Multi-model signaling procedures for RAN

Abstract

The present disclosure describes methods for utilizing information sharing based on a preconfigured AI / ML (artificial intelligence / machine learning)-based model in a wireless mobile communication system including a base station (e.g., gNB) and a mobile station (e.g., a UE). When an AI / ML model is applied to a radio access network, the signaling of the transmission / delivery of the model may be heavily congested due to the execution of multiple models with varying applications. Therefore, the operation of the model, e.g., transmission / retraining / updating of the model, between the network and the UE can be implemented by reducing the amount of multi-model signaling overhead.

Description

TECHNICAL FIELDThe present disclosure relates to sharing information based on an AI / ML-based model, wherein techniques for pre-configuring and signaling the specific information about the structure / characteristics of the model are presented.BACKGROUNDIn third generation partnership project (3GPP), one of the selected objects of investigation in the released release 18 packet is composed of artificial intelligence / machine learning (KI / ML) as described in related document (RP-213599) dealt with in meeting technical specification group (3GPP TSG) radio access network (RAN) #94e. The official name of the AI / ML subject of examination is "Study on AI / ML for NR Air Interface - Examination on AI / ML for NR Air Interface", and currently, the RAN WG1 (Working Group 1 - Working Group 1) and the WG2 actively operate on the specification. The goal of this subject matter is to identify a common AI / ML framework and areas where benefits are obtained using AI / ML-based techniques for use cases. According to 3GPP, the primary goal of this subject matter is to examine an AI / ML framework for an air interface with target use cases, taking into account performance, complexity, and a potential impact of the specification. In particular, the AI / ML model, terminology and description form one of the major work areas for identifying general and specific labels. Various aspects of the study are currently being considered with respect to the AI / ML framework, and one of the key points is life cycle management of an AI / ML model, wherein multiple stages for model training, model deployment, model inference, model monitoring, model update, etc. are included as obligatory. Previously, in 3GPP TR 37.817 for release 17, entitled Study to improve data collection for NR and EN-DC, UE (user equipment) Study on enhancement for Data Collection for NR and EN-DC, UE (user equipment), mobility was also considered one of the AI / ML use cases and one of the scenarios for training / inferencing the model is that both functions are located in a RAN node. Subsequently, in release 18, the new work item "Artificial Intelligence (AI) / Machine Learning (ML) for NG-RAN (Artificial Intelligence (AI) / Machine Learning (ML) for NG-RAN" was initiated to specify improvements in data collection and signaling support in existing NG-RAN interfaces and architecture. In the above active standardization works, the model identification (e.g., a model ID) for supporting a RAN-based AI / ML model is considered very essential for both the network and the UE to perform any desired model functions (e.g., training, inference, selection, switching, updating, monitoring the model, etc.). Model ID information may be signaled to couple both network and UE side models for different lifecycle management (LCM) functions.US2020374711A1 describes a representation of a local model of the first data endpoint of the radio access network with multiple common models for endpoints of the radio access network, selecting one of the multiple common models for the first data endpoint, and transmitting the selected common model to the first data endpoint, any other data endpoint, or any other external system utilizing the selected common model.US2021097428A1 describes a method for adapting a deep learning model to a local environment by collecting training data and training a general deep learning model, whereby the deep learning model is individualized using transfer learning based on labels specific to one of a plurality of local devices.WO2022003949A1 describes machine learning models trained using training data with a classifier configured to classify input data and output data.WO2022073167A1 describes a method for receiving, by means of wireless communications via radio resource control (RRC) signaling, configuration messages, wherein the configuration messages contain parameters of an artificial neural network and the configuration messages also contain a common configuration for base set signaling.However, if there are multiple models in parallel to be activated at one UE, the multi-model signaling effort may be very high, especially if multiple UEs have their own multiple intra-device models for activation between one base station (BS / gNB) and multiple UEs. Currently, no specification is known for signaling methods and network UE behaviors to manage model transmission based on multiple models and / or other related model signaling effort for LCM functions.According to a first aspect, the present disclosure relates to a method for multi-model signaling for a RAN, comprising configuring a model block structure having a cross-model block and a native block for an ML model in a wireless communication system, comprising configuring the ML model to split it into two blocks, a cross-model block and a native block; determining a cross-model block for a plurality of models to be shared and / or coupled across different models; determining a native block to be configured uniquely for the native ML model; establishing the pre-configured criteria wherein the specific threshold may be used to estimate applicability of the