Ue memory handling due to artificial intelligence operations

Abstract

A network node (170) transmits, to a User Equipment, UE (110), one or more requests to perform a plurality of model training data collection procedures. The UE (110) allocates memory of the UE (110) to one or more model training data collection procedures. The UE (110) collects model training data by performing the one or more model training data collection procedures. The 5 UE (110) transmits the collected model training data to a network node (170). The network node receives, from the UE (110) and responsive to the request, the model training data collected by the UE (110).figure for publication:

Description

[0001] UE MEMORY HANDLING DUE TO ARTIFICIAL INTELLIGENCE OPERATIONS

[0002] RELATED APPLICATIONS

[0003] This application claims priority to U.S. Provisional Patent Application Number 63 / 547,313 filed November s, 2023, the entire contents of which are incorporated by reference herein.

[0004] TECHNICAL FIELD

[0005] The present disclosure generally relates to the technical field of wireless communication and, more particularly, to managing memory used during model training data collection procedures.

[0006] BACKGROUND

[0007] Artificial Intelligence (Al) is a field of computer science that enables computers to mimic cognitive functions traditionally associated with human intelligence in order to solve problems. Machine Learning (ML) is a branch of Al that enables a computer to learn from experience in order to make improved predictions. Al often employs ML as a learning mechanism, e.g., so that a computer can act upon new data without having to be explicitly programmed to do so.

[0008] To enable Al, ML may use a statistical model to identify patterns in a dataset and generate results. Each result may have a corresponding probability of correctness (sometimes referred to as a confidence). A traditional ML approach often involves selecting and preparing a training dataset and applying an ML strategy (e.g., linear regression) to refine the training dataset. Through iterative refinement of the dataset, often with updated data and error checking, data quality generally increases and the model improves over the course of its lifetime. With more and better data, a sound ML model will make increasingly better predictions.

[0009] Al and ML have been investigated, both in academia and industry, as promising tools to improve the design of the air interface in wireless communication networks. Example use cases include using autoencoders for Channel State Information (CSI) compression to reduce the feedback overhead and improve channel prediction accuracy; using deep neural networks for classifying Line-of-Sight (LOS) and Non-LOS (NLOS) conditions to enhance positioning accuracy; using reinforcement learning for beam selection at the network side and / or the UE side to reduce signaling overhead and beam alignment latency; and using deep reinforcement learning to learn a high performance precoding policy for complex Multiple Input Multiple Output (MIMO) precoding problems.

[0010] In 3rd Generation Partnership Project (3GPP) New Radio (NR) standardization work, a new release 18 study item on AI / ML for the NR air interface started in May 2022. This study item is intended to explore the benefits of augmenting the air interface with features enabling improved support of AI / ML based algorithms for enhanced performance, reduced complexity, and / or reduced overhead. Through studying a few selected use cases (CSI feedback, beam management, and positioning), this study item intends to lay the foundation for future air interface use cases leveraging AI / ML techniques.

[0011] SUMMARY

[0012] The present disclosure generally relates to managing memory used by a UE during one or more model training data collection procedures. According to particular embodiments, a UE allocates memory for one or more model training data collection procedures and uses the allocated memory for storing the collected training data. The collected training data is transferred to the network for use in model training.

[0013] Embodiments of the present disclosure include a method implemented by a UE. The method comprises allocating memory of the UE to one or more model training data collection procedures. The method further comprises collecting model training data by performing the one or more model training data collection procedures. The method further comprises transmitting the collected model training data to a network node.

[0014] In some embodiments, allocating the memory comprises allocating a shared amount of memory between a plurality of the model training data collection procedures.

[0015] In some embodiments, allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures associated with a same model and / or model function.

[0016] In some embodiments, allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures that are associated with a same training node.

[0017] In some embodiments, allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures involving data collection in a same range of UE speeds.

[0018] In some embodiments, allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures involving data collection using a same frequency range.

[0019] In some embodiments, allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures involving data collection associated with a same RAN node configuration.

[0020] In some embodiments, allocating the memory comprises allocating a separate respective portion of the memory to each of a plurality of the model training data collection procedures.

[0021] In some embodiments, allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures having different functional purposes.

[0022] In some embodiments, allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures that are associated with different training nodes.

[0023] In some embodiments, allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures involving data collection in different ranges of UE speeds.

[0024] In some embodiments, allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures involving data collection using different frequency ranges.

[0025] In some embodiments, allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures involving data collection associated with different RAN node configurations.

[0026] In some embodiments, allocating the memory comprises allocating the memory exclusively to the model training data collection procedures. The method further comprises allocating a further amount of memory, separate from the amount of memory, to one or more non-model training data collection procedures.

[0027] In some embodiments, allocating the memory comprises allocating a first part of the memory to a first one or more of the model training data collection procedures that support UE-side data collection. Allocating the memory further comprises allocating a second part of the memory to a second one or more of the model training data collection procedures that support network-side data collection.

[0028] In some embodiments, the method further comprises withholding memory from one or more further model training data collection procedures until the collected model training data has been transmitted. The method further comprises notifying the network node of when the one or more further model training data collection procedures is estimated to begin.

[0029] In some embodiments, the method further comprises receiving a request to perform one or more of the model training data collection procedures from the network node. In some embodiments, the method further comprises transmitting a report indicating whether or not the one or more model training data collection procedures can be initiated based on the allocated memory.

[0030] In some embodiments, transmitting the report is responsive to the allocated memory being sufficient to initiate the one or more model training data collection procedures.

[0031] In some embodiments, transmitting the report is responsive to the allocated memory being insufficient to initiate the one or more model training data collection procedures.

