Method for advanced model adaptation of radio access network

Abstract

The present disclosure describes a method for model adaptation / handover for an AI / ML-based model in a wireless mobile communication system including a base station (e.g., a gNB) and a mobile station (e.g., a UE). Threshold-based mapping relationships are configured to enable model adaptation using trigger information in accordance with dynamic changes in the device-side model operating environment.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to AI / ML based model adaptation, in which techniques for pre-configuration and signaling mapping relationship information specific content are proposed, which mapping relationship information realizes model adaptation with the association between model and other index values. BACKGROUND

[0002] In 3GPP, as one of the selected study items of approved Release 18 package is AI / ML (Artificial Intelligence / Machine Learning), as described in the related document (RP-213599) submitted in 3GPP TSG RAN Meeting #94e. The formal name of the AI / ML study item is “Study on AI / ML for NR Air Interface”, and the current RAN WG1 and WG2 are actively developing specifications. The goal of this study item is to identify a common AI / ML framework and areas to use AI / ML based techniques and use cases to gain benefits.

[0003] According to 3GPP, the main goal of this study item is to study an AI / ML framework for the air interface by considering performance, complexity, and potential specification impact with target use cases. In particular, AI / ML models, terminology, and description for identifying common and specific characteristics of the framework will be one of the key work scopes. Regarding the AI / ML framework, investigations into various aspects are being considered, and one of the key items is about the lifecycle management of AI / ML models, in which multiple phases are mandatorily included for model training, model deployment, model inference, model monitoring, model update, etc.

[0004] Earlier, in 3GPP TR 37.817 of Release 17 named Study on enhancement for Data Collection for NR and EN-DC, UE mobility was also considered as one of the AI / ML use cases, and one of the scenarios for model training / inference is that both functions are located within the RAN node. Subsequently, in Release 18, a new work item “Artificial Intelligence (AI) / Machine Learning (ML) for NG-RAN” was initiated to specify data collection enhancements and signaling support within the existing NG-RAN interfaces and architecture.

[0005] For the above positive standardization work, there is currently no specification for the signaling method or gNB-UE behavior definition about supporting AI / ML model adaptation or model switching. When the activated model cannot run correctly on the UE with available resources, using mismatched models on both sides or one side can cause model performance degradation. Based on the above, it is observed that the model transfer signaling used by the network side for model adaptation / switching can be greatly increased in order to adapt to different dynamic conditions associated with UE capability changes. Therefore, due to the delay in traditional UE capability reporting, the applicable model at the UE can not match the reported UE capability.

[0006] EP3543917A1 describes training a deep neural network (DNN) using a low-precision method, i.e., a method using low-precision weights.

[0007] WO2020234902A1 describes a radio mapping architecture for applying machine learning techniques to wireless radio access networks.

[0008] WO2022033804A1 describes splitting an AI / ML model into multiple subparts and forming an aggregated block.

[0009] US20090170552A1 describes a switching profile method for a mobile device that utilizes detection of predetermined conditions. SUMMARY

[0010] The described problem is solved by embodiments of the present application. The first aspect is an advanced model adaptation method for a radio access network, which is implemented by generating a preconfigured mapping relationship between applicable sub-model IDs and different thresholds, the method comprising: predefining thresholds to match the activation triggers of different AI / ML sub-models; the sub-model ID is separated from the general model ID, where the sub-model can have lower model complexity / size compared to the general model ID; attribute data {UE device resource specification, model configuration, site / scene, application} for threshold calculation can be defined simultaneously considering different use cases / applications and implementation-specific environments and UE capabilities; the listed sub-models can be dynamically configured with the pre-defined parameter set change information sent from the network, where the determined model ID can be a macro model, and the listed multiple sub-models can be micro models, the macro model is a full-function model with the highest complexity / size, and the micro model is a partial-function model with lower complexity / size; the determined model ID itself can be one of the indexed sub-models depending on the actual use case; the size of the mapping table or the number of sub-models can adapt to specific model operation applications and / or environments.

