WO2024254864A1 - Dispositifs et procédés de communication - Google Patents

Dispositifs et procédés de communication Download PDF

Info

Publication number
WO2024254864A1
WO2024254864A1 PCT/CN2023/100799 CN2023100799W WO2024254864A1 WO 2024254864 A1 WO2024254864 A1 WO 2024254864A1 CN 2023100799 W CN2023100799 W CN 2023100799W WO 2024254864 A1 WO2024254864 A1 WO 2024254864A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
identity
procedure
processor
models
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2023/100799
Other languages
English (en)
Inventor
Peng Guan
Zhen He
Gang Wang
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
NEC Corp
Original Assignee
NEC Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by NEC Corp filed Critical NEC Corp
Priority to PCT/CN2023/100799 priority Critical patent/WO2024254864A1/fr
Publication of WO2024254864A1 publication Critical patent/WO2024254864A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/02—Arrangements for optimising operational condition

Definitions

  • Example embodiments of the present disclosure generally relate to the field of communication techniques and in particular, to devices and methods for timer-based life cycle management (LCM) of machine learning (ML) model.
  • LCD timer-based life cycle management
  • ML machine learning
  • LCM is one of the core parts for AI/ML related specification. LCM may include the following stages/phases/procedure: data collection, model training, model registration, model deployment, model configuration, model inference, model selection, model activation, model deactivation, model switching, and fallback operation, model monitoring, model update, model transfer, UE capability reporting and so on. If the model cannot be identified and released properly, the resource utilization may be decreased accordingly.
  • embodiments of the present disclosure provide a solution for timer-based LCM of ML model.
  • a first device comprising: a processor configured to cause the first device to: start an availability timer for a machine learning (ML) model known to the first device; and upon an expiry of the availability timer, transmit, to a second device, a request indicating the second device to release a model identity configured for the ML model associated with the availability timer.
  • ML machine learning
  • a second device comprising: a processor configured to cause the second device to: receive, from a first device, a request indicating the second device to release a model identity configured for a ML model known to the first device; and release the model identity based on the request.
  • a communication method performed by a first device.
  • the method comprises: starting an availability timer for a machine learning (ML) model known to the first device; and upon an expiry of the availability timer, transmitting, to a second device, a request indicating the second device to release a model identity configured for the ML model associated with the availability timer.
  • ML machine learning
  • a communication method performed by a second device. The method comprises: receiving, from a first device, a request indicating the second device to release a model identity configured for a ML model known to the first device; and releasing the model identity based on the request.
  • a computer readable medium having instructions stored thereon, the instructions, when executed on at least one processor, causing the at least one processor to carry out the method according to the third, or fourth aspect.
  • FIG. 1A illustrates an example communication environment in which example embodiments of the present disclosure can be implemented
  • FIG. 1B illustrates an example framework 100B of a ML model
  • FIG. 2A illustrates a signaling flow for communication in accordance with some embodiments of the present disclosure
  • FIG. 2B illustrates a hierarchical structure of global model identity
  • FIG. 2C illustrates a concatenation structure of global model ID
  • FIG. 3A and 3C illustrate signaling flows for identification procedures in accordance with some embodiments of the present disclosure
  • FIG. 4 illustrates a flowchart of a method implemented at a first device according to some example embodiments of the present disclosure
  • FIG. 5 illustrates a flowchart of a method implemented at a second device according to some example embodiments of the present disclosure
  • FIG. 6 illustrates a simplified block diagram of an apparatus that is suitable for implementing example embodiments of the present disclosure.
  • terminal device refers to any device having wireless or wired communication capabilities.
  • the terminal device include, but not limited to, user equipment (UE) , personal computers, desktops, mobile phones, cellular phones, smart phones, personal digital assistants (PDAs) , portable computers, tablets, wearable devices, internet of things (IoT) devices, Ultra-reliable and Low Latency Communications (URLLC) devices, Internet of Everything (IoE) devices, machine type communication (MTC) devices, devices on vehicle for V2X communication where X means pedestrian, vehicle, or infrastructure/network, devices for Integrated Access and Backhaul (IAB) , Space borne vehicles or Air borne vehicles in Non-terrestrial networks (NTN) including Satellites and High Altitude Platforms (HAPs) encompassing Unmanned Aircraft Systems (UAS) , eXtended Reality (XR) devices including different types of realities such as Augmented Reality (AR) , Mixed Reality (MR) and Virtual Reality (VR) , the unmanned aerial vehicle (UAV)
  • UE user equipment
  • the ‘terminal device’ can further has ‘multicast/broadcast’ feature, to support public safety and mission critical, V2X applications, transparent IPv4/IPv6 multicast delivery, IPTV, smart TV, radio services, software delivery over wireless, group communications and IoT applications. It may also incorporate one or multiple Subscriber Identity Module (SIM) as known as Multi-SIM.
  • SIM Subscriber Identity Module
  • the term “terminal device” can be used interchangeably with a UE, a mobile station, a subscriber station, a mobile terminal, a user terminal or a wireless device.
  • the term “network device” refers to a device which is capable of providing or hosting a cell or coverage where terminal devices can communicate.
  • a network device include, but not limited to, a Node B (NodeB or NB) , an evolved NodeB (eNodeB or eNB) , a next generation NodeB (gNB) , a transmission reception point (TRP) , a remote radio unit (RRU) , a radio head (RH) , a remote radio head (RRH) , an IAB node, a low power node such as a femto node, a pico node, a reconfigurable intelligent surface (RIS) , and the like.
  • the network device also may be an operation administration and maintenance (OAM) , a server, location management function (LMF) , access and mobility management function (AMF) , and other core network node.
  • OAM operation administration and maintenance
  • server location management function
  • AMF access and mobility management function
  • the terminal device or the network device may have Artificial intelligence (AI) or Machine learning capability. It generally includes a model which has been trained from numerous collected data for a specific function, and can be used to predict some information.
  • AI Artificial intelligence
  • Machine learning capability it generally includes a model which has been trained from numerous collected data for a specific function, and can be used to predict some information.
  • the terminal device or the network device may work on several frequency ranges, e.g., FR1 (e.g., 450 MHz to 6000 MHz) , FR2 (e.g., 24.25GHz to 52.6GHz) , frequency band larger than 100A GHz as well as Tera Hertz (THz) . It can further work on licensed/unlicensed/shared spectrum.
  • FR1 e.g., 450 MHz to 6000 MHz
  • FR2 e.g., 24.25GHz to 52.6GHz
  • THz Tera Hertz
  • the terminal device may have more than one connection with the network devices under Multi-Radio Dual Connectivity (MR-DC) application scenario.
  • MR-DC Multi-Radio Dual Connectivity
  • the terminal device or the network device can work on full duplex, flexible duplex and cross division duplex modes.
  • the embodiments of the present disclosure may be performed in test equipment, e.g., signal generator, signal analyzer, spectrum analyzer, network analyzer, test terminal device, test network device, channel emulator.
  • the terminal device may be connected with a first network device and a second network device.
  • One of the first network device and the second network device may be a master node and the other one may be a secondary node.
  • the first network device and the second network device may use different radio access technologies (RATs) .
  • the first network device may be a first RAT device and the second network device may be a second RAT device.
  • the first RAT device is eNB and the second RAT device is gNB.
  • Information related with different RATs may be transmitted to the terminal device from at least one of the first network device or the second network device.
  • first information may be transmitted to the terminal device from the first network device and second information may be transmitted to the terminal device from the second network device directly or via the first network device.
  • information related with configuration for the terminal device configured by the second network device may be transmitted from the second network device via the first network device.
  • Information related with reconfiguration for the terminal device configured by the second network device may be transmitted to the terminal device from the second network device directly or via the first network device.
  • the singular forms ‘a’ , ‘an’ and ‘the’ are intended to include the plural forms as well, unless the context clearly indicates otherwise.
  • the term ‘includes’ and its variants are to be read as open terms that mean ‘includes, but is not limited to. ’
  • the term ‘based on’ is to be read as ‘at least in part based on. ’
  • the term ‘one embodiment’ and ‘an embodiment’ are to be read as ‘at least one embodiment. ’
  • the term ‘another embodiment’ is to be read as ‘at least one other embodiment. ’
  • the terms ‘first, ’ ‘second, ’ and the like may refer to different or same objects. Other definitions, explicit and implicit, may be included below.
  • values, procedures, or apparatus are referred to as ‘best, ’ ‘lowest, ’ ‘highest, ’ ‘minimum, ’ ‘maximum, ’ or the like. It will be appreciated that such descriptions are intended to indicate that a selection among many used functional alternatives can be made, and such selections need not be better, smaller, higher, or otherwise preferable to other selections.
  • the term “resource, ” “transmission resource, ” “uplink resource, ” or “downlink resource” may refer to any resource for performing a communication, such as a resource in time domain, a resource in frequency domain, a resource in space domain, a resource in code domain, or any other resource enabling a communication, and the like.
  • a resource in both frequency domain and time domain will be used as an example of a transmission resource for describing some example embodiments of the present disclosure. It is noted that example embodiments of the present disclosure are equally applicable to other resources in other domains.
  • an AI/ML model may be equivalent to at least one of the following: a model, an ML model, an AI model, a data-driven, a data processing model, an algorithm, a functionality, a procedure, a process, an entity, a function, a feature, a feature group, a model ID, a functionality ID, a configuration ID, a scenario ID, a site ID, or a dataset ID.
  • a model an ML model
  • an AI model a data-driven
  • a data processing model an algorithm, a functionality, a procedure, a process, an entity, a function, a feature, a feature group, a model ID, a functionality ID, a configuration ID, a scenario ID, a site ID, or a dataset ID.
  • the AI/ML model may be represented by or associated with a channel, a resource, a resource set, a RS resource, a RS resource set, a RS port, a set of RS ports, a RS port ID, or a set of RS port IDs.
