Life-cycle-management and configuration of measurement, inference, training data acquisition, and performance monitoring of artificial intelligence models system
AI/ML models in wireless communication systems address reactive provisioning by associating network conditions with IDs, enabling proactive network management and reducing resource consumption.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- ZTE CORP
- Filing Date
- 2025-03-28
- Publication Date
- 2026-06-04
AI Technical Summary
Existing wireless communication systems rely on reactive network provisioning based on current conditions, lacking proactive prediction of future network changes, which leads to inefficiencies and resource-intensive mobility management.
Implementing AI/ML models for predictive network management by associating network conditions with IDs and configuring AI functionalities at the UE and network nodes, enabling proactive network provisioning and flexible AI configuration.
Enhances network provisioning by reducing resource consumption and improving mobility management through proactive prediction of network conditions, reducing measurement overhead and enhancing flexibility in AI model deployment.
Smart Images

Figure CN2025085776_04062026_PF_FP_ABST
Abstract
Description
LIFE-CYCLE-MANAGEMENT AND CONFIGURATION OF MEASUREMENT, INFERENCE, TRAINING DATA ACQUISITION, AND PERFORMANCE MONITORING OF ARTIFICIAL INTELLIGENCE MODELS SYSTEMTECHNICAL FIELD
[0001] This disclosure is directed generally to wireless communication networks and particularly to Life-Cycle-Management (LCM) and configuration of network measurement, network inference, training data acquisition, and performance monitoring of Artificial Intelligence (AI) or Machine Learning (ML) models.BACKGROUND
[0002] In a wireless communication system, mobility management and provisioning of wireless terminals may require lengthy and resource-intensive measurement of current network conditions as well as reporting of a large amount of such measurement. Such mobility management and provisioning may only be reactive in that it derives resource allocation and other network configuration for the mobility based on current network conditions without considering potential future change of the network conditions at the time mobility management is performed t. Utilization of AI models for measurement prediction may be desired in providing proactive network provisioning.SUMMARY
[0003] This disclosure is directed generally to wireless communication networks and particularly to Life-Cycle-Management (LCM) and configuration of network measurement, network inference, training data acquisition, and performance monitoring of Artificial Intelligence (AI) or Machine Learning (ML) models. As a particular example, for the various network conditions that the AI prediction is performed, schemes to associate such network conditions with an associated IDs and to specify network conditions using such associated IDs at the AI functionality, measurement objects, cell, cell set, or area are disclosed to provide flexible AI configuraiton.
[0004] In some example implementations, a method performed by a User Equipment (UE) in communication with a network node of a wireless communication network is disclosed. The method may include receiving, from the network node, an AI configuration; determining applicable AI functionalities according to at least the AI configuration; reporting information associated with the applicable AI functionality to the network node; further receiving, from the network node, a response containing an AI inference configuration unless a full AI inference has already been included in the AI configuration; and activating AI inference and / or monitoring of a performance of the applicable AI functionalities according to the AI configuration and the AI inference configuration.
[0005] In the example implementations above, the method may further include receiving a capability inquiry with respect to UE side Artificial Intelligence (AI) functionalities from the network node; transmitting the initial AI capability information to the network node in response to the capability inquiry to prompt the network node to transmit the AI configuration.
[0006] In any one of the example implementations above, the capability inquiry comprises a filter for the UE to selectively transmit the initial AI capability information in an adaptive manner.
[0007] In any one of the example implementations above, the capability inquiry comprises an indication of whether to report AI-related UE capabilities by the UE.
[0008] In any one of the example implementations above, the initial AI capability information is provided as UE AI capabilities for each supported frequency, frequency band, or frequency band combination, and comprises at least a maximum number of cells that the UE is capable of performing AI prediction on.
[0009] In any one of the example implementations above, the initial AI capability information comprises, for a supported AI functionality, supported measurement configuration / parameter values or value ranges including at least one of: a list of supported Measurement Reduction in Time-domain (Rate MRRT) or measurement Reduction Rate in Spatial-domain (MRRS) ; a maximum or minimum value of supported MRRT / MRRS; a list of supported time offsets between measurements and corresponding predictions in a training data collection procedure for the supported AI functionality; a maximum or minimum value of the supported time offsets; or a maximum number of cells that the UE can perform prediction on; or a list of required Signal / PBCH Block (SSB) periodicities for measurements and / or predictions; or a maximum OR minimum number of required SSB periodicities for measurements and / or predictions.
[0010] In any one of the example implementations above, the initial AI capability information comprises a maximum number of cells that the UE can perform prediction on per measurement object, per measurement ID, per frequency / frequency band / band combination, or for the UE.
[0011] In any one of the example implementations above, the AI configuration comprises an indication of whether to enable beam level prediction by the UE.
[0012] In any one of the example implementations above, the AI configuration comprises a partial AI inference configuration comprising or indicating at least one of: one or more measurement objects, carrier frequencies, or cells that the UE is allowed or requested to perform AI inference on; one or more AI functionalities that the UE is allowed or requested to perform; one or more measurement objects or carrier frequencies that can be used as input to the applicable AI functionalities; one or more associated IDs for indicating a set of network additional conditions; or one or more AI inference parameter sets for measurements and predictions.
[0013] In any one of the example implementations above, each of the one or more AI inference parameter sets comprises at least one of: a first measurement interval between two continuous actual measurements; a second prediction interval between two continuous AI prediction; a first number of continuous actual measurements; a second number of continuous AI predictions; a third number of beams used as input of the applicable AI functionalities; or a fourth number of beams used as output of the applicable AI functionalities.
[0014] In any one of the example implementations above, the one or more AI inference parameter sets comprises a maximum / minimum parameter set containing maximum or minimum values of a plurality of inference parameters comprising at least one of: a first maximum or minimum measurement interval between two continuous actual measurements; a second maximum or minimum measurement interval between two continuous AI predictions; a third maximum or minimum number of continuous actual measurements; a fourth maximum or minimum number of continuous AI predictions; a fifth maximum or minimum number of beams used as input to the applicable AI functionalities; or a sixth maximum or minimum number of beams used as output to the applicable AI functionalities.
[0015] In any one of the example implementations above, the one or more associated IDs are specified at: per cell level, indicating that network additional conditions corresponding to a same associated ID need not be the same in different cells; per area level, indicating that network additional conditions corresponding to a same associated ID are the same in different cells of a same area; or per prediction scenario level, wherein a particular associated ID indicates usage of one set of network additional conditions as a basis to predict cell results with another set of network additional conditions.
[0016] In any one of the example implementations above, each measurement object or frequency carrier corresponds to one of the one or more associated IDs; or each cell or cell set corresponds to one of the one or more associated IDs.
[0017] In any one of the example implementations above, an associated ID corresponding to cells in each cell set indicates network additional conditions of each cell and a spatial relationship between the cells.
[0018] In any one of the example implementations above, the response comprises a time instance configuration information for measurements or AI predictions, the time instance configuration information comprising at least one of: a first measurement interval between two continuous actual measurements; a second measurement interval between two continuous AI predictions; a first number of continuous actual measurements; a second number of continuous AI predictions; a third number of beams used as input to the applicable AI functionalities; or a fourth number of beams used as output of the applicable AI functionalities.
[0019] In any one of the example implementations above, the AI inference and / or monitoring is activated by and explicit activation command received from the network node.
[0020] In any one of the example implementations above, the AI inference and / or monitoring is activated by conditions associated with one or more predefined measurement events being met.
[0021] In any one of the example implementations above, the reported information associated with the applicable AI functionalities comprises at least one of: one or more IDs of measurement objects or carrier frequencies that the AI inference is applicable; one or more cell IDs that the AI inference is applicable; one or more IDs of measurement objects or carrier frequencies that the AI inference is not applicable; one or more cell IDs that the AI inference is not applicable; one or more applicable AI inference parameter sets; or one or more supported associated IDs indicating network additional conditions.
[0022] In any one of the example implementations above, the method may further include receiving a command from the network node for deactivating the AI inference and / or monitoring of the performance of the applicable AI functionalities with respect to one or more measurement objects, carrier frequencies, or cells, the command comprising IDs of the one or more measurement objects, carrier frequencies, or cells to be deactivated.
[0023] In any one of the example implementations above, cell level AI inferences and beam level deactivations are triggered by different deactivation commands.
[0024] In any one of the example implementations above, the AI configuration or the response comprises a monitoring configuration for the monitoring of the performance of the applicable AI functionalities.
[0025] In any one of the example implementations above, the monitoring of the performance of the applicable AI functionalities is configured at one of the following levels: measurement object level; UE level; or AI functionality level.
[0026] In any one of the example implementations above, the AI inference configuration and the monitoring configuration are included with different measurement objects for inference and measurement objects for measurements, respectively, and a mapping between the measurement objects for inference and the measurement objects for measurements are mapped by the network node.
[0027] In any one of the example implementations above, the method may further include generating a single report of the monitoring of the performance of the applicable AI functionalities, the single report comprising both of prediction results and corresponding actual measurement results.
[0028] In any one of the example implementations above, the method may further include generating separate reports of the monitoring of the performance of the applicable AI functionalities separately comprising prediction results and corresponding actual measurement results within one monitoring periodicity, wherein a key is configured to correlate the separate reports.
[0029] In any one of the example implementations above, the method may further include generating a monitoring report of the performance of the applicable AI functionality, the monitoring report indicating monitoring results for at least one of: serving cells and top K neighbor cells; top K cells; cells indicated by the wireless communication network; cells whose monitoring results are above a threshold configured by the wireless communication network; or top K beams for each monitored cell, wherein K is a positive integer configured by the wireless communication network.
[0030] In some other example implementations, methods performed by the wireless network node above corresponding to the methods above are further disclosed.
[0031] In some other implementations, a UE or wireless network node comprising at least one a processor and a memory is disclosed. The at least one processor may be configured to read computer code from the memory to implement any one of the methods above.
[0032] In yet some other implementations, a computer program product comprising a non-transitory computer-readable program medium with computer code stored thereupon is disclosed. The computer code, when executed by a processor, may cause the processor to implement any one of the methods above.