cross-model block across multiple ML models.In some embodiments, the method according to the first aspect may further comprise one or more of the following optional features, which are considered either alone or in any technically possible combination.In some embodiments of the method according to the first aspect, the method is characterized in that the model block is further segmented into smaller blocks, such as sub-blocks among the inter-model and the in-model blocks.In some embodiments of the method according to the first aspect, the method is characterized in that a finite number of sub-blocks included in the cross-model / in-model block is configured depending on varying properties used as criteria for a block-based division of the ML model.In some embodiments of the method according to the first aspect, the method is characterized in that the indication information about the cross-model block is transmitted with the associated model IDs, so that the cross-model block is applied to the indicated plurality of models when the ML configuration is transmitted to the UEs.In some embodiments of the method according to the first aspect, the method is characterized in that index-based indication information may be sent to apply a cross-model block to multiple ML models.In some embodiments of the method according to the first aspect, the method is characterized in that the cross-model and the intrinsic-model block may form part of the structure, the architecture, the layout, the functionality, etc. of the model depending on varying implementation application cases. For example, the cross-model block may be predefined to indicate a specific portion of specified information about the structure / parameters of the model that are commonly applicable across models of different UEs, while the in-model block may be predefined to indicate a specific portion of specified information about the structure / parameters of the model that are applicable only to the associated model of a UE. In other words, information included in the cross-model and the in-model blocks may be predefined for deployment at the network side based on various combinations of model-related configuration information in connection with applications of the model.In some embodiments of the method according to the first aspect, the method is characterized in that the cross-model block may take many different forms of applicable models depending on the applications, such as a group of functionalities, a group of sub-models or a group of parameters / levels, etc., based on structure / characteristics of the model.In some embodiments of the method according to the first aspect, the method is characterized in that a specific configuration of selecting properties of the cross-model block can be set in advance by the network side for different applications or usage scenarios of the model.In some embodiments of the method according to the first aspect, the method is characterized in that the cross-model block can be configured independently and applied to all applicable ML models, wherein the preconfigured cross-model block can be sent to UEs or known to UEs in advance.In some embodiments of the method according to the first aspect, the method is characterized in that a finite number of candidate cross-model blocks applicable to arbitrary ML models can be set with a pre-configuration and indication information about one or more specific candidate cross-model blocks is sent for use.In some embodiments of the method according to the first aspect, the method is characterized in that it can be determined that the specific model is the reference model having a cross-model block, and includes transmitting other multiple models together with the determined reference model; applying the indicated reference model to other multiple models transmitted for activation.According to a second aspect, the present disclosure relates to a wireless device comprising at least one memory and at least one processor configured to perform a method according to any of the embodiments of the first aspect.According to a third aspect, the present disclosure relates to a user equipment (UE) including a wireless device according to any one of the embodiments of the present disclosure.According to a fourth aspect, the present disclosure relates to a base station (BS) comprising at least one memory and at least one processor configured to perform a method according to any of the embodiments of the first aspect.According to a fifth aspect, the present disclosure relates to a wireless communication system including at least one base station according to any one of the embodiments of the present disclosure and at least one user equipment according to any one of the embodiments of the present disclosure.According to a sixth aspect, the present disclosure relates to a computer program product comprising instructions which, when executed by at least one processor, configure the at least one processor to perform a method according to the first aspect, wherein the at least one processor is intended to perform a method for exchanging data according to any of the embodiments of the present disclosure. The computer program product may use any programming language and may be in the form of source code, object code, or any form between source code and object code, such as in a partially compiled form, or in any other desired form.According to a sixth aspect, the present disclosure relates to a computer readable storage medium comprising instructions which, when executed by at least one processor, configure the at least one processor to perform a method according to any of the embodiments of the present disclosure.BRIEF DESCRIPTION OF THE DRAWINGSFIG. 