Criteria for UE reporting performance monitoring result for CSI prediction without ground-truth measurements

Abstract

A method, system and apparatus are disclosed. A method in a user equipment configured to communicate with a network node is described. The method includes generating predicted channel state information (CSI) samples via a trained artificial intelligence (AI) CSI prediction model, the AI CSI prediction model being trained based at least in part on a first training data set of samples of CSI measurements in a prediction window and transmitting a CSI performance monitoring (PM) report of CSI-PM results based on a performance metric estimator.

Description

[0001] CRITERIA FOR UE REPORTING PERFORMANCE MONITORING RESULT FOR CSI PREDICTION WITHOUT GROUND-TRUTH MEASUREMENTS

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to a user equipment (UE) and a method in a UE for generating predicted channel state information. It also relates to a network node and a method in a network node for transmitting a channel state information performance monitoring (CSI-PM) configuration.

[0004] BACKGROUND

[0005] The Third Generation Partnership Project (3GPP) has developed and is developing standards for Fourth Generation (4G) (also referred to as Long Term Evolution (LTE)) and Fifth Generation (5G) (also referred to as New Radio (NR)) wireless communication systems. Such systems provide, among other features, broadband communication between network nodes, such as base stations and mobile UEs, as well as communication between network nodes and between UEs. The 3GPP is also developing standards for Sixth Generation (6G) wireless communication networks.

[0006] CSI reporting in NR

[0007] In NR, a UE may be configured with one or multiple Channel State Information (CSI) Report Settings, each configured by a higher layer parameter CSI-ReportConfig. Each CSI-ReportConfig is associated with a Bandwidth Part (BWP) and includes one or more of the following:

[0008] • a CSI resource configuration for channel measurement;

[0009] • a CSI Interference Measurement (CSI-IM) resource configuration for interference measurement;

[0010] • reporting configuration type, i.e., aperiodic CSI (on Physical Uplink Shared Channel (PUSCH)), periodic CSI (on Physical Uplink Control Channel (PUCCH)), or semi-persistent CSI on PUCCH or PUSCH;

[0011] • report quantity specifying what is to be reported, such as Rank Indicator (RI), Precoding Matrix Indicator (PMI), Channel Quality Indicator (CQI);

[0012] • codebook configuration such as type I or type II CSI;

[0013] • frequency domain configuration, i.e., subband vs. wideband CQI or PMI, and subband size; and / or

[0014] • CQI table to be used.A UE may be configured with one or multiple CSI resource configurations for channel measurement and one or more CSI-IM resources for interference measurement. Each CSI resource configuration for channel measurement may include one or more Non Zero Power CSI Reference Signal (NZP CSI-RS) resource sets. For each NZP CSI-RS resource set, the CSI resource configuration may further include one or more NZP CSI-RS resources. A NZP CSI-RS resource may be periodic, semi-persistent, or aperiodic.

[0015] Similarly, each CSI-IM resource configuration for interference measurement may include one or more CSI-IM resource sets. For each CSI-IM resource set, the CSI-IM resource configuration may further include one or more CSI-IM resources. A CSI-IM resource may be periodic, semi-persistent, or aperiodic.

[0016] CSI reporting types and CSI-RS configuration types

[0017] In Table 1 below, a summary is provided for the CSI reporting types and CSI-RS configuration types supported in NR.

[0018] Table 1. The CSI reporting types and CSI-RS configuration types supported in NR CSI-RS Periodic CSI Semi-Persistent Aperiodic CSI Configuration Reporting CSI Reporting Reporting Periodic CSI-RS No dynamic For reporting on Triggered by DCI;

[0019] triggering / acti vati on PUCCH, the UE additionally,

[0020] receives an subselection activation indication is command; supported.

[0021] For reporting on

[0022] PUSCH, the UE

[0023] receives triggering

[0024] on Downlink

[0025] Control

[0026] Information (DCI)

[0027] Semi-Persistent Not Supported For reporting on Triggered by DCI; CSI-RS PUCCH, the UE additionally,

[0028] receives an subselection activation indication is command; supported.

[0029] For reporting on

[0030] PUSCH, the UE

[0031]

[0032] receives triggering

[0033] on DCI

[0034] Aperiodic CSI-RS Not Supported Not Supported Triggered by DCI;

[0035] additionally, subselection indication is supported.

[0036] Physical channel to Periodic CSI report Semi-Persistent Aperiodic CSI carry the CSI is carried on CSI report is report is carried report PUCCH carried on PUSCH on PUSCH

[0037] or PUCCH

[0038]

[0039] 3GPP NR Rel-18 time domain Type II CSI prediction at UE

[0040] In 3GPP NR Release 18 (3GPP Rel-18), channel measurement resource (CMR) enhancement for Type II CSI prediction at UE (i.e., Enhanced Type II predicted PMI) has been introduced, e.g., the measurement part shown in FIG. 1, where a burst of K E {4, 8, 12} same CSI-RS resources are configured to the UE in a single CSI-RS resource set. The burst of CSI-RS resources is aperiodically (AP) triggered using a single downlink control information (DCI). The K CSI-RS resources are used for the UE to extract time domain channel properties of the channel, based on which a future CSI may be predicted. Alternatively, the network may also configure a legacy periodic (P) or semi-persistent (SP) CSI-RS resource. The CSI-RS resources are uniformly spaced in time, separated by m E {1, 2} slots, within the resource set.

[0041] For the 3GPP Rel-18 Type II predicted PMI enhancement, a UE may be configured by gNB to report predicted PMIs for 1V4E {1, 2, 4, 8} time slots, see the 3GPP Rel-18 Type II PMI part in FIG. 1. The predicted 1V4PMIs are supposed to reflect the channels with d E {l,m} slots separation, starting from 8 E {—nCSI, 0,1,2} slots into the future relative to the UL slot in which the predicted CSI is reported. For an AP CSI-RS burst, m E {1,2}, while for P / SP CSI-RS, m is the CSI-RS periodicity. The spacing d between the 1V4PMIs and offset 6 relative to the UL slot for CSI reporting may be configured by the gNB via Radio Resource Control (RRC) signaling. The 1V4PMIs are compressed in a beam-frequency-Doppler domain, and the compressed PMI is reported to the gNB in a single CSI report. The codebookType for this Type II CSI prediction is set to 'typeII-Doppler-rl8' or 'typeII-Doppler-PortSelection-rl8'.

[0042] CSI processing unit (CPU) for computing CSI reportIn NR, the concept of CPU was introduced, where the number of CPUs, denoted as NCPU, is equal to the number of simultaneous CSI calculations supported by the UE. The UE indicates NCPUto the network as part of the UE capability. When the UE is triggered for a CSI report, a certain number of CPUs, denoted as OCPU, may be allocated to the UE from the available CPU pool, which may be occupied for a period of time (measured in symbols). If there are not enough CPUs for a given time instance, the newly triggered CSI report does not need to be calculated by the UE.

[0043] The number of occupied CPUs for a given CSI report depends on the content (configured by higher layer parameter ‘ reportQuantity actually the complexity, for calculating it. The followings options are based on the current 3 GPP NR Technical Standard (TS) 38.214 V18.5.0 (hereinafter referred to as “3GPP TS 38.214”):

[0044] • When ‘ reportQuantity ’ is set to ‘none’ and aperiodic Timing Reference Signal (TRS) is configured, then the TRS is mainly used for time and / or frequency synchronization at the UE, and nothing needs to be reported. In addition, the UE is assumed to have dedicated resources for TRS processing. Therefore, for this case, OCPU= 0;

[0045] • When "reportQuantity ’ is set to beam related parameters, such as ‘cri-RSRP’, ‘ssb-Index-RSRP’, etc., OCPU= 1, since beam related processing is usually not complex;

[0046] • When "reportQuantity ’ is set to non-beam related parameters, such as ‘cri-RI-PMI-CQT, ‘cri-RI-il’, etc., the CSI report typically occupies as many CPUs as the number of CSI-RS resources in the CSI-RS resource set for channel measurement;

[0047] • When ‘reportQuantity’ is set to 'cri-RI-PMI-CQI' and with codebookType set to 'typeII-Doppler-rl8' or 'typeII-Doppler-PortSelection-rl8':

[0048] o if the corresponding CSI-RS Resource Set for channel measurement is aperiodic and configured with K CSI-RS resources, OCPU= 8 for K = 12 and OCPU= Y • K for K < 12, where

[0049]

[0050] E {1, 2, 3} is reported by UE capability indication; and / or o if the corresponding CSI-RS Resource Set for channel measurement is periodic or semi-persistent and configured with a single CSI-RS resource, OCPU= 4 for N4= 1 and OCPU= max( Y2• N4, 4) for N4> 1, where the value of N4is configured by the higher layer parameter N4, and Y2E {1, 2, 3} is reported by UE capability indication, The period of time (measured by the number of symbols) for which the CPU is occupied for a given CSI report depends on the time domain behavior of the said CSI report. In general:• For periodic or semi-persistent CSI report ((excluding an initial semi-persistent CSI report on PUSCH after the PDCCH triggering the report and a semi-persistent CSI report on PUSCH configured with the higher layer parameter codebookType set to 'typeII-Doppler-rl8' or 'typeII-Doppler-PortSelection-rl8'), the CPU is occupied from the first symbol of the earliest CSI-RS / CSI-IM / SSB resource for channel or interference measurement, no later than the CSI-RS reference resource, until the last symbol of the configured PUSCH / PUCCH carrying the report. For the example in FIG. 2, one CSI-RS resource is configured to the UE for channel measurement (denoted by the first bar), then T' is the CPU occupancy period for periodic or semi-persistent CSI report;

[0051] • For aperiodic CSI report, the CPU is occupied from the first symbol after the PDCCH triggering the CSI report, until the last symbol of the scheduled PUSCH carrying the report. For the example in FIG. 2, T" is the CPU occupancy period for aperiodic CSI report;

[0052] • An initial semi-persistent CSI report on PUSCH after the PDCCH trigger occupies CPU(s) from the first symbol after the PDCCH until the last symbol of the scheduled PUSCH carrying the report. For the example in FIG. 2, T" is the CPU occupancy period for an initial semi-persistent CSI report; and / or

[0053] • A semi-persistent CSI report on PUSCH configured with the higher layer parameter codebookType set to 'typeII-Doppler-rl8' or 'typeII-Doppler-PortSelection-rl8' occupies CPU(s) from the first symbol of A>-th latest consecutive periodic / semi-persistent CSI-RS occasions no later than CSI reference resource, until the last symbol of the PUSCH carrying the report, where the value of KPE {1,2,4} is indicated by UE capability.

[0054] If a CSI-RS resource is referred N times by one or more CSI Reporting Settings not configured with higher layer parameter csi-ReportSubConfigToAddModList, the CSI-RS resource and the CSI-RS ports within the CSI-RS resource are counted N times.

[0055] For a periodic or semi-persistent CSI-RS resource in a CSI-RS resource set for channel measurement linked to a CSI-ReportConfig configured with the higher layer parameter codebookType set to 'typeII-Doppler-rl8' or 'typeII-Doppler-PortSelection-rl8', the CSI-RS resource and the CSI-RS ports within the CSI-RS resource are counted KPtimes, where the value of KPE {1,2,4} is indicated by UE capability.

[0056] General aspects for NR 3 GPP Rel-18 AI / ML for NR air interface

[0057] Artificial Intelligence (Al) and Machine Learning (ML) have been investigated, both in academia and industry, as promising tools to optimize the design of the airinterface in wireless communication networks. Example use cases include usingautoencoders CSI compression to reduce the feedback overhead and improve channel prediction accuracy; using deep neural networks for classifying Line-of-Sight (LOS) and Non-LOS (NLOS) conditions to enhance the positioning accuracy; and using reinforcement learning for beam selection at the network side and / or the UE side to reduce the signaling overhead and beam alignment latency; using deep reinforcement learning to learn an optimal precoding policy for complex Multiple Input Multiple Output (MIMO) precoding problems.

