Method, apparatus and computer program

Abstract

A user equipment comprises: means for receiving, from a base station, information about a duration for machine learning model identification; and means for performing an identification of one or more user equipment-side machine learning models from the plurality of user equipment-side machine learning models for a duration.

Description

Technical Field

[0001] Various exemplary embodiments of this disclosure relate to methods, apparatus, systems, and computer programs, and more specifically, non-exclusively, to the selection of machine learning models. Background Technology

[0002] A communication network can be viewed as a facility that enables communication between two or more communication devices, or as a facility that provides communication devices with access to a data network. Mobile or wireless communication networks are an example of communication networks. Communication devices may be served by application servers.

[0003] Such communication networks operate according to standards provided by organizations such as 3GPP (3rd Generation Partnership Project) or ETSI (European Telecommunications Standards Institute). An example of such a standard is the so-called 5G (5th generation) standard provided by 3GPP. Summary of the Invention

[0004] Some exemplary embodiments of this disclosure will be described with respect to certain aspects. These aspects are not intended to indicate key or essential features of embodiments of this disclosure, nor are they intended to be used to limit its scope. Other features, aspects, and elements will be apparent to those skilled in the art in light of this disclosure.

[0005] According to one aspect, an apparatus is provided, comprising: at least one processor and at least one memory storing instructions, the instructions, when executed by the at least one processor, causing the apparatus to at least: receive from a base station information about a duration for identifying a machine learning model; and perform identification of one or more user equipment-side machine learning models from a plurality of user equipment-side machine learning models during the duration.

[0006] One or more network-side conditions are stable over a period of time.

[0007] Receiving may include receiving from a base station information about one or more of the following: one or more cells associated with the duration of the machine learning model identification; one or more transmit / receive points associated with the duration of the machine learning model identification; one or more physical cell identifiers associated with the duration of the machine learning model identification; and / or the area associated with the duration of the machine learning model identification.

[0008] Receiving may include receiving information from the base station about the duration of the machine learning model identifier, either in a downlink control indicator or in a media access control element.

[0009] The execution may include: evaluating multiple user-device-side machine learning models among multiple user-device-side machine learning models for one or more network-side additional conditions during the duration used for machine learning model identification.

[0010] At least one processor can be configured to enable the device to receive information from a base station to assist in the identification of one or more user equipment-side machine learning models.

[0011] At least one processor can be configured to enable the device to receive one or more reference signals from a base station to assist in the identification of one or more user equipment-side machine learning models.

[0012] At least one processor may be configured to cause the device to report a first indicator to the base station, the first indicator indicating whether, during the duration of the machine learning model identification, one or more user equipment-side machine learning models suitable for one or more network-side additional conditions exist.

[0013] At least one processor can be configured to enable the device to determine whether one or more user-equipment-side machine learning models among the identified user-equipment-side machine learning models are associated with the new identifier.

[0014] At least one processor may be configured to cause the device to report a second indicator to the base station, the second indicator indicating the identification of one or more user equipment-side machine learning models that are suitable for one or more network-side additional conditions during the duration of the machine learning model identification.

[0015] At least one processor can be configured to cause the device to report additional information related to the second indicator.

[0016] Additional information includes at least one of the following: one or more durations associated with the second indicator; one or more cells associated with the second indicator; one or more transmit / receive points associated with the second indicator; one or more physical cell identifiers associated with the second indicator; and / or the area associated with the second indicator.

[0017] According to another aspect, a user equipment is provided, comprising: a component for receiving information from a base station regarding the duration of an identification for a machine learning model; and a component for performing identification of one or more user equipment-side machine learning models from a plurality of user equipment-side machine learning models during the duration.

[0018] According to one aspect, a method is provided, comprising: receiving from a base station information about the duration of an identification for a machine learning model; and performing identification of one or more user equipment-side machine learning models from a plurality of user equipment-side machine learning models during the duration.

[0019] One or more network-side conditions are stable over a period of time.

[0020] Receiving may include receiving from a base station information about one or more of the following: one or more cells associated with the duration of the machine learning model identification; one or more transmit / receive points associated with the duration of the machine learning model identification; one or more physical cell identifiers associated with the duration of the machine learning model identification; and / or the area associated with the duration of the machine learning model identification.

[0021] Receiving may include receiving information from the base station about the duration of the machine learning model identifier, either in a downlink control indicator or in a media access control element.

[0022] The execution may include: evaluating multiple user-device-side machine learning models among multiple user-device-side machine learning models for one or more network-side additional conditions during the duration used for machine learning model identification.

[0023] The method may include receiving information from a base station to assist in the identification of one or more user equipment-side machine learning models.

[0024] The method may include receiving one or more reference signals from a base station to assist in the identification of one or more user equipment-side machine learning models.

[0025] The method may include: reporting a first indicator to the base station, the first indicator indicating whether, during the duration of the machine learning model identification, one or more user equipment-side machine learning models suitable for one or more network-side additional conditions exist.

[0026] The method may include: determining whether one or more user device-side machine learning models among the identified user device-side machine learning models are associated with the new identifier.

[0027] The method may include: reporting a second indicator to the base station, the second indicator indicating: the identification of one or more user equipment-side machine learning models that are suitable for one or more network-side additional conditions during the duration of the machine learning model identification.

[0028] The method may include reporting additional information related to the second indicator.

[0029] Additional information includes at least one of the following: one or more durations associated with the second indicator; one or more cells associated with the second indicator; one or more transmit / receive points associated with the second indicator; one or more physical cell identifiers associated with the second indicator; and / or the area associated with the second indicator.

[0030] According to another aspect, a base station is provided, comprising: at least one processor and at least one memory storing instructions, which, when executed by the at least one processor, cause the means to at least: send information to a user equipment regarding the duration of identification for a machine learning model; and receive from the user equipment one or more reports regarding identification of one or more user equipment-side machine learning models from a plurality of user equipment-side machine learning models.

[0031] At least one processor can be configured to keep one or more network-side additional conditions stable over a period of time.

[0032] At least one processor may be configured to cause the device to send information about one or more of the following: one or more cells associated with the duration of the machine learning model identification; one or more transmit / receive points associated with the duration of the machine learning model identification; one or more physical cell identifiers associated with the duration of the machine learning model identification; and / or the area associated with the duration of the machine learning model identification.

[0033] At least one processor can be configured to cause the device to send information about the duration of the machine learning model identifier in a downlink control indicator or in a media access control element.

[0034] At least one processor can be configured to cause the device to send information to a user equipment to assist in the identification of one or more user equipment-side machine learning models.

[0035] At least one processor can be configured to cause the device to send one or more reference signals to assist in the identification of one or more user equipment-side machine learning models.

[0036] At least one processor may be configured to cause the device to receive a first indicator from the user equipment, the first indicator indicating whether, during the duration of the machine learning model identification, one or more user equipment-side machine learning models suitable for one or more network-side additional conditions exist.

[0037] At least one processor may be configured to cause the device to receive a second indicator from the user equipment, the second indicator indicating: the identification of one or more user equipment-side machine learning models that are suitable for one or more network-side additional conditions during the duration of the machine learning model identification.

[0038] At least one processor may be configured to enable the device to receive additional information related to the second indicator from the user equipment.

