the packet data unit session or another packet data unit session.

6. The method of claim 1 , comprising receiving the request by a core network component of the wireless communication network via a user plane tunnel.

7. A communications network configured to carry out a method according to any one of claims 1 to 6.

Images