[0058] In 3 GPP NR standardization work, a 3 GPP Release 18 study item on AI / ML for the NR air interface has been discussed. This study item explored the benefits of augmenting the air-interface with features enabling improved support of AI / ML based algorithms for enhanced performance and / or reduced complexity / overhead. Through studying a few selected use cases (CSI feedback, beam management, and positioning), this study item aims at laying the foundation for future air-interface use cases leveraging AI / ML techniques.

[0059] LCM operations for AI / ML for NR air interface

[0060] An important part in Al development and operation is the lifecycle management (LCM) of the AI / ML model (e.g., model training, model deployment, model inference, model monitoring, model updating) and AI / ML functionality.

[0061] In NR 3 GPP Rel-18 AI / ML for NR air interface study item, the LCM procedure is studied for the case that an AI / ML model has a model Identifier (ID) with associated information and / or for the case that a given functionality is provided by some AI / ML operations.

[0062] Two types of LCM operations were studied in NR 3 GPP Rel-18, functionalitybased LCM and model-ID based LCM:

[0063] • Functionality refers to an AI / ML-enabled Feature / FG enabled by configuration(s), where configuration(s) is(are) supported based on conditions indicated by UE capability. Correspondingly, functionality-based LCM operates based on, at least, one configuration of an AI / ML-enabled Feature / FG or specific configurations of an AI / ML-enabled Feature / FG. In functionality-based LCM, the network indicates activation / deactivation / fallback / switching of AI / ML functionality via 3GPP signaling (e.g., RRC, Medium Access Control (MAC) Control Element (CE), Downlink Control Information (DCI)). Models may not be identified at the network, and the UE may perform model-level LCM. Whether and how much awareness / interaction network should have about model-level LCM requires further study. For functionality identification, theremay be either one or more than one functionalities defined within an AI / ML-enabled feature, whereby AI / ML-enabled Feature refers to a Feature where AI / ML may be used; and

[0064] • In model-1 D-based LCM, models are identified at the network, and Network / UE may activate / deactivate / select / s witch individual AI / ML models via model ID. A model may be associated with specific configurations / conditions associated with UE capability of an AI / ML-enabled Feature / FG and additional conditions (e.g., scenarios, sites, and datasets) as determined / identified between the UE-side and the network-side. An AI / ML model identified by a model ID may be logical, and how it maps to physical AI / ML model(s) may be up to implementation.

[0065] Functional framework for AI / ML for NR air interface

[0066] FIG. 3 shows a functional framework that may be used for studying model LCM aspects for different Al for PHY use cases. The general framework consists of the following:

[0067] • Data Collection is a function that provides input data to the Model Training, Management, and Inference functions:

[0068] o Training Data: Data needed as input for the AI / ML Model Training function;

[0069] o Monitoring Data: Data needed as input for the Management of AI / ML models or AI / ML functionalities; and / or

[0070] o Inference Data: Data needed as input for the AI / ML Inference function;

[0071] • Model Training is a function that performs AI / ML model training, validation, and testing which may generate model performance metrics which may be used as part of the model testing procedure. The Model Training function is also responsible for data preparation (e.g., data pre-processing and cleaning, formatting, and transformation) based on Training Data delivered by a Data Collection function, if required:

[0072] o Trained / Updated Model: In case of having a Model Storage function, this is used to deliver trained, validated, and tested AI / ML models to the Model Storage function, or to deliver an updated version of a model to the Model Storage function;

[0073] • Management is a function that oversees the operation (e.g., selection / (de)activation / switching / fallback) and monitoring (e.g., performance) of AI / ML models or AI / ML functionalities. This function is also responsible for making decisions toensure the proper inference operation based on data received from the Data Collection function and the Inference function:

[0074] o Management Instruction: Information needed as input to manage the Inference function. Concerning information may include selection / (de)activation / switching of AI / ML models or AI / ML-based functionalities, fallback to non-AI / ML operation (i.e., not relying on inference process), etc.;

[0075] o Model Transfer / Delivery Request: Used to request model(s) to the Model Storage function; and / or

[0076] o Performance Feedback / Retraining Request: Information needed as input for the Model Training function, e.g., for model (re)training or updating purposes;

[0077] • Inference is a function that provides outputs from the process of applying AI / ML models or AI / ML functionalities, using the data that is provided by the Data Collection function (i.e., Inference Data) as an input. The Inference function is also responsible for data preparation (e.g., data pre-processing and cleaning, formatting, and transformation) based on Inference Data delivered by a Data Collection function, if required:

[0078] o Inference Output: Data used by the Management function to monitor the performance of AI / ML models or AI / ML functionalities; and / or

[0079] • Model Storage is a function responsible for storing trained / updated models that may be used to perform the Inference function:

[0080] o Note: The Model Storage function in FIG. 3 is only intended as a reference point (if any) when applicable for protocol terminations, model transfer / delivery, and related processes. It should be stressed that its purpose does not encompass restricting the actual storage locations of models. Therefore, the specification impact of all data / information / instruction flows (i.e., the arrows in FIG. 3) to / from this function should be studied case by case; and / or

[0081] o Model Transfer / Delivery: Used to deliver an AI / ML model to the Inference function.

[0082] 3GPP NR Rel-19 time-domain CSI prediction using UE-side AI / ML model In NR 3GPP Rel-18, Al-based UE-side CSI prediction was introduced as an Al for PHY use case and being specified in NR 3 GPP Rel-19 work item on AI / ML for the NR air interface.

[0083] One or more AI / ML models may be trained and deployed at a UE for the Al-based CSLprediction feature. During model inference, a UE is configured by the gNB tomeasure a set of historical CSI-RSs (e.g., the K CSI-RS measurements in the observation window shown in FIG. 1) and then report a predicted CSI in the scheduled UL slot for one or multiple future time instances (e.g., the 1V4future tine instances in FIG. 1) using its AI / ML model(s). The same codebookType (e.g., 'typeII-Doppler-rl8') that is defined in 3GPP Rel-18 for time domain Type II CSI prediction at UE may be reused for configuring the UE to send the predicted CSI using a UE-side AI / ML model, but a new codebookType may also be defined. The CSI report carrying the predicted CSI based on the inference output of a UE-sided AI / ML model may be referred to as the CSI-inference report.

[0084] Performance monitoring for CSI prediction

[0085] For CSI prediction using UE side AI / ML model use case studied in 3GPP Rel-18 AI / ML for NR air interface study item, at least the following aspects have been proposed by companies on performance monitoring for functionality -based LCM captured in TR 38.843:

[0086] • Type 1:

[0087] o UE calculates the performance melric(s)

[0088] o UE reports performance monitoring output that facilitates functionality fallback decision at the network:

[0089] ■ Performance monitoring output details may be further defined; and / or

[0090] ■ the network may configure threshold criterion to facilitate UE side performance monitoring (if needed); and / or

[0091] o the network makes decision(s) of functionality fallback operation (fallback mechanism to legacy CSI reporting);

[0092] • Type 2:

[0093] o UE reports predicted CSI and / or the corresponding ground-truth,' o The network node calculates the performance metrics,' o The network node makes decision(s) of functionality fallback operation (fallback mechanism to legacy CSI reporting);

[0094] • Type 3:

[0095] o UE calculates the performance metric(s), '

[0096] o UE reports performance metric(s) to the network; and / pr o The network node makes decision(s) of functionality fallback operation (fallback mechanism to legacy CSI reporting);

[0097] • Functionality selection / activation / deactivation / switching as defined forother UE side use cases may be reused, if applicable;

[0098] • Configuration and procedure for performance monitoring;

[0099] • CSI-RS configuration for performance monitoring;

[0100] • Performance metric including at least intermediate Key Performance Indicator (KPI) (e.g., Normalized Mean Square Error (NMSE) or Squared Generalized Cosine Similarity (SGCS));

[0101] • UE report, including periodic / semi-persistent / aperiodic reporting, and event driven report; and / or

[0102] • Note: The UE may make decision within the same functionality on model selection, activation, deactivation, switching operation transparent to the network node.

[0103] For monitoring the performance of a UE-sided AI / ML model, if the performance monitoring is performed at the UE-side (e.g., type 1 or type 3 performance monitoring), then, a UE may report the model performance monitoring results to the network node, so that the network node takes the UE reported model performance information into account when making model level or functionality level LCM decisions (e.g., fallback to non-AL / ML algorithm, functionality / model switching, etc.).

[0104] In 3GPP RANI #120 meeting, the following agreement was made regarding performance monitoring.

[0105] “Agreement For CSI prediction using UE-side model, if performance monitoring type 1 or 3 is supported, for calculation of monitoring metric, support

[0106] Based on intermediate KPI

[0107] o Down select between SGCS and NMSE by RANl#120bis o FFS on the definition of SGCS / NMSE and how to calculate monitoring metric ”

[0108] For monitoring the performance of a CSI prediction feature using UE side AI / ML model (e.g., type 1 or type 3 performance monitoring mentioned in section 2.2.6.1), a UE may be configured by the network node to compute the performance monitoring result(s) of the CSI prediction feature, and the performance monitoring result(s) may be sent to the network node in a CSI report.

[0109] For intermediate KPI based performance monitoring of an AI / ML-based CSI prediction feature, a monitoring data sample may consist of both the channel measurements within an observation window and the channel measurements within the associated prediction window. The channel measurements within the observation window are used for creating model input, which is then fed to the AI / ML model for generating amodel output (predicted CSI). The measurements within the prediction window are used for creating the ground-truth measurement. An intermediate KPI (e.g., NMSE or SGCS) per monitoring data sample may be derived by comparing a CSI prediction model output (i.e., predicted CSI for the one or more future time instances) with the corresponding ground-truth measurement (i.e., the channel measurements(s) corresponding to the one or more future time instances).

[0110] SUMMARY

[0111] Some embodiments advantageously provide methods, systems, and apparatuses for configuring a UE to report performance monitoring result(s) (e.g., CSI-PM report) of a CSI prediction feature using a UE-side AI / ML model, where the performance monitoring result(s) is obtained / approximated / predicted by the UE without the ground-truth measurements. Some embodiments provide methods to define the CSI processing criteria for the CSI-PM report.

[0112] Some embodiments provide:

[0113] • Methods for configuring the CSI-PM reporting with / without CSI-inference report; and / or

[0114] • Methods for determining the number of occupied CPUs and the CPU occupancy time that are required for the CSI-PM report with / without CSI-inference report.

[0115] Some embodiments provide methods to enable a UE to calculate the performance monitoring result(s) of an Al based CSI prediction feature without the ground-truth measurements, Hence, the UE may not need to buffer the predicted CSI for generating the performance monitoring results report. This may reduce the needed memory size and complexity at the UE side for deriving the performance monitoring results. The latency for the network node to obtain the performance monitoring report may also be reduced when compared to conventional methods. Some embodiments provide ways to define the CSI processing timeline for CSI reports, which may help the network node configure UE reporting performance monitoring result(s) of an Al based CSI prediction feature efficiently.

[0116] According to one aspect, a method in a user equipment, UE, configured to communicate with a network node is provided. The method includes generating predicted channel state information, CSI, samples via a trained artificial intelligence, Al, CSI prediction model, the Al CSI prediction model being trained based at least in part on a first training data set of samples of CSI measurements in a prediction window. The methodalso includes transmitting a CSI performance monitoring, PM, report of CSI-PM results based on a performance metric estimator.

[0117] According to this aspect, in some embodiments, training the Al CSI prediction model includes deriving ground truth labels for a performance metric based at least in part on the predicted CSI samples, and training the performance metric estimator based at least in part on the derived ground truth labels. In some embodiments, training the performance metric estimator includes obtaining a second training data set of samples of CSI measurements in an observation window. In some embodiments, training the performance metric estimator includes obtaining a third training data set of samples of predicted CSI. In some embodiments, training the performance metric estimator includes obtaining a fourth training data set of samples of CSI measurements in an observation window and predicted CSI. In some embodiments, the method includes receiving from the network node a CSI-PM report configuration, the transmitted CSI-PM report being based at least in part on the CSI-PM report configuration. In some embodiments, the method includes determining the CSI-PM report with or without a CSI inference report. In some embodiments, the method includes signaling a capability report, the capability report including a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report. In some embodiments, the method includes receiving a request for a CSI-PM report, and not computing a CSI-PM report when a number of CPUs is not available. In some embodiments, the UE is configured to report CSI inference reports and CSI PM reports, separately or jointly. In some embodiments, the predicted CSI samples include at least one of a predicted precoding matrix indicator, PMI, a predicted rank indicator, RI, a predicted reference signal received power, RSRP, a predicted channel quality indicator, CQI, a predicted CSI reference signal resource indicator, CRI, a predicted strongest beam of K strongest beams, and a predicted cell of top K cells. In some embodiments, first training data set of samples of CSI measurements include multiple CSI reference signal, RS, measurements spread over time on a set of K aperiodic CSI-RS resources. In some embodiments, the method includes: receiving from the network node a CSI-PM report configuration; applying the received configuration and receiving a trigger to predict and transmit a CSI-PM report based at least in part on the CSI-PM report configuration, without additional resources for measurement of ground truth labels. In some embodiments, the method includes transmitting the CSI-PM report in response to the received trigger.