[0011] In some embodiments of the method according to the first aspect, the method is characterized in that the UE monitors the model operation and the device resource status supporting the model operation together to detect a preconfigured threshold for triggering, and when the triggering is enabled with a specific threshold, the currently operating model autonomously switches to a sub-model associated with the matching threshold.

[0012] In some embodiments of the method according to the first aspect, the method is characterized in that the network side provides the mapping relationship and the triggering information with related configurations, including defining / generating the threshold based on the attribute data {UE device resource specification, model configuration, site / scene, application}; different numbers of mapping relationships and triggering information can be formed considering the UE capability at the same time for different use cases / applications and implementation-specific environments.

[0013] According to the second aspect, the disclosure relates to a method for forming a full model (fML) and a partial model (pML) for a radio access network, the method comprising: the network generating two model categories, so that the fML is an original model with a complete feature set, and the pML is a simplified model with lower complexity and / or a smaller feature set, for applying any specific model to the UE. The fML and the pML are preconfigured, so that for the partial model, the model complexity is lower and the feature set size is smaller, wherein the model configuration parameters such as the number of layers and the feature input can be adjusted to determine the limited set of full / partial models. The selection of the model from the fML and the pML depends on the UE ML capacity state to match the target model operation, so that any pML can run in the case of reduced UE device available resources. The scalable model structure supports models with different complexity by using parameter configurations.

[0014] In some embodiments of the method according to the first aspect and the second aspect, the method is characterized in that different numbers of UE groups can be configured, so that each group of UEs runs a model matching the reported UE ML capability information.

[0015] According to the third aspect, the disclosure relates to a device for a method of generating a preconfigured mapping relationship between applicable sub-model IDs and different thresholds and forming a full model (fML) and a partial model (pML) for a radio access network, the device comprising a wireless transceiver, a processor coupled with a memory, the memory storing computer program instructions configured to implement the steps of the method of the first aspect and the second aspect.

[0016] According to the fourth aspect, the disclosure relates to a user equipment comprising the device according to any one of the embodiments of the first aspect and the second aspect.

[0017] According to a fifth aspect, the disclosure relates to a base station comprising a device according to any one of the embodiments of the first and second aspects.

[0018] According to a sixth aspect, the disclosure relates to a wireless communication system, wherein a base station (gNB) comprises a processor coupled with a memory having stored therein computer program instructions configured to implement the steps of the methods of the first and second aspects, and wherein a user equipment (UE) comprises a processor coupled with a memory having stored therein computer program instructions configured to implement the steps of the methods of the first and second aspects. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is an exemplary table of threshold-based mapping relationship.

[0020] Figure 2 is an exemplary signaling of autonomous sub-model switching by the UE.

[0021] Figure 3 is a flowchart of mapping relationship configuration and triggering information at the network side.

[0022] Figure 4 is a flowchart of handling autonomous sub-model switching by the UE.

[0023] Figure 5 is an exemplary block diagram of using full / partial model. DETAILED DESCRIPTION

[0024] The detailed description set forth below, in connection with the appended drawings, is intended as a description of various configurations and is not intended to represent the only configurations in which the concepts described herein can 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 can be practiced without these specific details. In particular, although terminology from 3GPP 5G NR can be used in this disclosure to exemplify the embodiments herein, this should not be seen as limiting the scope of the application.

[0025] Some embodiments of the embodiments contemplated herein will now be described more fully with reference to the accompanying drawings. Other embodiments, however, are contained within the scope of the subject matter disclosed herein, the disclosed subject matter should not be construed as limited to the embodiments set forth herein; rather, they are provided by way of example to convey the scope of the subject matter to those skilled in the art.

[0026] Generally, all terms used herein are to be interpreted according to their ordinary meaning in the technical field of use, unless a different meaning is clearly given and / or is implied by the context in which it is used. All references to a / an / the item are to be interpreted openly as referring to one or more items unless otherwise indicated by particular circumstances. Unless otherwise stated, steps of any method disclosed herein do not have to be performed in the exact order disclosed. Any feature of any of the embodiments disclosed herein can be applied to any other embodiment, where suitable. Likewise, any advantage of any of the embodiments can apply to any other embodiment, and vice- versa. Other objectives, features and advantages of the enclosed embodiments will be apparent from the following description.