[0032] In some embodiments, the report indicates how much of the memory is allocated to the one or more model training data collection procedures.

[0033] In some embodiments, the report indicates how much of the memory is allocated to UE-side model training procedures and / or how much of the memory is allocated to network-side model training procedures.

[0034] In some embodiments, the method further comprises receiving, from the network node, a priority of at least one of the model training data collection procedures, wherein the priority is higher than a further priority of a further model training data collection procedure. The method further comprises refraining from including the further model training data collection procedure in the memory allocation.

[0035] In some embodiments, allocating the memory to the model training data collection procedures comprises allocating less of the memory to a lower priority model training data collection procedure than to a higher priority model training data collection procedure.

[0036] In some embodiments, the method further comprises freeing the memory for the allocation from one or more other model training data collection procedures that have completed.

[0037] Other embodiments include a UE configured to allocate memory of the UE to one or more model training data collection procedures. The UE is further configured to collect model training data by performing the one or more model training data collection procedures. The UE is further configured to transmit the collected model training data to a network node.

[0038] In some embodiments, the UE is further configured to perform any one of the methods described above.

[0039] In some embodiments, the UE comprises interface circuitry, memory, and processing circuitry communicatively connected to the interface circuitry and the memory. The processing circuitry is configured to perform any one of the methods described above.

[0040] Other embodiments include a computer program, comprising instructions which, when executed on processing circuitry of a UE, cause the processing circuitry to carry out any one of the methods described above. Other embodiments include a method implemented by a network node. The method comprises transmitting, to a UE, one or more requests to perform a plurality of model training data collection procedures. The method further comprises receiving, from the UE and responsive to the request, model training data collected by the UE performing the model training data collection procedures.

[0041] In some embodiments, the method further comprises transmitting, to the UE, a further request to perform one or more further model training data collection procedures. The method further comprises receiving, from the UE in response to the further request, notice of when the one or more further model training data collection procedures is estimated to begin.

[0042] In some embodiments, the method further comprises receiving, from the UE, a report indicating that the memory available at the UE is sufficient to initiate the one or more model training data collection procedures. In other embodiments, the method further comprises receiving, from the UE, a report indicating that the memory available at the UE is insufficient to initiate the one or more model training data collection procedures.

[0043] In some embodiments, the report further indicates how much of the memory is allocated to the one or more model training data collection procedures and / or UE-side model training procedures and / or network-side model training procedures.

[0044] In some embodiments, the method further comprises transmitting, to the UE, a priority of at least one of the model training data collection procedures, wherein the priority is higher than a further priority of a further model training data collection procedure. The method further comprises receiving the model training data comprises receiving the model training data before receiving further model training data collected by the UE performing the further model training data collection procedure.

[0045] Other embodiments include a network node configured to transmit, to a UE, one or more requests to perform a plurality of model training data collection procedures. The network node is further configured to receive, from the UE and responsive to the request, model training data collected by the UE performing the model training data collection procedures.

[0046] In some embodiments, the network node is further configured to perform any one of the network node methods described above.

[0047] In some embodiments, the network node comprises interface circuitry and processing circuitry communicatively coupled to the interface circuitry, wherein the processing circuitry is configured to perform any one of the network node methods described above.

[0048] Other embodiments include a computer program, comprising instructions which, when executed on processing circuitry of a network node cause the processing circuitry to carry out any one of the network node methods described above. Yet other embodiments include a carrier containing either of the computer programs described above. The carrier is one of an electronic signal, optical signal, radio signal, or computer readable storage medium.

[0049] BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Aspects of the present disclosure are illustrated by way of example and are not limited by the accompanying figures with like references indicating like elements.

[0051] FIG. 1 is a schematic block diagram illustrating an example of training and inference pipelines within a model LCM procedure, according to one or more embodiments of the present disclosure.

[0052] FIG. 2 is a schematic block diagram illustrating an example framework for studying aspects of an AI / ML model, according to one or more embodiments of the present disclosure.

[0053] FIG. 3 is a schematic block diagram illustrating example of CSI reporting using a two-sided model, according to one or more embodiments of the present disclosure.

[0054] FIG. 4 is a table illustrating an example mapping between functions and entities with regard to UE-side model training scenarios, according to one or more embodiments of the present disclosure.

[0055] FIG. 5 is a table illustrating an example mapping between functions and entities with regard to NW-side model training scenarios, according to one or more embodiments of the present disclosure.

[0056] FIG. 6 is a schematic block diagram illustrating an example wireless communication network, according to one or more embodiments of the present disclosure.

[0057] FIG. 7 is a flow diagram illustrating an example method implemented by a UE, according to one or more embodiments of the present disclosure.

[0058] FIG. 8 is a flow diagram illustrating an example method implemented by a network node, according to one or more embodiments of the present disclosure.

[0059] FIG. 9 is a schematic block diagram illustrating an example UE, according to one or more embodiments of the present disclosure.

[0060] FIG. 10 is a schematic block diagram illustrating an example network node, according to one or more embodiments of the present disclosure.

[0061] DETAILED DESCRIPTION

[0062] As used herein, the term “model” refers to software that has been trained for use in making a prediction from collected input data. The terms “model,” “ML model,” “Al model,” “AI / ML model,” and “Al and / or ML model” should be considered to have equivalent meanings to each other and therefore be interchangeable. As will be discussed in greater detail below, a model may be deployed, implemented, and / or configured in a User Equipment (UE), in a network node, or both.