[0039] Additional information may include at least one of the following: one or more durations associated with the second indicator; one or more cells associated with the second indicator; one or more transmit / receive points associated with the second indicator; one or more physical cell identifiers associated with the second indicator; and / or the area associated with the second indicator.

[0040] According to another aspect, a base station is provided, comprising: components for sending information to a user equipment regarding the duration of an identification for a machine learning model; and components for receiving from the user equipment one or more reports regarding identifications of one or more user equipment-side machine learning models from a plurality of user equipment-side machine learning models.

[0041] According to another aspect, a method is provided, comprising: sending information to a user device about the duration of an identifier used for a machine learning model; and receiving from the user device one or more reports about identifiers from one or more user device-side machine learning models among a plurality of user device-side machine learning models.

[0042] This method may include: keeping one or more network-side additional conditions stable during the time period.

[0043] The method may include sending information about one or more of the following: one or more cells associated with the duration of the machine learning model identification; one or more transmit / receive points associated with the duration of the machine learning model identification; one or more physical cell identifiers associated with the duration of the machine learning model identification; and / or the area associated with the duration of the machine learning model identification.

[0044] The method may include sending information about the duration of the machine learning model identifier in the downlink control indicator or in the media access control control element.

[0045] The method may include sending information to the user device to assist in the identification of one or more user device-side machine learning models.

[0046] The method may include sending one or more reference signals to assist in the identification of one or more user device-side machine learning models.

[0047] The method may include receiving a first indicator from a user equipment, the first indicator indicating whether, during the duration of the machine learning model identification, one or more user equipment-side machine learning models suitable for one or more network-side additional conditions exist.

[0048] The method may include receiving a second indicator from a user equipment, the second indicator indicating: an identifier of one or more user equipment-side machine learning models that are suitable for one or more network-side additional conditions during a duration for machine learning model identification.

[0049] The method may include receiving additional information related to the second indicator from the user equipment.

[0050] Additional information may include at least one of the following: one or more durations associated with the second indicator; one or more cells associated with the second indicator; one or more transmit / receive points associated with the second indicator; one or more physical cell identifiers associated with the second indicator; and / or the area associated with the second indicator.

[0051] According to one aspect, a non-transitory computer-readable medium is provided, comprising program instructions that, when executed by a device, cause the device to perform at least the method according to any one of the foregoing aspects.

[0052] Many different embodiments have been described above. It should be understood that other embodiments can be provided by combining any two or more of the above embodiments. Attached Figure Description

[0053] Some exemplary embodiments will now be described by way of non-limiting and illustrative example, with reference to the accompanying drawings, in which:

[0054] Figure 1 A brief overview of some embodiments is shown;

[0055] Figure 2a and Figure 2b Example processes of some embodiments are shown;

[0056] Figure 3 This illustrates a representation of a 5th generation communication system;

[0057] Figure 4 The illustration shows a method for using some example embodiments. Figure 3 The representation of a communication system device;

[0058] Figure 5A representation of an apparatus according to some example embodiments is shown;

[0059] Figure 6 A first method of some embodiments is shown;

[0060] Figure 7 A second method is shown in some embodiments;

[0061] Figure 8 A third method is shown in some embodiments; and

[0062] Figure 9 A fourth method is shown in some embodiments. Detailed Implementation

[0063] Some embodiments involve the use of artificial intelligence (AI) / machine learning (ML). AI / ML-based models can be used with respect to the air interface (i.e., the air interface between the user equipment and the access node). The use of these models can improve the performance of air interface functions. Some embodiments will be described by way of example with respect to a 5G system, wherein the access node is provided by a gNB. However, other embodiments may be provided in different systems.

[0064] AI / ML models are data-driven algorithms that use AI / ML techniques to generate output sets based on input sets.

[0065] User equipment can have multiple machine learning models. These can be UE-side models or two-side models.

[0066] A two-sided (AI / ML) model is a pair of AI / ML models on which inference is jointly executed. Joint inference includes AI / ML inference whose inference is jointly executed across the UE and the network, i.e., the first part of the inference is executed by the UE first, and then the remainder is executed by the gNB, or vice versa.

[0067] The UE-side (AI / ML) model is an AI / ML model in which inference is performed entirely at the UE.

[0068] Some embodiments involve model identification, wherein one or more models among the models on the UE are selected or identified.

[0069] Model identification is the process / method for identifying AI / ML models. This allows for mutual understanding between the network and the UE. Information about the AI / ML model can be shared during model identification.

[0070] Model selection is the process of choosing an AI / ML model from among multiple models that share the same AI / ML enabling features for activation.

[0071] The UE-side model can be associated with a set of one or more additional conditions.

[0072] Additional conditions may include one or more of the following: Training dataset information; Site-related information (e.g., scene, location / TRP (transmitter / receiver point) / area information, beam direction, codebook information, etc.); Time-related / timestamp information; gNB provides relevant information; UE implements relevant information; Statistical information (e.g., time delay spread, angle spread, LOS / NLOS (line-of-sight / non-line-of-sight) data and / or similar information); and / or Speed ​​and speed range information.

[0073] It should be understood that these conditions are merely examples, and in other embodiments, one or more other conditions may be used alternatively or additionally.

[0074] For UE-side or two-side models, model training can be mapped to a set of hypotheses considered for one or more additional conditions (such as those previously described). The UE-side or two-side model can be associated with one or more additional conditions.

[0075] One or more additional conditions can be considered as NW-side additional conditions (assumptions of the NW-side model in a two-sided model). One or more additional conditions can be considered as UE-side additional conditions (UE-side assumptions for the UE-side model or the UE-side model in a two-sided model). One or more additional conditions can be considered as common additional conditions (assumptions applicable to both the UE-side model and the NW-side model).

[0076] However, describing these additional conditions for online model identification is not straightforward. This can be problematic in scenarios where models can be identified via over-the-air signaling. Some embodiments can address or mitigate this issue.

[0077] Model identification is the process / method of identifying AI / ML models based on a shared understanding between NW and UE.

[0078] The model can be identified via over-the-air signaling. In some 5G standards, this is referred to as Type B.

[0079] Model identification can be initiated by the UE, and the NW assists with the remaining steps of model identification (if any). During model identification, the model can be assigned a model ID. In some 5G standards, this is referred to as Type B1.

[0080] Model identification can be initiated by the NW, and the UE responds to the remaining steps of model identification, if any, if applicable. During model identification, the model can be assigned a model ID. In some 5G standards, this is referred to as type B2.

[0081] refer to Figure 1 The diagram illustrates a schematic overview of some embodiments. UE 300 has three ML models, Model 1, Model 2, and Model 3. These models are stored in one or more memories on the UE. The UE may have two or more models. In some embodiments, the UE will have more than three models. UE 300 communicates with an access node (e.g., gNB100) via an air interface.

[0082] In some embodiments, an access node (e.g., a gNB) can query UE capabilities. Based on information about UE capabilities, the gNB can send queries to the user data store in the UE and / or the network. These capabilities may relate to ML / AI support. The user data store in the network may be provided by the AMF (Access and Mobility Management Function) and / or one or more other network functions.

[0083] After receiving information about the UE's capabilities, the gNB can enable the AMF to update its information about the UE to include information about the UE's capabilities.

[0084] UEs that support ML features can report one or more associated conditions to the gNB. These conditions can be provided in information about the UE's capabilities. (For example, BM Instance 1—the case of ML-based beam prediction in the spatial domain)

[0085] The UE may indicate its support for online model identification in the information about its UE capabilities. The UE may also indicate one or more conditions for online model identification in the information about its UE capabilities. These conditions may include a preferred minimum duration for which the UE wishes to be configured to support model identification. These conditions may include necessary assistance with DL RS (Downlink Reference Signal) transmission, associated functions, or conditions for model identification.

[0086] The UE can receive information about a model identification time period during which the UE identifies the UE-side model or the UE portion of the dual-side model that supports the required ML features. This defines the duration for model identification.

[0087] This model identifies time periods as either duration or length.

[0088] Alternatively, a signal can be received from the gNB to start or initiate a model identification time period, and another signal can be received from the gNB to terminate or end the model identification time period.

[0089] The UE may assume that one or more NW additional conditions remain stable during the duration used for model identification. During this period of model identification, the UE may select one or more UE models that are suitable or applicable to the NW additional conditions observed on the UE side.

[0090] The UE can be configured with information about the region applicable to the duration used for model identification. The information about the region may include information about one or more of the following: one or more cells; one or more TRPs (Transmit / Receive Points); one or more PCIs (Physical Cell Identifiers); and / or any other suitable information defining the region.

[0091] The UE may assume that one or more NW additional conditions for a region remain stable over the duration used for model identification. The UE may select one or more UE models suitable for the observed NW additional conditions for a region.

[0092] From the gNB side, the NW side additional conditions can be kept as stable as possible for the region during the time period used for model identification. For example, the network can delay any changes during this duration if possible, which could lead to changes in the NW additional conditions.

[0093] For example, for BM instance 1, the beam codebook and / or antenna configuration / settings associated with the region may remain unchanged for the duration used for model identification.

[0094] The duration for model identification can be configured by the gNB. The duration for model identification can be associated with one or more functions configured by the NW; one or more conditions reported by the UE; and / or configured independently for model identification.

[0095] In some embodiments, the duration for model identification can be predefined. The duration for model identification can be defined in one or more standards. The UE can use a fixed duration for model identification, or, if more than one fixed duration for model identification exists, the UE can use one of those fixed durations. In some embodiments, the UE can send a fixed duration for model identification to gNB signaling.

[0096] In some embodiments, when there are more than one fixed duration for model identification, the UE may use more than one fixed duration for model identification. The UE may assume that one or more NW additional conditions remain stable within one fixed duration for model identification, and that the same NW additional conditions are not necessarily applicable in other fixed durations for model identification. The UE may select one or more UE models that are suitable for or applicable to the observed NW additional conditions across more than one fixed duration for model identification.

[0097] One or more DCI (Downlink Control Information) or MAC-CE (Media Access Control-Control Element) messages may include information such as one or more start model identifiers and / or end model identifiers. DCI or MAC-CE messages may be sent by the gNB to the UE. This can be used when the UE can have variations for the model identifier over a duration, depending on when the NW wishes to keep the NW additional conditions unchanged or change.

[0098] During the duration of model identification, the UE can identify, monitor, and / or evaluate the UE-side applicability of (multiple) ML models or dual ML models. This can be done for a given scenario, region, and / or site-specific setting for the gNB program to use ML features.

[0099] Model evaluation and monitoring can be performed by the UE when ML functionality is supported. The UE can be configured or instructed to use this function for a specified duration for model identification. This allows the UE to evaluate the performance of different UE-side AI / ML models while supporting the required ML functionality.

[0100] Alternatively or additionally, model evaluation and monitoring can be performed by the UE even when ML functionality is not supported. During the duration used for model identification, the UE is not expected to operate using ML functionality, but the performance of different UE-side AI / ML models can still be evaluated.

[0101] Model evaluation and / or monitoring can be performed by the UE with or without support from the NW.

[0102] A logical AI / ML model can refer to a model that has been identified and assigned a model ID. A physical AI / ML model can refer to the actual implementation of such a model.

[0103] In some embodiments, the UE may have N ML models (logical models) that are developed assuming different NW / UE / additional conditions. These can be common additional conditions.

[0104] In other words, N different ML models can be associated with N different assumptions about additional conditions. However, all N models can be associated with the same functionality. For example, N models can involve the same RRC (Radio Resource Control) configuration configured by the gNB.

[0105] The UE can monitor channel characteristics, beam measurements, and / or one or more other aspects based on the received signal.

[0106] In the BM Example 1 example, the UE can monitor the propagation environment to provide channel characteristics. The UE can use one or more of the following to provide beam measurements: L1-RSRP (Layer 1 Reference Signal Received Power), L1-SINR (Layer 1 Signal-to-Interference-Noise Ratio), and / or beam direction of arrival. The UE can utilize one or more of the following to monitor one or more desired aspects: DL RS (Downlink Reference Signal), PDSCH (Physical Downlink Shared Channel), PDCCH (Physical Downlink Control Channel), and / or similar channels / signals.

[0107] When sufficient monitoring is available at the UE (during the duration used for model identification), the UE can be able to map these observations to NW additional conditions. Since the UE knows which NW additional conditions are assumed for a given ML model, the UE can select the ML model by comparing the observed NW additional conditions with the additional conditions associated with the ML model.

[0108] In some embodiments, the UE may employ an internal model monitoring process and may identify at least one model that can be applied to scene, region, and / or site-specific settings that the gNB plans to use for ML features. This can be transparent to the gNB.

[0109] For example, in BM Instance 1, if the UE is targeting a given PCI / TRP or cell measurement prediction beamset and measurement beamset, the UE may be able to monitor the performance of N ML models. In some embodiments, the UE may monitor performance on fewer than N ML models. In some embodiments, the UE may monitor two or more of the N ML models. The UE evaluates the performance on its side. Once the UE has evaluated, it can identify which of these N models is suitable for the NW additional conditions considered during the duration used for model identification. The UE can then provide the evaluation results.

[0110] In some embodiments, the UE and gNB can provide monitoring resources and necessary assistance for the model monitoring process performed by the UE. In BM Instance 1, for example, the UE can acquire dedicated DL RS and auxiliary information to monitor the ML model, enabling the UE to evaluate the performance of the ML model with the support of the NW.

[0111] During or after the duration used for model identification, the UE may report one or more of the following:

[0112] A first indicator is used to report whether a model (new or old) is suitable for NW additional conditions considered during the duration used for model identification. The same first indicator can also be used to report that no model is suitable for NW additional conditions considered during the duration used for model identification. The first indicator can be transmitted using a single field (e.g., a single bit). Other embodiments may use one or more fields. This field may include one or more bits. The first indicator may indicate whether a new identifier for the UE-side model exists; and / or

[0113] The second indicator reports the model ID of the corresponding model to be identified by the UE. This can be a new model or a matching model ID previously identified with the model ID.

[0114] New models applicable to NW additional conditions during the duration used for model identification can be reported by the UE. The UE can indicate the model ID from a predefined / configured bit field. This can be done in an ascending order of assigning new model IDs or in any other suitable manner.

[0115] When the predefined / configured bit field used for the second indicator is full, the UE can replace the existing model ID (from the list that has already been identified) and report a new model suitable for use under the NW additional conditions for the duration used for model identification.

[0116] When an older model (previously reported to the NW) is applicable to the NW's additional conditions for the duration used for model identification, the UE can report the already identified model ID.

[0117] In some embodiments, the UE may use a first indicator to indicate a change. For example, the first indicator may have two bits: one bit for indicating a new model, one bit for replacing an older model, one bit for an existing model, and one bit for indicating no model if no suitable model is available.

[0118] UE reports can be NW-initiated reports. For example, a gNB-initiated scheduling report may exist when the duration used for model identification ends. UE reports can also be UE-initiated reports. For example, the UE triggers a MAC-CE message to report the first and / or second indicator.

[0119] In some embodiments, the UE may report additional information associated with the second indicator. The additional information may include associated cell and / or TRP and / or PCI and / or area information.

[0120] In some embodiments, when the UE considers more than one fixed duration for model identification to identify a model, the UE may report additional information associated with the second indicator. The additional information may include the associated fixed duration for model identification, the associated cell and / or TRP and / or PCI and / or area information.

[0121] Upon receiving a UE report indicating the first indicator and / or the second indicator, the gNB determines whether any new model identifiers exist. If a new model identifier exists with a new model ID or a new model identifier with an older model ID, the gNB can associate the model ID with a set of assumptions used by cell and / or TRP and / or PCI and / or area and duration information.

[0122] For example, based on the information possessed by the gNB, it can associate the model ID_x with a set of conditions and a set of conditions known at the network. In the BM Instance 1 example, the conditions are time information_K, cell_L, beam codebook_M, antenna configuration_N, TRP_implementation_P, and / or similar information.

[0123] It should be noted that the UE can support single-cell measurements or multi-cell / TRP / PCI measurements. This means that model identification can be performed on a per-cell and / or TRP and / or PCI basis, or on a group of cells and / or TRPs and / or PCIs within the duration used for model identification. The latter option reduces the need to perform the model identification process in the event of TRP or cell changes.

[0124] The gNB can send the model ID, and optionally, timing information associated with the UE's model identification process, to, for example, the AMF. Information sets (e.g., gNB implementation-related information, such as the beamcodebook used by the gNB and the antenna panel / configuration used by the gNB) can be stored within the NW (e.g., the gNB), which has an associated model ID.

[0125] For example, gNB can report to AMF that model ID_x is associated with condition set, time information_K, cell_L, and TRP_M.

[0126] AMF can collect more than one model ID and associated NW additional conditions and criteria. Depending on the information level of the gNB, the gNB can operate on the UE using the model ID LCM.

[0127] When a UE detaches / reattaches to the network, the AMF can refresh the model ID obtained from the gNB and can restart online model identification for the UE. This can be performed after the UE reports new capability information.

[0128] NW additional conditions can be referred to as DL direction additional conditions. This can be a case where the UE mainly observes these additional conditions based on the DL channel / signal.

[0129] From the NW perspective, during the duration of model identification, the model ID identified in the above manner can be used to implicitly identify the NW or common additional conditions (or DL-direction additional conditions) of the ML model. The model ID can be considered by the gNB and UE for signaling purposes to align with additional conditions where the NW plans to use ML features. The model ID may not select an exact physical model at the UE. Generally, if the UE can maintain many ML models for the same NW / common additional conditions, the model ID reported by the UE can refer to one or more ML models on the UE side. If there are more than one ML model associated with a model ID, depending on the UE-side additional conditions considered at a given time, there may be other additional selection methods used by the UE to select an ML model from multiple ML models.

[0130] refer to Figure 2a and Figure 2b The process explained in the text.

[0131] As referenced in Figure 1, the UE initiates initial attachment, during which a UE supporting ML reports conditions associated with its UE ML characteristics to the gNB. Initial attachment can, for example, be a result of the UE being powered on. The UE can report that it can support online model identification. The UE can report relevant parameters / conditions for supporting online model identification.

[0132] gNB can send supported ML features and conditions to AMF.

[0133] As referenced in Figure 2, gNB determines one or more functions (e.g., configurations that enable ML features) based on reported conditions.

[0134] The gNB can determine the appropriate duration for model identification. This can be based on changes to the NW plan with NW add-ons and / or parameters and / or conditions of the UE associated with online model identification.

[0135] As referenced in Figure 3, the UE receives configuration for online model identification. This can be received in an RRC configuration or reconfiguration message. The configuration for online model identification can provide the duration of model identification. It can provide parameters such as which one or more cells, one or more TRPs, and / or one or more PCIs can be considered for online model identification. The configuration for online model identification can provide an indication of resources that the UE can use to measure / monitor the UE-side background ML model. The configuration for online model identification may depend on or relate to one or more functions.

[0136] As referenced in Figure 4, the duration for model identification begins at the UE and gNB.

[0137] As referenced in Figure 5, the gNB identifier does not change the NW-side additional conditions used for the associated one or more cells, one or more TRPs, and / or one or more PCIs.

[0138] As referenced in Figure 6, the UE begins to evaluate and / or monitor the ML model.

[0139] As referenced in Figure 7, the UE performs ML inference operations based on functional LCM and additionally performs model ID-LCM using a model whose start time is known by the model identifier duration.

[0140] AI / ML models need to be developed, deployed, and managed throughout their entire lifecycle—a process known as AI / ML model lifecycle management (LCM).

[0141] In function-based LCM, the network indicates the activation and / or deactivation and / or fallback / switching of AI / ML functions via signaling. A function refers to an AI / ML enabling feature or feature group facilitated by configuration. Multiple configurations are supported based on conditions indicated by UE capabilities.

[0142] In model ID-based LCM, the model is identified at the network level, and the network / UE can activate / deactivate / select / switch individual AI / ML models via the model ID. The model can be associated with specific and additional conditions related to UE capabilities and AI / ML enabling features or feature groups.

[0143] In 3GPP systems, signaling can be provided by RRC signaling, MAC-CE, and / or DCI.

[0144] As referenced in Figure 8, the gNB can assist the UE by transmitting DL RS or auxiliary information to the UE. This auxiliary information can be used for model evaluation and / or monitoring. DL RS can allow the UE to evaluate and / or observe additional NW conditions.

[0145] As referenced in Figure 9, the duration used for model identification ends.

[0146] As referenced in Figure 10, based on the evaluation or monitoring performed by the UE during the model identification period, the UE determines which ML model is suitable for the observed NW additional conditions. This could be a new model to be identified to the gNB or a model that has already been identified to the gNB.

[0147] As referenced in Figure 11, gNB can unfreeze NW additional conditions after the model identifier duration ends. For example, gNB can change to a different set of NW additional conditions than those applicable during the model identifier duration.

[0148] The timing of the process portions referenced by reference numerals 10 and 11 in the attached figures may occur sequentially or overlap in time.

[0149] As referenced in Figure 12, the UE can assign a new model ID (e.g., model ID_X) to a given ML model.

[0150] As referenced in Figure 13, the UE transmits the results of the model identification process to the gNB. The UE may provide a report that may include an indication of whether a new model has been identified (e.g., a newly identified model) or whether it has an identifier for a model that has previously been identified to the gNB. The report may specify an assigned model ID (e.g., modelID_X). The report may include a first indicator and / or a second indicator that have been previously described.

[0151] As referenced in Figure 14, the UE can store or update information used to determine the model ID and associated duration. This information may include: UE additional conditions, observed NW additional conditions, and / or other additional conditions. The UE can map the same model ID to specific conditions and additional conditions. Examples of additional conditions may be time information, cell information, observed NW additional conditions, and other UE additional conditions (e.g., UE antenna configuration, UE implementation-related conditions, etc.).

[0152] As referenced in Figure 15, upon receiving a UE report, the gNB assesses whether any new model identifier has occurred. The report may contain a first indicator and / or a second indicator. If a new model identifier exists, whether with a new model ID or an older model ID, the gNB may associate the model ID with an additional set of conditions related to one or more cells, one or more TRPs, one or more PCIs, and / or one or more areas. The gNB may associate the model ID with duration information.

[0153] For example, based on available information, the gNB can map a model ID (e.g., model ID_X) to specific and additional conditions. Examples of additional conditions could be time information, cell information, beamcodebook, antenna configuration, TRP implementation, etc.

[0154] As referenced in Figure 16, the gNB sends the model ID and associated time information related to the UE's model identification process to the AMF. The complete set of information (including proprietary details or additional conditions that cannot be explicitly specified) can be stored on the NW side, for example, at the gNB. This can be associated with the corresponding model ID. As a result of this method, the AMF can accumulate multiple model IDs along with their corresponding NW additional conditions and criteria. The range of information available to the gNB can vary.

[0155] As referenced in Figure 17, the gNB can operate on the UE using the Model ID-LCM method based on the information it possesses. This can be done as previously described.

[0156] refer to Figure 6 It illustrates methods of some embodiments.

[0157] The method can be executed by a device. The device can be a user equipment.

[0158] The apparatus may include suitable components, such as a circuit system for providing the method.

[0159] Alternatively or additionally, the apparatus may include at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to provide at least the following methods.

[0160] Alternatively or additionally, the device may be as described above. Figure 5 The apparatus under discussion.

[0161] The method can be provided by computer program code or computer executable instructions.

[0162] The method may include receiving information for a defined time period from a base station, as referenced by reference numeral A1 in the attached figure.

[0163] The method may include, as referenced by reference numeral A2 in the figure, selecting one or more user equipment side models from a plurality of user equipment side models within a time period.

[0164] The method may include, as referenced by reference numeral A3 in the accompanying drawings, sending information to the base station indicating one or more user equipment side models selected from a plurality of user equipment side models.

[0165] It should be understood that Figure 6 The methods outlined herein can be modified to include any previously described features.

[0166] refer to Figure 7 This illustrates another method for some embodiments.

[0167] The method can be executed by a device. This device can be located in or near a base station.

[0168] The apparatus may include suitable components, such as a circuit system for providing the method.

[0169] Alternatively or additionally, the apparatus may include at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to provide at least the following methods.

[0170] Alternatively or additionally, the device may be as described above. Figure 4 The apparatus under discussion.

[0171] The method can be provided by computer program code or computer executable instructions.

[0172] The method may include, as referenced by reference numeral B1 in the figure, sending information to the user equipment defining a time period during which one or more user equipment-side models from a plurality of user equipment-side models will be selected.

[0173] The method may include receiving information from the user equipment, as referenced by reference numeral B2, indicating one or more user equipment-side models selected from a plurality of user equipment-side models.

[0174] It should be understood that Figure 7 The methods outlined herein can be modified to include any previously described features.

[0175] refer to Figure 8 It illustrates methods of some embodiments.

[0176] This method can be executed by a device. The device can be a user equipment.

[0177] The device may include suitable components, such as a circuit system for providing the method.

[0178] Alternatively or additionally, the apparatus may include at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to provide at least the following methods.

[0179] Alternatively or additionally, the device may be as described above. Figure 5 The apparatus under discussion.

[0180] The method can be provided by computer program code or computer executable instructions.

[0181] The method may include receiving information from a base station about the duration of the identification used for the machine learning model, as referenced by reference numeral C1 in the figure.

[0182] The method may include, as referenced by reference numeral C2 in the figure, performing identification of one or more user device-side machine learning models from a plurality of user device-side machine learning models during the duration.

[0183] It should be understood that Figure 8 The methods outlined herein can be modified to include any previously described features.

[0184] refer to Figure 9 This illustrates another method for some embodiments.

[0185] The method can be executed by a device. This device can be located in or near a base station.

[0186] The apparatus may include suitable components, such as a circuit system for providing the method.

[0187] Alternatively or additionally, the apparatus may include at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to provide at least the following methods.

[0188] Alternatively or additionally, the device may be as described above. Figure 4 The apparatus under discussion.

[0189] The method can be provided by computer program code or computer executable instructions.

[0190] The method may include, as referenced by reference numeral D1 in the attached figure, sending information to the user equipment about the duration of the machine learning model identification.

[0191] The method may include, as referenced by reference numeral D2 in the figure, receiving from a user equipment one or more reports regarding the identification of one or more user equipment-side machine learning models from a plurality of user equipment-side machine learning models.

[0192] It should be understood that Figure 9 The methods outlined herein can be modified to include any previously described features.

[0193] According to the model identifier definition, the model identifier process precisely indicates how the NW and UE have a shared understanding of the "conditions" and "additional conditions" associated with the ML model. RAN1 agrees that the "conditions" associated with the ML model can be identified through a functional framework, where functions are configured based on conditions reported by the UE (conditions are reported in the UE capability report). On the other hand, there is no precise agreement regarding additional conditions. However, at least the following additional conditions have been discussed.

[0194] Example of additional conditions: ○ Training dataset information ○ Site-related information (e.g., scene, location / TRP / area information, beam direction / codebook information) ○ Time-related / timestamp information ○ gNB implementation information (explicit or implicit details for specific gNB implementation details) ○ UE implementation information (explicit or implicit details for specific UE implementations) ○ Statistical information (such as latency spread, angle spread, and LOS / NLOS data) ○ Speed ​​and speed range information

[0195] For the UE-side model (and for the dual-side model), model training can be associated with a set of assumptions that take into account the additional conditions listed above (e.g., actual deployment assumptions). Therefore, the trained UE-side model can be associated with certain assumptions that take into account these additional conditions. It is feasible to classify these additional conditions into NW-side additional conditions (assumptions specific to the NW side), UE-side additional conditions (assumptions specific to the UE side), and common additional conditions (assumptions common to both the UE and NW sides).

[0196] Through the model identifier, the NW and UE should gain a common understanding of the model-related additional conditions, and certain steps may be required to capture air signaling linked to the online model identifier. However, by reviewing the example list of additional conditions above, it should be clear that accurately describing these additional conditions (for online model identifier) ​​in the specification will be challenging.

[0197] Observation: Additional conditions associated with the model can be categorized into NW-side additional conditions, UE-side additional conditions, and common additional conditions. ○ NW-side additional conditions can refer to the NW assumptions made by the ML model when training the model (or generating the training dataset). These assumptions may not be known at the UE during the inference phase. ○ UE-side additional conditions can refer to the UE assumptions made by the ML model when training the model (or generating the training dataset). These assumptions may not be known at the gNB during the inference phase. ○ Public additional conditions can refer to public assumptions made by the ML model when training the model (or generating the training dataset). These assumptions can be known at gNB and UE during the inference phase.

[0198] Observation: Specifying details about additional conditions can be a challenging task.

[0199] Observation: For model identification, i.e., for additional conditions that identify ML models, it may be easier to consider the implicit mutual understanding of the associated NW-side and UE-side additional conditions than to define the explicit details of the additional conditions.

[0200] As mentioned in previous chapters, the NW should know whether the UE can support the online model identification type (which implies additional conditions associated with the identification and the model), and expects the UE to indicate whether it supports the online model identification and the conditions used for it. Here, the conditions for the online model identification may include the preferred minimum duration for which the UE wishes to be configured to support the model identification. Furthermore, the conditions may also include any necessary assistance, associated functions, or conditions related to DL RS transmissions used for the model identification.

[0201] The UE should receive durations for model identification from the gNB to identify (multiple) UE-side models. These durations for model identification can be sent by the gNB based on configuration, such as the duration and indications regarding when to start and stop sending online model identification. The purpose of the durations for model identification is to ensure that the UE can assume that NW-side additional conditions remain stable during the duration of the model identification, and that the UE can select (multiple) UE models suitable for the NW-side additional conditions observed during that duration. Details of configuring cell / TRP / PCI / area information should be feasible, where the model identification durations are applicable. The UE can assume that the NW-side additional conditions for those cells / TRP / PCI / areas remain stable during the duration of the model identification, and that the UE can select (multiple) UE models suitable for the observed NW-side additional conditions for those cells / TRP / PCI / areas.

[0202] In some embodiments, for online identification of the UE-side model, the gNB can configure or instruct the UE to assume the duration of model identification.

[0203] It can also provide the UE with further information about the cell / TRP / PCI / area information to which the model identifier duration applies.

[0204] It should be noted that, from the gNB's perspective, the cell / TRP / PCI / area associated with the model identification duration or indication should maintain "NW-side additional conditions" stability for those cells / sites / areas / TRPs. Furthermore, since functional identification and functional LCM operate as baseline operations for ML features, the model identification process can be linked to the functions or function sets associated with the ML model. Through this link, the associated conditions used for online model identification can be determined.

[0205] Since NW deployments can differ from one another, NWs can also have other considerations when determining additional conditions on the NW side. In a sense, more freedom may be needed to configure different durations. From this perspective, it may be feasible to control the model identification duration by sending dynamic signaling via DCI and MAC-CE signaling, rather than configuring a fixed duration via RRC.

[0206] In some embodiments, the initialization of the model identifier and the termination of the model identifier(s) can be indicated via DCI or MAC-CE messages, wherein the UE can have a variable duration for the model identifier depending on when the NW wishes to keep the NW conditional static or not.

[0207] During the duration defined by NW for model identification, the UE shall determine, monitor and evaluate the suitability of (multiple) UE-side ML models for a given scenario, cell or site-specific setting for ML features planned by the gNB.

[0208] In one option, model evaluation and monitoring can be performed by the UE while the ML function is supported. The UE can be configured or instructed to operate through this function during this duration, allowing it to evaluate the performance of different UE-side AI / ML models while supporting the function. In another option, model evaluation and monitoring can also be performed by the UE without supporting the ML function. During this duration, the UE is not expected to operate through this function, but the performance of different UE-side AI / ML models can still be evaluated.

[0209] In some embodiments, during the model identification period, the UE should be able to evaluate / monitor different ML models to select / identify a suitable model for the additional conditions on the NW side. The following options may be considered depending on the UE's capabilities: • Option 1: Support ML features in parallel based on the functional framework. • Option 2: ML features are not supported during the model identifier duration.

[0210] Furthermore, it is feasible to assume that model evaluation and monitoring can be performed by the UE with or without additional support from the NW. In one example, for BM instance 1, the UE can have N ML models (logical models) developed with different NW / UE / common additional conditions. In other words, N different ML models can be associated with N different assumptions about additional conditions. However, all N models can be associated with the same functionality (the same RRC configuration configured by the gNB). The UE can monitor channel characteristics (propagation environment), beam measurements (L1-RSRP, L1-SINR, beam arrival direction, etc.) and other aspects based on the DL RS / PDSCH / PDCCH received by the UE from the gNB. When a good level of monitoring is available at the UE (during the model identification duration), the UE can be able to map these observations to NW additional conditions. Since the UE knows which NW additional conditions it assumes for a given ML model, the UE can select the ML model by comparing the observed NW additional conditions with the additional conditions associated with the ML model. In another option, the gNB can provide monitoring resources and necessary assistance for the UE's model monitoring process (visible to the gNB), and can identify at least one model that can be applied to a scenario, cell, or site-specific setting that the gNB plans to use for ML features. For the same example assumed earlier, the UE can acquire dedicated DL RS and assistance information to monitor the background ML model, allowing the UE to evaluate the performance of the ML model with NW support. When the UE has monitoring results, it can identify which of these N models is suitable for the NW setting considered during the model identification period.

[0211] In some embodiments, in order to enable the evaluation / monitoring of different ML models to select / identify the appropriate model for additional conditions on the NW side, the NW can provide further assistance by sending the required monitoring resources (DL RS) and sending auxiliary information.

[0212] During or after the duration for model identification, the UE can report an indicator to indicate whether the model (new or old) is suitable for the NW additional conditions considered during the duration for model identification. The same indicator can also report that no model is suitable for the NW additional conditions considered during the duration for model identification. This indicator can be transmitted using the same fields (e.g., a single bit in the report).

[0213] In some embodiments, the UE may report an indicator after or during the model identifier duration that highlights whether a new model identifier exists within the associated model identifier duration.

[0214] Additionally, the UE may also report another indicator, primarily reporting the model ID identified by the UE (new model) or a matching model ID previously identified using the model ID. New models applicable to the NW additional conditions within the model identification duration can be reported by the UE. Here, the UE is expected to indicate the model IDs from a predefined / configured bit field in ascending order of new model ID allocation. When an older model is applicable to the NW additional conditions within the model identification duration, the UE may report the already identified model ID. When a new model is applicable to the NW additional conditions within the model identification duration, the UE may report the new model ID. In this case, the UE is expected to indicate the model IDs from a predefined / configured bit field in ascending order of new model ID allocation. When the predefined / configured bit field for the second indicator is full, the UE may replace the existing model ID (from the already identified list) and report a new model suitable for the NW additional conditions used within the model identification duration.

[0215] UE reports can be NW-initiated reports (e.g., when the model identifier duration ends, there may be a gNB-initiated scheduling report) or UE-initiated reports (e.g., the UE triggers a MAC-CE message to report the first indicator and the second indicator).

[0216] In some embodiments, the UE may report a second indicator after or during the model identification duration, the second indicator identifying the model ID used for the model identification duration.

[0217] Upon receiving (multiple) UE reports, the gNB determines whether any new model identifiers exist. If a new model identifier exists with a new model ID or an older model ID, the gNB can associate the model ID with a set of assumptions used by the cell / TRP / PCI / area and duration information. For example, based on the information the gNB has, it can associate model_x with a set of "conditions" and a set of "additional conditions," where the additional conditions are time information_K, cell_L, beamcodebook_M, antenna configuration_N, TRP_implementation_P, etc.

[0218] From the NW perspective, the model ID identified in the above manner can only be used to implicitly identify the NW or common additional conditions (or additional conditions in the DL direction) of the ML model. The model ID is primarily considered by the gNB and UE for signaling purposes to align with additional conditions where the NW plans to use ML features. It does not always select the exact physical model at the UE. Generally, if the UE can maintain many ML models for the same NW / common additional conditions, the model ID reported by the UE can refer to one or more ML models on the UE side. If there are more than one ML model associated with a model ID, depending on the UE-side additional conditions considered at a given time, other additional selection methods may exist for the UE to select an ML model from multiple ML models.

[0219] In some embodiments, for model ID-based LCMs or functional LCMs using model IDs, the identified model ID can be reused by the NW and UE for signaling purposes, wherein the model ID refers to additional conditions assumed during the duration of the corresponding model identification.

[0220] Some implementations can be provided in 5G scenarios. (See reference) Figure 3 , Figure 4 and Figure 5 Briefly explain the 5th generation communication system (5GS), its access network and core network (5GC), and communication equipment.

[0221] Figure 3 A schematic representation of a 5G communication system (5GS) is shown. The 5GS may include a user equipment (UE) 300, an access network such as a 5G radio access network (5G-RAN) or a next-generation radio access network (NG-RAN), a 5G core network (5GC), and one or more application functions. Application functions may be deployed as trusted application functions within the 5GS, or they may be deployed or hosted on one or more application servers in the data network. Such application functions are untrusted application functions. The 5GS connects the UE to the data network via the access network and the 5GC (e.g., the 5GC's UPF).

[0222] 5G-RAN may include one or more radio access nodes, such as gNodeB (gNB) 100. gNB may include one or more gNodeB (gNB) distributed units connected to one or more gNodeB (gNB) centralized units.

[0223] 5GC can include the following network functions: Network Slice Selection Function (NSSF); Network Open Function; Network Repository Function (NRF); Policy Control Function (PCF); Unified Data Management (UDM); Application Function (AF); Authentication Server Function (AUSF); Access and Mobility Management Function (AMF); Session Management Function (SMF); and User Plane Function (UPF). Figure 3 Various interfaces (N1, N2, etc.) that can be implemented between various components of the system are also shown.

[0224] Figure 4 An example of device 200 is illustrated. The device may be a gNB or a portion thereof. Device 200 may include at least one random access memory (RAM) 211a, at least one read-only memory (ROM) 211b, at least one processor 212, 213, and a network interface 214. At least one processor 212, 213 may be coupled to RAM 211a and ROM 211b. At least one processor 212, 213 may be configured to execute appropriate software code 215. Execution of software code 215 may, for example, cause the device to perform operations for controlling the gNB. Software code 215 may be stored in ROM 211b.

[0225] Figure 5 An example of a communication device 300 is illustrated, such as Figure 1 Figure 2 and Figure 3 The UE illustrated above. Communication device 300 can be provided by any device capable of transmitting and receiving radio signals. Non-limiting examples of communication device 300 include user equipment, mobile station (MS), or mobile device such as a mobile phone or so-called "smartphone," a computer configured with a wireless interface card or other wireless interface facility (e.g., a USB dongle), a personal data assistant (PDA) or tablet computer configured with wireless communication capabilities, machine-type communication (MTC) devices, Internet of Things (IoT) type communication devices, or any combination thereof. Communication device 300 may include a transceiver for transmitting and / or receiving, for example, wireless signals carrying communication (e.g., radio signals). Communication can be one or more of voice, email, text messages, multimedia data, machine data, etc.

[0226] Communication device 300 can receive wireless signals (e.g., radio signals) via air or radio interface 307 through suitable means for receiving, and can transmit wireless signals via suitable means for transmitting. Figure 5In this diagram, the transceiver is schematically designated by block 306. Transceiver 306 may include, for example, a radio section and an associated antenna arrangement. The antenna arrangement may be located inside or outside the mobile device and may include one or more antenna elements. The antenna arrangement may be a multiple-input multiple-output (MIMO) antenna.

[0227] The communication device 300 may be configured with at least one processor 301, at least one ROM 302a, at least one RAM 302b, and other possible components 303 for use in software and hardware-assisted execution of tasks it is designed to perform, including control of access networks (e.g., Figure 3 The diagram illustrates access to and communication with 5G-RAN (or NG-RAN) and other communication devices. At least one processor 301 is coupled to RAM 302b and ROM 302a. At least one processor 301 can be configured to execute appropriate software code 308. The software code 308, for example, can allow the execution of one or more operations of the communication device. The software code 308 can be stored in ROM 302a.

[0228] Processors, ROM and RAM, transceivers, and other circuitry (e.g., modems) of the communication device may be provided on a circuit board, in a chipset, or in a system-on-a-chip. Circuit boards, chipsets, or systems-on-a-chip are indicated by reference numeral 304. The communication device 300 may optionally have a user interface, such as a keyboard 305, a touchscreen or pad, or a combination thereof. Depending on the type of communication device, one or more of a display, speaker, and microphone may optionally be provided.

[0229] It should be noted that while some embodiments have been described with respect to 5G networks, similar principles can be applied to other networks and communication systems. Therefore, although some embodiments have been described above by way of example with reference to certain example architectures used in wireless networks, technologies, and standards, these embodiments can be applied to any other suitable form of communication system besides the communication systems described herein.

[0230] It should also be noted that although exemplary embodiments have been described above, several changes and modifications can be made to the disclosed solutions without departing from the scope of the invention.

[0231] As used in this document, “at least one of the following: ” and “at least one of ” and similar wording, where the list of two or more elements is connected by “and” or “or”, means at least any one of the elements, or at least any two or more of the elements, or at least all of the elements.

[0232] Generally, various embodiments can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects of this disclosure can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, although this disclosure is not limited thereto. While various aspects of this disclosure may be illustrated and described as block diagrams, flowcharts, or using some other graphical representation, it should be understood that, by way of non-limiting example, these blocks, apparatuses, systems, techniques, or methods described herein can be implemented in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0233] As used herein, the term "circuit system" may refer to one or more of the following: (a) Hardware circuit implementation only (such as implementation only in analog and / or digital circuit systems) and (b) A combination of hardware circuitry and software, such as (if applicable): (i) A combination of (multiple) analog and / or digital hardware circuits with software / firmware, and (ii) Any part of a hardware processor (including digital signal processors), software, and memory (including multiple memory), which work together to enable a device (such as a mobile phone or server) to perform various functions, and (c) Multiple hardware circuits and / or multiple processors (such as multiple microprocessors or portions of multiple microprocessors) that require software (e.g., firmware) for operation, but may be absent when no software is required for operation.

[0234] This definition of circuit system applies to all uses of the term herein, including in any claim. As another example, as used herein, the term circuit system also covers only hardware circuitry or a processor (or processors) or portions thereof, and implementations of their accompanying software and / or firmware. For example, if applicable to a particular claim element, the term circuit system also covers baseband integrated circuits or processor integrated circuits for mobile devices, or similar integrated circuits in servers, cellular network devices, or other computing or networking devices.

[0235] Embodiments of this disclosure can be implemented by computer software executable by a data processor of a communication device, such as in a processor entity, or by hardware, or by a combination of software and hardware. Computer software or programs (also referred to as program products), including software routines, applets, and / or macros, can be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. A computer program product may include one or more computer-executable components that, when the program is run, are configured to perform the embodiments. The one or more computer-executable components may be at least one piece of software code or a portion thereof.