  • the AI/ML model may comprise a set of weights values that may be learned during training, for example for a specific architecture or configuration, where a set of weights values may also be called a parameter set.
  • the AI/ML model may be used to predict a target cell, or measurements of a set of beams of a set of candidate cells in future based on at least historical measurements (e.g., L1-RSRP, L1-SINR) of a set of beams of a set of candidate cells.
  • at least historical measurements e.g., L1-RSRP, L1-SINR
  • an input of the AI/ML model may refer to the input of a model and indicate data inputted into the model, which may be equivalent to data, input data or input data item.
  • an output of AI/ML model may refers to the output of a model and indicate result (s) outputted by the model, which is equivalent to label, data, output data, or label data.
  • CR compression ratio
  • CR output dimension of encoder/input dimension of encoder.
  • the input dimension of encoder may refer to the number of ports ⁇ the number of sub-bands
  • the output dimension of encoder may refer to the number of compressed bits or the number of compressed bits /the number of quantization bits
  • CR 2 ⁇ output dimension of encoder /input dimension of encoder. 2 may stand for real and imaginary part of a complex value. 2 may stand for dual polarizations. 2 may also be replaced by other constants;
  • CR the number of bits required to report codebook-based CSI /the number of bits required to report compressed CSI.
  • the codebook-based CSI may refer to type I single panel codebook, type I multi-panel codebook, type II codebook, type II port selection codebook, enhanced type II codebook, enhanced type II port selection codebook, and further enhanced type II port selection codebook.
  • CR1 ⁇ CR2 means “CRI is a higher compression ratio than CR2” and “with CR1, less bits are required to compress the same amount of information” .
  • CR may be not a specific value of CR and may be an index value indicating CR.
  • CR index 1 may correspond to 1/2
  • CR index 2 may correspond to 1/4, etc.;
  • CR may comprise no compression.
  • CR index 0 may correspond to “no compression” .
  • a compression may be associated with at least one of the following parameters: a CR, the number of compressed bits (or the payload size of compressed CSI) , the number of quantization bits, output or input dimension of encoder, output or input dimension of decoder, the number of layers in an encoder or decoder, dimension or size of each layer, activation function, and so on.
  • Each of these parameters may be set per frequency unit, per time unit or per spatial unit.
  • a CR per frequency unit refers to the number of compressed bits per sub-band, per PRB, etc.
  • a CR per time unit refers to the number of compressed bits per slot, per frame, etc..
  • a CR per spatial unit refers to the number of compressed bits per port, per beam, etc..
  • quantization information may refer to at least one of the following:
  • the number of quantization bits such as 1-bit, 2-bit, 4-bit, etc.
  • the method of quantization such as uniform quantization, non-uniform quantization, vector quantization, scale quantization;
  • Aceil, floor or round operation applied in quantization function such as ceil/floor/round value of (x ⁇ 2 (B-1) ) / (2 (B-1) ) if B-bit uniform quantization is used;
  • the higher compression ratio means more bits for quantization
  • the quantization information may be not a specific value and may be an index value indicating the quantization information.
  • quantization index 0 may correspond to 1-bit quantization
  • quantization index 1 may correspond to 2-bit quantization, etc..
  • LCM is one of the core parts for AI/ML related specification.
  • two types of LCM are proposed to be supported, i.e., functionality based LCM and model-identity (ID) based LCM.
  • ID model-identity
  • functionality/model identification may be needed accordingly.
  • both the network device/terminal device may initiate the identification procedure.
  • an AI/ML model may have a model ID with associated information and/or model functionality at least for some AI/ML operations.
  • a model may be identified by a model ID.
  • a model ID may be used to identify which AI/ML model is being used in LCM including model delivery, and the model ID may be used to identify a model (or models) during model selection/activation/deactivation/switching.
  • indication of activation/deactivation/switching/fallback may be implemented based on individual AI/ML functionality.
  • indication of model selection/activation/deactivation/switching/fallback may be implemented based on individual model IDs.
  • the UE may have one AI/ML model for the functionality, Alternatively, in some embodiments, the UE may have multiple AI/ML models for the functionality.
  • UE-side models and UE-part of two-sided models 1) for AI/ML functionality identification legacy 3GPP framework of features may be reused, UE may indicate the supported functionalities/functionality for a given sub-use-case, and the UE capability reporting may be taken as starting point; 2) for AI/ML model identification, models may be identified by model ID at the Network and UE may indicate the supported AI/ML models; 3) in functionality-based LCM, network may indicate activation/deactivation/fallback/switching of AI/ML functionality via a signaling (e.g., a radio resource control (RRC) signalling, downlink control information (DCI) ) ; models may not be identified at the Network, and UE may perform model-level LCM; 4) in model-ID-based LCM, models are identified at the network, and network/UE may activate/deactivate/select/switch individual AI/ML models via model ID.
  • RRC radio resource control
  • DCI downlink control information
  • functionality identification there may be either one or more than one functionality defined within an AI/ML-enabled feature.
  • the model-ID-based LCM may operate based on identified models, where a model may be associated with specific configurations/conditions associated with UE capability of an AI/ML-enabled feature/feature group and additional conditions (e.g., scenarios, sites, and datasets) as determined/identified between UE-side and network-side.
  • additional conditions e.g., scenarios, sites, and datasets
  • model identification types may be categorized as follows:
  • Model is identified to network (if applicable) and UE (if applicable) without over-the-air signaling, where the model may be assigned with a model ID during the model identification, which may be referred/used in over-the-air signaling after model identification;
  • Type B Model is identified via over-the-air signaling
  • Model identification initiated by the UE, and network assists the remaining steps (if any) of the model identification and the model may be assigned with a model ID during the model identification;
  • Type B2 model identification initiated by the network, and UE responds (if applicable) for the remaining steps (if any) of the model identification, and the model may be assigned with a model ID during the model identification.
  • UE may indicate the supported AI/ML model IDs for a given AI/ML-enabled feature/feature group in a UE capability report as starting point.
  • model ID should be unique “globally” , e.g., in order to manage test certification, each retrained version needs to be identified. However, it is still pending whether a global model ID is needed for model inference in model activation/deactivation/selection/switching and so on since for these purposes it may not be efficient to use global model ID. In this event, it is expected that a local (or temporary) model ID may be configured for the model, which may be used for model management after identification/registration.
  • the first device (such as, a terminal device) starts an availability timer for a ML model known to the first device. Then, upon an expiry of the availability timer, first device transmits, to a second device (such as, a network device/another terminal device) , a request indicating the second device to release a model identity configured for the ML model associated with the availability timer.
  • a second device such as, a network device/another terminal device
  • the LCM of a ML model known to the first device when the LCM of a ML model known to the first device is not active (or the ML model is not in use) , another device may release the model identity configured for the ML model in time. As a result, a recycling rate pf the available model identity is increased. Further, due to the timely release, the resources (such as, UE memory, or battery) used for maintaining models which are not in use may be saved accordingly.
  • “Conditions” referred to as configurations supported indicated via UE capability reporting (e.g., component field) , related to model training, model inference, performance monitoring, validation procedure, fallback, of an AI/ML model/functionality or a group of AI/ML models/functionalities;
  • UE capability reporting e.g., component field
  • Additional conditions including but not limited to, application conditions, scenarios, datasets, cell ID, timestamp and SNR, and so on;
  • UE internal conditions including but not limited to, memory, battery, computation resource, overheating and other hardware limitations;
  • AI/ML-enabled feature referred to as a feature where AI/ML may be used;
  • Physical AI/ML model (s) referred to as an actual implementation of such a model
  • a logical AI/ML model referred to as a model that is identified and assigned a model ID;
  • a local model index refers to a temporary model index for model control after Model identification
  • Local Model index may be allocated between network and UE and utilized for various LCM signalling purposes such as activation/deactivation/selection/switching;
  • Physical model ID used for identifying a real AI/ML model, a real implementation
  • Global model ID usually used in radio access network (RAN) 2 solutions
  • Logical model ID associated with one or a group of physical models for the same purpose;
  • Local model ID RAN1 solutions usually do not require global ID;
  • a group of models referred to as a group of physical/logical/global/local models
  • functionality-based LCM in case of AI/ML functionality identification and functionality-based LCM of UE-side models and/or UE-part of two-sided models, functionality is referred to an AI/ML-enabled Feature/feature group enabled by configuration (s) , where configuration (s) is (are) supported based on conditions indicated by UE capability.
  • functionality-based LCM may operate based on, at least, one configuration of AI/ML-enabled Feature/feature group , or specific configurations of an AI/ML-enabled Feature/Feature group.
  • ML model ML model
  • AI model ML function
  • AI function ML function
  • model “functionality” and “model/functionality” may be used interchangeably.
  • model and “model group” may be used interchangeably.
  • ID identifier
  • index identifier
  • identifier identifier
  • wording of “in response to/upon a model inference/model activation/switching-to/selection/model update/fine-tuning/model retraining” may refer to “in response to/upon a start, initiation or completion of a model inference/model activation/switching-to/selection/model update/fine-tuning/model retraining” ;
  • wording of “in response to/upon a positive/negative assessment result” may refer to “in response to/upon detecting/determining a positive/negative assessment result” ;
  • wording of “in response to/upon a message/indication” may refer to “in response to/upon receiving a message/indication” or “after a pre-configured period from receiving a message/indication” .
  • FIG. 1A illustrates a schematic diagram of an example communication environment 100A in which example embodiments of the present disclosure can be implemented.
  • a plurality of communication devices including a first device 110 and a second device 120, can communicate with each other.
  • both the first device 110 and the second device 120 may include a terminal device.