1 is an exemplary block diagram of an ML model block structure. FIG. 2 is an exemplary block diagram of types of an ML model block. FIG. 3 is an example block diagram of multiple ML models having a model block structure. FIG. 4 is an example block diagram of assigning an inter-model block to multiple ML models. FIG. 5 is an example block diagram of transferring multiple ML models with a model block structure. FIG. 6 is an exemplary flowchart of network-side behavior for applying an ML model block. FIG. 7 is an exemplary flowchart of a UE-side behavior for applying an ML model block. FIG. 8 is an exemplary signaling flow of applying an ML model block with an activation decision on the UE side. FIG. 9 is an exemplary signaling flow of applying an ML model block with an activation decision on the network side.DETAILED DESCRIPTIONThe detailed description below with reference to the accompanying drawings is intended to be a description of various configurations and not to represent the only configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of comprehensively understanding the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. In particular, although terminology 3GPP 5G NR may be used in this disclosure to explain embodiments herein, this should not be considered limiting of the scope of the invention.Now, some of the embodiments contemplated herein will be described in more detail with reference to the accompanying drawings. Other embodiments are, however, included within the scope of the subject matter disclosed herein, which disclosed subject matter is not to be construed as limited only 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.In general, all terms used herein should be interpreted according to their usual meaning in the respective technical field, unless another meaning is clearly indicated and / or arises from the context in which it is used. All references to a / the / the / the element, device, component, means, step, etc. are to be construed open to reference at least one instance of the element, device, component, means, step, etc., unless expressly stated otherwise. The steps of any methods 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 wherein it is implicit that a step must follow or preceding another step. Any feature of any of the embodiments disclosed herein may be applied to any other embodiment whenever appropriate. Likewise, any advantage of any of the embodiments may be applicable to any other embodiments, and vice versa. Other objects, features, and advantages of the included embodiments will become apparent from the following description.In some embodiments, a more general term "network node" may be used and may correspond to any type of radio network node or 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 MCG or SCG, a base station (BS), a multi-standard radio (MSR) radio node such as MSR BS, eNodeB, gNodeB, a network controller, a radio network controller (RNC), a base station controller (BSC), a relay, a donor node control relay, a base transceiver station (BTS), an access point (AP), transmission points, transmission nodes, RRU, RRH, Nodes in a distributed antenna system (DAS), a core network node (e.g., a mobile switching center (MSC, mobility management entity (MME), etc.), operation and maintenance (O&M), an operation support system (OSS), a self-optimized network (SON), a positioning node (e.g., (about:) an enhanced mobile location centre (E-SMLC)) serving mobile location centre), Minimization of Quick Change Reception Quality (MDT) testing, test equipment (physical node or software), etc.In some embodiments, the non-limiting term user equipment (UE) or wireless device may be used for and 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 a UE are a target device, a device to device (D2D) UE, a machine UE, or a UE capable of machine to machine (M2M) communication, a PDA, a PAD, a tablet, mobile terminals, a smartphone, laptop embedded equipment (LEE), laptop mounted equipment (LME), USB dongles, a UE category M1, a UE category M2, a ProSe UE, a V2V UE, a V2X UE, etc.In addition, terminology such as base station / gNodeB and UE should be considered non-limiting and, in particular, do not imply a particular hierarchical relationship between the two; in general, "gNodeB" could be considered device 1 and "UE" could be considered device 2 and these two devices communicate with each other over any radio channel. And in the following, the transmitter or receiver could be either the gNodeB (gNB) or the UE.As will be appreciated by one of ordinary skill in the art, aspects of the embodiments may be embodied as a system, apparatus, method, or program product. Accordingly, embodiments may take the form of a fully hardware implemented embodiment, a fully software implemented embodiment (including firmware, resident software, microcode, etc.), or an embodiment that combines software and hardware aspects.For example, the disclosed embodiments may be implemented as a hardware circuit including very-large-scale integration ("VLSI") circuits or gate arrays, commercially available semiconductors such as logic chips, transistors, or other discrete components. The disclosed embodiments may also be implemented in programmable hardware devices such as field programmable gate arrays, programmable logic arrays, programmable logic devices, or the like. As another example, the disclosed embodiments may include one or more physical or logical blocks of executable code, which may be organized as, for example, an object, procedure, or function.Moreover, 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, referred to herein as code. The storage devices may be tangible, non-transitory, and / or non-transitory. The memory devices may not embody signals. In a particular embodiment, the memory devices only use signals to access code.Any combination of one or more computer readable media may be utilized. The computer readable medium may be a computer readable storage medium. The computer readable storage medium may