[0027] In some embodiments, the more general term "network node" can be used, and the term can correspond to any type of radio network node or any network node that communicates with a UE (directly or via another node) and / or with another network node. Examples of network nodes are NodeB, MeNB, ENB, a network node belonging to a MCG or SCG, base station (BS), multi-standard radio (MSR) radio node such as MSR BS, eNodeB, gNodeB, network controller, radio network controller (RNC), base station controller (BSC), relay, donor node controlling relay, base transceiver station (BTS), access point (AP), transmission points, transmission nodes, RRU, RRH, nodes in distributed antenna system (DAS), core network node (e.g., mobile switching center (MSC), mobility management entity (MME), etc.), operation and maintenance (O&M) system, operation support system (OSS), self-optimizing network (SON), positioning node (e.g., Evolved Serving Mobile Location Center (E-SMLC)), minimization of drive testing (MDT) system, test equipment (physical node or software), etc.

[0028] In some embodiments, the non-limiting term user equipment (UE) or wireless device can be used and can refer to any type of wireless device communicating with a network node and / or with another UE in a cellular or mobile communication system. Examples of UE are target device, device to device (D2D) UE, machine type UE or UE capable of machine to machine (M2M) communication, PDA, PAD, Tablet, mobile terminal, smart phone, laptop mounted with embedded equipment (LEE), laptop mounted with equipment (LME), USB dongle, UE category Ml, UE category M2, ProSe UE, V2V UE, V2X UE, etc.

[0029] Additionally, the terms such as base station / gNodeB and UE should be considered as non-limiting and in particular do not imply some hierarchical relation between the two; generally, “gNodeB” can be considered as device 1 and “UE” can be considered as device 2 and these two devices communicate with each other over some radio channel. Also, in the following, transmitter or receiver can be gNodeB (gNB) or UE.

[0030] As those skilled in the art will appreciate, the aspects of the embodiments can be embodied as a system, device, method or program product. Accordingly, the embodiments can take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects.

[0031] For example, the disclosed embodiments can be implemented as hardware circuitry, including custom 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 or the like. As another example, the disclosed embodiments can include one or more physical or logical blocks of executable code, which may, for example, be organized as an object, procedure, or function.

[0032] Furthermore, embodiments can take the form of a program product embodied in one or more computer readable storage devices storing machine-readable code, computer readable code, and / or program code, referred to herein as code. The storage devices can be tangible, non-transitory, and / or non-transmission. The storage devices can not embody signals. In a certain embodiment, the storage devices only employ a signal to access the code.

[0033] Any combination of one or more computer-readable media can be utilized. The computer-readable media can be computer-readable storage media. The computer-readable storage media can be storage devices. Storage devices can be, for example and without limitation, electronic, magnetic, optical, electromagnetic, infrared, holographic, micromechanical, or semiconductor systems, apparatuses, or devices, or any suitable combination of the foregoing.

[0034] More specific examples (a non-exhaustive list) of the storage device would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory ("RAM"), a read-only memory ("ROM"), an erasable programmable read-only memory ("EPROM" or Flash memory), a portable compact disc read-only memory ("CD-ROM"), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer-readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0035] Code for carrying out operations for embodiments can be any number of lines and any combination of programming languages, including an object oriented programming language such as Python, Ruby, Java, Smalltalk, C++, or the like, and conventional procedural programming languages, such as the "C" programming language, or the like, and / or machine languages such as assembly languages. The code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network ("LAN"), wireless LAN ("WLAN"), or a wide area network ("WAN"), or the like, or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider ("ISP")). In some embodiments, electronic circuitry including, for example, a