  • the first device 110 and the second device 120 may communicate with each other via sidelink communication.
  • the first device 110 may include a terminal device and the second device 120 may include a network device serving the terminal device.
  • a link from the first device 110 to the second device 120 is referred to as uplink, while a link from the second device 120 to the first device 110 is referred to as a downlink.
  • the second device 120 is a transmitting (TX) device (or a transmitter) and the first device 110 is a receiving (RX) device (or a receiver) , and the second device 120 may transmit downlink transmission to the first device 110.
  • the second device 120 is an RX device (or a receiver) and the first device 110 is a TX device (or a transmitter) , and the first device 110 may transmit uplink transmission to the second device 120.
  • FIG. 1A one or more ML models may be deployed at the first device 110.
  • FIG. 1B illustrates an example framework 100B of a ML model.
  • the example framework 100B consists of: (i) Data Collection, (ii) Model Training, (iii) Model Management, (iv) Model Inference, and (v) Model Storage.
  • the framework 100B may be suitable for both the model based and/or functionality based LCM.
  • Model Storage in the FIG. 1B is only intended as a reference point (if any) for protocol terminations etc for model transfer/delivery etc. It is not intended to limit where models are actually stored. It is to be understood that the example framework 100B should is given for illustrative purpose only, which should not be interpreted as any limitation to the present disclosure.
  • the communication environment 100A may include any suitable number of devices configured to implementing example embodiments of the present disclosure.
  • the first device 110 and the second device 120 may communicate with each other via a channel such as a wireless communication channel on an air interface (e.g., Uu interface) or a PC5 interface.
  • the wireless communication channel may comprise a sidelink, a physical uplink control channel (PUCCH), a physical uplink shared channel (PUSCH), a physical random-access channel (PRACH), a physical downlink control channel (PDCCH), a physical downlink shared channel (PDSCH) and a physical broadcast channel (PBCH) .
  • PUCCH physical uplink control channel
  • PUSCH physical uplink shared channel
  • PRACH physical random-access channel
  • PDCCH physical downlink control channel
  • PDSCH physical downlink shared channel
  • PBCH physical broadcast channel
  • any other suitable channels are also feasible.
  • the communications in the communication environment 100A may conform to any suitable standards including, but not limited to, Global System for Mobile Communications (GSM), Long Term Evolution (LTE), LTE-Evolution, LTE-Advanced (LTE-A), New Radio (NR), Wideband Code Division Multiple Access (WCDMA), Code Division Multiple Access (CDMA), GSM EDGE Radio Access Network (GERAN), Machine Type Communication (MTC) and the like.
  • GSM Global System for Mobile Communications
  • LTE Long Term Evolution
  • LTE-Evolution LTE-Advanced
  • LTE-A LTE-A
  • New Radio NR
  • WCDMA Wideband Code Division Multiple Access
  • CDMA Code Division Multiple Access
  • GERAN GSM EDGE Radio Access Network
  • MTC Machine Type Communication
  • Examples of the communication protocols include, but not limited to, the first generation (1G), the second generation (2G), 2.5G, 2.75G, the third generation (3G), the fourth generation (4G), 4.5G, the fifth generation (5G) communication protocols, 5.5G, 5G-Advanced networks, or the sixth generation (6G) networks.
  • FIG. 2A illustrates a signaling flow 200A for communication in accordance with some embodiments of the present disclosure.
  • the signaling flow 200 will be discussed with reference to FIG. 1A and FIG. 1B, for example, by using the first device 110 and the second device 120.
  • the operations at the first device 110 and the second device 120 should be coordinated.
  • the second device 120 and the first device 110 should have common understanding about configurations, parameters and so on. Such common understanding may be implemented by any suitable interactions between the second device 120 and the first device 110 or both the second device 120 and the first device 110 applying the same rule/policy.
  • the corresponding operations should be performed by the second device 120.
  • the corresponding operations should be performed by the first device 110.
  • some of the same or similar contents are omitted here.
  • some interactions are performed among the first device 110 and the second device 120 (such as, exchanging first and second information and so on) . It is to be understood that the interactions may be implemented either in one single signaling/message/configuration or multiple signaling/messages/configurations, including system information, a radio resource control (RRC) signalling, downlink control information (DCI) , uplink control information (UCI) , media access control (MAC) control element (CE) and so on.
  • RRC radio resource control
  • DCI downlink control information
  • UCI uplink control information
  • CE media access control element
  • the first device 110 may be operated as a terminal device and the second device 120 may be operated as a network device.
  • both the first and the second devices are terminal devices.
  • the first device 110 may determine a first identity (such as, a UE side maintained model ID if the first device 110 is a UE) , where the first identity may be one of the following: a global model identity or a physical model identity.
  • the second device 120 may configure a second identity for the ML model (such as, a network assigned model ID if the second device 120 is a network) , where the second identity may be one of the following: a local model identity or a logical model identity.
  • the second identity configured for the ML model also may be the ordinal number of the ML model in a configured ML model list.
  • the second identity may have a lower signalling overhead.
  • the first device 110 and the second device 120 may perform 220 an identification procedure.
  • this identification procedure a mapping between the first identity and the second identity may be established at the first device 110 and the second device 120.
  • one ML model may correspond to the first identity and the second identity.
  • FIGS. 3A to 3C illustrate signaling flows for identification procedures 300A, 300B and 300C in accordance with some embodiments of the present disclosure.
  • the first device 110 may transmit 310 first information to the second device 120, where the first information indicates a set of first identities of a first set of ML models comprising one or more ML models.
  • the first device 110 may receive 320 second information from the second device 120, where the second information indicates a set of second identities of a second set of ML models.
  • the second device 120 may only select those ML models suitable to configure the second identity.
  • the set of ML models with the first identity may be larger than the set of ML models with the second identity. That is, in some embodiments, the second set of ML models may be a subset of the first set of ML models.
  • the second device 120 is equipped with 64 Tx beams, the second device 120 may only select ML models trained for 64 Tx beams and configure to first device 110. The other models supporting 16 beams or 128 beams are not useful for this the second device 120.
  • the first device 110 and the second device 120 may obtain mappings between first identities and second identities.
  • the set of first identities may be ⁇ first identity #1, first identity #2, first identity #3, first identity #4 ⁇ corresponding to model #1, model #2, model #3 and model #4, respectively
  • the set of second identities may be ⁇ second identity #1, second identity #2 ⁇ corresponding to model #1, model #2, respectively.
  • the first device 110 and the second device 120 may determine a mapping of ⁇ first identity #1, second identity #1 ⁇ for model #1, and a mapping of ⁇ first identity #2, second identity #2 ⁇ for model #2.
  • the second identity may be determined at least partially based on the first information.
  • the value of the second identity may be a function of the value of the first identity.
  • the first information may contain a second identity (ies) suggested by the first device 110.
  • the first identity and the second identity may be used for different LCM procedure set.
  • the first identity may be used for identifying the ML model during at least of the following procedure: a model transfer procedure, a model delivery procedure, a model update procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the second identity may be used for identifying the ML model during at least of the following procedure: a model inference procedure, a model management procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the first identify may be used.
  • the second identity may be used for identifying the ML model to reduce the signalling overhead.
  • the first device 110 may receive, from the second device 120, a model-related configuration associated with a second identity of the set of second identities, where the model-related configuration is configured for at least one of the following: a model training procedure, a model inference procedure, a model monitoring procedure or a data collection and associated with at least one of the following: measurement, reporting and behaviors of the first device 110.
  • the second identity may be associated with configurations of measurement, report and UE behavior specified for model training, model inference, model monitoring, data collection and so on.
  • Table 2 illustrates an example to show the relationship between first and second identity, and the configurations.
  • Table 2 an example to show the relationship between first and second identities, and the configurations
  • Second identity 0 and First identity “X1-1, 1” are confiugued for the ML model 1, further, the Configuration 0 may be associated with the Second identity 0. Further, the Configuration 0 may be configured for at least one of the following: a model training procedure of the ML model 1, a model inference procedure of the ML model 1, a model monitoring procedure or a data collection of the ML model 1 and comprises at least one of: 0 th set of resources for measurement, report, and related UE behavior.
  • Second identity 1/2/3 and First identity “X1-1, 3” / “X1-1, 5” / “X1-1, 7” are confiugued for the ML model 3/5/7
  • the Configuration 1/2/3 may be associated with the Second identity 1/2/3.
  • the Configuration 1/2/3 may be configured for at least one of the following: a model training procedure of the ML model 3/5/7, a model inference procedure of the ML model 3/5/7, a model monitoring procedure or a data collection of the ML model 3/5/7 and comprises at least one of: 1 th /2 th 3 th set of resources for measurement, report, and related UE behavior.
  • the identification procedure also may be implemented in other manners.
  • the first device 110 may transmit 330 the first information to the second device 120, where the first information indicates a set of first identities of a first set of ML models comprising the ML model. Then the first device 110 may receive 340 third information from the second device 120, where the third information indicates a set of temporary second identities (which may be a network assigned model ID if the second device 120 is a network device) of a second set of ML models comprising the ML model, wherein the temporary second identity is configured by the second device 120.
  • the third information indicates a set of temporary second identities (which may be a network assigned model ID if the second device 120 is a network device) of a second set of ML models comprising the ML model, wherein the temporary second identity is configured by the second device 120.
