Model management method and apparatus, and terminal and network device

Abstract

Provided in the present disclosure are a model management method and apparatus, and a terminal and a network device. The method is applied to the terminal, and comprises: sending first information and second information to a network device, wherein the first information is used for indicating identifier information of at least one model supported by a terminal, and the second information is used for indicating a first terminal-side condition that corresponds to the terminal; or, sending third information to the network device, the third information being used for indicating at least one model supported by the terminal, wherein the model meets one or more of the following constraint conditions: under any terminal-side condition, the target performance of the model being greater than or equal to a first threshold value, under any terminal-side condition, the degradation degree of the target performance of the model being less than or equal to a second threshold value, and the model meeting the first terminal-side condition of the terminal.

Description

Model management method, device, terminal and network equipment

[0001] This disclosure claims priority to the Chinese patent application filed with the China Patent Office on April 3, 2024, with application number 202410401088.1 and application name “A model management method, device, terminal and network device”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] The present disclosure relates to the field of communication technology, and in particular to a model management method, apparatus, terminal, and network equipment. Background Art

[0003] In communication networks, communication devices such as base stations, terminal devices, or network-side devices can deploy and run artificial intelligence (AI) / machine learning (ML) models to support services such as positioning, beam management, and channel state information (CSI) feedback.

[0004] Limited by the scope of the training data set, an AI / ML model can usually only learn the logical relationship between input and output under certain conditions. At present, the above-mentioned certain conditions are divided into conditions and additional conditions. Among them, conditions usually refer to the configurations that can be applied to AI / ML models or functions, such as inputs and outputs configured through Radio Resource Control (RRC) signaling. User Equipment (UE) can report the configurations supported by its own AI / ML models or functions through UE capability reporting; additional conditions are not reflected in UE capability reporting, but are conditions assumed during the AI / ML model training process. Additional conditions are divided into additional conditions on the network (NW) side and additional conditions on the UE side. The consistency of additional conditions of an AI / ML model or function between the model training stage and the model processing stage may significantly affect the performance of the AI / ML model.

[0005] Currently, terminal-side models are primarily managed through the network, and only the consistency of additional conditions on the network side of the model can be guaranteed, resulting in inability to optimally manage model performance. Therefore, optimizing model management to ensure model performance is a technical issue that needs to be addressed. Summary of the Invention

[0006] The purpose of the present disclosure is to provide a model management method, apparatus, terminal and network device to solve the problem of how to optimize model management to ensure the performance of the model.

[0007] In a first aspect, in order to solve the above technical problems, the present disclosure provides a model management method applied to a terminal, comprising:

[0008] Sending first information and second information to a network device; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal;

[0009] or,

[0010] Sending third information to the network device, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0011] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0012] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0013] The model satisfies a first terminal-side condition of the terminal.

[0014] In some optional embodiments, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0015] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0016] In some optional embodiments, the above model management method further includes:

[0017] The meta information of the model is sent to the network device, where the meta information includes fourth information, and the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training phase.

[0018] In a second aspect, in order to solve the above technical problems, the present disclosure provides a model management method applied to a network device, including:

[0019] Receiving first information and second information sent by a terminal; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal;

[0020] or,

[0021] Receive third information sent by a terminal, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0022] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0023] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0024] The model satisfies a second terminal-side condition corresponding to the terminal.

[0025] In some optional embodiments, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0026] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0027] In some optional embodiments, the above model management method further includes:

[0028] Receive metadata of the model sent by the terminal, where the metadata includes fourth information, and the fourth information is used to indicate a second terminal-side condition corresponding to the at least one model during the training phase.

[0029] In a third aspect, in order to solve the above technical problems, the present disclosure provides a model management method applied to a terminal, including:

[0030] Receiving first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, where the at least one model is a subset of models supported by the terminal;

[0031] Determine a target model used by the terminal according to the first indication information or the second indication information.

[0032] In some optional embodiments, determining a first network-side condition corresponding to the network device according to the at least one model indicated by the second indication information;

[0033] The target model used by the terminal is determined according to the first network-side condition corresponding to the network device, the first terminal-side condition corresponding to the terminal, the second terminal-side condition and the second network-side condition corresponding to the at least one model in the training phase.

[0034] In some optional embodiments, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0035] In some optional embodiments, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0036] Applicable to the target community;

[0037] Applicable to cells belonging to the same cell group;

[0038] Applicable to communities belonging to the same manufacturer;

[0039] Applicable to cells belonging to the same operator;

[0040] Applicable to any community.

[0041] In some optional embodiments, models of different functions correspond to different network-side conditions during the training phase.

[0042] In some optional embodiments, the above model management method further includes:

[0043] Sending identification information of the target model used by the terminal to the network device.

[0044] In a fourth aspect, in order to solve the above technical problems, the present disclosure provides a model management method applied to a network device, including:

[0045] Sending first indication information or second indication information to the terminal; wherein, the first indication information is used to indicate a first network side condition corresponding to the network device; the second indication information is used to indicate identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

[0046] In some optional embodiments, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0047] In some optional embodiments, after sending the first indication information or the second indication information to the terminal, the method further includes:

[0048] Receive identification information of a target model sent by the terminal, where the target model is determined by the terminal according to the first indication information or the second indication information.

[0049] In some optional embodiments, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0050] Applicable to the target community;

[0051] Applicable to cells belonging to the same cell group;

[0052] Applicable to communities belonging to the same manufacturer;

[0053] Applicable to cells belonging to the same operator;

[0054] Applicable to any community.

[0055] In some optional embodiments, models of different functions correspond to different network-side conditions during the training phase.

[0056] In a fifth aspect, to solve the above technical problems, an embodiment of the present disclosure provides a terminal, comprising: a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process:

[0057] Sending first information and second information to a network device; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal;

[0058] or,

[0059] Sending third information to the network device, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0060] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0061] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0062] The model satisfies a first terminal-side condition of the terminal.

[0063] In some optional embodiments, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0064] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0065] In some optional embodiments, the processor is further configured to read a program in a memory and execute the following process:

[0066] The meta information of the model is sent to the network device, where the meta information includes fourth information, and the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training phase.

[0067] In a sixth aspect, to solve the above technical problems, an embodiment of the present disclosure provides a network device, comprising: a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process:

[0068] Receiving first information and second information sent by a terminal; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal;

[0069] or,

[0070] Receive third information sent by a terminal, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0071] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0072] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0073] The model satisfies a second terminal-side condition corresponding to the terminal.

[0074] In some optional embodiments, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0075] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0076] In some optional embodiments, the processor is further configured to read a program in a memory and execute the following process:

[0077] Receive metadata of the model sent by the terminal, where the metadata includes fourth information, and the fourth information is used to indicate a second terminal-side condition corresponding to the at least one model during a training phase.

[0078] In a seventh aspect, to solve the above technical problems, an embodiment of the present disclosure provides a terminal, comprising: a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process:

[0079] Receiving first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, where the at least one model is a subset of models supported by the terminal;

[0080] Determine a target model used by the terminal according to the first indication information or the second indication information.

[0081] In some optional embodiments, the processor is specifically configured to read a program in a memory and execute the following process:

[0082] Determining a first network-side condition corresponding to the network device according to the at least one model indicated by the second indication information;

[0083] The target model used by the terminal is determined according to the first network-side condition corresponding to the network device, the first terminal-side condition corresponding to the terminal, the second terminal-side condition and the second network-side condition corresponding to the at least one model in the training phase.

[0084] In some optional embodiments, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0085] In some optional embodiments, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0086] Applicable to the target community;

[0087] Applicable to cells belonging to the same cell group;

[0088] Applicable to communities belonging to the same manufacturer;

[0089] Applicable to cells belonging to the same operator;

[0090] Applicable to any community.

[0091] In some optional embodiments, models of different functions correspond to different network-side conditions during the training phase.

[0092] In some optional embodiments, the processor is further configured to read a program in a memory and execute the following process:

[0093] Sending identification information of the target model used by the terminal to the network device.

[0094] In an eighth aspect, to solve the above technical problems, an embodiment of the present disclosure provides a network device, comprising: a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process:

[0095] Sending first indication information or second indication information to the terminal; wherein, the first indication information is used to indicate a first network side condition corresponding to the network device; the second indication information is used to indicate identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

[0096] In some optional embodiments, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0097] In some optional embodiments, the processor is further configured to read a program in a memory and execute the following process:

[0098] Receive identification information of a target model sent by the terminal, where the target model is determined by the terminal according to the first indication information or the second indication information.

[0099] In some optional embodiments, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0100] Applicable to the target community;

[0101] Applicable to cells belonging to the same cell group;

[0102] Applicable to communities belonging to the same manufacturer;

[0103] Applicable to cells belonging to the same operator;

[0104] Applicable to any community.

[0105] In some optional embodiments, models of different functions correspond to different network-side conditions during the training phase.

[0106] In a ninth aspect, in order to solve the above technical problems, an embodiment of the present disclosure provides a model management device, applied to a terminal, comprising:

[0107] A first sending module is configured to send first information and second information to a network device, wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, to send third information to the network device, wherein the third information is used to indicate at least one model supported by the terminal, wherein the model satisfies one or more of the following constraints:

[0108] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0109] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0110] The model satisfies a first terminal-side condition of the terminal.

[0111] In a tenth aspect, in order to solve the above technical problems, an embodiment of the present disclosure provides a model management device, applied to a network device, comprising:

[0112] A first receiving module is configured to receive first information and second information sent by a terminal, wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, to receive third information sent by the terminal, wherein the third information is used to indicate at least one model supported by the terminal, wherein the model satisfies one or more of the following constraints:

[0113] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0114] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0115] The model satisfies a second terminal-side condition corresponding to the terminal.

[0116] In an eleventh aspect, in order to solve the above technical problems, an embodiment of the present disclosure provides a model management device, applied to a terminal, comprising:

[0117] a second receiving module, configured to receive first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, wherein the at least one model is a subset of models supported by the terminal;

[0118] A determination module is used to determine the target model used by the terminal according to the first indication information or the second indication information.

[0119] In a twelfth aspect, in order to solve the above technical problems, an embodiment of the present disclosure provides a model management device, applied to a network device, comprising:

[0120] The second sending module is used to send the first indication information or the second indication information to the terminal; wherein the first indication information is used to indicate the first network side condition corresponding to the network device; the second indication information is used to indicate the identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

[0121] In the thirteenth aspect, in order to solve the above-mentioned technical problems, an embodiment of the present disclosure provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute the steps of the method described in the first aspect, the second aspect, the third aspect, or the fourth aspect above.

[0122] In the thirteenth aspect, in order to solve the above technical problems, the embodiment of the present disclosure provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method described in the first aspect, the second aspect, the third aspect, or the fourth aspect above.

[0123] The beneficial effects of the above technical solutions disclosed herein are as follows:

[0124] In the above solution, the terminal sends first and second information to the network device; the first information is used to indicate the identification information of at least one model supported by the terminal, and the second information is used to indicate the first terminal-side condition corresponding to the terminal. Alternatively, the terminal sends third information to the network device, and the third information is used to indicate at least one model supported by the terminal; the model satisfies one or more of the following constraints: under any terminal-side condition, the target performance of the model is greater than or equal to the first threshold value; under any terminal-side condition, the degree of degradation of the target performance of the model is less than or equal to the second threshold value; the model satisfies the first terminal-side condition of the terminal. In this way, the network device can perform model management based on the first and second information, or perform model management based on the third information, to optimize model management and ensure model performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0125] FIG1 is a flow chart of a model management method according to an embodiment of the present disclosure;

[0126] FIG2 is a second flow chart of the model management method according to an embodiment of the present disclosure;

[0127] FIG3 is a structural block diagram of a model management device according to an embodiment of the present disclosure;

[0128] FIG4 is a second structural block diagram of the model management device according to an embodiment of the present disclosure;

[0129] FIG5 is a schematic diagram of a hardware structure of a terminal according to an embodiment of the present disclosure;

[0130] FIG6 is a schematic diagram of a hardware structure of a network device according to an embodiment of the present disclosure;

[0131] FIG7 is a third flow chart of the model management method according to an embodiment of the present disclosure;

[0132] FIG8 is a fourth flow chart of the model management method according to an embodiment of the present disclosure;

[0133] FIG9 is a third structural block diagram of the model management device according to an embodiment of the present disclosure;

[0134] FIG10 is a fourth structural block diagram of the model management device according to an embodiment of the present disclosure;

[0135] FIG11 is a second schematic diagram of the hardware structure of the terminal according to an embodiment of the present disclosure;

[0136] FIG12 is a second schematic diagram of the hardware structure of the network device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0137] The following will be combined with the accompanying drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the embodiments described are only part of the embodiments of the present disclosure and not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0138] In the embodiments of the present disclosure, the term "and / or" describes the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0139] In the embodiments of the present disclosure, the term "plurality" refers to two or more than two, and other quantifiers are similar thereto.

[0140] The terminal devices involved in the embodiments of the present disclosure may be devices that provide voice and / or data connectivity to users, handheld devices with wireless connection capabilities, or other processing devices connected to wireless modems. In different systems, the names of terminal devices may also be different. For example, in the fifth generation mobile communication technology (5G) system, the terminal device may be called user equipment (UE). Wireless terminal devices can communicate with one or more core networks (CN) via a radio access network (RAN). Wireless terminal devices can be mobile terminal devices, such as mobile phones (or "cellular" phones) and computers with mobile terminal devices. For example, they can be portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile devices that exchange voice and / or data with the radio access network. For example, personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), and other devices. The wireless terminal device may also be referred to as a system, a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, an access point, a remote terminal device, an access terminal device, a user terminal device, a user agent, or a user device, but is not limited in the embodiments of the present disclosure.

[0141] The network device involved in the embodiments of the present disclosure may be a base station, which may include multiple cells providing services to terminals. Depending on the specific application scenario, the base station may also be called an access point, or may be a device in an access network that communicates with a wireless terminal device through one or more sectors on an air interface, or may be called another name. The network device may be used to interchange received air frames with Internet Protocol (IP) packets, acting as a router between the wireless terminal device and the rest of the access network, wherein the rest of the access network may include an Internet Protocol (IP) communication network. The network device may also coordinate the attribute management of the air interface. For example, the network device involved in the embodiments of the present disclosure may be a base transceiver station (BTS) in the Global System for Mobile communications (GSM) or code division multiple access (CDMA), a network device (NodeB) in wide-band code division multiple access (WCDMA), an evolutionary Node B (eNB or e-NodeB) in the long term evolution (LTE) system, a 5G base station (gNB) in the 5G network architecture (next generation system), a home evolved Node B (HeNB), a relay node, a femto, a pico, etc., and is not limited in the embodiments of the present disclosure. In some network structures, the network device may include a centralized unit (CU) node and a distributed unit (DU) node, and the centralized unit and the distributed unit may also be geographically separated.

[0142] Network devices and terminal devices can each use one or more antennas for Multiple Input Multiple Output (MIMO) transmission. MIMO transmission can be single-user MIMO (SU-MIMO) or multi-user MIMO (MU-MIMO). Depending on the configuration and number of antenna combinations, MIMO transmission can be two-dimensional MIMO (2D-MIMO), three-dimensional MIMO (3D-MIMO), full-dimensional MIMO (FD-MIMO), or massive MIMO. It can also use diversity transmission, precoding, or beamforming.

[0143] The following first introduces the contents involved in the solution provided by the embodiment of the present disclosure.

[0144] Currently, the terminal-side model is mainly managed by the network side, but the network side cannot obtain the current terminal-side additional conditions (i.e., the terminal-side conditions in the model inference phase). It can only ensure the consistency of the network-side conditions in the model training phase and the model inference phase, and therefore cannot guarantee the performance of the model.

[0145] For ease of understanding, the following examples are used to illustrate this. Assume that a terminal initiates model recognition on a network device, and the network identifies four models, each applicable to different network-side additional conditions (such as the network-side beam pattern) and terminal-side additional conditions (such as the terminal's mobility). The applicability of these additional conditions is communicated to the network through the model identifier (ID) or meta information. These additional conditions refer to the conditions assumed during the model training phase.

[0146] Exemplarily, the additional network-side conditions and terminal-side conditions assumed by different models during the training phase are shown in Table 1 below.

[0147] Table 1 - Different models correspond to different additional conditions on the network side and terminal side

[0148] Assuming the network's current beam pattern is NW beam pattern 1, Table 1 above indicates that the network will use either model #1a or model #1b, while models #2a and #2b will not be used. However, since the network does not know the UE's speed, it cannot determine which of these two models to use. Therefore, the network can only randomly select one of these, and cannot guarantee the consistency of the terminal's additional conditions during model training and inference, thus failing to guarantee model performance.

[0149] It should be noted that the above examples are for reference only. It is also possible that a model is trained using data collected from multiple network sides, so a model can correspond to multiple additional network conditions, and the same applies to additional conditions on the UE side.

[0150] Based on the above, the embodiments of the present disclosure provide a model management method to solve the problem of how to optimize model management to ensure the performance of the model.

[0151] Referring to Figure 1 , an embodiment of the present disclosure provides a model management method, which is applied to a terminal. The model may include an artificial intelligence (AI) model, or AI function, or a machine learning (ML) model, or ML function; alternatively, the model may be referred to as an AI / ML model, or AI / ML function. The terminal may be a user equipment (UE).

[0152] Specifically, the model management method includes the following steps:

[0153] Step 11: Sending first information and second information to the network device; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or,

[0154] Sending third information to the network device, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0155] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0156] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0157] The model satisfies a first terminal-side condition of the terminal.

[0158] Among them, network equipment includes but is not limited to: base stations, fifth-generation mobile communication system base stations (next Generation Node B, gNB), transmission reception points (Transmission Reception Point, TRP) and location management function (Location Management Function, LMF) entities.

[0159] It should be noted that the first terminal-side condition refers to the terminal-side condition corresponding to the model inference phase, and can be understood as the terminal-side condition that the terminal possesses during the model inference phase. This is different from the second terminal-side condition corresponding to the model training phase, which is the terminal-side condition assumed during the model training phase, that is, the terminal-side condition actually used during the model training process. The first terminal-side condition and the second terminal-side condition may include, but are not limited to, the terminal's speed, the terminal's receive antenna selection, and the terminal's receive beam pattern.

[0160] The following describes a solution for sending the first information and the second information to the network device.

[0161] As an example, the identification information of at least one model supported by the terminal indicated by the first information may be a model ID.

[0162] For example, in the model training phase, model training is performed in UE#1; the ID of the trained model is recorded as model ID#1, and the model may be deployed to multiple terminals, not limited to UE#1, such as deployed to UE#2; wherein, the training data of the model may also come from several different UEs or NWs; in the model inference preparation phase, UE#2 initiates model identification to NW (such as NW#2), and UE#2 sends first information to NW#2 to indicate the model ID of at least one model supported by UE#2.

[0163] As another example, the identification information of at least one model supported by the terminal indicated by the first information is a network side identifier (such as NW ID), and one network side identifier represents one network side condition or a combination of multiple network side conditions.

[0164] It should be noted that the network-side conditions here refer to those imposed by network equipment during the model training phase. These are additional conditions that are not reflected in terminal capability reporting but are assumed during model training. These network-side conditions may include, but are not limited to, the gNB's beam shape and beam pattern.

[0165] For example, during the data collection phase, the network device indicates an ID (such as NW ID#1) to UE#1, associating it with the data collection configuration / process. The data collected by UE#1 is associated with NW ID#1, which represents a network-side condition or a combination of multiple network-side conditions. During the model inference preparation phase, UE#2 initiates model identification to the NW (such as NW#2), and UE#2 sends first information to NW#2 indicating the NW ID#1 corresponding to at least one model supported by UE#2. It should be noted that NW ID#1 is used here only as an example and may be represented in other ways, as long as it can serve as identification information representing network-side conditions / additional network-side conditions.

[0166] In a specific implementation, the first terminal-side condition corresponding to the terminal indicated by the second information may be a condition ID or a specific physical meaning. In the case of a condition ID, the terminal reports one or more condition IDs representing the first terminal-side condition to the network device; in the case of a specific physical meaning, the terminal may report to the network device the specific values ​​of various physical meanings or combinations of physical meanings, such as the specific values ​​of one or more of the following: the terminal's speed, the terminal's receiving antenna selection, the terminal's receiving antenna style, the terminal's timing offset, the terminal's sampling frequency deviation, and the like.

[0167] In this solution, the terminal sends first information and second information to the network device, so that the network device can obtain at least one model supported by the terminal and the first terminal-side condition corresponding to the terminal in the model inference stage based on the first information and the second information, so that the network device can manage at least one model reported by the terminal based on the first terminal-side condition, as well as the first network-side condition corresponding to the model inference stage and the second network-side condition corresponding to the model training stage known to the network device itself. The management operation may include one or more of the following: activating / deactivating / selecting / switching / rolling back the at least one model. The specific management operation can be determined based on the model management indication information sent by the network device. For example, the terminal receives the model management indication information sent by the network device; according to the model management indication information, the target model is selected from the at least one model reported by the terminal, so that the performance of the target model meets the first terminal-side condition, and the first network-side condition is ensured to match the second network-side condition, so as to ensure the performance of the terminal-side model.

[0168] In some optional embodiments, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information; wherein, the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training stage.

[0169] It should be noted that the second terminal-side condition is an additional condition that is not reflected in the terminal capability report and is assumed during the model training process. The second terminal-side condition may include, but is not limited to: terminal speed, terminal receive antenna selection, terminal receive beam pattern, terminal sampling frequency deviation, etc.

[0170] It should be understood that the association relationship between the first information and the fourth information may indicate a mapping relationship between the first information and the fourth information, that is, the fourth information may be determined based on the first information, or the fourth information may be determined based on the first information and / or other information.

[0171] Exemplarily, the implementation of part or all of the first information as the fourth information may include: a part of the model ID field is composed of the condition ID corresponding to the first terminal side condition. It should be noted that this example is applicable to the case where the second terminal side condition is indicated by the condition ID.

[0172] In the above embodiment, the network device can obtain at least one model supported by the terminal and the second terminal-side condition corresponding to the terminal in the model training phase through the first information sent by the terminal device, so that the network device can manage at least one model reported by the terminal based on the second terminal-side condition, combined with the first terminal-side condition corresponding to the model reasoning phase reported by the terminal, as well as the first network-side condition corresponding to the model reasoning phase and the second network-side condition corresponding to the model training phase known to the network device itself. The management operation may include one or more of the following: activating / deactivating / selecting / switching / rolling back the at least one model. Among them, the specific management operation can be determined based on the model management indication information sent by the network device.

[0173] For example, the terminal receives model management indication information sent by the network device; according to the model management indication information, a target model is selected from at least one model reported by the terminal, so that the first terminal side condition matches the second terminal side condition, and the first network side condition matches the second network side condition, so as to ensure the usage performance of the terminal side model.

[0174] It should be noted that the matching of the first terminal side condition and the second terminal side condition may include: the first terminal side condition and the second terminal side condition are exactly the same, or partially the same, or similar. Alternatively, the performance of a model (such as a reference model) under the first terminal side condition and the second terminal side condition is the same or close, and if the performance difference is within a predefined range, then the first terminal side condition is considered to match the second terminal side condition. Similarly, the matching of the first network side condition and the second network side condition may include: the first network side condition and the second network side condition are exactly the same, or partially the same, or similar. Alternatively, the performance of a model (such as a reference model) under the first network side condition and the second network side condition is the same or close, and if the performance difference is within a predefined range, then the first network side condition is considered to match the second network side condition.

[0175] It should be pointed out that in the solution of managing the terminal side model through network equipment, the first network side condition corresponding to the model inference stage and the second network side condition corresponding to the network equipment in the model training stage are known to the network equipment, and the network equipment does not need to obtain additional information from the terminal side.

[0176] In some optional embodiments, the above model management method further includes:

[0177] The meta information of the model is sent to the network device, where the meta information includes fourth information, and the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training phase.

[0178] In specific implementations, during the model recognition process, the terminal sends the model's metadata to the network device. This metadata can also be referred to as model description information or related information. A model's metadata often describes the model's attributes, characteristics, and purpose.

[0179] In the above embodiment, the network device can obtain the second terminal-side condition corresponding to the terminal in the model training phase through the metadata of the model sent by the terminal device, so that the network device can manage at least one model reported by the terminal based on the second terminal-side condition, while combining the first terminal-side condition corresponding to the model reasoning phase reported by the terminal through the second information, as well as the first network-side condition corresponding to the model reasoning phase and the second network-side condition corresponding to the model training phase known to the network device itself. The management operation may include one or more of the following: activating / deactivating / selecting / switching / rolling back the at least one model. Among them, the specific management operation can be determined based on the model management indication information sent by the network device.

[0180] For example, the terminal receives model management indication information sent by the network device; according to the model management indication information, a target model is selected from at least one model reported by the terminal, so that the first terminal-side condition corresponding to the terminal in the model inference stage matches the second terminal-side condition assumed by the target model in the training stage, and the first network-side condition corresponding to the network device in the model inference stage matches the second network-side condition assumed by the target model in the training stage.

[0181] In this way, the network device side can not only ensure that the second network side condition corresponding to the selected target model in the model training phase matches the first network side condition corresponding to the network device in the model inference phase, but also ensure that the second terminal side condition corresponding to the selected target model in the model training phase matches the first terminal side condition corresponding to the terminal in the model inference phase, thereby enabling the most appropriate management of the terminal side model and ensuring the model's performance.

[0182] The following describes an embodiment of sending the third information to the network device.

[0183] In a specific implementation, during the model identification process, at least one model reported by the terminal to the network device satisfies at least one of the following three constraints:

[0184] Constraint 1: Under any terminal-side conditions, the target performance of the model is greater than or equal to a first threshold.

[0185] Optionally, the first threshold is a preset value; or the first threshold is indicated by the network device configuration. It should be noted that the first threshold may be different depending on the target performance of the model.

[0186] The target performance of the model includes but is not limited to: squared general cosine similarity (SGCS) of channel state information (CSI), system throughput, probability of optimal beam, deviation of predicted position, deviation of intermediate measurement quantity, etc.

[0187] For example, a model for CSI feedback enhancement may ensure that, under any terminal-side conditions, the SGCS of the CSI corresponding to the model's predicted output is not lower than a certain threshold value; or, under any terminal-side conditions, the system throughput corresponding to the model's predicted output is not lower than a certain threshold value.

[0188] For example, in a model for beam management (BM), under any terminal-side condition, the probability that the best K beams corresponding to the output predicted by the model are optimal beams is not less than a first threshold.

[0189] For example, a model for positioning, under any terminal side condition, the deviation between the UE position corresponding to the output predicted by the model and the actual position of the UE is not higher than a certain threshold value with a probability of X%, that is, the position accuracy corresponding to the position predicted by the model is greater than or equal to the first threshold value, where X is a preset value; or, under any terminal side condition, the deviation between the intermediate measurement amount related to positioning corresponding to the output predicted by the model and the intermediate measurement amount corresponding to the actual position of the UE is not higher than a certain threshold value with a probability of Y%, that is, the measurement deviation corresponding to the intermediate measurement amount predicted by the model is greater than or equal to the first threshold value, where Y is a preset value.

[0190] In this way, if the model reported by the terminal meets constraint condition 1, although the network device can only manage the terminal-side model under the condition of meeting the consistency of the network-side conditions in the model training phase and the model inference phase, it can also guarantee a certain model performance without the need for additional information transmission overhead.

[0191] Constraint 2: Under any terminal-side condition, the target performance degradation of the model is less than or equal to a second threshold value.

[0192] Optionally, the second threshold is a preset value; or the second threshold is indicated by the network device configuration. It should be noted that the second threshold may be different depending on the target performance of the model.

[0193] Specifically, the target performance of the terminal-side model indicated by the network device, under any terminal-side conditions, does not degrade by more than a second threshold value compared to an ideal situation where a first terminal-side condition corresponding to the terminal in the model inference phase is consistent with a second terminal-side condition corresponding to the model training phase. The target performance of the model includes, but is not limited to, SGCS of CSI, system throughput, probability of optimal beam, deviation of predicted position, deviation of intermediate measurement quantities, and the like.

[0194] In this way, if the model reported by the terminal meets constraint condition 2, although the network device can only manage the terminal-side model under the condition of consistency of the network-side conditions in the model training phase and the model inference phase, it can also guarantee a certain model performance without the need for additional information transmission overhead.

[0195] Constraint 3: The model satisfies the first terminal-side condition of the terminal.

[0196] It can be understood that the model is trained under the assumption that the second terminal side condition is met. If the model reported by the terminal meets the first terminal side condition corresponding to the terminal in the model reasoning stage, then the model reported by the terminal can meet the consistency of the terminal side conditions in the model training stage and the model reasoning stage.

[0197] In a specific implementation, a network device receives at least one model reported by a terminal and manages the terminal-side model by determining whether the first network condition matches the second network condition based on a first network condition corresponding to the model inference phase and a second network condition corresponding to the model training phase. The management operation may include one or more of the following: activating / deactivating / selecting / switching / rolling back the at least one model.

[0198] For example, to ensure model performance, the network device selects and activates only the target model that meets the network-side conditions for consistency during the model training and model inference phases from at least one model reported by the terminal. This ensures consistency between the network-side and terminal-side conditions during both the model training and model inference phases, guaranteeing the performance of the terminal-side model.

[0199] For example, using Table 1 above as an example, assume that four models are deployed on the terminal side, and the terminal is a static terminal placed on the warehouse cargo, that is, the terminal is in a low-speed state. In the model recognition phase, the terminal only identifies model#1a and model#2a to the network device, or only indicates to the network device that it supports model#1a and model#2a. In this case, regardless of which model#1a or model#2a the network device indicates, the terminal-side consistency conditions for the model training phase and the model inference phase are met. Assuming the network device's beam pattern is 2, the network device instructs the terminal to use model#2a.

[0200] Referring to FIG2 , an embodiment of the present disclosure provides a model management method applied to a network device, including:

[0201] Step 21: Receive first information and second information sent by the terminal; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or,

[0202] Receive third information sent by a terminal, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0203] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0204] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0205] The model satisfies a second terminal-side condition corresponding to the terminal.

[0206] In the above embodiments, in one solution, the terminal sends the first information and the second information to the network device, so that the network device can obtain, based on the first information and the second information, at least one model supported by the terminal and the first terminal-side condition corresponding to the terminal in the model reasoning stage, so that the network device can manage at least one model reported by the terminal according to the first terminal-side condition, as well as the first network-side condition corresponding to the model reasoning stage and the second network-side condition corresponding to the model training stage known to the network device itself. The management operation may include one or more of the following: activate / deactivate / select / switch / rollback the at least one model. Among them, the specific management operation can be determined based on the model management indication information sent by the network device. In another solution, the network device side can not only ensure that the second network-side condition corresponding to the selected target model in the model training stage matches the first network-side condition corresponding to the network device in the model reasoning stage, but also ensure a certain model performance without the need for additional information transmission overhead.

[0207] In some optional embodiments, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information; wherein, the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training stage.

[0208] It should be noted that the explanation examples here are consistent with the terminal side embodiments, and will not be repeated here to avoid repetition.

[0209] In the above embodiment, the network device can obtain at least one model supported by the terminal and the second terminal-side condition corresponding to the terminal in the model training phase through the first information sent by the terminal device, so that the network device can manage at least one model reported by the terminal based on the second terminal-side condition, combined with the first terminal-side condition corresponding to the model reasoning phase reported by the terminal, as well as the first network-side condition corresponding to the model reasoning phase and the second network-side condition corresponding to the model training phase known to the network device itself. The management operation may include one or more of the following: activating / deactivating / selecting / switching / rolling back the at least one model. Among them, the specific management operation can be determined based on the model management indication information sent by the network device.

[0210] In some optional embodiments, the above model management method further includes:

[0211] Receive metadata of the model sent by the terminal, where the metadata includes fourth information, and the fourth information is used to indicate a second terminal-side condition corresponding to the at least one model during the training phase.

[0212] It should be noted that the explanation examples here are consistent with the terminal side embodiments, and will not be repeated here to avoid repetition.

[0213] In the above embodiment, the network device can obtain the second terminal-side condition corresponding to the terminal in the model training phase through the metadata of the model sent by the terminal device, so that the network device can manage at least one model reported by the terminal based on the second terminal-side condition, while combining the first terminal-side condition corresponding to the model reasoning phase reported by the terminal through the second information, as well as the first network-side condition corresponding to the model reasoning phase and the second network-side condition corresponding to the model training phase known to the network device itself. The management operation may include one or more of the following: activating / deactivating / selecting / switching / rolling back the at least one model. Among them, the specific management operation can be determined based on the model management indication information sent by the network device.

[0214] It should be noted that the above-described embodiments of the present disclosure are primarily applicable to 5G NR systems, including network equipment and terminal devices. However, the present invention can also be applied to other systems where terminal-side AI / ML models are required. For example, the system includes multiple UEs, including UE1 and UE2, requesting wireless network connection services; the gNB provides wireless services for them. The gNB and UE1 and UE2 exchange and transmit data via wireless communication. For example, the gNB provides AI / ML-related services to UE1 and UE2, including transmitting various indication information to UE1 and UE2.

[0215] The following introduces the model management method of the terminal side model managed by the network device, which can mainly include the following two solutions.

[0216] Option 1:

[0217] In step 1, during the data collection phase, the network device (possibly NW#1) indicates an ID (e.g., NW ID#1) to UE#1, and associates it with the data collection configuration / process. The data collected by UE#1 is also associated with NW ID#1. NW ID#1 represents the second network-side condition.

[0218] In step 2, during the model training phase, UE#1 performs model training. The ID of the trained terminal-side model is recorded as model ID#1. The model may be deployed to multiple UEs, not just UE#1, such as UE#2. The model training data may also come from several different UEs or NWs.

[0219] Step 3: In the model inference preparation phase, UE#2 initiates model identification to the network device (such as NW#2) and sends first information and second information to NW#2. The first information is used to indicate the model ID(s) of one or more models supported by UE#2, and the second information is used to indicate the first terminal-side condition corresponding to UE#2 in the model inference phase.

[0220] Optionally, model ID#1 is associated with the second terminal side condition of the model in the training phase, or a portion of the fields of model ID#1 is composed of the second terminal side condition of the model in the training phase.

[0221] Optionally, the second terminal-side condition of the model during the training phase is indicated to NW#2 through meta information (meta info) of the model.

[0222] Model ID#1 may be the NW ID#1 indicated by NW#1 in step 1, or there may be a predefined association between model ID#1 and NW ID#1, or NW ID#1 may be indicated to NW#2 via model meta information (meta info).

[0223] Step 4: In the model inference phase, NW#2 receives the model ID reported by UE#2, the second terminal side condition corresponding to the model in the training phase, and the first terminal side condition corresponding to the terminal in the inference phase; by judging whether the first network side condition (NW ID#2) corresponding to the network device in the model inference phase and the second network side condition (NW ID#1) corresponding to the model training phase match, and whether the first terminal side condition corresponding to the terminal in the model inference phase matches the second terminal side condition assumed in the model training phase, select which target models to use for the terminal from the model ID reported by UE#2.

[0224] Through the above-mentioned solution 1, the network device can not only ensure the consistency of the network-side conditions of the terminal-side model, but also ensure the consistency of the terminal-side conditions, thereby enabling the most appropriate management of the terminal-side model. Among them, the consistency of the network-side conditions can be understood as the second network-side condition in the model training phase matches the first network-side condition of the network device in the model inference phase, and the consistency of the terminal-side conditions can be understood as the second terminal-side condition corresponding to the model training phase matches the first terminal-side condition corresponding to the terminal in the model inference phase.

[0225] Option 2:

[0226] In step 1, during the data collection phase, the network device (possibly NW#1) indicates an ID (e.g., NW ID#1) to UE#1, and associates it with the data collection configuration / process. The data collected by UE#1 is also associated with NW ID#1. NW ID#1 represents the second network-side condition.

[0227] In step 2, during the model training phase, UE#1 performs model training. The ID of the trained terminal-side model is recorded as model ID#1. The model may be deployed to multiple UEs, not just UE#1, such as UE#2. The model training data may also come from several different UEs or NWs.

[0228] Step 3: In the model inference preparation phase, UE#2 initiates model identification to a network device (e.g., NW#2) and sends third information to NW#2. The third information is used to indicate the model ID(s) of one or more models supported by UE#2.

[0229] Model ID#1 may be the NW ID#1 indicated by NW#1 in step 1, or there may be a predefined association between model ID#1 and NW ID#1, or NW ID#1 may be indicated to NW#2 via model meta information (meta info).

[0230] The model supported by UE#2 satisfies one or more of the following constraints:

[0231] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0232] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0233] The first terminal-side condition corresponding to the terminal in the model inference phase is satisfied.

[0234] Step 4: In the model inference phase, NW#2 can instruct the UE which target models to use by judging whether the first network side condition (NW ID#2) corresponding to the network device in the model inference phase and the second network side condition (NW ID#1) corresponding to the model training phase match based on the model ID reported by UE#2 and the corresponding network side conditions behind it.

[0235] With Solution 2, although the network device can only manage the terminal-side model while ensuring consistency of network-side conditions, it can still guarantee certain model performance without requiring additional information transmission overhead. Consistency of network-side conditions can be understood as matching the second network-side condition during the model training phase with the first network-side condition of the network device during the model inference phase.

[0236] 3 , an embodiment of the present disclosure provides a model management device 300, which is applied to a terminal and includes:

[0237] The first sending module 301 is configured to send first information and second information to a network device, wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, to send third information to the network device, wherein the third information is used to indicate at least one model supported by the terminal, wherein the model satisfies one or more of the following constraints:

[0238] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0239] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0240] The model satisfies a second terminal-side condition of the terminal.

[0241] Optionally, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0242] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0243] Optionally, the first sending module 301 is further configured to:

[0244] The meta information of the model is sent to the network device, where the meta information includes fourth information, and the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training phase.

[0245] It should be noted here that the above-mentioned device 300 provided in the embodiment of the present disclosure can implement all the method steps implemented by the method embodiment applied to the terminal side as shown in the above-mentioned Figure 1, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0246] 4 , an embodiment of the present disclosure provides a model management apparatus 400, which is applied to a network device and includes:

[0247] The first receiving module 401 is configured to receive first information and second information sent by a terminal, wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, to receive third information sent by the terminal, wherein the third information is used to indicate at least one model supported by the terminal, wherein the model satisfies one or more of the following constraints:

[0248] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0249] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0250] The model satisfies a second terminal-side condition corresponding to the terminal.

[0251] Optionally, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0252] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0253] Optionally, the first receiving module 401 is further configured to:

[0254] Receive metadata of the model sent by the terminal, where the metadata includes fourth information, and the fourth information is used to indicate a second terminal-side condition corresponding to the at least one model during the training phase.

[0255] It should be noted here that the above-mentioned device 400 provided in the embodiment of the present disclosure can implement all the method steps implemented by the method embodiment applied to the network device side as shown in the above-mentioned Figure 2, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0256] Referring to Figure 5 , an embodiment of the present disclosure provides a terminal, comprising: a processor 510; and a memory 520 connected to the processor 510 via a bus interface. The memory 520 is used to store programs and data used by the processor 510 when performing operations, and the processor 510 calls and executes the programs and data stored in the memory 520. A transceiver 500 is connected to the bus interface and is used to receive and send data under the control of the processor 510. The processor 510 is used to read the program in the memory 520 and perform the following processes:

[0257] Sending first information and second information to a network device; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or,

[0258] Sending third information to the network device, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0259] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0260] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0261] The model satisfies a second terminal-side condition of the terminal.

[0262] Optionally, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0263] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0264] Optionally, the processor 510 is further configured to read a program in the memory 520 and execute the following process:

[0265] The meta information of the model is sent to the network device, where the meta information includes fourth information, and the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training phase.

[0266] In FIG5 , the bus architecture may include any number of interconnected buses and bridges, specifically various circuits connected together by one or more processors represented by processor 510 and memory represented by memory 520. The bus architecture may also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not described further herein. The bus interface provides an interface. The transceiver 500 may be a plurality of components, including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, such as a wireless channel, a wired channel, an optical cable, and the like. For different user devices, the user interface 530 may also be an interface capable of connecting external or internal devices as required, and the connected devices include but are not limited to a keypad, a display, a speaker, a microphone, a joystick, and the like.

[0267] The processor 510 is responsible for managing the bus architecture and general processing, and the memory 520 can store data used by the processor 510 when performing operations.

[0268] In some embodiments, the processor 510 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

[0269] The processor calls the computer program stored in the memory to execute any of the methods provided by the embodiments of the present disclosure according to the obtained executable instructions. The processor and the memory can also be arranged physically separately.

[0270] Referring to Figure 6, an embodiment of the present disclosure provides a network device, including: a processor 610; and a memory 620 connected to the processor 610 through a bus interface, wherein the memory 620 is used to store programs and data used by the processor 610 when performing operations, and the processor 610 calls and executes the programs and data stored in the memory 620.

[0271] The transceiver 600 is connected to the bus interface and is used to receive and send data under the control of the processor 610; the processor 610 is used to read the program in the memory 620 and execute the following process:

[0272] Receiving first information and second information sent by a terminal; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or,

[0273] Receive third information sent by a terminal, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints:

[0274] Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value;

[0275] Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value;

[0276] The model satisfies a second terminal-side condition corresponding to the terminal.

[0277] Optionally, the first information and the fourth information are associated with each other, or part or all of the first information is the fourth information;

[0278] The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

[0279] In some embodiments, the processor 610 is further configured to read a program in the memory 620 and execute the following process:

[0280] Receive metadata of the model sent by the terminal, where the metadata includes fourth information, and the fourth information is used to indicate a second terminal-side condition corresponding to the at least one model during the training phase.

[0281] In FIG6 , the bus architecture may include any number of interconnected buses and bridges, specifically various circuits linked together by one or more processors represented by processor 610 and memory represented by memory 620. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 600 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium, such as a wireless channel, a wired channel, an optical cable, and the like. The processor 610 is responsible for managing the bus architecture and general processing, and the memory 620 may store data used by the processor 610 when performing operations.

[0282] The processor 610 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

[0283] Referring to FIG. 7 , an embodiment of the present disclosure provides a model management method, which is applied to a terminal. The model may include an artificial intelligence (AI) model, or an AI function, or a machine learning (ML) model, or an ML function. Alternatively, the model may be referred to as an AI / ML model, or an AI / ML function. The terminal may be a user equipment (UE).

[0284] Specifically, the model management method includes the following steps:

[0285] Step 31: Receive first indication information or second indication information sent by the network device; wherein, the first indication information is used to indicate a first network side condition corresponding to the network device; the second indication information is used to indicate identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

[0286] Among them, network equipment includes but is not limited to: base stations, gNBs, TRPs and LMFs.

[0287] It should be noted that the second network side condition corresponding to at least one model indicated by the second indication information during the training phase matches the first network side condition.

[0288] The first network-side condition and the second network-side condition match each other, which may include: the first network-side condition and the second network-side condition are completely identical, partially identical, or similar. Alternatively, the first network-side condition and the second network-side condition are considered to match each other if the performance of a model (such as a reference model) under the first network-side condition and the second network-side condition are identical or similar, and if the performance difference is within a predefined range.

[0289] Exemplarily, the model corresponds to an NW indication ID (i.e., the ID of the second network side condition) in the training stage. If the NW indication ID (i.e., the ID of the first network side condition) in the model inference stage is the same as the NW indication ID corresponding to the model training stage, it is determined that the consistency of the network side condition is met, that is, it is determined that the first network side condition matches the second network side condition.

[0290] Exemplarily, the model corresponds to multiple NW indication IDs during the training phase, i.e., the ID of the second network side condition, including NW IDs corresponding to multiple different network side conditions collected in the training data set. If the current NW indication ID (i.e., the ID of the first network side condition) is at least partially the same as the NW indication ID corresponding to the model training phase, it is determined that the consistency of the network side condition is met, i.e., it is determined that the first network side condition matches the second network side condition.

[0291] Exemplarily, the ID of the first network-side condition is different from the ID of the second network-side condition, but through a predefined mapping relationship, it can be determined that the first network-side condition and the second network-side condition are identical, partially identical, or similar, and then the first network-side condition is determined to match the second network-side condition. For example, if the difference between the IDs of the two network-side conditions is less than a certain threshold value, then the first network-side condition is determined to match the second network-side condition; for another example, if the mapped IDs obtained by the two network condition IDs through a predefined operation mapping (such as performing a modulo operation on a certain value) are the same, then the first network-side condition is determined to match the second network-side condition; for another example, if the IDs of the two network conditions both belong to the same predefined or preconfigured ID group, then the first network-side condition is determined to match the second network-side condition.

[0292] It should be noted that the first network-side conditions described above are those encountered by network devices during the model inference phase. These conditions are distinct from the second network-side conditions encountered during the model training phase, which are assumed during the training phase and are the actual network-side conditions used during the training process. These first and second network-side conditions may include, but are not limited to, the gNB's beam shape and beam pattern.

[0293] Step 32: Determine the target model used by the terminal according to the first indication information or the second indication information.

[0294] In the above embodiment, in the scheme of receiving the first indication information sent by the network device and determining the target model used by the terminal based on the first indication information, the terminal does not need to initiate a model recognition process to the network device, nor does it need to report the models supported by itself to the network device. Instead, based on the first indication information sent by the network device, the first network-side condition corresponding to the network device in the model inference stage is determined, and combined with the second network-side condition corresponding to the multiple models supported by itself in the model training stage, the target model can be selected from the multiple models supported by itself, so that the target model satisfies the first terminal-side condition and matches the second terminal-side condition. Preferably, the terminal can also combine the second terminal-side condition corresponding to the model training stage and the first terminal-side condition corresponding to the terminal in the model inference stage to select the target model from the multiple models supported by itself, so that the target model simultaneously satisfies the first terminal-side condition and matches the second terminal-side condition, as well as the first network-side condition and matches the second network-side condition.

[0295] In a solution that receives a second indication message sent by a network device and determines the target model used by the terminal based on the second indication message, after the terminal reports the models it supports to the network device, the network device selects at least one model that satisfies the first network-side condition and matches the second network-side condition, and instructs the terminal through the second indication message; in this way, the terminal selects a target model from the at least one model indicated by the network device, so that the selected target model satisfies the first network-side condition and matches the second network-side condition. Preferably, the terminal can combine the matching of the terminal-side conditions in the model training phase and the model inference phase, and select a target model from the at least one model indicated by the network device, so that the target model satisfies the first terminal-side condition and matches the second terminal-side condition, as well as the first network-side condition and matches the second network-side condition.

[0296] Through these two solutions, the terminal can not only ensure that the terminal-side conditions used by the model match the model training and model inference phases, but also ensure that the network-side conditions match the model training and model inference phases, thereby effectively managing the terminal-side model. Moreover, by having the terminal manage its own model, the complexity of model management is reduced.

[0297] It should be noted that in the above-mentioned solution for managing terminal-side models through the terminal, the first terminal-side condition corresponding to the model inference phase and the second terminal-side condition corresponding to the model training phase are already known to the terminal and do not need to be obtained from the network. The second network-side condition corresponding to the model training phase is obtained by the terminal during the training data collection phase.

[0298] In some optional embodiments, determining the target model used by the terminal according to the second indication information includes:

[0299] Determining a first network-side condition corresponding to the network device according to the at least one model indicated by the second indication information;

[0300] The target model used by the terminal is determined according to the first network-side condition corresponding to the network device, the first terminal-side condition corresponding to the terminal, the second terminal-side condition and the second network-side condition corresponding to the at least one model in the training phase.

[0301] For example, still taking Table 1 above as an example, assume that the terminal identifies or indicates to the network device that it supports the four models in the table, and assume that the current beam pattern of the network device is pattern 2. Based on this assumption, it can be determined that model#2a or model#2b can meet the matching of the network-side conditions in the model training phase and the model inference phase. However, since the network device cannot determine the terminal-side conditions, it can only randomly indicate one of model#2a and model#2b to the terminal. Assuming that the model ID indicated by the network device to the terminal is model#2a, the terminal can determine the beam pattern 2 corresponding to the network device in the model inference phase. Therefore, the terminal determines that both model#2a and model#2b can ensure the consistency of the network-side conditions. Then, based on the terminal-side conditions corresponding to the terminal in the model inference phase, if the terminal is in a high-speed state at this time, the terminal can select and use model#2b from model#2a and model#2b.

[0302] In the above embodiment, during the process of model management by the terminal, the second indication information sent by the terminal through the network device can indirectly determine the first network side condition corresponding to the network device in the model inference stage. Combined with the second network side condition known in the model training stage, the matching of the network side conditions in the model training stage and the model inference stage can be guaranteed.

[0303] In some optional embodiments, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0304] Item 1: Applicable to the target community;

[0305] That is, it applies to the NW ID of a specific cell. This also means that the determination of whether the first network-side condition indicated by the first indication information matches the second network-side condition corresponding to the model training phase is only applicable within the same target cell and not across cells. For example, the network-side condition corresponding to NW ID = 1 of cell #1 is not necessarily the same as the network-side condition corresponding to NW ID = 1 of cell #2.

[0306] The second item: applies to cells belonging to the same cell group;

[0307] That is, the NW ID applies to a specific cell group. This also means that the judgment of whether the first network-side condition indicated by the first indication information is the same as the second network-side condition corresponding to the model training phase is only applicable within the same cell group and not across cells. For example, the network-side condition corresponding to NW ID = 1 of cell group #1 is not necessarily the same as the network-side condition corresponding to NW ID = 1 of cell group #2.

[0308] Among them, which cells belong to the same cell group may be predefined or indicated by the network device, for example, by indicating the cell group to which the network device belongs through broadcast, multicast or unicast.

[0309] Item 3: Applicable to communities belonging to the same manufacturer;

[0310] For example, the network side condition corresponding to NW ID=1 indicated by the network device belonging to manufacturer #1 is not necessarily the same as the network side condition corresponding to NW ID=1 indicated by the network device belonging to manufacturer #2.

[0311] Item 4: Applicable to cells belonging to the same operator;

[0312] Optionally, cells belonging to the same operator may include cells with the same PLMN number.

[0313] For example, the network side condition corresponding to NW ID=1 indicated by the network equipment belonging to operator #1 is not necessarily the same as the network side condition corresponding to NW ID=1 indicated by the network equipment belonging to operator #2.

[0314] Item 5: Applicable to any community.

[0315] This also means that the judgment of whether the first network side condition indicated by the first indication information is the same as the second network side condition corresponding to the model training phase is applicable to any cell in the world; as long as the NW ID of cell#1 is 1 and the NW ID of cell#2 is 1, it means that their corresponding network side conditions are the same.

[0316] In some optional embodiments, models of different functions correspond to different network-side conditions during the training phase.

[0317] For example, the network conditions indicated by the first indication information of an AI / ML model used for CSI feedback and an AI / ML model used for BM are not necessarily the same. This means that only when the functions or features of the AI / ML models are the same (for example, both are used for CSI feedback) can the matching of the network conditions be determined by the NW ID corresponding to the network conditions.

[0318] In some optional embodiments, the above model management method further includes:

[0319] Sending identification information of the target model used by the terminal to the network device.

[0320] In this embodiment, after the network device indicates the model ID corresponding to at least one model to the terminal, the terminal determines the target model to be selected or used and provides feedback to the network device on the model ID of the target model actually being used. This assists the network device in determining the terminal-side conditions corresponding to the model inference phase and the specific model being used, facilitating subsequent model management for the terminal or other terminals, or enabling the network device to accurately monitor the performance of terminal-side models.

[0321] In an optional embodiment, the second network-side condition corresponding to the model in the training phase is associated with the configuration or process of training data collection and will not be configured or indicated separately; the first network-side condition corresponding to the model inference phase is configured or indicated separately by the network device and is not associated with the configuration or process of training data collection.

[0322] Optionally, the format of the first NW ID corresponding to the first network side condition and the second NW ID corresponding to the second network side condition are the same.

[0323] Referring to Figure 8, an embodiment of the present disclosure provides a model management method applied to network devices, wherein the network devices include but are not limited to: base stations, gNBs, TRPs, and LMF entities.

[0324] The method comprises the following steps:

[0325] Step 41: Send a first indication message or a second indication message to the terminal; wherein the first indication message is used to indicate a first network side condition corresponding to the network device; the second indication message is used to indicate identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

[0326] Among them, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0327] In the above embodiment, by sending the first indication information to the terminal, the terminal determines the first network-side condition corresponding to the network device in the model inference stage according to the first indication information sent by the network device, and the terminal selects the target model from the multiple models supported by itself in combination with the second network-side condition corresponding to the multiple models supported by itself in the model training stage, so that the target model satisfies the first terminal-side condition and matches the second terminal-side condition. The terminal does not need to initiate the model recognition process to the network device, nor does it need to report the models supported by itself to the network device. Preferably, the terminal can also select the target model from the multiple models supported by itself in combination with the second terminal-side condition corresponding to the model training stage and the first terminal-side condition corresponding to the terminal in the model inference stage, so that the target model satisfies the first terminal-side condition and matches the second terminal-side condition, as well as the first network-side condition and matches the second network-side condition, thereby ensuring the performance of the model.

[0328] In the above embodiment, after the terminal reports the models it supports to the network device, the network device selects at least one model that satisfies the first network-side condition and matches the second network-side condition, and indicates it to the terminal through the second indication information; in this way, the terminal selects the target model from the at least one model indicated by the network device, so that the selected target model satisfies the first terminal-side condition and matches the second terminal-side condition. Preferably, the terminal can combine the matching of the terminal-side conditions in the model training phase and the model inference phase, and select the target model from the at least one model indicated by the network device, so that the target model satisfies the first terminal-side condition and matches the second terminal-side condition, as well as the first network-side condition and matches the second network-side condition, to ensure the performance of the model.

[0329] Through these two solutions, the terminal can not only ensure that the terminal-side conditions used by the model match the model training and model inference phases, but also ensure that the network-side conditions match the model training and model inference phases, thereby effectively managing the terminal-side model. Moreover, by having the terminal manage its own model, the complexity of model management is reduced.

[0330] In some optional embodiments, after sending the first indication information or the second indication information to the terminal, the method further includes:

[0331] Receive identification information of a target model sent by the terminal, where the target model is determined by the terminal according to the first indication information or the second indication information.

[0332] In this embodiment, after the network device indicates the model ID corresponding to at least one model to the terminal, the terminal determines the target model to be selected or used and provides feedback to the network device on the model ID of the target model actually being used. This assists the network device in determining the terminal-side conditions corresponding to the model inference phase and the specific model being used, facilitating subsequent model management for the terminal or other terminals, or enabling the network device to accurately monitor the performance of terminal-side models.

[0333] In some optional embodiments, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0334] Applicable to the target community;

[0335] Applicable to cells belonging to the same cell group;

[0336] Applicable to communities belonging to the same manufacturer;

[0337] Applicable to cells belonging to the same operator;

[0338] Applicable to any community.

[0339] It should be pointed out that this embodiment is the same as the embodiment on the terminal side and can achieve the same technical effect. For details, please refer to the explanation of the embodiment on the terminal side.

[0340] In some optional embodiments, models of different functions correspond to different network-side conditions during the training phase.

[0341] For example, the network conditions indicated by the first indication information of an AI / ML model used for CSI feedback and an AI / ML model used for BM are not necessarily the same. This means that only when the functions or features of the AI / ML models are the same (for example, both are used for CSI feedback) can the matching of the network conditions be determined by the NW ID corresponding to the network conditions.

[0342] It should be noted that the above-described embodiments of the present disclosure are primarily applicable to 5G NR systems, including network equipment and terminal devices. However, the present invention can also be applied to other systems where terminal-side AI / ML models are required. For example, the system includes multiple UEs, including UE1 and UE2, requesting wireless network connection services; the gNB provides wireless services for them. The gNB and UE1 and UE2 exchange and transmit data via wireless communication. For example, the gNB provides AI / ML-related services to UE1 and UE2, including transmitting various indication information to UE1 and UE2.

[0343] The following introduces the model management method of the terminal-side model managed by the terminal, which can mainly include the following two solutions.

[0344] Option 1:

[0345] In step 1, during the data collection phase, a network device (e.g., NW#1) indicates an ID (e.g., NW ID#1) to UE#1, and associates the configuration / process of data collection. The data collected by UE#1 is also associated with NW ID#1. NW ID#1 represents the second network-side condition.

[0346] In step 2, during the model training phase, UE#1 performs model training. The ID of the trained terminal-side model is recorded as model ID#1. The model may be deployed to multiple UEs, not just UE#1, such as UE#2. The model training data may also come from several different UEs or NWs.

[0347] Step 3: In the model inference preparation phase, the network device (eg, NW#2) sends a first indication message to UE#2, indicating a first network-side condition corresponding to the network device in the model inference phase.

[0348] Step 4: In the model inference phase, UE#2 can determine the consistency of the network side conditions of a model in the current network based on whether the first network side conditions match the second network side conditions corresponding to the model in the training phase.

[0349] Through the above-mentioned solution 1, the terminal can not only ensure the consistency of the network-side conditions of the terminal-side model, but also ensure the consistency of the terminal-side conditions, thereby enabling the most appropriate management of the terminal-side model and ensuring the performance of the communication system. The consistency of the network-side conditions can be understood as the second network-side condition in the model training phase matching the first network-side condition of the network device in the model inference phase, and the consistency of the terminal-side conditions can be understood as the second terminal-side condition corresponding to the model training phase matching the first terminal-side condition corresponding to the terminal in the model inference phase.

[0350] Option 2:

[0351] In step 1, during the data collection phase, the network device (possibly NW#1) indicates an ID (e.g., NW ID#1) to UE#1, and associates it with the data collection configuration / process. The data collected by UE#1 is also associated with NW ID#1. NW ID#1 represents the second network-side condition.

[0352] In step 2, during the model training phase, UE#1 performs model training. The ID of the trained terminal-side model is recorded as model ID#1. The model may be deployed to multiple UEs, not just UE#1, such as UE#2. The model training data may also come from several different UEs or NWs.

[0353] Step 3: Model inference preparation phase: UE#2 initiates model identification to the network device (e.g., NW#2) and reports to NW#2 the model ID(s) of one or more models supported by UE#2.

[0354] Model ID#1 may be the NW ID#1 indicated by NW#1 in step 1, or there may be a predefined association between model ID#1 and NW ID#1, or NW ID#1 may be indicated to NW#2 through model meta information (meta info).

[0355] Step 4. In the model inference phase, NW#2 sends a second indication message to UE#2, which is used to indicate the identification information of at least one model, where the at least one model is a subset of the models supported by the terminal, and the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0356] UE#2 can indirectly determine the first network side condition corresponding to NW#2 through the model ID indicated by NW#2.

[0357] Through the above-mentioned second solution, the terminal can not only ensure the consistency of the network-side conditions of the terminal-side model, but also ensure the consistency of the terminal-side conditions, thereby enabling the most appropriate management of the terminal-side model and ensuring the performance of the communication system. The consistency of the network-side conditions can be understood as the second network-side condition in the model training phase matching the first network-side condition of the network device in the model inference phase, and the consistency of the terminal-side conditions can be understood as the second terminal-side condition corresponding to the model training phase matching the first terminal-side condition corresponding to the terminal in the model inference phase.

[0358] 9 , an embodiment of the present disclosure provides a model management device 900, which is applied to a terminal and includes:

[0359] The second receiving module 901 is configured to receive first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, where the at least one model is a subset of models supported by the terminal.

[0360] The determination module 902 is configured to determine the target model used by the terminal according to the first indication information or the second indication information.

[0361] Optionally, the determining module 902 is specifically configured to:

[0362] Determining a first network-side condition corresponding to the network device according to the at least one model indicated by the second indication information;

[0363] The target model used by the terminal is determined according to the first network-side condition corresponding to the network device, the first terminal-side condition corresponding to the terminal, the second terminal-side condition and the second network-side condition corresponding to the at least one model in the training phase.

[0364] Optionally, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0365] Optionally, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0366] Applicable to the target community;

[0367] Applicable to cells belonging to the same cell group;

[0368] Applicable to communities belonging to the same manufacturer;

[0369] Applicable to cells belonging to the same operator;

[0370] Applicable to any community.

[0371] Optionally, models with different functions correspond to different network-side conditions during the training phase.

[0372] Optionally, the model management device 900 further includes:

[0373] A fourth sending module is configured to send identification information of the target model used by the terminal to the network device.

[0374] It should be noted here that the above-mentioned device 900 provided in the embodiment of the present disclosure can implement all the method steps implemented by the method embodiment applied to the terminal side as shown in Figure 7, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0375] 10 , an embodiment of the present disclosure provides a model management apparatus 1000, which is applied to a network device and includes:

[0376] The second sending module 1001 is used to send the first indication information or the second indication information to the terminal; wherein the first indication information is used to indicate the first network side condition corresponding to the network device; the second indication information is used to indicate the identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

[0377] Optionally, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0378] Optionally, the model management device 1000 further includes:

[0379] The fourth receiving module is configured to receive identification information of a target model sent by the terminal, where the target model is determined by the terminal according to the first indication information or the second indication information.

[0380] Optionally, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0381] Applicable to the target community;

[0382] Applicable to cells belonging to the same cell group;

[0383] Applicable to communities belonging to the same manufacturer;

[0384] Applicable to cells belonging to the same operator;

[0385] Applicable to any community.

[0386] Optionally, models with different functions correspond to different network-side conditions during the training phase.

[0387] It should be noted here that the above-mentioned device 1000 provided in the embodiment of the present disclosure can implement all the method steps implemented by the method embodiment applied to the network device side as shown in Figure 8, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0388] Referring to Figure 11 , an embodiment of the present disclosure provides a terminal, comprising: a processor 1110; and a memory 1120 connected to the processor 1110 via a bus interface. The memory 1120 is configured to store programs and data used by the processor 1110 when performing operations, and the processor 1110 calls and executes the programs and data stored in the memory 1120. A transceiver 1100 is connected to the bus interface and is configured to receive and transmit data under the control of the processor 1110. The processor 1110 is configured to read the program in the memory 1120 and perform the following processes:

[0389] Receiving first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, where the at least one model is a subset of models supported by the terminal;

[0390] Determine a target model used by the terminal according to the first indication information or the second indication information.

[0391] Optionally, the processor 1110 is specifically configured to read a program in the memory 1120 and execute the following process:

[0392] Determining a first network-side condition corresponding to the network device according to the at least one model indicated by the second indication information;

[0393] The target model used by the terminal is determined according to the first network-side condition corresponding to the network device, the first terminal-side condition corresponding to the terminal, the second terminal-side condition and the second network-side condition corresponding to the at least one model in the training phase.

[0394] Optionally, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0395] Optionally, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0396] Applicable to the target community;

[0397] Applicable to cells belonging to the same cell group;

[0398] Applicable to communities belonging to the same manufacturer;

[0399] Applicable to cells belonging to the same operator;

[0400] Applicable to any community.

[0401] Optionally, models with different functions correspond to different network-side conditions during the training phase.

[0402] Optionally, the processor 1110 is further configured to read a program in the memory 1120 and execute the following process:

[0403] Sending identification information of the target model used by the terminal to the network device.

[0404] In FIG11 , the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 1110 and memory represented by memory 1120. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 1100 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium, such as a wireless channel, a wired channel, an optical cable, and the like. For different user devices, the user interface 1130 may also be an interface capable of connecting external or internal devices as required, and the connected devices include but are not limited to a keypad, a display, a speaker, a microphone, a joystick, and the like.

[0405] The processor 1110 is responsible for managing the bus architecture and general processing, and the memory 1120 can store data used by the processor 510 when performing operations.

[0406] In some embodiments, the processor 1110 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

[0407] The processor calls the computer program stored in the memory to execute any of the methods provided by the embodiments of the present disclosure according to the obtained executable instructions. The processor and the memory can also be arranged physically separately.

[0408] Referring to Figure 12, an embodiment of the present disclosure provides a network device, including: a processor 1210; and a memory 1220 connected to the processor 1210 through a bus interface, the memory 620 is used to store programs and data used by the processor 1210 when performing operations, and the processor 1210 calls and executes the programs and data stored in the memory 1220.

[0409] The transceiver 1200 is connected to the bus interface and is used to receive and send data under the control of the processor 1210; the processor 1210 is used to read the program in the memory 1220 and perform the following processes:

[0410] Sending first indication information or second indication information to the terminal; wherein, the first indication information is used to indicate a first network side condition corresponding to the network device; the second indication information is used to indicate identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

[0411] Optionally, the second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

[0412] Optionally, the processor 1210 is configured to read a program in the memory 1220 and execute the following process:

[0413] Receive identification information of a target model sent by the terminal, where the target model is determined by the terminal according to the first indication information or the second indication information.

[0414] Optionally, the first network-side condition corresponding to the network device is applicable to at least one of the following application scopes:

[0415] Applicable to the target community;

[0416] Applicable to cells belonging to the same cell group;

[0417] Applicable to communities belonging to the same manufacturer;

[0418] Applicable to cells belonging to the same operator;

[0419] Applicable to any community.

[0420] Optionally, models with different functions correspond to different network-side conditions during the training phase.

[0421] In FIG12 , the bus architecture may include any number of interconnected buses and bridges, specifically various circuits linked together by one or more processors represented by processor 1210 and memory represented by memory 1220. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 1200 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, such as a wireless channel, a wired channel, an optical cable, and the like. The processor 1210 is responsible for managing the bus architecture and general processing, and the memory 1220 may store data used by the processor 1210 when performing operations.

[0422] The processor 1210 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

[0423] The present disclosure also provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute the above-mentioned model management method on the terminal side or the network device side.

[0424] The embodiment of the present disclosure also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the various processes of the above-mentioned method embodiment are implemented and the same technical effect can be achieved. To avoid repetition, they are not described here.

[0425] The processor-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as compact discs (CD), digital video discs (DVD), Blu-ray discs (BD), high-definition versatile discs (HVD), etc.), and semiconductor memory (such as ROM, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), non-volatile memory (NAND (Non-volatile Memory Device) FLASH), solid-state drives (SSD)), etc.

[0426] It should be noted that the division of units in the embodiments of the present disclosure is schematic and is merely a logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of the present disclosure may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0427] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of the present disclosure is essentially or the part that contributes to the relevant technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0428] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage and optical storage, etc.) containing computer-usable program code.

[0429] These processor-executable instructions may also be stored in a processor-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the processor-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0430] These processor-executable instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0431] In addition, it should be noted that, in the apparatus and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. Moreover, the steps of performing the above-mentioned series of processing can naturally be performed in chronological order according to the order of description, but it is not necessary to perform them in chronological order, and some steps can be performed in parallel or independently of each other. For those of ordinary skill in the art, it will be understood that all or any steps or components of the method and apparatus of the present invention can be implemented in any computing device (including processors, storage media, etc.) or a network of computing devices in hardware, firmware, software or a combination thereof, which can be achieved by those of ordinary skill in the art using their basic programming skills after reading the description of the present invention.