[0118] According to another aspect, a method in a network node configured tocommunicate with a user equipment, UE, is provided. The method includes transmitting to the UE a channel state information performance monitoring, CSI-PM, report configuration. The method includes receiving a CSI PM report of CSI-PM results, the CSI-PM results being based at least on a performance metric estimator, the performance metric estimator being based at least in part on ground truth labels derived from predicted CSI samples generated by an artificial intelligence, Al, CSI prediction model trained by a first training data set of CSI measurement samples.

[0119] According to this aspect, in some embodiments, the CSI-PM report configuration configures the UE to determine the CSI-PM report with or without a CSI inference report. In some embodiments, the CSI-PM report configuration configures the UE to determine a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report. In some embodiments, the CSI-PM report configuration configures the UE to report CSI inference reports and CSI PM reports, separately or jointly. In some embodiments, the CSI-PM report configuration configures the UE to report CSI-PM without reporting an associated CSI inference report. In some embodiments, the method includes triggering the UE to compute and transmit the CSI PM report.

[0120] According to yet another aspect, a user equipment, UE, configured to communicate with a network node is provided. The UE includes processing circuitry is configured to generate predicted channel state information, CSI, samples via a trained artificial intelligence, Al, CSI prediction model, the Al CSI prediction model being trained based at least in part on a first training data set of samples of CSI measurements in a prediction window. The processing circuitry is configured to transmit a CSI performance monitoring, PM, report of CSI-PM results based on a performance metric estimator.

[0121] In some embodiments, the processing circuitry is configured to receive from the network node a CSI-PM report configuration, the transmitted CSI-PM report being based at least in part on the CSI-PM report configuration. In some embodiments, the processing circuitry is configured to determine the CSI-PM report with or without a CSI inference report. In some embodiments, the processing circuitry is configured to signal a capability report, the capability report including a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report. In some embodiments, the processing circuitry is configured to report CSI inference reports and CSI PM reports, separately or jointly. In some embodiments, the processing circuitry is further configured to: receive from the network node a CSI-PMreport configuration; apply the received configuration; receive a trigger to predict and transmit a CSI-PM report based at least in part on the CSI-PM report configuration, without additional resources for measurement of ground truth labels; and transmitting the CSI-PM report in response to the received trigger.

[0122] According to another aspect, a network node configured to communicate with a user equipment, UE, is provided. The network node includes processing circuitry configured to: transmit to the UE a channel state information performance monitoring, CSI-PM, report configuration; and receive a CSI PM report of CSI-PM results, the CSI-PM results being based at least on a performance metric estimator, the performance metric estimator being based at least in part on ground truth labels derived from predicted CSI samples generated by an artificial intelligence, Al, CSI prediction model trained by a first training data set of CSI measurement samples.

[0123] In some embodiments, the processing circuitry is configured to trigger the UE to compute and transmit the CSI PM report.

[0124] BRIEF DESCRIPTION OF THE DRAWINGS

[0125] A more complete understanding of the present embodiments, and the attendant advantages and features thereof, will be more readily understood by reference to the following detailed description when considered in conjunction with the accompanying drawings wherein:

[0126] FIG. 1 shows an example CMR enhancement for 3GPP Rel-18 Type II CSI prediction at UE;

[0127] FIG. 2 shows an example CPU occupancy period;

[0128] FIG. 3 shows a functional framework for AI / ML for NR Air Interface;

[0129] FIG. 4 is a schematic diagram of an example network architecture illustrating a communication system according to principles disclosed herein;

[0130] FIG. 5 is a block diagram of a network node in communication with a user equipment over a wireless connection according to some embodiments of the present disclosure;

[0131] FIG. 6 is a schematic diagram of another example network architecture illustrating a communication system according to principles disclosed herein;

[0132] FIG. 7 is a flowchart of an example process in a UE according to some embodiments of the present disclosure;FIG. 8 is a flowchart of an example process in a network node according to some embodiments of the present disclosure;

[0133] FIG. 9 is a flowchart of another example process in a network node according to some embodiments of the present disclosure;

[0134] FIG. 10 is a flowchart of another example process in a UE according to some embodiments of the present disclosure;

[0135] FIG. 11 shows example signaling for CSI-inference reporting with AP CSI-RS resources according to some embodiments of the present disclosure;

[0136] FIG. 12 shows example signaling for CSI-inference reporting with P or SP CSI-RS resources according to some embodiments of the present disclosure;

[0137] FIG. 13 shows example single DCI triggering for a joint CSI-inference and CSI-PM reporting with AP CSI-RS resources according to some embodiments of the present disclosure;

[0138] FIG. 14 shows example single DCI triggering for a joint CSI-inference and CSI-PM reporting with P or SP CSI-RS resource according to some embodiments of the present disclosure;

[0139] FIG. 15 shows two example DCIs triggering for measurement and CSI report separately, with AP CSI-RS resources according to some embodiments of the present disclosure;

[0140] FIG. 16 shows two example DCIs triggering for measurement and CSI report separately, with P or SP CSI-RS resources according to some embodiments of the present disclosure;

[0141] FIG. 17 shows an example DCI triggering for a CSI-PM reporting with AP CSI-RS resources according to some embodiments of the present disclosure;

[0142] FIG. 18 shows an example of two separate DCI signaling for CSI-inference reporting and CSI-PM reporting with AP CSI-RS resource, PM is calculated per sample according to some embodiments of the present disclosure;

[0143] FIG. 19 shows an example of two separate DCI signaling for CSI-inference reporting and CSI-PM reporting with AP CSI-RS resource according to some embodiments of the present disclosure; and

[0144] FIG. 20 shows an example of single DCI signaling for CSI-inference reporting and CSI-PM reporting with AP CSI-RS resource according to some embodiments of the present disclosure.DETAILED DESCRIPTION

[0145] Before describing in detail exemplary embodiments, it is noted that the embodiments reside primarily in combinations of apparatus components and processing steps related to criteria for a user equipment or device to report performance monitoring result for channel state information (CSI) prediction without ground-truth measurements. Accordingly, components have been represented where appropriate by conventional symbols in the drawings, showing only those specific details that are pertinent to understanding the embodiments so as not to obscure the disclosure with details that will be readily apparent to those of ordinary skill in the art having the benefit of the description herein.

[0146] As used herein, relational terms, such as “first” and “second,” “top” and “bottom,” and the like, may be used solely to distinguish one entity or element from another entity or element without necessarily requiring or implying any physical or logical relationship or order between such entities or elements. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the concepts described herein. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and / or “including” when used herein, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0147] In embodiments described herein, the joining term, “in communication with” and the like, may be used to indicate electrical or data communication, which may be accomplished by physical contact, induction, electromagnetic radiation, radio signaling, infrared signaling or optical signaling, for example. One having ordinary skill in the art will appreciate that multiple components may interoperate and modifications and variations are possible of achieving the electrical and data communication.

[0148] In some embodiments described herein, the term “coupled,” “connected,” and the like, may be used herein to indicate a connection, although not necessarily directly, and may include wired and / or wireless connections.

[0149] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the concepts described herein. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural formsas well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and / or “including” when used herein, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0150] The term “network node” used herein may be any kind of network node comprised in a radio network which may further comprise any of base station (BS), radio base station, base transceiver station (BTS), base station controller (BSC), radio network controller (RNC), g Node B (gNB), evolved Node B (eNB or eNodeB), Node B, multistandard radio (MSR) radio node such as MSR BS, multi-cell / multicast coordination entity (MCE), relay node, donor node controlling relay, radio access point (AP), transmission points, transmission nodes, Remote Radio Unit (RRU) Remote Radio Head (RRH), a core network node (e.g., mobile management entity (MME), self-organizing network (SON) node, a coordinating node, positioning node, MDT node, etc.), an external node (e.g., 3rd party node, a node external to the current network), nodes in distributed antenna system (DAS), a spectrum access system (SAS) node, an element management system (EMS), etc. The network node may also comprise test equipment. The term “radio node” used herein may be used to also denote a user equipment (UE) such as a wireless device (WD) or a radio network node.

[0151] In some embodiments, the non-limiting terms wireless device (WD) or a user equipment (UE) are used interchangeably. The UE herein may be any type of user equipment capable of communicating with a network node or another UE over radio signals, such as a wireless device (WD). The UE may also be a radio communication device, target device, device to device (D2D) UE, machine type UE or UE capable of machine to machine communication (M2M), low-cost and / or low-complexity UE, a sensor equipped with UE, Tablet, mobile terminals, smart phone, laptop embedded equipped (LEE), laptop mounted equipment (LME), USB dongles, Customer Premises Equipment (CPE), an Internet of Things (loT) device, or a Narrowband loT (NB-IOT) device etc.

[0152] Also, in some embodiments the generic term “radio network node” is used. It may be any kind of a radio network node which may comprise any of base station, radio base station, base transceiver station, base station controller, network controller, RNC, evolved Node B (eNB), Node B, gNB, Multi-cell / multicast Coordination Entity (MCE), relay node, access point, radio access point, Remote Radio Unit (RRU) Remote Radio Head (RRH).Note that although terminology from one particular wireless system, such as, for example, 3GPP LTE and / or New Radio (NR) and / or 6G, may be used in this disclosure, this should not be seen as limiting the scope of the disclosure to only the aforementioned system. It is contemplated that other 3GPP systems may make use of the concepts and arrangements disclosed herein. For example, a disclosure relating to NR may also be implementable in a 6G system and / or an LTE system, a disclosure relating to 6G may also be implementable in a NR and / or LTE system, and a disclosure relating to LTE may also be implementable in a NR and / or 6G system. Other wireless systems, including without limitation Wide Band Code Division Multiple Access (WCDMA), Worldwide Interoperability for Microwave Access (WiMax), Ultra Mobile Broadband (UMB) and Global System for Mobile Communications (GSM), may also benefit from exploiting the ideas covered within this disclosure.

[0153] Note further, that functions described herein as being performed by a user equipment or a network node may be distributed over a plurality of user equipments and / or network nodes. In other words, it is contemplated that the functions of the network node and user equipment described herein are not limited to performance by a single physical device and, in fact, may be distributed among several physical devices.

[0154] In some embodiments, the term “ground-truth measurement” is used and may refer to a ground-truth label. In some other embodiments, a ground-truth measurement (or ground truth label) may be a measurement of what the UE is supposed to predict. For example, ground-truth measurements may include ground-truth measurements for ALCSI prediction model training, ground-truth measurements for ALCSI prediction monitoring, ground-truth measurements for performance metric, etc.

[0155] In some embodiments, the difference between ground truth-measurements for AL CSI prediction model training and ground-truth measurements for ALCSI prediction monitoring may only be conceptual, in what they are used for. “Ground-truth measurements for ALCSI prediction model training” and “Ground-truth measurements for ALCSI prediction monitoring” are channels / measurements / processed measurements from the instances in a prediction window. In other words, if a model is trained to predict something in time, then the ground-truth measurement is a measurement of those timeinstances. This may also be applicable for monitoring, where a prediction is made, and then how good the prediction was may be measured and compared.

[0156] Further, there may be “formats” for the “Ground-truth measurements for ALCSI prediction model training” and “Ground-truth measurements for ALCSI predictionmonitoring.” The formats may be related to or further described with respect to “A ground-truth measurement for Al CSI prediction model training”, and “A ground-truth measurement for Al CSI performance monitoring”.