  • the first device 120 needs to determine whether the temporary second identities are acceptable, e.g., the first device 110 needs to check whether the temporary second identity (ies) has been used. Based on the determination, the first device 110 may transmit the feedback information to the second device 120. In some embodiments, the first device 110 may transmit 350 confirmation information which is used for confirming at least one temporary second identity to the second device 120. As the result, the temporality of the temporary second identity may be removed, i.e., the temporary second identity will be used as the second identity.
  • the first device 110 may transmit 350 contention information to the second device 120, where the contention information indicates that at least temporary second identity is not applicable.
  • the second device 120 may optionally transmit 360 contention resolution information to the first device 110, where the contention resolution information indicates at least one newly-configured second identity corresponding to the unacceptable at least temporary second identity.
  • the first device 110 may transmit 350 contention information to the second device 120, where the contention information indicates at least one recommended second identity in case that at least one temporary second identity comprised is not applicable.
  • the second device 120 may transmit 360 contention resolution information to the first device 110, where the contention resolution information may indicate whether the at least one recommended second identity is confirmed by the second device 120.
  • the first device 110 and the second device 120 also may determine the mapping between the first identity and the second identity.
  • the ML model may be developed at the first device 110.
  • a third device (a third party, such as, either (or both) AI/ML server for the second device 120 or/and AI/ML server for the first device 110) may deliver an AI/ML model with a third identity to either (or both) of the second device 120 or/and the first device 110.
  • the second device 120 may receive 375 a third identity of the ML model configured by the third device, and the first device 110 also may receive 380 a third identity of the ML model configured by the third device.
  • the third identity of a ML model may include one or more of the following information: global model, local model (optional) , network ID, UE ID, configuration ID, dataset ID and so on.
  • the third identity may be represented in a hierarchical structure (as illustrated in FIG. 2B) or a concatenation structure (as illustrated in FIG. 2C) to identify a global model.
  • the second device 120 may develop three different versions for the ML model.
  • the first deice 110 also may develop three different versions for the ML model.
  • the mode 0, model 1 and model 2 at the first device 110 they only need to know the mode 0 at the second device 120 , and the other version of the ML model may be transparent for them. That is, the second device 120 may only notify the mode 0 at the second device to the first device 110. In this way, by using this hierarchical structure to identify the ML model, the version maintenance at the related device would be more flexible.
  • the identity consists of sequentially connected multiple parts, for example, a global model, a local model, a network (NW) ID, a UE ID, a configuration ID, a dataset ID and so on.
  • NW network
  • UE ID a configuration ID
  • dataset ID a dataset ID and so on.
  • the second device 120 and/or the first device 110 may develop different versions of AI/ML models based on the one reference AI/ML model with a global ID.
  • the second device 120/the first device 110 versions are transparent to the third party.
  • the second device 120/the first device 110 versions are not transparent to the third party.
  • Table 3 illustrates example identities.
  • example value of “aaaaaaa” / “aaaaffff” / “fffffff” may be an 8 hexadecimal numbers.
  • the first device 110 may transmit the first information comprising the third identity to the second device 120.
  • the ML model may be better identified, i.e., the third identity may be used to align the model deployed at second device 120, especially for a two sided model.
  • the first device 110 may receive second information (indicating a set of second identities of a second set of ML models) from the second device 120 as discussed with reference to FIG. 3A, or receive the third information (indicating a set of temporary second identities of a second set of ML models) from the second device 120 as discussed with reference to FIG. 3B 15.
  • second information indicating a set of second identities of a second set of ML models
  • third information indicating a set of temporary second identities of a second set of ML models
  • the first identity and the second identity are transmitted via an RRC signaling, MAC CE, UCI or a capacity report.
  • the second identity and the second identity may be transmitted via an RRC signaling, MAC CE or DCI.
  • mapping relationship among different types of AI/ML model identities may be established in model identification procedure.
  • the second identity may be used to identify an AI/ML model and in subsequent AI/ML operations between the first device 110 and the second device 120.
  • the first device 110 when transmitting the first identity via a capacity report, may transmit 210 capability-related information of the first device 110 to the second device 120 and via at least one feature or at least one feature group, as illustrated in FIG. 2A.
  • the capability-related information indicates a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device 110.
  • the capability-related information also may indicate model-related information, including but not limited to, the number of ML models supported for a specific feature or a feature group, or at least one model-related parameter related to a specific feature or a feature group.
  • a UE capability reporting for AI/ML-enabled feature/feature group is supported, i.e., a UE capability report may include the member of AI/ML models supported for a feature/feature group and/or details of AI/ML models supported for a feature/feature group.
  • the first device 110 may indicate supported AI/ML models to the second device 120, and the second device may then perform correct configuration based on the UE capability report.
  • Table 4 illustrates an example structure of the UE capability report.
  • Table 4 an example structure of the UE capability report
  • Feature Index (X1, X2, 7) may be replaced with any integer number.
  • Feature Index (X1, X2, ...) may be a type of model identity, functionality identity or a type of the first identity.
  • Field names may be other names according to the components of the feature/feature group.
  • the Candidate values may have a different range.
  • one feature may correspond to more than one Feature group.
  • each feature may be associated with a sub use case being one of the following: ML model-based channel state information (CSI) feedback, ML model-based time-domain CSI prediction, ML model-based spatial-domain beam prediction, ML model-based time-domain beam prediction, ML model-based direct positioning, or ML model-based assisted positioning.
  • the feature group may be one of the following: a CSI compression, a CSI prediction, a downlink transmitter beam prediction, or a beam pair prediction, and wherein the capability-related information indicates at least one of the following.
  • each feature group may correspond one component in which related parameter associated with this feature group may be configured.
  • the UE capability report may also include: configurations, conditions, additional conditions, UE internal conditions, and so on.
  • details of AI/ML models may be also reported via this feature/feature group report, in addition to the number.
  • a ML model identity or a group of identity may be included, and is associated with different inputs/outputs/other information.
  • Table 4 illustrates an example using feature/feature group to report AI/ML related UE capability for CSI compression.
  • Table 5 an example using feature/feature group to report ML related UE capability for CSI compression
  • the first device 110 may determine the ML model is known to the first device 110 if:
  • At least part of a model structure is known to the first device 110, e.g.,
  • the AI/ML model is known if the partial or full model structure is known and/or the partial or full of the model parameters are known;
  • the AI/ML model is known at the first device 110 if partial or full structure/parameters of UE part of the two sided model are known at the first device 110;
  • model parameters are known to the first device 110,
  • a time length timed from a time point of the last model inference procedure of the ML model is less than or equal to a threshold time, i.e., during a period from the last model inference using this AI/ML model;
  • a time length timed from the last model identification procedure of the ML model is less than or equal to a threshold time, i.e., During a period from the last model identification for this AI/ML model;
  • a time length timed from the last performance monitoring procedure of the ML model is less than or equal to a threshold time, i.e., during a period from the last performance monitoring of this AI/ML model;
  • model fine-tuning procedure or model validation procedure of the ML model is less than or equal to a threshold time, i.e., during a period from the last training/fine-tuning/validation completion of this AI/ML model;
  • a time length timed from the last model indication, model measurement, or model reporting related to the ML model is less than or equal to a threshold time, i.e., during a period from the last indication/measurement/reporting related to this AI/ML model; a time length timed from the last model activation, switching or selection of the ML model is less than or equal to a threshold time, i.e., during a period from the last model activation/switching-to/selection for this AI/ML model;
  • the positive performance monitoring results means that satisfy a certain condition, which may be one or many of the following
  • the ML model may be determined to be unknow to the first device 110. Further, during a period from the last model deactivation/switching-from for this ML model, the ML model may be determined to be unknow to the first device 110.
  • the first device 110 may perform at least one of the following:
  • requesting another ML model with the second device 120 e.g., requesting a new model, or request to switch to another AI/ML model;
  • the first device 110 may apply 230 one or more identified models. That is, the first device 110 may additionally perform a further selection, for example, based on current application condition, like Signal to Noise Ratio (SNR) , mobility, line of sight (LOS) /non line of sight (NLOS) and so on for example, based on current application condition, like SNR, mobility, LOS/NLOS and so on.
  • SNR Signal to Noise Ratio
  • LOS line of sight
  • NLOS non line of sight
  • the second device 120 also may transmit 240 a message used for configurating (such as, configuring the second identity and related resource as discussed above) or activating one or more ML models deployed at the first device 110. .
  • the second device 120 may switch/select the ML model.
  • one or more ML models may be running at the first device 110.
  • the first device 110 may assess 250 the one or more running ML models, such as, monitoring the model inference/model activation/switching-to/selection/model update/fine-tuning/model retraining, or assessing the performance and so on.
  • the assessment results may be used for maintaining an availability timer for a ML model.
  • the availability timer may be configured for a group of ML models including one or more ML model.