[0432] It should be noted that it should be understood that the division of the above modules is merely a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. Moreover, these modules can all be implemented in the form of software called by a processing element; or they can all be implemented in the form of hardware; or some modules can be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. For example, a module can be a separately established processing element, or it can be integrated into a chip of the above-mentioned device. In addition, it can also be stored in the memory of the above-mentioned device in the form of program code, and called by a processing element of the above-mentioned device to perform the functions of the above-mentioned module. The implementation of other modules is similar. In addition, these modules can all or partly be integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. During implementation, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0433] For example, each module, unit, sub-unit or sub-module can be one or more integrated circuits configured to implement the above method, such as one or more application-specific integrated circuits (ASICs), one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented by scheduling program code through a processing element, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0434] The terms "first," "second," and the like in the specification and claims of the present disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of the present disclosure described herein may be implemented in a sequence other than that illustrated or described herein. In addition, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units need not be limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or devices. In addition, the use of "and / or" in the specification and claims to indicate at least one of the connected objects, for example, A and / or B and / or C, means that seven situations are included: A alone, B alone, C alone, both A and B present, both B and C present, both A and C present, and all A, B, and C present. Similarly, the use of "at least one of A and B" in the specification and claims should be understood to mean "A alone, B alone, or both A and B present."

[0435] Obviously, those skilled in the art may make various changes and modifications to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure fall within the scope of the claims of the present disclosure and their equivalents, the present disclosure is intended to include these modifications and variations.

Claims

1. A model management method, applied to a terminal, comprising: Sending first information and second information to a network device; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, Sending third information to the network device, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints: Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value; Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value; The model satisfies a first terminal-side condition of the terminal.

2. The model management method according to claim 1, wherein: There is an association relationship between the first information and the fourth information, or part or all of the first information is the fourth information; The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

3. The model management method according to claim 1, wherein: The method further comprises: The meta information of the model is sent to the network device, where the meta information includes fourth information, and the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training phase.

4. A model management method, applied to a network device, comprising: Receiving first information and second information sent by a terminal; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, Receive third information sent by a terminal, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints: Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value; Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value; The model satisfies a second terminal-side condition corresponding to the terminal.

5. The model management method according to claim 4, wherein: There is an association relationship between the first information and the fourth information, or part or all of the first information is the fourth information; The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

6. The model management method according to claim 4, wherein: The method further comprises: Receive metadata of the model sent by the terminal, where the metadata includes fourth information, and the fourth information is used to indicate a second terminal-side condition corresponding to the at least one model during a training phase.

7. A model management method, applied to a terminal, comprising: Receiving first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, where the at least one model is a subset of models supported by the terminal; Determine a target model used by the terminal according to the first indication information or the second indication information.

8. The model management method according to claim 7, wherein: Determining, according to the second indication information, a target model used by the terminal, including: Determining a first network-side condition corresponding to the network device according to the at least one model indicated by the second indication information; The target model used by the terminal is determined according to the first network-side condition corresponding to the network device, the first terminal-side condition corresponding to the terminal, the second terminal-side condition and the second network-side condition corresponding to the at least one model in the training phase.

9. The model management method according to claim 7, wherein: The second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

10. The model management method according to claim 7, wherein: The first network-side condition corresponding to the network device applies to at least one of the following application scopes: Applicable to the target community; Applicable to cells belonging to the same cell group; Applicable to communities belonging to the same manufacturer; Applicable to cells belonging to the same operator; Applicable to any community.

11. The model management method according to claim 7, wherein: Models with different functions correspond to different network-side conditions during the training phase.

12. The model management method according to claim 7, wherein: The method further comprises: Sending identification information of the target model used by the terminal to the network device.

13. A model management method, applied to a network device, comprising: Sending first indication information or second indication information to the terminal; wherein, the first indication information is used to indicate a first network side condition corresponding to the network device; the second indication information is used to indicate identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

14. The model management method according to claim 13, wherein: The second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

15. The model management method according to claim 13, wherein: After sending the first indication information or the second indication information to the terminal, the method further includes: Receive identification information of a target model sent by the terminal, where the target model is determined by the terminal according to the first indication information or the second indication information.

16. The model management method according to claim 13, wherein: The first network-side condition corresponding to the network device applies to at least one of the following application scopes: Applicable to the target community; Applicable to cells belonging to the same cell group; Applicable to communities belonging to the same manufacturer; Applicable to cells belonging to the same operator; Applicable to any community.

17. The model management method according to claim 13, wherein: Models with different functions correspond to different network-side conditions during the training phase.

18. A terminal comprising: A transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process: Sending first information and second information to a network device; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, Sending third information to the network device, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints: Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value; Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value; The model satisfies a first terminal-side condition of the terminal. The terminal according to claim 18 , wherein: There is an association relationship between the first information and the fourth information, or part or all of the first information is the fourth information; The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

20. The terminal according to claim 18, wherein The processor is further configured to read a program in the memory and execute the following processes: The meta information of the model is sent to the network device, where the meta information includes fourth information, and the fourth information is used to indicate the second terminal side condition corresponding to the at least one model in the training phase.

21. A network device comprising: A transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process: Receiving first information and second information sent by a terminal; wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, Receive third information sent by a terminal, where the third information is used to indicate at least one model supported by the terminal; wherein the model satisfies one or more of the following constraints: Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value; Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value; The model satisfies a second terminal-side condition corresponding to the terminal.

22. The network device according to claim 21, wherein: There is an association relationship between the first information and the fourth information, or part or all of the first information is the fourth information; The fourth information is used to indicate the second terminal side condition corresponding to the at least one model during the training phase.

23. The network device according to claim 21, wherein: The processor is further configured to read a program in the memory and execute the following processes: Receive metadata of the model sent by the terminal, where the metadata includes fourth information, and the fourth information is used to indicate a second terminal-side condition corresponding to the at least one model during a training phase.

24. A terminal comprising: A transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process: Receiving first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, where the at least one model is a subset of models supported by the terminal; Determine a target model used by the terminal according to the first indication information or the second indication information.

25. The terminal according to claim 24, wherein: The processor is specifically configured to read a program in a memory and execute the following process: Determining a first network-side condition corresponding to the network device according to the at least one model indicated by the second indication information; The target model used by the terminal is determined according to the first network-side condition corresponding to the network device, the first terminal-side condition corresponding to the terminal, the second terminal-side condition and the second network-side condition corresponding to the at least one model in the training phase.

26. The terminal according to claim 24, wherein: The second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

27. The terminal according to claim 24, wherein: The first network-side condition corresponding to the network device applies to at least one of the following application scopes: Applicable to the target community; Applicable to cells belonging to the same cell group; Applicable to communities belonging to the same manufacturer; Applicable to cells belonging to the same operator; Applicable to any community.

28. The terminal according to claim 24, wherein: Models with different functions correspond to different network-side conditions during the training phase.

29. The terminal according to claim 24, wherein: The processor is further configured to read a program in the memory and execute the following process: Sending identification information of the target model used by the terminal to the network device.

30. A network device comprising: A transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor is configured to read the program in the memory and execute the following process: Sending first indication information or second indication information to the terminal; wherein, the first indication information is used to indicate a first network side condition corresponding to the network device; the second indication information is used to indicate identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

31. The network device according to claim 30, wherein: The second network side condition corresponding to the at least one model in the training phase matches the first network side condition.

32. The network device according to claim 30, wherein: The processor is further configured to read a program in the memory and execute the following process: Receive identification information of a target model sent by the terminal, where the target model is determined by the terminal according to the first indication information or the second indication information.

33. The network device according to claim 30, wherein: The first network-side condition corresponding to the network device applies to at least one of the following application scopes: Applicable to the target community; Applicable to cells belonging to the same cell group; Applicable to communities belonging to the same manufacturer; Applicable to cells belonging to the same operator; Applicable to any community.

34. The network device according to claim 30, wherein: Models with different functions correspond to different network-side conditions during the training phase.

35. A model management device, applied to a terminal, comprising: A first sending module is configured to send first information and second information to a network device, wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, to send third information to the network device, wherein the third information is used to indicate at least one model supported by the terminal, wherein the model satisfies one or more of the following constraints: Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value; Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value; The model satisfies a first terminal-side condition of the terminal.

36. A model management device, applied to a network device, comprising: A first receiving module is configured to receive first information and second information sent by a terminal, wherein the first information is used to indicate identification information of at least one model supported by the terminal, and the second information is used to indicate a first terminal-side condition corresponding to the terminal; or, to receive third information sent by the terminal, wherein the third information is used to indicate at least one model supported by the terminal, wherein the model satisfies one or more of the following constraints: Under any terminal-side condition, the target performance of the model is greater than or equal to a first threshold value; Under any terminal-side condition, the target performance degradation degree of the model is less than or equal to a second threshold value; The model satisfies a second terminal-side condition corresponding to the terminal.

37. A model management device, applied to a terminal, comprising: a second receiving module, configured to receive first indication information or second indication information sent by a network device; wherein the first indication information is used to indicate a first network-side condition corresponding to the network device; and the second indication information is used to indicate identification information of at least one model, wherein the at least one model is a subset of models supported by the terminal; A determination module is used to determine the target model used by the terminal according to the first indication information or the second indication information.

38. A model management device, applied to a network device, comprising: The second sending module is used to send the first indication information or the second indication information to the terminal; wherein the first indication information is used to indicate the first network side condition corresponding to the network device; the second indication information is used to indicate the identification information of at least one model, and the at least one model is a subset of the models supported by the terminal.

39. A processor-readable storage medium storing a computer program, wherein the computer program is used to cause the processor to execute the steps of the model management method according to any one of claims 1 to 3, or the steps of the model management method according to any one of claims 4 to 6, the steps of the model management method according to any one of claims 7 to 12, or the steps of the model management method according to any one of claims 13 to 17.

40. A computer program product, comprising computer instructions, which, when executed by a processor, implement the steps of the model management method according to any one of claims 1 to 3, or the steps of the model management method according to any one of claims 4 to 6, the steps of the model management method according to any one of claims 7 to 12, or the steps of the model management method according to any one of claims 13 to 17.

Images