[0033] The above embodiments and other aspects and alternatives of their implementations are described in greater detail in the drawings, the descriptions, and the claims below.BRIEF DESCRIPTION OF THE DRAWINGS
[0034] FIG. 1 illustrates an example wireless communication network including a wireless access network, a core network, and data networks.
[0035] FIG. 2 illustrates an example wireless access network including a plurality of mobile stations / terminals or User Equipments (UEs) and a wireless access network node in communication with one another via an over-the-air radio communication interface.
[0036] FIG. 3 shows an example radio access network (RAN) architecture.
[0037] FIG. 4 shows an example communication protocol stack in a wireless access network node or wireless terminal device including various network layers.
[0038] FIG. 5 illustrates an example flow for configuring and performing UE-side AI functionalities.
[0039] FIG. 6 illustrates an example flow configuring and performing training data collection for UE-side AI functionalities.
[0040] FIG. 7 illustrates and example configuration message of FIG. 5.
[0041] FIG. 8 illustrates and example configuration message of FIG. 5.
[0042] FIG. 9 illustrates and example configuration message of FIG. 5.
[0043] FIG. 10 illustrates an example flow for configuring and performing performance monitoring of UE-side AI models.
[0044] FIG. 11 illustrates an example configuration for measurements for monitoring UE-side AI models.
[0045] FIG. 12 illustrates an example configuraiton structure for monitoring UE-side AI models.
[0046] FIG. 13 illustrates an example flow for configuring and performing network-side AI functionalities.
[0047] FIG. 14 illustrates and example configuration of time gaps for measurements in monitoring network-side AI functionalities.
[0048] FIG. 15 illustrates an example configuration for associated ID for network additional conditions.
[0049] FIG. 16 illustrates another example configuration for associated ID for network additional conditions.
[0050] FIG. 17 illustrates yet another example configuration for associated ID for network additional conditions.DETAILED DESCRIPTION
[0051] The present disclosure will now be described in detail hereinafter with reference to the accompanied drawings, which form a part of the present disclosure, and which show, by way of illustration, specific examples of embodiments. The present disclosure may, however, be embodied in a variety of different forms and, therefore, the covered or claimed subject matter is intended to be construed as not being limited to any of the embodiments to be set forth below.
[0052] Throughout the specification and claims, terms may have nuanced meanings suggested or implied in context beyond an explicitly stated meaning. Likewise, the phrase “in one embodiment” or “in some embodiments” as used herein does not necessarily refer to the same embodiment and the phrase “in another embodiment” or “in other embodiments” as used herein does not necessarily refer to a different embodiment. The phrase “in one implementation” or “in some implementations” as used herein does not necessarily refer to the same implementation and the phrase “in another implementation” or “in other implementations” as used herein does not necessarily refer to a different implementation. It is intended, for example, that claimed subject matter includes combinations of exemplary embodiments or implementations in whole or in part.
[0053] In general, terminology may be understood at least in part from usage in context. For example, terms, such as “and” , “or” , or “and / or, ” as used herein may include a variety of meanings that may depend at least in part upon the context in which such terms are used. Typically, “or” if used to associate a list, such as A, B or C, is intended to mean A, B, and C, here used in the inclusive sense, as well as A, B or C, here used in the exclusive sense. In addition, the term “one or more” or “at least one” as used herein, depending at least in part upon context, may be used to describe any feature, structure, or characteristic in a singular sense or may be used to describe combinations of features, structures or characteristics in a plural sense. Similarly, terms, such as “a” , “an” , or “the” , again, may be understood to convey a singular usage or to convey a plural usage, depending at least in part upon context. In addition, the term “based on” or “determined by” may be understood as not necessarily intended to convey an exclusive set of factors and may, instead, allow for existence of additional factors not necessarily expressly described, again, depending at least in part on context.
[0054] This disclosure is directed generally to wireless communication networks and particularly to Life-Cycle-Management (LCM) and configuration of network measurement, network inference, training data acquisition, and performance monitoring of Artificial Intelligence (AI) or Machine Learning (ML) models. As a particular example, for the various network conditions that the AI prediction is performed, schemes to associate such network conditions with an associated IDs and to specify network conditions using such associated IDs at the AI functionality, measurement objects, cell, cell set, or area are disclosed to provide flexible AI configuraiton
[0055] Wireless Network Overview
[0056] An example wireless communication network, shown as 100 in FIG. 1, may include wireless terminal devices or user equipment (UE) 110, 111, and 112, a carrier network 102, various service applications 140, and other data networks 150. The wireless terminal devices or UEs, may be alternatively referred to as wireless terminals. The carrier network 102, for example, may include access network nodes 120 and 121, and a core network 130. The carrier network 110 may be configured to transmit voice, data, and other information (collectively referred to as data traffic) among UEs 110, 111, and 112, between the UEs and the service applications 140, or between the UEs and the other data networks 150. The access network nodes 120 and 121 may be configured as various wireless access network nodes (WANNs, alternatively referred to as wireless base stations) to interact with the UEs on one side of a communication session and the core network 130 on the other. The term “access network” may be used more broadly to refer a combination of the wireless terminal devices 110, 111, and 112 and the access network nodes 120 and 121. A wireless access network may be alternatively referred to as Radio Access Network (RAN) . The core network 130 may include various network nodes configured to control communication sessions and perform network access management and traffic routing. The service applications 140 may be hosted by various application servers deployed outside of but connected to the core network 130. Likewise, the other data networks 150 may also be connected to the core network 130.
[0057] In the example wireless communication network of 100 of FIG. 1, the UEs may communicate with one another via the wireless access network. For example, UE 110 and 112 may be connected to and communicate via the same access network node 120. The UEs may communicate with one another via both the access networks and the core network. For example, UE 110 may be connected to the access network node 120 whereas UE 111 may be connected to the access network node 121, and as such, the UE 110 and UE 111 may communicate to one another via the access network nodes 120 and 121, and the core network 130. The UEs may further communicate with the service applications 140 and the data networks 150 via the core network 130. Further, the UEs may communicate to one another directly via side link communications, as shown by 113.
[0058] FIG. 2 further shows an example system diagram of the wireless access network 120 including a WANN 202 serving UEs 110 and 112 via the over-the-air interface 204. The wireless transmission resources for the over-the-air interface 204 include a combination of frequency, time, and / or spatial resource. Each of the UEs 110 and 112 may be a mobile or fixed terminal device installed with mobile access units such as SIM / USIM modules for accessing the wireless communication network 100. The UEs 110 and 112 may each be implemented as a terminal device including but not limited to a mobile phone, a smartphone, a tablet, a laptop computer, a vehicle on-board communication equipment, a roadside communication equipment, a sensor device, a smart appliance (such as a television, a refrigerator, and an oven) , or other devices that are capable of communicating wirelessly over a network. As shown in FIG. 2, each of the UEs such as UE 112 may include transceiver circuitry 206 coupled to one or more antennas 208 to effectuate wireless communication with the WANN 120 or with another UE such as UE 110. The transceiver circuitry 206 may also be coupled to a processor 210, which may also be coupled to a memory 212 or other storage devices. The memory 212 may be transitory or non-transitory and may store therein computer instructions or code which, when read and executed by the processor 210, cause the processor 210 to implement various ones of the methods described herein.
[0059] Similarly, the WANN 120 may include a wireless base station or other wireless network access point capable of communicating wirelessly via the over-the-air interface 204 with one or more UEs and communicating with the core network 130. For example, the WANN 120 may be implemented, without being limited, in the form of a 2G base station, a 3G nodeB, an LTE eNB, a 4G LTE base station, a 5G NR base station of a 5G gNB, a 5G central-unit base station, or a 5G distributed-unit base station. Each type of these WANNs may be configured to perform a corresponding set of wireless network functions. The WANN 202 may include transceiver circuitry 214 coupled to one or more antennas 216, which may include an antenna tower 218 in various forms, to effectuate wireless communications with the UEs 110 and 112. The transceiver circuitry 214 may be coupled to one or more processors 220, which may further be coupled to a memory 222 or other storage devices. The memory 222 may be transitory or non-transitory and may store therein instructions or code that, when read and executed by the one or more processors 220, cause the one or more processors 220 to implement various functions of the WANN 120 described herein.
[0060] Data packets in a wireless access network such as the example described in FIG. 2 may be transmitted as protocol data units (PDUs) . The data included therein may be packaged as PDUs at various network layers wrapped with nested and / or hierarchical protocol headers. The PDUs may be communicated between a transmitting device or transmitting end (these two terms are used interchangeably) and a receiving device or receiving end (these two terms are also used interchangeably) once a connection (e.g., a radio resource control (RRC) connection) is established between the transmitting and receiving ends. Any of the transmitting device or receiving device may be either a wireless terminal device such as device 110 and 120 of FIG. 2 or a wireless access network node such as node 202 of FIG. 2. Each device may both be a transmitting device and receiving device for bi-directional communications.
[0061] The core network 130 of FIG. 1 may include various network nodes geographically distributed and interconnected to provide network coverage of a service region of the carrier network 102. These network nodes may be implemented as dedicated hardware network nodes. Alternatively, these network nodes may be virtualized and implemented as virtual machines or as software entities. These network nodes may each be configured with one or more types of network functions which collectively provide the provisioning and routing functionalities of the core network 130.
[0062] Returning to wireless radio access network (RAN) , FIG. 3 illustrates an example RAN 340 in communication with a core network 310 and wireless terminals UE1 to UE7. The RAN 340 may include one or more various types of wireless base station or WANNs 320 and 321 which may include but are not limited to gNB, eNodeB, NodeB, or other type of base stations. The RAN 340 may be backhauled to the core network 310. The WANNs 320, for example, may further include multiple separate access network nodes in the form of a Central Unit (CU) 322 and one or more Distributed Unit (DU) 324 and 326. The CU 322 is connected with DU1 324 and DU2 326 via various interfaces, for example, an F1 interface. The F1 interface, for example, may further include an F1-C interface and an F1-U interface, which may be used to carry control plane information and user plane data, respectively. In some embodiments, the CU may be a gNB Central Unit (gNB-CU) , and the DU may be a gNB Distributed Unit (gNB-DU) . While the various implementations described below are provided in the context of a 5G cellular wireless network, the underlying principles described herein are applicable to other types of radio access networks including but not limited to other generations of cellular network, as well as Wi-Fi, Bluetooth, ZigBee, and WiMax networks.