  • all the discussions made to the ML model should also may applicable for the group of ML model, such as, “a positive assessment result of the ML model” may be replaced by “a positive assessment result of at least one ML model in the group of ML model” , “a first indication indicating the first device 110 to continue applying the ML model” may be replaced by “a first indication indicating the first device 110 to continue applying at least one ML model in the group of ML model” , “a model inference/model activation/switching-to/selection/model update/fine-tuning/model retraining of the ML model” may be replaced by “a model inference/model activation/switching-to/selection/model update/fine-tuning/model retraining of at least one ML model in the group of
  • the first device 110 starts 260 an availability timer for a ML model known to the first device 110.
  • first device 110 transmits 282 a request to the second device 120, where the request indicates the second device 120 to release a model identity configured for the ML model associated with the availability timer.
  • the second device 120 may release the model identity configured for the ML model. In this way, a recycling rate pf the available model identity is increased.
  • a time length of the availability timer may be defined as a default value or determined by the first device 110 or the second device 120.
  • a same or different timer (s) may be associated with different ML models/model groups.
  • the first device 110 may start 260 the availability timer upon receiving a first message from the second device 120, where the first message is used for configuring or activating the ML model.
  • the first device 110 may start the availability timer. In this event, it implies that the started availability timer is only associated with configured/activated ML models.
  • the first device 110 may start 260 the availability timer upon initiating a model identification procedure for the ML model with the second device 120.
  • the first device 110 may restart 270 the availability timer in response to at least one of the following:
  • the length of the availability timer should be greater than the time configured for performance monitoring
  • receiving a first indication indicating the first device 110 to continue applying the ML model such as, receiving an indication from the second device 120/server to apply this AI/ML model for model inference: e.g., model activation/switching-to/selection,
  • the first device 110 may determine 280 that the availability timer expiries upon one of the following:
  • radio resource management e.g., handover, radio link failure and so on.
  • model-related configuration associated with the ML model, the model-related configuration associated with at least one of the following: measurement, reporting and behaviors of the first device 110,
  • the first device 110 and the second device 120 may release the model identity.
  • the first device 110 and the second device 120 may release the model identity by releasing a correspondence between a first identity of the ML model and a second identity of the ML model, wherein the second identity is configured by the second device 120 for the ML model and the first identity is determined by the first device 110.
  • the first device 110 and the second device 120 may release the model identity by remapping the second identity previously configured for the ML model to another ML model.
  • the first device 110 and the second device 120 may release the model identity by removing the model identity from a list of ML model identity .
  • the same second identity may correspond to different AI/ML models.
  • Table 6 illustrates an example to show that model 5 is released.
  • table 6 an example to show that model 5 is released
  • the first device 110/the second device 120 may inform the latest/updated model identity mapping (s) to the second device 120/the first device 110.
  • the second identity 2 is mapped to the model 5
  • the second identity 2 is mapped to the model 11. That is, the second identity 2 is re-assigned/mapped to a new ML model.
  • a new identification for the model 11 is initiated and the second device 120 re-configure the second identity 2 to the model 11.
  • FIG. 4 illustrates a flowchart of a communication method 400 implemented at a first device in accordance with some embodiments of the present disclosure. For the purpose of discussion, the method 400 will be described from the perspective of the first device 110 in FIG. 1A.
  • the first device 110 starts an availability timer for a machine learning (ML) model known to the first device.
  • ML machine learning
  • the first device 110 transmits to a second device, a request indicating the second device to release a model identity configured for the ML model associated with the availability timer.
  • the first device 110 may start the availability timer upon one of the following: receiving a first message from the second device, the first message used for configuring or activating the ML model, or initiating a model identification procedure for the ML model with the second device.
  • the first device 110 may restart the availability timer upon one of the following: a positive assessment result of the ML model, a first indication indicating the first device to continue applying the ML model, a model update procedure of the ML model, a model fine-tuning procedure of the ML model, or a model retraining procedure of the ML model.
  • the first device 110 may determine that the availability timer expiries upon one of the following: a negative assessment result of the ML model, a second indication indicating the first device not to continue applying the ML model, a change of configuration information configured for the ML model, a change of an application condition for the ML model at the first device, or a predefined event associated with radio resource management.
  • the first device 110 may fall back to a default ML model or an operation mode without ML model, report the falling back to the second device, release the model identity configured for the ML model, release a model-related configuration associated with the ML model, the model-related configuration associated with at least one of the following: measurement, reporting and behaviors of the first device, initiate a model identification procedure to identify another ML model, initiate a data collection procedure for the ML model, initiate a model update procedure for the ML model, initiate a model retraining procedure for the ML model, initiate a model deactivation procedure for the ML model, or determine that the ML mode is unknown to the first device.
  • the first device 110 may release the model identity by at least one of the following: releasing a correspondence between a first identity of the ML model and a second identity of the ML model, wherein the second identity is configured by the second device for the ML model and the first identity is determined by the first device; remapping the second identity previously configured for the ML model to another ML model; or removing the model identity from a list of ML model identities.
  • the first device 110 may transmit, to the second device, first information indicating a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device; and receive, from the second device, second information indicating a set of second identities of a second set of ML models comprising the ML model, wherein the second identity is configured by the second device.
  • the first device 110 may receive, from the second device, a model-related configuration associated with a second identity of the set of second identities, the model-related configuration being configured for at least one of the following: a model training procedure, a model inference procedure, a model monitoring procedure or a data collection and associated with at least one of the following: measurement, reporting and behaviors of the first device.
  • the second set of ML models is a subset of the first set of ML models.
  • the first identity is one of the following: a global model identity or a physical model identity, , or the first identity is used for identifying the ML model during at least of the following procedure: a model transfer procedure, a model delivery procedure, a model update procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the second identity is one of the following: a local model identity or a logical model identity, or the second identity is used for identifying the ML model during at least of the following procedure: a model inference procedure, a model management procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the first device 110 may transmit, to the second device, first information indicating a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device; receive, from the second device, third information indicating a set of temporary second identities of a second set of ML models comprising the ML model, wherein the temporary second identity is configured by the second device; and transmit, one of the following to the second device: confirmation information used for confirming at least one temporary second identity, contention information indicating at least one the following: at least temporary second identity is not applicable, or at least one recommended second identity in case that at least one temporary second identity comprised is not applicable.
  • the first device 110 may after transmitting the contention information, receive, from the second device, contention resolution information indicating at least one of the following: whether the at least one recommended second identity is confirmed by the second device, or at least one newly-configured second identity.
  • the first device 110 may prior to transmitting the first information, receive, from a third device, a third identity of the ML model configured by the third device; and transmit, to the second device, the first information comprising the third identity.
  • the third identity is represented in a hierarchical structure or a concatenation structure.
  • the first device 110 may determine whether the ML model is known to the first device based on at least of the following: cognition degree on the ML model, availability of model parameters or model structure, or an operation status of the ML mode.
  • a time length of the availability timer is defined as a default value or determined by the first device or the second device.
  • FIG. 5 illustrates a flowchart of a communication method 500 implemented at a second device in accordance with some embodiments of the present disclosure. For the purpose of discussion, the method 500 will be described from the perspective of the second device 120 in FIG. 1A.
  • the second device receives, from a first device, a request indicating the second device to release a model identity configured for a ML model known to the first device.
  • the second device release the model identity based on the request.
  • releasing the model identity comprising at least one of the following: releasing a correspondence between a first identity of the ML model and a second identity of the ML model, wherein the second identity is configured by the second device for the ML model and the first identity is determined by the first device; remapping the second identity previously configured for the ML model to another ML model; or removing the model identity from a list of ML model identities.
  • the second device may receive, from the first device, first information indicating a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device; and transmit, to the first device, second information indicating a set of second identities a second set of ML models comprising the ML model, wherein the second identity is configured by the second device for the ML model.
  • the second device may transmit, to the first device, a model-related configuration configured for at least one of the following: a model training procedure, a model inference procedure, a model monitoring procedure or a data collection and associated with a second identity of the set of second identities, the model-related configuration associated with at least one of the following: measurement, reporting and behavior of the first device.
  • the second set of ML models is a subset of the first set of ML models.