[0063] Building a model may include several development steps. The actual training of the model may be just one step in a training pipeline. An important part of model development is model LCM. FIG. 1 is an illustration of training and inference pipelines 10, 20 and their interactions within a model LCM procedure 30. The model LCM procedure 30 typically comprises a training pipeline 10 (which may or may not be used for retraining, either partly or wholly), a model deployment stage 16, an inference pipeline 20, and a drift detection stage 25.

[0064] The training pipeline 10 may include data ingestion 11 , data preprocessing 12, model training 13, model evaluation 14, and / or model registration 15 stages.

[0065] Data ingestion 11 refers to gathering raw data (e.g., training data) from data storage. After data ingestion 11 , there may also be a step that controls the validity of the gathered data.

[0066] Data preprocessing 12 refers feature engineering that is applied to the gathered data. For example, data preprocessing 12 may include data normalization and / or data transformation required for the input data to the model.

[0067] Model training 13 refers to the actual model training steps.

[0068] Model evaluation 14 refers to benchmarking model performance to some model baseline. The iterative steps of model training 13 and model evaluation 14 may continue until an acceptable level of performance is achieved.

[0069] Model registration 15 refers to registering the model, including any corresponding metadata that provides information on how the model was developed, and possibly model evaluation performance outcomes.

[0070] The model deployment stage 16 makes the trained (e.g., retrained) model part of the inference pipeline 20.

[0071] The inference pipeline 20 may include data ingestion 21 , data preprocessing 22, model operation 23, and data and / or model monitoring 24 stages.

[0072] Data ingestion 21 refers to gathering raw data (e.g., inference data) from a data storage.

[0073] Data preprocessing 22 for the inference pipeline 20 is substantially similar to the corresponding processing that occurs in the training pipeline 10.

[0074] Model operation 23 refers to using the trained and deployed model in an operational mode.

[0075] Data and model monitoring 24 refers to validating that the inference data are from a distribution that aligns well with the training data, as well as monitoring model outputs for detecting any performance, or operational, drifts.

[0076] The drift detection stage 25 informs about any drifts in the model operations. FIG. 2 illustrates a functional framework for studying model LCM aspects. The framework may, for example, be used for studying different network (NW)-UE collaboration levels for physical layer use cases.

[0077] The models being discussed in the Rel-18 study item on AI / ML for the NR air interface can be categorized into either a one-sided AI / ML model or a two-sided AI / ML model.

[0078] One example of a one-sided AI / ML model is a UE-sided model in which inferences are performed entirely at the UE. Another example of a one-sided AL / ML model is a NW-sided model in which inferences are performed entirely at the NW.

[0079] A two-sided AI / ML model refers to a paired model in which inferences are jointly performed across the UE and the NW. That is, a first part of the inference is performed by a UE and a remaining part of the inference is performed by a network node (e.g., at a next generation Node B (gNB)), or vice versa.

[0080] FIG. 3 shows an example autoencoder (AE)-based two-sided CSI compression use case. In this example, a UE uses an encoder 42 (i.e. , the UE-part of the two-sided AE model 40) operated at a UE to compress measured CSI 41 for a wireless channel. The output of the encoder 42 (i.e., compressed CSI 43) is reported from the UE to a gNB. The gNB uses a decoder 44 (i.e., the NW-part of the two-sided AE model 40) to generate reconstructed CSI 45 for the wireless channel.

[0081] In functionality-based LCM, 3GPP signaling may be used by the network to indicate activation, deactivation, fallback, and / or switching of AI / ML functionality. Such 3GPP signaling may include, e.g., Radio Resource Control (RRC), Medium Access Control (MAC) Control Element (MAC-CE), and / or Downlink Control Information (DCI), among others. In many cases, models may not be identified at the network and the UE may perform model-level LCM. Whether (and how much) awareness the network should have about such model-level LCM currently requires further study within the industry. For functionality identification, there may be one or more functionalities defined within an AI / ML-enabled feature (i.e., a feature in which AI / ML may be used). Notably, a UE may have one AI / ML model for such functionality, or the UE may have multiple AI / ML models for the functionality, depending on the embodiment.

[0082] For AI / ML functionality identification and functionality-based LCM of UE-side models (and / or the UE-part of two-sided models), AI / ML functionality refers to an AI / ML-enabled feature or Feature Group (FG) enabled by one or more configurations, wherein the one or more configurations are supported based on conditions indicated by UE capability. Correspondingly, functionality-based LCM generally operates based on at least one configuration of AI / ML-enabled feature (or FG). After functionality identification, mechanisms for the UE to report one or more identified functionalities may be performed. The identified functionalities may be a subset of all functionalities that are available or configured.

[0083] In model-l D-based LCM, models are identified at the network, and the network or the UE may activate, deactivate, select, and / or switch individual AI / ML models using the model ID.

[0084] For AI / ML model identification and model-ID-based LCM of UE-side models (and / or the UE- part of two-sided models), model-ID-based LCM operates based on identified models, where a model may be associated with specific configurations and / or conditions associated with UE capability of an AI / ML-enabled Feature / FG and additional conditions (e.g., scenarios, sites, and datasets) as determined / identified between UE-side and NW-side.

[0085] From the perspective of the physical layer of the Radio Access Network (RAN) (i.e. , RAN1), an AI / ML model identified by a model ID may be logical, and how it maps to one or more physical AI / ML models may be up to implementation. When distinction is necessary for discussion purposes, companies may use the term “logical AI / ML model” to refer to a model that is identified and assigned a model ID and the term “physical AI / ML model” to refer to an actual implementation of such a model.

[0086] Model identification may be categorized into multiple types. For purposes of this disclosure, a Type A model will refer to a model that is identified to the network (if applicable) and the UE (if applicable) without over-the-air signaling. The model may be assigned with a model ID during model identification, which may be referred to and used in over-the-air signaling after model identification.