[0063] The UEs may be connected to the network via the WANNs 320 over an air interface. The UEs may be served by at least one cell. Each cell is associated with a coverage area. These cells may be alternatively referred to as serving cells. The coverage areas between cells may partially overlap. Each UE may be actively communicating with at least one cell while may be potentially connected or connectable to more than one cell. In the example of FIG. 1, UE1, UE2, and UE3 may be served by cell1 330 of the DU1, whereas UE4 and UE5 may be served by cell2 332 of the DU1, and UE6 and UE7 may be served by cell3 associated with DU2. In some implementations, a UE may be served simultaneously by two or more cells. Each of the UE may be mobile and the signal strength and quality from the various cells at the UE may depend on the UE location and mobility.
[0064] In some example implementations, the cells shown in FIG. 3 may be alternatively referred to as serving cells. The serving cells may be grouped into serving cell groups (CGs) . A serving cell group may be either a Master CG (MCG) or Secondary CG (SCG) . Within each type of cell groups, there may be one primary cell and one or more secondary cells. A primary cell in a MSG, for example, may be referred to as a PCell, whereas a primary cell in a SCG may be referred to as PScell. Secondary cells in either an MCG or an SCG may be all referred to as SCell. The primary cells including PCell and PScell may be collectively referred to as spCell (special Cell) . All these cells may be referred to as serving cells or cells. The term “cell” and “serving cell” may be used interchangeably in a general manner unless specifically differentiated. The term “serving cell” may refer to a cell that is serving, will serve, or may serve the UE. In other words, a “serving cell” may not be currently serving the UE. While the various embodiment described below may at times be referred to one of the types of serving cells above, the underlying principles apply to all types of serving cells in both types of serving cell groups.
[0065] FIG. 4 further illustrates a simplified view of the various network layers involved in transmitting user-plane PDUs from a transmitting device 402 to a receiving device 404 in the example wireless access network of FIGs. 1-3. FIG. 4 is not intended to be inclusive of all essential device components or network layers for handling the transmission of the PDUs. FIG. 4 illustrates that the data packaged by upper network layers 420 at the transmitting device 402 may be transmitted to corresponding upper layer 430 (such as radio resource control or RRC layer) at the receiving device 304 via Packet Data Convergence Protocol layer (PDCP layer, not shown in FIG. 4) and radio link control (RLC) layer 422 and of the transmitting device, the physical (PHY) layers of the transmitting and receiving devices and the radio interface, as shown as 406, and the media access control (MAC) layer 434 and RLC layer 432 of the receiving device. Various network entities in each of these layers may be configured to handle the transmission and retransmission of the PDUs.
[0066] In FIG. 4, the upper layers 420 may be referred as layer-3 or L3, whereas the intermediate layers such as the RLC layer and / or the MAC layer and / or the PDCP layer (not shown in FIG. 4) may be collectively referred to as layer-2, or L2, and the term layer-1 is used to refer to layers such as the physical layer and the radio interface-associated layers. In some instances, the term “low layer” may be used to refer to a collection of L1 and L2, whereas the term “high layer” may be used to refer to layer-3. In some situations, the term “lower layer” may be used to refer to a layer among L1, L2, and L3 that are lower than a current reference layer. Control signaling may be initiated and triggered at each of L1 through L3 and within the various network layers therein. These signaling messages may be encapsulated and cascaded into lower layer packages and transmitted via allocated control or data over-the-air radio resources and interfaces. The term “layer” generally includes various corresponding entities thereof. For example, a MAC layer encompasses corresponding MAC entities that may be created. The layer-1, for example, encompasses PHY entities. The layer-2, for another example encompasses MAC layers / entities, RLC layers / entities, service data adaptation protocol (SDAP) layers and / or PDCP layers / entities.
[0067] AI / ML Assisted Wireless Network Provisioning and Configuration
[0068] AI and Machine Learning (ML) (alternatively referred to AI generally) may facilitate more efficient configuration and provisioning in wireless networks. At the core of a general AI framework are various AI models. An AI model generally contains a large number of model parameters that are determined through a training process where correlations in a set of training data are learned and embedded in the trained model parameters. The trained model parameters may thus be used to generate inference or predictions from a set of input datasets. AI models are particularly suitable for situations where there is few trackable deterministic or analytical derivation paths between input data and output but correlations in historical data may be identified and may be embedded into the AI models via training processes.
[0069] AI models may be constructed and trained for various inference or predictive functionalities. Inferences or predictions generated by these AI models may facilitate various aspect of the configuration and provisioning of the over-the-air interface of a wireless network, including but not limited to resource management, resource allocation / selection (including spatial, radio frequency, and time resources) , mobility management, and radio link failure management.
[0070] Merely as an example, AI assisted mobility management and radio link failure management may be particularly beneficial. Specifically, traditional beam, cell, and radio frequency provisioning involved in mobility and radio link failure management is inflexible and consumes much overhead. For example, traditional beam management typically relies on exhaustive beam sweeping / scanning and measurements. A UE may be configured to monitor and measure each reference signal and then report the measurement results to the NW for the NW to decide the best beam for the UE to switch to during mobility. This process is resource and power intensive and time consuming. Trained AI models that embed learned correlation between various network condition parameters can at least help reduce the number of measurements (or requiring fewer reference signals) via inference capability of these AI models. For example, in some implementations, AI models may help identify inferences of best candidate beams using other network conditions and / or reduced number of beam measurement, and then only sweep and measure a smaller number of candidate beams to select the beam for use in current communication.
[0071] The traditional non-AI approaches to mobility and / or radio failure management are further only reactive in that they do not take historical time correlations into consideration and thus can only react to current measurements in relation to predetermined or configured threshold values. Such traditional approaches thus cannot predict future network conditions and unable to provide proactive mobility and radio link failure management. As such, in situations where cell quality changes, e.g., degrades, rapidly (e.g., when UE mobility is high) , the network may not be able to know the degraded cell quality in time based on the traditional reactive approaches, leading to frequent handover failure or radio link failure during mobility. Therefore, it may be desirable in at least in such situations to implement a proactive rather than reactive scheme. A proactive approach may also be based on AI techniques described above. In an AI-based proactive scheme for mobility and / or radio link failure management, for example, the network can know the beam / cell quality change in advance (even before the change actually happens) via Artificial Intelligence / Machine Learning (AI / ML) algorithm. For example, the UE / NW may be able to predict whether the beams / cells will become worse in a future time instance via temporal-domain prediction.
[0072] In some example implementations of AI / ML assisted network provisioning including but not limited to mobility and radio link failure management, AI models may reside on the UE side. In some other implementations, AI model may instead reside on the network (NW) side. AI models may be provided as a service. AI models may be retrained and updated as needed. The UE or the NW may determine what models to retrieve or retain and how they are updated and configured for assisting in the various aspects of wireless communications.
[0073] AI / ML Model Life Management Cycle
[0074] AI / ML models may be in general subject to a Life Cycle Management (LCM) , that provision the configuration, training, selection, deployment, updating, distribution, and performance monitoring of the AI / ML model in the wireless network. The LCM of AI / ML models for measurement prediction and for measurement event prediction to assist in mobility provisioning, however, may be specifically treated in a manner that is tailored to the purposes of these AI / ML models for measurement prediction and that benefits the purposes of effective mobility management and provisioning. For example, for L3 measurement, in addition to measurements on a serving cell measurement, neighbor cell measurement predications are also critical for mobility provisioning. As such, LCM of the AI / ML model with respect to L3 measurement prediction, including MO (measurement object) configuration, reporting configuration and other aspects of the LCM, may need to be adjusted from per cell level to per frequency level.
[0075] The configuration of the AI functionality by the network represents a basic aspect of the LCM of the AI in wireless network. FIG. 5 illustrates an example flow for configuring and activation of UE side AI functionalities (or models) . FIG. 5 include the following example steps for configuring the UE side AI functionalities and for activation / deactivation of such AI functionalities.
[0076] In Step 1 of FIG. 5, the network (via e.g., a base station, or access network node) sends an inquiry message, referred to as a UECapabilityEnqiry message, to initiate a procedure for a UE to report its AI / ML supported functionalities.
[0077] In Step 2 of FIG. 5, the UE, in response to the inquiry message of Step 1, prepares and sends a response message, referred to as a UECapablityInformation message, to the network, containing information about supported AI functionalities at the UE side.
[0078] In Step 3 of FIG. 5, the network, in response to the AI functionality information provided by the UE in Step 2, may prepare and transmit an AI configuration message to the UE. Such an AI configuration message may include one or more of the following AI configuration information, referred to as reference AI configuration information:
[0079] · Indication that UE is allowed to do UAI (UE Assistance Information) reporting via another configuration, referred to as OtherConfig;
[0080] · Network-side additional condition provided by the network; or
[0081] · An optional Inference configuration.
[0082] Between “Step 3” above and “Step 4” below, the UE may further decide the applicable AI functionalities based on NW-side additional conditions (if provided in Step 3 above) , UE-side additional conditions (internally known by UE) , model availability in the UE, and other information that facilitate the UE in determining applicable AI functionalities and their configuration.
[0083] In some example implementations, if the inference configuration is provided in the Step 3, for periodic Channel State Information (CSI) reporting, the UE may autonomously activate the applicable AI functionalities upon reporting applicable functionalities in Step 4 below.
[0084] In Step 4 of FIG. 5, the UE reports applicable functionalities to the network.
[0085] In Step 5 of FIG. 5, the network may configure inference configuration to UE in response to and after analyzing the applicable AI functionalities reported from the UE, if such inference configuration has not been sent to the UE in Step 3 above.