  • the first identity is one of the following: a global model identity or a physical model identity, or the first identity is used for identifying the ML model during at least of the following procedure: a model transfer procedure, a model delivery procedure, a model update procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the second identity is one of the following: a local model identity or a logical model identity, or the second identity is used for identifying the ML model during at least of the following procedure: a model inference procedure, a model management procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the second device may transmit, to the first device contention resolution information indicating at least one of the following: whether the at least one recommended second identity is confirmed by the second device, or at least one newly-configured second identity.
  • the second device may receive, from a third device, a fourth identity of the ML model configured by the third device; and receive, from the first device, the first information comprising a third identity of the ML model configured by the third device, wherein the third and the fourth identity are the same or different.
  • any of the third and fourth identity is represented in a hierarchical structure or a concatenation structure.
  • the second device may receive, from the first device and via at least one feature or at least one feature group, capability-related information of the first device, wherein the capability-related information indicates at least one of the following: a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device, the number of ML models supported for a specific feature or a feature group, or at least one model-related parameter related to a specific feature or a feature group.
  • the first device is a terminal device and the second device is a network device.
  • FIG. 6 is a simplified block diagram of a device 600 that is suitable for implementing embodiments of the present disclosure.
  • the device 600 can be considered as a further example implementation of any of the devices as shown in FIG. 1A. Accordingly, the device 600 can be implemented at or as at least a part of the first device 16 or the second device 120.
  • the device 600 includes a processor 610, a memory 620 coupled to the processor 610, a suitable transceiver 640 coupled to the processor 610, and a communication interface coupled to the transceiver 640.
  • the memory 630 stores at least a part of a program 630.
  • the transceiver 640 may be for bidirectional communications or a unidirectional communication based on requirements.
  • the transceiver 640 may include at least one of a transmitter 642 and a receiver 644.
  • the transmitter 642 and the receiver 644 may be functional modules or physical entities.
  • the transceiver 640 has at least one antenna to facilitate communication, though in practice an Access Node mentioned in this application may have several ones.
  • the communication interface may represent any interface that is necessary for communication with other network elements, such as X2/Xn interface for bidirectional communications between eNBs/gNBs, S1/NG interface for communication between a Mobility Management Entity (MME) /Access and Mobility Management Function (AMF) /SGW/UPF and the eNB/gNB, Un interface for communication between the eNB/gNB and a relay node (RN) , or Uu interface for communication between the eNB/gNB and a terminal device.
  • MME Mobility Management Entity
  • AMF Access and Mobility Management Function
  • RN relay node
  • Uu interface for communication between the eNB/gNB and a terminal device.
  • the program 630 is assumed to include program instructions that, when executed by the associated processor 610, enable the device 600 to operate in accordance with the embodiments of the present disclosure, as discussed herein with reference to FIGS. 1 to 6.
  • the embodiments herein may be implemented by computer software executable by the processor 610 of the device 600, or by hardware, or by a combination of software and hardware.
  • the processor 610 may be configured to implement various embodiments of the present disclosure.
  • a combination of the processor 610 and memory 620 may form processing means 650 adapted to implement various embodiments of the present disclosure.
  • the memory 620 may be of any type suitable to the local technical network and may be implemented using any suitable data storage technology, such as a non-transitory computer readable storage medium, semiconductor based memory devices, magnetic memory devices and systems, optical memory devices and systems, fixed memory and removable memory, as non-limiting examples. While only one memory 620 is shown in the device 600, there may be several physically distinct memory modules in the device 600.
  • the processor 610 may be of any type suitable to the local technical network, and may include one or more of general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs) and processors based on multicore processor architecture, as non-limiting examples.
  • the device 600 may have multiple processors, such as an application specific integrated circuit chip that is slaved in time to a clock which synchronizes the main processor.
  • a first device comprising a circuitry.
  • the circuitry is configured to: start an availability timer for a machine learning (ML) model known to the first device; and upon an expiry of the availability timer, transmit, to a second device, a request indicating the second device to release a model identity configured for the ML model associated with the availability timer.
  • the circuitry may be configured to perform any method implemented by the first device as discussed above.
  • a second device comprising a circuitry.
  • the circuitry is configured to: receive, from a first device, a request indicating the second device to release a model identity configured for a ML model known to the first device; and release the model identity based on the request.
  • the circuitry may be configured to perform any method implemented by the second device as discussed above.
  • circuitry used herein may refer to hardware circuits and/or combinations of hardware circuits and software.
  • the circuitry may be a combination of analog and/or digital hardware circuits with software/firmware.
  • the circuitry may be any portions of hardware processors with software including digital signal processor (s) , software, and memory (ies) that work together to cause an apparatus, such as a terminal device or a network device, to perform various functions.
  • the circuitry may be hardware circuits and or processors, such as a microprocessor or a portion of a microprocessor, that requires software/firmware for operation, but the software may not be present when it is not needed for operation.
  • the term circuitry also covers an implementation of merely a hardware circuit or processor (s) or a portion of a hardware circuit or processor (s) and its (or their) accompanying software and/or firmware.
  • a first apparatus comprises means for starting an availability timer for a machine learning (ML) model known to the first device; and means for upon an expiry of the availability timer, transmitting, to a second device, a request indicating the second device to release a model identity configured for the ML model associated with the availability timer.
  • the first apparatus may comprise means for performing the respective operations of the method 400.
  • the first apparatus may further comprise means for performing other operations in some example embodiments of the method 400.
  • the means may be implemented in any suitable form.
  • the means may be implemented in a circuitry or software module.
  • a second apparatus comprises means for receiving, from a first device, a request indicating the second device to release a model identity configured for a ML model known to the first device; and means for releasing the model identity based on the request.
  • the second apparatus may comprise means for performing the respective operations of the method 500.
  • the second apparatus may further comprise means for performing other operations in some example embodiments of the method 500.
  • the means may be implemented in any suitable form.
  • the means may be implemented in a circuitry or software module.
  • embodiments of the present disclosure provide the following aspects.
  • a first device comprising: a processor configured to cause the first device to: start an availability timer for a machine learning (ML) model known to the first device; and upon an expiry of the availability timer, transmit, to a second device, a request indicating the second device to release a model identity configured for the ML model associated with the availability timer.
  • ML machine learning
  • the processor is further configured to cause the first device to: restart the availability timer upon one of the following: a positive assessment result of the ML model, a first indication indicating the first device to continue applying the ML model, a model update procedure of the ML model, a model fine-tuning procedure of the ML model, or a model retraining procedure of the ML model.
  • the processor is further configured to cause the first device to: determine that the availability timer expiries upon one of the following: a negative assessment result of the ML model, a second indication indicating the first device not to continue applying the ML model, a change of configuration information configured for the ML model, a change of an application condition for the ML model at the first device, or a predefined event associated with radio resource management.
  • the processor is further configured to cause the first device to: fall back to a default ML model or an operation mode without ML model, report the falling back to the second device, release the model identity configured for the ML model, release a model-related configuration associated with the ML model, the model-related configuration associated with at least one of the following: measurement, reporting and behaviors of the first device, initiate a model identification procedure to identify another ML model, initiate a data collection procedure for the ML model, initiate a model update procedure for the ML model, initiate a model retraining procedure for the ML model, initiate a model deactivation procedure for the ML model, or determine that the ML mode is unknown to the first device.
  • the processor is further configured to cause the first device to: release the model identity by at least one of the following: releasing a correspondence between a first identity of the ML model and a second identity of the ML model, wherein the second identity is configured by the second device for the ML model and the first identity is determined by the first device; remapping the second identity previously configured for the ML model to another ML model; or removing the model identity from a list of ML model identities.
  • the processor is further configured to cause the first device to: transmit, to the second device, first information indicating a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device; and receive, from the second device, second information indicating a set of second identities of a second set of ML models comprising the ML model, wherein the second identity is configured by the second device.
  • the processor is further configured to cause the first device to: receive, from the second device, a model-related configuration associated with a second identity of the set of second identities, the model-related configuration being configured for at least one of the following: a model training procedure, a model inference procedure, a model monitoring procedure or a data collection and associated with at least one of the following: measurement, reporting and behaviors of the first device.
  • the second set of ML models is a subset of the first set of ML models.