[0157] In some embodiments, ground truth measurements for performance metric are achieved by comparing the predicted CSI to the ground-truth measurements for AI-CSI prediction monitoring, and getting a performance metric, e.g., NMSE or SGCS, that a later model may then be trained to predict.

[0158] In some embodiments, the term “CSI prediction feature” is used and may refer to device or node being capable to predict a CSI based on inputs such as measurements, information, parameters, etc. In some embodiments, a CSI prediction feature may be a configuration used by the device or node to perform CSI predictions.

[0159] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It will be further understood that terms used herein should be interpreted as having a meaning that is consistent with their meaning in the context of this specification and the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0160] Referring again to the drawing figures, in which like elements are referred to by like reference numerals, there is shown in FIG. 4 a schematic diagram of a communication system 10, according to an embodiment, such as a 3 GPP -type cellular network that may support standards such as LTE and / or NR (5G) and / or 6G, which comprises an access network 12, such as a radio access network, and a core network 14. The core network 14 includes one or more network nodes 15. The access network 12 comprises a plurality of network nodes 16a, 16b, 16c (referred to collectively as network nodes 16), such as NBs, eNBs, gNBs or other types of wireless access points, each defining a corresponding coverage area 18a, 18b, 18c (referred to collectively as coverage areas 18). Each network node 16a, 16b, 16c is connectable to the core network 14 over a wired or wireless connection 20. A first user equipment (UE) 22a located in coverage area 18a is configured to wirelessly connect to, or be paged by, the corresponding network node 16a. A second UE 22b in coverage area 18b is wirelessly connectable to the corresponding network node 16b. While a plurality of UEs 22a, 22b (collectively referred to as user equipments 22) are illustrated in this example, the disclosed embodiments are equally applicable to a situation where a sole UE is in the coverage area or where a sole UE is connecting to the corresponding network node 16. Note that although only two UEs 22 and three networknodes 16 are shown for convenience, the communication system may include many more UEs 22 and network nodes 16.

[0161] As one example, in certain embodiments, access network 12 may contain some access network nodes 16 that support 3 GPP radio access technologies (RAT), such as LTE or NR, while other access network nodes 16 support (or the same access network nodes 16 additionally support) non-3GPP RATs, such as Wi-Fi or a proprietary RAT. As another example, communication system 10 may support multiple generations of related communication standards (e.g., 4G, 5G and 6G 3GPP communication standards) and, as a result, may include an access network 12 and / or a core network 14 that supports multiple different standard generations or may include multiple access networks 12 and / or multiple core networks 14 with individual networks supporting different standards generations.

[0162] Also, it is contemplated that a UE 22 may be in simultaneous communication and / or configured to separately communicate with more than one network node 16 and more than one type of network node 16. For example, a UE 22 may have dual connectivity with a network node 16 that supports LTE and the same or a different network node 16 that supports NR. As an example, UE 22 may be in communication with an eNB for LTEZE-UTRAN, a gNB for NR / NG-RAN (i.e. being configured for multiradio dual connectivity (MR-DC), such as E-UTRAN (Evolved-UMTS Terrestrial Radio Access Network) New Radio - Dual Connectivity (EN-DC) and / or Wi-Fi.

[0163] A network node 16 is configured to include a node management unit 24 which is configured to perform any step and / or task and / or process and / or method and / or feature described in the present disclosure, e.g., network node functions. A user equipment 22 is configured to include a UE management unit 26 which is configured to perform any step and / or task and / or process and / or method and / or feature described in the present disclosure, e.g., UE functions.

[0164] Example implementations, in accordance with an embodiment, of the UE 22 and network node 16 discussed in the preceding paragraphs will now be described with reference to FIG. 5.

[0165] The communication system 10 includes a network node 16 provided in a communication system 10 and including hardware 28 enabling it to communicate with the UE 22. The hardware 28 may include a communication interface 29 comprising a radio interface 30 for setting up and maintaining at least a wireless connection 32 with a UE 22 located in a coverage area 18 served by the network node 16. The radio interface 30 may be formed as or may include, for example, one or more RF transmitters, one or more RFreceivers, and / or one or more RF transceivers. The radio interface 30 includes an array of antennas 34 to radiate and receive signal(s) carrying electromagnetic waves.

[0166] In the embodiment shown, the hardware 28 of the network node 16 further includes processing circuitry 36. The processing circuitry 36 may include a processor 38 and a memory 40. In particular, in addition to or instead of a processor, such as a central processing unit, and memory, the processing circuitry 36 may comprise integrated circuitry for processing and / or control, e.g., one or more processors and / or processor cores and / or FPGAs (Field Programmable Gate Array) and / or ASICs (Application Specific Integrated Circuitry) adapted to execute instructions. The processor 38 may be configured to access (e.g., write to and / or read from) the memory 40, which may comprise any kind of volatile and / or nonvolatile memory, e.g., cache and / or buffer memory and / or RAM (Random Access Memory) and / or ROM (Read-Only Memory) and / or optical memory and / or EPROM (Erasable Programmable Read-Only Memory).

[0167] Thus, the network node 16 further has software 42 stored internally in, for example, memory 40, or stored in external memory (e.g., database, storage array, network storage device, etc.) accessible by the network node 16 via an external connection. The software 42 may be executable by the processing circuitry 36. The processing circuitry 36 may be configured to control any of the methods and / or processes described herein and / or to cause such methods, and / or processes to be performed, e.g., by network node 16.

[0168] Processor 38 corresponds to one or more processors 38 for performing network node 16 functions described herein. The memory 40 is configured to store data, programmatic software code and / or other information described herein. In some embodiments, the software 42 may include instructions that, when executed by the processor 38 and / or processing circuitry 36, causes the processor 38 and / or processing circuitry 36 to perform the processes described herein with respect to network node 16. For example, processing circuitry 36 of the network node 16 may include a node management unit 24 which is configured to perform any step and / or task and / or process and / or method and / or feature described in the present disclosure, e.g., network node functions.

[0169] The network node 16 may be composed of multiple distinct network entities (e.g., a NodeB entity and a RNC entity, or a BTS entity and a BSC entity, etc.), which may each have or utilize their own respective physical components. In certain scenarios in which the network node 16 comprises multiple such entities (e.g., BTS and BSC), one or more of the separate entities may be shared among several network nodes. For example, a single RNC may control multiple NodeB s. In such a scenario, each unique NodeB and RNC pair, mayin some instances be considered a single separate network node. In some embodiments, the network node 16 may be configured to support multiple radio access technologies (RATs). In such embodiments, some components may be duplicated (e.g., separate memories 40 or portions of memory 40 for different RATs) and some components may be reused (e.g., a same antenna may be shared by different RATs). The network node 16 may also include multiple sets of the various illustrated components for different wireless technologies integrated into network node 16, for example GSM, WCDMA, LTE, NR, Wi-Fi (e.g., according to an IEEE 802.11 family standard), Zigbee, Z-wave, LoRaWAN, Radio Frequency Identification (RFID) or Bluetooth wireless technologies. These wireless technologies may be integrated into the same or different chip or set of chips and other components within network node 16.

[0170] In certain alternative embodiments, network node 16 may be capable of wireless communication but does not include separate radio front-end circuitry, instead, the processing circuitry 36 includes radio front-end circuitry and is connected to the antenna 34. Similarly, in some embodiments, all or some of the RF receivers, transmitters and / or transceivers are part of the radio interface 30. In still other embodiments, the communication interface 29 includes one or more ports or terminals, the radio interface 30, and the RF receiver, transmitter and / or transceiver, and the communication interface 31 communicates with baseband processing circuitry, which is part of a digital unit (not shown).

[0171] The antenna 34 may include one or more antennas, or antenna arrays, configured to send and / or receive wireless signals. The antenna 34 may be coupled to the radio front-end circuitry in radio interface 30 and may be any type of antenna capable of transmitting and receiving data and / or signals wirelessly. In certain embodiments, the antenna 34 is separate from the network node 16 and connectable to the network node 16 through one or more interfaces or ports.

[0172] Network node 15 may include one or more components described above with respect to network node 16, e.g., communication interface 29, radio interface 30, antenna 34, ports, processing circuitry 36, processor 38, memory 40 and software 42. These elements of network node 15 may be arranged such that network node 15 may perform various core network functions. Network node 15 may communicate wirelessly or via a wired connection with network nodes 16 via communication link 59.

[0173] The communication system 10 further includes the UE 22 already referred to. The UE 22 may have hardware 44 that may include a radio interface 46 configured to set upand maintain a wireless connection 32 with a network node 16 serving a coverage area 18 in which the UE 22 is currently located. The radio interface 46 may be formed as or may include, for example, one or more RF transmitters, one or more RF receivers, and / or one or more RF transceivers. The radio interface 46 includes an array of antennas 48 to radiate and receive signal(s) carrying electromagnetic waves.

[0174] Communication functions of the radio interface 46 may include cellular communication, Wi-Fi communication (e.g., according to an IEEE 802.11 family standard), LPWAN communication, data communication, voice communication, multimedia communication, short-range communications such as Bluetooth, near-field communication, location-based communication such as the use of the global positioning system (GPS) to determine a location, another like communication function, or any combination thereof. Communications may be implemented according to one or more communication protocols and / or standards, such as IEEE 802.11, Code Division Multiplexing Access (CDMA), Wideband Code Division Multiple Access (WCDMA), GSM, LTE, New Radio (NR), UMTS, WiMax, Ethernet, transmission control protocol / intemet protocol (TCP / IP), synchronous optical networking (SONET), Asynchronous Transfer Mode (ATM), QUIC, Hypertext Transfer Protocol (HTTP), and so forth.

[0175] The hardware 44 of the UE 22 further includes processing circuitry 50. The processing circuitry 50 may include a processor 52 and memory 54. In particular, in addition to or instead of a processor, such as a central processing unit, and memory, the processing circuitry 50 may comprise integrated circuitry for processing and / or control, e.g., one or more processors and / or processor cores and / or FPGAs (Field Programmable Gate Array) and / or ASICs (Application Specific Integrated Circuitry) adapted to execute instructions. The processor 52 may be configured to access (e.g., write to and / or read from) memory 54, which may comprise any kind of volatile and / or nonvolatile memory, e.g., cache and / or buffer memory and / or RAM (Random Access Memory) and / or ROM (Read-Only Memory) and / or optical memory and / or EPROM (Erasable Programmable Read-Only Memory).

[0176] Thus, the UE 22 may further comprise software 56, which is stored in, for example, memory 54 at the UE 22, or stored in external memory (e.g., database, storage array, network storage device, etc.) accessible by the UE 22. The software 56 may be executable by the processing circuitry 50. The software 56 may include a client application 58. The client application 58 may be operable to provide a service to a humanor non-human user via the UE 22.

[0177] The processing circuitry 50 may be configured to control any of the methods and / or processes described herein and / or to cause such methods, and / or processes to be performed, e.g., by UE 22. The processor 52 corresponds to one or more processors 52 for performing UE 22 functions described herein. The UE 22 includes memory 54 that is configured to store data, programmatic software code and / or other information described herein. In some embodiments, the software 56 and / or the client application 58 may include instructions that, when executed by the processor 52 and / or processing circuitry 50, causes the processor 52 and / or processing circuitry 50 to perform the processes described herein with respect to UE 22. For example, the processing circuitry 50 of the user equipment 22 may include a UE management unit 26 which is configured to perform any step and / or task and / or process and / or method and / or feature described in the present disclosure, e.g., UE functions.

[0178] In some embodiments, the inner workings of the network node 16 and UE 22 may be as shown in FIG. 5 and independently, the surrounding network topology may be that of FIG. 4.

[0179] The wireless connection 32 between the UE 22 and the network node 16 is in accordance with the teachings of the embodiments described throughout this disclosure. More precisely, the teachings of some of these embodiments may improve the data rate, latency, and / or power consumption and thereby provide benefits such as reduced user waiting time, relaxed restriction on file size, better responsiveness, extended battery lifetime, etc. In some embodiments, a measurement procedure may be provided for the purpose of monitoring data rate, latency and other factors on which the one or more embodiments improve.

