Methods for a UE and a network component, UE and network component

Abstract

Some aspects of the present disclosure relate to a method for user equipment, UE, comprising obtaining (320) from a network component of a cellular mobile communication system, information on an inference operation to be performed by the UE, wherein the information on the inference operation to be performed comprises at least one of information on a threshold used to trigger the inference operation or information on a timing window for performing the inference operation, performing (330) the inference operation according to the information on the inference operation to be performed, and providing (340) a result of the inference operation to the network component.

Description

[0001] 202407065

[0002] 1

[0003] TITLE

[0004] Methods for a UE and a Network Component, UE and Network Component

[0005] TECHNNICAL FIELD

[0006] The 3rd Generation Partnership Project (3GPP) is a standards organization which develops protocols for mobile telephony, and is known for the development and maintenance of various standards, including second generation (2G), third generation (3G), fourth generation (4G), Long Term Evolution (LTE), and fifth generation (5G) standards.

[0007] The 5G network has been designed as a Service Based Architecture (SBA), that is, a system architecture in which system functionality is achieved by a set of network functions that provide services to other authorized network functions, thereby enabling them to access those services. The 5G network may comprise a plurality of base stations (e.g., Next Generation NodeB (gNB), etc.) that serve multiple cells across a particular area.

[0008] The present disclosure relates to a method for a User Equipment (UE), a UE performing said method, a method for a gNodeB, and a gNodeB performing said method.

[0009] BACKGROUND

[0010] User Equipment (UE) can leverage AI / ML models to process network measurements locally before providing enhanced information to the gNodeB. This approach enables intelligent compression, feature extraction, and prediction of radio conditions directly at the device level. By processing raw measurements such as Reference Signal Received Power (RSRP), Signal-to-lnterference-plus-Noise Ratio (SINR), and Channel State Information (CSI) through trained ML models, UEs can generate compact, higher-quality representations that reduce reporting overhead while maintaining or improving accuracy for network optimization tasks. 202407065

[0011] 2

[0012] The processed outputs from UE-side AI / ML models can support various network functions at the gNodeB, including beam management, mobility optimization, and resource allocation. For instance, a UE might use an ML model to predict future channel quality or beam preferences based on historical measurements and movement patterns, then report these predictions to the gNodeB. This enables more proactive network decisions and reduces the latency associated with reactive measurement reporting.

[0013] Standardization efforts in 3GPP have been addressing the framework for UE-side AI / ML model usage, including model delivery mechanisms, performance monitoring, and feedback procedures between UE and network. Key considerations include ensuring model compatibility across different UE implementations, managing computational complexity and power consumption at the device, and establishing procedures for model updates and lifecycle management.

[0014] As outlined above, the results provided by such models can improve the decisionmaking on part of the gNodeB in specific cases. However, performing inference using such models can tax the batteries of the UEs while in many cases, particularly if the UE is not moving, the benefit to the quality of decision-making by the gNodeB is negligible.

[0015] There may be a desire to provide an improved concept for performing UE-side inference operations.

[0016] SUMMARY

[0017] This desire is addressed by the subject matter of the independent claims.

[0018] Various examples of the present disclosure are based on the finding that inference operations being performed by UE, while providing a potential benefit to the decision-making of a gNodeB, can tax the UE’s battery, so inference should be performed in specific scenarios (only). Moreover, UEs usually have limited time to 202407065

[0019] 3 make decisions, so they oftentimes cannot wait until a UE has reported the results of an inference operation before making these decisions. The present disclosure provides an improved concept for controlling the performance of inference operations performed by the UE. In particular, the UE may be provided with a trigger that helps the UE decide in which scenarios / circumstances the inference operation is to be performed. Moreover, the UE may be provided with a time limit, so it does not perform inference operations that are ultimately useless to the gNodeB. This way, the efficiency of the UE performing inference operations can be improved.

[0020] Some aspects of the present disclosure relate to a method for a user equipment (UE). For example, the method may be performed by the UE. The method comprises obtaining, from a network component of a cellular mobile communication system (e.g., from a gNodeB), information on an inference operation to be performed by the UE. The information on the inference operation to be performed comprises at least one of information on a threshold used to trigger the inference operation or information on a timing window for performing the inference operation. The method comprises performing the inference operation according to the obtained information on the inference operation to be performed. The method comprises providing a result of the inference operation to the network component. By receiving this information from the network component, the UE can perform the inference operation in cases that are actually relevant to the gNodeB, and conserve battery power and computational resources otherwise.

[0021] The method may further comprise obtaining updated information on the inference operation to be performed from the network component, after providing the result of the inference operation to the network component. This updated information is then used for a subsequent instance (or subsequent instances) of performing the inference operation. This adaptive mechanism ensures that the UE’s inference operations are continuously improved based on network conditions and evolving requirements.

[0022] In some examples, the threshold used to trigger the inference operation may be related to a relative movement or distance between the UE and a serving gNB 202407065

[0023] 4 serving the UE. By utilizing movement or distance as a trigger, the inference operation can be proactively initiated when the UE’s position changes, which oftentimes leads to scenarios in which the results of the inference operation are useful for the gNodeB’s decision making.

[0024] For example, the threshold may be set to trigger the inference operation when the UE is moving away from the serving gNB. This allows the system to anticipate potential handover scenarios and prepare accordingly, improving the user experience.

[0025] In practice, it is expected that the AI / ML (Artificial Intelligence / Machine Learning) models used to perform the inference operation are vendor-specific (i.e., specific to a vendor of the UE), which leads to a large variety of AI / ML models being available at the different UEs. Moreover, there is a large variety of computational capabilities across different handsets. To provide information on the inference operation to be performed that is targeted at the AI / ML models available at the UE and the UE’s computational capabilities, the UE may first provide information on its AI / ML-related capabilities to the gNodeB. In other words, the method may comprise, prior to obtaining the information on the inference operation to be performed, providing information on a capability of the UE with respect to performing inference operations to the network component. This information allows the network component to tailor the inference tasks to the UE’s specific capabilities.

[0026] The information on the capability of the UE with respect to performing inference operations may comprise at least one of information on machine learning models available at the UE, or information of a computational capability of the UE. Providing details about available machine learning models and computational capacity allows the network component to select the most appropriate inference tasks for the UE.

[0027] Various communication mechanisms may be employed to communicate the respective information between the UE and the network component (e.g., gNodeB). For example, the UE may obtain the information on the inference operation to be 202407065

[0028] 5 performed via at least one of L1 signaling, L2 signaling, L3 signaling, or Radio Resource Control (RRC) configuration or reconfiguration.

[0029] In scenarios requiring more detailed information, the method may comprise providing information on machine learning model parameters associated with the inference operation to the network component. For example, the machine learning model parameters may include the weights of a neural network, or the inference-time parameters being used to perform inference with the machine learning model (such as a confidence threshold for interpreting the output, etc.). Providing model parameters allows the network component to better understand the inference process and improve its performance.