be a storage device storing the code. The memory device may be, for example, but is not limited to, one / e electronic / s, magnetic / s, optical / s, electromagnetic / s, infrared, holographic / s, micromechanical / s, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing.More specific examples (a non-exhaustive list) of the memory 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), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or apparatus.Code for performing operations for embodiments may consist of any number of lines and may be written in any combination of one or more programming languages, including an object oriented programming language such as python, ruby, Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or the like, and / or machine languages, such as assembly languages. The code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN), a wireless LAN (WLAN), or a wide area network (WAN), or the connection may be made to an external computer (for example, via the Internet using an Internet Service Provider (ISP).Moreover, the described features, structures, or characteristics of the embodiments may be combined in any suitable manner. In the following description, numerous specific details are provided, such as examples of programming, software modules, user selections, network transmissions, database requests, database structures, hardware modules, hardware circuits, hardware chips, etc., to provide a thorough understanding of the embodiments. However, those skilled in the art will appreciate that embodiments may be practiced without one or more of the specific details, or with other methods, components, materials, and so forth. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of an embodiment. References throughout this specification to "one (single) embodiment," "one embodiment," or similar language means that a specific feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of the phrases "in a (single) embodiment," "in an embodiment," and similar language throughout this specification may, but are not necessarily, all referring to the same embodiment, but are intended to mean "one or more, but not all embodiments," unless expressly stated otherwise. The terms "including", "comprising", "having" and variants thereof mean "including, but not limited to" unless expressly stated otherwise. A enumerated listing of items does not imply that any or all of the items are mutually exclusive unless expressly stated otherwise. The terms "a", "an" and "the / s" also refer to "one or more" unless expressly stated otherwise.Aspects of the embodiments are described below with reference to schematic flowcharts and / or schematic block diagrams of methods, apparatus, systems, and program products according to embodiments. It is understood that each block of the schematic flowcharts and / or schematic block diagrams, as well as combinations of blocks in the schematic flowcharts and / or schematic block diagrams, may be implemented by code. This code may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, produce means for implementing the functions / acts specified in the flowcharts and / or block diagrams.The code may also be stored in a storage device that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the storage device produce an article of manufacture including instructions that implement the function / act indicated in the flowcharts and / or block diagrams.The code may be loaded into a computer, other programmable data processing apparatus, or other devices to cause a series of operations to be performed on the computer, other programmable apparatus, or other devices to produce a computer-implemented process such that the code executed on the computer or other programmable apparatus provides processes for implementing the functions / operations indicated in the flowcharts and / or block diagrams.The flowcharts and / or block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of the 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 that includes one or more executable instructions of the code for implementing the one or more logical functions.It is also noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may in fact be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending on the functionality involved. Other steps and methods are conceivable that are equivalent in function, logic, or effect to one or more blocks, or portions thereof, of the illustrated figures.Although various types of arrows as well as types of lines may be employed in the flowchart and / or block diagrams, it should be understood that they do not limit the scope of the respective embodiments. Indeed, some arrows or other connectors may be used to indicate only the logical flow of the depicted embodiment. For example, an arrow may indicate a wait or monitoring period of unspecified duration between enumerated steps of the depicted embodiment. It is also noted that each block of the block diagrams and / or flowcharts, as well as combinations of blocks in the block diagrams and / or flowcharts, may be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and code.The description of elements in each figure may refer to elements from previous figures. Like numerals refer to like elements throughout the figures, including alternative embodiments of like elements.The detailed description below with reference to the figures is intended to be a description of various configurations and not to represent the only configurations in which the concepts described herein may be practiced. The detailed description contains specific details for the purpose of comprehensively understanding 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 terminology 3GPP, e.g., from 5G NR, may be used in this disclosure to explain embodiments herein, this should not be considered limiting of the scope of the present disclosure.In general, all terms used herein should be interpreted according to their usual meaning in the respective technical field, unless another meaning is clearly indicated and / or arises from the context in which it is used. All references to a / the / the / the element, device, component, means, step, etc. are to be construed open to reference at least one instance of the element, device, component, means, step, etc., unless expressly stated otherwise. Also, the order of steps of any methods disclosed herein, particularly in the figures, is provided for illustrative purposes only and does not mean to limit the present disclosure, which may be applied with the same steps performed in a different order and / or with all or a portion of the steps performed in parallel or together, unless a step is expressly described as following or preceding a step and / or where it is implicit that a step must follow or preceding another step. Steps represented in a figure, which are surrounded by a dashed line, are likewise to be considered as optional for the embodiment represented in this figure. Any feature of any of the embodiments disclosed herein may be applied to any other embodiment whenever appropriate. Likewise, any advantage of any of the embodiments may be applicable to any other embodiments, and vice versa. Other objects, features, and advantages of the included embodiments will become apparent from the following description.The disclosure relates to a wireless communication system, which may be, for example, a 5G NR wireless communication system. In particular, it represents a RAN of the wireless communication system used for exchanging data with UEs via radio signals. The RAN may, for example, send data to the UEs (DL), for example data received from a core network (CN). The RAN may also receive data from the UEs (UL) and this data may be forwarded to the CN.In the illustrated examples, the RAN includes a base station (BS). Of course, the RAN may include more than one BS to increase the coverage area of the wireless communication system. Each of these BSs may be referred to as NB, eNodeB (or eNB), gNodeB (or gNB in the case of a 5G NR wireless communication system), an access point, or the like, depending on the wireless communication standard or standards implemented.The UEs are located in a coverage area of the BS. The coverage area of the BS corresponds to, for example, the area in which UEs can decode a PDCCH transmitted by the BS.An example of a wireless device suitable for implementing any method discussed in the present disclosure performed at a UE corresponds to a means that provides wireless connectivity to the RAN of the wireless communication system and that can be used to exchange data with the RAN. Such a wireless device may be included in a UE. The UE may be, for example, a mobile phone, a wireless modem, a wireless communication device, a handheld device, a laptop computer, or the like. The UE may also be an Internet of Things (IoT) device, such as a wireless camera, smart sensor, smart glasses, a vehicle (manned or unmanned), a global positioning system device, etc., or any other device that can execute applications where data needs to be exchanged with remote receivers via the wireless device.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 nonvolatile memory (magnetic hard disk, solid state memory, optical disk, electronic memory, etc.). The one or more memories may store a computer program product in the form of a set of program-encoded instructions to be executed by the one or more processors to implement all or part of the steps of a method for exchanging data performed on the side of a UE according to any of the embodiments disclosed herein.The wireless device may also include a main radio (MR) unit. The MR unit corresponds to a wireless communication main unit of the wireless device used for exchanging data with BSs 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, etc. transceiver or the like. In preferred embodiments, the MR unit corresponds to a 5G NR unit for wireless communication.The following discussion provides the detailed description of the mechanism for pre-configuring an AI / ML-based model prior to the occurrence of handover of the wireless mobile communication system that includes a base station (e.g., a gNB) and a mobile station (e.g., a UE).The AI / ML life cycle may be divided into various stages, such as data collection / preprocessing, model training, model testing / validation, model deployment / update, model monitoring, model switching / selection, etc., each stage being equally important to achieve target performance in one or more specific models. When applying an AI / ML model to any application or application, one of the challenges is to manage the life cycle of the AI / ML model. This is mainly because the data / model deviations occur during deployment / inference of the model and result in performance degradation of the AI / ML model. Basically, the statistical changes to the data set take place after the model is deployed, and the model inference capability is also affected by unscreened data as input. In a similar aspect, the statistical characteristics of the dataset and the ratio between input and output for the trained model may change as deviations occur. Then, model adaptation is required to support operations such as model changeover, retraining, alternate solutions, etc. When employing an AI / ML model-capable wireless communication network, it is then important to consider how to accomplish an adjustment of the AI / ML model during operation, such as training, inference, monitoring, updating, etc. of the model. Based on a specific network-UE-ML interaction in deployment scenarios (e.g., in a UE mobility case), conditions applicable to ML for lifecycle management (LCM) functions may be substantially changed in various use cases and environmental characteristics. Currently, KI / ML-based techniques are applied to many different applications and 3GPP has also begun to work on its engineering study to apply it to multiple use cases based on the observed potential advantages.In a similar aspect, the statistical characteristics of the dataset and the ratio between input and output for the trained model may change as deviations occur. In this context, the performance of the model, such as inferencing and / or training, depends on different model execution environments with varying configuration parameters.To overcome this problem, interaction between UE and gNB is extremely important to track the performance of the model and reconfigure the model according to different environments between UE and different gNBs. When employing an AI / ML model-capable wireless communication network, it is then important to consider how to manipulate the AI / ML model in activation with a reconfiguration for wireless devices in functions such as training, inference, update, etc. of the model. In other words, by previously reconfiguring any model that operates with various gNBs, such as a primary node (MN) and a secondary node (SN), performance degradations can be minimized compared to when using each individual connection link for the operation of the model.In this method, UE connections to multiple gNBs, such as a MN and a SN, based on a master cell group (MCG) and a secondary cell group (SCG), respectively, are enabled to support decentralized multi-model operation that is activated via a MN and / or one or more SNs, so that an ML model signaling exchange may be distributed to one or more SNs due to an LCM-based function of the model. First, one or more candidate SNs are informed by the MN of information about an ML configuration as well as an LCM-based model. To trigger decentralized multi-model operation, there may be multiple conditions (e.g., preconfigured metrics or thresholds), such as the condition when the current traffic load in the current cell becomes high and / or additional model operation through separate connections (e.g., SN UE) parallel to the activated connection (e.g., MN UE) becomes necessary. The configured model requested by the MN (e.g., an LCM function in the MN UE connection) may be duplicated in an SN UE connection after the establishment of the SN connection for activation. Or, separately configured model operation may also be enabled in an SN UE connection as requested by the MN.The following discussion provides the detailed description of the mechanism of pre-configuring and signaling the specific information about model selection using a linkage of models and index values.Currently, KI / ML-based techniques are applied to many different applications and 3GPP has also begun to work on its engineering study to apply it to multiple use cases based on the observed potential advantages. The AI / ML life cycle may be divided into various stages, such as data collection / preprocessing, model training, model testing / validation, model deployment / update, model monitoring, etc., each stage being equally important to achieve target performance at one or more specific models.When applying an AI / ML model to any application or application, one of the challenges is to manage the life cycle of the AI / ML model. This is mainly because the data / model deviations occur during deployment / inference of the model and result in performance degradation of the AI / ML model. Basically, the statistical changes to the data set take place after the model is deployed, and the model inference capability is also affected by unscreened data as input. In a similar aspect, the statistical characteristics of the dataset and the ratio between input and output for the trained model may change as deviations occur. In this context, the model selection is one of the core topics for maintaining the performance of the model, such as inferencing and / or training depending on different model execution environments with varying configuration parameters. To overcome this problem, interaction between UE and gNB is extremely important to track the performance of the model and reconfigure the model according to different environments. An AI / ML model requires model monitoring after deployment because performance of the model cannot be continuously maintained due to fluctuations, and then update feedback is provided to retraine / update the model or select an alternative model. When employing an AI / ML model-capable wireless communication network, it is then important to consider how to handle the AI / ML model in activation in reconfiguration for wireless devices in functions such as training, inference, update, etc. of the model. For multiple parallel KI / ML models for varying LCM functions on one UE device or across multiple UEs, both the signaling effort and the increase in computing power for model transmission / delivery and / or retraining / updating a model may be relatively significant. In this method, it is determined that each ML model is divided into two blocks, such as an inter-model block and a native block, wherein the inter-model block may be shared and / or coupled with one or more other ML models, and the native block is configured uniquely for the associated ML model. The cross-model block, it is part of an ML model that can be used for other ML models based on the pre-configured criteria, wherein the specific threshold value can be used to assess the applicability of the cross-model block over several ML models, since the threshold value is configured from a similarity measure, for example. Or, the cross-model block may be configured independently and applied to all applicable ML models, wherein the preconfigured cross-model block may be sent to UEs or known to the UEs in advance. In this case, there may be a finite number of candidate cross-model blocks applicable to any ML models to be set with a pre-configuration if necessary when sending indication information about one or more specific candidate cross-model blocks for use. In the implementation aspect, inter-model and native blocks may form part of the model's structure, architecture, layout, functionality, etc., depending on varying implementation application cases. When an inter-model block is used across different ML models for multiple UEs and / or multiple applications on one device, index-based indication information may be sent to apply the inter-model block to multiple ML models, where the index may be configured in advance by the network side. For example, if an inter-model block is valid for an application to any specific ML model, the index value for the indicated inter-model block of the ML model is configured such that other ML models reference the configured index indicating the specific inter-model block and the indexed inter-model block is then applied to multiple ML models. The main benefit of this is that the cross-model block is sent once and applied to multiple ML models across multiple applications and / or multiple UEs. Otherwise, each ML model including the cross-model block that can be used for multiple models would have to be sent individually. In another aspect, the common part of the ML model for multiple models may be saved in signaling overhead if it may be preconfigured with an index-based indication for UEs. When sending out cross-model block indication information of an ML model, an ID list of all associated models must be sent together with it so that the identified models can be indicated to apply the indexed cross-model block across different models, if applicable. Alternatively, the network side performs model transfer of a plurality of models such as three models M 1, M 2, and M 3. When these models share a common part of each model (e.g., a submodel), the UE applies the indexed submodel (configured by the network side) to these three models when an indication of index information is received.FIG. 1 shows an example block diagram of an ML model block structure. In this example, the ML model may be structured to be divided into a cross-model block and a native block, where the block-by-block segmentation may be based on varying characteristics such as group of layers, group of sub-models, group of model functionalities / parameters, etc. Depending on the ML model characteristics, either the cross-model block or the native block may not be available. For example, a model may have only one model's own block.FIG. 2 shows an exemplary block diagram of types of an ML model block. In this example, the model block may be further segmented into smaller blocks, such as sub-blocks among the cross-model block and the native model block. Depending on varying properties used as criteria for block-based partitioning of the ML model, a finite number of sub-blocks included in the cross-model / in-model block may be configured.FIG. 3 shows an example block diagram of multiple ML models with a model block structure. In this example, for a finite number of ML models, an intermodel block is configured for all models. Different models may have different model sizes and / or redundancies, while a cross-model block is collectively included across K different models. Therefore, for model transmission / delivery of these K different models, the cross-model block need not be sent separately for each model because it can be reused for all models by sharing. As an example, a specific model may be determined to be a reference model that includes a cross-model block so that it is sent along with other multi-model transmission information and the specified reference model is applied to other multiple models transmitted for activation.FIG. 4 shows an example block diagram of assigning an inter-model block to multiple ML models. In this example, the cross-model block may take many different forms of applicable models depending on applications such as a set of functionalities, a set of sub-models, a set of parameters / levels, etc. based on the model structure / characteristics. A specific configuration of selecting properties of the cross-model block may be set by the network side in advance for various applications or deployment scenarios of the model.FIG. 5 shows an example block diagram of transferring multiple ML models with a model block structure. In this example, there are two independent ML models and the cross-model block has been identified as applicable to both models. When both models are transmitted over the air, the particular cross-model block is transmitted along with the on-model block of each model, and the cross-model block does not need to be transmitted separately at each model transmission because it can be duplicated after reception. As the number of ML models having an inter-model block increases, the advantage in reducing signaling overhead may be relatively high.FIG. 6 shows an example flowchart of network side behavior for applying an ML model block. On the network side, it is determined that the ML configuration having a model block structure determines, for all applicable models, an inter-model and a native-model block of each model. The specific threshold may be used to identify a cross-model block across different models based on the preconfigured criteria, such as similarity measure. When the ML configuration is transmitted to the UEs, the inter-model block indication information is transmitted with the associated model IDs, so that the inter-model block can be applied to the indicated plurality of models.FIG. 7 shows an example flowchart of a UE-side behavior for applying an ML model block. On the UE side, there may be multiple models for activation in parallel, and the identified cross-model block indicated by the network side may be used to apply