  • the first identity is one of the following: a global model identity or a physical model identity, , or the first identity is used for identifying the ML model during at least of the following procedure: a model transfer procedure, a model delivery procedure, a model update procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the second identity is one of the following: a local model identity or a logical model identity, or the second identity is used for identifying the ML model during at least of the following procedure: a model inference procedure, a model management procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the processor is further configured to cause the first device to: transmit, to the second device, first information indicating a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device; receive, from the second device, third information indicating a set of temporary second identities of a second set of ML models comprising the ML model, wherein the temporary second identity is configured by the second device; and transmit, one of the following to the second device: confirmation information used for confirming at least one temporary second identity, contention information indicating at least one the following: at least temporary second identity is not applicable, or at least one recommended second identity in case that at least one temporary second identity comprised is not applicable.
  • the processor is further configured to cause the first device to: prior to transmitting the first information, receive, from a third device, a third identity of the ML model configured by the third device; and transmit, to the second device, the first information comprising the third identity.
  • the third identity is represented in a hierarchical structure or a concatenation structure.
  • the processor is further configured to cause the first device to: determine whether the ML model is known to the first device based on at least of the following: cognition degree on the ML model, availability of model parameters or model structure, or an operation status of the ML mode.
  • the processor is further configured to cause the first device to: determine the ML model is known to the first device if: at least part of a model structure is known to the first device, at least part of model parameters are known to the first device, a time length timed from a time point of the last model inference procedure of the ML model is less than or equal to a threshold time, a time length timed from the last model identification procedure of the ML model is less than or equal to a threshold time, a time length timed from the last performance monitoring procedure of the ML model is less than or equal to a threshold time, a time length timed from a completion of the last model training procedure, model fine-tuning procedure or model validation procedure of the ML model is less than or equal to a threshold time, a time length timed from the last model indication, model measurement, or model reporting related to the ML model is less than or equal to a threshold time, a time length timed from the last model activation, switching or selection of the ML model is less than or equal to
  • the processor is further configured to cause the first device to: in accordance with a determination that the ML model is unknown to the first device, performing at least one of the following: requesting another ML model with the second device, initiating a model identification procedure for another ML model, or applying a longer timer for a model inference procedure of the ML model.
  • the processor is further configured to cause the first device to: transmit, to the second device and via at least one feature or at least one feature group, capability-related information of the first device, wherein the capability-related information indicates at least one of the following: a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device, the number of ML models supported for a specific feature or a feature group, or at least one model-related parameter related to a specific feature or a feature group.
  • a time length of the availability timer is defined as a default value or determined by the first device or the second device.
  • the first device is a terminal device and the second device is a network device.
  • a second device comprising: a processor configured to cause the second device to: receive, from a first device, a request indicating the second device to release a model identity configured for a ML model known to the first device; and release the model identity based on the request.
  • releasing the model identity comprising at least one of the following: releasing a correspondence between a first identity of the ML model and a second identity of the ML model, wherein the second identity is configured by the second device for the ML model and the first identity is determined by the first device; remapping the second identity previously configured for the ML model to another ML model; or removing the model identity from a list of ML model identities.
  • the processor is further configured to cause the second device to: receive, from the first device, first information indicating a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device; and transmit, to the first device, second information indicating a set of second identities a second set of ML models comprising the ML model, wherein the second identity is configured by the second device for the ML model.
  • the processor is further configured to cause the second device to: transmit, to the first device, a model-related configuration configured for at least one of the following: a model training procedure, a model inference procedure, a model monitoring procedure or a data collection and associated with a second identity of the set of second identities, the model-related configuration associated with at least one of the following: measurement, reporting and behavior of the first device.
  • the second set of ML models is a subset of the first set of ML models.
  • the first identity is one of the following: a global model identity or a physical model identity, or the first identity is used for identifying the ML model during at least of the following procedure: a model transfer procedure, a model delivery procedure, a model update procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the second identity is one of the following: a local model identity or a logical model identity, or the second identity is used for identifying the ML model during at least of the following procedure: a model inference procedure, a model management procedure, a model monitoring procedure, a model training procedure, or a data collection procedure.
  • the processor is further configured to cause the second device to: receive, from the first device, first information indicating a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device; transmit, to the first device, third information indicating a set of temporary second identities of a second set of ML models comprising the ML model, wherein the temporary second identity is configured by the second device; and receive, one of the following from the first device: confirmation information used for confirming at least one temporary second identity, contention information indicating at least one the following: at least temporary second identity is not applicable, or at least one recommended second identity in case that at least one temporary second identity comprised is not applicable.
  • the processor is further configured to cause the second device to: after receiving the contention information, transmit, to the first device contention resolution information indicating at least one of the following: whether the at least one recommended second identity is confirmed by the second device, or at least one newly-configured second identity.
  • the processor is further configured to cause the first device to: prior to receiving the first information, receive, from a third device, a fourth identity of the ML model configured by the third device; and receive, from the first device, the first information comprising a third identity of the ML model configured by the third device, wherein the third and the fourth identity are the same or different .
  • any of the third and fourth identity is represented in a hierarchical structure or a concatenation structure.
  • the processor is further configured to cause the second device to: receive, from the first device and via at least one feature or at least one feature group, capability-related information of the first device, wherein the capability-related information indicates at least one of the following: a set of first identities of a first set of ML models comprising the ML model, wherein the first identity is determined by the first device, the number of ML models supported for a specific feature or a feature group, or at least one model-related parameter related to a specific feature or a feature group.
  • the first device is a terminal device and the second device is a network device.
  • a first device comprises: at least one processor; and at least one memory coupled to the at least one processor and storing instructions thereon, the instructions, when executed by the at least one processor, causing the device to perform the method implemented by the first device discussed above.
  • a computer readable medium having instructions stored thereon, the instructions, when executed on at least one processor, causing the at least one processor to perform the method implemented by the first device discussed above.
  • a computer program comprising instructions, the instructions, when executed on at least one processor, causing the at least one processor to perform the method implemented by the second device discussed above.
  • various embodiments of the present disclosure may be implemented in hardware or special purpose circuits, software, logic or any combination thereof. Some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software which may be executed by a controller, microprocessor or other computing device. While various aspects of embodiments of the present disclosure are illustrated and described as block diagrams, flowcharts, or using some other pictorial representation, it will be appreciated that the blocks, apparatus, systems, techniques or methods described herein may be implemented in, as non-limiting examples, hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof.
  • the present disclosure also provides at least one computer program product tangibly stored on a non-transitory computer readable storage medium.
  • the computer program product includes computer-executable instructions, such as those included in program modules, being executed in a device on a target real or virtual processor, to carry out the process or method as described above with reference to FIGS. 1 to 6.
  • program modules include routines, programs, libraries, objects, classes, components, data structures, or the like that perform particular tasks or implement particular abstract data types.
  • the functionality of the program modules may be combined or split between program modules as desired in various embodiments.
  • Machine-executable instructions for program modules may be executed within a local or distributed device. In a distributed device, program modules may be located in both local and remote storage media.
  • the above program code may be embodied on a machine readable medium, which may be any tangible medium that may contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
  • the machine readable medium may be a machine readable signal medium or a machine readable storage medium.
  • a machine readable medium may include but not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing.
  • machine readable storage medium More specific examples of the machine readable storage medium would include 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) , an optical fiber, a portable compact disc read-only memory (CD-ROM) , an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
  • RAM random access memory
  • ROM read-only memory
  • EPROM or Flash memory erasable programmable read-only memory
  • CD-ROM portable compact disc read-only memory
  • magnetic storage device or any suitable combination of the foregoing.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