[0236] In this regard, it should also be noted that any box in the logic flow diagram can represent a program step, or interconnected logic circuits, boxes, and functions, or a combination of program steps and logic circuits, boxes, and functions. Software can be stored on physical media such as memory chips or memory blocks implemented within a processor, magnetic media such as hard disks or floppy disks, and optical media such as DVDs and their data variants, and CDs. Physical media are non-transitory media.

[0237] As used herein, the term “non-transient” refers to a limitation on the medium itself (i.e., tangible, not signal), rather than a limitation on the persistence of data storage (e.g., RAM vs. ROM).

[0238] The memory can be of any type suitable for the local technical environment and can be implemented using any suitable data storage technology, such as semiconductor-based memory devices, magnetic memory devices and systems, optical memory devices and systems, fixed memory, and removable memory. The data processor can be of any type suitable for the local technical environment and can include one or more of the following: as non-limiting examples, general-purpose computers, special-purpose computers, microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), FPGAs, gate-level circuits, and processors based on multi-core processor architectures.

[0239] The various example embodiments of this disclosure can be practiced in a variety of components, such as integrated circuit modules. The design of integrated circuits is largely a highly automated process. Complex and powerful software tools can be used to transform logic-level designs into semiconductor circuit designs ready to be etched and formed on semiconductor substrates.

[0240] The scope of protection sought by the various exemplary embodiments of this disclosure is set forth in the independent claims. Any exemplary embodiments and features described in this disclosure that do not fall within the scope of the independent claims should be interpreted as examples useful for understanding the various exemplary embodiments of this disclosure.

[0241] The foregoing description has provided a complete and informative description of various exemplary embodiments of the present disclosure by way of non-limiting and illustrative examples. However, various modifications and adaptations will become apparent to those skilled in the art when read in conjunction with the accompanying drawings and claims, given the foregoing description. Nevertheless, all such and similar modifications to the teachings will still fall within the scope of the various exemplary embodiments of the present disclosure as set forth in the claims. Further exemplary embodiments exist by way of non-limiting and illustrative examples, which include combinations of one or more exemplary embodiments with any other exemplary embodiments previously discussed.

Claims

1. A user equipment, comprising: A component for receiving information from the base station about the duration of the identifier used in the machine learning model; as well as A component for performing an identification of one or more user device-side machine learning models from a plurality of user device-side machine learning models during the said duration.

2. The user equipment according to claim 1, wherein one or more network-side additional conditions are stable during the time period.

3. The user equipment according to claim 1 or 2, wherein the receiving component is configured to receive information from the base station regarding one or more of the following: One or more cells associated with the duration used for machine learning model identification; One or more send / receive points associated with the duration used for machine learning model identification; One or more physical cell identifiers associated with the duration used for machine learning model identification; and The region associated with the duration used for machine learning model identification.

4. The user equipment according to any one of the preceding claims, comprising: In the downlink control indicator or in the media access control element, the information about the duration used for machine learning model identification is received from the base station.

5. The user equipment according to any one of the preceding claims, wherein the component for execution is configured to: evaluate a plurality of user equipment-side machine learning models among the plurality of user equipment-side machine learning models for one or more network-side additional conditions during the duration for machine learning model identification.

6. The user equipment according to any one of the preceding claims, wherein the component for receiving is configured to: receive information from the base station to assist in the identification of one or more user equipment-side machine learning models.

7. The user equipment according to any one of the preceding claims, wherein the receiving component is configured to: receive one or more reference signals from the base station to assist in the identification of one or more user equipment-side machine learning models.

8. The user equipment according to any one of the preceding claims, comprising: A component for reporting a first indicator to the base station, the first indicator indicating whether, during the duration for machine learning model identification, one or more user equipment-side machine learning models suitable for one or more network-side additional conditions exist.

9. The user equipment according to any one of the preceding claims, comprising: A component for determining whether one or more of the identified user equipment-side machine learning models are associated with a new identifier.

10. The user equipment according to any one of the preceding claims, comprising: A component for reporting a second indicator to the base station, the second indicator indicating: the identification of one or more user equipment-side machine learning models suitable for one or more network-side additional conditions during the duration for machine learning model identification.

11. The user equipment according to any one of the preceding claims, comprising: A component used to report additional information related to the second indicator.

12. The user equipment of claim 11, wherein the additional information includes at least one of the following: One or more durations associated with the second indicator; One or more cells associated with the second indicator; One or more transmit / receive points associated with the second indicator; One or more physical cell identifiers associated with the second indicator; and The area associated with the second indicator.

13. A base station, comprising: A component used to send information to the user equipment about the duration of the identification used for the machine learning model; as well as A component for receiving from the user equipment one or more reports regarding the identification of one or more user equipment-side machine learning models from a plurality of user equipment-side machine learning models.

14. The base station according to claim 13, further comprising: Components used to maintain the stability of one or more network-side additional conditions during the time period.

15. The base station according to claim 13 or 14, wherein the component for transmitting is used to transmit information about one or more of the following: One or more cells associated with the duration used for machine learning model identification; one or more transmit / receive points associated with the duration used for machine learning model identification; One or more physical cell identifiers associated with the duration used for machine learning model identification; and the region associated with the duration used for machine learning model identification.

16. The base station according to any one of the preceding claims, comprising: In the downlink control indicator or in the media access control element, the information regarding the duration used for machine learning model identification is sent.

17. The base station according to any one of the preceding claims, wherein the component for transmitting is configured to: transmit information to the user equipment to assist in the identification of one or more user equipment-side machine learning models.

18. The base station of claim 17, wherein the component for transmitting is configured to: transmit one or more reference signals to assist in the identification of one or more user equipment-side machine learning models.

19. The base station according to any one of the preceding claims, wherein the component for receiving comprises: A first indicator is received from the user equipment, the first indicator indicating whether, during the duration for machine learning model identification, one or more user equipment-side machine learning models suitable for one or more network-side additional conditions exist.

20. The base station according to any one of the preceding claims, wherein the component for receiving comprises: Receive a second indicator from the user equipment, the second indicator indicating: during the duration for machine learning model identification, the identification of one or more user equipment-side machine learning models suitable for one or more network-side additional conditions.

21. The base station according to any one of the preceding claims, wherein the component for receiving comprises: Receive additional information related to the second indicator from the user equipment.

22. The base station of claim 21, wherein the additional information includes at least one of the following: One or more durations associated with the second indicator; One or more cells associated with the second indicator; One or more transmit / receive points associated with the second indicator; One or more physical cell identifiers associated with the second indicator; and The area associated with the second indicator.

23. A method comprising: Receive information from the base station about the duration of the identifier used for the machine learning model; as well as During the duration, the identification of one or more user device-side machine learning models from a plurality of user device-side machine learning models is performed.

24. The method of claim 23, comprising: A first indicator is reported to the base station, the first indicator indicating whether, during the duration used for machine learning model identification, one or more user equipment-side machine learning models suitable for one or more network-side additional conditions exist.

25. The method according to claim 23 or 24, comprising: A second indicator is reported to the base station, the second indicator indicating the identification of one or more user equipment-side machine learning models suitable for one or more network-side additional conditions during the duration used for machine learning model identification.

Images