to models for activation, if valid.FIG. 8 shows an example signaling flow of applying an ML model block with an activation decision on the UE side. In this example, it is determined that the cross-model block is to be applied to multiple models based on the ML configuration supporting information received from the network side. The criteria for determining the cross-model block, such as a threshold, may be provided in advance by the network side. After applying the cross-model block, the indication message about the activation of applying the cross-model block is sent to the network side so that both sides can be aligned for ML reconfiguration. General ML configuration information may be sent via system information or RRC signaling. Information related to a model block may be sent via L1 / L2 or L3 signaling.FIG. 9 shows an example signaling flow of applying an ML model block with an activation decision on the network side. In this example, it is determined on the network side that the cross-model block is to be applied to a plurality of models based on the ML configuration supporting information received from the UE side. The indication information about the cross-model block is transmitted with the associated model IDs, so that the cross-model block can be applied to the indicated plurality of models when an ML configuration is transmitted to UEs.References included in the specificationThis list of documents cited by the applicant has been produced in an automated manner and is only included for the better information of the reader. The list is not part of the German patent application or utility model application. The DPMA does not take any adhesion for any faults or omissions.Patent Literature citedUS 2020374711A1

[0003] US 2021097428A1

[0004] WO 2022003949A1

[0005] WO 2022073167A1

[0006]

Claims

A method for multi-model signaling for a RAN, comprising configuring a model block structure having a cross-model block and a native block for an ML model in a wireless communication system, comprising: • configuring the ML model to be divided into two blocks, a cross-model block and a native block; • determining that the cross-model block is shared and / or coupled across different models for multiple models; • determining that the native block is configured uniquely for the associated ML model; • setting the criteria that have been pre-configured, wherein the specific threshold may be used to assess applicability of the cross-model block across multiple ML models.The method of claim 1, wherein the model block is further segmented into smaller blocks, such as sub-blocks among the cross-model and the in-model blocks.The method according to any of the preceding claims, wherein a finite number of sub-blocks comprised in a cross-model and / or in-model block is configured depending on varying properties used for a block-based division of an ML model as criteria.The method of any preceding claim, wherein the indication information about the cross-model block is transmitted with the associated model IDs, such that the cross-model block is applied to the indicated plurality of models when the ML configuration is transmitted to the UEs.The method of the preceding claim 4, wherein index-based indication information can be sent to apply a cross-model block to multiple ML models.The method according to any of the preceding claims, wherein the cross-model block and the native-model block may form part of the structure and / or the architecture and / or the layout, the functionality of the model depending on varying implementation application cases.Method according to any of the preceding claims, wherein the cross-model block may take different forms of applicable models depending on applications, such as a group of functionalities and / or a group of sub-models and / or a group of parameters / levels based on the model structure and / or the model properties.The method according to any of the preceding claims, wherein a specific configuration of the selection of properties of the cross-model block is pre-determined by the network side for different applications or deployment scenarios of the model.The method of any preceding claim, wherein the cross-model block can be independently configured and applied to all applicable ML models, wherein the preconfigured cross-model block can be sent to UEs or known in advance to UEs.The method of any preceding claim, wherein a finite number of candidate cross-model blocks applicable to any ML models can be specified with a pre-configuration and indication information about one or more specific candidate cross-model blocks is sent for use.The method according to any of the preceding claims, wherein a specific model can be determined to be a reference model having an inter-model block comprising: • Other multiple models are transmitted together with the determined reference model; • The specified reference model is applied to other multiple models transmitted for activation.A wireless device comprising at least one memory and at least one processor configured to perform a method according to any preceding claim.A user equipment (UE) comprising a wireless device according to claim 12.A base station (BS) comprising at least one memory and at least one processor configured to perform a method according to any one of claims 1 to 11.A wireless communication system comprising at least one base station according to claim 14 and at least one user equipment according to claim 12.A computer program product comprising instructions which, when executed by at least one processor, configure the at least one processor to perform a method according to any one of claims 1 to 11.A computer readable storage medium comprising instructions which, when executed by at least one processor, configure the at least one processor to perform a method (40) according to any one of claims 1 to 11.

Images