[0180] Although FIGS. 4 and 5 show various “units” such as node management unit 24 and UE management unit 26 as being within a respective processor, it is contemplated that these units may be implemented such that a portion of the unit is stored in a corresponding memory within the processing circuitry. In other words, the units may be implemented in hardware or in a combination of hardware and software within the processing circuitry.

[0181] FIG. 6 is another example of a communication system 10 according to some embodiments. As used herein, the communication system 10 of FIG. 6 includes multiple access points (APs) 60 (with four example APs 60a, 60b, 60c, and 60d being depicted) and multiple wireless devices, referred to in the context of communication system 10 of FIG. 6 as stations (STAs) 62 (referred to individually as STA 62a, STA 62b, STA 62c, STA 62d,and STA 62e). STA 62a is served by AP 60a in a first basic service set (BSS) 64a. STA 60b and STA 60c are served by AP 60b in a second BSS, BSS 64b. STA 62d is served by AP 60c in a third BSS, BSS 64c. STA 62e is served by AP 60d in a fourth BSS, BSS 64d. Stations 62 may be non-AP STAs and correspond to various kinds of wireless devices, for example, user terminals, such as mobile or stationary computing devices like smartphones, laptop computers, desktop computers, tablet computers, gaming devices, head-mounted displays (HMDs) for Augmented Reality (AR) or Virtual Reality (VR), or the like, including UEs 22 that are shown and described with respect to FIGS. 4 and 5. In other words, in some embodiment, STA 62 is a UE 22. Further, stations 62 may, for example, correspond to other kinds of equipment like smart home devices, printers, multimedia devices, data storage devices, or the like.

[0182] Each of STAs 62 may connect through a radio link to one of APs 60. For example, depending on location or channel conditions experienced by a given STA 62, the STA may select an appropriate AP and BSS for establishing the radio link. The radio link may be based on one or more orthogonal frequency-division multiplexing (OFDM) carriers from a frequency spectrum that is shared on the basis of a contention-based mechanism, e.g., an unlicensed or license exempt band like 2.4 GHz Industrial, Scientific, and Medical (ISM) band, the 5 GHz band, the 6 GHz band, or the 60 GHz band.

[0183] Each AP 60 may provide data connectivity to STAs 62 connected to a particular AP 60. As illustrated, APs 60 may be connected to a data network 66. In this way, APs 60 may also provide data connectivity between STAs 62 and other entities, e.g., to one or more servers, service providers, data sources, data sinks, user terminals, or the like.

[0184] Accordingly, the radio link established between a given STA 62 and its serving AP 60 may be used for providing various kinds of services to STA 62, e.g., a voice service, a multimedia service, or other data service. Such services may be based on applications that are executed on STA 62 and / or on a device linked to STA 62. By way of example, FIG. 6 illustrates an application service platform 68 provided in data network 66. The application(s) executed on STA 62 and / or on one or more other devices linked to STA 62 may use the radio link for data communication with one or more other STA 62 and / or the application service platform 68, thereby enabling utilization of the corresponding service(s) at STA 62.

[0185] FIG. 7 is a flowchart of an example process in a UE 22 according to some embodiments of the present disclosure. One or more blocks described herein may be performed by one or more elements of user equipment 22 such as by one or more ofprocessing circuitry 50 (including the UE management unit 26), processor 52, and / or radio interface 46. User equipment 22 such as via processing circuitry 50 and / or processor 52 and / or radio interface 46 is configured to determine (Block SI 00) one or more performance monitoring results of a channel state information (CSI) prediction feature using an artificial intelligence and / or machine learning (AI / ML) model. The one or more performance monitoring results are determined without one or more ground-truth measurements. The one or more ground-truth measurements are one or more measurements associated with an expected UE prediction. As used herein, “determination without one or more ground-truth measurements” may mean, in some embodiments, that a determination is made based on one or more measurements other than ground-truth measurements. The UE 22 is further configured to transmit (Block SI 02) a CSI performance monitoring (CSI-PM) report to the network node, the CSI-PM report including the one or more performance monitoring results. In some embodiments, the method may optionally include that the AI / ML model is trained using a ground-truth measurement.

[0186] In some embodiments, the method further includes receiving from the network node a CSI prediction performance monitoring report configuration, where the one or more performance monitoring results are based on the CSI prediction performance monitoring report configuration.

[0187] In some other embodiments, the method further includes determining the CSI-PM report with or without a C Si-inference report.

[0188] In some embodiments, the method further includes determining a quantity of occupied CSI processing units (CPUs) and a CPU occupancy time required for the CSI-PM report with or without the C Si-inference report.

[0189] In some other embodiments, the CSI prediction feature includes one or more of predicted precoding matrix indicators (PMI), one or more predicted raw channels, one or more predicted rank indicators (RIs), one or more predicted reference signal receive power (RSRP) values, one or more predicted channel quality indicators (CQIs), one or more predicted CSI reference signal (RS) resource indicator (CRIs), one or more predicted topic strongest beams, and one or more predicted top-K cells.

[0190] In some embodiments, the AI / ML model uses a model input sample for CSI prediction including multiple CSLRS measurements spread over time on a set of K aperiodic CSLRS resources.

[0191] In some other embodiments, the AI / ML model is configured to generate a modeloutput sample for CSI prediction including predicted CSI for one or more prediction time instances.

[0192] In some embodiments, the method further includes training the AI / ML model using a ground-truth measurement obtained using measured CSI for corresponding one or more prediction time instances.

[0193] In some other embodiments, the method further includes performing performance monitoring using the ground-truth measurement.

[0194] In some embodiments, the method further includes one or both of: (A) determining a training data sample for a prediction of the AI / ML model, where the training data sample includes a model input sample and a corresponding ground-truth measurement; and (B) determining an estimated performance metric for an associated prediction inference output using a performance metric estimator. The estimated performance metric includes an intermediate key performance indicator.

[0195] In some other embodiments, the UE is configured to report CSI-inference and CSL PM via separate CSI reports or jointly via a single CSI report, or the UE is configured to report CSI-PM without reporting an associated CSI-inference report.

[0196] In some embodiments, the one or more ground-truth measurements include one or more of ground-truth measurements for ALCSI prediction model training, ground-truth measurements for ALCSI prediction monitoring, ground-truth measurements for a performance metric.

[0197] FIG. 8 is a flowchart of an example process in a network node 16. One or more blocks described herein may be performed by one or more elements of network node 16 such as by one or more of processing circuitry 36 (including the node management unit 24), processor 38, and / or radio interface 30. Network node 16 such as via processing circuitry 36 and / or processor 38 and / or radio interface 30 is configured to transmit (Block SI 04) a channel state information (CSI) prediction performance monitoring report configuration to the UE and receive (Block SI 06) a CSI performance monitoring (CSI-PM) report from the UE. The CSI-PM report includes one or more performance monitoring results of a CSI prediction feature associated with an artificial intelligence and / or machine learning (AI / ML) model. The one or more performance monitoring results are determined by the UE without one or more ground-truth measurements. The one or more ground-truth measurements are one or more measurements associated with an expected UE prediction. The network node 16 is also configured to perform (Block SI 08) one or more action based on the CSI-PM report.FIG. 9 is a flowchart of another example process in a network node 16. One or more blocks described herein may be performed by one or more elements of network node 16 such as by one or more of processing circuitry 36 (including the node management unit 24), processor 38, and / or radio interface 30. Network node 16 such as via processing circuitry 36 and / or processor 38 and / or radio interface 30 is configured to transmits to the UE 22 a channel state information performance monitoring, CSI-PM, report configuration (Block SI 10). The method includes receiving a CSI PM report of CSI-PM results, the CSI-PM results being based at least on a performance metric estimator, the performance metric estimator being based at least in part on ground truth labels derived from predicted CSI samples generated by an artificial intelligence, Al, CSI prediction model trained by a first training data set of CSI measurement samples (Block SI 12).

[0198] According to this aspect, in some embodiments, the CSI-PM report configuration configures the UE 22 to determine the CSI-PM report with or without a CSI inference report. In some embodiments, the CSI-PM report configuration configures the UE 22 to determine a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report. In some embodiments, the CSI-PM report configuration configures the UE 22 to report CSI inference reports and CSI PM reports, separately or jointly. In some embodiments, the CSI-PM report configuration configures the UE 22 to report CSI-PM without reporting an associated CSI inference report. In some embodiments, the method includes triggering the UE 22 to compute and transmit the CSI PM report.

[0199] FIG. 10 is a flowchart of another example process in a UE 22 according to some embodiments of the present disclosure. One or more blocks described herein may be performed by one or more elements of user equipment 22 such as by one or more of processing circuitry 50 (including the UE management unit 26), processor 52, and / or radio interface 46. User equipment 22 such as via processing circuitry 50 and / or processor 52 and / or radio interface 46 is configured to generate predicted channel state information, CSI, samples via a trained artificial intelligence, Al, CSI prediction model, the Al CSI prediction model being trained based at least in part on a first training data set of samples of CSI measurements in a prediction window (Block SI 14). The method also includes transmitting a CSI performance monitoring, PM, report of CSI-PM results based on a performance metric estimator (Block SI 16).

[0200] According to this aspect, in some embodiments, training the Al CSI prediction model includes deriving ground truth labels for a performance metric based at least in parton the predicted CSI samples, and training the performance metric estimator based at least in part on the derived ground truth labels. In some embodiments, training the performance metric estimator includes obtaining a second training data set of samples of CSI measurements in an observation window. In some embodiments, training the performance metric estimator includes obtaining a third training data set of samples of predicted CSI. In some embodiments, training the performance metric estimator includes obtaining a fourth training data set of samples of CSI measurements in an observation window and predicted CSI. In some embodiments, the method includes receiving from the network node 16 a CSI-PM report configuration, the transmitted CSI-PM report being based at least in part on the CSI-PM report configuration. In some embodiments, the method includes determining the CSI-PM report with or without a CSI inference report. In some embodiments, the method includes signaling a capability report, the capability report including a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report. In some embodiments, the method includes receiving a request for a CSI-PM report, and not computing a CSI-PM report when a number of CPUs is not available. In some embodiments, the UE 22 is configured to report CSI inference reports and CSI PM reports, separately or jointly. In some embodiments, the predicted CSI samples include at least one of a predicted precoding matrix indicator, PMI, a predicted rank indicator, RI, a predicted reference signal received power, RSRP, a predicted channel quality indicator, CQI, a predicted CSI reference signal resource indicator, CRI, a predicted strongest beam of K strongest beams, and a predicted cell of top K cells. In some embodiments, first training data set of samples of CSI measurements include multiple CSI reference signal, RS, measurements spread over time on a set of K aperiodic CSI-RS resources. In some embodiments, the method includes: receiving from the network node 16 a CSI-PM report configuration; applying the received configuration and receiving a trigger to predict and transmit a CSI-PM report based at least in part on the CSI-PM report configuration, without additional resources for measurement of ground truth labels. In some embodiments, the method includes transmitting the CSI-PM report in response to the received trigger.

[0201] Having described the general process flow of arrangements of the disclosure and having provided examples of hardware and software arrangements for implementing the processes and functions of the disclosure, the sections below provide details and examples of arrangements for determining criteria for a user equipment or device to report performance monitoring result for channel state information (CSI) prediction withoutground-truth measurements.

[0202] The network signals CSI prediction performance monitoring report configuration for a CSI prediction feature to a UE 22, based on which the UE 22 derives one or more performance monitoring result(s), and reports the derived performance monitoring result(s) to the network. The Al based CSI prediction feature model / algorithm is implemented at the UE 22 using UE side AI / ML model(s) (e.g., one or more neural network models trained by using a large dataset and then deployed at the UE 22).

[0203] The network may refer to the gNB (network node 16), e.g. the gNB-CU or the gNB-DU, the 0AM, or a core network node 15, e.g. the Network Data Analytics Function (NWDAF). In the following, the term network may refer to any of the aforementioned entity. The entities may also be corresponding 6G nodes.

[0204] The CSI prediction feature is not limited to predicting the PMIs or raw channels for one or multiple future time instances. Examples of predicted CSI include predicted PMI(s), predicted raw channel(s), predicted RI(s), predicted RSRP(s), predicted CQI(s), predicted CRI(s), predicted top-K strongest beams, predicted top-K cells.

[0205] In some embodiments, the term intermediate KPI estimator is used. Other similar terms may also be applicable to this disclosure, such as intermediate KPI predictor, intermediate KPI approximator, etc.

[0206] In the following is an example case where the predicted CSI is predicted PMI(s) to describe the methods of configuring the CSI-PM report and the methods of defining the CPU criteria for this CSI-PM report.

[0207] An example of CSI prediction use case, where the predicted CSI is predicted PMI(s):

[0208] A model input sample for CSI prediction:

[0209] A model input sample for CSI prediction comprises multiple CSI-RS measurements spread over time on a set of K aperiodic CSI-RS resources (e.g., the measured CSIs on the aperiodic K CSI-RS resources within the observation window as shown in FIG. 1). An exact model input format may be up to UE implementation, the measurement occasions are specified in the standard. Different AI / ML model design may take different model input formats, e.g.:

[0210] a) The model input may be estimated raw channel per measurement time instance for all the K CSI-RS resources within the observation window;

[0211] b) The model input may be a channel Tx-correlation matrix per measurement time instance for all the K CSI-RS resources within the observation window;c) The model input may be a PMI in a free-format approximating an eigenvector of the channel Tx-correlation, per measurement time instance for all the K CSI-RS resources within the observation window;

[0212] d) The model input may be a PMI following the 3GPP Rel-16 type II CSI codebook, per measurement time instance for all the K CSI-RS resources within the observation window; and / or

[0213] e) The model input may be a PMI following the 3GPP Rel-18 “typell-Doppler-rl8” codebook except for the special (possibly non-standard) setting 1V4= K, derived from the measurements within the observation window (the K CSI-RS resources within the observation window).

[0214] A model output sample for CSI prediction:

[0215] A model output sample for CSI prediction comprises predicted CSI for the 1V4prediction time instances (e.g., the predicted CSIs on the 1V4time slots within the prediction window as shown in FIG. 11).

[0216] For CSI prediction using UE-sided AI / ML models, the UE 22 may report predicted PMI in a CSI report (denoted as CSI-inference report in this disclosure) to the network node 16 in the format of the 3GPP Rel-18 “typeII-Doppler-rl8” codebook as defined in 3GPP TS 38.214. The model output format may be up to UE implementation, whereas the CSI report format may be specified in the standard. Different AI / ML model designs may result in different model output formats, e.g.:

[0217] a) The model output may be predicted raw channel per prediction time instance for all the A4prediction time instances. The predicted raw channels may then be used for calculating predicted PMIs according to the 3GPP Rel-18 “typeII-Doppler-rl8” codebook as defined in 3GPP TS 38.214;

[0218] b) The model output may be a predicted channel Tx-correlation matrix per prediction time instance. The predicted channel Tx-correlation matrices may then be used for calculating predicted PMIs according to the 3GPP Rel-18 “typeII-Doppler-rl8” codebook as defined in 3GPP TS 38.214;

[0219] c) The model output may be a predicted PMI in a free-format approximating an eigenvector of the channel Tx-correlation, per time instance. The predicted free-format PMIs may then be used for calculating predicted PMIs according to the 3GPP Rel-18 “typeII-Doppler-rl8” codebook as defined in 3GPP TS 38.214;

[0220] d) The model output may be predicted PMI following the 3 GPP Rel-16 type II CSI codebook per prediction time instance for all the A4prediction time instances. Thepredicted 3 GPP Rel-16 type II PMIs may then be compressed into a PMI following the 3GPP Rel-18 “typeII-Doppler-rl8” codebook as defined in 3GPP TS 38.214; and / or e) The model may directly output a predicted PMIs following the 3 GPP Rel-18 “typeII-Doppler-rl8” codebook for the 1V4prediction time instances.

[0221] Further, if the model output is not in the 3GPP Rel-18 “typeII-Doppler-rl8” codebook, as defined in 3 GPP TS 38.214, then the UE 22 may do some post-processing to the model output to place it into the format of the 3GPP Rel-18 “typeII-Doppler-rl8” codebook before the UE 22 may transmit the CSI report with the predicted CSI to the network node 16.

[0222] A ground-truth measurement for Al CSI prediction model training

[0223] A ground-truth measurement for model training may be obtained using measured CSI(s) for the corresponding one or multiple prediction time instances (e.g., the measured CSIs on the 1V4time slots in the prediction window as shown in FIG. 1).

[0224] As the details of the AI / ML model in the UE 22 may be up to UE implementation, so is the training pipeline. Thus, the ground-truth measurement for model training may be different depending on different AI / ML model designs, data collection capabilities, and training strategies. Examples of strategies may include:

[0225] 1. Train in an end-to-end fashion, in which case one may define the groundtruth measurement, for model training, as “typeII-Doppler-rl8” codebook, regardless which model output format is used for the AI / ML model. The ground-truth measurement in the format of “typeII-Doppler-rl8” codebook may be obtained by measuring CSI on the / V4time instances in the prediction window, and compressing the measured CSIs on the / V4prediction time instances according to section 5.2.2.2.10 of 3GPP TS 38.214;

[0226] 2. Train in a manner the minimizes the processing of the AI / ML model output before the loss is calculated. Then the ground-truth measurement, for model training, may be defined to match the output of the AI / ML model. Continuing the example from the previous paragraph, the following may be considered:

[0227] a) The ground-truth measurement may be the measured raw channel per time instance for all the A4measured time instances in the prediction window;

[0228] b) The ground-truth measurement may be a computed channel Tx-correlation matrix per time instance for all the A4measured time instances in the prediction window;

[0229] c) The ground-truth measurement may be a computed PMI in a firee-format approximating an eigenvector of the channel Tx-correlation, per time instance forall the N4measured time instances in the prediction window;

[0230] d) The ground-truth measurement may be a computed PMI following the 3GPP Rel-16 type II CSI codebook per time instance for all the 1V4measured time instances in the prediction window; and / or

[0231] e) The ground-truth measurement may be the computed PMIs following the 3GPP Rel-18 “typeII-Doppler-rl8” codebook for all the 1V4measured time instances in the prediction window. This is the same format as in strategy 1.

[0232] Combination of strategies are also possible, e.g., a loss function taking both strategy 1 and 2 into account, e.g., by being a combination of two different loss functions, each taking a separate model output / ground-truth measurement format.

[0233] Training may also happen in a staged approach were one strategy is used initially and then a second strategy is used, e.g., training for channel prediction in a first stage (strategy 2-a) and then for end-to-end performance of the feature (strategy 1).

[0234] The collection of these ground-truth measurements may be the same. For example, one may derive the ground-truth measurements described for strategy 2-b trough 2-e, from measurements of the ground-truth measurement described in strategy 2-a, if those are of high enough quality.

[0235] A ground-truth measurement for Al CSI performance monitoring

[0236] A ground-truth measurement for performance monitoring may be obtained using measured CSI(s) for the corresponding one or multiple prediction time instances (e.g., the measured CSIs on the N4 time slots in the prediction window as shown in FIG. 1).

[0237] In some embodiments, the ground-truth measurement format for performance monitoring may follow the ground-truth measurement format for model training.

[0238] However, for the UE-side CSI prediction functionality, the predicted CSI(s) for 1V4prediction time instances may be reported, from the UE 22 to the network node 16, in the format of “typeII-Doppler-rl8” codebook. Hence, to acquire accurate performance metrics for the reported CSI, it may be beneficial to define the ground-truth measurement format, for performance monitoring, as “typeII-Doppler-rl8” codebook for calculating the intermediate KPI, regardless which model output format is used for the AI / ML model. The ground-truth measurement in the format of “typeII-Doppler-rl8” codebook may be obtained by measuring CSI on the 1V4time instances in the prediction window, and compressing the measured CSIs on the 1V4prediction time instances according to section 5.2.2.2.10 of 3GPP Technical Standard (TS) 38.214.

[0239] A training data sample for Al CSI prediction model trainingA training data sample for Al CSI prediction model training may consists of {a model input sample, the corresponding ground-truth measurement}, where an example of a model input sample is the measured CSIs on the aperiodic K CSI-RS resources within the observation window as shown in FIG. 1, while an example of the corresponding ground-truth measurement is the measured CSIs on the 1V4time slots in the prediction window as shown in FIG. 1. Details, with examples, around input- and ground-truthformats have been discussed above.

[0240] An example of developing a performance metric estimator, which directly outputs the estimated performance metric for performance monitoring for an associated Al CSI prediction inference output.

[0241] Various performance metrics may be used for performance monitoring. For example, final KPI, such as system level throughput, or intermediate KPI, such as SGCS or NMSE may be used. In some embodiments, intermediate KPI may be used as the performance metric, as an example. It should be noted that some embodiments also apply to other types of performance metrics and performance metric estimators.

[0242] Definition of intermediate KPI per model inference

[0243] An intermediate KPI per model inference may be defined as the SGCS or NMSE between the predicted CSI and the ground-truth measurement for performance monitoring, where the predicted CSI is generated based on the model inference output and the associated ground-truth measurement is obtained using measured CSI(s) for the corresponding one or multiple 1V4time instances in the prediction window.

[0244] Develop a performance metric estimator

[0245] At the UE 22, a performance metric estimator may be used to directly generate an estimated performance metric using the information related to the CSI measurements in the observation window (i.e., associated to the Al CSI prediction model input) and / or the predicted CSI (i.e., associated to the Al CSI prediction model output), without the need of ground-truth measurement (i.e., without the need of measuring CSIs for the corresponding one or multiple prediction time instances).

[0246] This performance metric estimator algorithm / model may be developed / trained together with the Al CSI prediction model at the UE-side. An example is shown below with 4 steps.

[0247] 1. Training an Al CSI prediction model:

[0248] The UE may collect a first training dataset that consists of multiple training data samples of {CSI measurements in the observation window (model input), CSImeasurements in the prediction window (ground-truth measurement for model training)} and uses the training dataset to train an Al CSI prediction model.

[0249] 2. Generating predicted CSI samples with the trained model:

[0250] The UE 22 may feed each of the model input samples to the trained Al CSI prediction model to generate the corresponding predicted CSI samples.

[0251] 3. Deriving ground-truth measurements for performance metric:

[0252] The UE 22 may use the predicted CSI samples and the associated ground-truth measurements for performance monitoring to derive the corresponding ground-truth measurements for performance metric (e.g., intermediate-KPI (SGCS, NMSE, etc.) samples).

[0253] 4. Training a performance metric estimator:

[0254] Based on the derived ground-truth measurements for performance metric, and some model input, a performance metric estimator algorithm / model, e.g., an intermediate KPI estimator, may be developed / trained using one of the following methods:

[0255] • In some embodiments, the UE may create a second training data set that consists of multiple training data samples of {CSI measurements in the observation window (Al CSI prediction model input), ground-truth measurements for performance metric}, which may be used for training an estimator, whose algorithm / model input is the CSI measurements in the observation window, and algorithm / model output is an estimated intermediate KPI, or

[0256] • In some embodiments, the UE may create a third training data set that consists of multiple training data samples of {predicted CSI (Al CSI prediction model output), ground-truth measurements for performance metric}, which may be used for training an estimator, whose algorithm / model input is the predicted CSI (Al CSI prediction model output) and algorithm / model output is an estimated intermediate KPI, or • In some embodiments, the UE may create a fourth training data set that consists of multiple training data samples of {CSI measurements in the observation window (Al CSI prediction model input), predicted CSI (Al CSI prediction model output), ground-truth measurements for performance metric }, which may be used for training an estimator, whose algorithm / model input comprises of {CSI measurements in the observation window(AI CSI prediction model input) and predicted CSI (Al CSI prediction model output)}, and the algorithm / model output is an estimated intermediate KPI.