[0036] Furthermore, the described features, structures, or characteristics of the embodiments can be combined in any suitable manner. In the following description, numerous specific details are provided, such as examples of programming, software modules, user selections, network transactions, database queries, database structures, hardware modules, hardware circuits, hardware chips, etc., to provide a thorough understanding of the embodiments. One skilled in the relevant art will recognize, however, that the embodiments can be practiced without one or more of the specific details, or with other methods, components, materials, and so forth. In other instances, well-known structures, materials, or operations are not shown or described in detail in order to avoid obscuring aspects of the embodiments. Reference throughout this specification to “an embodiment,” “embodiments,” or similar language means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, appearances of the phrase “in one embodiment,” “in embodiments,” and similar language throughout this specification may, but do not necessarily, all refer to the same embodiment, but mean “one or more but not all embodiments.” Unless otherwise noted, the terms “including,” “comprising,” “having,” and variations thereof are meant to encompass the items listed thereafter and equivalents thereof as well as additional items. Unless otherwise noted, the enumerated listing of items does not imply that any or all of the items are mutually exclusive. Unless otherwise expressly specified, the terms “a,” “an,” and “the” do not imply that a particular element is singular or that, alone and without duplicates, it is the only one of its kind.

[0037] The code can also be stored in a storage device that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the storage device produce an article of manufacture including instructions which implement the function / act specified in the flowchart and / or block diagram block or blocks.

[0038] The code can also be stored in a storage device that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the storage device produce an article of manufacture including instructions which implement the function / act specified in the flowchart and / or block diagram block or blocks.

[0039] The code can also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the code which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0040] The flow diagrams and / or block diagrams in the drawings are illustrations of architectures, functional processes and operations, and non-limiting embodiments, in accordance with various embodiments. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that

[0041] It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending on the functionality involved. Other steps and methods can be conceived that are equivalent in function, logic, or effect to those illustrated, with the scope of the present disclosure intended to include all such steps and methods.

[0042] Although various arrow types and line types can be employed in the flow diagrams and / or block diagrams, these are merely meant as an aid to better understand the various embodiments of the present disclosure. One having ordinary skill in the art will recognize that some of the arrows, as well as / other connectors between blocks, can be used to indicate one or more desired timing mechanisms. For example, arrows can indicate a desired sequence of operation, flows and / or priority in processes having generally parallel concurrent flow sequences.

[0043] The description of elements in each of the figures can refer to elements in previous figures. The same numbers in all the figures refer to the same elements, including alternative embodiments of the same element.

[0044] The detailed description set forth below in connection with the appended drawings is intended as a description of various configurations and is not intended to represent the only configurations in which the concepts described herein can be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of various concepts. However, it will be apparent to those skilled in the art that these concepts can be practiced without these specific details. For example, while 3GPP terminology from, e.g., 5G NR, can be used in this disclosure to exemplify embodiments herein, this should not be seen as limiting the scope of the present disclosure.

[0045] The present disclosure relates to a wireless communication system, which can be, e.g., a 5G NR wireless communication system. More specifically, it denotes a RAN of a wireless communication system for exchanging data with UEs via radio signals. For example, the RAN can transmit data (downlink, DL) to UEs, e.g., data received from a core network (CN). The RAN can also receive data (uplink, UL) from UEs, which can be forwarded to the CN.

[0046] In the illustrated example, the RAN comprises one base station, BS. Of course, the RAN can comprise more than one BS to increase the coverage of the wireless communication system. Each of these BSs can be referred to as an NB, an eNodeB (or eNB), a gNodeB (or gNB in the case of a 5G NR wireless communication system), an access point, etc., depending on the wireless communication standard that is implemented.

[0047] The UE is located in the coverage of the BS. The coverage of the BS corresponds for example to an area in which the UE can decode a PDCCH transmitted by the BS.

[0048] Examples of a wireless device suitable for implementing any of the methods performed at the UE discussed in the present disclosure correspond to devices providing wireless connectivity with a RAN of a wireless communication system and that can be used to exchange data with said RAN. Such a wireless device can be comprised in a UE. The UE can be for example a cellular phone, a wireless modem, a wireless communication device, a handheld device, a laptop or the like. The UE can also be an Internet of Things (IoT) equipment such as a wireless camera, a smart sensor, a smart meter, smart glasses, a vehicle (manned or unmanned), a global positioning system device, etc., or any other equipment that can run an application requiring the exchange of data with a remote recipient via a wireless device.