[0030] To further enhance adaptation, the method may comprise obtaining updated machine learning model parameters from the network component after providing the result of the inference operation to the network component. These updated parameters are then used for a subsequent instance of performing the inference operation, providing continuous improvement of the inference process.

[0031] As outlined above, the proposed concept is particularly useful in cases in which the UE performs a measurement, processes the result of the measurement in the inference operation, and provides the result to the network component, so the network component can make decisions (e.g., with respect to beam management, handover, modulation etc.) to improve or maintain the performance of the UE’s radio communication. For example, the inference operation may be related to performing a radio resource measurement or to performing a channel state information measurement. Results of such inference operation can help the network component improve the UE’s radio communication.

[0032] Various examples of the present disclosure relate to a UE comprising a wireless transceiver, a processor coupled with a memory in which computer program instructions are stored, said instructions being configured to implement the above method. This enables the UE to perform the described inference operations in coordination with the network component. 202407065

[0033] 6

[0034] Some aspects of the present disclosure relate to a method for a network component of a cellular mobile communication system, such as a gNodeB. For example, the method may be performed by the network component. The method comprises providing, to a user equipment (UE), the information on the inference operation to be performed by the UE. The information on the inference operation to be performed comprises at least one of the information on a threshold used to trigger the inference operation, or the information on the timing window for performing the inference operation. The method comprises obtaining a result of the inference operation from the UE. For example, the network component may use the result for various options, such as beam management, handover decision-making, or selection of communication parameters to be used for communicating with the UE.

[0035] To dynamically adjust to changing conditions, the method may comprise determining, after obtaining the result of the inference operation from the UE, updated information on the inference operation to be performed. The method may comprise providing the updated information on the inference operation to the UE. This closed-loop feedback mechanism ensures that the inference operations can be improved for the current network conditions and / or for the current situation of the UE.

[0036] The updated information on the inference operation may comprise at least one of an updated threshold or an updated timing window. Utilizing updated thresholds or timing windows allows the network component to fine-tune the inference operations based on changes in the UEs behavior, e.g., movement relative to the network component.

[0037] To account for the variability of models available across different UEs as well as their different computational capabilities, the method may comprise obtaining information on a capability of the UE with respect to performing inference operations from the UE, and providing the information on the inference operation to be performed by the UE based on the information on the capability of the UE with respect to performing inference operations. This ensures that the inference tasks are assigned to the UEs in a manner that is tailored to the UE’s capabilities. 202407065

[0038] 7

[0039] In some examples, the network component (e.g., gNodeB) may attempt to improve the inference operation(s) performed by the UE by improving the machine learning model parameters being used by the UE. For example, the method may comprise obtaining information on machine learning model parameters associated with the inference operation from the UE, and providing, after obtaining the result of the inference operation from the UE, updated machine learning model parameters from the network component. The updated machine learning model parameters are to be used for a subsequent instance of performing the inference operation. This closed- loop feedback mechanism allows for continuous improvement of the inference process.

[0040] Various examples of the present disclosure relate to a gNB comprising a wireless transceiver, a processor coupled with a memory in which computer program instructions are stored, said instructions being configured to implement the above method.

[0041] Some examples relate to a computer program having a program code for performing at least one of the above methods, when the computer program is executed on a computer, a processor, or a programmable hardware component.

[0042] Some examples relate to a non-transitory, computer-readable medium comprising a program code that, when the program code is executed on a processor, a computer, or a programmable hardware component, causes the processor, computer, or programmable hardware component to perform at least one of the above methods.

[0043] BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figs. 1a to 1c show different training paradigms for training models used in cellular mobile communication systems;

[0045] Fig. 2 shows an illustration of a challenging scenario; 202407065

[0046] 8

[0047] Fig. 3a shows a block diagram of a UE, a gNodeB, and a cellular mobile communication system comprising the UE and the gNodeB;

[0048] Fig. 3b shows a flowchart of a method for the UE;

[0049] Fig. 4 shows a flowchart of a method for a network component, such as a gNodeB;

[0050] Fig. 5a shows a schematic diagram of a UE moving relative to a gNodeB;

[0051] Fig. 5b shows a sequence diagram of a gNodeB configuring an inference operation at UE;

[0052] Fig. 6 shows a schematic diagram of a UE moving relative to a gNodeB, with two different thresholds being configured;

[0053] Fig. 7a shows a flowchart of one or more operations performed by a gNodeB; and

[0054] Fig. 7b shows a flowchart of one or more operations performed by UE.

[0055] DETAILED DESCRIPTION

[0056] The detailed description set forth below, with reference to annexed drawings, is intended as a description of various configurations and is not intended to represent the only configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. In particular, although terminology from 3GPP 5G NR may be used in this disclosure to exemplify embodiments herein, this should not be seen as limiting the scope of the invention. 202407065

[0057] 9

[0058] Some of the embodiments contemplated herein will now be described more fully with reference to the accompanying drawings. Other embodiments, however, are contained within the scope of the subject matter disclosed herein. The disclosed subject matter should not be construed as limited to only the embodiments set forth herein; rather, these embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art.

[0059] Generally, all terms used herein are to be interpreted according to their ordinary meaning in the relevant technical field, unless a different meaning is clearly given and / or is implied from the context in which it is used. All references to a / an / the element, apparatus, component, means, step, etc. are to be interpreted openly as referring to at least one instance of the element, apparatus, component, means, step, etc., unless explicitly stated otherwise. The steps of any methods disclosed herein do not have to be performed in the exact order disclosed, unless a step is explicitly described as following or preceding another step, or it is implicit that a step must follow or precede another step. Any feature of any of the embodiments disclosed herein may be applied to any other embodiment, wherever appropriate. Likewise, any advantage of any of the embodiments may apply to any other embodiments, and vice versa. Other objectives, features and advantages of the enclosed embodiments will be apparent from the following description.

[0060] In some embodiments, a more general term “network node” may be used and may correspond to any type of radio network node or any network node, which communicates with a UE (directly or via another node) and / or with another network node. Examples of network nodes include NodeB, MeNB, ENB, a network node belonging to MCG or SCG, base station (BS), multi-standard radio (MSR) radio node such as MSR BS, eNodeB, gNodeB, network controller, radio network controller (RNC), base station controller (BSC), relay, donor node controlling relay, base transceiver station (BTS), access point (AP), transmission points, transmission nodes, RRU, RRH, nodes in distributed antenna system (DAS), core network node (e.g. Mobile Switching Center (MSC), Mobility Management Entity (MME), etc.), Operations & Maintenance (O&M), Operations Support System (OSS), Self Optimized Network (SON), positioning node (e.g. Evolved- Serving Mobile Location 202407065

[0061] 10

[0062] Centre (E-SMLC)), Minimization of Drive Tests (MDT), test equipment (physical node or software), etc.

[0063] In some embodiments, the non-limiting term user equipment (UE) or wireless device may be used and may refer to any type of wireless device communicating with a network node and / or with another UE in a cellular or mobile communication system. Examples of UE are target device, device to device (D2D) UE, machine type UE or UE capable of machine to machine (M2M) communication, PDA, PAD, Tablet, mobile terminals, smartphone, laptop embedded equipped (LEE), laptop mounted equipment (LME), USB dongles, UE category Ml, UE category M2, ProSe UE, V2V UE, V2X UE, etc.

[0064] Additionally, terminologies such as base station / gNodeB and UE should be considered non-limiting and do in particular not imply a certain hierarchical relation between the two; in general, “gNodeB” could be considered as device 1 and “UE” could be considered as device 2 and these two devices communicate with each other over some radio channel. And in the following the transmitter or receiver could be either gNodeB (gNB), or UE.

[0065] As will be appreciated by one skilled in the art, aspects of the embodiments may be embodied as a system, apparatus, method, or program product. Accordingly, embodiments may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects.

[0066] For example, the disclosed embodiments may be implemented as a hardware circuit comprising custom very-large-scale integration (“VLSI”) circuits or gate arrays, off- the-shelf semiconductors such as logic chips, transistors, or other discrete components. The disclosed embodiments may also be implemented in programmable hardware devices such as field programmable gate arrays, programmable array logic, programmable logic devices, or the like. As another example, the disclosed embodiments may include one or more physical or logical 202407065

[0067] 11 blocks of executable code which may, for instance, be organized as an object, procedure, or function.

[0068] Furthermore, embodiments may take the form of a program product embodied in one or more computer readable storage devices storing machine readable code, computer readable code, and / or program code, referred hereafter as code. The storage devices may be tangible, non- transitory, and / or non-transmission. The storage devices may not embody signals. In a certain embodiment, the storage devices only employ signals for accessing code.

[0069] Any combination of one or more computer readable media may be utilized. The computer readable medium may be a computer readable storage medium. The computer readable storage medium may be a storage device storing the code. The storage device may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, holographic, micromechanical, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing.

[0070] More specific examples (a non-exhaustive list) of the storage device would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random-access memory (“RAM”), a read-only memory (“ROM”), an erasable programmable read-only memory (“EPROM” or Flash memory), a portable compact disc read-only memory (“CD-ROM”), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0071] Code for carrying out operations for embodiments may be any number of lines and may be written in any combination of one or more programming languages including an object- oriented programming language such as Python, Ruby, Java, Smalltalk, C++, or the like, and conventional procedural programming languages, such as the “C” programming language, or the like, and / or machine languages such as assembly languages. The code may execute entirely on the user’s computer, partly on the 202407065

[0072] 12 user’s computer, as a stand-alone software package, partly on the user’s computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user’s computer through any type of network, including a local area network (“LAN”), wireless LAN (“WLAN”), or a wide area network (“WAN”), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider (“ISP”)).

[0073] Furthermore, the described features, structures, or characteristics of the embodiments may be combined in any suitable manner. In the following description, numerous specific details are provided, such as examples of programming, software modules, user selections, network transactions, database queries, database structures, hardware modules, hardware circuits, hardware chips, etc., to provide a thorough understanding of embodiments. One skilled in the relevant art will recognize, however, that embodiments may be practiced without one or more of the specific details, or with other methods, components, materials, and so forth. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of an embodiment. Reference throughout this specification to “one embodiment,” “an embodiment,” or similar language means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, appearances of the phrases “in one embodiment,” “in an embodiment,” and similar language throughout this specification may, but do not necessarily, all refer to the same embodiment, but mean “one or more but not all embodiments” unless expressly specified otherwise. The terms “including,” “comprising,” “having,” and variations thereof mean “including but not limited to,” unless expressly specified otherwise. An enumerated listing of items does not imply that any or all of the items are mutually exclusive, unless expressly specified otherwise. The terms “a,” “an,” and “the” also refer to “one or more” unless expressly specified otherwise.

[0074] Aspects of the embodiments are described below with reference to schematic flowchart diagrams and / or schematic block diagrams of methods, apparatuses, systems, and program products according to embodiments. It will be understood that 202407065

[0075] 13 each block of the schematic flowchart diagrams and / or schematic block diagrams, and combinations of blocks in the schematic flowchart diagrams and / or schematic block diagrams, can be implemented by code. This code may be provided to a processor of a general-purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart diagrams and / or block diagrams.

[0076] The code may also be stored in a storage device that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the storage device produce an article of manufacture, including instructions which implement the function / act specified in the flowchart diagrams and / or block diagrams.

[0077] The code may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus, or other devices to produce a computer implemented process such that the code which executes on the computer or other programmable apparatus provides processes for implementing the functions / acts specified in the flowchart diagrams and / or block diagrams.

[0078] The flowchart diagrams and / or block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of apparatuses, systems, methods, and program products according to various embodiments. In this regard, each block in the flowchart diagrams and / or block diagrams may represent a module, segment, or portion of code, which includes one or more executable instructions of the code that implement the specified logical function(s).

[0079] It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the Figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the 202407065

[0080] 14 functionality involved. Other steps and methods may be conceived that are equivalent in function, logic, or effect to one or more blocks, or portions thereof, of the illustrated Figures.

[0081] Although various arrow types and line types may be employed in the flowchart and / or block diagrams, they are understood not to limit the scope of the corresponding embodiments. Indeed, some arrows or other connectors may be used to indicate only the logical flow of the depicted embodiment. For instance, an arrow may indicate a waiting or monitoring period of unspecified duration between enumerated steps of the depicted embodiment. It will also be noted that each block of the block diagrams and / or flowchart diagrams, and combinations of blocks in the block diagrams and / or flowchart diagrams, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and code.

[0082] The description of elements in each figure may refer to elements of preceding figures. Like numbers refer to like elements in all figures, including alternate embodiments of like elements.

[0083] The detailed description set forth below, with reference to annexed drawings, is intended to describe various configurations and is not intended to represent the only configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. Although terminology from 3GPP 5G NR may be used in this disclosure to exemplify embodiments herein, this should not be seen as limiting the scope of the invention.

[0084] Various examples of the present disclosure relate to a method of optimally configuring UE for model inference. Within the 3GPP standardization process, various types of models have been discussed - UE side, gNB side and two-sided models. For inference of UE-side models, to ensure consistency between training and inference regarding network-side additional conditions (if identified), the following 202407065

[0085] 15 options can be taken as potential approaches (when feasible and necessary). For example, model identification may be used to achieve alignment on the network-side additional condition between NW-side and UE-side. Model training may be performed at the network, and the model may be transferred to the UE, where it has been trained under the additional condition. Information and / or indication on network-side additional conditions may be provided to UE. Consistency may be assisted by monitoring (by UE and / or network) the performance of UE-side candidate models / functionalities to select a model / functionality. Other approaches are not precluded. For example, different approaches may achieve the same function.

[0086] In various examples, vendor-specific UE models may be the first to market, as network side model upgrades will take time. Therefore, the UE side models may be developed in a way that they can help improve the overall network and enhance the user experience. It may be important for the network to know the UE model details, so that the network can better configure the UE side model to perform improved optimal inference. For example, the network may be aware of the UE’s requirements to configure the measurement resources needed for RRM (Radio Resource Measurement) measurement prediction in an improved or optimal manner.

[0087] Fig. 1 a to 1 c show the different training paradigms for training models used in cellular mobile communication systems. In Fig. 1 a, a first type of training paradigm is shown, denoted “joint training at one side”. A training entity at the network side performs training, and a training entity at the UE side performs training. In Fig. 1 b, a second type of training paradigm is shown, denoted “joint training at two sides”. A training entity for a UE-side model provides forward activation information to a training entity for the network-side model, and the training entity for the network-side model provides backward gradient information for the training entity for the UE-side model. In Fig. 1 c, a third type of training paradigm is shown, denoted “separate training at two sides”. On top, at 1 ., the training entity for the UE-side model trains the UE-side model. At 2., the training entity for the UE-side model shares the training dataset with the training entity for the network-side model. At 3., the training entity for the networkside model trains the network-side model. On the bottom, at 1 ., the training entity for the network-side model trains the network-side model. At 2., the training entity for the 202407065

[0088] 16 network-side model shares the training dataset with the training entity for the UE-side model. At 3., the training entity for the UE-side model trains the UE-side model.

[0089] For the case of UE side models and two-sided models, it may be considered important that the network is aware of the UE model configuration before configuring the UE for model inference. Various examples of the present disclosure relate to How to configure a UE side model for improved or optimal model inference.

[0090] Fig. 2 shows an illustration of the challenge. A UE comprises a UE-sided model (e.g., an Artificial Neural Network). If the gNB performs model configuration without model knowledge, this may lead to a configuration of model parameters that may not be optimal for the UE-sided model.

[0091] According to the proposed concept, the gNB configures an AI / ML threshold and an AI / ML inference timer, e.g., upon receiving a UE capability indication that comprises information on the AI / ML functionality supported by the UE.

[0092] Fig. 3a shows a block diagram of a UE 30, a gNodeB 40 (as example of a network component of a cellular mobile communication system), and the cellular mobile communication system comprising the UE 30 and the gNodeB 40.

[0093] The UE 30 comprises a wireless transceiver 32, a processor 34, and a memory 36. The processor 34 is coupled with the wireless transceiver 32 and with the memory 36. For example, the UE 30 may use the wireless transceiver to communicate with terrestrial and non-terrestrial radio access network components, such as the nonterrestrial gNodeB 40 of the cellular mobile communication system. The functionality of the UE 30 may be provided by the processor 34, which may be configured to execute computer program instructions stored in the memory 36 to provide its functionality. For example, the processor 34 may use the wireless transceiver 32 to communicate in the cellular mobile communication system. The UE, e.g., the processor 34, is configured to perform the method shown in Fig. 3b. 202407065

[0094] 17

[0095] Similarly, the non-terrestrial serving gNodeB 40 comprises a wireless transceiver 42, a processor 44, and a memory 46. The processor 44 is coupled with the wireless transceiver 42 and with the memory 46. For example, the non-terrestrial gNodeB 40 may use the wireless transceiver to communicate with UEs and other entities of the cellular mobile communication system, such as a feeder station. The functionality of the non-terrestrial gNodeB may be provided by the processor 44, which may be configured to execute computer program instructions stored in the memory 46 to provide its functionality. For example, the processor 44 may use the wireless transceiver 42 to communicate in the cellular mobile communication system. The non-terrestrial gNodeB, e.g., the processor 44, is configured to perform the method of Fig. 4.

[0096] Fig. 3b shows a flowchart of a method for the UE 30. The method comprises obtaining 320 from a network component of a cellular mobile communication system (e.g., from the gNodeB 40), information on an inference operation to be performed by the UE 30. The information on the inference operation to be performed comprises at least one of information on a threshold used to trigger the inference operation or information on a timing window for performing the inference operation. The method comprises performing 330 the inference operation according to the information on the inference operation to be performed. The method comprises providing 340 a result of the inference operation to the network component.

[0097] Fig. 4 shows a flowchart of a method for a network component, such as the gNodeB 40. The method comprises providing 420 to the UE 30, the information on the inference operation to be performed by the UE. The method comprises obtaining 430 the result of the inference operation from the UE.

[0098] The proposed concept is based on the finding that the act of performing an inference operation is computationally expensive for UEs. On the other hand, in many use cases, the result of such an inference operation can yield highly valuable information that helps the network improve the wireless operation of the UE (and potentially other UEs), at least if the result is provided within a timeframe in which the network can still incorporate the information in its decision making. For example, the inference 202407065

[0099] 18 operation may be used as part of radio resource management (RRM) or channel state information (CSI) measurement, e.g., to project the current measurement into the future. In other words, the inference operation may be related to performing a radio resource measurement or to performing a channel state information measurement. This information is highly valuable for the network. Therefore, a balance may be struck that aims to provide the benefit without overly draining the UE’s battery. This can be achieved by specifying conditions for performing the inference operation - (a) in which cases the inference operation is to be performed, and (b) how long the result of the inference operation is useful to the network, so the inference operation only needs to be performed if the result can be provided within the specified time window.

[0100] The first condition (a) generally depends on the scenario the UE finds itself in. In particular, when the UE is moving relative to the gNodeB, the radio condition experienced by the UE changes, up to the point where the UE might be handed over to a different cell. Therefore, the threshold used to trigger the inference operation may be related to the movement of the UE relative to the gNodeB. In other words, the threshold used to trigger the inference operation may be related to a relative movement or distance between the UE and a serving gNB that serves the UE. In this context, the movement may be detected, by the UE, by tracking the position of the UE, e.g., using triangulation of radio signals from different radio sources, or using a satellite positioning system. Additionally, or alternatively, the movement may be detected, by the UE, using an accelerometer or gyroscope of the UE. Additionally, or alternatively, the movement may be detected, by the UE, based on a signal strength or signal level of the radio signals emitted by the serving gNB. In general, any movement that surpasses a threshold (in terms of distance, speed, change in signal strength / level etc.) may trigger the inference operation being performed. In particular, the threshold may be set to trigger the inference operation when the UE is moving away from the serving gNB and the threshold is surpassed.

[0101] The second condition (b) helps to avoid situations in which the UE performs the inference operation, but the inference operation is disregarded by the network, because it is no longer relevant. To avoid such scenarios, the timing window for 202407065

[0102] 19 performing the inference operation may be defined. For example, the timing window may start at the point in time when the first condition (a) is met, or from the time the information on the inference operation to be performed. Alternatively, the information on the timing window may define a fixed point in time at which the timing window ends (or fixed points in time defining the timing window). If the UE determines that it cannot perform the inference operation within the timing window, it may refrain from performing the inference operation at all to conserve power.

[0103] In the context of the present disclosure, inference is the process of using a trained machine learning model to make predictions or decisions based on new input data. The result of inference is the output of the model, which may be a prediction, classification, or estimated value based on the patterns learned during training. The inputs to inference are the new data samples or features that are fed into the trained model to generate outputs. For example, in CSI (Channel State Information) prediction for 5G NR, inference uses a trained neural network model where the inputs may be historical CSI measurements (such as previous CQI (Channel Quality Indicator), PMI (Precoding Matrix Indicator), and Rl (Rank Indicator) reports from a UE), the inference process applies the trained model to predict future channel conditions, and the result is the predicted CSI that helps the gNB perform proactive Radio Resource Management decisions like beam selection or modulation and coding scheme adaptation.

[0104] After the gNodeB 40 has received the result of the inference, it may use it for radio resource management and similar tasks. In addition, it may use the result, as well as metadata on the result (i.e. , how fast the UE reached the threshold), to adapt the parameters transmitted to the UEs. For example, the method of Fig. 4 may comprise determining 440, after obtaining the result of the inference operation from the UE, updated information on the inference operation to be performed, and providing 450, the updated information on the inference operation to be performed to the UE. In particular, the updated information on the inference operation may comprise at least one of an updated threshold or an updated timing window. For example, the timing window may be shortened if the UE is moving faster than a speed threshold.

[0105] Similarly, if the UE is reaching a critical distance from the gNodeB (e.g., reaching the 202407065

[0106] 20 edges of the cell), the threshold may be lowered. The method of Fig. 3b may comprise obtaining 350, after providing the result of the inference operation to the network component, the updated information on the inference operation to be performed from the network component. For example, the UE may use the updated information on the inference operation to be performed for a subsequent instance of performing the inference operation.

[0107] The machine learning model(s) used for performing the respective inference operations may be vendor-specific - e.g., each UE vendor (or OS platform vendor, or baseband chip vendor) may supply one or more machine learning models for performing the inference operation with the respective UE. Moreover, depending on the hardware version, software version, build date, etc., different machine learning models may be deployed. Finally, different UEs may have different computational capabilities, e.g., due to the neural accelerators I tensor processing units being used to perform the inference operation. All this leads to a fragmented and heterogeneous environment. Therefore, prior to instructing the UE to perform the inference operation, the network may gather information on the UE’s capabilities with respect to performing inference operations. For example, the method of Fig. 3b may comprise, prior to obtaining the information on the inference operation to be performed, providing 310 information on a capability of the UE with respect to performing inference operations to the network component. In particular, the information on the capability of the UE with respect to performing inference operations may comprise at least one of information on machine learning models available at the UE (e.g., one or more of information on a purpose of the respective machine learning model, information on an output of the respective machine learning model, information on a version of the respective machine learning model, information on one or more inputs used for the machine learning model, or information on a time required for performing the inference, etc.) or information on a computational capability of the UE (e.g., availability of a neural accelerator / tensor processing unit, computational capability in terms of (matrix) operations per second, etc.). On the side of the network, the method of Fig. 4 may comprise obtaining 410 the information on the capability of the UE with respect to performing inference operations from the UE, and providing 420 the information on the inference operation 202407065

[0108] 21 to be performed by the UE based on the information on the capability of the UE with respect to performing inference operations. For example, the network component may use a data structure with information on which machine learning models are useful for a given task to select a machine learning model to use for the inference operation. Moreover, the network component may calculate the time window based on the time it takes the UE to perform the inference operation, including some additional time to account for delays in messaging, the UE being busy, etc.

[0109] In some cases, the UE and the gNodeB may attempt to further improve the inference performance by exchanging machine learning model parameters associated with the inference operation. In other words, the method of Fig. 3b may comprise providing 310, 340 (e.g., as part of the information on the capability, and / or with the result) information on machine learning model parameters associated with the inference operation to the network component. Accordingly, the method of Fig. 4 may comprise obtaining 410, 430 the information on machine learning model parameters associated with the inference operation from the UE. For example, the machine learning model parameters may comprise a portion of (or the entirety of) the neural network weights. Additionally, or alternatively, the machine learning parameters may include the inference-time parameters being used to perform inference using the machine learning model (such as a confidence threshold for interpreting the output, a time step in the input or output data, etc.). The network component may process these machine learning model parameters to determine updated (i.e. , improved) machine learning model parameters that can be deployed to the UE. Thus, the method of Fig. 4 may comprise providing 450, after obtaining the result of the inference operation from the UE, updated machine learning model parameters from the network component, the updated machine learning model parameters are used for a subsequent instance of performing the inference operation. Accordingly, the method of Fig. 3b may comprise, after providing the result of the inference operation to the network component, obtaining 350 updated machine learning model parameters from the network component. The UE may use the updated machine learning model parameters for a subsequent instance of performing the inference operation. 202407065

[0110] 22

[0111] As outlined above, the UE and the gNodeB may cooperate to further enhance inference accuracy and adaptability by exchanging one or more machine learning (ML) model parameters associated with a particular inference operation. In other words, the method of Fig. 3b may further comprise, at operations 310 and / or 340, providing to the network component information indicative of the ML model parameters used locally by the UE for inference (i.e. , the information on machine learning model parameters associated with the inference operation).

[0112] Correspondingly, the method of Fig. 4 may comprise, at operations 410 and / or 430, obtaining such ML model parameter information from the UE for analysis or refinement.

[0113] The ML model parameters may include, by way of non-limiting examples, a subset or the entirety of the neural network weight tensors, bias vectors, or layer-specific activation parameters of a model deployed at the UE (e.g., a convolutional neural network for object detection, or a recurrent model for mobility prediction). In certain implementations, the parameters may include hyperparameters associated with inference execution, such as a confidence threshold, a learning-rate scaling factor, an input normalization coefficient, a beam selection parameter, a context validity timeout, or a temporal sampling interval used for sensor data fusion. For lightweight implementations, the UE may provide compressed representations of the parameters, such as quantized weight matrices (e.g., 8-bit or 4-bit) or hash-encoded parameter differentials, thereby reducing uplink overhead.

[0114] The UE may embed these parameters or summaries thereof within one or more uplink control or data transmissions, such as a MAC Control Element (MAC CE), RRC Assistance Information message, or a dedicated logical channel configured for AI / ML feedback. The network component (e.g., gNodeB, edge server, or centralized training controller) may process the received parameter information to determine updated or improved / optimized model parameters. For example, the network may (i) perform federated parameter averaging across multiple UEs to produce globally improved weights, (ii) fine-tune the model using network-wide context or labeled datasets unavailable to the UE, or (iii) adjust inference-time thresholds to minimize false-positive rates detected across a population of devices. 202407065

[0115] 23

[0116] After such refinement, the method of Fig. 4 may further comprise, at operation 450, transmitting the updated model parameters (i.e., the updated machine learning model parameters) to the UE, wherein the updated model parameters are to be used in a subsequent instance of the inference operation. The update may be delivered via downlink broadcast, unicast signaling, or edge-caching delivery, depending on the system’s capability. In some embodiments, the network may transmit only the delta parameters (e.g., changes exceeding a threshold), or may specify a version identifier (Model-ID and Revision-Counter) to ensure synchronization between the UE and network models.

[0117] Accordingly, the method of Fig. 3b may further comprise, at operation 350, receiving and applying the updated ML model parameters, thereby producing a refined model instance for future inference tasks. This cyclical exchange allows for progressive performance enhancement without requiring complete model re-download, and can adapt dynamically to contextual variations such as radio environment, vehicular speed, user behavior, or sensor drift. For instance, a vehicular UE performing lanelevel positioning may periodically transmit its PRS-assisted inference thresholds and receive network-optimized updates that improve positioning accuracy by leveraging joint statistics across multiple UEs operating in the same cell. Similarly, in an Al- native beam management use case, the UE may report inference errors related to beam mis-classification, allowing the gNodeB to adjust model weights to reduce beam prediction latency or improve Top-K beam accuracy.

[0118] Such parameter exchange enables a closed-loop, distributed learning framework between the UE and network, wherein both sides iteratively contribute to optimizing the inference pipeline while maintaining compliance with 3GPP air-interface signaling principles. This mechanism reduces overall retraining cost, enhances model generalization under non-stationary environments, and achieves inference consistency across heterogeneous device classes.

[0119] Various messaging mechanisms may be used depending on which context the inference operation is performed in. For example, the UE may obtain the information 202407065

[0120] 24 on the inference operation to be performed (and other information provided by the network component to the UE) via at least one of L1 signaling (e.g., as part of Downlink Control Information (DCI)), L2 signaling (e.g., as part of Medium Access Control - Control Element (MAC-CE)), or L3 signaling (e.g., RRC signaling). For example, on layer 3, the information on the inference operation may be provided as RRC configuration or reconfiguration.

[0121] The wireless transceiver(s) 32, 42 may serve as an interface for communicating in the cellular mobile communication system. The wireless transceiver(s) 32, 42 may correspond to one or more inputs and / or outputs for receiving and / or transmitting information, which may be in digital (bit) values or analog according to a specified code or protocol within a module, between modules, or between modules of different entities. For example, a wireless transceiver may comprise interface circuitry configured to receive and / or transmit information. In examples, a wireless transceiver may comprise any means for obtaining, receiving, transmitting, or providing analog or digital signals or information, e.g., any connector, contact, pin, register, input port, output port, conductor, lane, etc., which allows providing or obtaining signals or information. The wireless transceiver(s) 32, 42 may be configured to communicate (transmit, receive, or both) in a wireless manner. The wireless transceiver(s) 32, 42 may comprise further components to enable communication in a (mobile) communication system or network; such components may include transceiver (transmitter and / or receiver) components, such as one or more Low-Noise Amplifiers (LNAs), one or more Power-Amplifiers (PAs), one or more duplexers, one or more diplexers, one or more filters or filter circuitry, one or more converters, one or more mixers, accordingly adapted radio frequency components, one or more antennas, etc. For example, the respective wireless transceiver(s) 32, 42 may enable radio communication with UEs and communication between base stations, which can be directly and / or indirectly wired and / or wireless, respectively.

[0122] For example, the processor(s) 34, 44 can be implemented using one or more processing units, processing devices, or any means for processing, such as a processor, a computer, or a programmable hardware component equipped with appropriately adapted software. Thus, the described function of the processor(s) 34, 202407065

[0123] 25

[0124] 44 can be executed in software running on one or more programmable hardware components. Such components may include a general-purpose processor, a Digital Signal Processor (DSP), a microcontroller, and more.

[0125] In at least some embodiments, the memory / memories 36, 46 may comprise at least one element of the group of a computer-readable storage medium, such as a magnetic or optical storage medium, e.g. a hard disk drive, a flash memory, Floppy Disk, Random Access Memory (RAM), Programmable Read Only Memory (PROM), Erasable Programmable Read Only Memory (EPROM), an Electronically Erasable Programmable Read Only Memory (EEPROM), or a network storage.

[0126] In examples, the network component may be terrestrial or non-terrestrial, and it may also be referred to as a base station, network node, etc. It may belong to an access network or to a core network. A non-terrestrial base station may be implemented in an aircraft, a satellite, or a High-Altitude Platform System (HAPS). A network component, e.g., a non-terrestrial base station (implemented at a satellite, platform, airplane, etc.) or a terrestrial base station, may generate cells of a cellular system. A network component may correspond to a remote radio head, a transmission point, an access point, a macro cell, a small cell, a micro cell, a pico cell, a femto cell, or a metro cell. The term small cell may refer to any cell smaller than a macro cell, e.g., a micro cell, a pico cell, a femto cell, or a metro cell. A network component / base station can be a wireless interface of a wired network, enabling transmission and reception of radio signals to a communication device. Such a radio signal may comply with radio signals, for example, standardized by 3GPP, or, generally, in line with one or more of the above-listed systems. Thus, a network component may be a base station and may correspond to a NodeB, an eNodeB, an ngNB, a gNB (gNodeB), a BTS (Base Transceiver Station), or an access point, all of which may be implemented in a satellite, plane, HAPS, etc. In case of a moving implementation in a satellite, an airplane, etc., the link towards a core network of the communication system may also be implemented in a wireless manner.

[0127] The mobile communication system may hence be cellular. The term cell refers to a coverage area of radio services provided by a transmission point, a remote unit, a 202407065

[0128] 26 remote head, a remote radio head, a communication device, a network component, or a NodeB, an eNodeB, an ngNB, a gNB, a beam, or a satellite, respectively. The terms cell and base station may be used synonymously; a base station may generate multiple cells and it may be implemented in a high-altitude platform, a plane, a drone, a satellite, etc. A wireless communication device, e.g., the UE, can be registered or associated with at least one cell (e.g., the network component); for example, it can be associated with a cell such that data can be exchanged between the network and the mobile in the coverage area of the associated cell using a dedicated channel, connection, or link.

[0129] In general, the UE may be a communication device that is capable of communicating wirelessly. In particular, however, the communication device may be a mobile communication device, e.g., a communication device that is suitable for being carried around by a user. For example, the communication device may be a User Terminal (UT) or UE within the meaning of the respective communication standards being used for mobile communication. For example, the communication device may be a mobile phone, such as a smartphone, a network access device embedded in a vehicle, ship, or airplane, or another type of mobile communication device, such as a computer, a laptop computer, a tablet computer, and so forth.

[0130] For example, the communication device and the network component may be configured to communicate in a cellular mobile communication system. Accordingly, the communication device and the network component may be configured to communicate in a cellular mobile communication system, for example, in a Sub- 6GHz-based cellular mobile communication system (covering frequency bands between 400 MHz and, in the meantime, 7 GHz), in a mmWave-based cellular mobile communication system (covering frequency bands between 24 GHz and 71 GHz), or in the so-called mid-bands (covering frequency bands between 7 GHz and 24 GHz). For example, the communication device and the network component may be configured to communicate in a mobile communication system / cellular mobile communication system. 202407065

[0131] 27

[0132] In general, the mobile communication system may, for example, correspond to one of the 3GPP-standardized mobile communication networks, where the term mobile communication system is used synonymously with mobile communication network. The mobile communication system may correspond to, for example, a 6th Generation system (6G), a 5th Generation system (5G), a New Radio (NR) system, a Long-Term Evolution (LTE), an LTE-Advanced (LTE-A), High Speed Packet Access (HSPA), a Universal Mobile Telecommunication System (UMTS) or a UMTS Terrestrial Radio Access Network (UTRAN), an evolved-UTRAN (e-UTRAN), a Global System for Mobile communication (GSM) or Enhanced Data rates for GSM Evolution (EDGE) network, a GSM / EDGE Radio Access Network (GERAN), or mobile communication networks with different standards, for example, an Orthogonal Frequency Division Multiple Access (OFDMA) network, a Time Division Multiple Access (TDMA) network, a Code Division Multiple Access (CDMA) network, a Wideband-CDMA (WCDMA) network, a Frequency Division Multiple Access (FDMA) network, or a Spatial Division Multiple Access (SDMA) network, etc.

[0133] For example, a serving base station is a terrestrial or non-terrestrial base station. A serving cell, often used interchangeably with the serving base station (such as an eNodeB in LTE or gNodeB in 5G), is the specific network cell that a mobile device or UE is currently connected to for all of its communication needs. When a UE is in an active state (specifically, RRC_CONNECTED mode), it establishes a dedicated radio link with one particular cell. This cell becomes the single point of contact responsible for managing the UE’s connection. It provides all the necessary radio resources for both sending (uplink, UL) and receiving (downlink, DL) data, handles the scheduling of these resources, and manages the UE’s mobility. Essentially, the serving cell is the primary cell that controls the radio connection, sends system information, and facilitates the handover process to another cell when the UE moves or radio conditions change.

[0134] More details and aspects of the UE 30 and gNodeB 40 are mentioned in connection with the proposed concept, or one or more examples described above or below (e.g., Figs. 1a to 2, 5a to 7b). The UE 30 and non-terrestrial gNodeB 40 may comprise one 202407065

[0135] 28 or more additional optional features corresponding to one or more aspects of the proposed concept, or one or more examples described above or below.

[0136] Fig. 5a shows a schematic diagram of a UE 30 moving relative to a gNodeB 40. The outer elliptical shape indicates the cell served by the gNodeB 40, while the inner elliptical shape indicates a threshold. Fig. 5b shows a sequence diagram of a gNodeB configuring an inference operation at UE. The process starts with the gNB 40 receiving a UE capability message from the UE 30 (optional). The gNB 40 then configures an AI / ML inference timer and / or threshold and provides it to the UE 30. The UE 30 provides the inference to the network if the inference is computed within the timer period. The gNB 40 dynamically updates the timer based on the UE’s proximity to the threshold, and provides the timer update to the UE 30.

[0137] In the proposed concept, the network configures at least one of an AI / ML threshold and AI / ML inference timer for the UE (i.e., a UE indicating support for AI / ML functionality). The UE may provide information on its support of AI / ML functionality, type of AI / ML model, AI / ML compute capabilities etc. The UE may provide this information in the UE capability message, UAI etc. The network (e.g., gNodeB) configures the UE with an AI / ML threshold and / or AI / ML inference timer. For example, the network may configure the AI / ML threshold and AI / ML timer to the UE via at least one of a RRC (Radio Resource Control) Configuration message, L1 signaling, L2 signaling, or L3 signaling. The network may also adjust the AI / ML timer for dynamic UE side model optimization via the RRC Reconfiguration signal, MAC- CE (Medium Access Control - Control Element), DCI (Downlink Control Information) etc.

[0138] The UE receives the AI / ML inference timer and / or the AI / ML threshold. The UE, upon receiving these, performs AI / ML model inference based on its AI / ML model. The AI / ML model may be initiated to perform inference once the threshold is crossed. Once the UE crosses the threshold, the timer starts, and the UE may perform the inference within the timing window. Once the UE completes the AI / ML inference before the timer expires, it may provide the gNB with the inference (e.g., the updated UE AI / ML model parameters). The updated UE’s AI / ML parameters may be the 202407065

[0139] 29 model parameters used for determining the inference. Additionally, UE’s AI / ML parameters may be the model parameters that it may want to use for subsequent inference, for a better inference depending on the inference in the current measurement cycle.

[0140] The gNB configures the AI / ML inference timer and / or the AI / ML threshold for the UE. The AI / ML threshold defines when the UE will initiate the AI / ML model inference. The AI / ML timer defines the time within which the UE must perform the AI / ML inference. Upon receiving the AI / ML model inference from the UE, the gNB may either dynamically update the UE’s model parameter, or signal the UE to perform the inference with the same model parameters.

[0141] Fig. 6 shows a schematic diagram of a UE moving relative to a gNodeB, with two different thresholds (T1 , T2) being configured. T1 and T2 are the thresholds defined by the network for some use-cases where UE speed has impact for the AI / ML inference. Based on how quickly the UE crosses the threshold, a computation of the UE’s speed is performed. Based on this, e.g., in subsequent cycles, the gNB may dynamically adapt the AI / ML parameters for the UE to perform AI / ML inference.

[0142] Fig. 7a shows a flowchart of one or more operations performed by the gNodeB 40. For example, the gNodeB configures 420 the AI / ML model threshold and / or the AI / ML inference timer (see also Fig. 4).

[0143] Fig. 7b shows a flowchart of one or more operations performed by the UE 30. For example, the UE may provide 310 the UE capability information along with Al model parameters. The UE performs inference 330 within the configured timer period. The UE provides 340 the inference along with UE AI / ML model parameters to the gNodeB (see also Fig. 3).