[0257] • The above three examples may serve as examples of the main framework (main training data samples and model input-output) for the intermediate KPI estimator.Additional information may be relevant or beneficial to train the model such as the distance between the observation window and the prediction window (e.g., the distance between the last measurement occasion in the observation window and the first prediction occasion in the prediction window or the alike), the configured codebook parameters, the UE 22 (estimated) speed, past estimated or measured performance metric values, Acknowledgement (ACK) and / or Negative ACK (NACK) history from downlink transmissions, etc.

[0258] Definition of performance monitoring metric(s)

[0259] Examples of performance monitoring metric(s) for UE-sided CSI prediction use case include:

[0260] • Option 1 : intermediate KPI per monitoring data sample; and / or

[0261] • Option 2: Statistics of the intermediate KPI over Ml monitoring data samples, e.g., the mean and variance of the intermediate KPI associated to the collected monitoring data samples within a time window; percentage of monitoring data samples that fulfill certain condition(s); x.th percentile, etc. In some embodiments, the statistic may be in terms of intermediate KPIs of a subset of monitoring data samples (e.g., the first prediction occasion and the last prediction occasion in each monitoring window).

[0262] Option 1 may be useful when the UE 22 may feedback the predicted CSI and the corresponding intermediate KPI to the network node 16 in the same CSI report. This enables network node 16 to know the quality of the received predicted CSI, thus, making better decisions on scheduling and downlink transmission for the future time slots. For example, if the UE 22 reports uncertainty with the quality of the prediction, the network node 16 may favor scheduling SU-MIMO, over MU-MIMO, for this particular UE 22.

[0263] An intermediate KPI of a single monitoring data sample may not indicate whether the CSI prediction model is functioning properly. A low SGCS value of a monitoring data sample may be due to the reason that the channel is hard to predict (i.e., the SGCS value is expected to be low) or the Al model is not working well (i.e., the SGCS value should be high). Hence, to enable the network node 16 to make reliable functionality -based LCM operation decisions, e.g., fallback to legacy CSI reporting and root cause analysis in case the UE performance drop, statistics of the intermediate KPI over multiple monitoring data samples, i.e., Option 2, may be needed.

[0264] Definition of performance monitoring output

[0265] The performance monitoring output is determined based on performance metric, and additionally, baseline and / or threshold criterion if configured. Examples ofperformance monitoring output include:

[0266] • An intermediate KPI range indicator, indicating whether the intermediate KPI of a single monitoring data sample fulfills a certain condition. For example, two SGCS threshold values, thr l and thr_2, may be configured / predetermined per MIMO layer or common for all MIMO layers for performance monitoring. The intermediate KPI range indicator is set to 00, 01 or 10, if the SGCS of this layer for the monitoring data sample is within in the range of [0, thr l ], [thr l , thr_2] or [thr_2, 1], respectively. Note that the min value of 0 and max value of 1 here may serve as an example. The min value and the max value may also be determined, e.g., in the standard text.

[0267] • An intermediate KPI statistics range indicator, indicating whether the statistics of the intermediate KPI over multiple samples in a monitoring window fulfills a certain condition. For example, two SGCS threshold values, mean thr l and mean_thr_2, are configured / predetermined per MIMO layer or common for all MIMO layers for performance monitoring. The intermediate KPI statistics range indicator may be set to 00, 01 or 10, if the mean SGCS of this layer for the monitoring data sample is within in the range of [0, thr l ], [thr l , thr_2] or [thr_2, 1], respectively. Note that the min value of 0 and max value of 1 here may serve as an example. The min value and the max value may also be determined, e.g., in the standard text.

[0268] A performance monitoring result(s) report (denoted by CSI-PM report) may include one or more performance metric(s) and / or one or more performance monitoring output(s). In some embodiments, the one or more performance metric(s) and / or one or more performance monitoring output(s) may be referred to as performance monitoring results. The performance monitoring results may be derived based on one or more intermediate KPIs, which may be generated using an intermediate KPI estimator described above.

[0269] Configuration and CSI processing criteria for computing a CSI-inference report

[0270] As shown in FIG. 11, a CSI-inference report configuration points to a channel measurement resource (CMR) configuration that contains a set of aperiodic CSI-RS burst consisting of K CSI-RS resources, with K>1. The measurements on the K CSI-RS resources are used for creating a model input for CSI prediction. The UE 22 uses the model input and the AI / ML CSI prediction model / algorithm to generate a CSI-inference report.

[0271] In FIG. 12, the CSI-inference report is triggered via a DCI signal such that the CSI-inference report is reported by the UE 22 in slot n. The KpP / SP CSI-RS occasions(e.g., the KpP / SP CSI-RS occasions in green slots) are used for channel measurement for the purpose of generating the predicted CSI. The measurements on the KpP / SP CSI-RS occasions are used for creating a model input for CSI prediction. The UE 22 uses the model input and the AI / ML CSI prediction model / algorithm to generate a CSI-inference report.

[0272] Hence, the CSI-inference reporting for a CSI prediction feature using UE side AI / ML model use case is similar to the legacy 3GPP Rel-18 predicted CSI reporting when the codebookType is set to 'typeII-Doppler-rl8', except for that a dedicated hardware may be used at UE 22 for model inference. In some embodiments, a new CPU type, e.g., AI-CPU, may be defined for a CSI-inference report. Otherwise, the legacy CPU type (as defined in 3GPP TS 38.214) may be reused for the CSI-inference report. For both the cases of CSI-inference reporting in FIGS. 11 and 12, the occupation time of the consumed CPU(s) / AI-CPU(s) is from the first symbol after the PDCCH triggering the CSI-inference report, until the last symbol of the scheduled PUSCH carrying the CSI-inference report.

[0273] In the following, the term “processing unit (PU)” may be used, to describe the CSI processing criteria for computing a CSI-PM report with / without a CSI-inference report. The processing unit may be a legacy CPU, or a new CPU type (e.g., ALCPU) defined for Al based CSI reporting features.

[0274] Configuration and CSI processing criteria for computing a CSI-PM report As described before, at the UE 22, a performance metric estimator, such as an intermediate KPI estimator algorithm / model, may be used to directly generate an estimated intermediated KPI using the information related to the CSI measurements in the observation window (i.e., associated to the Al CSI prediction model input) and / or the predicted CSI (i.e., associated to the Al CSI prediction model output), without the need of ground-truth measurement (i.e., without the need of measuring CSIs for the corresponding one or multiple prediction time instances).

[0275] Case 1: the UE 22 is configured to report CSI-inference and CSI-PM jointly via a single CSI report.

[0276] In the example shown in FIG. 13, the CSI-inference and the CSI-PM are scheduled to be jointly reported over the same UL slot / PUSCH.

[0277] The UE 22 uses the CSI measurements on the K CSI-RS resources to create an Al CSI prediction model input sample. The UE 22 uses the model input sample and the AI / ML CSI prediction model / algorithm to generate a CSI-inference report (i.e., predicted CSI).

[0278] In the example shown in FIG. 14, the CSI-inference and the CSI-PM are scheduledto be jointly reported over the same UL slot / PUSCH.

[0279] The UE 22 uses the CSI measurements on the Kpoccasions of the P or SP CSI-RS resource to create an Al CSI prediction model input sample. The UE 22 uses the model input sample and the AI / ML CSI prediction model / algorithm to generate a CSI-inference report (i.e., predicted CSI).

[0280] Compared to the CSI processing for generating a CSI-inference report shown in FIGS. 11 and 12, the additional processing steps required for generating a CSI-PM report include computing the performance monitoring result estimates in the examples in FIGS.

[0281] 13 and 14. Depending on the design of the performance metric estimator, the UE 22 may:

[0282] • Option 1 : Use the Al CSI prediction model input sample and the performance metric estimator model / algorithm, together with one or more network node 16 configured conditions (e.g., network node RS related configuration, network node Tx power related configuration, network node antenna port patten related configuration, etc.) if needed, to generate a CSI-PM report;

[0283] • Option 2: Use the Al CSI-inference report and the performance metric estimator model / algorithm, together with one or more network node configured conditions if needed, to generate a CSI-PM report; and / or

[0284] • Option 3 : Use the Al CSI prediction model input sample, the Al CSI-inference report, and the performance estimator model / algorithm, together with one or more network node configured conditions if needed, to generate a CSI-PM report.

[0285] In some embodiments, for CSI predicting using UE 22-side AI / ML model, the UE 22 is configured to report CSI-inference and CSI-PM jointly via a single CSI report, where the CSI report configuration points to a CMR configuration that contains a set of aperiodic CSI-RS burst consisting of K CSI-RS resources, with K>1.

[0286] In some embodiments, for CSI predicting using UE-side AI / ML model, the UE 22 is configured to report CSI-inference and CSI-PM jointly via a single CSI report, where the CSI report configuration points to a CMR configuration that contains a single periodic or semi-persistent CSI-RS resource with periodicity P.

[0287] In some embodiments, the number of occupied PUs for the single report carrying both CSI-inference and CSI-PM, denoted as OPU-CSI -PM-inferenceis defined as

[0288] • PU -CSI -PM-infrence O

[0289]

[0290] pu-CSI-inference E Opu-delta> ^nd the

[0291] Opu-csi-PM-infrence PU(s) are occupied from the first symbol after the PDCCH triggering the joint CSI-inference and CSI-PM report, until the last symbol of the scheduled PUSCH carrying the single CSI report. Here, OPU-CSI -inferenceis the numberof occupied PUs when the UE 22 is triggered for the CSI-inference report only. The value of OPU-deitaaccounts for the additional processing needed for generating the CSI-PM report. The value of OPU-deita, which may depend on UE capability, is specified in 3GPP specifications. Note that instead of OPU-deita, the UE capability may also indicate information on the supported OPU-CSI-PM-infrence

[0292] o In some embodiments, OPU-deUa= S ■ OPU-CSI-inferencewherein S is a positive scaling factor;

[0293] o In some embodiments, OPU-deita= n, where n is a positive number. OPU-deitaaccounts for the extra delay for calculating the performance metric for performance monitoring;

[0294] o In some embodiments, OPU-deitais determined by scaling the number of PMI prediction time occasions 1V4by a positive integer Ydeitawhere Ydeitais a positive number that depends on UE capability. In some embodiments, OPU-deita= max Ydeita■ N4, 0m) where Omis a maximum bound on the number of PUs for calculating the performance metric for performance monitoring;

[0295] o In some embodiments, OPU-deitais determined by scaling the number of AP CSI-RS resources K by a positive integer Ydeitawhere Ydeitais a positive number that depends on UE capability. In some embodiments, OPU-deita= max Ydeita■ K, 0m) where Omis a maximum bound on the number of PUs for calculating the performance metric for performance monitoring; and / or

[0296] o In some embodiments, OPU-deitais determined by scaling the number of P or SP CSI-RS resource occasions Kpby a positive integer Ydeitawhere Ydeitais a positive number that depends on UE capability. In some embodiments, OPU-deita= max(Ydeita■ Kp, 0m) where Omis a maximum bound on the number of PUs for calculating the performance metric for performance monitoring.

[0297] In some embodiments, the number of occupied PUs for the single report carrying both CSI-inference and CSI-PM, denoted as OPU-CSI-PM-inferenceis defined to consists of two parts, where the first part is associated to processing for CSI measurements, and the second part is associated to processing for generating the single CSI report based on the CSI measurements.

[0298] • In some embodiments, the PU(s) in the first part,

[0299] O

[0300]

[0301] pu-csi-PM-in / rence-Parti, is / are occupied from the first symbol after the PDCCH triggering the joint CSI-inference and CSI-PM report, until the last symbol of the CSI-RSresource based on the CMR configuration. In some cases, the OPU-CSI_pM-infrence -PartlPUs are occupied until the last symbol of the CSI-RS resource based on the CMR configuration plus an offset. The offset may account for, for example, time needed for estimating the channel.

[0302] • In some embodiments, the PU(s) in the second part,

[0303] 0

[0304]

[0305] pu-cs / -PM-in / rence-Part2< is / are occupied from the last symbol of the CSI-RS resource based on the CMR configuration, until the last symbol of the scheduled PUSCH carrying the single CSI report. In some cases, the OPU-CSI -PM-infrence -Part2PUs are occupied starting from the last symbol of the CSI-RS resource based on the CMR configuration plus and offset. The offset in part 2 may be the same as the offset in part 1.