[0049] The wireless device comprises one or more processors and one or more memories. The one or more processors can comprise 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 can comprise any type of computer-readable volatile and non-volatile memory (magnetic hard disk, solid state disk, optical disk, electronic memory, etc.). The one or more memories can store a computer program product in the form of sets of program code instructions to be executed by the one or more processors to implement all or part of the steps of the method for exchanging data performed at the UE side according to any of the embodiments disclosed herein.

[0050] The wireless device can also comprise a main radio, MR, unit. The MR unit corresponds to the main wireless communication unit of the wireless device for exchanging data with the BSs of the RAN using radio signals. The MR unit can implement one or more wireless communication protocols and can be for example a 3G, 4G, 5G, NR, WiFi, WiMax, etc. transceiver, etc. In a preferred embodiment, the MR unit corresponds to a 5G NR wireless communication unit.

[0051] The following explanation will provide detailed description on the mechanism of provisioning and signaling specific information on mapping relationship information using the association between the model for dataset information and other index values. AI / ML based techniques are currently being applied to many different applications and 3GPP also started its technical study based on the observed potential benefits to be applied to multiple use cases. The AI / ML life cycle can be split into several phases such as data collection / preprocessing, model training, model testing / validation, model deployment / update, model monitoring, model switching / selection, etc., where each phase is equally important to achieve the target performance of any specific model.

[0052] One of the challenging issues in applying AI / ML models to any use case or application is to manage the life cycle of AI / ML models. This is mainly because of the data / model drift that occurs during model deployment / inference and it results in the degradation of the performance of AI / ML models. Fundamentally, there is a change in the statistics of the dataset after the model is deployed and the model inference capability is also affected by the unseen data as input. In a similar aspect, the statistical properties of the dataset and the relationship between the input and output of the trained model can change with the occurrence of drift.

[0053] When deploying a wireless communication network that supports AI / ML models, it is important to consider how to handle AI / ML model / dataset delivery under operations such as model training, inference, monitoring, update, etc. When AI / ML models are delivered to the UE, if the model size and / or complexity is higher than the ML capability supported by the UE, the delivered model cannot run completely. For example, each UE has many different levels of ML capability. If the minimum ML capability of all UEs cannot be guaranteed due to limited device resources such as computing power / memory size / battery power consumption, etc., the general model to be delivered can be valid for some UEs but not for others. The ML support capability at the UE can change dynamically because the applicable conditions (such as scenarios / sites) for any specific model / functionality are not static but dynamic. Therefore, when the UE ML capability is dynamic, depending on the UE ML capability state change, the configured model operation with specific functionality can be unstable and the performance quality degrades. The model delivery signaling at the network side can be greatly increased in order to adapt to different dynamic conditions associated with the UE capability change. Therefore, due to the delay in the traditional UE capability reporting, the applicable model at the UE can not match the reported UE capability.

[0054] As a new mechanism to solve this problem, a threshold-based mapping relationship between applicable sub-model IDs and different thresholds is pre-configured / provided by the network side. For example, when a threshold is triggered at the UE during model operation, the UE can autonomously switch between pre-configured sub-models that match the indicated threshold. The threshold-based mapping relationship and trigger configuration information are sent through system information or RRC signaling.

[0055] Figure 1 An exemplary table of threshold-based mapping relationships is shown. In this table, thresholds are predefined to match the activation triggers of different AI / ML sub-models. Sub-model IDs are separated from general model IDs, where sub-models can have lower model complexity / size compared to general model IDs. Thresholds are pre-configured to reflect the available resources of UE capabilities running the model, where each indexed sub-model can be correctly executed based on available UE capability resources according to the indication of thresholds. Thresholds can be defined for different use cases / applications and implementation-specific environments while taking into account UE capabilities such as computing power, memory size, and battery power, etc. (e.g., derived from usage attribute data {UE device resource specification, model configuration, site / scenario, application}), where although there are different types of devices with model support capabilities, general or standardized capability requirements for running models can be used for threshold calculation. Depending on different applicable conditions, this threshold-based mapping relationship needs to be updated and sent to the UE (e.g., through RRC signaling).

[0056] Based on the determined model ID for use, the listed sub-models can be dynamically configured with pre-defined parameter set variation information sent from the network, where the determined model ID can be a macro model and the listed multiple sub-models can be micro models, the macro model being a full-function model with the highest complexity / size and the micro model being a partial-function model with lower complexity / size. Depending on the actual use case, the determined model ID itself can be one of the indexed sub-models. Therefore, any separate model delivery for each sub-model is not needed, where any sub-model can be dynamically switched at the UE side based on the pre-defined parameter set variation information. The size of the mapping table or the number of sub-models can be adapted to the specific model operation application and / or environment.

[0057] Figure 2 An exemplary signaling of autonomous sub-model switching by the UE is shown. Without receiving model switching / adaption indication from the network side, the UE can autonomously switch / adapt the model according to the trigger rule based on the mapping table and pre-configured information about the sub-models. After sub-model switching / adaption, the UE sends the ML capability status and model switching update. This example is a method of implementing model switching / adaption based on event triggering. However, model switching / adaption can also be performed periodically using pre-configured sub-models.

[0058] Figure 3 A flow diagram showing mapping relationship configuration and triggering information at the network side. At the network side, the mapping relationship and triggering information are configured together. The threshold is predefined to match the activation triggering value of different AI / ML sub-models. The threshold is pre-configured to reflect the available resources of UE capability running the model, where each indexed sub-model can be correctly executed based on the available UE capability resources according to the indication of the threshold. The threshold can be defined for different use cases / applications and implementation-specific environments while considering UE capability (e.g., derived from usage attribute data {UE device resource specification, model configuration, site / scenario, application}).

[0059] Figure 4 A flow diagram showing UE processing autonomous sub-model switching. After receiving the mapping relationship configuration and triggering information, the UE monitors the model operation and device resource state supporting the model operation together to detect the pre-configured threshold for triggering. When the triggering is enabled with a specific threshold, the currently operating model switches to the sub-model associated with the matching threshold.

[0060] Figure 5 An exemplary block diagram showing the use of full / partial models. In this example, there is one full model (fML) and two partial models (pML). When any specific model is applied to the UE, the network generates two model categories so that the fML is the original model with a full feature set and the pML is a simplified model with lower complexity and / or a smaller feature set. Therefore, the main difference between the fML and the pML is that the model complexity is lower and the feature set size is smaller for the partial model, where the model configuration parameters such as the number of layers and feature input can be adjusted to determine the limited set of full / partial models. The selection of the model from the fML and the pML depends on the UE ML capacity state to match the target model operation, so that any pML can be run with reduced UE device available resources. In this example, the scalable model structure then supports models with different complexity by using parameter configurations. Based on the reported UE ML capability information, two or more groups of UEs that match the reported UE ML capability information can also be determined, such as UE group #1 that can support full model delivery and UE group #2 that can support partial model delivery. The number of UE groups can be configurable. Therefore, based on the scalable model set, UE grouping-based model adaptation can be applied.