[0086] In Step 6 of FIG. 5, the UE-side AI functionalities and / or AI inference performance monitoring may be activated, either upon receiving full AI functionality configuration from the network in previous steps or by receiving an explicit activation command from the network.
[0087] FIG. 6 further illustrates an example flow for configuring and collecting training data from the network for the training of the UE-side AI functionalities. The example flow of FIG. 6 includes the following example steps for configuring training data and their collection.
[0088] In Step 1 of FIG. 5 (which may be optional) , the network can configure whether the UE is allowed to initiate request for data collection.
[0089] In Step 2 of FIG. 5 (which may be optional) , for data collection configuration for UE-side model training, the UE may send a request message to the network for training data collection.
[0090] In Step 3 of FIG. 5, the network may provide the data collection configuration in response to receiving the request message of Step 2 from the UE. In some other example implementations, the network may provide the UE with the data collection configuraiton at any point in time, with or without any UE request. In some example implementations, in addition to data collection related configuration (s) , associated ID (s) may also be included in the training data collection configuration transmitted from the network to the UE. The various implementations related to the associated IDs are explained in further detail below.
[0091] In Steps 4-6 of FIG. 5, the network can decide when to start / stop / release the data collection and instruct the UE of its decision. The UE may perform data collection
[0092] UE-Side Model / Functionality Identification / Configuraiton
[0093] Step 1 of FIG. 5 is intended for the network to inquire general AI capability information from the UE. In order to limit signaling overhead, the inquiry message sent by the network to the UE may include information that limit the scope of AI capability information that is necessary fo the network to determine the AI configuration to send to the UE later in Step 3 and / or in step 5. As such, through this inquiry message, referred to as UECapabilityEnqiry message, the network may configure some filtering in order to avoid excessive and / or unnecessary transmission by the UE. As an example, the inquiry message may one or more of:
[0094] 1) A set of high-level AI features (e.g., AI for beam management, AI for mobility, and the like) that the network would like to request the UE to report, so that the UE only need to investigate its AI capabilities with respect to these relevant features when responding to the network in the next step;
[0095] 2) A set of specific AI functionalities that the network is concerned (e.g., L3 cell temporal domain prediction, L3 beam temporal domain prediction, L3 cell level frequency domain prediction) , so that the UE only need to investigate its AI capabilities with respect to these relevant functionalities when responding to the network in the next step; or
[0096] 3) frequencies / bands / band combinations that the network would like to request the UE to report its relevant AI-related capability.
[0097] In some example implementations, the inquiry message may include an indication to the UE as to whether to report AI-related capability. If the indication is negative, the UE may not need to report any AI capabilities and may disable UE-side AI functionalities until further reconfiguration by the network.
[0098] Step 2 of FIG. 5 is intended for the UE to report its AI-related capability to the network in response to the inquiry message in Step 1. In generating the report of AI-related capabilities, the UE may apply the filtering indicated in the inquiry message above in order to only provide desired information and to reduce communication overhead. This response by the UE may be included in a reply message, referred to as UECapabilityInformation message. Such a message may include or indicate one or more of:
[0099] · For each supported frequency / band / band combination, one or more supported AI / LM functionalities (e.g. L3 cell frequency domain prediction, L3 beam frequency domain prediction) .
[0100] · For each supported frequency / band / band combination, a set of supported network additional conditions. In some example implementations, each set of supported network additional conditions may be explicitly signaled in this response message. Alternatively, each set of supported network additional conditions may be represented and indicated by an associated ID. Each associated ID, as described in further detail below, may linked to a predefined or configurable set of network additional conditions. Merely as an example, the additional network conditions may include base station or cell attribution / configuration, e.g., ISD or antenna height or deployment information.
[0101] · For each frequency / band / band combination, a maximum number of cell that the UE is capable of performing AI prediction on.
[0102] · For each supported AI / ML feature / functionality / model, supported measurement configuration / parameter values or supported measurement configuration / parameter ranges (e.g., maximum value or minimum value) , which may include or indicate one or more of the following:
[0103] o A list of the supported measurement time interval between two continuous measurements, or a maximum supported measurement time interval and / or a minimum supported measurement time interval.
[0104] o A list of the supported AI prediction time interval between two continuous AI predictions, or a maximum supported AI prediction time interval and / or a minimum supported AI prediction time interval.
[0105] o A list of the supported number of continuous AI-predicted results, or a maximum / minimum supported number of continuous AI-predicted results.
[0106] o A list of the supported number of continuous actual measurement results, or a maximum / minimum supported number of continuous actual measurement results.
[0107] o A list of supported Measurement Reduction Rate in Time domain (MRRT) or Measurement Reduction Rate in Spatial domain (MRRS) , or a maximum / minimum value of supported MRRT / MRRS.
[0108] o A list of supported time offsets between measurements on a frequency for measurement purposes, and the measurement on the frequency for prediction purposes in the training data collection procedure, or maximum / minimum supported time offset.
[0109] o A maximum number of the cells that the UE is capable of perform AI prediction. In some example implementations, such maximum number may be specified per measurement object, per measurement ID, per band combination, or per UE.
[0110] o A list of required SSB periodicity for a frequency for measurement and / or a frequency for AI prediction, or a maximum / minimum required SSB periodicity for a frequency for measurement and / or a frequency for AI prediction.
[0111] o For measurement event predictions, a list of, a maximum of, or a minimum of one or more of supported TTT, supported event threshold (e.g., A1-threshold) , or supported event offset (e.g., A3-offset) .
[0112] o A list of filtering coefficients, or a maximum filtering coefficient or a minimum filtering coefficient.
[0113] o One or more reference configurations, e.g., SSB or CSI-RS configurations.
[0114] o One or more supported relationships between the cells used as AI model input and the cells used as AI model output in inter-cell prediction (e.g., the relationship being co-located)
[0115] o A list of numbers of cells / beams used as AI model input, or maximum or minimum number of cells / beams used as AI model input.
[0116] o A list of the number of cells / beams used as AI model output, or maximum or minimum number of cells / beams used as AI model output.
[0117] Step 3 of FIG. 5 is intended for the network to configure the UE with respect to available and relevant AI capabilities. Such a configuration message may be generated by the network in response to receiving the UE capability information of Step 2.
[0118] Such a configuration message, for example, may include one or more measurements thresholds (e.g. RSRP threshold) for the cell (s) whose applicable AI functionality is to be reported in Step 4 below. Under such configuration, the UE may only report applicability of AI functionalities applicable to only the cells whose measurements are above the configured thresholds. Specifically, if the RSRP value of cell A is below the threshold, the UE would not need to consider this cell when reporting applicability of AI functionality to the network.
[0119] In some example implementations, the configuration message of Step 3 may include an indication as to whether to enable the applicability reporting for beam level predictions. If the configuration indicate that the beam level prediction is not enabled, then the UE would not need to consider beam level prediction when reporting applicability of AI functionality to the network.
[0120] In some example implementations, the configuration message of Step 3 may include one or more of the following configuration information items in order to avoid overly frequent reporting by the UE:
[0121] · A timer to avoid unnecessary reporting. For example, an AI functionality may become applicable only for a duration. The UE may report such functionality as applicable only in that duration. Once the timer associated with that duration expires, the AI deems the functionality non-applicable and thus would not report it.Such a timer may be configured and activated by the network.
[0122] · A minimum time difference between two reporting of applicable AI functionalities in order to avoid overly-frequent reporting of AI functionality by the UE for the applicable AI functionalities.
[0123] In some example implementations, the configuration message of Step 3 may include partial inference configuration for the AI models that may have already been determined by the network without needing additional information from the UE. Such inference information may be applied by the UE to determine whether the UE has available functionalities and / or how should inference of measurements is to be made using the applicable AI functionalities. Such partial inference configuration, for example, may include or indicate one or more the following:
[0124] · One or more measurement objects (e.g., via measurement ID (measId) or measurement object ID (measObjectId) or carrier frequency or cell ID) that the UE is allowed or requested to perform inference on. In some example implementations, the one or more measurement object may be provided as a list and the list may be provided be per UE level, per feature level, per functionality level, or the like.
[0125] · one or more AI functionalities (e.g., L3 beam temporal domain prediction, L3 cell temporal domain prediction…) that UE is allowed or requested to perform inference with. In some example implementations, the AI functionalities may be reported per UE level, or per measId level, or per measObject level, or per frequency level, and the like. For example, for each measObject, a list of AI functionalities may be provided in the configuration message above to indicate which AI functionalities are allowed on this measObject.
[0126] · One or more measId or measObjectId or frequency whose measurement / prediction results can be used as AI model input.
[0127] · One or more associated ID linking to corresponding sets of network additional conditions for applicable AI functionalities. The motivation of the associated id is to enable the consistency between the training and inference. Details are provided below.
[0128] · One or more of the following configurations provided by the network for the network to adopt a try-and-error approach to AI inference:
[0129] o One or more inference parameter set, where each parameter set may include one or more of these parameters:
[0130] · A measurement interval between two continuous actual measurements;
[0131] · A measurement interval between two continuous AI predictions;
[0132] · A number of continuous actual measurements;
[0133] · A number of continuous AI predictions;
[0134] · A number of beams used as AI model input;
[0135] · A number of beams used as AI model output;
[0136] o One or more maximum / minimum parameter sets to specify boundaries for AI inference when performing the applicable AI functionalities. Each of such parameter sets may include one or more of:
[0137] · A maximum or minimum measurement interval between two continuous actual measurements;
[0138] · A maximum or minimum prediction interval between two continuous AI predictions;
[0139] · A maximum or minimum number of continuous actual measurements;
[0140] · A maximum or minimum number of continuous AI predictions;
[0141] · A maximum or minimum number of beams used as AI model input;
[0142] · A maximum or minimum number of beams used as AI model output.
[0143] In some example implementations, the measurements or prediction as configured above may be L1 measurements / predictions or L3 measurements / predictions. In some example implementations, the partial configurations above are included in one or more RRC signaling (e.g., otherConfig in the RRCReconfiguration message) .
[0144] In some example implementations of the associated ID above, such associated ID may be configured per cell level, which means that the same ID included for different cell may link to different sets of network additional conditions. In other words, the associated ID is defined separately for each cell.
[0145] For example, as shown in FIG. 15, associated ID 1 in the cell 1 refers to the network additional condition (e.g. beam pattern) 1, while associated ID 1 in the cell 2 refers to the network additional condition 2. The network additional conditions are different in the cell 1 and the cell 2 even though their associated IDs are the same.
[0146] In some example implementations, the associated IDs above may be configured at per area level, which means that associated ID space is defined for each area which may have multiple cells, and for different cells in the same area, the same ID would represent the same corresponding network additional condition, and the same ID may link to different of network additional conditions across cells of different areas. The per area level associated ID means the referred network additional conditions with the same associated ID are the same in different cells that belonging to the same area. An example is shown in FIG. 16. where the network additional conditions in the cell 1 and cell 2 are the same (indicated by the same associated ID in the same area) .
[0147] The area above, may be geographic area, e.g. within one city, or logic area, e.g. at the same frequency or belonging to the same PLMN.
[0148] In some example implementations, the associated IDs may be configured at per scenario level. The per scenario level associated ID implementations refers to the scheme to predict the results with the network additional condition B based on the results with the network additional condition A. For example, associated ID 1 may refer to predict the results with network additional condition 2 based on the network additional conditional 1. Associated id 2 may refer to predict the results with network additional 3 based on the network additional 1, as shown inf FIG. 17. In other words, each association ID represents a linking between one network additional conditional (or set of network additional conditions) to another network additional condition (or another set of network additional conditions) .
[0149] For both three options above (per cell, per area, and per scenario associated IDs) , in some example implementations, network node 1 can send its cell (s) and the corresponding associated ID (s) to network node 2. For example, base station 1 has cell 1, cell 2 and cell 3. The base station 1 can indicate to base station 2 (e.g., neighbor base station) its cell information and the corresponding associated ID.
[0150] In some example implementations, this can be triggered by the network node 1 proactively; or requested by the network node 2 explicitly. In some example implementations, the request may be per cell level (e.g. the request is for each cell) or per network node level (e.g. the request is for all cells within this network node) .
[0151] In some example implementations, when the associated ID of the cell in the base station 1 is changed, the base station 1 can inform the the base station 2 proactively about the updated associated id information and the corresponding cell information. Alternatively, base station 1 may indicate to base station 2 that there is a change, and based on the request from the base station 2 to determine whether report the updated associated ID.
[0152] In some example implementations with DU-CU split in the base station, the DU can indicate to the CU the associated ID value or associated ID change above.
[0153] In some example implementations, one associated IDs may be configured for each measurement object or measurement ID (or at the measurement object level) . In some example implementations, one associated ID may be configured for each cell or cell set such that each entire cell or cell set is linked to a set of network additional conditions with respect to the AI functionalities. In some other example implementations, the cells in such a cell set sharing the same associated ID may relate to one another in certain manners (e.g., by some particular spatial relationship, e.g., neighboring cells, or one associated ID may be configured to a cell set, which means Optionally, one cell set corresponds to one associated id means the associated id indicates the additional condition for each cell in this cell set and their relationship (e.g. spatial relationship, such as neighbor cells, and the like) .
[0154] An example configuration structure of Step 3 above that may be sent from the network to the UE is illustrated in FIG. 7. In the example of FIG. 7, the associated ID is configured at per area level for each measurement object. The cell 1~3 and 7, 9 are within the same area. The cell 1, cell 2 and cell 3 has the same associated id, which refers that their network additional condition are the same; the cell 7and cell 9 has the same associated id, which refers that their network additional condition are the same. The configuration structure of FIG. 7, also specify for the measurement object 1 the applicable or requested AI functionality, and a list of inference parameter sets.
[0155] FIG. 8 shows another example configuration structure of Step 3 above. The example of FIG. 8 configures each AI functionality as to its applicability to measurement objects with other inference parameters. For each measurement object, Each cell corresponds to an associated id
[0156] In some example implementations, the configuration message of Step 3 above may include full inference configurations (rather than partial inference configurations) if the network has sufficient information determine such full inference configurations at the time. In some example implementations, such full inference configurations may include at least one of:
[0157] · One or more measObject / measId for AI predictions. In some example implementations, an indication in the measObject or measurement identity configuration IE can be introduced to indicate that AI predictions are to be performed on this measObject or measId. In another implementation, a new list in a measurement configuration IE, referred to as MeasConfig, may be introduced to include the measObject or measurement identity with predictions.
[0158] · One or more measObject / measId whose results can be used to predict other results. In some examples implementations, an indication in the measObject or measurement identity configuration IE may be introduced to indicate that the measurement results can be used to predict other results. In some other example implementations, a new list in a measurement configuration IE for the measurement objects, referred to as measConfig, may be introduced to include the measObject or measurement identity whose results can be used to predict other results.
[0159] · A mapping relationship between the measObject / measId with predictions above and the measObject / measId whose results can be used to predict other results.
[0160] · A mapping relationship between the cell (s) in the AI model output and the cell (s) in the AI model input (for inter-cell prediction) .
[0161] · A configuration for time instance for measurements or predictions including one or more of:
[0162] o A measurement interval between two continuous actual measurements;
[0163] o A measurement interval between two continuous predictions;
[0164] o A number of continuous actual measurements;
[0165] o A number of continuous AI prediction;
[0166] o A number of beams used as AI model input;
[0167] o A number of beams used as AI model output.
[0168] In some example implementations, the configuration above can be per measObject or per MeasId or per functionality level.
[0169] FIG. 9 illustrates an example configuration structure indicating that results on the measObject 1 is predicted based on the measurement results of the measObject 2.
[0170] In some example implementations, upon receiving only the partial inference configuration above, the UE may not start any inference, but instead, may wait for the full inference configuraiton.
[0171] In some example implementations, upon receiving the full inference configuration, the corresponding AI inference may be implicitly considered as being activated at the UE, e.g., after the UE reports the applicable AI functionalities in Step 4 below.
[0172] In some example implementations, activation of the AI inference at the UE may need to be explicitly indicated by a further activation command from the network. Such activation command, for example, may be provided in RRC or MAC CE or lower layer signaling that is sent by RRC or MAC.
[0173] In some example implementations, the activation command above may be issued at per measurement object (measObject) level or per measId level, or per carrier frequency level, or per cell level.
[0174] In some example implementations, when AI inference starts, the UE may stop actual measurements on corresponding measId / measObject automatically.
[0175] In some example implementations, an indication may be included in the configuration of Step 3 above to indicate whether the UE is allowed to perform inference immediately after it reports applicable AI functionalities in Step 4. In some example implementations, such an indication may be provided in the measObject IE at per measurement object level.
[0176] In some example implementations, the AI inference at the UE may be considered as being activated after one or more pre-defined conditions are met. Such pre-defined conditions may include one or more of:
[0177] · A measurement event is triggered;
[0178] · Measurement results for serving cell (s) or neighbor cell (s) are above or below a threshold; or
[0179] · The UE is near a cell edge or at the cell center.
[0180] In some example implementations, the pre-defined conditions above are provided at per measId or per measObject level.
[0181] If no inference configuration or the partial inference configuration and / or full partial inference configuration is configured in Step 3 above (rather than full inference configuration) , the reporting in Step 4 by the UE may include one or more of:
[0182] · one or more measId / measObjectId or carrier frequency that the AI inference is applicable. Optionally, the UE may report one or more measId / measObjectId or carrier frequency that a cell level inference is applicable for. In some example implementations, the UE may report one or more measId / measObjectId or carrier frequency that a beam level inference is applicable for.
[0183] · one or more cell (s) that the AI inference is applicable. In some example implementations, the UE may report one or more cell (s) that the cell level AI inference is applicable. In some example implementations, the UE may report one or more cell (s) that the beam level inference is applicable.
[0184] · one or more measId / measObjectId or carrier frequency that the AI inference is non-applicable for. In some example implementations, the UE may report one or more measId / measObjectId or carrier frequency that the cell level inference is non-applicable. In some example implementations, the UE may report one or more measIds / measObjectIds or carrier frequencies that the beam level inference is non-applicable. In some example implementations, for each non-applicable measId / measObjectId or carrier frequency, the UE may report one cause value (e.g., for indicating that the AI model is not available, or that the inference is of low accuracy) .
[0185] · one or more cell (s) that the inference is non-applicable. In some example implementations, the UE may report one or more cell (s) that the cell level inference is non-applicable. In some example implementations, the UE may report one or more cell (s) that the beam level inference is non-applicable. In some example implementations, for each non-applicable cell, the UE may report one cause value (e.g., for indicating that the AI model is not available, or that the inference is of low accuracy) .
[0186] · one or more applicable inference parameter set (that comply with the above Step 3 parameter from network) . In some example implementations, the applicable inference parameter set can be indicated via the index in the configured inference parameter set list. In some example implementations, the UE may explicitly report the one or more applicable inference parameter sets, which may include one or more of:
[0187] o A measurement interval between two continuous actual measurements;
[0188] o A measurement interval between two continuous AI predictions;
[0189] o A number of continuous actual measurements;
[0190] o A number of continuous AI predictions;
[0191] o A number of beams used as AI model input;
[0192] o A number of beams used as AI model output; or
[0193] o One or more supported associated IDs.
[0194] In some example implementations, the applicable functionality reporting above in Step 4 of FIG. 5, if needed, can be sent via RRC signaling (e.g., RRCReconfigurationComplete or UAI)
[0195] For Step 5 of FIG. 5 above, the inference configuration to be sent from the network to the UE is similar as in step 3. If no inference configuration or only partial inference configuration was sent in Step 3 from the network to the UE, then Step 5 may be needed in response to the applicable functionality reporting of Step 4 of FIG. 5. Such inference configuration, if still needed, may include similar information items and configuration as described above for Step 3 and Step 4.
[0196] For Step 6 of FIG. 5 above, in some example implementations for inference deactivation, the deactivation command may include or indicate one or more measId or measObjectId or carrier frequency or cell id that the inference is deactivated. In other words, the deactivation may be selective and directed to particular measurement objects or at other levels.
[0197] In some example implementations, the inference deactivation command can be RRC signaling or MAC CE or DCI or lower layer signaling that is sent by RRC or MAC.
[0198] In some example implementations, the deactivation for cell level prediction and beam level prediction may be separately. In one example, one indication is used to indicate whether the deactivation command is for beam level prediction or cell level prediction.
[0199] In some example implementations, the UE may deactivate the AI inference on one measObject or measId, and may then resume normal measurements on this measObject or measId.
[0200] For Step 6 of FIG. 5 above, in some example implementations for inference activation, the activation command from the network may include or indicate one or more measId or measObjectId or carrier frequency or cell id that the inference is to be activated.
[0201] In some example implementations, the activation command may be provided via RRC signaling or MAC CE or DCI or lower layer signaling that sent RRC or MAC.
[0202] In some example implementations, the activation for cell level AI prediction and beam level AI prediction may be separately provided. In one example, one indication may be used to indicate whether the activation command is for beam level prediction or cell level prediction.
[0203] In some example implementations, the UE may activate the AI inference on one measObject or measId. The UE autonomously stop normal measurement on this measObject or measId after inference activation.
[0204] Model monitoring
[0205] FIG. 11 illustrates example steps related to monitoring the performance of AI functionalities described above, including the following.
[0206] Step 1: The network sends monitoring configurations to the UE (this step may be optional) .
[0207] Step 2: The UE performs prediction performance monitoring based on the network configuration. For example, a KPI (e.g., RSRP difference between predicted result and monitoring result) may be performed.
[0208] Step 3: The UE sends a monitoring report to the network.
[0209] For step 1 above, the monitoring configuration sent to the UE, by the network may include one or more of:
[0210] · A monitoring periodicity for the prediction performance;
[0211] · A time offset between two measurements / prediction for monitoring within a monitoring periodicity; or
[0212] · A number of continuous measurements / predictions for monitoring within a monitoring periodicity.
[0213] An example is illustrated in FIG. 11, where the open and shaded squares in the first row indicates time instances of measurements and AI predictions of a measurement object, respectively. For performance monitoring purposes, a number of continuous measurements are made (second row) in each monitoring period corresponding to the predictions in order to determine the performance of the predictions. FIG. 11 illustrates a configured monitoring periodicity, and number of continuous measurements for monitoring, and a time gap between the number of continuous monitoring measurements.
[0214] In some example implementations, the monitoring requirements or monitoring configuration above may be provided at per measId or measObject level, or other levels.
[0215] In some example implementations, the monitoring configuration and AI inference configuration are included in the measurement identity or measurement object configuration. In other words, the monitoring configuration and the inference configuration may be provided in a same configuration structure, as an example.
[0216] In some example implementations, the monitoring configuration and inference configuration may be included in the separate measurement identity or measurement object. In that case, the mapping between the measId / measObejct for inference and measId / measObject for performance monitoring measurement may be configured by the network. In some example implementations, the possible schemes for such mapping may include one of the followings:
[0217] · An ID may be included for the measurement identity for inference to indicate the corresponding measurement identity for monitoring;
[0218] · An ID may be included for in the measurement identity for monitoring to indicate the measurement identity for inference;
[0219] · A list may be introduced to explicit indicate the mapping relationship between measurement object for inference and measurement object for measurement monitoring.
[0220] The mapping above may be at per UE level, or per AI functionality level. In some example implementations, the KPI calculation may be performed by the network using reported monitoring information from the UE. Alternatively, the KPI calculation for the performance of AI functionalities may be performed by the UE and then reported to the network.
[0221] For Step 3 of FIG. 10, the monitoring reporting may include prediction results and actual measurement results within one monitoring period for the network to compare to calculate the KPI. In some example implementations, the predicted measurement results and monitoring results are included in one report or a single report. In some example implementations, the predicted measurement results and monitoring results may be included in the separate reports. In some example implementations, a mapping relationship between the report for prediction and the report for monitoring may be configured by the network. In some example implementations, a key may be used for such mapping. If the key in the report for prediction and the key in the report for monitoring results are the same or matches, they are considered as being related and mapped.
[0222] For Example, when the UE reports the prediction results and monitoring results for the first time, the key for the separate reports may be set to 1, and then each time the prediction results and monitoring results are reported for monitoring purposes, the key may be incremented to key+1.
[0223] In some example implementations, the UE may report the monitoring results and prediction results for all the cells / beams for prediction together in one single report or in two separate reports.
[0224] In some example implementations, the UE may report monitoring results and predictions results for one or more of:
[0225] 1) Serving cell (s) + top K neighbor cell, K being a positive integer that may be predefined or configured by the network;
[0226] 2) Top K cells, K being a positive integer that may be predefined or configured by the network;
[0227] 3) Cells indicated by the network;
[0228] 4) Cells whose results are above a threshold, the threshold may be predefined or configured by the network, or
[0229] 5) Top-K beams results (predicted and monitored) for each reported cell,
[0230] For (1) , (2) and (5) above:
[0231] · the UE may report the cell / beam results for the top-K cells based on the predicted results.
[0232] · The UE reports the cell / beam results for the top-K cells based on the monitoring results.
[0233] · For predicted results, the UE reports the cell / beam results for the top-K cells based on the predicted results, and for monitoring results, the UE reports the cell / beam results for top-K cells based on the monitoring results.
[0234] · The UE reports the cell results for top-K cells based on the predicted results U top-K cells based on the monitoring results (U represents union here) .
[0235] For (4) above:
[0236] · The UE reports the cell results whose predicted results are above a threshold;
[0237] · The UE reports the cell results whose monitoring results are above a threshold;
[0238] · For predicted results, the UE reports the cell results for the cell whose predicted results are above a threshold; For monitoring results, the UE reports the cell whose monitoring results are above a threshold.
[0239] · The UE reports the results whose predicted results or monitoring results are above a threshold.
[0240] In some example implementations, the monitoring results may be reported via RRC signaling (e.g., MeasurementReport or UAI (UE Assistant Information) or a new RRC message) or MAC CE or lower layer signaling that sent by RRC or MAC.
[0241] An example monitoring report structure is shown in FIG. 12. In the example report of FIG. 12, each result (e.g., measurement object) is reported at cell level, where for each cell, prediction result and corresponding measurements result are reported.
[0242] In some example implementations, the above monitoring is first activated upon receiving monitoring configuration. In some example implementations, activation / deactivation commands may be used and may include or indicate one or more measId or measObjectId or carrier frequency or cell id that the monitoring is activated / deactivated.
[0243] In some example implementations, the activation / deactivation command may be transmitted from the network via RRC signaling or MAC CE or DCI or lower layer signaling that sent by RRC or MAC.
[0244] In some example implementations, the activation / deactivation for cell level prediction and beam level prediction monitoring may be provided separately.
[0245] Training data collection for UE-Side AI Models
[0246] The UE-side AI model training data configuration and collection procedure shown in FIG. 6 above is described in further detail below.
[0247] In Step 1 of FIG. 6, the network configuration for training data collection may include or indicate one or more of:
[0248] · An indication to indicate whether the UE is allowed to initiate the request for data collection.
[0249] · One or more AI feature the UE is allowed to initiate the data collection request for.
[0250] · One or more functionalities (e.g. L3 cell level temporal domain prediction) the UE is allowed to initiate the data collection request for.
[0251] · One or more measId / measObjectId / carrier frequency / cell (s) that the UE is allowed to initiate the data collection request for.
[0252] · One or more associated IDs that the UE is allowed to indicate the data collection request for and optionally the corresponding cell information.
[0253] · SSB periodicity for the serving cell and / or neighbor cell (s) .
[0254] · A minimum time gap for continuous measurements.
[0255] · Supported time gap between measurements on the frequency for prediction and measurements on the frequency for actual measurement.
[0256] · A relationship (e.g., spatial relationship such as co-located or non-located) between two frequencies.
[0257] In some example implementations, one or more of the information items above may be carried via RRC signaling (e.g. Otherconfig or measConfig) .
[0258] In Step 2 of FIG. 6, the content of the request for data collection configuration may include one or more of:
[0259] · a list of features / functionalities that the training data collection is for. For each feature / functionality, the UE may report more detailed information of the AI model to be trained to the network, including one or more of:
[0260] o Prediction window length;
[0261] o The number of predicted results within the prediction window;
[0262] o Time gap between two predicted results;
[0263] o Observation window length;
[0264] o The number of actual results within the observation window;
[0265] o Time gap between two actual results;
[0266] o Model type (cell-based approach or cluster-based approach) .
[0267] · a list of associated IDs that the training data collection is for
[0268] · requested / expected measurement configuration, which may include one or more of:
[0269] o reference signaling (e.g. SSB, CSI-RS) ;
[0270] o the L1 / L3 measurement interval;
[0271] o a list of frequency that the UE needs to measure;
[0272] o the frequency and time configurations for the measurement resource;
[0273] o measurement gap configuration.
[0274] In some example implementations, the request above in Step 2 of FIG. 6 may be sent to the network via RRC signaling (e.g. UAI) .
[0275] In Step 3 of FIG. 6, the training data collection configuration may include or indicate at least one of:
[0276] · One or more MO / MeasId / cell that the UE is allowed to collect the training data.
[0277] · One or more associated id for the measId / measObect / cell (s) / beam (s) .
[0278] · One or more info for AI model output and / or AI model input suggested by the network including one or more of:
[0279] o The prediction window length;
[0280] o The time gap between two continuous predicted results;
[0281] o The number of continuous predicted results;
[0282] o The observation window length;
[0283] o The time gap between two actual measurement;
[0284] o The number of actual results within the observation window.
[0285] · An indicator to indicate whether to report the corresponding measurement results to the network.
[0286] In Step 4-6 of FIG. 6, the start / stop of the training data collection may be performed.
[0287] In some example implementations, the training data collection configuration may be first activated upon receiving training data collection configuration.
[0288] In some example implementations, the start / stop command may include or indicate one or more measId or measObjectId or carrier frequency or cell ID that the training data collection starts on.
[0289] In some example implementations, the start / stop command may be transmitted from the network via RRC signaling or MAC CE or DCI or lower layer signaling that sent by RRC or MAC.
[0290] In some example implementations, the start / stop for cell level prediction and beam level prediction may be provided separately.
[0291] Network Side Model
[0292] Training:
[0293] FIG. 13 illustrates a general procedure for configuring and training network-side models, including the following general steps.
[0294] In Step 1, the network configures the training data collection for AI-mobility, which may include one or more the followings:
[0295] · Measurement quantity (RSRP / RSRQ / SINR) and / or result type (L3 filtered cell results, L3 unfiltered cell results, L1 beam result, and the like) .
[0296] · A time offset between the measurement on the frequency 1 and the measurement on the frequency 2. In some example implementations, the UE may log the measurement results on the frequency 1 and the frequency 2 satisfying the requirement of the time offset or the measurement time are close. In some example implementations, the frequency 1 and frequency 2 may be indicated or configured by the network. For example, the time offset may be configured as 40 ms.The UE may perform measurement on the frequency 1 at t=0, 200ms, 400ms and 600ms, and the UE may perform measurement on the frequency 2 at t =40ms, 440ms. In this case, the UE logs the measurement results at frequency 1 at t=0 and 400ms and the UE logs the measurement results at frequency 1 at t=40 and 440ms. In some example implementations, the measurement can be L1 measurement or L3 measurement.
[0297] · One or more measurement identity / measObject / cell that the network collects the data on.
[0298] In some example implementations, the training data collection are configured via RRC signaling (e.g. LoggedMeasurementConfiguration or measConfig) in Step 1 above for FIG. 13.
[0299] In Step 2, which may be optional, the UE performs the training data collection upon receiving start command or start indication. The start command can be implicitly. For example, upon receiving measurement configuration for training data collection, the UE may start data collection immediately. For another example, an explicitly indication in the training data configuration can be used to indicate whether the UE shall start data collection immediately. For another example, the start command may be an explicit command from the network. The explicit start command may be used to reconfigure the explicit indication in the configuraiton above. The explicit start command may be provided from, the network via some other RRC signaling (e.g. otherConfig) or the like.
[0300] In some example implementations, the start indication above can be per measObject level per measId level or per cell level or per UE level.
[0301] In some example implementations, the start indication above can be indicated via RRC signaling or MAC CE or lower layer signaling that sent by RRC or MAC.
[0302] In some example implementations, for monitoring performance of network side AI models, measurements may be reported from the UE to the network in order for the network to monitor the performance of the network-side AI models.
[0303] Optionally, when UE reporting measurement results, the UE also report the corresponding time stamp for each measurement result in the measurement report.
[0304] Inference or Monitoring
[0305] In some example implementations, a measurement configuration for inference or monitoring may include or indicate one or more of the following information:
[0306] · Measurement quantity (RSRP / RSRQ / SINR) and / or the result type (L3 filtered cell results, L3 unfiltered cell results, L1 beam result, and the like) .
[0307] · Expected time gap between two continuous measurements and an optional tolerance.
[0308] · Expected time gap between two non-continuous measurement, as shown by gap 2 in the example of FIG. 14 (where open squares represent measurements, and shaded squares represent predictions) .
[0309] · Measurement periodicity.
[0310] · Reporting interval.
[0311] · Number of historical cell results within one measurement report.
[0312] · Number of historical time instance within one measurement report. For temporal domain prediction case A, the number may be 1. For temporal domain prediction, the number may be configured by the network.
[0313] Optionally, when UE reporting measurement results, the UE also report the corresponding time info for each measurement result in the measurement report.
[0314] In some example implementations, the measurement report may be upon the expiry of the report interval above, or upon obtaining the number of continuous measurement results specified above. In some example implementations, if some measurements are missed, the UE may resume to perform continuous measurement.
[0315] The description and accompanying drawings above provide specific example embodiments and implementations. The described subject matter may, however, be embodied in a variety of different forms and, therefore, covered or claimed subject matter is intended to be construed as not being limited to any example embodiments set forth herein. A reasonably broad scope for claimed or covered subject matter is intended. Among other things, for example, subject matter may be embodied as methods, devices, components, systems, or non-transitory computer-readable media for storing computer codes. Accordingly, embodiments may, for example, take the form of hardware, software, firmware, storage media or any combination thereof. For example, the method embodiments described above may be implemented by components, devices, or systems including memory and processors by executing computer codes stored in the memory.
[0316] Throughout the specification and claims, terms may have nuanced meanings suggested or implied in context beyond an explicitly stated meaning. Likewise, the phrase “in one embodiment / implementation” as used herein does not necessarily refer to the same embodiment and the phrase “in another embodiment / implementation” as used herein does not necessarily refer to a different embodiment. It is intended, for example, that claimed subject matter includes combinations of example embodiments in whole or in part.
[0317] Reference throughout this specification to features, advantages, or similar language does not imply that all of the features and advantages that may be realized with the present solution should be or are included in any single implementation thereof. Rather, language referring to the features and advantages is understood to mean that a specific feature, advantage, or characteristic described in connection with an embodiment is included in at least one embodiment of the present solution. Thus, discussions of the features and advantages, and similar language, throughout the specification may, but do not necessarily, refer to the same embodiment.
[0318] Furthermore, the described features, advantages and characteristics of the present solution may be combined in any suitable manner in one or more embodiments. One of ordinary skill in the relevant art will recognize, in light of the description herein, that the present solution can be practiced without one or more of the specific features or advantages of a particular embodiment. In other instances, additional features and advantages may be recognized in certain embodiments that may not be present in all embodiments of the present solution.
Claims
1.A method performed by a User Equipment (UE) in communication with a network node of a wireless communication network, comprising:receiving, from the network node, an AI configuration;determining applicable AI functionalities according to at least the AI configuration ;reporting information associated with the applicable AI functionality to the network node;further receiving, from the network node, a response containing an AI inference configuration unless a full AI inference has already been included in the AI configuration; andactivating AI inference and / or monitoring of a performance of the applicable AI functionalities according to the AI configuration and the AI inference configuration.2.The method of claim 1, further comprising:receiving a capability inquiry with respect to UE side Artificial Intelligence (AI) functionalities from the network node;transmitting the initial AI capability information to the network node in response to the capability inquiry to prompt the network node to transmit the AI configuration.3.The method of claim 2, wherein the capability inquiry comprises a filter for the UE to selectively transmit the initial AI capability information in an adaptive manner.4.The method of claim 2, wherein the capability inquiry comprises an indication of whether to report AI-related UE capabilities by the UE.5.The method of claim 2, wherein the initial AI capability information is provided as UE AI capabilities for each supported frequency, frequency band, or frequency band combination, and comprises at least a maximum number of cells that the UE is capable of performing AI prediction on.6.The method of claim 2, wherein the initial AI capability information comprises, for a supported AI functionality, supported measurement configuration / parameter values or value ranges including at least one of:a list of supported Measurement Reduction in Time-domain (Rate MRRT) or measurement Reduction Rate in Spatial-domain (MRRS) ;a maximum or minimum value of supported MRRT / MRRS;a list of supported time offsets between measurements and corresponding predictions in a training data collection procedure for the supported AI functionality;a maximum or minimum value of the supported time offsets; ora maximum number of cells that the UE can perform prediction on; ora list of required Signal / PBCH Block (SSB) periodicities for measurements and / or predictions; ora maximum OR minimum number of required SSB periodicities for measurements and / or predictions.7.The method of claim 2, wherein the initial AI capability information comprises a maximum number of cells that the UE can perform prediction on per measurement object, per measurement ID, per frequency / frequency band / band combination, or for the UE.8.The method of claim 2, wherein the AI configuration comprises an indication of whether to enable beam level prediction by the UE.9.The method of claim 2, wherein the AI configuration comprises a partial AI inference configuration comprising or indicating at least one of:one or more measurement objects, carrier frequencies, or cells that the UE is allowed or requested to perform AI inference on;one or more AI functionalities that the UE is allowed or requested to perform;one or more measurement objects or carrier frequencies that can be used as input to the applicable AI functionalities;one or more associated IDs for indicating a set of network additional conditions; orone or more AI inference parameter sets for measurements and predictions.10.The method of claim 9, wherein each of the one or more AI inference parameter sets comprises at least one of:a first measurement interval between two continuous actual measurements;a second prediction interval between two continuous AI prediction;a first number of continuous actual measurements;a second number of continuous AI predictions;a third number of beams used as input of the applicable AI functionalities; ora fourth number of beams used as output of the applicable AI functionalities.11.The method of claim 9, wherein the one or more AI inference parameter sets comprises a maximum / minimum parameter set containing maximum or minimum values of a plurality of inference parameters comprising at least one of:a first maximum or minimum measurement interval between two continuous actual measurements;a second maximum or minimum measurement interval between two continuous AI predictions;a third maximum or minimum number of continuous actual measurements;a fourth maximum or minimum number of continuous AI predictions;a fifth maximum or minimum number of beams used as input to the applicable AI functionalities; ora sixth maximum or minimum number of beams used as output to the applicable AI functionalities.12.The method of claim 9, wherein the one or more associated IDs are specified at:per cell level, indicating that network additional conditions corresponding to a same associated ID need not be the same in different cells;per area level, indicating that network additional conditions corresponding to a same associated ID are the same in different cells of a same area; orper prediction scenario level, wherein a particular associated ID indicates usage of one set of network additional conditions as a basis to predict cell results with another set of network additional conditions.13.The method of claim 9, wherein:each measurement object or frequency carrier corresponds to one of the one or more associated IDs; oreach cell or cell set corresponds to one of the one or more associated IDs.14.The method of claim 13, wherein an associated ID corresponding to cells in each cell set indicates network additional conditions of each cell and a spatial relationship between the cells.15.The method of claim 2, wherein the response comprises a time instance configuration information for measurements or AI predictions, the time instance configuration information comprising at least one of:a first measurement interval between two continuous actual measurements;a second measurement interval between two continuous AI predictions;a first number of continuous actual measurements;a second number of continuous AI predictions;a third number of beams used as input to the applicable AI functionalities; ora fourth number of beams used as output of the applicable AI functionalities.16.The method of claim 2, wherein the AI inference and / or monitoring is activated by and explicit activation command received from the network node.17.The method of claim 2, wherein the AI inference and / or monitoring is activated by conditions associated with one or more predefined measurement events being met.18.The method of claim 2, wherein the reported information associated with the applicable AI functionalities comprises at least one of:one or more IDs of measurement objects or carrier frequencies that the AI inference is applicable;one or more cell IDs that the AI inference is applicable;one or more IDs of measurement objects or carrier frequencies that the AI inference is not applicable;one or more cell IDs that the AI inference is not applicable;one or more applicable AI inference parameter sets; orone or more supported associated IDs indicating network additional conditions.19.The method of claim 2, further comprising receiving a command from the network node for deactivating the AI inference and / or monitoring of the performance of the applicable AI functionalities with respect to one or more measurement objects, carrier frequencies, or cells, the command comprising IDs of the one or more measurement objects, carrier frequencies, or cells to be deactivated.20.The method of claim 19, wherein cell level AI inferences and beam level deactivations are triggered by different deactivation commands.21.The method of claim 2, wherein the AI configuration or the response comprises a monitoring configuration for the monitoring of the performance of the applicable AI functionalities.22.The method of claim 21, wherein the monitoring of the performance of the applicable AI functionalities is configured at one of the following levels:measurement object level;UE level; orAI functionality level.23.The method of claim 21, wherein the AI inference configuration and the monitoring configuration are included with different measurement objects for inference and measurement objects for measurements, respectively, and a mapping between the measurement objects for inference and the measurement objects for measurements are mapped by the network node.24.The method of claim 21, further comprising generating a single report of the monitoring of the performance of the applicable AI functionalities, the single report comprising both of prediction results and corresponding actual measurement results.25.The method of claim 21, further comprising generating separate reports of the monitoring of the performance of the applicable AI functionalities separately comprising prediction results and corresponding actual measurement results within one monitoring periodicity, wherein a key is configured to correlate the separate reports.26.The method of claim 21, further comprising generating a monitoring report of the performance of the applicable AI functionality, the monitoring report indicating monitoring results for at least one of:serving cells and top K neighbor cells;top K cells;cells indicated by the wireless communication network;cells whose monitoring results are above a threshold configured by the wireless communication network; ortop K beams for each monitored cell,wherein K is a positive integer configured by the wireless communication network.27.The UE comprising a memory for storing instructions and at least one processor for executing the instructions to perform the method of any one of claims 1 to 26.28.A non-transitory computer-readable storage medium for storing instructions, the instructions, when executed by at least one processor, are configured to cause the at least one processor to perform the method of any one of claims 1 to 26.29.A method performed by a network node in communication with a User Equipment (UE) in a wireless communication network, comprising:transmitting, to the UE, an AI configuration;receiving reporting information associated with the applicable AI functionality as determined by the UE according to at least the AI configuration;transmitting, to the UE, a response containing an AI inference configuration unless a full AI inference has already been included in the AI configuration; andactivating AI inference and / or monitoring of a performance of the applicable AI functionalities at the UE according to the AI configuration and the AI inference configuration.30.The method of claim 29, further comprising:transmitting a capability inquiry with respect to UE side Artificial Intelligence (AI) functionalities to the UE;receiving the initial AI capability information from the UE in response to the capability inquiry to prompt the network node to transmit the AI configuration.31.The method of claim 20, wherein the capability inquiry comprises a filter for the UE to selectively transmit the initial AI capability information in an adaptive manner.32.The method of claim 30, wherein the capability inquiry comprises an indication of whether to report AI-related UE capabilities by the UE.33.The method of claim 30, wherein the initial AI capability information is provided as UE AI capabilities for each supported frequency, frequency band, or frequency band combination, and comprises at least a maximum number of cells that the UE is capable of performing AI prediction on.34.The method of claim 30, wherein the initial AI capability information comprises, for a supported AI functionality, supported measurement configuration / parameter values or value ranges including at least one of:a list of supported Measurement Reduction in Time-domain (Rate MRRT) or measurement Reduction Rate in Spatial-domain (MRRS) ;a maximum or minimum value of supported MRRT / MRRS;a list of supported time offsets between measurements and corresponding predictions in a training data collection procedure for the supported AI functionality;a maximum or minimum value of the supported time offsets; ora maximum number of cells that the UE can perform prediction on; ora list of required Signal / PBCH Block (SSB) periodicities for measurements and / or predictions; ora maximum OR minimum number of required SSB periodicities for measurements and / or predictions.35.The method of claim 30, wherein the initial AI capability information comprises a maximum number of cells that the UE can perform prediction on per measurement object, per measurement ID, per frequency / frequency band / band combination, or for the UE.36.The method of claim 30, wherein the AI configuration comprises an indication of whether to enable beam level prediction by the UE.37.The method of claim 30, wherein the AI configuration comprises a partial AI inference configuration comprising or indicating at least one of:one or more measurement objects, carrier frequencies, or cells that the UE is allowed or requested to perform AI inference on;one or more AI functionalities that the UE is allowed or requested to perform;one or more measurement objects or carrier frequencies that can be used as input to the applicable AI functionalities;one or more associated IDs for indicating a set of network additional conditions; orone or more AI inference parameter sets for measurements and predictions.38.The method of claim 37, wherein each of the one or more AI inference parameter sets comprises at least one of:a first measurement interval between two continuous actual measurements;a second prediction interval between two continuous AI prediction;a first number of continuous actual measurements;a second number of continuous AI predictions;a third number of beams used as input of the applicable AI functionalities; ora fourth number of beams used as output of the applicable AI functionalities.39.The method of claim 37, wherein the one or more AI inference parameter sets comprises a maximum / minimum parameter set containing maximum or minimum values of a plurality of inference parameters comprising at least one of:a first maximum or minimum measurement interval between two continuous actual measurements;a second maximum or minimum measurement interval between two continuous AI predictions;a third maximum or minimum number of continuous actual measurements;a fourth maximum or minimum number of continuous AI predictions;a fifth maximum or minimum number of beams used as input to the applicable AI functionalities; ora sixth maximum or minimum number of beams used as output to the applicable AI functionalities.40.The method of claim 37, wherein the one or more associated IDs are specified at:per cell level, indicating that network additional conditions corresponding to a same associated ID need not be the same in different cells;per area level, indicating that network additional conditions corresponding to a same associated ID are the same in different cells of a same area; orper prediction scenario level, wherein a particular associated ID indicates usage of one set of network additional conditions as a basis to predict cell results with another set of network additional conditions.41.The method of claim 37, wherein:each measurement object or frequency carrier corresponds to one of the one or more associated IDs; oreach cell or cell set corresponds to one of the one or more associated IDs.42.The method of claim 41, wherein an associated ID corresponding to cells in each cell set indicates network additional conditions of each cell and a spatial relationship between the cells.43.The method of claim 30, wherein the response comprises a time instance configuration information for measurements or AI predictions, the time instance configuration information comprising at least one of:a first measurement interval between two continuous actual measurements;a second measurement interval between two continuous AI predictions;a first number of continuous actual measurements;a second number of continuous AI predictions;a third number of beams used as input to the applicable AI functionalities; ora fourth number of beams used as output of the applicable AI functionalities.44.The method of claim 30, wherein the AI inference and / or monitoring is activated by and explicit activation command received from the network node.45.The method of claim 30, wherein the AI inference and / or monitoring is activated by conditions associated with one or more predefined measurement events being met.46.The method of claim 30, wherein the reported information associated with the applicable AI functionalities comprises at least one of:one or more IDs of measurement objects or carrier frequencies that the AI inference is applicable;one or more cell IDs that the AI inference is applicable;one or more IDs of measurement objects or carrier frequencies that the AI inference is not applicable;one or more cell IDs that the AI inference is not applicable;one or more applicable AI inference parameter sets; orone or more supported associated IDs indicating network additional conditions.47.The method of claim 40, further comprising transmitting a command to the UE for deactivating the AI inference and / or monitoring of the performance of the applicable AI functionalities with respect to one or more measurement objects, carrier frequencies, or cells, the command comprising IDs of the one or more measurement objects, carrier frequencies, or cells to be deactivated.48.The method of claim 47, wherein cell level AI inferences and beam level deactivations are triggered by different deactivation commands.49.The method of claim 40, wherein the AI configuration or the response comprises a monitoring configuration for the monitoring of the performance of the applicable AI functionalities.50.The method of claim 49, wherein the monitoring of the performance of the applicable AI functionalities is configured at one of the following levels:measurement object level;UE level; orAI functionality level.51.The method of claim 49, wherein the AI inference configuration and the monitoring configuration are included with different measurement objects for inference and measurement objects for measurements, respectively, and a mapping between the measurement objects for inference and the measurement objects for measurements are mapped by the network node.52.The method of claim 49, further comprising generating a single report of the monitoring of the performance of the applicable AI functionalities, the single report comprising both of prediction results and corresponding actual measurement results.53.The method of claim 49, further comprising receiving separate reports of the monitoring of the performance of the applicable AI functionalities separately comprising prediction results and corresponding actual measurement results within one monitoring periodicity, wherein a key is configured to correlate the separate reports.54.The method of claim 49, further comprising receiving a monitoring report of the performance of the applicable AI functionality, the monitoring report indicating monitoring results for at least one of:serving cells and top K neighbor cells;top K cells;cells indicated by the wireless communication network;cells whose monitoring results are above a threshold configured by the wireless communication network; ortop K beams for each monitored cell,wherein K is a positive integer configured by the wireless communication network.55.The network node comprising a memory for storing instructions and at least one processor for executing the instructions to perform the method of any one of claims 29 to 54.56.A non-transitory computer-readable storage medium for storing instructions, the instructions, when executed by at least one processor, are configured to cause the at least one processor to perform the method of any one of claims 29 to 54.