Perception of user equipment side model training by radio access network

Abstract

A User Equipment, UE, (110) and a network node (190) within a wireless communication network exchange messages to perform Life Cycle Management, LCM, of an Artificial Intelligence, AI, model. The model is a UE-side model, and reasoning for the UE-side model is performed at the UE (110). The UE (110) performs data collection to support a UE-side model.

Description

[0001] Related applications This application claims priority to U.S. Provisional Patent Application No. 63 / 539387, filed September 20, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This disclosure generally relates to the field of wireless communication technology, and more specifically to lifecycle management (LCM) of artificial intelligence (AI) and / or machine learning (ML) models that perform inference at a user equipment (UE). Background Technology

[0003] AI and ML have been studied in both academia and industry as promising tools for optimizing air interface design in wireless communication networks. Example use cases include: using autoencoders for channel state information (CSI) compression to reduce feedback overhead and improve channel prediction accuracy; using deep neural networks for line-of-sight (LOS) and non-LOS (NLOS) condition classification to enhance positioning accuracy; using reinforcement learning for beam selection on the network side and / or UE side to reduce signaling overhead and beam alignment delay; and using deep reinforcement learning to learn optimal precoding strategies for complex multiple-input multiple-output (MIMO) precoding problems.

[0004] In the 3GPP New Radio (NR) standardization work, the new version 18 research project on AI / ML for the NR air interface was initiated in May 2022. This research project will explore the benefits of enhancing the air interface through features that enable improved support for AI / ML-based algorithms to enhance performance and / or reduce complexity / overhead. By studying several selected use cases (CSI feedback, beam management, and positioning), this research project aims to lay the foundation for future air interface use cases leveraging AI / ML technologies. Summary of the Invention

[0005] This disclosure generally relates to the LCM of a model in which inference is performed at the UE.

[0006] The embodiments described herein include a first method implemented by a UE. The method includes: transmitting an instruction to a network node indicating that the UE needs to perform UE-side model training. The method further includes: performing data collection for UE-side model training.

[0007] In some embodiments, the instruction includes a request to perform UE-side model training. Performing data collection is in response to receiving the instruction from the network node. The response includes configuration for the UE to use for data collection and / or acceptance of the instruction.

[0008] In some embodiments, the method further includes: receiving a notification from the network node or a different network node that data collection for UE-side model training has been stopped or suspended. The method further includes: transmitting to the network node or different network nodes a further indication that data collection for UE-side model training has been stopped or suspended.

[0009] In some embodiments, the method further includes notifying the network node or different network nodes that data collection for UE-side model training has been completed; or that it needs to be resumed and / or restarted and / or reconfigured.

[0010] Other embodiments include a UE, which includes processing circuitry and memory. The memory contains instructions executable by the processing circuitry, thereby configuring the UE to transmit instructions to a network node indicating that the UE needs to perform UE-side model training. The UE is further configured to perform data collection for UE-side model training.

[0011] In some embodiments, the UE is further configured to perform any of the methods described in the first method above.

[0012] Other embodiments include a second method performed by the UE. The method includes: receiving a first message from a first entity, the first message indicating that UE-side model training needs to be performed for one or more lower-layer functions and / or one or more higher-layer functions. The one or more lower-layer functions include PHY and / or MAC layer functions.

[0013] In some embodiments, one or more lower-level functions include beam management, providing CSI, and / or positioning. One or more higher-level functions include RRM measurement and / or L3 mobility functions.

[0014] In some embodiments, the method further includes: in response to receiving a first message, indicating in a second message a request to the RAN node to perform UE-side model training. The method further includes: receiving a third message from the RAN node, the third message indicating whether the UE is permitted to perform UE-side model training. The method further includes: in response to the third message indicating permission for the UE to perform UE-side model training, initiating data collection for UE-side model training at one or more lower layers. The method further includes: in response to receiving the third message, sending a fourth message to a first entity, the fourth message indicating whether the request has been accepted or rejected.

[0015] In some embodiments, the method further includes: informing a first entity that data collection has begun.

[0016] Other embodiments include a UE, which includes processing circuitry and memory. The memory contains instructions executable by the processing circuitry, thereby configuring the UE to receive a first message from a first entity, the first message indicating that UE-side model training needs to be performed for one or more lower-layer functions and / or one or more higher-layer functions. The one or more lower-layer functions include PHY and / or MAC layer functions.

[0017] In some embodiments, the UE is further configured to perform any of the methods in the second method described above.

[0018] Other embodiments include a third method implemented by the UE. The method includes: performing data collection for UE-side model training. The method further includes: receiving a first message from the RAN node indicating a halt to data collection.

[0019] In some embodiments, the method further includes: in response to receiving a first message, stopping data collection for UE-side model training at one or more lower layers. The one or more lower layers include a PHY layer and / or a MAC layer.

[0020] In some embodiments, the method further includes: transmitting a second message to a first entity indicating that data collection for UE-side model training has been stopped.

[0021] In some embodiments, the method further includes: receiving from the RAN node a third message indicating the resumption of stopped data collection. The method further includes: in response to receiving the third message, resuming data collection for UE-side model training at one or more lower layers. The one or more lower layers include a PHY layer and / or a MAC layer.

[0022] In some embodiments, the method further includes: transmitting a fourth message to a first entity indicating that data collection for UE-side model training has been resumed.

[0023] Other embodiments include a UE, which includes processing circuitry and memory. The memory contains instructions executable by the processing circuitry, thereby configuring the UE to perform data collection for UE-side model training. The UE is further configured to receive a first message from the RAN node indicating a halt to data collection.

[0024] In some embodiments, the UE is further configured to perform any of the methods described in the third method above.

[0025] Other embodiments include a fourth method implemented by the UE. The method includes: stopping data collection for UE-side model training. The method further includes: transmitting a first message to the RAN node indicating that data collection for UE-side model training has been stopped.

[0026] In some embodiments, stopping data collection for UE-side model training includes stopping data collection at one or more lower layers. The one or more lower layers include the PHY layer and / or the MAC layer.

[0027] In some embodiments, the method further includes: transmitting a second message to a first entity indicating that data collection for UE-side model training has been stopped.

[0028] In some embodiments, the method further includes: determining that data collection for UE-side model training can be resumed. The method further includes: transmitting a third message to the RAN node indicating that data collection for UE-side model training can be resumed.

[0029] In some embodiments, the method further includes: receiving a fourth message from a RAN node indicating whether data collection has been resumed. The method further includes: resuming data collection for UE-side model training at one or more lower layers in response to the fourth message indicating the resumption of data collection. The method further includes: transmitting a fifth message to a first entity, the fifth message indicating whether data collection for UE-side model training has been resumed according to the fourth message.

[0030] Other embodiments include a UE, which includes processing circuitry and memory. The memory contains instructions executable by the processing circuitry, thereby configuring the UE to stop data collection for UE-side model training. The UE is further configured to transmit a first message to the RAN node indicating that data collection for UE-side model training has been stopped.

[0031] In some embodiments, the UE is further configured to perform any of the methods in the fourth method described above.

[0032] Other embodiments include a computer program comprising instructions that, when executed on the processing circuitry of the UE, cause the processing circuitry to perform any of the methods described in the UE method.

[0033] Other embodiments include a method implemented by a network node. The method includes receiving from a UE an instruction indicating that the UE needs to perform UE-side model training. The method further includes: in response to receiving the instruction, transmitting a response to the UE indicating that the UE is permitted to perform data collection for UE-side model training, or should perform data collection for UE-side model training. The response includes accepting the instruction and / or providing the UE with configuration for data collection.

[0034] In some embodiments, the instruction includes a request from the UE to perform UE-side model training. The method further includes determining whether to accept or reject the request.

[0035] In some embodiments, the method further includes receiving from the UE: a request to stop or suspend data collection for UE-side model training, or a notification that the UE has stopped or suspended data collection for UE-side model training, or an instruction to resume and / or restart and / or reconfigure data collection for UE-side model training.

[0036] In some embodiments, the method further includes: transmitting to the UE an instruction to stop or pause data collection for the UE-side model.

[0037] Other embodiments include a network node that includes processing circuitry and memory. The memory contains instructions executable by the processing circuitry, thereby configuring the network node to receive from the UE an instruction indicating that the UE needs to perform UE-side model training. The network node is further configured to, in response to receiving the instruction, transmit a response to the UE indicating that the UE is permitted to perform data collection for UE-side model training, or should perform data collection for UE-side model training. The response includes accepting the instruction and / or providing configuration for the UE to use for data collection.

[0038] In some embodiments, the network node is further configured to perform any of the methods described above in the network node method.

[0039] Other embodiments include a second method implemented by a network node. The method includes: receiving from the UE a request to perform data collection for UE-side model training. The method further includes: determining whether to accept or reject the request. The method further includes: indicating to the UE whether the request is accepted or rejected.

[0040] In some embodiments, the method further includes: determining that the UE should stop data collection for UE-side model training. The method further includes: instructing the UE to stop data collection for UE-side model training. The method further includes: determining that the UE should resume data collection. The method further includes: instructing the UE to resume data collection.

[0041] In some embodiments, the method further includes receiving from the UE: a notification indicating that data collection for UE-side model training has been stopped or paused; and / or a notification that data collection for UE-side model training can be resumed. The method further includes informing the UE whether data collection should be resumed.

[0042] Other embodiments include a network node that includes processing circuitry and memory. The memory contains instructions executable by the processing circuitry, thereby configuring the network node to receive a request from the UE to perform data collection for UE-side model training. The network node is further configured to determine whether to accept or reject the request. The network node is further configured to indicate to the UE whether the request is accepted or rejected.

[0043] In some embodiments, the network node is further configured to perform any of the methods described in the second network node method above.

[0044] Other embodiments include a computer program comprising instructions that, when executed on the processing circuitry of a network node, cause the network node to perform any of the methods described above for network node methods.

[0045] Other embodiments include a carrier that contains any of the computer programs described above. The carrier is one of an electronic signal, an optical signal, a radio signal, or a computer-readable storage medium. Attached Figure Description

[0046] Various aspects of this disclosure are shown by way of example and are not limited to the accompanying drawings, in which similar references indicate similar elements. In general, the use of reference numerals should be regarded as referring to the subject matter depicted according to one or more embodiments, and a discussion of specific instances of the illustrated elements will be followed by letter names (e.g., a general discussion of computing device 110 relative to a discussion of specific instances 110a, 110b of the computing device).

[0047] Figure 1 This is a schematic block diagram illustrating an example model LCM process according to one or more embodiments of the present disclosure.

[0048] Figure 2 This is a schematic block diagram illustrating an example framework for studying aspects of the LCM model according to one or more embodiments of the present disclosure.

[0049] Figure 3 This is a schematic block diagram illustrating an example automatic encoder for CSI according to one or more embodiments of the present disclosure.

[0050] Figure 4 This is a schematic block diagram illustrating an example wireless communication network according to one or more embodiments of the present disclosure.

[0051] Figures 5-11 This is a signaling diagram illustrating an example of signaling exchanged according to one or more embodiments of the present disclosure.

[0052] Figures 12-15 This is a flowchart illustrating an example method implemented by a UE according to one or more embodiments of this disclosure.

[0053] Figures 16-17 This is a flowchart illustrating an example method implemented by a network node according to one or more embodiments of the present disclosure.

[0054] Figure 18 Example UEs are shown according to one or more embodiments of this disclosure.

[0055] Figure 19 An example network node is shown according to one or more embodiments of this disclosure. Detailed Implementation

[0056] As used herein, the term "model" refers to one or more data structures and / or algorithms used to generate predictions from collected input data. The terms "model," "ML model," "AI model," "AI / ML model," and "AI and / or ML model" should be considered equivalent to each other and are therefore interchangeable. As will be discussed in more detail below, a model may be deployed, implemented, and / or configured in a UE, a network node, or both.

[0057] In one example, the model may receive a measurement of a reference signal at time instance t0 (e.g., a measurement of a synchronization signal block (SSB)) as input and provide a prediction of the reference signal at time t0+T as output. In another example, the model may receive a measurement of a reference signal transmitted on a first beam as input and provide a prediction of another reference signal transmitted on a second beam as output.

[0058] Another example is a model used for auxiliary channel state information (CSI) estimation. In such examples, the model may include a UE-side specific model and a network (NW)-side specific model, which operate jointly. The UE-side model may compress the channel input, while the NW-side model may decompress the output received from the UE.

[0059] Other examples can be similarly applied to localization. For instance, the input to the model could be a channel pulse associated with a time reference point. The model's NW side could detect different peaks within the impulse response corresponding to different reception directions of the radio signal on the UE side. Another example related to localization is inputting multiple sets of measurements into an ML network and deriving an estimated locality based on these.

[0060] Another example of a model is one that assists the UE in channel estimation (or interference estimation for channel estimation). Channel estimation may, for example, be for the Physical Downlink Shared Channel (PDSCH) and is associated with a specific set of reference signaling patterns transmitted from the NW to the UE. This model may be part of the receiver chain within the UE and may not be directly visible within the reference signaling patterns, and may be configured or scheduled for use between the NW and the UE.

[0061] Another example of a model used for CSI estimation is predicting a suitable future Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), Rank Indicator (RI), or similar value. The future could be a certain number of time slots after the UE has performed its last measurement, or a specific time slot within a future time period.

[0062] The UE is connected to a network (e.g., it can receive and transmit data and / or control information). The UE may further be in the RRC_CONNECTED state and configured to use the model for a specific function. The specific function may include one or more of the following "functional areas": - CSI Report - Beam management - RRM Measurement L3 mobility - Conditional switching - Lower-level triggered mobility (LTM) - HARQ teleportation - Data transmission - Data reception - Power control.

[0063] Examples of RRM measurements include mobility measurements such as Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), Received Signal Strength Indicator (RSSI), Radio Link Failure (RLF) prediction, and other aspects related to radio link failure. As will be discussed further below, a measurement framework can be used to perform RRM measurements. The measurement framework governs how the UE performs measurements (e.g., through measurement configuration), what triggers measurement reporting (e.g., whether measurement reports are event-triggered or periodically sent), and what content is included in the measurement reports.

[0064] A “UE-side model” is a model in which the UE performs inference. Training of the UE-side model can be performed at the UE and / or outside the mobile network, such as in an over-the-top (OTT) server (e.g., managed by the UE manufacturer). According to one or more embodiments of this disclosure, the UE indicates that it needs to perform data collection for UE-side model training. The data is collected by the UE, and it can be used by the UE to perform training and / or passed to an OTT server so that the OTT server performs model training. The model can then be provided back to that UE and / or other UEs, for example, to update existing models at (one or more) UEs.

[0065] Building a model can involve several development steps. The actual training of the model may only be one step in the training pipeline. A crucial part of model development is the model's lifecycle management (LCM). Figure 1This is a diagram of the training pipeline 10 and the inference pipeline 20, and their interactions within the model LCM process 30. The model LCM process 30 typically includes the training pipeline 10 (which may be used partially or entirely for retraining or not), the model deployment phase 16, the inference pipeline 20, and the drift detection phase 25.

[0066] The training pipeline 10 may include the following stages: data ingestion 11, data preprocessing 12, model training 13, model evaluation 14, and / or model registration 15.

[0067] Data ingestion 11 refers to collecting raw data (e.g., training data) from a data store. Following data ingestion 11, there may be steps to control the validity of the collected data.

[0068] Data preprocessing 12 refers to the feature engineering applied to the collected data. For example, data preprocessing 12 may include data normalization and / or data transformation required for the model's input data.

[0069] Model training 13 refers to the actual model training steps.

[0070] Model evaluation 14 refers to benchmarking the model's performance against a certain model baseline. The iterative steps of model training 13 and model evaluation 14 can continue until an acceptable performance level is reached.

[0071] Model registration 15 refers to registering a model, including any corresponding metadata that provides information on how the model was developed, as well as possible model evaluation performance results.

[0072] The model deployment phase 16 makes the trained (e.g., retrained) model part of the inference pipeline 20.

[0073] The inference pipeline 20 may include the following stages: data ingestion 21, data preprocessing 22, model operation 23, and data and / or model monitoring 24.

[0074] Data ingestion 21 refers to the collection of raw data (e.g., inference data) from data storage.

[0075] The data preprocessing 22 used in the inference pipeline 20 is essentially similar to the corresponding processing that occurs in the training pipeline 10.

[0076] Model operation 23 refers to using a trained and deployed model in the operation mode.

[0077] Data and model monitoring 24 refers to verifying that the inference data comes from a distribution that is well aligned with the training data, and monitoring the model output to detect any performance or operational drift.

[0078] Drift detection phase 25 notifies the model of any drift during operation.

[0079] Figure 2 This illustrates a functional framework for studying LCM models. This framework can, for example, be used to study different network (NW)-UE cooperation levels for physical layer use cases.

[0080] The models being discussed in the Rel-18 research project on AI / ML for NR air interfaces can be divided into two types: one-sided AI / ML models and two-sided AI / ML models.

[0081] A one-sided AI / ML model can be a UE-side model (where inference is performed entirely at the UE) or an NW-side model (where inference is performed entirely at the NW).

[0082] A two-sided AI / ML model refers to a pairwise model that performs joint inference across the UE and NW. That is, the first part of the inference is performed by the UE, while the remaining part is performed by the network node (e.g., at the next-generation node B (gNB)), or vice versa.

[0083] Figure 3 An example of a two-sided CSI compression use case based on an autoencoder (AE) is shown. In this example, the UE uses encoder 42 (i.e., the UE portion of the two-sided AE model 40) operated in UE operation to compress the CSI 41 measured for the radio channel. The output of encoder 42 (i.e., the compressed CSI 43) is reported from the UE to the gNB. The gNB uses decoder 44 (i.e., the NW portion of the two-sided AE model 40) to generate the reconstructed CSI 45 for the radio channel.

[0084] When applying AI and / or ML to air interface use cases, different levels of collaboration between network nodes and UEs can be considered. In one example, there is no collaboration between the network node and the UE. In this case, a proprietary model that operates with the existing standard air interface is applied at one end of the communication chain (e.g., on the UE side), and model LCM (e.g., model selection / training, model monitoring, model retraining, model updating) is completed at this node without inter-node assistance (e.g., assistance information provided by the network node).

[0085] In another example, for a one-sided model, there is limited collaboration between the network node and the UE. In this case, the model operates at one end of the communication chain (e.g., on the UE side), but this node receives some assistance from one or more nodes at the other end of the communication chain (e.g., on the gNB) for its model LCM (e.g., for training / retraining AI models, model updates, model monitoring, model selection, fallback, and / or switching).

[0086] In another example, for a two-sided model, there is joint operation between the network node and the UE. In this case, it is assumed that the model is split, with one part located on the NW side and the other on the UE side. Therefore, the model requires joint inference between the NW and the UE, and the model LCM involves both ends of the communication chain.

[0087] Several methods exist for UE-side model monitoring. One approach involves training the UE-side model on the UE itself. That is, the UE performs both training and inference. However, this method can be overly complex in practice. For example, considering the UE's limited computing resources and the potentially enormous computational complexity of training, having the UE perform either or both of the training and inference processes may be impractical. Furthermore, if the model depends on location and / or region, a single UE is unlikely to be able to cover the entire coverage area. In reality, the model trained by the UE itself will be limited to the area where the UE is moving. Therefore, whenever the UE enters a new area, its trained model may become outdated.

[0088] In view of the above, alternative methods for training the UE-side model include the following possibilities: network nodes, such as radio access network (RAN) nodes (e.g., gNB) or core network (CN) nodes (e.g., network data analysis function (NWDAF)), collect data from the UE and train the model, which should be transferred to the UE or other UEs at some point, and the UE will then apply it.

[0089] According to another example, an over-the-top (OTT) server outside the 3GPP environment can be responsible for performing the training. This server could be, for example, a UE vendor-specific server. This latter approach is likely a reasonable candidate because, for optimal performance, the training dataset should be adapted to the inference operations at the device, which can depend on the UE vendor-specific implementation (e.g., software / hardware attributes / capabilities).

[0090] In view of the above, Figure 4 This is a schematic diagram illustrating an example wireless communication network 100, which includes a RAN node 120, a core network node 140, an over-the-top (OTT) node 150, and a UE 110. The RAN node 120 provides cell 130 to the UE 110. Cell 130 supports RAT (e.g., NR, 6G), which provides the UE 110 with access to a core network 160, which includes core network node 140. The core network 160 provides access to an OTT server 150, which can be located outside the core network 160.

[0091] If UE-side model training is performed by a network node outside RAN 130—for example, in CN node 140, outside the 3GPP network (such as on a UE vendor-specific server), on OTT server 150, or by UE 110 itself (e.g., in the UE's application layer)—then under the current 3GPP specification, RAN node 120 is not aware of this operation. This can lead to undesirable behavior that ultimately affects overall system performance (and potentially UE performance). In some scenarios, it may be possible that the training operations performed by the UE require specific configurations and settings at the UE, which may conflict with the configurations provided by the NW. For example, NW 100 may expect UE 110 to perform measurements with a certain accuracy or performance (e.g., a relaxed mode), while a training request from OTT server 150 may require a different accuracy. In other scenarios, NW may require specific operations (e.g., for energy-saving purposes) that may affect training performance at the UE. Therefore, without RAN awareness, training operations at UE 110 may lead to unexpected and / or inaccurate behavior and / or results for either or both UE 110 and network nodes.

[0092] Another example of training impacting performance is if UE 110 needs to train a model, which requires measurements on frequencies not configured in network 100 for typical Radio Resource Management (RRM) measurements to support mobility decisions and carrier aggregation—for example, frequencies that are not serving frequencies or adjacent frequencies for which measurements are configured. In this case, UE 110 may need to leave the serving cell to switch to the frequencies requiring measurements, and UE 110 may not be configured with measurement gaps or the measurement gaps may be too short. This can lead to frequent handovers or reconfigurations, impacting performance within network 100.

[0093] Furthermore, it is anticipated that at some point, for example, when data collection is complete, UE 110 may need to upload the collected data to the node performing the training, such as to a UE vendor-specific server. Uploading this collected data could impact RAN 130 system performance in terms of available uplink (UL) radio resources, and the delivery of the collected data itself may be delayed if cell congestion occurs.

[0094] Embodiments of this disclosure include a method implemented by a UE. The method includes transmitting an instruction to a network node (e.g., RAN node 120, such as a gNB) indicating that the UE needs to perform data collection for model training of at least one function (e.g., beam management, CSI, positioning, L3 mobility, RRM measurement) for UE-side model training. In this context, performing data collection may include performing model training itself at the UE. Throughout this disclosure, "data collection for UE-side model training" refers to data collection for UE-side model training of at least one function.

[0095] In some embodiments (and as will be discussed further below), the indication may correspond to a request from the UE to the network node, or an indication for notifying the network node.

[0096] In some embodiments, in response to a transmitted instruction, the UE receives a response from the network node, such as an acceptance instruction, a rejection instruction, or a configuration for the UE to use for data collection. When the UE receives an acceptance instruction, the UE performs data collection for UE-side model training.

[0097] In some embodiments, when the UE (e.g., in response to an acceptance instruction) is performing data collection for UE-side model training, the UE transmits an instruction to stop and / or pause data collection for the UE-side model. Alternatively, the UE transmits an instruction that the UE has stopped and / or paused data collection for the UE-side model. This instruction is transmitted to a network node (e.g., the same network node to which the UE has transmitted an instruction that it needs to perform data collection for UE-side model training, or a different network node).

[0098] In some embodiments, after the UE (e.g., in response to receiving an instruction) has started performing data collection for UE-side model training, the UE receives an instruction to stop and / or pause data collection for the UE-side model from a network node (e.g., the same network node to which the UE has transmitted an instruction that it needs to perform data collection for UE-side model training, or a different network node).

[0099] In some embodiments, after the UE has started data collection for UE-side model training (e.g., in response to an acceptance instruction), the UE transmits an instruction notifying the NW that data collection for UE-side model training has been completed.

[0100] In some embodiments, after the UE has stopped (e.g., in response to accepting an instruction) performing data collection for UE-side model training, the UE transmits an instruction that notifies the NW that data collection for UE-side model training needs to be resumed and / or restarted and / or reconfigured.

[0101] It should be noted that in this context, UE generally refers to a mobile terminal or a device that includes a UE performing one or more specified functions.

[0102] The embodiments also include methods for network nodes (e.g., RAN node 120) to determine whether a UE can start, stop, or resume data collection for UE-side model training and for transmitting responses to indications transmitted from the UE.

[0103] Therefore, a network node may become aware of whether UE 110 needs to perform model training by performing data collection, which may require UE 110 to perform one or more measurements that can be used as inputs for model training. Thus, if a UE request to perform UE-side model training is accepted, the network may, for example, provide the UE with the necessary configuration to perform appropriate UE-side model training, such as enabling the UE to perform data collection for UE-side model training purposes, and / or understand whether it expects performance degradation due to the data collection process performed by the UE.

[0104] Furthermore, the solution proposed in this paper enables the network to avoid one or more side effects or unintended consequences of model training on the performance of the UE and / or the network, for example, by avoiding conflicts between the network's expected configuration / behavior at the UE and the AI / ML-based training activities at the UE.

[0105] Figure 5 A first example of signaling according to one or more embodiments of this disclosure is shown. Figure 5 As shown, the UE transmits an instruction to the network node indicating that the UE needs to perform UE-side model training (step 1510). The action of "model training" (step 1520) may include the UE performing measurements and / or performing data collection (which may include measurements already performed) for the purpose of training one or more UE-side AI / ML models, for example, in the case where data is collected at the UE and reported to an OTT server for use in AI / ML model training and / or in the case where data is collected at the UE and model training is performed by the UE itself.

[0106] Network nodes may correspond to radio access network (RAN) nodes (such as gNodeB, eNodeB, 6G radio access network nodes), centralized units in the RAN, servers operating baseband, and / or core network (CN) nodes (such as authentication and mobility functions (AMF)).

[0107] The instruction transmitted by the UE may correspond to a request to the network node to perform UE-side model training. In other words, the UE does not perform UE-side model training and / or data collection until it receives a response from the network node.

[0108] In one option, the UE starts a monitoring timer when it sends a request to the network node. While the timer is running, the UE expects a response, and if no response is received and the timer expires, the UE terminates the process and does not perform data collection and / or UE-side model training.

[0109] In one option, for example, where the network node corresponds to RAN node 120 (such as gNodeB or a 6G RAN node), the request includes an RRC message. In this case, the RRC message may be a UE assistance information message containing instructions that the UE needs to perform UE-side training of the AI / ML model and / or data collection for AI / ML model training.

[0110] A request to a network node to perform UE-side model training and / or data collection for AI / ML model training may include one or more of the following information: • An indication of whether a reduction in capability is anticipated or necessary, wherein the indication may specify, for example, the need for the UE to perform lenient measurements at certain frequencies, the need to reduce the MIMO layer, or the need to avoid DAPS handover; • Instructions for the desired configuration used to perform appropriate data collection, such as the desired DRX configuration, desired CSI-RS / SSB resources, and desired frequency; • The UE needs to execute instructions on the frequency of data collection for UE-side model training; • The UE needs to specify the time period or interval for data collection used for UE-side model training; • Time periods or intervals may be provided as one or more time units (such as: the number of radio frames and / or subframes and / or OFDM symbols, seconds, minutes, hours, etc.); • The timing indication may include the initial time unit in which the UE expects to begin data collection, for example, the first radio frame and / or subframe and / or OFDM symbol in which the UE expects to begin data collection; • The UE needs to provide instructions for executing DRB, QoS stream, or PDU sessions for data collection used for UE-side model training; • Indications identifying training use cases and / or AI / ML functionality, such as training for channel state information (CSI) compression, positioning, or beam management. These indications may be a list specifying different training requests for UE-side model training received via an over-the-top server; • The UE expects an indication of the type of data collected for training AI / ML models. Examples of data that can be specified are SSB-related data, such as measurements (e.g., SS-RSRP, SS-RSRQ, SS-SINR as defined in TS 38-215), SSB indexes, Physical Cell Identifiers (PCIs), information obtained from system information, etc.; another example is CSI-RS-related measurements; • An indication of the duration for which the request is valid. Using such an indication, the UE can avoid sending multiple requests during this duration (if the network wants to accept / reject the request for a shorter duration than the duration requested by the UE).

[0111] Figure 6 and Figure 7 Different examples of signaling according to embodiments of this disclosure are shown. This instruction may correspond to a request to the network node to perform UE-side model training (step 1610), and the UE may receive an acceptance instruction (such as...) as a response. Figure 6 Step 1620 shown) or rejection instruction (such as Figure 7 Step 1720 shown in the figure.

[0112] In one option, the UE receives a response with an acceptance indication, and if the UE request is accepted, the indication specifies the RRC configuration to be used for the training duration, wherein the RRC configuration may include CSI-RS / SSB resources to be used for measurement and data collection, DRX configuration, radio bearer reconfiguration, MIMO configuration, frequencies on which data collection is permitted, duration of data collection, and time point at which data collection may begin.

[0113] In one option, if the UE request is accepted, the indication specifies the RRC configuration to be used for the training duration, wherein the RRC configuration may include CSI-RS / SSB resources to be used for measurement and data collection, DRX configuration, radio bearer reconfiguration, MIMO configuration, frequencies on which data collection is permitted, duration of data collection, and the point in time when data collection may begin.

[0114] In one option, if the request is rejected, the reason indicates the reason for the rejection, such as network overload, lack of radio resources, or high-priority service being performed at the UE.

[0115] In one option, the accept or reject indication is a set of indications for multiple training requests received from the UE, with each entry specifying the network's response to each training request. Each training request can be identified by a specific identifier (ID).

[0116] In one option, an acceptance or rejection indication is provided in the DL MAC CE or PHY layer signaling from the network node to the UE. This option enables the training request / response process to be operated more frequently based on the L1 operational state at a lower layer.

[0117] One option provides an acceptance or rejection indication in the RRC message. Compared to MAC / PHY-based response options, this option allows the network to make the training request / response process slower, and thus potentially reduce the need for frequent communication between the UE and network nodes.

[0118] In one option, the UE receives multiple accept / reject decisions for a single request it has transmitted. This type of implementation is particularly useful when the UE includes an indication of the duration for which it intends to perform training. Based on such a 'duration' indication in the request from the UE, the network node can respond with accept / reject multiple times until such a timer expires.

[0119] In one option, the UE receives an accept / reject decision from the network node, but such an accept / reject decision comes with a validity duration that indicates how long such an accept / reject decision is valid.

[0120] In one option, when the response indicates a rejection, the response message contains a timer value, based on which the UE starts a timer (set to the received value) and: i) while the timer is running, the UE is not allowed to send another indication that UE-side model training needs to be performed; ii) when the timer expires, the UE is allowed to send another indication that UE-side model training needs to be performed; iii) the timer is stopped under one or more conditions such as entering RRC_IDLE or RRC_INACTIVE.

[0121] In one option, when the response indicates acceptance, the response message contains a timer value, based on which the UE starts a timer (set to the received value) and: i) while the timer is running, the UE is allowed to perform UE-side model training; ii) when the timer expires and the UE-side model training is not completed, the UE is allowed to send another instruction that UE-side model training needs to be performed; iii) the timer is stopped under one or more conditions such as training completion.

[0122] In response to receiving an instruction from a network node to accept a UE request, the UE performs data collection for UE-side model training.

[0123] In response to receiving an indication from a network node to reject a UE request, the UE does not perform data collection for UE-side model training.

[0124] In one option, the indication to deny the request includes a timer value, based on which the UE initiates a first time (using the specified value). While the timer is running, the UE does not transmit another request. After the timer expires, the UE is allowed to transmit another request for data collection for UE-side model training. In some such embodiments, the UE stops the timer once an event such as a handover occurs, or upon receiving a message from the network indicating that the UE is allowed to perform data collection and / or model training for AI / ML model training.

[0125] In some embodiments, in response to a first message from a first entity containing an indication that UE-side model training needs to be performed (e.g., at a lower layer, PHY / MAC, or at a higher layer of the RAN, e.g., for RRM or Layer 3 mobility), the UE triggers a transmission to a network node indicating that data collection for model training needs to be performed.

[0126] The first entity may include one or more higher layers 50 of the UE (e.g., the UE application layer), such as Figure 8 As shown in the diagram. For example, a higher layer 50 of the UE may be responsible for performing UE-side model training. In this case, there may be internal communication within the UE, where an OTT client indicates a need for data collection within the UE (e.g., to the UE RAN layer, such as the PHY / MAC / RRC layer), causing the UE to trigger a transmission of the data collection requirement instruction (e.g., by one or more lower layers 55 of the UE) to the network node (step 1810).

[0127] In another embodiment, if the OTT server is responsible for performing UE-side model training, then the first entity may be the OTT server (e.g., such as...). Figure 9 (As shown in the example).

[0128] In another embodiment, if the core network node is responsible for performing UE-side model training, the first entity may be such a core network node, for example, NWDAF. In a later embodiment, the instruction that UE-side model training needs to be performed at a lower layer 55 is signaled to the UE via NAS signaling.

[0129] In one embodiment, the UE performing UE-side model training implies that a lower layer 55 performs data collection (e.g., for CSI / beam prediction or localization) or a higher layer RAN, such as for RRM or layer 3 mobility.

[0130] In one embodiment, before transmitting a request / instruction, the UE may determine whether certain conditions for transmitting the request / instruction can be met. For example, the conditions may be: Whether an instruction to start or resume data collection for UE-side model training has been received from the first entity (OTT server, core network node, UE application layer). The first entity may provide this instruction to the UE in response to any of the following: • A UE has entered a certain area (e.g., a geographic area, or one or more cells controlled by a gNB) and needs another UE-side model training session in that area. For example, a first entity does not have a trained model available for the area of ​​interest. Therefore, the first entity may request the UE to begin data collection for UE-side model training. The first entity can learn about the area from the location information provided to it by the UE. • The UE has entered a certain area (e.g., a geographical area, or a cell or multiple cells controlled by a gNB) where UE-side model training is permitted, for example, the target gNB supports providing configurations for UE-side model training. Therefore, the first entity can request the UE to initiate data collection for UE-side model training. • Whether the UE has been configured with the necessary resources to perform training. For example, if the UE needs to perform data collection on a frequency or certain CSI-RS / SSB resources on which it is already configured to perform measurements, the UE may not send a request because it is already capable of performing data collection for UE-side model training. In another embodiment, if the UE is not configured with the necessary CSI-RS / SSB resources or frequencies for performing training as instructed by the first entity, the UE may send a request to the gNB. • If the UE begins data collection for UE-side model training, is a reduction in capabilities expected? For example, in some cases, the impact on normal operation can be expected in order to perform data collection for UE-side model training. For instance, the UE may need to perform relaxed measurements on certain frequencies, or it may need a different antenna configuration (reduced MIMO layer), or it may need a different DRX configuration, or it may need to be configured with a different number of serving cells, or it may need to have MR-DC deactivated. • Whether there is sufficient battery power remaining. For example, if the UE has limited battery power remaining, the lower layer 55 of the UE may reject the request from the first entity to begin data collection for UE-side model training. Otherwise, the UE may send a request to the gNB to perform UE-side model training. • Whether the UE is performing certain user plane or control plane operations that prevent the UE from starting UE-side model training. For example, when the UE is configured with certain high-priority radio bearers, or when the UE may be performing a PCell HO, or performing an RRC reconstruction procedure, or performing a fast MCG link recovery procedure, or performing an SCG fault report, it may receive an instruction from a first entity indicating that UE-side model training needs to be performed at a lower layer 55. • Does RAN node 120 provide a message indicating that AIML training operations are supported in the cell, and that one or more specific sets of UEs can request resources for UE-side model training? This message could be provided, for example, in SIB signaling. For instance, this message could also include an indication of a specific training session that training can begin / resume, where each training session could be associated with a specific set of training resource configurations or a specific model training (in this case, the indication of a specific data collection could be a model ID), or associated with a specific AIML functionality (such as beam management, CSI prediction, positioning prediction, etc.) (in this case, the indication of a specific data collection could be a functionality ID). This message could also include an indication of a specific UE vendor (UE set) to which training is permitted. In such cases, all UEs from a specific UE vendor (i.e., the same UE set) are permitted to use for UE-side model training in that cell. • Does RAN node 120 provide a message indicating that AIML training operations are supported for this specific UE? For example, the UE may indicate to RAN node 120 that it is capable of AIML training. RAN node 120 can then provide a dedicated RRC message indicating whether the UE can request resources for UE-side model training. For example, this message may also include an indication of a specific training session that can be started / resumed, where each training session may be associated with a specific set of training resource configurations or a specific model training (in this case, the indication of a specific data collection could be a model ID), or associated with a specific AIML functionality (such as beam management, CSI prediction, positioning prediction, etc.) (in this case, the indication of a specific data collection could be a functionality ID). This message may also include an indication associated with a specific UE vendor that is permitted to train with. In such cases, all UEs from a specific UE vendor are permitted to use for UE-side model training in that cell. • Whether the UE is within the appropriate / expected coverage for AI / ML training operations. For example, if the measured RSRP / RSRQ / SINR / RSSI is determined to be below a certain threshold (which may be configured), the UE should not start UE-side model training, or if certain timers (such as T310 / T312 / T304) are running, the UE should not start data collection for UE-side model training. • Whether the UE is connected to the PLMN in which the UE may perform data collection for UE-side model training. • Whether the UE is connected to it. The UE may perform radio access technology (e.g., NR) for data collection for UE-side model training.

[0131] Once the UE has begun collecting data for model training, the UE monitors for one or more indications from the network, such as... Figure 10 As shown in the diagram. In some such embodiments, the UE may receive an indication from a network node (e.g., RAN node 120, e.g., gNB) in a first message (e.g., under RAN overload conditions) to stop / pause initiated data collection (step 1530). This indication may include multiple indications, each indicating to the UE to stop / pause each of the AI / ML training operations in progress at the UE. Each indication may identify the AI / ML training operation by an identifier.

[0132] In some embodiments, in response to receiving a stop / pause data collection instruction from RAN node 120 in a first message, the UE may stop performing data collection for UE-side model training (step 1540). Upon receiving multiple stop / pause instructions for multiple training operations at the UE, the UE may apply the stop / pause to data collection for each specified AI / ML training model.

[0133] In some embodiments, the UE may transmit an indication to the first entity in a second message that data collection for UE-side model training has been stopped / paused, for example, such as Figure 11 As shown in (step 1560).

[0134] In some embodiments, the UE may receive an instruction from RAN node 120 (e.g., gNB) in a third message to resume stopped / paused data collection. This instruction may contain multiple instructions, each instructing the UE to resume each of the training operations that were stopped / paused at the UE. Each instruction may be identified by an identifier for the AI / ML training operation.

[0135] In some embodiments, in response to receiving an instruction to resume data collection from RAN node 120 in a third message, the UE may resume performing data collection at a lower layer 55 (e.g., PHY / MAC) for UE-side model training. Upon receiving multiple resumption instructions for multiple stopped / paused training operations at the UE, the UE may resume data collection for each of the specified training models.

[0136] In some embodiments, the UE may transmit an indication to the first entity in a fourth message that data collection for UE-side model training has been resumed.

[0137] In view of the above, the embodiments include methods performed by the UE, such as... Figure 11 As shown in the diagram. The method includes transmitting an indication to RAN node 120 (e.g., gNB) in a first message that data collection for UE-side model training has been stopped. The transmission of the first message may be in response to receiving an indication from a first entity to stop data collection for UE-side model training. Whether the UE is within suitable / expected coverage for AI / ML training operations. For example, if it is determined that the measured RSRP / RSRQ / SINR / RSSI is below a certain threshold (which may be configured), the UE stops UE-side model training, or if certain timers (such as T310 / T312 / T304) are running, model training stops.

[0138] The method may further include stopping the data collection at a lower layer 55 (e.g., PHY / MAC) for UE-side model training.

[0139] The method may further include transmitting an indication to the first entity in a second message that data collection for UE-side model training has been stopped.

[0140] The method may further include determining data collection that can be recovered for UE-side model training.

[0141] The method may further include transmitting in a third message an instruction to RAN node 120 (e.g., gNB) that data collection for UE-side model training can be resumed.

[0142] The method may further include receiving, in a fourth message, an instruction from RAN node 120 (e.g., gNB) to resume or not resume data collection that has been stopped.

[0143] The method may further include resuming data collection at a lower layer 55 (e.g., PHY / MAC) for UE-side model training in response to receiving an instruction to resume data collection from RAN node 120 in a fourth message.

[0144] The method may further include, in response to receiving an indication of data collection from RAN node 120 in a fourth message, transmitting to the first entity in a fifth message an indication of whether data collection for UE-side model training has been resumed or not, based on the response in the fourth message.

[0145] The method may further include receiving, in a first message, an indication from a network node (e.g., RAN node 120, e.g., gNB) to stop / pause initiated data collection (e.g., under RAN overload conditions). This indication may include multiple indications, each instructing the UE to stop / pause each operation within the ongoing AI / ML training operation at the UE. Each indication may specify the AI / ML training operation via an identifier.

[0146] Other embodiments include another method performed by the UE. This method includes stopping data collection for UE-side model training in response to receiving an instruction to stop / pause data collection for UE-side model training from a first entity in a first message. In the case of receiving multiple stop / pause instructions for multiple training operations at the UE, the UE may apply the stop / pause to data collection for each specified AI / ML training model.

[0147] The first entity may transmit the first message in response to one or more of the following conditions being met in the first entity: • A UE has entered a certain area (e.g., a geographical area, or a cell or multiple cells controlled by a gNB) where another UE-side model training session is needed, for example, if the first entity does not have a trained model available for the area of ​​interest. Therefore, the first entity may request the UE to stop the current data collection session used for UE-side model training and may start a new data collection session for training another UE-side model. • The UE has entered a certain area (e.g., a geographical area, or a cell or multiple cells controlled by a gNB) where the UE-side model training session that has been started is not valid, for example, the results of the UE-side model training are not applicable to such an area. • The UE has entered a certain area (e.g., a geographical area, or a cell or multiple cells controlled by a gNB) where the resources (CSI-RS / SSB, real data) necessary for the UE to perform the UE-side model training that has already begun are unavailable. • The UE has entered an area (e.g., a geographical area, or a cell or multiple cells controlled by a gNB) where UE-side model training is not permitted, for example, the target gNB does not support providing configurations for UE-side model training. Therefore, the first entity may request the UE to stop data collection for UE-side model training.

[0148] The method may further include transmitting an indication to the network node in a second message that data collection for UE-side model training has been stopped / paused.

[0149] The method may further include receiving an instruction from the first entity in a third message to resume stopped / suspended data collection. This instruction may include multiple instructions, each instructing the UE to resume each operation within the stopped / suspended AI / ML training operations at the UE. Each instruction may identify the AI / ML training operation via an identifier.

[0150] The method may further include, in response to receiving an instruction from a first entity in a third message to resume data collection for UE-side model training, transmitting an instruction to the network node in a fourth message that data collection for UE-side model training can be resumed. In the case of receiving multiple resumption instructions for multiple stopped / paused training operations at the UE, the UE may resume data collection for each of the specified AI / ML training models.

[0151] The method may further include receiving an instruction from the network node in the fifth message to resume stopped / suspended data collection.

[0152] The method may further include resuming data collection performed at a lower layer 55 (e.g., PHY / MAC / RRC) for UE-side model training. In the event of receiving multiple resumption instructions for multiple stopped / paused training operations at the UE, the UE may resume data collection for each of the specified training models.

[0153] The method may further include transmitting an indication to the first entity in a sixth message that data collection for UE-side model training has been resumed.

[0154] Other embodiments include a further method performed by the UE. This method includes clearing stored collected data associated with a UE-side model training session when data collection for that UE-side model training session is stopped. Alternatively, the method includes retaining stored collected data associated with a UE-side model training session when data collection for that UE-side model training session is stopped.

[0155] Correspondingly, other methods include those performed by RAN node 120. This method includes receiving a request from the UE in the first message for UE-side model training to be performed.

[0156] The method may further include, in response to receiving a request in a first message, determining whether to accept or reject the UE request. The determination may depend on any of the following: • Load conditions in the community; • The UE needs the availability of radio resources (e.g., CSI-RS, SSB resources) to perform UE-side model training; • The number of UEs in the cell that have already undergone UE-side model training; • Whether the UE has been configured with the resources necessary to perform training. For example, if the UE needs to perform data collection on a frequency or certain CSI-RS / SSB resources on which it is configured to perform measurements, the RAN node 120 may accept a UE request to perform data collection for UE-side model training. In another embodiment, if the UE is not configured with the necessary CSI-RS / SSB resources or frequencies for performing training as indicated by the first entity, the RAN node 120 may reject the request. • If the UE begins data collection for UE-side model training, is a reduction in capabilities expected? For example, in some cases, the impact on normal operation can be expected in order to perform data collection for UE-side model training. For instance, the UE may need to perform lenient measurements on certain frequencies, or it may need a different antenna configuration (reduced MIMO layer), or it may need a different DRX configuration, or it may need to be configured with a different number of serving cells, or it may need to have MR-DC deactivated. • Whether the UE is performing certain user plane or control plane operations that prevent the UE from starting UE-side model training, for example, being configured with a higher priority DRB or SRB; • This depends on the UE's radio conditions. For example, when a UE is near the cell edge or the most recent RRM measurement report indicates poor UE coverage, it may receive a UE request to perform UE-side model training. In such cases, the UE-side model training request can be rejected.

[0157] In some embodiments, the method further includes transmitting an indication to the UE in a second message whether the UE request is accepted or rejected.

[0158] In some embodiments, the method further includes determining that the UE should stop the data collection that has been initiated for UE-side model training. The determination may depend on any of the following: • Load conditions in the community; • The UE needs the availability of radio resources (e.g., CSI-RS, SSB resources) to perform UE-side model training; • The number of UEs in the cell that have already undergone UE-side model training; • Does the UE already need to be configured with different CSI-RS / SSB resources? • If the UE continues to be used for data collection for UE-side model training, is a reduction in capability expected? • Whether the UE is configured with certain user plane or control plane operations that prevent the UE from continuing UE-side model training, for example, being configured with higher priority DRB or SRB; • This depends on the UE's radio conditions. For example, if the network determines that the UE is near the cell edge or the most recent RRM measurement report indicates poor UE coverage, the network may request the UE to stop training.

[0159] The method may further include transmitting an instruction to the UE to stop the data collection that has already begun.

[0160] The method may further include determining that the UE should resume data collection for UE-side model training that has been stopped, wherein the conditions for resuming the stopped data collection may be the same as the conditions for starting data collection.

[0161] The method may further include transmitting an instruction to the UE to resume the data collection that has already begun.

[0162] The method may further include receiving an indication from the UE that data collection for UE-side model training has been stopped.

[0163] The method may further include receiving instructions from the UE that can recover data collection used for UE-side model training.

[0164] The method may further include transmitting to the UE an indication of whether the UE can resume the data collection that has already begun.

[0165] Further network node methods may include providing a message indicating that model training operations are supported in the cell and that one or more specific sets of UEs can request resources for UE-side model training. This message may be provided, for example, in SIB signaling. For instance, the message may also include an indication of a specific training session that training can begin / resume, where each training session may be associated with a specific set of training resource configurations or with a specific model training (in which case, the indication of a specific data collection could be a model ID), or with a specific AIML functionality (such as beam management, CSI prediction, location prediction, etc.) (in which case, the indication of a specific data collection could be a functionality ID). The message may also include an indication associated with a specific UE vendor (UE set) that is permitted to train. In such cases, all UEs from a specific UE vendor (i.e., the same UE set) are permitted to use for UE-side model training in the cell.

[0166] The network node method may additionally or alternatively include providing a message indicating that model training operations are supported for this particular UE. For example, the UE may indicate to RAN node 120 that it is capable of model training. RAN node 120 may then provide a dedicated RRC message indicating whether the UE can request resources for UE-side model training. For example, this message may also contain an indication of a specific training session that can be started / resumed, where each training session may be associated with a specific set of training resource configurations or with a specific model training (in this case, the indication of a specific data collection may be a model ID), or with a specific AIML functionality (such as beam management, CSI prediction, location prediction, etc.) (in this case, the indication of a specific data collection may be a functionality ID). This message may also contain an indication of a specific UE vendor that is permitted to train with. In such cases, all UEs from a specific UE vendor are permitted to use for UE-side model training within the cell.

[0167] In view of the above, Figure 12 This is a flowchart illustrating an example method 200 implemented by UE 110. Method 200 includes transmitting an instruction to network node 120 indicating that UE 110 needs to perform UE-side model training (box 210). In some embodiments, method 200 further includes performing data collection for UE-side model training (box 220).

[0168] Figure 13 This is a flowchart illustrating another example method 300 implemented by UE 110. Method 300 may be appended to or performed in place of method 200. Method 300 includes receiving a first message from a first entity, the first message indicating that UE-side model training needs to be performed, for example, for one or more lower-layer functions and / or one or more higher-layer functions (box 210). The one or more lower-layer functions may include physical PHY and / or media access control (MAC) layer functions. In some embodiments, method 300 further includes performing data collection for UE-side model training (box 220).

[0169] Figure 14 This is a flowchart illustrating another example method 400 implemented by UE 110. Method 300 may be appended to or replace any of methods 200 and 300, or both of these methods, for execution. Method 400 includes performing data collection for UE-side model training (box 410). Method 400 further includes receiving a first message from RAN node 120 indicating the cessation of data collection (box 420).

[0170] Figure 15This is a flowchart illustrating another example method 500 implemented by UE 110. Method 300 may be appended to or replace any one or more of methods 200, 300, and 400. Method 500 includes stopping data collection for UE-side model training (box 510). Method 500 further includes transmitting a first message to RAN node 120 indicating that data collection for UE-side model training has been stopped (box 520).

[0171] Figure 16 This is a flowchart illustrating an example method 800 implemented by network node 190. Method 800 includes receiving an instruction from UE 110 indicating that UE 110 needs to perform UE-side model training (block 810). Method 800 further includes, in response to receiving the instruction, transmitting a response to UE 110 indicating that UE 110 is permitted to perform data collection for UE-side model training, or should perform data collection for UE-side model training (block 820). The response includes accepting the instruction and / or providing configuration for UE 110 to use for data collection.

[0172] Figure 17 This is a flowchart illustrating another example method 900 implemented by network node 190. Method 900 may be appended to or performed in place of method 800. Method 900 includes receiving a request from UE 110 to perform data collection for UE-side model training (box 910). Method 900 further includes determining whether to accept or reject the request (box 920). Method 900 further includes indicating to UE 110 whether the request is accepted or rejected (box 930).

[0173] For example, UE 110 can be as follows Figure 18 The example illustrates the implementation. Figure 18The UE 110 includes processing circuitry 610, memory circuitry 620, and interface circuitry 630. Processing circuitry 610 is communicatively coupled to memory circuitry 620 and interface circuitry 630, for example, via bus 604. Processing circuitry 610 may include one or more microprocessors, microcontrollers, hardware circuitry, discrete logic circuitry, hardware registers, digital signal processors (DSPs), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), or combinations thereof. For example, processing circuitry 610 may be programmable hardware capable of executing software instructions, such as a machine-readable computer program 640, stored in memory circuitry 620. The memory circuitry 620 in various embodiments may include any non-transitory machine-readable medium known or developed in the art, whether volatile or non-volatile, including but not limited to solid-state media (e.g., SRAM, DRAM, DDRAM, ROM, PROM, EPROM, flash memory, solid-state drives, etc.), removable storage devices (e.g., Secure Digital (SD) cards, miniSD cards, microSD cards, Memory Sticks, thumb drives, USB flash drives, ROM cartridges, universal media disks), fixed drives (e.g., magnetic hard disk drives), and so on (all or in any combination).

[0174] Interface circuitry 630 may be a controller hub configured to control the input and output (I / O) data paths of UE 110. Such I / O data paths may include data paths for exchanging signals over a network. Interface circuitry 630 may be implemented as a single physical component or as multiple physical components arranged adjacently or separately, wherein any component may be communicatively coupled to or communicate with any other component via processing circuitry 610. For example, interface circuitry 630 may include a transmitter 632 configured to transmit wireless communication signals and a receiver 634 configured to receive wireless communication signals.

[0175] UE 110 can be configured to perform any one or more of the UE methods 200, 300, 400, and 500 described above. In one example, memory 620 contains instructions that can be executed by processing circuitry 610, thereby configuring UE 110.

[0176] Other embodiments include a control program 640 that includes instructions that, when executed on the processing circuitry 610 of the UE 110, cause the UE 110 to perform any of the methods described in the UE methods 200, 300, 400, and 500.

[0177] Other embodiments include a carrier containing control program 640. This carrier is one of an electronic signal, an optical signal, a radio signal, or a computer-readable storage medium.

[0178] Correspondingly, network node 190 can be as follows Figure 19 The example illustrates the implementation. Figure 19 Network node 190 includes processing circuitry 710, memory circuitry 720, and interface circuitry 730. Processing circuitry 710 is communicatively coupled to memory circuitry 720 and interface circuitry 730, for example, via bus 704. Processing circuitry 710 may include one or more microprocessors, microcontrollers, hardware circuitry, discrete logic circuitry, hardware registers, digital signal processors (DSPs), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), or combinations thereof. For example, processing circuitry 710 may be programmable hardware capable of executing software instructions, such as a machine-readable computer program 740, stored in memory circuitry 720. The memory circuitry 720 in various embodiments may include any non-transitory machine-readable medium known or developed in the art, whether volatile or non-volatile, including but not limited to solid-state media (e.g., SRAM, DRAM, DDRAM, ROM, PROM, EPROM, flash memory, solid-state drives, etc.), removable storage devices (e.g., Secure Digital (SD) cards, miniSD cards, microSD cards, Memory Sticks, thumb drives, USB flash drives, ROM cartridges, universal media disks), fixed drives (e.g., magnetic hard disk drives), and so on (all or in any combination).

[0179] Interface circuitry 730 may be a controller hub configured to control the input and output (I / O) data paths of network node 120. Such I / O data paths may include data paths for exchanging signals over the network. Interface circuitry 730 may be implemented as a single physical component or as multiple physical components arranged adjacently or separately, wherein any component may be communicatively coupled to or communicate with any other component via processing circuitry 710. For example, interface circuitry 730 may include a transmitter 732 configured to transmit wireless communication signals and a receiver 734 configured to receive wireless communication signals.

[0180] Network node 190 can be configured to perform the above-described method 300. According to a specific embodiment, processing circuitry 710 is configured to perform any of the methods described in the above-described network node method.

[0181] Other embodiments include a control program 740 that includes instructions that, when executed on the processing circuitry 710 of the network node 120, cause the network node 120 to perform any of the methods described above in the network node method.

[0182] Other embodiments include a carrier containing control program 740. This carrier is one of an electronic signal, an optical signal, a radio signal, or a computer-readable storage medium.

[0183] While the various communication devices described herein may include the illustrated combinations of hardware components, other embodiments may include computing and / or communication hardware having different combinations of components. It should be understood that these computing devices may include any suitable combination of hardware and / or software required to perform the tasks, features, functions, and methods disclosed herein. Furthermore, although components are depicted as single boxes located within larger boxes or nested within multiple boxes, in practice, the devices described herein may include multiple different physical components constituting a single illustrated component, and functionality may be partitioned among the separate components.

[0184] While the computing devices described herein (e.g., UE, network node, host) may include the illustrated combinations of hardware components, other embodiments may include computing devices having different combinations of components. It should be understood that these computing devices may include any suitable combination of hardware and / or software required to perform the tasks, features, functions, and methods disclosed herein. The determination, calculation, acquisition, or similar operations described herein may be performed by processing circuitry that processes information by, for example, converting acquired information into other information, comparing the acquired or converted information with information stored in a network node, and / or performing one or more operations based on the acquired or converted information, and making a determination as a result of said processing. Furthermore, although components are depicted as single boxes located within larger boxes or nested within multiple boxes, in practice, computing devices may include multiple different physical components constituting a single illustrated component, and functionality may be partitioned between separate components. For example, a communication interface may be configured to include any of the components described herein, and / or the functionality of the components may be partitioned between processing circuitry and the communication interface. In another example, non-computationally intensive functions of any component in such a component may be implemented in software or firmware, while computationally intensive functions may be implemented in hardware.

Claims

1. A method (200) implemented by a user equipment (UE) (110), the method comprising: Transmit (210) an instruction to the network node (190) indicating that the UE (110) needs to perform UE-side model training; as well as Perform (220) data collection for training the UE-side model.

2. The method of claim 1, wherein: The instruction includes a request to perform the UE-side model training; as well as The data collection is performed in response to receiving the instruction from the network node (190); The response includes configuration and / or acceptance instructions for the UE (110) to use the data collection.

3. The method according to any one of claims 1-2, further comprising: Receive a notification from the network node (190) or a different network node to stop or suspend the data collection for the UE-side model; as well as Further instructions indicating that the data collection used for training the UE-side model have been stopped or paused are transmitted to the network node (190) or the different network nodes.

4. The method as described in any one of claims 1-3, further comprising notifying the network node (190) or different network nodes of the data collection for the UE-side model training: It has been completed; or It needs to be restored and / or restarted and / or reconfigured.

5. A user equipment (UE) (110), comprising: The processing circuit (610) and memory (620), the memory (620) containing instructions executable by the processing circuit (610), thereby configuring the UE (110) to: Transmit to network node (190) an instruction indicating that the UE (110) needs to perform UE-side model training; and Perform data collection for training the UE-side model.

6. The UE as described in the preceding claims is further configured to perform the method (200) as described in any one of claims 2-4.

7. A method (300) performed by a user equipment (UE) (110), the method comprising: Receive (310) a first message from the first entity, the first message indicating that UE-side model training needs to be performed for one or more lower-level functions and / or one or more higher-level functions; The one or more lower-layer functions include physical PHY and / or media access control (MAC) layer functions.

8. The method of claim 7, wherein: The one or more lower-level functions include beam management, providing channel state information (CSI) and / or positioning; and The one or more higher-level functions include Radio Resource Management (RRM) measurement and / or L3 mobility functions.

9. The method of any one of claims 7-8, further comprising: In response to receiving the first message, a second message is sent to the radio access network (RAN) node requesting the execution of the UE-side model training. A third message is received from the RAN node (120), the third message indicating whether the UE (110) is allowed to perform the UE-side model training; In response to the third message indicating permission for the UE (110) to perform the UE-side model training, data collection for the UE-side model training is initiated at one or more lower layers (55); as well as In response to receiving the third message, a fourth message is sent to the first entity, the fourth message indicating whether the request has been accepted or rejected.

10. The method of any one of claims 7-9, further comprising informing the first entity that data collection has commenced.

11. A user equipment (UE) (110), comprising: The processing circuit (610) and memory (620) contain instructions that can be executed by the processing circuit (610), thereby configuring the UE (110) to receive a first message from a first entity, the first message indicating that UE-side model training needs to be performed for one or more lower-level functions and / or one or more higher-level functions; The one or more lower-layer functions include physical PHY and / or media access control (MAC) layer functions.

12. The UE as described in the preceding claims is further configured to perform the method (300) as described in any one of claims 8-10.

13. A method (400) implemented by a user equipment (UE) (110), the method comprising: Perform (410) data collection for UE-side model training; as well as Receive (420) a first message from the radio access network RAN ​​node indicating the cessation of the data collection.

14. The method of claim 13, further comprising: In response to receiving the first message, the data collection for UE-side model training in one or more lower layers (55) is stopped, wherein the one or more lower layers (55) include the physical PHY layer and / or the media access control (MAC) layer.

15. The method of any one of claims 13-14, further comprising: A second message is sent to the first entity indicating that the data collection used for UE-side model training has been stopped.

16. The method of any one of claims 13-15, further comprising: Receive a third message from the RAN node (120) indicating the resumption of stopped data collection; as well as In response to receiving the third message, the data collection for UE-side model training is resumed in one or more lower layers (55), wherein the one or more lower layers (55) include a PHY layer and / or a MAC layer.

17. The method of any one of claims 13-16, further comprising: A fourth message is sent to the first entity indicating that the data collection used for UE-side model training has been resumed.

18. A user equipment (UE) (110), comprising: The processing circuit (610) and memory (620), the memory (620) containing instructions executable by the processing circuit (610), thereby configuring the UE (110) to: Perform data collection for UE-side model training; and Receive a first message from the radio access network (RAN) node indicating the cessation of the data collection.

19. The UE as described in the preceding claims is further configured to perform the method (400) as described in any one of claims 14-17.

20. A method (500) implemented by a user equipment (UE) (110), the method comprising: Stop (510) data collection for UE-side model training; as well as A first message (520) indicating that the data collection for the UE-side model training has been stopped is transmitted to the radio access network RAN ​​node (120).

21. The method of claim 20, wherein: Stopping the data collection used for training the UE-side model includes stopping the data collection at one or more lower layers (55); as well as The one or more lower layers (55) include the physical PHY layer and / or the media access control (MAC) layer.

22. The method of any one of claims 20-21, further comprising: A second message is sent to the first entity indicating that the data collection used for training the UE-side model has been stopped.

23. The method of any one of claims 20-22, further comprising: Determine that the data collection used for training the UE-side model can be resumed; as well as A third message is transmitted to the RAN node (120) indicating that the data collection used for training the UE-side model can be resumed.

24. The method of any one of claims 20-23, further comprising: Receive a fourth message from the RAN node (120) indicating whether to resume the data collection; In response to the fourth message indicating the resumption of the data collection, the data collection used for UE-side model training is resumed at one or more lower layers (55); as well as A fifth message is transmitted to the first entity, the fifth message indicating whether the data collection for training the UE-side model has been resumed according to the fourth message.

25. A user equipment (UE) (110), comprising: The processing circuit (610) and memory (620), the memory (620) containing instructions executable by the processing circuit (610), thereby configuring the UE (110) to: Stop collecting data for UE-side model training; and A first message is transmitted to the radio access network RAN ​​node (120) indicating that the data collection used for the training of the UE-side model has been stopped.

26. The UE as described in the preceding claims is further configured to perform the method as described in any one of claims 21-24.

27. A computer program (640) comprising instructions that, when executed on a processing circuit (610) of a user equipment (UE) (110), cause the processing circuit (610) to perform the method as claimed in any one of claims 1-4, 7-10, 13-17 or 20-24.

28. A method (800) implemented by a network node (190), the method comprising: Receive (810) an instruction from the user equipment (UE) (110) indicating that the UE (110) needs to perform UE-side model training; as well as In response to receiving the instruction, a response (820) is transmitted to the UE (110), the response indicating the UE (110): The data collection for training the UE-side model is permitted; or The data collection for training the model on the UE side should be performed; The response includes accepting instructions and / or providing the UE (110) with configurations for the data collection.

29. The method of claim 28, wherein: The instruction includes a request from the UE (110) to perform model training on the UE side; and The method further includes determining whether to accept or reject the request.

30. The method of any one of claims 28-29, further comprising receiving from the UE (110): Stop or pause the request for data collection used for training the UE-side model; or The UE (110) has stopped or suspended the notification of data collection for model training on the UE side; or Instructions are required to resume and / or restart and / or reconfigure the data collection used for training the UE-side model.

31. The method of any one of claims 28-30, further comprising: Send an instruction to the UE (110) to stop or pause the data collection for the UE-side model.

32. A network node (190), comprising: A processing circuit (710) and a memory (720) containing instructions executable by the processing circuit (720) are provided, thereby configuring the network node (190) to: Receive from user equipment (UE) (110) an instruction indicating that the UE (110) needs to perform UE-side model training; and In response to receiving the instruction, a response is transmitted to the UE (110), the response indicating the UE (110): The data collection for training the UE-side model is permitted; or The data collection for training the model on the UE side should be performed; The response includes accepting instructions and / or providing the UE (110) with configurations for the data collection.

33. The network node as described in the preceding claims is further configured to perform the method (800) as described in any one of claims 29-31.

34. A method (900) implemented by a network node (190), comprising: Receive (910) a request from the user equipment (UE) (110) to perform data collection for UE-side model training; Determine (920) whether to accept or reject the request; as well as Indicate to the UE (110) whether the request is accepted or rejected.

35. The method of claim 34, further comprising: It is determined that the UE (110) should cease the data collection used for model training on the UE side; Instruct the UE (110) to stop the data collection used for UE-side model training; It is determined that the UE (110) should resume the data collection; as well as Instruct the UE (110) to resume the data collection.

36. The method of any one of claims 34-35, further comprising: Received from the UE (110): A notification indicating that the data collection used for training the model on the UE side has been stopped or paused; and / or The data collection used for UE-side model training is a recoverable notification; as well as Indicate to the UE (110) whether to resume the data collection.

37. A network node (190), comprising: A processing circuit (710) and a memory (720) containing instructions executable by the processing circuit (710) are provided, thereby configuring the network node (190) to: Receive a request from the user equipment (UE) (110) to perform data collection for UE-side model training; Determine whether to accept or reject the request; as well as Indicate to the UE (110) whether the request is accepted or rejected.

38. The network node as described in the preceding claims is further configured to perform the method (900) as described in any one of claims 35-37.

39. A computer program (740) comprising instructions that, when executed on a processing circuit (710) of a network node (190), cause the network node (190) to perform the method (900) as described in any one of claims 28-31 or 34-36.

40. A carrier comprising the computer program as described in claim 27 or 39, wherein, The carrier is one of electronic signals, optical signals, radio signals, or computer-readable storage media.

Images