Method of multi-training model operation signaling

Abstract

The disclosure describes a method of using a pre-configured AI / ML (artificial intelligence / machine learning) based model training operation type in a wireless mobile communication system including a base station (e.g., gNB, TN, NTN) and a mobile station (e.g., UE). When an AI / ML model is applied to a radio access network, model training performance can be degraded depending on UE ML condition or capability change. Accordingly, model operation (e.g., model training / inference / monitoring / update) can be established between the network and the UE by configuring a model training operation type.

Description

Technical Field

[0001] This disclosure relates to AI / ML-based model training operation types, wherein techniques are proposed for pre-configuring and signaling specific information about model training operation types applicable to radio access networks. Background Technology

[0002] Within 3GPP (3rd Generation Partnership Project), one of the selected research projects as part of the approved Release 18 package is AI / ML (Artificial Intelligence / Machine Learning), as described in the relevant document (RP-213599) submitted at 3GPP TSG (Technical Specification Group) RAN (Radio Access Networks) meeting #94e. The formal title of this AI / ML research project is "Study on AI / ML for NR Air Interface." The goal of this research project is to establish a general AI / ML framework and, in conjunction with use cases, identify areas where AI / ML-based technologies can yield benefits. According to 3GPP, the primary objective of this research project is to study an AI / ML framework for the air interface using target use cases, taking into account performance, complexity, and potential specification impacts. Specifically, AI / ML models, terminology, and descriptions used to establish common and specific characteristics of the framework will be included as one of the key work scopes. Regarding the AI / ML framework, various aspects are being researched, and one key project concerns the lifecycle management of AI / ML models, which mandates multiple phases including model training, deployment, inference, monitoring, and updates. Furthermore, in 3GPP, a two-sided (AI / ML) model is defined as a paired AI / ML model on which joint inference is performed, where joint inference includes AI / ML inference jointly performed across the UE and network. Conversely, for a one-sided (AI / ML) model, a UE-side (AI / ML) model is defined as an AI / ML model whose inference is entirely performed at the UE, and a network-side (AI / ML) model is defined as an AI / ML model whose inference is entirely performed at the network. Currently, the AI / ML specification work is in the discussion phase regarding the work item for version 19. Earlier, in 3GPP TR 37.817, Release 17, entitled "Study on enhancement for Data Collection for NR and EN-DC," UE (User Equipment) mobility was also considered as one of the AI / ML use cases, and one of the scenarios for model training / inference was that both functions reside within the RAN node. Subsequently, in Release 18, a new work project, "Artificial Intelligence (AI) / Machine Learning (ML) for NG-RAN," was initiated to specify data collection enhancements and signaling support within the existing NG-RAN interface and architecture.Regarding the aforementioned positive standardization efforts, RAN-based AI / ML models are considered crucial for both the network and UE to meet any desired model operations (e.g., model training, inference, selection, handover, updates, monitoring, etc.). Model information can be signaled to pair the network-side and UE-side models for various lifecycle management (LCM) operations. However, the signaling overhead for indicating model information can be very high, especially when processing model-based LCM between the base station (BS / gNB) and multiple UEs.

[0003] In LCM, model training is one of the most critical parts of model deployment. Currently, there are no defined specifications for signaling methods and network-UE behavior to establish the datasets required for model updates / retraining, as any activated model can be affected by model / data drift. When ML conditions change, the performance of the activated AI / ML model may be impacted by data / model drift. In such cases, model retraining / update can be performed.

[0004] WO2023144831A1 describes a method for lifecycle management of at least one machine learning (ML) model for telecommunications capacity planning in a network, wherein the method includes performing actions to determine that the performance of the current ML model is unacceptable for the prediction of telecommunications capacity planning, and selecting a minimum information dataset.

[0005] WO2023173296A1 describes a method for selectively including or excluding a device from an online training process based on its currently reported learning capabilities, in order to provide a trade-off between reduced overhead and training performance.

[0006] US2022116764A1 shows that the base station receives machine learning processing capability reports from each of multiple UEs, and the base station groups the multiple UEs according to the machine learning processing capability reports to receive gradient updates of the machine learning model.

[0007] US2022360973A1 indicates that the UE may transmit one or more of the following indications to the network based on a request to report UE capabilities: AI capabilities, ML capabilities, radio capabilities associated with at least one of the AI ​​or ML processes, or core network capabilities associated with at least one of the AI ​​or ML processes.

[0008] WO2022008037A1 indicates that if the terminal indicates the capability and the terminal is in a capability-incapable state, the terminal notifies the network that the terminal is in a capability-incapable state, wherein, in a capability-incapable state, the terminal is unable to execute and / or train a machine learning model, or the terminal is unable to execute and / or train a machine learning model at least with predefined performance.

[0009] For AI / ML model operations involving one-sided or two-sided models between the network and the UE, there are various types of model training methods, such as joint training or individual training. However, different UEs have model training constraints to support any particular model training method (e.g., private models are not allowed for joint training). Due to the training capability constraints of the UEs, the number of target UEs for training is limited, so training convergence may be slow or it may not be possible to efficiently apply any particular type of model training.

[0010] This problem is solved by configuring multiple types of model training operations so that two or more types of model training operations can work together to train the same target model.

[0011] This solution advantageously leads to an increase in the target UE used for model training, regardless of the variety of model training types, and improves model training performance, such as convergence speed.

[0012] This disclosure addresses the cited problems through the proposed embodiments and, as a first aspect, describes a method for configuring multi-training model operation signaling for multiple types of model training operations in a wireless communication system, comprising: grouping target UEs for specific types of model training operations; defining information for exchange between model training operations; determining specific bits of signaling for selective / combined model training; and pre-setting mapping information to support various types of model training operations.

[0013] In some embodiments of the method according to the first aspect, the method is characterized in that two or more types of model training operations can be performed collaboratively to train the same target model.

[0014] According to the method of any one of the preceding claims, the grouping of target UEs for model training can be based on device training capability information for a specific model training operation type.

[0015] In some embodiments of the method according to the first aspect, the method is characterized in that, depending on network-side decision or UE autonomous decision, the UE can be assigned to operate two or more types of model training operations.

[0016] In some embodiments of the method according to the first aspect, the method is characterized in that two or more model training operation types can be enabled simultaneously to perform single-sided or dual-sided model training between the network side and the UE side.

[0017] In some embodiments of the method according to the first aspect, the method is characterized in that model training updates can be exchanged with each other across different UE groups.

[0018] In some embodiments of the method according to the first aspect, the method is characterized by using N-bit signaling to enable selective model training or combined model training.

[0019] In some embodiments of the method according to the first aspect, the method is characterized in that selective model training allows a single type of model training operation to be activated between the network and the UE.

[0020] In some embodiments of the method according to the first aspect, the method is characterized in that combined model training allows multiple types of model training operations to be activated between the network and the UE, enabling two or more different types of model training operations to be executed in parallel.

[0021] In some embodiments of the method according to the first aspect, the method is characterized in that model training information exchanged across those training types can be combined for model training.

[0022] In some embodiments of the method according to the first aspect, the method is characterized in that the specific content of the training update information may vary depending on different model training operation types and / or application scenarios.

[0023] In some embodiments of the method according to the first aspect, the method is characterized in that the exchange of training update information can be configured using ML configuration information via L1 / L2 or RRC signaling.

[0024] In some embodiments of the method according to the first aspect, the method is characterized in that the mapping relationship between the associated set of parameter values ​​for information exchange and the model training operation type can be preset to be available based on ML deployment scenarios with different model applications / functions via system information or dedicated RRC signaling.

[0025] In some embodiments of the method according to the first aspect, the method is characterized in that any update to the mapping table (including the associated set of parameter values ​​and model training operation types for information exchange) can be indicated via L1 / L2 or RRC signaling.

[0026] In some embodiments of the method according to the first aspect, the method is characterized in that a set of model training operation types can be configured by including single-sided model scenarios and dual-sided model scenarios, so that model training can be performed jointly or independently between the network side and the UE side.

[0027] In some embodiments of the method according to the first aspect, the method is characterized in that the set of parameter values ​​associated with each type of model training operation can be predefined in a mapping table.

[0028] In some embodiments of the method according to the first aspect, the method is characterized in that the UE can provide the network side with configured parameter value information related to each model training operation type based on a mapping table format.

[0029] In some embodiments of the method according to the first aspect, the method is characterized in that different model training operation types can be determined for different UE groups.

[0030] In some embodiments of the method according to the first aspect, the method is characterized in that the network side can coordinate which combinations of model training types are enabled across multiple UE groups for model training.

[0031] In some embodiments of the method according to the first aspect, the method is characterized in that an indication message can be sent via L1 / L2 or RRC signaling using broadcast or multicast transmission to enable a specific model training operation type for each UE group.

[0032] In some embodiments of the method according to the first aspect, the method is characterized in that, based on a pre-configured model training operation type, a single network-side model can be enabled to perform bilateral model training with a group of UEs having a UE-side model.

[0033] In some embodiments of the method according to the first aspect, the method is characterized in that the UE can enable two or more model training operation types to support the configured two-sided model training.

[0034] In some embodiments of the method according to the first aspect, the method is characterized in that the number of UE groups can be flexible to determine the target UE and its associated UE-side model for use.

[0035] In some embodiments of the method according to the first aspect, the method is characterized in that multiple network-side models can be enabled to perform bilateral model training with a group of UEs having UE-side models.

[0036] In some embodiments of the method according to the first aspect, the method is characterized in that multiple network-side models configured to support a specific model training operation type can be located at the same network-side location or distributed in different locations via network interfaces.

[0037] In some embodiments of the method according to the first aspect, the method is characterized in that each network-side model can activate a specific model training operation type for a determined UE group.

[0038] In some embodiments of the method according to the first aspect, the method is characterized by the ability to collaboratively train models across different network-side models associated with their own UE groups.

[0039] In some embodiments of the method according to the first aspect, the method is characterized in that the mapping relationship between the associated set of parameter values ​​for information exchange and the model training operation type can be preset on the network side based on ML deployment scenarios with different model applications / functions.

[0040] In some embodiments of the method according to the first aspect, the method is characterized in that the grouping of target UEs for model training based on device training capability information can be determined by the network side.

[0041] In some embodiments of the method according to the first aspect, the method is characterized in that two or more model training operation types can be enabled simultaneously to perform single-sided or dual-sided model training between the network side and the UE side.

[0042] In some embodiments of the method according to the first aspect, the method is characterized in that the UE can receive ML model training configuration information from the network side, including a mapping relationship between an associated set of parameter values ​​and model training operation types for information exchange.

[0043] In some embodiments of the method according to the first aspect, the method is characterized in that, after activating a specific model training operation type, the UE can provide the network side with configured parameter value information based on a mapping table.

[0044] According to the second aspect, this disclosure addresses the cited problems through the proposed embodiments and is described by means of a device for configuring multiple types of model training operations signaling in a wireless communication system, the device including a wireless transceiver and a processor coupled to a memory storing computer program instructions configured to perform the steps of the first aspect.

[0045] According to the third aspect, this disclosure addresses the cited problems through the proposed embodiments and is described through a user equipment that includes the device according to the second aspect.

[0046] According to the fourth aspect, this disclosure addresses the cited problems through the proposed embodiments and is described by way of a gNB that includes the device according to the second aspect.

[0047] According to a fifth aspect, this disclosure relates to a wireless communication system for configuring multi-training model operation signaling for multiple types of model training operations, wherein these wireless communication systems include at least one user equipment according to a third aspect and at least one gNB according to a fourth aspect, wherein the user equipment and the gNB each include a processor coupled to a memory in which computer program instructions are stored.

[0048] According to a sixth aspect, this disclosure relates to a computer program product comprising instructions that, when executed by at least one processor, configure the at least one processor to perform the method according to the first aspect, and to perform a method for exchanging data according to any of the embodiments of this disclosure. The computer program product may use any programming language and may be in the form of source code, object code, or any intermediate form between source code and object code, such as a partially compiled form, or any other desired form.

[0049] According to a sixth aspect, this disclosure relates to a computer-readable storage medium including instructions that, when executed by at least one processor, configure the at least one processor to perform any of the methods described in embodiments of this disclosure. Attached Figure Description

[0050] Figure 1 This is an exemplary block diagram of exchanging training update information between Type 1 and Type 2 model training operations.

[0051] Figure 2 This is an exemplary table showing the mapping between model training operation types and the associated set of parameter values ​​used for information exchange.

[0052] Figure 3 This is an exemplary block diagram for assigning model training types to multiple UE groups.

[0053] Figure 4 This is an exemplary block diagram for enabling model training types for a single network-side model and multiple UE-side models.

[0054] Figure 5 This is an exemplary block diagram for enabling model training types for multiple network-side models and multiple UE-side models.

[0055] Figure 6 This is an example flowchart for configuring model training operation types on the network side.

[0056] Figure 7 This is an example flowchart for configuring model training operation types on the UE side. Detailed Implementation

[0057] The detailed description set forth below with reference to the accompanying drawings is intended as a description of various configurations and is not intended to represent only the configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. Specifically, although terms from 3GPP 5G NR may be used in this disclosure to exemplify embodiments herein, this should not be construed as limiting the scope of the invention.

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

[0059] Generally, all terms used herein should be interpreted according to their ordinary meaning in the relevant art, unless a different meaning is expressly given and / or implied in the context of their use. Unless otherwise expressly stated, all references to a / an / element, device, component, means, step, etc., should be interpreted openly as referring to at least one instance of that element, device, component, means, step, etc. The steps of any method disclosed herein need not be performed in the exact order disclosed, unless a step is explicitly described as following or preceding another step, and / or it is implied that a step must follow or precede another step. Where appropriate, any feature of any embodiment of the embodiments disclosed herein may be applied to any other embodiment. Similarly, any advantage of any embodiment may be applied to any other embodiment, and vice versa. Other objects, features, and advantages of the appended embodiments will become apparent from the following description.

[0060] In some embodiments, the more general term "network node" may be used, and this term may correspond to any type of radio network node or any network node that communicates with the UE (directly or via another node) and / or another network node. Examples of network nodes are NodeB, MeNB, ENB, network nodes belonging to MCG or SCG, base stations (BS), multi-standard radio (MSR) radio nodes (such as MSR BS), eNodeB, gNodeB, network controller, radio network controller (RNC), base station controller (BSC), repeater, donor node controlling repeater, base transceiver station (BTS), access point (AP), transport point, transport node, RRU, RRH, nodes in distributed antenna system (DAS), core network nodes (e.g., mobile switching center (MSC), mobility management entity (MME), etc.), operations and maintenance (O&M), operations support system (OSS), self-optimizing network (SON), location nodes (e.g., evolved servicing mobile location center (E-SMLC)), minimized drive test (MDT), test equipment (physical node or software), etc.

[0061] In some embodiments, the non-limiting terms User Equipment (UE) or Wireless Device may be used, and the term may refer to any type of wireless device that communicates with a network node and / or with another UE in a cellular or mobile communication system. Examples of UEs are target devices, device-to-device (D2D) UEs, machine-type UEs or UEs capable of machine-to-machine (M2M) communication, PDAs, PADs, tablet computers, mobile terminals, smartphones, laptop embedded devices (LEE), laptop mounted devices (LME), USB dongles, UE class M1, UE class M2, ProSe UE, V2V UE, V2X UE, etc.

[0062] Additionally, terms such as base station / gNodeB and UE should be considered non-limiting and, in particular, do not imply any hierarchical relationship between the two; generally, "gNodeB" can be considered device 1 and "UE" can be considered device 2, and the two devices communicate with each other via a radio channel. Furthermore, in the following text, a transmitter or receiver can be either a gNodeB (gNB) or a UE.

[0063] As those skilled in the art will understand, aspects of the embodiments can be embodied as a system, device, method, or program product. Therefore, embodiments can take the form of an all-hardware embodiment, an all-software embodiment (including firmware, resident software, microcode, etc.), or an embodiment combining software and hardware aspects.

[0064] For example, the disclosed embodiments can be implemented as hardware circuitry, including custom-designed very large-scale integration (“VLSI”) circuitry or gate arrays, off-the-shelf semiconductors (such as logic chips, transistors, or other discrete components). The disclosed embodiments can also be implemented in programmable hardware devices, such as field-programmable gate arrays, programmable array logic, programmable logic devices, etc. As another example, the disclosed embodiments may include one or more physical or logical blocks of executable code, which may, for example, be organized as objects, procedures, or functions.

[0065] Furthermore, embodiments may take the form of a program product embodied in one or more computer-readable storage devices that store machine-readable code, computer-readable code, and / or program code (hereinafter referred to as code). The storage device may be tangible, non-transitory, and / or non-transferable. The storage device may not contain signals. In one embodiment, the storage device uses only signals to access the code.

[0066] Any combination of one or more computer-readable media may be used. A computer-readable medium may be a computer-readable storage medium. A computer-readable storage medium may be a storage device for storing code. A storage device may be, for example, but not limited to, electronic, magnetic, optical, electromagnetic, infrared, holographic, micromechanical, or semiconductor systems, devices, or apparatuses, or any suitable combination of the foregoing.

[0067] More specific examples of storage devices (a non-exhaustive list) will include the following: electrical connections having one or more wires, portable computer floppy disks, hard disks, random access memory (“RAM”), read-only memory (“ROM”), erasable programmable read-only memory (“EPROM” or flash memory), portable optical disc read-only memory (“CD-ROM”), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. In the context of this document, a computer-readable storage medium can be any tangible medium that can contain or store programs for use by or in conjunction with an instruction execution system, device, or apparatus.

[0068] The code used to perform the operations of the embodiments can be any number of lines and can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Python, Ruby, Java, Smalltalk, C++, etc., as well as conventional procedural programming languages ​​such as the "C" programming language, and / or machine languages ​​such as assembly language. The code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer via any type of network, including a local area network ("LAN"), a wireless LAN ("WLAN"), or a wide area network ("WAN"), or can be connected to an external computer (e.g., via the Internet through an Internet service provider ("ISP").

[0069] Furthermore, the features, structures, or characteristics described in the embodiments can be combined in any suitable manner. Numerous specific details, such as examples of programming, software modules, user selection, network transactions, database queries, database structures, hardware modules, hardware circuits, hardware chips, etc., are provided in the following description to provide a thorough understanding of the embodiments. However, those skilled in the art will recognize that the embodiments can be practiced without one or more of these specific details or using other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the embodiments. References to “an embodiment,” “embodiment,” or similar language throughout the specification mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment. Therefore, unless expressly stated otherwise, the phrases “in one embodiment,” “in an embodiment,” and similar language appearing throughout the specification may, but not necessarily all, refer to the same embodiment, but rather mean “one or more, but not all, embodiments.” Unless expressly stated otherwise, the terms “including,” “comprising,” “having,” and variations thereof mean “including, but not limited to.” Unless expressly stated otherwise, the enumerated list of items does not imply that any or all of these items are mutually exclusive. Unless otherwise expressly stated, the terms “a,” “an,” and “the / said” also mean “one or more.”

[0070] The following description refers to schematic flowcharts and / or schematic block diagrams of methods, apparatus, systems, and program products according to embodiments. It should be understood that each block of the schematic flowcharts and / or schematic block diagrams, and combinations of blocks of the schematic flowcharts and / or schematic block diagrams, can be implemented by code. This code can be provided to a processor of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus to produce a machine such that instructions executable via the processor of the computer or other programmable data processing apparatus establish components for implementing the functions / actions specified in the flowcharts and / or block diagrams.

[0071] The code may also be stored in a storage device that can instruct a computer, other programmable data processing device or other device to operate in a particular manner, such that the instructions stored in the storage device produce an article of art, which includes instructions that implement the functions / actions specified in the flowchart and / or block diagram.

[0072] The code may also be loaded onto a computer, other programmable data processing device or other apparatus to cause a series of operational steps to be performed on the computer, other programmable device or other apparatus, thereby producing a computer-implemented process, such that the code executing on the computer or other programmable device provides a process for implementing the functions / actions specified in the flowchart and / or block diagram.

[0073] The flowcharts and / or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, systems, methods, and program products according to various embodiments. In this regard, each block in the flowcharts and / or block diagrams may represent a module, segment, or portion of code, which includes one or more executable instructions for implementing a specified logical function.

[0074] It should also be noted that in some alternative implementations, the functions indicated within the boxes may not appear in the order shown in the figures. For example, depending on the functions involved, two consecutively shown boxes may actually be executed substantially simultaneously, or these boxes may sometimes be executed in reverse order. Other steps and methods that are functionally, logically, or effectively equivalent to one or more boxes or portions thereof in the illustrated figures are conceivable.

[0075] While various arrow and line types may be used in flowcharts and / or block diagrams, they should be understood not to limit the scope of the corresponding embodiments. In fact, some arrows or other connectors may be used only to indicate the logical flow of the depicted embodiment. For example, an arrow may indicate a wait or monitoring period of unspecified duration between enumeration steps in a depicted embodiment. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, may be implemented by a system based on dedicated hardware or a combination of dedicated hardware and code that performs the specified function or action.

[0076] The description of the elements in each figure can be referenced to the elements in the preceding figures. Throughout all figures, the same numbers refer to the same elements, including alternative embodiments of the same elements.

[0077] AI / ML models are data-driven algorithms that use AI / ML techniques to generate output sets based on input sets.

[0078] AI / ML model delivery is a general term that refers to delivering an AI / ML model from one entity to another in any way. Note that "entity" can mean network node / function (e.g., gNB, LMF, etc.), UE, proprietary server, etc.

[0079] AI / ML model inference is the process of using a trained AI / ML model to produce an output set based on an input set.

[0080] AI / ML model testing is a sub-process of training that evaluates the performance of the final AI / ML model using a different dataset than that used for model training and validation. Unlike AI / ML model validation, testing does not assume subsequent tuning of the model.

[0081] AI / ML model training is the process of training an AI / ML model in a data-driven manner [by learning input / output relationships] and obtaining a trained AI / ML model for inference.

[0082] AI / ML model delivery is conducted via the air interface in a manner opaque to 3GPP signaling. The delivery consists of parameters representing a model structure known at the receiving end, or a new model with parameters. The delivery may include a complete model or a partial model.

[0083] AI / ML model validation is a subprocess of training that evaluates the quality of an AI / ML model using a different dataset than the one used for model training. This helps in selecting model parameters that generalize to datasets other than those used for model training.

[0084] Data collection is the process by which network nodes, management entities, or user-defined users (UEs) collect data for the purpose of AI / ML model training, data analysis, and inference.

[0085] Federated learning / federated training is a machine learning technique that trains AI / ML models across multiple decentralized edge nodes (e.g., UE, gNB), with each node performing local model training using local data samples. This technique requires the models to interact multiple times, but does not require exchanging local data samples.

[0086] Function identification is the process / method of identifying AI / ML functions to achieve a common understanding between the NW and UE. Note that information about AI / ML functions can be shared during function identification. The location where AI / ML functions reside depends on the specific use case and sub-use case.

[0087] Model activation means enabling AI / ML models for specific AI / ML-supporting features.

[0088] Model deactivation means disabling the AI / ML model for specific AI / ML-supporting features.

[0089] Model download refers to the transfer of a model from the network to the UE.

[0090] Model identification is the process / method of identifying AI / ML models to achieve a common understanding between the NW and UE. The process / method of model identification may or may not be applicable, and information about the AI / ML model can be shared during model identification.

[0091] Model monitoring is the process of monitoring the inference performance of AI / ML models.

[0092] Model parameter update is the process of updating the model parameters.

[0093] Model selection is the process of choosing an AI / ML model to activate based on the same AI / ML-supporting features from among multiple models. Model selection may or may not be performed concurrently with model activation.

[0094] Model switching involves deactivating the currently active AI / ML model and activating different AI / ML models for specific AI / ML-supporting features.

[0095] Model update is the process of updating the model parameters and / or model structure.

[0096] Model upload is the transfer of the model from the UE to the network.

[0097] A network-side (AI / ML) model is an AI / ML model in which inference is performed entirely at the network.

[0098] Offline field data is data collected from the field and used for offline training of AI / ML models.

[0099] Offline training is an AI / ML training process in which a model is trained on a collected dataset, and the trained model is later used or delivered for inference. Note that this definition is for reference only. There may be cases that do not perfectly fit this definition but can still be classified as offline training according to accepted conventions.

[0100] Online field data refers to data collected from the field and used for online training of AI / ML models.

[0101] Online training is an AI / ML training process in which the model used for inference is trained (near) real-time (usually continuously) as new training samples arrive. It's worth noting that the concepts of (near) real-time and non-real-time depend on the context and are related to the inference timescale. This definition is for guidance only.

[0102] There may be cases that don't perfectly fit the definition but can still be categorized as online training according to accepted conventions. Note that fine-tuning / retraining can be done via online or offline training. This note can be removed when we define the term "fine-tuning".

[0103] Reinforcement learning (RL) is the process of training an AI / ML model in an environment in which the model interacts, based on inputs (also known as states) and feedback signals (also known as rewards) generated by the model's outputs (also known as actions).

[0104] Semi-supervised learning is the process of training a model using a mixture of labeled and unlabeled data.

[0105] Supervised learning is the process of training a model based on the input and its corresponding labels.

[0106] A two-sided (AI / ML) model is a paired AI / ML model on which joint inference is performed, where joint inference includes AI / ML inference performed jointly across the UE and the network, i.e., the first part of the inference is performed by the UE first, and then the remaining part is performed by the gNB, or vice versa.

[0107] The UE-side (AI / ML) model is an AI / ML model in which inference is performed entirely at the UE.

[0108] Unsupervised learning is the process of training a model without using labeled data.

[0109] From a 3GPP perspective, proprietary format models are vendor / device-specific proprietary format ML models. They are not mutually identifiable across vendors and hide model design information from other vendors when shared.

[0110] From 3GPP's perspective, open format models are formatted ML models that are mutually identifiable across vendors and allow interoperability. They are mutually identifiable between vendors and do not hide model design information from other vendors when shared.

[0111] The detailed description set forth below with reference to the accompanying drawings is intended as a description of various configurations and is not intended to represent only the configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. For example, although 3GPP terms from, for example, 5G NR may be used in this disclosure to exemplify embodiments herein, this should not be considered as limiting the scope of this disclosure.

[0112] This disclosure relates to a wireless communication system, which may be, for example, a 5G NR wireless communication system. More specifically, it refers to a RAN (Radio Access Network) within the wireless communication system for exchanging data with a UE via radio signals. For example, the RAN may transmit data to the UE (Downlink DL), such as data received from the core network (CN). The RAN may also receive data from the UE (Uplink UL), which may be forwarded to the CN.

[0113] In the illustrated example, the RAN includes a base station (BS). Of course, the RAN may include more than one BS to increase the coverage of the wireless communication system. Depending on the implemented wireless communication standard, each of these BSs may be referred to as an NB, eNodeB (or eNB), gNodeB (or gNB in ​​the case of a 5G NR wireless communication system), access point, etc.

[0114] The UE is located within the coverage area of ​​the BS. The coverage area of ​​the BS corresponds, for example, to the region where the UE can decode the PDCCH transmitted by the BS.

[0115] Examples of wireless devices suitable for implementing any of the methods discussed in this disclosure at the UE correspond to devices that provide wireless connectivity to a RAN (Radio Range) of a wireless communication system and can be used to exchange data with that RAN. Such wireless devices may be included in the UE. The UE may be, for example, a cellular phone, a wireless modem, a wireless communication device, a handheld device, a laptop computer, etc. The UE may also be an Internet of Things (IoT) device, such as a wireless camera, a smart sensor, a smart meter, smart glasses, a (manned or unmanned) vehicle, a GPS device, etc., or any other device capable of running applications that require the exchange of data with a remote receiver via a wireless device.

[0116] The wireless device includes one or more processors and one or more memories. The one or more processors may include, for example, a central processing unit (CPU), a digital signal processor (DSP), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), etc. The one or more memories may include any type of computer-readable volatile and non-volatile memory (magnetic hard disk, solid-state drive, optical disk, electronic storage, etc.). The one or more memories may store a computer program product, which is in the form of a set of program code instructions, executed by the one or more processors to implement all or part of the steps of a method for exchanging data performed on the UE side according to any embodiment of the embodiments disclosed herein.

[0117] The wireless device may also include a main radio (MR) unit. The MR unit corresponds to the main wireless communication unit of the wireless device, which is used to exchange data with the BS of the RAN using radio signals. The MR unit may implement one or more wireless communication protocols and may be, for example, a 3G, 4G, 5G, NR, WiFi, WiMax, or other transceivers. In a preferred embodiment, the MR unit corresponds to a 5G NR wireless communication unit.

[0118] The following explanation provides a detailed description of the mechanism for pre-configuring and signaling specific information about online model training by configuring a set of UE behaviors. AI / ML-based technologies are currently applied to many different applications, and based on observed potential benefits, 3GPP has also begun its technology research for application to multiple use cases. The AI / ML lifecycle can be broken down into several phases, such as data collection / preprocessing, model training, model testing / validation, model deployment / update, and model monitoring, each of which is equally important for achieving the target performance of any particular model. One of the challenging issues when applying AI / ML models to any use case or application is managing the lifecycle of the AI / ML model. This is primarily because data / model drift occurs during model deployment / inference, and it can lead to a degrade in AI / ML model performance. Fundamentally, statistical changes in the dataset occur after the model is deployed, and the model's inference capabilities are also affected by unseen data as input. Similarly, the statistical properties of the dataset and the relationship between the inputs and outputs of the trained model can change with drift. In this context, model training or retraining is one of the key issues in maintaining model performance, as model performance, such as inference and / or training, depends on different model execution environments with varying configuration parameters. To address this issue, collaboration between the UE and gNB is crucial for tracking model performance and reconfiguring models to suit different environments. Since model performance cannot be consistently maintained due to drift, AI / ML models require model monitoring after deployment, followed by updated feedback to retrain / update the model or select an alternative model.

[0119] When deploying wireless communication networks that support AI / ML models, it is crucial to consider how to use reconfiguration for wireless devices to handle active AI / ML models during operations such as model training, inference, and updates. For AI / ML model operations on one-sided or two-sided models between the network and the UE, model training performance (e.g., training in LCM) may degrade depending on changes in ML conditions.

[0120] In this approach, multiple types of model training operations are configured, enabling two or more types of model training operations to collaboratively train the same target model. Grouping of target UEs for model training can be based on device training capability information for specific model training operation types, where, depending on network-side decisions or UE-autonomous decisions, a UE can be assigned to operate two or more types of model training operations. Two or more model training operation types can be enabled simultaneously to perform single-sided or dual-sided model training between the network side and the UE side.

[0121] Model training updates can be exchanged between different UE groups. N-bit signaling (e.g., 1 bit) is used to enable selective or combined model training. Specifically, selective model training allows the network to choose a single type of model training operation to activate between the network and the UE. For example, when performing a type 1 model training operation, it can be switched to a type 2 model training operation based on pre-configured thresholds or metric information. On the other hand, combined model training allows the network to choose multiple types of model training operations to activate between the network and the UE. For example, two or more different types of model training operations can be executed in parallel, allowing the model training information exchanged across those training types to be combined for model training. The specific content of the training update information may vary depending on the different model training operation types and / or application scenarios.

[0122] However, the exchange of training update information is configured using ML configuration information via L1 / L2 or RRC signaling. The mapping between the associated parameter value set and model training operation type used for information exchange is preset based on ML deployment scenarios with different model applications / functions. The mapping table can be sent via system information or dedicated RRC signaling. Furthermore, any updates to the mapping table can be indicated via L1 / L2 or RRC signaling. A set of model training operation types can be configured by including single-sided and dual-sided model scenarios, where model training can be performed jointly or independently between the network side and the UE side. The parameter value set associated with each model training operation type can also be predefined in the mapping table, where the parameter values ​​can be related to or be a part of {input dataset / features, training output, target / reference training model, etc.}.

[0123] On the UE side, the UE provides the configured parameter value information to the network side based on the mapping table.

[0124] Different model training operation types can be determined for different UE groups.

[0125] Network-side coordination can enable combinations of model training types for model training across multiple UE groups. To enable a specific model training operation type for each UE group, an indication message is sent via L1 / L2 or RRC signaling using broadcast or multicast transmission. Based on pre-configured model training operation types, a single network-side model is enabled to perform bilateral model training with a UE group that has a UE-side model. A UE can enable two or more model training operation types to support the configured bilateral model training. The number of UE groups can be flexible, depending on the target UE and its associated UE-side models available. Based on pre-configured model training operation types, multiple network-side models can be enabled to perform bilateral model training with a UE group that has a UE-side model. In this scenario, the configured multiple network-side models can be located in the same network-side location or distributed in different locations via network interface connections. Each network-side model can activate a specific model training operation type for the identified UE group. Model training can be performed collaboratively across different network-side models associated with their own UE groups.

[0126] The mapping relationship between the associated parameter value set used for information exchange and the model training operation type is preset on the network side based on ML deployment scenarios with different model applications / functions. The grouping of target UEs for model training based on device training capability information can also be determined by the network side. Two or more model training operation types can be enabled simultaneously to perform single-sided or dual-sided model training between the network side and the UE side. The UE receives ML model training configuration information from the network side, including the mapping relationship between the model training operation type and the associated parameter value set used for information exchange. After activating a specific model training operation type, the UE provides the configured parameter value information to the network side based on the mapping table.

[0127] Figure 1 This diagram illustrates an exemplary block diagram for exchanging training update information between Type 1 and Type 2 model training operations. In this example, two model training operations are enabled in parallel, such that training update information from Type 1 is provided to Type 2 model training operation, and vice versa. The specific content of the training update information may vary depending on the different model training operation types and / or application scenarios. However, the exchange of training update information is configured using ML configuration information via L1 / L2 or RRC signaling.

[0128] Figure 2An exemplary table illustrating the mapping between associated parameter value sets and model training operation types used for information exchange is provided. In this example, the mapping between associated parameter value sets and model training operation types for information exchange is pre-defined based on ML deployment scenarios with different model applications / functions. The mapping table can be sent via system information or dedicated RRC signaling. Furthermore, any updates to the mapping table can be indicated via L1 / L2 or RRC signaling. A set of model training operation types can be configured by including single-sided and dual-sided model scenarios, where model training can be performed jointly or independently between the network side and the UE side. A set of parameter values ​​associated with each model training operation type can also be predefined in the mapping table, where parameter values ​​may be related to or a part of {input dataset / features, training output, target / reference training model, etc.}. On the UE side, the UE provides the configured parameter value information to the network side based on the mapping table.

[0129] Figure 3 An exemplary block diagram is shown for assigning model training types to multiple UE groups. In this example, different model training operation types can be determined for different UE groups. Network-side coordination can enable which combinations of model training types to perform model training across multiple UE groups. To enable a specific model training operation type for each UE group, an indication message is sent via L1 / L2 or RRC signaling using broadcast or multicast transmission.

[0130] Figure 4 The diagram illustrates an exemplary block diagram for enabling model training types for a single network-side model and multiple UE-side models. In this example, a single network-side model is enabled to perform bilateral model training with a group of UEs that have UE-side models, based on pre-configured model training operation types. A UE may enable two or more model training operation types to support the configured bilateral model training. The number of UE groups can be flexible, depending on the target UE and its associated UE-side models available.

[0131] Figure 5 This diagram illustrates an exemplary block diagram for enabling model training types for multiple network-side models and multiple UE-side models. In this example, multiple network-side models can be enabled to perform bilateral model training with a group of UEs that have UE-side models, based on pre-configured model training operation types. In this scenario, the configured multiple network-side models can be located in the same network-side location or distributed in different locations via network interface connections. Each network-side model can activate a specific model training operation type for a determined UE group. Model training can be conducted collaboratively across different network-side models associated with their own UE groups.

[0132] Figure 6An exemplary flowchart illustrating the configuration of model training operation types on the network side is shown. In this example, the mapping between the associated set of parameter values ​​used for information exchange and the model training operation types is preset on the network side based on ML deployment scenarios with different model applications / functions. The grouping of target UEs for model training based on device training capability information can also be determined by the network side. Two or more model training operation types can be enabled simultaneously to perform one-sided or two-sided model training between the network side and the UE side.

[0133] Figure 7 An exemplary flowchart illustrating the configuration of model training operation types on the UE side is shown. In this example, the UE receives ML model training configuration information from the network side, including a mapping between the associated set of parameter values ​​and model training operation types for information exchange. After activating a specific model training operation type, the UE provides the configured parameter value information to the network side based on the mapping table.

[0134] This application provides a fundamental mechanism for supporting AI / ML interoperability and data flow in radio access network collaboration, particularly in supporting model operations based on model training operation types. Based on the proposed method, the behavior of gNB-UE for supporting AI / ML operations in wireless communication using configured model training operation types is significantly improved in potential scenarios.

Claims

1. A method for configuring multiple types of model training operations in a wireless communication system using multi-training model operation signaling executed by a network node, comprising: • For specific types of model training operations, group the target UEs; • Define the information to be exchanged between model training operations; • Determine the specific bits of the signaling used for selective / combinatorial model training; • Pre-set mapping information to support various types of model training operations.

2. The method according to claim 1, wherein two or more types of model training operations are performed collaboratively to train the same target model.

3. The method according to any one of the preceding claims, wherein the grouping of the target UE for model training is based on device training capability information for a specific model training operation type.

4. The method according to any one of the preceding claims, wherein the UE is assigned to operate two or more types of model training operations, depending on network-side decision or UE autonomous decision.

5. The method according to any one of the preceding claims, wherein two or more model training operation types are enabled simultaneously to perform single-sided or dual-sided model training between the network side and the UE side.

6. The method according to any one of the preceding claims, wherein model training updates are exchanged with each other across different UE groups.

7. The method according to any one of the preceding claims, wherein N-bit signaling is used to enable selective model training or combined model training.

8. The method according to any one of the preceding claims, wherein selective model training allows for the selection of a single type of model training operation to be activated between the network and the UE.

9. The method according to any one of the preceding claims, wherein combined model training allows multiple types of model training operations to be activated between the network and the UE, enabling two or more different types of model training operations to be executed in parallel.

10. The method according to any one of the preceding claims, wherein the model training information exchanged across those training types is combined for model training.

11. The method according to any one of the preceding claims, wherein the specific content of the training update information can vary depending on different model training operation types and / or application scenarios.

12. The method according to any one of the preceding claims, wherein the exchange of training update information is configured using ML configuration information via L1 / L2 or RRC signaling.

13. The method according to any one of the preceding claims, wherein the mapping relationship between the associated set of parameter values ​​for information exchange and the model training operation type is preset to be available based on ML deployment scenarios with different model applications / functions via system information or dedicated RRC signaling.

14. The method according to any of the preceding claims, wherein any update to the mapping table (including the associated set of parameter values ​​and model training operation type for information exchange) is indicated via L1 / L2 or RRC signaling.

15. The method according to any one of the preceding claims, wherein a set of model training operation types is configured by including single-sided model scenarios and dual-sided model scenarios, such that model training is performed jointly or independently between the network side and the UE side.

16. The method according to any one of the preceding claims, wherein the set of parameter values ​​associated with each model training operation type is predefined in a mapping table.

17. The method according to any one of the preceding claims, wherein the UE provides the network side with configured parameter value information related to each model training operation type based on a mapping table format.

18. The method according to any one of the preceding claims, wherein different model training operation types are determined for different UE groups.

19. The method according to any one of the preceding claims, wherein the network side is capable of coordinating which combinations of model training types are enabled across multiple UE groups for model training.

20. The method according to any one of the preceding claims, wherein an indication message is transmitted via broadcast or multicast transmission using L1 / L2 or RRC signaling to enable a specific model training operation type for each UE group.

21. The method according to any one of the preceding claims, wherein a single network-side model is enabled to perform bilateral model training with a group of UEs having a UE-side model based on a pre-configured model training operation type.

22. The method according to any one of the preceding claims, wherein the UE is capable of enabling two or more model training operation types to support configured two-sided model training.

23. The method according to any one of the preceding claims, wherein the number of UE groups is flexible to determine the target UE and its associated UE-side model for use.

24. The method according to any one of the preceding claims, wherein multiple network-side models are enabled to perform two-sided model training with a group of UEs having UE-side models.

25. The method according to any one of the preceding claims, wherein multiple network-side models configured to support a specific model training operation type are located at the same network-side location or distributed at different locations via network interfaces.

26. The method according to any one of the preceding claims, wherein each network-side model is capable of activating a specific model training operation type for the determined UE group.

27. The method according to any one of the preceding claims, wherein collaborative model training is performed across different network-side models, the different network-side models being associated with their own UE groups.

28. The method according to any one of the preceding claims, wherein the mapping relationship between the associated set of parameter values ​​for information exchange and the model training operation type is preset on the network side based on ML deployment scenarios with different model applications / functions.

29. The method according to any one of the preceding claims, wherein the grouping of target UEs for model training based on device training capability information is determined by the network side.

30. The method according to any one of the preceding claims, wherein two or more model training operation types are enabled simultaneously to perform single-sided or dual-sided model training between the network side and the UE side.

31. The method according to any one of the preceding claims, wherein the UE is capable of receiving ML model training configuration information from the network side, including a mapping relationship between the associated set of parameter values ​​and the model training operation type for information exchange.

32. The method according to any one of the preceding claims, wherein after activating a specific model training operation type, the UE is able to provide the configured parameter value information to the network side based on a mapping table.

33. An apparatus for configuring multi-training model operation signaling for multiple types of model training operations in a wireless communication system, the apparatus comprising a wireless transceiver, a processor coupled to a memory, wherein computer program instructions are stored in the memory, the instructions being configured to implement the steps as claimed in claims 1 to 32.

34. A user equipment comprising the device according to claim 33.

35. A gNB comprising the device according to claim 33.

36. A wireless communication system for configuring multi-training model operation signaling for multiple types of model training operations, wherein the wireless communication system includes at least one user equipment according to claim 34 and at least one gNB according to claim 35, wherein the user equipment and the gNB each include a processor coupled to a memory in which computer program instructions are stored.

Images