[0087] A Type B model will refer to a model that is identified via over-the-air signaling. In particular, a Type B1 model will refer to a model in which model identification is initiated by the UE. The network may assist with one or more remaining steps of the model identification, if any. In contrast, a Type B2 model will refer to a model in which model identification is initiated by the network, and UE may, if applicable, respond for one or more remaining steps of the model identification, if any. Type B models may be assigned with a model ID during model identification, whether they are of Type B1 or B2.

[0088] Once models are identified, the UE can indicate one or more supported AI / ML model IDs for a given AI / ML-enabled feature (or FG) in a UE capability report. Such model identification using a capability report may be performed for any model type (i.e., Type A, Type B1 , Type B2) depending on the embodiment.

[0089] A Model ID may or may not be globally unique, and different types of model IDs may be created for a single model for various LCM purposes. For functionality / model-ID based LCM, once functionalities / models are identified, the same or similar procedures may be used for their activation, deactivation, switching, fallback, and monitoring.

[0090] The way in which the UE-side of a model interacts with other entities may vary, depending on the embodiment. In one approach, the training of a UE-side model is performed at the UE itself. That is, the UE performs both training and inference procedures. However, this approach might be too complex in practice to be feasible given the limited computational resources of the UE and the large computational complexity that training might impose. Also, if models are dependent on location and / or region, a single UE is unlikely to cover an entire coverage area. A model in which the UE trains by itself would be limited to the areas the UE moves around. Accordingly, every time the UE enters a new area, its trained AI / ML models may be outdated.

[0091] As such, alternative approaches for training UE-sided models include embodiments in which a network node (e.g., a radio access node such as a gNB or a core network node such as a Network Data Analytics Function (NWDAF)) collects data from a UE and trains a model that, at some point, is provided to one or more UEs that will apply it. Further, an Over-the-Top (OTT) server, outside of the 3GPP networking environment, may be in charge of performing the training. Such a server could be, for example, a UE-vendor specific server. This latter approach might be reasonable because in order to have high performance, the trained data set should fit the inference operations at the device which may depend on UE-vendor specific implementations (e.g. software / hardware properties / capabilities).

[0092] Irrespective of whether the UE-side model training is performed by a node outside the RAN (e.g., in a core network node) or even outside of the 3GPP network, a certain amount of data typically needs to be collected by a UE in order to enable the training node to perform model training. That is because, for many use cases (e.g., Al-based CSI compression, Al-based CSI prediction, Al-based beam management, Al-based positioning, Al-based mobility predictions, Al- based traffic predictions, etc.), the training node needs to receive inputs from the UE. Accordingly, in some embodiments a UE does training for a certain amount of time, collects data, and once data collection is completed, the UE transfers the collected data to a training node. In some such embodiments, the UE performs this training upon receiving a trigger from the training node. FIG. 4 is a table illustrating an example mapping between functions and entities with regard to UE-side model training scenarios.

[0093] As for NW-side model training, 3GPP has historically assumed that a gNB and / or Operations and Maintenance (OAM) server would be in charge of that. If the gNB is responsible, the gNB may configure the UE with a set of resources (e.g., CSI-Reference Signal (RS) resources or Synchronization Signal Block (SSB) resource sets) in which the UE should collect measurements for a certain amount of time. The UE reports the measurements to the gNB, e.g., via Radio Resource Control (RRC) signaling. The training can then be performed in the gNB itself or in another node controlled by the gNB-vendor (e.g., an OTT server handled by the gNB-vendor).

[0094] A similar approach may apply when the CAM server does the NW-side training. In this case, the CAM may request that the gNB provide, to the UE, a certain configuration according to which the UE should perform certain measurements and collect data. Once the data collection is completed, the UE transfers the collected data to the OAM, e.g., using a Minimization of Drive Test (MDT) framework such as the immediate MDT or the logged MDT. FIG. 5 is a table illustrating an example mapping between functions and entities with regard to NW-side model training scenarios.

[0095] FIG. 6 is a schematic diagram illustrating an example network 100 comprising a RAN node 120, a core network node 140 (e.g., an OAM server), an Over-the-Top (OTT) node 150, and a UE 110. The RAN node 120 serves a cell 130 to the UE 110. The cell 130 supports a Radio Access Technology (RAT) (e.g., NR) that provides the UE 110 with access to a core network 160 that comprises the core network node 140. The core network 160 provides access to the OTT server 150, which may be external to the core network 160.

[0096] As used herein, the term “network node” may refer to a RAN node 120 or a core network node 140. A RAN node 120 as discussed herein may refer to a node of a RAN for any generation of the 3GPP standard (e.g., a gNB, a 6G RAN node). Examples of a core network node 140 include a User Plane Function (UPF), an Access and Mobility Function (AMF), and an OAM server.

[0097] The UE 110 may be required to initiate data collection for one or more Al-based functionalities. Examples of such Al-based functionalities include Al-based CSI compression, Al- based CSI prediction, Al-based beam management, Al-based positioning, Al-based mobility predictions, and Al-based traffic predictions. Further, the list of Al-based functionalities for which a UE may be required to collect data is likely to increase in future 3GPP releases. For each of these use cases, the UE 110 may need to perform training data collection independent of one another since each use case may require the UE 110 to perform measurements on different resource types, monitor different types of Key Performance Indicators (KPIs), and collect different type of measurements.

[0098] When the UE 110 performs data collection, the UE 110 will likely need to store the data collected until the collected data is transferred to the training entity or other entity (e.g., a network node). As used herein, the term “training node” will refer to a network node (e.g., a RAN node 120 or a core network node 140) that trains a model using the collected data.