Des modes de réalisation de la présente divulgation concernent une solution pour une gestion de cycle à base de temporisateur (LCM) d'un modèle d'apprentissage automatique (ML). Dans cette solution, le premier dispositif (tel qu'un dispositif terminal) met en marche un temporisateur de disponibilité pour un modèle d'apprentissage automatique (ML) connu du premier dispositif. Ensuite, lors d'une expiration du temporisateur de disponibilité, un premier dispositif transmet, à un second dispositif (tel qu'un dispositif de réseau/un autre dispositif terminal), une demande indiquant le second dispositif pour libérer une identité de modèle configurée pour le modèle ML associé au temporisateur de disponibilité.
PCT/CN2023/100799 2023-06-16 2023-06-16 Dispositifs et procédés de communication Ceased WO2024254864A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/100799 WO2024254864A1 (fr) 2023-06-16 2023-06-16 Dispositifs et procédés de communication

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/100799 WO2024254864A1 (fr) 2023-06-16 2023-06-16 Dispositifs et procédés de communication

Publications (1)

Publication Number Publication Date
WO2024254864A1 true WO2024254864A1 (fr) 2024-12-19

Family

ID=93851225

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/100799 Ceased WO2024254864A1 (fr) 2023-06-16 2023-06-16 Dispositifs et procédés de communication

Country Status (1)

Country Link
WO (1) WO2024254864A1 (fr)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021219201A1 (fr) * 2020-04-28 2021-11-04 Nokia Technologies Oy Commande d'opérations assistées par un apprentissage machine
CN113765957A (zh) * 2020-06-04 2021-12-07 华为技术有限公司 一种模型更新方法及装置
WO2022013093A1 (fr) * 2020-07-13 2022-01-20 Telefonaktiebolaget Lm Ericsson (Publ) Gestion d'un dispositif sans fil utilisable pour se connecter à un réseau de communication
WO2022077202A1 (fr) * 2020-10-13 2022-04-21 Qualcomm Incorporated Procédés et appareil de gestion de modèle de traitement ml
WO2022222089A1 (fr) * 2021-04-22 2022-10-27 Qualcomm Incorporated Rapport, repli et mise à jour d'un modèle d'apprentissage machine pour communications sans fil

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021219201A1 (fr) * 2020-04-28 2021-11-04 Nokia Technologies Oy Commande d'opérations assistées par un apprentissage machine
CN113765957A (zh) * 2020-06-04 2021-12-07 华为技术有限公司 一种模型更新方法及装置
WO2022013093A1 (fr) * 2020-07-13 2022-01-20 Telefonaktiebolaget Lm Ericsson (Publ) Gestion d'un dispositif sans fil utilisable pour se connecter à un réseau de communication
WO2022077202A1 (fr) * 2020-10-13 2022-04-21 Qualcomm Incorporated Procédés et appareil de gestion de modèle de traitement ml
WO2022222089A1 (fr) * 2021-04-22 2022-10-27 Qualcomm Incorporated Rapport, repli et mise à jour d'un modèle d'apprentissage machine pour communications sans fil

Similar Documents

Publication Publication Date Title
WO2024159533A1 (fr) Dispositif et procédé de communication
WO2025030519A1 (fr) Dispositifs et procédés de communication
WO2024197743A1 (fr) Dispositifs et procédés de communication
US20240414810A1 (en) Method, device and computer storage medium of communication
WO2026065165A1 (fr) Dispositifs et procédés de communication
WO2025171501A1 (fr) Dispositifs et procédés de communication
WO2025231672A1 (fr) Dispositifs et procédés de communication
WO2025199901A1 (fr) Dispositifs et procédés de communication
WO2026076680A1 (fr) Dispositifs et procédés de communication
WO2026065341A1 (fr) Dispositifs et procédés de communication
WO2025217780A1 (fr) Dispositifs et procédés de communication
WO2026060601A1 (fr) Dispositifs et procédés de communication
WO2024148476A1 (fr) Dispositifs et procédés de communication
WO2026007022A1 (fr) Dispositifs et procédés de communication
WO2025030281A1 (fr) Dispositifs et procédés de communication
WO2024174070A1 (fr) Procédés, dispositifs et support lisible par ordinateur destinés à la communication
WO2024212112A1 (fr) Dispositif et procédé de communication
WO2025217887A1 (fr) Dispositif, procédé et support lisible par ordinateur pour communication
WO2026081159A1 (fr) Dispositifs et procédés d'entraînement de modèle
WO2025199915A1 (fr) Procédé et appareil pour déterminer des faisceaux prédits à rapporter
WO2025260341A1 (fr) Dispositifs et procédés de communication
WO2024221241A1 (fr) Dispositifs et procédés de communication
WO2024187475A9 (fr) Dispositifs et procédés de communication
WO2026102611A1 (fr) Dispositif, procédé et support lisible par ordinateur pour la communication
WO2025199763A1 (fr) Dispositifs et procédés de communication

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23941097

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE