Wireless terminal and method thereof

Abstract

To enable a network (e.g., a radio access network (RAN) node) to control whether a wireless terminal is permitted to take action on the basis of a prediction or decision made by a trained machine learning model.SOLUTION: A wireless terminal receives control information from a network indicating whether the wireless terminal is authorized to perform a specific action on the basis of the inference results of the machine learning-based artificial intelligence. The wireless terminal makes a prediction or decision using the trained machine learning model. When the received control information permits it, the wireless terminal performs a specific action triggered by the prediction or decision made by the machine learning model. The specific action may include, for example, transmitting assistance information to the network indicating the results of the prediction or decision made using the trained machine learning model.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to wireless communication networks, and more particularly to the application of artificial intelligence (AI) to wireless communication networks. [Background technology]

[0002] The 3rd Generation Partnership Project (3GPP®) is discussing the application or introduction of AI or machine learning (ML) to 5G. AI / ML can be considered for both network internal functions and the air interface (i.e., Uu). In 3GPP Release 17, the Radio Access Network (RAN) Working Group #3 (RAN3) is discussing network-based AI / ML without User Equipment (UE) involvement, where examples of targets include energy saving, load balancing, and mobility optimization (see, for example, Non-Patent Documents 1 and 2). In network-based AI / ML, the network performs AI / ML inference. AI / ML inference refers to prediction or decision based on a trained machine learning model. The AI / ML inference function is expected to be applied to Next Generation Radio Access Networks (NG-RANs) (e.g., The machine learning model may be deployed in the NG-RAN (e.g., gNB). Training of the machine learning model may be performed in the NG-RAN. Alternatively, Operation, Administration and Maintenance (OAM) may train the machine learning model and provide the trained machine learning model (i.e., trained parameters) to the NG-RAN (e.g., gNB).

[0003] Furthermore, for 3GPP Release 18, network-based AI / ML with UE involvement and UE-based AI / ML have been proposed (see, for example, Non-Patent Documents 3-5). In UE-based AI / ML, the UE performs AI / ML inference. Potential use cases of AI / ML for the air interface include Channel State Information (CSI) feedback compression, beam management, positioning, Reference Signal (RS) overhead reduction, and mobility. In UE-based AI / ML, the UE runs an AI model (i.e., a trained machine learning model) and obtains the AI ​​inference results locally. For example, the UE can predict future events or measurements based on past measurements. The UE can feed back the predicted results (e.g., mobility or beam predictions) to the network (e.g., gNB). [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] CMCC, "Revised SID: Study on enhancement for data collection for NR and ENDC", RP-201620, 3GPP TSG-RAN Meeting #89e, Electronic meeting, September 14-18, 2020 [Non-patent document 2] 3GPP TR 37.817 V0.3.0 (2021-08) "3rd Generation Partnership Project; Technical Specification Group RAN; Evolved Universal Terrestrial Radio Access (E-UTRA) and NR; Study on enhancement for Data Collection for NR and EN-DC (Release 17)", September 23, 2021 [Non-patent document 3] Ericsson, "Views on Rel-18 AI / ML on Air-Interface", RP-212346, 3GPP TSG-RAN Meeting#93e, Electronic meeting, September 13-17, 2021 [Non-patent document 4] ZTE, "Support of Artificial Intelligence Applications for 5G Advanced", RP-212383, 3GPP TSG-RAN Meeting#93e, Electronic meeting, September 13-17, 2021 [Non-patent document 5] Xiaomi, "Mobility enhancement by UE based AI", RP-211787, 3GPP TSG-RAN Meeting#93e, Electronic meeting, September 13-17, 2021 Summary of the Invention [Problem to be solved by the invention]

[0005] The present inventors have studied UE-based AI / ML and found various issues. One of these issues relates to the execution conditions of UE-based AI / ML. For example, a UE performs some action based on a prediction or decision made using a machine learning model. However, if this UE action may result in an interaction with the network or affect network performance, it may not be appropriate for the UE to perform the action freely. Alternatively, it may not be appropriate for the UE to perform an action triggered by AI inference instead of being triggered by a rule (or criterion or formula) defined in a 3GPP specification. Actions triggered by a rule (or criterion or formula) defined in a 3GPP specification include, for example, sending a measurement report for handover, performing conditional mobility, beam selection, and cell (re)selection.

[0006] One of the objectives to be achieved by the embodiments disclosed in this specification is to provide an apparatus, a method, and a program that contribute to solving at least one of the multiple problems related to UE-based AI / ML, including the problem described above. It should be noted that this objective is only one of the multiple objectives to be achieved by the multiple embodiments disclosed in this specification. Other objectives or objectives and novel features will become apparent from the description of this specification or the accompanying drawings. [Means for solving the problem]

[0007] In a first aspect, a wireless terminal includes at least one memory and at least one processor coupled to the at least one memory. The at least one processor is configured to receive control information from a network indicating whether the wireless terminal is authorized to perform a specific action based on an inference result of a machine learning-based artificial intelligence. The at least one processor is configured to make a prediction or decision using a trained machine learning model. The at least one processor is configured to perform the specific action triggered by the prediction or decision if the control information authorizes performance of the specific action based on the inference result.

[0008] In a second aspect, a method performed by a wireless terminal includes the following steps: (a) receiving control information from a network indicating whether the wireless terminal is authorized to perform a particular action based on an inference result of a machine learning-based artificial intelligence; (b) making predictions or decisions using trained machine learning models; and (c) taking the specific action triggered by the prediction or decision, if the control information permits the execution of the specific action based on the inference result.

[0009] In a third aspect, a Radio Access Network (RAN) node includes at least one memory and at least one processor coupled to the at least one memory, the at least one processor configured to transmit control information to the wireless terminal indicating whether the wireless terminal is authorized to perform a specific action based on an inference result of a machine learning-based artificial intelligence, and if the control information authorizes performance of the specific action based on the inference result, the control information causes the wireless terminal to perform the specific action triggered by a prediction or decision made using a trained machine learning model.

[0010] In a fourth aspect, a method performed by a RAN node includes transmitting control information to a wireless terminal indicating whether the wireless terminal is authorized to take a particular action based on an inference result of a machine learning-based artificial intelligence, and if the control information authorizes performance of the particular action based on the inference result, the control information causes the wireless terminal to take the particular action triggered by a prediction or decision made using a trained machine learning model.

[0011] A fifth aspect is directed to a program, which includes a group of instructions (software code) that, when loaded into a computer, causes the computer to perform the method according to the second or fourth aspect. [Effects of the Invention]

[0012] According to the above-described aspects, it is possible to provide an apparatus, a method, and a program that contribute to solving at least one of a plurality of problems related to UE-based AI / ML. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a diagram illustrating an example of the configuration of a wireless communication system according to an embodiment. [Figure 2] FIG. 10 is a sequence diagram illustrating an example of the operation of a radio terminal and a radio access network node according to the embodiment. [Figure 3] 10 is a flowchart illustrating an example of an operation of the wireless terminal according to the embodiment. [Figure 4A] FIG. 10 is a diagram showing a specific example of a format of control information according to the embodiment. [Figure 4B] FIG. 10 is a diagram showing a specific example of a format of control information according to the embodiment. [Figure 5] FIG. 10 is a sequence diagram illustrating an example of the operation of a radio terminal and a radio access network node according to the embodiment. [Figure 6] FIG. 10 is a diagram showing a specific example of a format of an AI interest display according to an embodiment. [Figure 7]10 is a flowchart illustrating an example of an operation of a radio access network node according to the embodiment. [Figure 8] FIG. 2 is a block diagram illustrating a configuration example of a wireless terminal according to the embodiment. [Figure 9] FIG. 2 is a block diagram illustrating a configuration example of a radio access network node according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0014] Hereinafter, specific embodiments will be described in detail with reference to the drawings. In each drawing, the same or corresponding elements are designated by the same reference numerals, and for clarity of explanation, duplicate explanations will be omitted as necessary.

[0015] The multiple embodiments described below can be implemented independently or in appropriate combination. These multiple embodiments have different novel features. Therefore, these multiple embodiments contribute to solving different purposes or problems and to achieving different effects.

[0016] Although the following embodiments will be described mainly with respect to the 3GPP fifth generation mobile communication system (5G system), these embodiments may also be applied to other wireless communication systems.

[0017] As used herein, depending on the context, "if" may be construed to mean "when," "at or around the time," "after," "upon," "in response to determining," "in accordance with a determination," or "in response to detecting." These expressions may be construed to have the same meaning, depending on the context.

[0018] First, the configurations and operations of multiple network elements common to multiple embodiments will be described. Figure 1 illustrates an example configuration of a wireless communication system according to multiple embodiments. In the example of Figure 1, the wireless communication system includes a wireless terminal (i.e., UE) 1 and a radio access network (RAN) node (e.g., gNB) 2. Each element (network function) illustrated in Figure 1 can be implemented, for example, as a network element on dedicated hardware, as a software instance running on dedicated hardware, or as a virtualized function instantiated on an application platform.

[0019] UE1 has at least one radio transceiver and is configured to perform wireless communication with RAN node 2. UE1 is connected to RAN node 2 via an air interface 101. RAN node 2 is configured to manage a cell and perform wireless communication with multiple UEs, including UE1, using a cellular communication technology (e.g., NR Radio Access Technology (RAT)). UE1 may be simultaneously connected to multiple RAN nodes for dual connectivity (DC).

[0020] The RAN node 2 may be a Central Unit (e.g., gNB-CU) in a cloud RAN (C-RAN) deployment, or a combination of a CU and one or more Distributed Units (e.g., gNB-DUs). C-RAN is also referred to as a CU / DU split. Furthermore, a CU may include a Control Plane (CP) Unit (e.g., gNB-CU-CP) and one or more User Plane (UP) Units (e.g., gNB-CU-UP). Thus, the RAN node 2 may be a CU-CP or a combination of a CU-CP and a CU-UP. The CU may be a logical node that hosts the Radio Resource Control (RRC), Service Data Adaptation Protocol (SDAP), and Packet Data Convergence Protocol (PDCP) protocols of the gNB (or the RRC and PDCP protocols of the gNB). The DU may be a logical node that hosts the Radio Link Control (RLC), Medium Access Control (MAC), and Physical (PHY) layers of the gNB.

[0021] The UE1 may locally perform AI / ML inference. The AI / ML inference may relate to radio access network optimization. The UE1 may run AI inference on a trained machine learning model and take one or more actions according to a prediction or decision based on the AI ​​inference. The machine learning model may be any model known in the field of machine learning, including deep learning. The machine learning model may be, for example, but not limited to, a neural network model, a support vector machine model, a decision tree model, a random forest model, or a K-nearest neighbor model.

[0022] By way of example and not limitation, the AI-based prediction or decision by UE 1 and the resulting one or more actions may relate to beam management, mobility, or both. The one or more actions may include, but are not limited to, at least one of cell reselection, sending a measurement report, performing conditional mobility, and selecting a downlink beam. The downlink beam may be a Synchronization Signal (SS) / Physical Broadcast Channel (PBCH) block (SSB) beam. For example, the machine learning model may output a predicted result of a candidate cell for cell reselection, a target cell or node for handover, a candidate cell or node for conditional mobility, a candidate beam for beam selection, or a UE trajectory. Additionally or alternatively, the machine learning model may predict or determine the timing of performing an action for mobility or beam management.

[0023] Cell reselection may occur when UE1 is in RRC_IDLE or RRC_INACTIVE.

[0024] Mobility may occur when UE1 is in RRC_CONNECTED. Mobility in RRC_CONNECTED may be handover. The handover may be a Dual Active Protocol Stack (DAPS) handover. Additionally or alternatively, mobility in RRC_CONNECTED may relate to various mobilities in the DC. Specifically, the mobility may be a change of a primary cell of a Master Cell Group (MCG) in the DC, an inter-Master Node (MN) handover in the DC, a secondary node change in the DC, or an addition or change of a primary cell of a secondary cell group in the DC.

[0025] Conditional mobility may be performed when UE1 is RRC_CONNECTED. The conditional mobility may be a conditional handover. Additionally or alternatively, the conditional mobility may relate to various mobility in the DC. Specifically, the conditional mobility may be a change of a primary cell of a Master Cell Group (MCG) in the DC, an inter-Master Node (MN) handover in the DC, a secondary node change in the DC, or an addition or change of a primary cell of a secondary cell group in the DC.

[0026] The selection of the downlink beam may be performed in a Beam Failure Recovery (BFR) procedure.

[0027] The training of the machine learning model for AI / ML inference by the UE 1 may be performed by the UE 1 or by the network (e.g., OAM, RAN node 2). The training method may be offline learning, online learning, or a combination of these.

[0028] Similarly, the RAN node 2 may perform AI / ML inference. This AI / ML inference may relate to radio access network optimization. The RAN node 2 may run AI inference on a trained machine learning model and take one or more actions according to a prediction or decision based on the AI ​​inference. The machine learning model may be any model known in the field of machine learning, including deep learning. The machine learning model may be, for example, but not limited to, a neural network model, a support vector machine model, a decision tree model, a random forest model, or a k-nearest neighbor model.

[0029] By way of example and not limitation, the AI ​​reasoning-based prediction or decision by the RAN node 2 and the one or more actions triggered thereby may relate to at least one of energy saving, load balancing, mobility optimization, enhanced CSI feedback, and enhanced positioning accuracy. For example, the machine learning model's prediction or decision may relate to one or both of an energy saving strategy and a mobility strategy. With respect to a mobility strategy, the machine learning model may output a predicted result of a target cell or node for handover, a candidate cell or node for conditional mobility, or a UE trajectory.

[0030] The training of the machine learning model for AI / ML inference by the RAN node 2 may be performed by the RAN node 2 or by the OAM. The training method may be offline learning, online learning, or a combination of these.

[0031] First Embodiment A configuration example of the wireless communication system according to this embodiment may be the same as the example shown in FIG. 1. FIG. 2 shows an example of the operation of UE1 and RAN node 2 for UE-based AI / ML, i.e., for UE1 to perform AI inference using a machine learning model. In step 201, RAN node 2 transmits control information to UE1. The control information may be transmitted via broadcast (e.g., system information). The system information may be any system information block. Alternatively, the control information may be transmitted via dedicated signaling for UE1. The dedicated signaling may be dedicated RRC signaling, such as an RRC Reconfiguration message, an RRC Reestablishment message, an RRC Resume message, or an RRC Setup message. The control information may be an information element included in an SIB or the dedicated RRC signaling. The name of the information element may be, but is not limited to, "AI configuration."

[0032] The control information indicates whether UE1 is authorized to perform a specific action based on the inference result of the machine learning-based artificial intelligence. In other words, the control information indicates whether UE1 is authorized to perform a specific action triggered by a prediction or decision using the trained machine learning model. If the control information indicates authorization, the control information causes UE1 to perform the specific action triggered by a prediction or decision using the trained machine learning model. RAN node 2 may transmit the control information (step 201) only if UE1 is authorized to perform the specific action based on the AI / ML inference result. In this case, the transmission of the control information may implicitly indicate authorization for UE1 to perform the specific action based on the AI / ML inference result.

[0033] The term "machine learning-based artificial intelligence inference result" may be rephrased as an inference result by artificial intelligence, a self-learning inference result (by a UE), a self-correction inference result, or an optimization (algorithm) inference result. The expression "performing a specific action based on a machine learning-based artificial intelligence inference result" may be rephrased as applying an artificial intelligence inference result to a specific action, performing self-optimization (by a UE) to a specific action, or applying an optimization (algorithm) to a specific action. The term "radio access network optimization" may refer to, for example, optimization of the function or processing of a radio network (e.g., one or more RAN nodes) or optimization of the values ​​or settings of radio parameters that a radio network configures for a radio terminal (e.g., one or more UEs).

[0034] Specific examples of predictions or decisions based on AI inference by UE1 and one or more actions triggered thereby are as described above. By way of example and not limitation, specific actions triggered by predictions or decisions using a machine learning model in UE1 relate to at least one of beam management, mobility, CSI feedback, and positioning (or location estimation). Specific actions include, but are not limited to, at least one of cell reselection, transmission of a measurement report, execution of conditional mobility, and selection of a downlink beam. Transmission of a measurement report may trigger a decision by the network (e.g., RAN node 2) to initiate mobility (e.g., handover).

[0035] By way of example and not limitation, the specific action may relate to CSI feedback. The specific action may be adjusting the timing (e.g., periodicity) of CSI reporting. Additionally or alternatively, the specific action may be reducing or compressing the amount of information in CSI reporting (e.g., beam information, channel matrix, precoding matrix).

[0036] By way of example and not limitation, the particular action may relate to the positioning (or location estimation) of UE 1. The particular action may be a correction to information used for location estimation or a correction to the location estimation result.

[0037] Additionally or alternatively, the specific action may include transmitting assistance information (e.g., UE assistance information) indicating a result of a prediction or decision using the first machine learning model in UE1 to a network (e.g., RAN node 2). The assistance information may be used to train a second machine learning model related to radio access network optimization. The assistance information may be used to perform inference on the second trained machine learning model related to radio access network optimization. The second machine learning model and the second trained machine learning model may be located in RAN node 2 or OAM. Additionally or alternatively, the assistance information may be used to train (e.g., offline reinforcement learning) the first machine learning model located in UE1.

[0038] The control information may indicate whether or not execution of a specific action based on the AI ​​inference result is permitted, instead of or in addition to taking a specific action based on a predefined rule (or criterion or formula). The predefined rule may be a rule predefined in a 3GPP specification. Alternatively, the predefined rule may be a rule preconfigured by a network (e.g., RAN node 2). More specifically, the predefined rule may be a cell reselection rule (or criterion or formula), a measurement reporting rule (or criterion or formula), a conditional mobility execution rule (or criterion or formula), or a beam selection rule (or criterion or formula) preconfigured in a 3GPP specification or configured by the network. These rules (or criterion or formula) are typically based on comparing cell or beam quality (e.g., Reference Signal Received Power (RSRP)) with a threshold.

[0039] The control information may indicate to UE1 whether machine learning-based artificial intelligence (i.e., UE-based AI / ML) is permitted for multiple actions, including a specific action. Alternatively, the control information may indicate to UE1 individually whether each of multiple actions, including the specific action, is permitted. The control information may indicate actions or categories (or groups) of actions for which UE-based AI / ML is permitted. The control information may indicate functions or features (e.g., mobility, power saving) for which UE-based AI / ML is permitted. The control information may indicate subfunctions or subfeatures (e.g., cell selection, handover, beam management) for which UE-based AI / ML is permitted. The control information may indicate procedures (e.g., RRC re-establishment, beam failure recovery (BFR)) for which UE-based AI / ML is permitted. The control information may also indicate RRC configurations (e.g., information element or field level in Abstract Syntax Notation One (ASN.1)) for which UE-based AI / ML is permitted.

[0040] The control information may indicate conditions under which a specific action based on the AI / ML inference result is permitted to be performed. In this case, UE1 may perform the specific action triggered by the prediction or decision of the machine learning model (i.e., UE-based AI / ML) only if the indicated conditions are met. The conditions under which the specific action is permitted to be performed may indicate restrictions on at least one of frequency band, location, and time.

[0041] For example, the conditions may indicate a location (e.g., a geographic area, a cell, or a set of cells) where execution of a particular action based on the AI / ML inference results is permitted. The conditions may indicate a time or period where execution of a particular action based on the AI / ML inference results is permitted. The conditions may indicate one or more frequency bands where execution of a particular action based on the AI / ML inference results is permitted.

[0042] Additionally or alternatively, the condition may be that a consistent failure is detected. For example, the condition may allow UE1 to perform a specific action based on the AI / ML inference result if a predetermined failure is repeatedly detected in the same or similar situation (e.g., location, cell, cell pair, time, frequency band, time). The predetermined failure may be, for example, a handover failure or a beam failure. The situation may be at least one of a location, a cell, a cell pair, time, a frequency band, and time.

[0043] Additionally or alternatively, the condition may be that UE 1 receives a signal indicating a predetermined identifier from the network (e.g., RAN node 2). The predetermined identifier may be, for example, but not limited to, a Config Set Number or ID, an AI Set Number or ID, or a Combination Set Number or ID. The predetermined identifier may be set by the network to represent the same or similar conditions (e.g., transmit power, antenna tilt, number of downlink beams, or other physical layer settings of RAN node 2).

[0044] FIG. 3 illustrates an example of the operation of UE1. In step 301, UE1 receives the above-described control information. In step 302, UE1 makes a prediction or decision using the trained machine learning model. In step 303, if the control information indicates that it is permitted, UE1 performs a specific action triggered by the prediction or decision based on the machine learning model. In step 303, UE1 may consider a condition indicated in the control information. As described above, the condition may be a condition for UE1 to be permitted to perform a specific action based on the AI / ML inference result. In this case, UE1 may perform a specific action triggered by the prediction or decision of the machine learning model (i.e., UE-based AI / ML) only if the indicated condition is met.

[0045] Figure 4A shows an example of the format of the control information (steps 201 and 301). The Ai-configuration information element (IE) 401 shown in Figure 4A may be sent from the RAN node 2 to the UE 1 in a SIB or a dedicated RRC message. The Ai-configuration IE 401 may include an AI-ML-ConfigList IE 402. The AI-ML-ConfigList IE 402 includes one or more AI-ML-Config IEs 403. Each AI-ML-Config IE 403 may include one or any combination of a targetFeature IE 404, a targetArea IE 405, and a condition IE 406.

[0046] The targetFeature IE 404 may indicate one or more categories (or groups) of actions for which UE-based AI / ML is permitted to be applied, such as one or more of "mobilityConnected", "mobilityIdleInactive", "beamManagement", "energySaving", "rrmMeasurement", "csiFeedback", and "positioningAccuracy".

[0047] If the targetFeature IE404 is set to "mobilityConnected", this indicates that, for example, UE1 is permitted to apply UE-based AI / ML to the mobility function in the RRC_Connected state. The application of UE-based AI / ML to the mobility function in the RRC_Connected state may be a handover-related adjustment or a conditional handover (CHO)-related adjustment. The handover-related adjustment may be an adjustment of the reporting timing of a measurement report (MR) or an adjustment of an offset value or threshold of an MR event. The CHO-related processing adjustment may be an adjustment of an offset value or threshold of a CHO execution condition. Additionally or alternatively, the application of UE-based AI / ML to the mobility function in the RRC_Connected state may be an adjustment related to a PSCell addition or change in Multi-Radio Dual Connectivity (MR-DC) or an adjustment of a conditional PSCell addition or change (CPAC). PSCell stands for Primary SCell or Primary Secondary Cell Group (SCG) Cell.

[0048] If the targetFeature IE 404 is set to "mobilityIdleInactive", this indicates that, for example, UE1 is permitted to apply UE-based AI / ML to the mobility function in either or both of the RRC_IDLE and RRC_INACTIVE states. Applying UE-based AI / ML to the mobility function in the RRC_IDLE or RRC_INACTIVE state may mean adjustments related to cell reselection. The adjustments related to cell reselection may be adjustments of parameters used in the cell reselection process. For example, this may be adjustments of offset values ​​or thresholds, adjustments of per-frequency or inter-frequency priorities, or adjustments of per-network slice or inter-network slice priorities.

[0049] If the targetFeature IE404 is set to "beamManagement," this indicates, for example, that UE1 is permitted to apply UE-based AI / ML to the beam management function. Applying UE-based AI / ML to the beam management function may mean adjustments related to UE1's beam selection, beam failure detection (BFD) (or beam failure recovery (BFR)). The beam selection-related adjustments may be adjustments to offsets or thresholds for radio quality (e.g., RSRP, Reference Signal Received Quality (RSRQ), Received Signal Strength Indicator (RSSI)), or adjustments to priorities per beam or between beams. The beam failure detection-related adjustments may be adjustments to thresholds for determining BFD. The beam failure recovery-related adjustments may be adjustments to beam selection criteria for BFR (e.g., thresholds or reference signal (RS) type (e.g., SSB or CSI-RS)).

[0050] If the targetFeature IE 404 is set to "energySaving", this indicates, for example, that UE1 is permitted to apply UE-based AI / ML to features related to power saving. Applying UE-based AI / ML to features related to power saving may involve adjusting the timing (e.g., period, target frequency) of measurements of either or both of the serving cell and neighbor cells to result in (or are expected to result in) a reduction in power consumption. Such measurements may be either or both of Radio Link Monitoring (RLM) measurements and Radio Resource Management (RRM) measurements.

[0051] If the targetFeature IE 404 is set to "rrmMeasurement," this indicates, for example, that UE1 is permitted to apply UE-based AI / ML to the RRM measurement function. Applying UE-based AI / ML to the RRM measurement function may involve adjusting the timing of measurements to reduce power consumption as described above. Additionally or alternatively, this may involve adjusting the frequency of RRM measurements or the criteria for target relaxation (i.e., RRM relaxation). For example, the threshold for determining not-at-cell edge or low-mobility may be adjusted. For example, if UE1 determines itself to be not-at-cell edge or low-mobility, UE1 may be permitted to lengthen the RRM measurement period or reduce the target frequency.

[0052] If the targetFeature IE 404 is set to "csiFeedback", this indicates, for example, that UE 1 is allowed to apply UE-based AI / ML to the CSI feedback function. Applying UE-based AI / ML to the CSI feedback function may be adjusting the timing (e.g., period) of CSI reporting to RAN node 2. Additionally or alternatively, applying UE-based AI / ML to the CSI feedback function may reduce or compress the amount of information in the CSI report (e.g., beam information, channel matrix, precoding matrix).

[0053] If targetFeature IE404 is set to "positioningAccuracy", this indicates that, for example, UE1 is allowed to apply UE-based AI / ML to functions related to improving positioning accuracy. Applying UE-based AI / ML to functions related to improving positioning accuracy may mean correcting the information used for position estimation or correcting the position estimation result.

[0054] The targetArea IE 405 may indicate one or more areas where UE-based AI / ML application is permitted (i.e., permitted target functional areas). The area may include, for example, one or more of "intraFreq", "interFreq", "intraAndInterFreq", "interRAT", and "any". "IntraFreq" means cell reselection or handover within the same frequency band. "InterFreq" means cell reselection or handover between different frequency bands. "IntraAndInterFreq" means cell reselection or handover within the same frequency band and between different frequency bands. "InterRAT" means cell selection or handover between different radio access technologies. "any" means no restriction on the area. Alternatively, the absence of the targetArea IE 405 in the AI-ML-Config IE 403 may mean no restriction on the area. In addition, targetArea may be defined as targetFunc (targetFunction), targetScope, or applicableTarget, which may respectively indicate one or more target functions (target functions), one or more scopes (target ranges), or one or more application targets (applicable targets) to which UE-based AI / ML is permitted to be applied.

[0055] The condition IE406 may indicate a situation in which application of UE-based AI / ML is permitted. For example, the situation may include one or more of "consistentFailure," "predictable," "lowBattery," "gnssAvailable," and "nlos." "ConsistentFailure" refers to a situation in which a specific failure is repeatedly detected under the same or similar circumstances. "Predictable" refers to a situation that can be predicted based on history. "LowBattery" refers to a situation in which UE1's battery is low. "gnssAvailable" refers to a situation in which location information can be acquired using a Global Navigation Satellite System (GNSS). The GNSS may be, for example, a Global Positioning System (GPS), Galileo, or Global Navigation Satellite System (GLONASS). "nlos" refers to a situation in which UE1's radio wave environment (or communication environment) is a non-line-of-sight (NLOS) environment.

[0056] 4B shows another example of the format of the control information (steps 201 and 301). The Ai-configuration information element (IE) 421 shown in FIG. 4B may be transmitted from the RAN node 2 to the UE 1 in a SIB or a dedicated RRC message. The Ai-configuration IE 421 may include an AI-ML-ConfigList IE 422. The AI-ML-ConfigList IE 422 includes one or more AI-ML-Config IEs 423. Each AI-ML-Config IE 423 may include one or any combination of a mobility IE 424, an energySaving IE 425, and a positioningAccuracy IE 426.

[0057] The mobility IE 424 specifies a target function (targetFuncMob IE) and a target scope (targetScopeMob IE) related to mobility. The target function may indicate one or more categories (or groups) of actions for which UE-based AI / ML is permitted to be applied. Action categories include, for example, one or more of "mobilityConnected", "mobilityIdleInactive", "beamManagement", and "rrmMeasurement". The target scope may indicate one or more scopes for which UE-based AI / ML is permitted to be applied. The target scope may include, for example, one or more of "intraFreq", "interFreq", "intraAndInterFreq", "interRAT", and "any".

[0058] The energySaving IE 425 specifies target functions (targetFuncES IE) and target scopes (targetScopeES IE) related to power consumption reduction. The target functions may indicate one or more categories (or groups) of actions for which UE-based AI / ML is permitted to be applied. The action categories include, for example, one or more of "rrmMeasurement" and "csiFeedback". The target scope may indicate one or more scopes for which UE-based AI / ML is permitted to be applied. The target scope may include, for example, one or more of "intraFreq", "interFreq", "intraAndInterFreq", "interRAT", "any", "pCell", "sCell", and "servCell". "pCell" refers to the primary cell of carrier aggregation or the MCG or SCG primary cell of DC. "sCell" refers to the DC or the secondary cell of carrier aggregation. "servCell" refers to the serving cell of UE1.

[0059] The positioningAccuracy IE 426 specifies a target function (targetFuncPosi IE) and a target scope (targetScopePosi IE) related to positioning accuracy. The target function may indicate one or more categories (or groups) of actions for which UE-based AI / ML is permitted to be applied. The action categories include, for example, one or more of "nlosMitigation", "multipathMitigation", and "propagationDelayCompensation". "nlosMitigation" means correcting the positioning result (or UL transmission timing) when UE1 is in an NLOS environment. "multipathMitigation" means correcting the positioning result (or UL transmission timing) when UE1 is in a multipath environment. "propagationDelayCompensation" means compensating for the effects of propagation delay (e.g., UL transmission timing). The target scope may indicate one or more scopes for which UE-based AI / ML is permitted to be applied. The target scope includes, for example, one or more of "positioning" and "timingAdvance." "Positioning" refers to the acquisition of location information by a positioning function. "TimingAdvance" refers to the calculation of a transmission timing adjustment value (Timing Advance (TA)) that UE1 applies when transmitting on the uplink. The result of the TA calculation by UE1 may be reflected in the UL transmission timing of UE1 in a Non-Terrestrial Network (NTN), for example, or the result of the TA calculation may be reported from UE1 to RAN node 2.

[0060] According to the operations of UE1 and RAN node 2 described with reference to Figures 2 to 4B, RAN node 2 can control whether UE1 is permitted to take an action based on a prediction or decision made by UE-based AI / ML. For example, if an action of UE1 triggered by AI inference may result in an interaction with the network or may affect network performance, it may not be appropriate for UE1 to freely take that action. Alternatively, it may not be appropriate for UE1 to freely take an action that is triggered by AI inference instead of being triggered by a rule (or criterion or formula) defined in a 3GPP specification. The operations of UE1 and RAN node 2 described with reference to Figures 2 to 4B can contribute to solving these problems.

[0061] <Second embodiment> A configuration example of the wireless communication system according to this embodiment may be similar to the example shown in Fig. 1. Fig. 5 shows an example of the operation of UE1 and RAN node 2 for UE-based AI / ML, i.e., for UE1 to perform AI inference using a machine learning model. In step 501, UE1 transmits an AI interest indication to RAN node 2. In one example, UE1 may transmit the AI ​​interest indication during a procedure for setting up, resuming, or re-establishing an RRC connection. RAN node 2 may request UE1 to transmit the AI ​​interest indication.

[0062] The AI ​​interest indication may indicate to the RAN node 2 that the UE 1 supports machine learning-based artificial intelligence. The AI ​​interest indication may indicate to the RAN node 2 that the UE 1 is interested in running UE-based AI / ML. The AI ​​interest indication may indicate to the RAN node 2 that the UE 1 requests permission to run UE-based AI / ML. The AI ​​interest indication may be included in the UE Capability Information.

[0063] Additionally or alternatively, the AI ​​interest indication may indicate to the RAN node 2 that the UE 1 wishes to perform a particular action based on the inference results of the machine learning-based artificial intelligence. In other words, the AI ​​interest indication may indicate to the RAN node 2 that the execution of UE-based AI / ML by the UE 1 is expected to be beneficial for improving performance.

[0064] The AI ​​interest indication may indicate a category of AI / ML estimation that the UE 1 wishes to perform. In other words, the AI ​​interest indication may indicate an action or category of action for which the UE 1 wishes to be allowed to apply UE-based AI / ML. The AI ​​interest indication may indicate a function or feature (e.g., mobility, power saving) for which the UE 1 wishes to be allowed to apply UE-based AI / ML. The AI ​​interest indication may indicate a sub-function or sub-feature (e.g., cell selection, handover, beam management) for which the UE 1 wishes to be allowed to apply UE-based AI / ML. The AI ​​interest indication may indicate a procedure (e.g., RRC re-establishment, beam failure recovery (BFR)) for which the UE 1 wishes to be allowed to apply UE-based AI / ML. The AI ​​interest indication may indicate an RRC configuration (e.g., at the level of an information element or field in Abstract Syntax Notation One (ASN.1)) for which the UE 1 wishes to be allowed to apply UE-based AI / ML.

[0065] The AI ​​interest indication may indicate the magnitude or level of effect expected from UE1's performance of AI / ML estimation.

[0066] Step 502 is the same as step 201 in Figure 2. The RAN node 2 transmits control information to the UE 1. Details of the control information (e.g., transmission method, content) are the same as those described in the first embodiment, so description of the control information will be omitted here. The RAN node 2 may decide whether to transmit the control information (502) based on the AI ​​interest indication (501). The RAN node 2 may decide the content of the control information (502) based on the AI ​​interest indication (501).

[0067] Figure 6 shows an example of the format of an AI interest indication. The ai-ML-Assistance IE 602 shown in Figure 6 corresponds to the AI ​​interest indication. The ai-ML-Assistance IE 602 may be included in the UEAssistanceInformation IE 601. The UEAssistanceInformation IE 601 may be sent to the RAN node 2 via an RRC message (e.g., RRC Setup Request, RRC Setup Complete, RRC Resume Request, RRC Resume Complete, RRC Re-establishment Request, RRC Re-establishment Complete). The ai-ML-Assistance IE 602 may include the ai-ML-InterestIndication IE 603 ​​or 604.

[0068] The ai-ML-InterestIndication IE603 indicates, for example, that UE1 is interested in performing UE-based AI / ML. When UE1 desires permission to perform UE-based AI / ML, UE1 may include the ai-ML-InterestIndication IE603 in UEAssistanceInformation IE601.

[0069] Meanwhile, the ai-ML-InterestIndication IE604 indicates the category of AI / ML estimation that the UE1 wishes to perform. In other words, the ai-ML-InterestIndication IE604 indicates the action or category of action that the UE1 wishes to be allowed to apply UE-based AI / ML to. The category may include, for example, one or more of "mobilityConnected", "mobilityIdleInactive", "beamManagement", "energySaving", "rrmMeasurement", "csiFeedback", and "positioningAccuracy". The meaning (indication) of each of these may be as described above.

[0070] 7 is a flowchart showing an example of the operation of the RAN node 2. In step 701, the RAN node 2 receives an AI interest indication from the UE1. In step 702, the RAN node 2 determines whether the UE1 is permitted to perform a specific action based on the inference result of the machine learning-based artificial intelligence. In step 703, the RAN node 2 transmits control information to the UE1. The control information in step 703 indicates whether the UE1 is permitted to perform a specific action based on the inference result of the machine learning-based artificial intelligence. Details of the control information (e.g., transmission method, content) are the same as those described in the first embodiment, so a description of the control information will be omitted here.

[0071] According to the operations of UE1 and RAN node 2 described with reference to Figures 5 to 7, RAN node 2 can receive an AI interest indication (steps 501 and 701) from UE1 before transmitting control information for permission to execute UE-based AI / ML (steps 502 and 703). Thus, for example, RAN node 2 can determine whether to allow UE1 to execute UE-based AI / ML based on the AI ​​interest indication received from UE1. Alternatively, RAN node 2 can determine one or more actions (or categories of actions) that UE1 is permitted to apply in UE-based AI / ML based on the AI ​​interest indication received from UE1.

[0072] Next, exemplary configurations of a UE 1 and a RAN node 2 according to the above-described embodiments will be described below. FIG. 8 is a block diagram illustrating an exemplary configuration of a UE 1. A radio frequency (RF) transceiver 801 performs analog RF signal processing for communication with a RAN node. The RF transceiver 801 may include multiple transceivers. The analog RF signal processing performed by the RF transceiver 801 includes frequency up-conversion, frequency down-conversion, and amplification. The RF transceiver 801 is coupled to an antenna array 802 and a baseband processor 803. The RF transceiver 801 receives modulation symbol data (or OFDM symbol data) from the baseband processor 803, generates a transmit RF signal, and provides the transmit RF signal to the antenna array 802. The RF transceiver 801 also generates a baseband receive signal based on the receive RF signal received by the antenna array 802 and provides the baseband receive signal to the baseband processor 803. The RF transceiver 801 may include an analog beamformer circuit for beamforming. The analog beamformer circuitry includes, for example, multiple phase shifters and multiple power amplifiers.

[0073] The baseband processor 803 performs digital baseband signal processing (data plane processing) and control plane processing for wireless communication. Digital baseband signal processing includes (a) data compression / decompression, (b) data segmentation / concatenation, (c) transmission format (transmission frame) generation / decomposition, (d) transmission path coding / decoding, (e) modulation (symbol mapping) / demodulation, and (f) generation of OFDM symbol data (baseband OFDM signal) using Inverse Fast Fourier Transform (IFFT). Meanwhile, control plane processing includes communication management for Layer 1 (e.g., transmit power control), Layer 2 (e.g., radio resource management and hybrid automatic repeat request (HARQ) processing), and Layer 3 (e.g., signaling related to attachment, mobility, and call management).

[0074] For example, the digital baseband signal processing by the baseband processor 803 may include signal processing of a Service Data Adaptation Protocol (SDAP) layer, a Packet Data Convergence Protocol (PDCP) layer, a Radio Link Control (RLC) layer, a Medium Access Control (MAC) layer, and a Physical (PHY) layer. Also, the control plane processing by the baseband processor 803 may include processing of a Non-Access Stratum (NAS) protocol, a Radio Resource Control (RRC) protocol, MAC Control Elements (CEs), and Downlink Control Information (DCIs).

[0075] The baseband processor 803 may perform Multiple Input Multiple Output (MIMO) encoding and precoding for beamforming.

[0076] The baseband processor 803 may include a modem processor (e.g., a Digital Signal Processor (DSP)) that performs digital baseband signal processing and a protocol stack processor (e.g., a Central Processing Unit (CPU) or a Micro Processing Unit (MPU)) that performs control plane processing. In this case, the protocol stack processor that performs control plane processing may be shared with the application processor 804, which will be described later.

[0077] The application processor 804 is also referred to as a CPU, MPU, microprocessor, or processor core. The application processor 804 may include multiple processors (multiple processor cores). The application processor 804 executes a system software program (operating system (OS)) and various application programs (e.g., a call application, a web browser, a mailer, a camera operation application, and a music playback application) read from the memory 806 or a memory not shown, thereby realizing various functions of the UE1.

[0078] In some implementations, the baseband processor 803 and the application processor 804 may be integrated on a single chip, as indicated by the dashed line (805) in Figure 8. In other words, the baseband processor 803 and the application processor 804 may be implemented as a single System on Chip (SoC) device 805. An SoC device may also be called a system Large Scale Integration (LSI) or a chipset.

[0079] The memory 806 is volatile memory, nonvolatile memory, or a combination thereof. The memory 806 may include multiple physically independent memory devices. The volatile memory may be, for example, static random access memory (SRAM), dynamic RAM (DRAM), or a combination thereof. The nonvolatile memory may be mask read only memory (MROM), electrically erasable programmable ROM (EEPROM), flash memory, a hard disk drive, or any combination thereof. For example, the memory 806 may include an external memory device accessible from the baseband processor 803, the application processor 804, and the SoC 805. The memory 806 may also include an internal memory device integrated within the baseband processor 803, the application processor 804, or the SoC 805. Furthermore, the memory 806 may include memory within a universal integrated circuit card (UICC).

[0080] The memory 806 may store one or more software modules (computer programs) 807 including instructions and data for performing the processes by the UE 1 described in the above embodiments. In some implementations, the baseband processor 803 or the application processor 804 may be configured to read and execute the software modules 807 from the memory 806 to perform the processes by the UE 1 described in the above embodiments with reference to the drawings.

[0081] It should be noted that the control plane processing and operations performed by UE1 described in the above embodiment can be realized by elements other than the RF transceiver 801 and the antenna array 802, namely, at least one of the baseband processor 803 and the application processor 804, and the memory 806 storing the software module 807.

[0082] FIG. 9 is a block diagram showing an example configuration of a RAN node 2 according to the above embodiment. Referring to FIG. 9, the RAN node 2 includes a radio frequency transceiver 901, a network interface 903, a processor 904, and a memory 905. The RF transceiver 901 performs analog RF signal processing for communicating with UEs, including UE1. The RF transceiver 901 may include multiple transceivers. The RF transceiver 901 is coupled to an antenna array 902 and the processor 904. The RF transceiver 901 receives modulation symbol data from the processor 904, generates a transmit RF signal, and provides the transmit RF signal to the antenna array 902. The RF transceiver 901 also generates a baseband receive signal based on the receive RF signal received by the antenna array 902 and provides the baseband receive signal to the processor 904. The RF transceiver 901 may include an analog beamformer circuit for beamforming. The analog beamformer circuit may include, for example, multiple phase shifters and multiple power amplifiers.

[0083] The network interface 903 is used to communicate with network nodes (e.g., other RAN nodes, and control and forwarding nodes of the core network), and may include, for example, a network interface card (NIC) compliant with the IEEE 802.3 series.

[0084] The processor 904 performs digital baseband signal processing (data plane processing) and control plane processing for wireless communication. The processor 904 may include multiple processors. For example, the processor 904 may include a modem processor (e.g., a Digital Signal Processor (DSP)) that performs digital baseband signal processing and a protocol stack processor (e.g., a Central Processing Unit (CPU) or a Micro Processing Unit (MPU)) that performs control plane processing. The processor 904 may include a digital beamformer module for beamforming. The digital beamformer module may include a Multiple Input Multiple Output (MIMO) encoder and precoder.

[0085] The memory 905 is configured by a combination of volatile memory and non-volatile memory. The volatile memory is, for example, Static Random Access Memory (SRAM), Dynamic RAM (DRAM), or a combination thereof. The non-volatile memory is, for example, Mask Read Only Memory (MROM), Electrically Erasable Programmable ROM (EEPROM), flash memory, or a hard disk drive, or any combination thereof. The memory 905 may include storage located remotely from the processor 904. In this case, the processor 904 may access the memory 905 via the network interface 903 or an I / O interface (not shown).

[0086] The memory 905 may store one or more software modules (computer programs) 906 including instructions and data for performing the processing by the RAN node 2 described in the above embodiments. In some implementations, the processor 904 may be configured to read and execute the software modules 906 from the memory 905 to perform the processing by the RAN node 2 described in the above embodiments.

[0087] Note that if the RAN node 2 is a CU (e.g., gNB-CU) or a CU-CP (e.g., gNB-CU-CP), the RAN node 2 may not include the RF transceiver 901 (and the antenna array 902).

[0088] As described with reference to FIGS. 8 and 9, each of the processors included in the UE 1 and the RAN node 2 according to the above-described embodiments can execute one or more programs including instructions for causing a computer to perform the algorithms described with reference to the drawings. The programs include instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more functions described in the embodiments. The programs may be stored on a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, computer-readable media or tangible storage media include random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technology, CD-ROM, digital versatile disk (DVD), Blu-ray® disc or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The programs may also be transmitted on a transitory computer-readable medium or a communication medium. By way of example and not limitation, transitory computer-readable media or communication media include electrical, optical, acoustic, or other forms of propagated signals.

[0089] <Other embodiments> The above-described embodiments may be applied to a Non-Terrestrial Network (NTN), a vehicle-to-everything (V2X), high-speed trains (HSTs), unmanned aerial vehicles (UAVs), uncrewed aerial vehicles (UAVs), urban air mobility (UAM), and various other new applications.

[0090] The above-described embodiment may be applied to a secondary cell group (SCG) in dual connectivity (e.g., MR-DC). In this case, the RAN node 2 may be a secondary node (SN). In this case, the SN may transmit and receive RRC messages to and from the UE 1 directly using a signaling radio bearer (SRB3) in the SCG, or may transmit and receive RRC messages via a master node (MN). Additionally or alternatively, the RAN node 2 may be the master node (MN) and may play a role in transferring RRC messages between the SN and the UE 1.

[0091] The RAN node 2 in the above-described embodiment may be implemented in a C-RAN configuration. For example, the RAN node 2 may include a CU (e.g., gNB-CU) and a DU (e.g., gNB-DU). The CU may determine whether to permit UE1 to perform a specific action based on the inference result of the machine learning-based artificial intelligence, and transmit control information indicating the result of the decision to UE1 (via the DU). Additionally or alternatively, the DU may determine whether to permit the radio terminal to perform a specific action based on the inference result of the machine learning-based artificial intelligence, and transmit control information indicating the result of the decision to the CU, which then transmits the control information to UE1. Note that the above-described controls by the CU and DU may be selected or combined depending on the object, type, or content of the object that UE1 is permitted to perform.

[0092] Furthermore, the above-described embodiments are merely examples of application of the technical ideas obtained by the inventors of the present invention. In other words, the technical ideas are not limited to the above-described embodiments, and various modifications are possible.

[0093] For example, some or all of the above embodiments can be described as, but are not limited to, the following supplementary notes.

[0094] (Appendix 1) A wireless terminal, at least one memory; at least one processor coupled to the at least one memory; The at least one processor: receiving control information from a network indicating whether the wireless terminal is authorized to perform a particular action based on the inference results of the machine learning-based artificial intelligence; Making predictions or decisions using the trained machine learning model; and performing the specific action triggered by the prediction or decision if the control information permits the execution of the specific action based on the inference result; The wireless terminal is configured to: (Appendix 2) the control information indicates whether execution of the specific action based on the inference result is permitted, instead of or in addition to taking the specific action based on a predefined rule; 1. A wireless terminal as defined in claim 1. (Appendix 3) The predefined rules are rules predefined in a Third Generation Partnership Project (3GPP) specification or rules preconfigured by a network. 2. A wireless terminal as defined in claim 1. (Appendix 4) the control information individually indicates whether each of a plurality of actions including the specific action is permitted; 4. The wireless terminal according to claim 1. (Appendix 5) the control information collectively indicates permission to use the machine learning-based artificial intelligence for multiple actions, including the specific action; 4. The wireless terminal according to claim 1. (Appendix 6) the control information indicates a condition under which execution of a specific action based on the inference result is permitted; the at least one processor is configured to perform the specific action triggered by the prediction or decision only if the condition is met. 6. The wireless terminal according to any one of Supplementary notes 1 to 5. (Appendix 7) The conditions indicate constraints on at least one of a frequency band, a location, and a time. 7. A wireless terminal as defined in claim 6. (Appendix 8) the control information is transmitted by the network if execution of the specific action based on the inference result is permitted. 8. The wireless terminal according to any one of Supplementary notes 1 to 7. (Appendix 9) the at least one processor is configured to inform the network that the wireless terminal supports machine learning-based artificial intelligence prior to receiving the control information. 9. The wireless terminal according to any one of Supplementary notes 1 to 8. (Appendix 10) the at least one processor is configured to notify the network prior to receiving the control information that the wireless terminal desires to perform the specific action based on an inference result of the machine learning-based artificial intelligence. 10. The wireless terminal according to any one of Supplementary notes 1 to 9. (Appendix 11) The specific action relates to beam management or mobility. 11. The wireless terminal according to claim 1. (Appendix 12) The specific action includes at least one of cell reselection, sending a measurement report, performing conditional mobility, and selecting a downlink beam. 12. A wireless terminal according to any one of Supplementary notes 1 to 11. (Appendix 13) The specific action includes transmitting assistance information to the network indicating the results of a prediction or decision made using the trained machine learning model. 13. A wireless terminal according to any one of Supplementary notes 1 to 12. (Appendix 14) the assistance information is used to train a second machine learning model related to radio access network optimization or to perform inference on the second trained machine learning model related to radio access network optimization; 14. The wireless terminal of claim 13. (Appendix 15) 1. A method performed by a wireless terminal, comprising: receiving control information from a network indicating whether the wireless terminal is authorized to take a particular action based on an inference result of the machine learning-based artificial intelligence; Using the trained machine learning model to make predictions or decisions; and taking the specific action triggered by the prediction or decision if the control information permits the execution of the specific action based on the inference result; A method for providing (Appendix 16) A program for causing a computer to perform a method for a wireless terminal, comprising: The method comprises: receiving control information from a network indicating whether the wireless terminal is authorized to take a particular action based on an inference result of the machine learning-based artificial intelligence; Using the trained machine learning model to make predictions or decisions; and taking the specific action triggered by the prediction or decision if the control information permits the execution of the specific action based on the inference result; A program that includes: (Appendix 17) At least one memory; at least one processor coupled to the at least one memory and configured to send control information to the wireless terminal indicating whether the wireless terminal is authorized to perform a particular action based on an inference result of the machine learning-based artificial intelligence; Equipped with If the control information authorizes the execution of the specific action based on the inference result, the control information causes the wireless terminal to take the specific action triggered by the prediction or decision using the trained machine learning model. Radio access network node. (Appendix 18) the control information indicates whether execution of the specific action based on the inference result is permitted instead of or in addition to taking the specific action based on a predefined rule; 18. A radio access network node according to claim 17. (Appendix 19) The predefined rules are rules predefined in a Third Generation Partnership Project (3GPP) specification or rules preconfigured by a network. 19. A radio access network node according to claim 18. (Appendix 20) the control information individually indicates whether each of a plurality of actions including the specific action is permitted; 20. A radio access network node according to any one of Supplementary notes 17 to 19. (Appendix 21) the control information collectively indicates permission to use the machine learning-based artificial intelligence for multiple actions, including the specific action; 20. A radio access network node according to any one of Supplementary notes 17 to 19. (Appendix 22) the control information indicates a condition under which execution of a specific action based on the inference result is permitted; the control information causes the wireless terminal to perform the particular action triggered by the prediction or decision only if the condition is met. A radio access network node according to any one of Supplementary notes 17 to 21. (Appendix 23) The conditions indicate constraints on at least one of a frequency band, a location, and a time. 23. A radio access network node according to claim 22. (Appendix 24) the at least one processor is configured to transmit the control information if execution of the specific action based on the inference result is permitted. A radio access network node according to any one of Supplementary Notes 17 to 23. (Appendix 25) the at least one processor is configured to be notified by the wireless terminal that the wireless terminal supports machine learning-based artificial intelligence prior to transmitting the control information. A radio access network node according to any one of Supplementary notes 17 to 24. (Appendix 26) the at least one processor is configured to be notified by the wireless terminal prior to transmitting the control information that the wireless terminal desires to perform the specific action based on an inference result of the machine learning-based artificial intelligence. A radio access network node according to any one of Supplementary notes 17 to 25. (Appendix 27) The specific action relates to beam management or mobility. 27. A radio access network node according to any one of Supplementary notes 17 to 26. (Appendix 28) The specific action includes at least one of cell reselection, sending a measurement report, performing conditional mobility, and selecting a downlink beam. A radio access network node according to any one of Supplementary notes 17 to 27. (Appendix 29) the specific action includes transmitting assistance information to the radio access network node indicating a result of a prediction or decision using the trained machine learning model. A radio access network node according to any one of Supplementary notes 17 to 28. (Appendix 30) the assistance information is used to train a second machine learning model related to radio access network optimization or to perform inference on the second trained machine learning model related to radio access network optimization; 29. A radio access network node according to claim 29. (Appendix 31) transmitting control information to the wireless terminal indicating whether the wireless terminal is permitted to perform a particular action based on an inference result of the machine learning-based artificial intelligence; If the control information authorizes the execution of the specific action based on the inference result, the control information causes the wireless terminal to take the specific action triggered by the prediction or decision using the trained machine learning model. A method performed by a radio access network node. (Appendix 32) 1. A program for causing a computer to perform a method for a radio access network node, comprising: The method comprises transmitting control information to the wireless terminal indicating whether the wireless terminal is authorized to perform a particular action based on an inference result of the machine learning-based artificial intelligence; If the control information authorizes the execution of the specific action based on the inference result, the control information causes the wireless terminal to take the specific action triggered by the prediction or decision using the trained machine learning model. program.

[0095] This application claims priority based on Japanese Patent Application No. 2021-182104, filed on November 8, 2021, the disclosure of which is incorporated herein in its entirety. [Explanation of symbols]

[0096] 1 UE 2. RAN Node 803 Baseband Processor 804 Application Processor 806 memory 807 modules 904 processor 905 memory 906 Modules

Claims

1. A wireless terminal, means for receiving control information from a network indicating whether the wireless terminal is authorized to perform a specific action based on an inference result of the machine learning-based artificial intelligence; means for making predictions or decisions using the trained machine learning model; means for performing the specific action triggered by the prediction or decision if the control information permits the execution of the specific action based on the inference result; Equipped with The specific action includes transmitting assistance information to the network indicating the results of a prediction or decision made using the trained machine learning model. Wireless terminal.

2. the control information collectively indicates permission to use the machine learning-based artificial intelligence for multiple actions, including the specific action; The wireless terminal of claim 1 .

3. the control information is transmitted by the network if execution of the specific action based on the inference result is permitted. The wireless terminal of claim 1 .

4. the assistance information is used for training a second machine learning model related to radio access network optimization or for performing inference on the second trained machine learning model related to radio access network optimization; The wireless terminal of claim 1 .

5. 1. A method performed by a wireless terminal, comprising: receiving control information from a network indicating whether the wireless terminal is authorized to perform a particular action based on an inference result of the machine learning-based artificial intelligence; Using the trained machine learning model to make predictions or decisions; and performing the specific action triggered by the prediction or decision if the control information permits the execution of the specific action based on the inference result; The specific action includes transmitting assistance information to the network indicating the results of a prediction or decision made using the trained machine learning model. method.

6. the control information collectively indicates permission to use the machine learning-based artificial intelligence for multiple actions, including the specific action; The method of claim 5.

7. the control information is transmitted by the network if execution of the specific action based on the inference result is permitted. The method of claim 5.

8. the assistance information is used for training a second machine learning model related to radio access network optimization or for performing inference on the second trained machine learning model related to radio access network optimization; The method of claim 5.

Images