[0061] Abbreviations

[0062] BWP Bandwidth Part

[0063] CBG Code Block Group

[0064] CLI Cross-Link Interference

[0065] CP Cyclic Prefix

[0066] CQI Channel Quality Indicator

[0067] CPU CSI Processing Unit

[0068] CRB Common Resource Block

[0069] CRC Cyclic Redundancy Check

[0070] CRI CSI-RS Resource Indicator

[0071] CSI Channel State Information

[0072] CSI-RS Channel State Information Reference Signal

[0073] CSI-RSRP CSI Reference Signal Received Power

[0074] CSI-RSRQ CSI Reference Signal Received Quality

[0075] CSI-SINR CSI Signal to Interference and Noise Ratio

[0076] CW Code Word

[0077] DCI Downlink Control Information

[0078] DL Downlink

[0079] DM-RS Demodulation Reference Signal

[0080] DRX Discontinuous Reception

[0081] EPRE Energy Per Resource Element

[0082] IAB-MT Integrated Access and Backhaul-Mobile Termination

[0083] L1-RSRP Layer 1 Reference Signal Received Power

[0084] LI Layer Indicator

[0085] MCS Modulation Coding Scheme

[0086] PDCCH Physical Downlink Control Channel

[0087] PDSCH Physical Downlink Shared Channel

[0088] PSS Primary Synchronization Signal

[0089] PUCCH Physical Uplink Control Channel

[0090] QCL Quasi Co-Location

[0091] PMI precoding matrix indicator

[0092] PRB physical resource block

[0093] PRG precoding resource block group

[0094] PRS positioning reference signal

[0095] PT-RS phase tracking reference signal

[0096] RB resource block

[0097] RBG resource block group

[0098] RI rank indicator

[0099] RIV resource indication value

[0100] RS reference signal

[0101] SCI sidelink control information

[0102] SLIV start and length indication value

[0103] SR scheduling request

[0104] SRS sounding reference signal

[0105] SS synchronization signal

[0106] SSS secondary synchronization signal

[0107] SS-RSRP SS reference signal received power

[0108] SS-RSRQ SS reference signal received quality

[0109] SS-SINR SS signal to interference and noise ratio

[0110] TB transport block

[0111] TCI transmission configuration indicator

[0112] TDM time division multiplexing

[0113] UE user equipment

[0114] UL uplink

Claims

1. A high-level model adaptation method for radio access networks, which achieves this by generating a pre-configured mapping relationship between applicable sub-model IDs and different thresholds, comprising: • Predefined thresholds to match the activation triggering of different AI / ML sub-models. • The sub-model ID is separated from the general model ID, where the sub-model can have lower model complexity / size compared to the general model ID. • It can define attribute data {UE device resource specifications, model configuration, site / scenario, application} for threshold calculation, taking into account different use cases / applications and implementation methods, as well as specific environments and UE capabilities. • The listed sub-models can be dynamically configured using predefined parameter set change information sent from the network. The determined model ID can be a macro-model, and the listed sub-models can be micro-models. The macro-model is the full-featured model with the highest complexity / size, while the micro-model is a partially functional model with lower complexity / size. • Depending on the actual use case, the determined model ID itself can be one of the indexed sub-models. • The size of the mapping table or the number of sub-models can be adapted to specific model operation applications and / or environments.

2. The method of claim 1, wherein the UE monitors the model operation and the device resource status supporting the model operation together to detect a pre-configured threshold for triggering, and when triggering is enabled using a specific threshold, the currently operating model autonomously switches to a sub-model associated with the matching threshold.

3. The method according to any of the preceding claims, wherein the network side provides the mapping relationship and triggering information with related configuration, including: • Thresholds are defined / generated based on attribute data {UE device resource specifications, model configuration, site / scenario, application}. • It can generate different numbers of mapping relationships and triggering information for different use cases / applications and specific environments, while taking into account UE capabilities.

4. A method for forming a complete model (fML) and a partial model (pML) for a radio access network, comprising: • The network generates two model categories: fML is the original model with the full feature set, and pML is a simplified model with lower complexity and / or a smaller feature set, for use in applying any particular model to the UE. • Pre-configure fML and pML to achieve lower model complexity and smaller feature set size for partial models, where model configuration parameters, such as the number of layers and feature inputs, can be adjusted to determine the finite set of the full / partial model. • The selection of a model from fML and pML depends on the UE ML capacity status to match the target model's operation, ensuring that any pML can operate even with reduced available resources on the UE device. • Scalable model architecture supports models with varying degrees of complexity through parameter configuration.

5. The method according to any of the preceding claims, wherein different numbers of UE groups can be configured such that each group of UEs runs a model that matches the reported UE ML capability information.

6. An apparatus for generating a pre-configured mapping relationship between applicable sub-model IDs and different thresholds, the apparatus comprising a wireless transceiver and a processor coupled to a memory, the memory storing computer program instructions configured to implement the steps of the method as claimed in claims 1 to 5.

7. A user equipment comprising the device according to claim 6.

8. A base station comprising the device according to claim 6.

9. A wireless communication system, wherein a gNB includes a processor coupled to a memory, the memory storing computer program instructions configured to implement the steps of the method as claimed in claims 1 to 5: The user equipment (UE) includes a processor coupled to a memory storing computer program instructions configured to implement the steps of the method as described in claims 1 to 5.

Images