[0099] For example, in UE-side model training, the UE 110 collects data and then transfers the collected data to the node performing the UE-side training, e.g. the core network node 140, or the OTT server 150 (which may be outside of the 3GPP networking environment). Similarly, for NW- side training, the UE 110 collects data and then transfers the collected data to the training node performing the NW-side training. Additionally, the UE 110 may need to allocate memory for legacy operations, such as for logged MDT.

[0100] All these data collection sessions may pose challenges on how the UE 110 allocates and handles memory. For example, if too much memory is allocated by the UE 110 to perform data collection for UE-side model training, then there may not be sufficient memory available to perform data collection for NW-side model training, thereby limiting the ability of the network to train a NW- side model properly.

[0101] In view of the above, proper handling of UE memory, coordination with the training node, and coordination with the node configuring the resources to perform the training may be advantageous so that training may be performed efficiently, and AI / ML techniques may be more fully exploited.

[0102] Accordingly, embodiments of the present disclosure include a UE 110 that allocates UE 110 memory for data collection purposes (e.g., to collect model training data). In some embodiments, the UE 110 may collect or refrain from data collection based on UE memory availability. In some particular embodiments, the UE 110 may report whether or not it is capable of data collection based on UE memory availability.

[0103] Particular embodiments may allow the UE 110 to determine how to allocate memory for the data collection purposes. In some such embodiments, the UE 110 may allocate memory such that model training data is collected for different model training purposes. Additionally, the training node and / or the node configuring data collection resources may determine whether or not the UE 110 is able to perform such data collection based on UE memory availability.

[0104] As used herein, “data collection” may, for example, refer to one or more procedures performed for any one or more of the following reasons:

[0105] • NW-side model specific training data collection

[0106] • NW-side use case specific training data collection

[0107] • UE-side model specific training data collection

[0108] • UE-side use case specific training data collection

[0109] • Two-sided model specific training data collection

[0110] • Two-sided use case specific training data collection

[0111] • Logged MDT-related data collection

[0112] • Immediate MDT-related data collection

[0113] • Quality of Experience (QoE)-related data collection

[0114] To enable one or more embodiments discussed herein, the UE 110 manages memory allocation associated with more than one measurement logging functionality and communicates with a network node regarding the impact of such memory allocation. The total available memory may be shared between a plurality of the data collection purposes, e.g., as discussed above. For example, UE memory may be shared between model training related data collection (e.g., for one or more use cases) and MDT related measurement collection procedures. UE memory may additionally or alternatively be shared with QoE related measurement collection procedures.

[0115] According to a particular example, a UE 110 can allocate ‘X’ megabytes of memory for data collection procedures. This ‘X’ megabytes can be used entirely by one of the above-mentioned data collection procedures or for more than one of the above mentioned data collection procedures. When only one data collection procedure is configured at the UE 110, the UE 110 may collect and transmit one or more measurements whose size is up to ‘X’ megabytes. When more than one data collection procedure is configured at the UE 110, the UE 110 may still collect and transmit up to ‘X’ megabytes. However, in some embodiments the ‘X’ megabytes that are transmitted may be shared across the configured data collection procedures. In other embodiments, the ‘X’ megabytes that are transmitted may relate to the same data collection procedure. The UE 110 may decide how to allocate memory and transmit accordingly, depending on the embodiment.

[0116] For example, in some embodiments, a first amount of shared memory is allocated for data collection sessions associated with UE-side model training and a second, separate amount of shared memory is allocated for data collection sessions associated with NW-side model training. In some such embodiments, the UE may indicate, to the network, that it is capable of differentiated memory allocation between the UE-side model training and the NW-side model training.

[0117] In another embodiment, shared memory is allocated for data collection sessions associated with training one or more models that have the same features, functionalities or groups thereof. For example, for CSI prediction, the UE 110 may initiate multiple data collection sessions to train different models (e.g., different models valid in different geographical areas or different gNBs). All these data collection sessions are associated with training different models with respect to the same functionality. The UE 110 may allocate memory that is shared between these data collection sessions.

[0118] In another embodiment, shared memory is allocated for data collection sessions associated with the same training node. For example, for the data collection sessions associated with a first specific network node (e.g., a first gNB performing the training), the UE 110 may allocate a first amount of memory that is separate from a second amount of memory allocated for data collection sessions associated with a second network node (e.g., a second gNB performing the training). Similarly, the allocated memory for the data collections sessions associated with the first and / or second network node may be separate from memory allocated for the data collection sessions associated with a core network node 140 (e.g., an OAM server). In another embodiment, shared memory is allocated for each of a plurality of different vendors that are associated with the node initiating the training data collection procedure.

[0119] In another embodiment, shared memory is allocated for each of a plurality of different operating conditions at the UE 110. Such operating conditions may include, e.g., one or more of:

[0120] • A range of speeds of the UE as measured using a sensor or standardized classification (e.g., as used in cell reselection).

[0121] • One or more frequency ranges (e.g., one or more frequency bands supported by the UE 110 or one or more of the configured serving cell frequencies).

[0122] • A gNB configuration. For example, different gNBs having the same configuration may be associated with the same model and data collected for such gNBs may correspondingly share the same memory.

[0123] • NW deployments. For example, data collected under the same NW deployment (e.g., a macro, micro, or indoor deployment) may share the same memory.

[0124] In yet another embodiments, the shared memory is allocated for logged MDT procedure and a model training data collection procedure.

[0125] Any of the embodiments discussed above may include further features relating to signaling between the UE 110 and the network node with respect to memory management for data collection performed at the UE 110, as described below.

[0126] In some embodiments, the UE 110 transmits, to a network node, a report indicating whether or not a data collection session can be initiated based on UE memory availability. This report may be transmitted by the UE 110 in response to the UE 110 being configured by a network node to perform data collection for NW-side model training. For example, the network node may transmit a request to the UE 110 that requests availability of the UE to perform data collection for NW-side model training. The UE 110 may reply to the request indicating the memory availability based on the current memory allocation. In some particular examples, the UE 110 may indicate that memory is not available for the requested data collection.

[0127] In another embodiment, the report is transmitted by the UE 110 in response to the UE 110 determining that sufficient memory is available. For example, upon terminating data collection and upon transmitting the collected data to the training node, the UE 110 may release some memory that may now be available for data collection. The UE 110 may report this availability. Similarly, in an alternative example, the UE 110 may transmit a memory availability report upon determining that memory is no longer available.

[0128] The memory availability report may comprise an indication of any one or more of, e.g., whether or not UE memory is available, how much memory is available (or how much is unavailable), how much memory is allocated (e.g., for a particular purpose, such as data collected for UE-side model training or for the NW-side model training), for what purpose UE memory is allocated, and / or the memory allocated for each purpose.

[0129] In some embodiments, the network node may indicate the priorities amongst the data collection procedures. For example, the network node can indicate, to the UE 110, that data collection for NW-side training data collection is to be prioritized over ongoing data collection for UE-sided training data collection.

[0130] In some embodiments, the UE 110 discards data collected for a lower priority data collection procedure upon receiving a request to collect data for a higher priority data collection procedure. Such an embodiment may be useful for increasing data collection for higher priority data.

[0131] Alternatively, in some embodiments, the UE 110 does not discard data already collected for a lower priority data collection procedure but refrains from collecting further data for the lower priority data collection procedure until the higher priority data collection procedure session is established. Such an embodiment may be useful to avoid wasting lower priority data collection efforts.

[0132] In some embodiments, the UE 110 may indicate the priorities of the data collection procedures as part of the UE capabilities. For example, the UE 110 may indicate that the UE prioritizes NW-sided training data collection over UE-sided training data collection.

[0133] In some embodiments, the UE prioritizes signaling-based training data collection procedures over management-based training data collection procedures in a similar way to that of MDT procedures. To do so, a data collection request from the network node may include an indication of whether the configured data collection procedure is a signaling-based data collection procedure or a management-based data collection procedure. That is, in some embodiments, the data collection request may indicate a type of data collection being requested and the UE 110 may apply a priority that corresponds to the type of data collection indicated.

[0134] In some embodiments, if the UE 110 receives a data collection request from a network node that has lower priority than an ongoing data collection procedure, then the UE 110 may send a rejection notification to the network node in response to the request.

[0135] In some such embodiments, such a rejection message includes an indication that the UE is already configured with a higher priority data collection procedure. In some such embodiments, the UE 110 indicates the priority of the higher priority data collection procedure. In some such embodiments, the UE 110 indicates whether or not the higher priority data collection procedure is currently running. In some such embodiments in which the higher priority data collection procedure is not currently running, the UE 110 may indicate whether any higher priority data has yet been collected. In some embodiments, in response to a request, from a network node, for lower priority data collection when higher priority data collection is already running, the UE 110 may indicate when the UE 110 will be able to initiate the requested lower priority data collection. In some such embodiments, the UE 110 indicates a time when the higher priority data collection is expected to end.

[0136] It should be noted that any combination of the above signaling and / or memory allocation embodiments may be combined. For example, FIG. 7 is a flow diagram illustrating an example method 200 implemented by a UE 110, according to one or more embodiments of the present disclosure. The method 200 comprises allocating memory of the UE 110 to one or more model training data collection procedures (block 210). The method 200 further comprises collecting model training data by performing the one or more model training data collection procedures (block 220). The method 200 further comprises transmitting the collected model training data to a network node (block 230).

[0137] FIG. 8 is a flow diagram illustrating an example method 300 implemented by a network node, according to one or more embodiments of the present disclosure. The method 300 comprises transmitting, to a UE 110, one or more requests to perform a plurality of model training data collection procedures (block 310). The method 300 further comprises receiving, from the UE 110 and responsive to the request, model training data collected by the UE 110 performing the model training data collection procedures (block 320).

[0138] The UE 110 may, for example, be implemented as schematically illustrated in the example of FIG. 9. The UE 110 of FIG. 9 comprises processing circuitry 610, memory circuitry 620, and interface circuitry 630. The processing circuitry 610 is communicatively coupled to the memory circuitry 620 and the interface circuitry 630, e.g., via a bus 604. The processing circuitry 610 may comprise one or more microprocessors, microcontrollers, hardware circuits, discrete logic circuits, hardware registers, digital signal processors (DSPs), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), or a combination thereof. For example, the processing circuitry 610 may be programmable hardware capable of executing software instructions stored, e.g., as a machine-readable computer program 640 in the memory circuitry 620. The memory circuitry 620 of the various embodiments may comprise any non-transitory machine- readable media known in the art or that may be developed, 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 drive, etc.), removable storage devices (e.g., Secure Digital (SD) card, miniSD card, microSD card, memory stick, thumb-drive, USB flash drive, ROM cartridge, Universal Media Disc), fixed drive (e.g., magnetic hard disk drive), or the like, wholly or in any combination. The interface circuitry 630 may be a controller hub configured to control the input and output (I / O) data paths of the UE 110. Such I / O data paths may include data paths for exchanging signals over a network. The interface circuitry 630 may be implemented as a unitary physical component, or as a plurality of physical components that are contiguously or separately arranged, any of which may be communicatively coupled to any other, or may communicate with any other via the processing circuitry 610. For example, the interface circuitry 630 may comprise a transmitter 632 configured to send wireless communication signals and a receiver 634 configured to receive wireless communication signals.

[0139] The UE 110 may be configured (e.g., by the processing circuitry 610) to perform the method 200 described above.

[0140] Still other embodiments include a control program 640 comprising instructions that, when executed on processing circuitry 610 of a UE 110, cause the UE 110 to carry out the method 200 described above.

[0141] Yet other embodiments include a carrier containing the control program 640. The carrier is one of an electronic signal, optical signal, radio signal, or computer readable storage medium.

[0142] Correspondingly, a network node 170 may be implemented as schematically illustrated in the example of FIG. 10. The network node 170 of FIG. 10 comprises processing circuitry 710, memory circuitry 720, and interface circuitry 730. The processing circuitry 710 is communicatively coupled to the memory circuitry 720 and the interface circuitry 730, e.g., via a bus 704. The processing circuitry 710 may comprise one or more microprocessors, microcontrollers, hardware circuits, discrete logic circuits, hardware registers, digital signal processors (DSPs), field- programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), or a combination thereof. For example, the processing circuitry 710 may be programmable hardware capable of executing software instructions stored, e.g., as a machine-readable computer program 740 in the memory circuitry 720. The memory circuitry 720 of the various embodiments may comprise any non-transitory machine-readable media known in the art or that may be developed, 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 drive, etc.), removable storage devices (e.g., Secure Digital (SD) card, miniSD card, microSD card, memory stick, thumb-drive, USB flash drive, ROM cartridge, Universal Media Disc), fixed drive (e.g., magnetic hard disk drive), or the like, wholly or in any combination.

[0143] The interface circuitry 730 may be a controller hub configured to control the input and output (I / O) data paths of the network node 170. Such I / O data paths may include data paths for exchanging signals over a network. The interface circuitry 730 may be implemented as a unitary physical component, or as a plurality of physical components that are contiguously or separately arranged, any of which may be communicatively coupled to any other or may communicate with any other via the processing circuitry 710. For example, the interface circuitry 730 may comprise a transmitter 732 configured to send wireless communication signals and a receiver 734 configured to receive wireless communication signals.

[0144] The network node 170 may be configured (e.g., by the processing circuitry 710) to perform the method 300 described above.

[0145] Still other embodiments include a control program 740 comprising instructions that, when executed on processing circuitry 710 of a network node 170, cause the network node 170 to carry out the method 300 described above.

[0146] Yet other embodiments include a carrier containing the control program 740. The carrier is one of an electronic signal, optical signal, radio signal, or computer readable storage medium.

[0147] Although the various communication devices described herein may include the illustrated combination of hardware components, other embodiments may comprise computing and / or communication hardware with different combinations of components. It is to be understood that these computing devices may comprise any suitable combination of hardware and / or software needed to perform the tasks, features, functions, and methods disclosed herein. Further, while components are depicted as single boxes located within a larger box, or nested within multiple boxes, in practice, the devices described herein may comprise multiple different physical components that make up 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 the processing circuitry and the communication interface. In another example, non-computationally intensive functions of any of such components may be implemented in software or firmware and computationally intensive functions may be implemented in hardware.

[0148] In certain embodiments, some or all of the functionality described herein may be provided by processing circuitry executing instructions stored on in memory, which in certain embodiments may be a computer program product in the form of a non-transitory computer-readable storage medium. In alternative embodiments, some or all of the functionality may be provided by the processing circuitry without executing instructions stored on a separate or discrete device-readable storage medium, such as in a hard-wired manner. In any of those particular embodiments, whether executing instructions stored on a non-transitory computer-readable storage medium or not, the processing circuitry can be configured to perform the described functionality. The benefits provided by such functionality are not limited to the processing circuitry alone or to other components of the device but are enjoyed by the device as a whole, and / or by end users and a wireless network generally.

Claims

CLAIMSWhat is claimed is:

1. A method (200), implemented by a User Equipment, UE (110), the method comprising: allocating (210) memory of the UE (110) to one or more model training data collection procedures; collecting (220) model training data by performing the one or more model training data collection procedures; and transmitting (230) the collected model training data to a network node (170).

2. The method of claim 1 , wherein allocating the memory comprises allocating a shared amount of memory between a plurality of the model training data collection procedures.

3. The method of claim 2, wherein allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures associated with a same model and / or model function.

4. The method of claim 2, wherein allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures that are associated with a same training node.

5. The method of claim 2, wherein allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures involving data collection in a same range of UE speeds.

6. The method of claim 2, wherein allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures involving data collection using a same frequency range.

7. The method of claim 2, wherein allocating the shared amount of memory between the plurality of the model training data collection procedures comprises allocating the shared amount of memory to model training data collection procedures involving data collection associated with a same RAN node configuration.

8. The method of any one of claims 1-7, wherein allocating the memory comprises allocating a separate respective portion of the memory to each of a plurality of the model training data collection procedures.

9. The method of claim 8, wherein allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures having different functional purposes.

10. The method of claim 8, wherein allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures that are associated with different training nodes.11 . The method of claim 8, wherein allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures involving data collection in different ranges of UE speeds.

12. The method of claim 8, wherein allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures involving data collection using different frequency ranges.

13. The method of claim 8, wherein allocating the separate respective portion of the memory to each of the plurality of the model training data collection procedures comprises allocating separate respective portions of the memory to model training data collection procedures involving data collection associated with different RAN node configurations.