[0144] The proposed concept enables energy savings at the UE while improving the robustness of model inference. 202407065

[0145] 30

[0146] AI / ML is an important technology being actively discussed in 3GPP. AI / ML is one of the key topics for Rel-19 in RAN2. AI / ML is considered highly relevant for network optimization and energy efficient networks. The present invention relates to 3GPP AI / ML Rel-19 Wl for Mobility. It is intended to provide fundamental mechanisms of interworking, and data information flow in radio access network collaboration for AI / ML support. Based on the proposed gNB-UE collaboration operation for AI / ML support, AI / ML performance for wireless communication can be improved.

[0147] 202407065

[0148] 31

[0149] Abbreviations:

[0150] 2G: Second Generation

[0151] 3G: Third Generation

[0152] 3GPP: 3rd-Generation Partnership Project

[0153] 4G: Fourth Generation

[0154] 5G: Fifth Generation

[0155] AI / ML: Artificial Intelligence I Machine Learning

[0156] AP: Access Point

[0157] BSC: Base Station Controller

[0158] BTS: Base Transceiver Station

[0159] CDMA: Code Division Multiple Access

[0160] CQI: Channel Quality Indicator

[0161] CSI: Channel State Information

[0162] D2D: Device-to-Device

[0163] DAS: Distributed Antenna System

[0164] DCI: Downlink Control Information eNB: Evolved Node B

[0165] E-SMLC: Evolved-Serving Mobile Location Centre

[0166] FDMA: Frequency Division Multiple Access gNodeB (gNB): Next Generation NodeB (also known as gNB)

[0167] L1 : Layer 1

[0168] L2: Layer 2

[0169] L3: Layer3

[0170] LME: Laptop Mounted Equipment

[0171] LTE: Long Term Evolution

[0172] M2M: Machine-to-Machine

[0173] MAC-CE: Medium Access Control - Control Element

[0174] MeNB: Mobility-enhanced Node B

[0175] MME: Mobility Management Entity

[0176] MSC: Mobile Switching Center ngNB: Next Generation Node B

[0177] NodeB: Base Station (2G / 3G) 202407065

[0178] 32

[0179] NR: New Radio

[0180] OFDMA: Orthogonal Frequency Division Multiple Access

[0181] OSS: Operations Support System

[0182] PMI: Precoding Matrix Indicator

[0183] ProSe UE: Proximity Services User Equipment

[0184] Rl: Rank Indicator

[0185] RNC: Radio Network Controller

[0186] RRC: Radio Resource Control

[0187] RRH: Remote Radio Head

[0188] RRM: Radio Resource Management

[0189] RRU: Remote Radio Unit

[0190] RSRP: Reference Signal Received Power

[0191] SBA: Service-Based Architecture

[0192] SDMA: Spatial Division Multiple Access

[0193] SINR: Signal-to-lnterference-plus-Noise Ratio

[0194] SON: Self-Optimised Network

[0195] TDMA: Time Division Multiple Access

[0196] UE: User Equipment

[0197] V2V UE: Vehicle-to-Vehicle User Equipment

[0198] V2X UE: Vehicle-to-X User Equipment

[0199] WCDMA: Wideband Code-Division Multiple Access

Claims

20240706533CLAIMS1 . A method for a user equipment, UE, (30) comprising: obtaining (320), from a network component of a cellular mobile communication system, information on an inference operation to be performed by the UE, wherein the information on the inference operation to be performed comprises at least one of information on a threshold used to trigger the inference operation or information on a timing window for performing the inference operation; performing (330) the inference operation according to the information on the inference operation to be performed; and providing (340) a result of the inference operation to the network component.

2. The method according to claim 1 , wherein the method comprises obtaining (350), after providing the result of the inference operation to the network component, updated information on the inference operation to be performed from the network component, wherein the updated information on the inference operation to be performed is used for a subsequent instance of performing the inference operation.

3. The method according to one of the claims 1 or 2, wherein the threshold used to trigger the inference operation is related to a relative movement or distance between the UE and a serving gNB serving the UE.

4. The method according to claim 3, wherein the threshold is set to trigger the inference operation when the UE is moving away from the serving gNB.

5. The method according to one of the claims 1 to 4, wherein the method comprises, prior to obtaining the information on the inference operation to be performed, providing (310) information on a capability of the UE with respect to performing inference operations to the network component.

6. The method according to claim 5, wherein the information on the capability of the UE with respect to performing inference operations comprises at least one20240706534 of information on machine learning models available at the UE or information of a computational capability of the UE.

7. The method according to one of the claims 1 to 6, wherein the UE obtains the information on the inference operation to be performed via at least one of L1 signaling, L2 signaling, L3 signaling, or Radio Resource Control, RRC, configuration or reconfiguration.

8. The method according to one of the claims 1 to 7, wherein the method comprises providing (310, 340) information on machine learning model parameters associated with the inference operation to the network component.

9. The method according to claim 8, wherein the method comprises, after providing the result of the inference operation to the network component, obtaining (350) updated machine learning model parameters from the network component, wherein the updated machine learning model parameters are used for a subsequent instance of performing the inference operation.

10. The method according to one of the claims 1 to 9, wherein the inference operation is related to performing a radio resource measurement or to performing a channel state information measurement.11 . A method for a network component of a cellular mobile communication system, comprising: providing (420), to a user equipment, UE, information on an inference operation to be performed by the UE, wherein the information on the inference operation to be performed comprises at least one of information on a threshold used to trigger the inference operation or information on a timing window for performing the inference operation; and obtaining (430) a result of the inference operation from the UE.

12. The method according to claim 11 , wherein the method comprises determining (440), after obtaining the result of the inference operation from the UE, updated information on the inference operation to be performed, and providing20240706535(450) the updated information on the inference operation to be performed to the UE.

13. The method according to one of the claims 11 or 12, wherein the method comprises obtaining (410) information on a capability of the UE with respect to performing inference operations from the UE, and providing (420) the information on the inference operation to be performed by the UE based on the information on the capability of the UE with respect to performing inference operations, and / or wherein the method comprises obtaining (410; 430) information on machine learning model parameters associated with the inference operation from the UE, and providing (450), after obtaining the result of the inference operation from the UE, updated machine learning model parameters from the network component, wherein the updated machine learning model parameters are used for a subsequent instance of performing the inference operation.

14. A UE (30) comprising a wireless transceiver (32), a processor (34) coupled with a memory (36) in which computer program instructions are stored, said instructions being configured to implement the method of one of the claims 1 to 10.

15. A gNB (40) comprising a wireless transceiver (42), a processor (44) coupled with a memory (46) in which computer program instructions are stored, said instructions being configured to implement the method of one of the claims 11

Images