14. The method of any one of claims 1-13, wherein: allocating the memory comprises allocating the memory exclusively to the model training data collection procedures; and the method further comprises allocating a further amount of memory, separate from the amount of memory, to one or more non-model training data collection procedures.

15. The method of any one of claims 1-14, wherein allocating the memory comprises allocating: a first part of the memory to a first one or more of the model training data collection procedures that support UE-side data collection; and a second part of the memory to a second one or more of the model training data collection procedures that support network-side data collection.

16. The method of any one of claims 1-15, further comprising: withholding memory from one or more further model training data collection procedures until the collected model training 1data has been transmitted; and notifying the network node (170) of when the one or more further model training data collection procedures is estimated to begin.

17. The method of any one of claims 1-16, further comprising receiving a request to perform one or more of the model training data collection procedures from the network node (170).

18. The method of any one of claims 1-17, further comprising transmitting a report indicating whether or not the one or more model training data collection procedures can be initiated based on the allocated memory.

19. The method of claim 18, wherein transmitting the report is responsive to the allocated memory being sufficient to initiate the one or more model training data collection procedures.

20. The method of claim 18, wherein transmitting the report is responsive to the allocated memory being insufficient to initiate the one or more model training data collection procedures.21 . The method of any one of claims 18-20, wherein the report indicates how much of the memory is allocated to the one or more model training data collection procedures.

22. The method of any one of claims 18-21 , wherein the report indicates how much of the memory is allocated to UE-side model training procedures and / or how much of the memory is allocated to network-side model training procedures.

23. The method of any one of claims 1-22, further comprising: receiving, from the network node (120), a priority of at least one of the model training data collection procedures, wherein the priority is higher than a further priority of a further model training data collection procedure; and refraining from including the further model training data collection procedure in the memory allocation.

24. The method of claim 23, wherein allocating the memory to the model training data collection procedures comprises allocating less of the memory to a lower priority model training data collection procedure than to a higher priority model training data collection procedure.

25. The method of any one of claims 1-24, further comprising freeing the memory for the allocation from one or more other model training data collection procedures that have completed.

26. A User Equipment, UE (110), configured to: allocate memory of the UE (110) to one or more model training data collection procedures; collect model training data by performing the one or more model training data collection procedures; and transmit the collected model training data to a network node (170).

27. The UE of the preceding claim, further configured to perform the method of any one of claims 2-25.

28. A User Equipment, UE (110), comprising: interface circuitry (630), memory (620), and processing circuitry (610) communicatively connected to the interface circuitry (630) and the memory (620), wherein the processing circuitry (610) is configured to: allocate the memory (620) of the UE (110) to one or more model training data collection procedures; collect model training data by performing the one or more model training data collection procedures; and transmit the collected model training data to a network node (170) via the interface circuitry (630).

29. The UE of the preceding claim, wherein the processing circuitry (610) is further configured to perform the method of any one of claims 2-25.

30. A computer program, comprising instructions which, when executed on processing circuitry (610) of a User Equipment, UE (110), cause the processing circuitry (610) to carry out the method according to any one of claims 1-23.31 . A method (300), implemented by a network node (170), the method comprising: transmitting (310), to a User Equipment, UE (110), one or more requests to perform a plurality of model training data collection procedures; receiving (320), from the UE (110) and responsive to the request, model training data collected by the UE (110) performing the model training data collection procedures.

32. The method of claim 31 , further comprising: transmitting, to the UE (110), a further request to perform one or more further model training data collection procedures; and receiving, from the UE (110) in response to the further request, notice of when the one or more further model training data collection procedures is estimated to begin.

33. The method of any one of claims 31-32, further comprising receiving, from the UE (110), a report indicating that the memory available at the UE (110) is sufficient to initiate the one or more model training data collection procedures.

34. The method of any one of claims 31-32, further comprising receiving, from the UE (110), a report indicating that the memory available at the UE (110) is insufficient to initiate the one or more model training data collection procedures.

35. The method of any one of claims 33-34, wherein the report further indicates how much of the memory is allocated to: the one or more model training data collection procedures; and / orUE-side model training procedures; and / or network-side model training procedures.

36. The method of any one of claims 31-35, further comprising: transmitting, to the UE (110), a priority of at least one of the model training data collection procedures, wherein the priority is higher than a further priority of a further model training data collection procedure; and receiving the model training data comprises receiving the model training data before receiving further model training data collected by the UE (110) performing the further model training data collection procedure.

37. A network node (170) configured to: transmit, to a User Equipment, UE (110), one or more requests to perform a plurality of model training data collection procedures; receive, from the UE (110) and responsive to the request, model training data collected by the UE (110) performing the model training data collection procedures.

38. The network node of the preceding claim, further configured to perform the method of any one of claims 32-36.

39. A network node (170) comprising: interface circuitry (730) and processing circuitry (710) communicatively coupled to the interface circuitry (730), wherein the processing circuitry (710) is configured to: transmit, to a User Equipment, UE (110) via the interface circuitry (730), one or more requests to perform a plurality of model training data collection procedures; receive, from the UE (110) via the interface circuitry (730) and responsive to the request, model training data collected by the UE (110) performing the model training data collection procedures.

40. The network node of the preceding claim, wherein the processing circuitry (710) is further configured to perform the method of any one of claims 32-36.41 . A computer program, comprising instructions which, when executed on processing circuitry (710) of a network node (170) cause the processing circuitry (710) to carry out the method according to any one of claims 31-36.

42. A carrier containing the computer program of claim 30 or 41 , wherein the carrier is one of an electronic signal, optical signal, radio signal, or computer readable storage medium.

Images