[0306] • As an example, O

[0307]

[0308] PU_P$I —PM—infrence—Partl Opu— CSI— measurements’ where 0PU-CSI -measurementsis the number of occupied PUs when the UE 22 is triggered for performing the associated CSI measurements, and 0PU-CSI -PM-infrence -Part2= Opu_csi-inference + 0PU-deita, where 0PU-deUais the additional number of PU(s) needed for generating the CSI-PM report. The value of 0PU-deitais specified in 3GPP specifications. The value of 0PU-deitamay depend on UE capability.

[0309] o In one example, 0PU-delta= S ■ 0PU-CSI -inferencewherein S is a positive scaling factor.

[0310] • As another example, 0P[j—csi —PM—infrence—Partl

[0311] 0Pu —csi —measurements’ where OPu_Pg]_measrUgmgnfSis the number of occupied CPUs when the UE 22 is triggered for performing the associated CSI measurements, and O

[0312]

[0313] pu-CSI -PM-infrence-Part2 Opu-CSI-inference-report + Opu-CSI -PM-report> where Opu -csi -inference -report is the number of occupied PUs for the UE 22 to compute the CSI-inference report using the CSI measurements, and 0PU-CSI -PM-reportis the number of occupied PUs for computing the associated CSI-PM report. The values of

[0314] O

[0315]

[0316] pu- CSI- measurements ■> Opu-CSI-inference-report> ^nd OPu_Pgj -pM-report are Specified in 3GPP specifications. The values may depend on UE capability.

[0317] o In one example, 0Pu — si —measurements U K wherein U is a positive scaling factor value the UE 22 reports as part of UE capability reporting, and K is the number of CSI-RS resources configured in the CMR configuration;

[0318] Opu -csi -inference -report = Vi ' K + V2■ N4wherein

[0319]

[0320] and V2are non-negative scaling factor values the UE reports as part of UE capability reporting, and 1V4is the number of prediction time instances; 0PU-CSI -PM-report= ■ K + S2■ / V4wherein and S2arenon-negative scaling factor values the UE reports as part of UE capability reporting. o In another example, 0PU-CSI-measurements = U ■ K wherein U is a positive scaling factor value the UE reports as part of UE capability reporting, and K is the number of CSI-RS resources configured in the CMR configuration;

[0321] O

[0322]

[0323] pu-CSI-inference-report $1 ' ^CPU-CSI-inf erence > and Opu-csi -pM-report S20 CPU —csi -inference’ wherein 0CPU-CSI -inferenceis the number of CPUs for legacy CSI reporting when the ‘reportQuantity’ is set to 'cri-RI-PMI-CQI' and with codebookType set to 'typeII-Doppler-rl8', and S and S2are positive scaling factor values the UE reports as part of UE capability reporting.

[0324] o For the above examples, additional offset, either positive or negative, may be added on top of at least one of 0PU-CSI-measurements,

[0325] 0 pu— csi —inf erence— report and 0 P —CSI —PM— report • For example, if an offset is added to 0Pu —csi —measurements’ thenOpy_ csi— measurements K 71, where 71 is a non-zero offset.

[0326] Due to additional processing steps, the value of OPU-CSI -PM-inferenceis larger than OPU-CSI -inference. In some occasion, it may be possible that the UE 22 is configured / indicated to report CSI (and the monitoring results) between

[0327] Opu-csi-inference and OPU-CSI -PM-inferenceafter receiving the indication. In some embodiments, the UE 22 may ignore the reporting indication completely and not transmitting both the CSI report and the monitoring results. In some embodiments, the UE 22 may transmit only the CSI report and not the monitoring results.

[0328] In some embodiments, the measurement / computation and the reporting of the single report carrying both CSI-inference and CSI-PM are indicated / triggered via different DCIs as shown in FIG. 15, for the case of using aperiodic CSI-RS resources. As shown in the figure, DCI1 triggers measurement / computation related to the CSI-inference and CSI-PM reports. DCI2 triggers the joint reporting of the CSI-inference and CSI-PM reports. In this embodiment, the number of occupied PUs for the single report carrying both CSI-inference and CSI-PM, denoted a

[0329]

[0330] s OPu-csi—PM—inference-> where 0Pu— csi —PM— inf erence is associated to processing for CSI measurements and computation of the CSI-inference and CSI-PM reports. In some embodiments, the PU(s) is / are occupied from the first symbol of the PDCCH carrying DCI1 until CSI-RS resource K in the aperiodic CSI-RS burst (referring to the example in FIG. 15). In some embodiments, the PU(s) is / are occupied from the first symbol of the PDCCH carrying DCI1 until CSI-RS resource K in theaperiodic CSI-RS burst plus an offset (referring to the example in FIG. 15). The offset may account for, for example, time needed for any one or more of estimating the channel, computing the CSI-inference reporting content, and computing CSI-PM reporting content. In some embodiments, the UE 22 expects that DCI2 is received after the PU occupation period such that the UE 22 has the computed CSI-inference and CSI-PM reporting contented ready to report in slot n2. In some embodiments, the UE 22 expects that slot n2 is after the PU occupation period such that the UE 22 has the computed CSI-inference and CSI-PM reporting contented ready to report in slot n2.

[0331] Another example case is shown in FIG. 16, where the CSI report configuration points to a CMR configuration that contains a single periodic or semi-persistent CSI-RS resource with periodicity P, and a UE 22 is configured to report a single CSI report carrying both CSI-inference and CSI-PM. DCI1 triggers measurement / computation related to the CSI-inference and CSI-PM reports. DCI2 triggers the joint reporting of the CSI-inference and CSI-PM reports. Similar to the AP CSI-RS resource based example case shown in FIG. 15, in this example, the number of occupied PUs for the single report carrying both CSI-inference and CSI-PM, denoted as OPU-CSI-PM-inference, where O

[0332]

[0333] pu-csi-PM-in / erence is associated to processing for CSI measurements and computation of the CSI-inference and CSI-PM reports. In some embodiments, the PU(s) is / are occupied from the first symbol of the PDCCH carrying DCI1 until in the Kp-th occasion of the P or SP CSI-RS resource (referring to the example in FIG. 16). In some embodiments, the PU(s) is / are occupied from the first symbol of the PDCCH carrying DCI1 until the Kp-th occasion of the P or SP CSI-RS resource plus an offset (referring to the example in FIG.

[0334] 16). The offset may account for, for example, time needed for any one or more of estimating the channel, computing the CSI-inference reporting content, and computing CSI-PM reporting content. In some embodiments, the UE 22 expects that DCI2 is received after the PU occupation period such that the UE 22 has the computed CSI-inference and CSI-PM reporting contented ready to report in slot n2. In some embodiments, the UE 22 expects that slot n2 is after the PU occupation period such that the UE 22 has the computed CSI-inference and CSI-PM reporting contented ready to report in slot n2.

[0335] Case 2: the UE 22 is configured to report CSI-PM without reporting the associated CSI-inference report.

[0336] In the example shown in FIG. 17, the CSI-PM is scheduled to be reported over the UL slot n2 without reporting the corresponding CSI-inference report.Depending on the design of the performance metric estimator (e.g., an intermediate KPI estimator), the UE 22 may:

[0337] • Option 1 : Use the CSI measurements on the K CSI-RS resources to create an Al CSI prediction model input sample, then, use the Al CSI prediction model input sample and the performance metric estimator model / algorithm, together with one or more network node configured conditions if needed, to generate a CSI-PM report;

[0338] • Option 2: Use the CSI measurements on the K CSI-RS resources to create an Al CSI prediction model input sample, then, use the model input sample and the AI / ML CSI prediction model / algorithm to generate a CSI-inference output (i.e., predicted CSI). Finally, it uses the Al CSI-inference output and the performance metric estimator model / algorithm, together with one or more network node configured conditions if needed, to generate a CSI-PM report; and / or

[0339] • Option 3 : Use the CSI measurements on the K CSI-RS resources to create an Al CSI prediction model input sample, then, use the model input sample and the AI / ML CSI prediction model / algorithm to generate a CSI-inference output (i.e., predicted CSI). Finally, it uses the Al CSI prediction model input sample, the Al CSI-inference output, and the performance metric estimator model / algorithm, together with one or more network node configured conditions if needed, to generate a CSI-PM report.

[0340] For option 1, key steps of the CSI-PM processing include CSI-measurements and performance metric estimation. For option 2 and option 3, key steps of the CSI-PM processing include CSI-measurements, CSI prediction inference operation, and performance metric estimation.

[0341] In some embodiments, for CSI predicting using UE-side AI / ML model, the UE 22 is configured to report a CSI-PM report without reporting the corresponding CSI-inference report, where the CSI-PM report configuration points to a CMR configuration that contains a set of aperiodic CSI-RS burst consisting of K CSI-RS resources, with K>1.

[0342] In some embodiments, the number of occupied PUs for the CSI-PM report is defined as OPU-CSI-PM, and the OPU-CSI-PMPU(s) are occupied from the first symbol after the PDCCH triggering the CSI-PM report, until the last symbol of the scheduled PUSCH carrying the CSI-PM report. The value of OPU-CSI-PMis specified in 3GPP specifications. The value of OPU-CSI-PMmay depend on UE capability.

[0343] • In some embodiments, the value of OPU-CSI-PMmay depend on the OPU-csi -inference value where OPU-CSI-inferenceis the number of occupied PUs when the UE 22 is triggered for the CSI-inference report only. In an example of implementation,Opu-csi-PM = S ’ 0PU-CSI-inference, wherein S is a positive scaling factor.. In another example of implementation, 0PU-CSi_PM= OPU-CSi_in^erence+ OPU-cieita.

[0344] In some embodiments, the number of occupied PUs for the CSI-PM report, denoted as 0PU-CSI-PMis defined to consists of two parts, where the first part is associated to processing for CSI measurements over the measurement window, and the second part is associated to processing for generating the CSI-PM report based on the CSI measurements.

[0345] • In some embodiments, the PU(s) in the first part, OPU-CSI-PM-Partl, is / are occupied from the first symbol after the PDCCH triggering the CSI-PM report, until the last symbol of the CSI-RS resource based on the CMR configuration. In some cases, the OPu-csi-PM-Parti PUs are occupied until the last symbol of the CSI-RS resource based on the CMR configuration plus an offset. The offset may account for, for example, time needed for estimating the channel.

[0346] • In some embodiments, the PU(s) in the second part, OPU-CSI-PM-Part2, is / are occupied from the last symbol of the CSI-RS resource based on the CMR configuration, until the last symbol of the scheduled PUSCH carrying the CSI-PM report. In some cases, the OPU-CSI-PM-Part2PUs are occupied starting from the last symbol of the CSI-RS resource based on the CMR configuration plus and offset. The offset in part 2 may be the same as the offset in part 1.

[0347] • In Some embodiments, 0 pu—csi —PM— Parti PU— CSI —PM— measurements’ where OPU-CSI-PM-measruementsis the number of occupied CPUs when the UE 22 is triggered for performing the associated CSI measurements, and OPU-CSI-PM-Part2= 0Pu —csi —PM —report’ where 0PPj — si —PM —report the number of occupied PUs for computing the CSI-PM report using the obtained CSI measurements. The values of 0

[0348]

[0349] pu —csi —PM— measurements^ and 0 py—csi —PM— report are specified in 3GPP specifications. The values may depend on UE capability.

[0350] o In some embodiments, OPU-csl-PM-measurements= U ■ K wherein U is a positive scaling factor value the UE 22 reports as part of UE capability reporting, and K is the number of CSI-RS resources configured in the CMR configuration;

[0351] Opu-csi -PM-report =

[0352]

[0353] ■ K + S2■ N4wherein

[0354]

[0355] and S2are non-negative scaling factor values the UE reports as part of UE capability reporting, K is the number of CSI-RS resources configured in the CMR configuration, and N4is the number of prediction time instances.o In some embodiments, the value of 0PU-CSI-PM-reportmay depend on the OPU-CSI-inferencevalue where OPU-CSI-inferenceis the number of occupied PUs when the UE 22 is triggered for the CSI-inference report only. In an example of implementation, 0PU-CSI-PM-report= S ■ OPU-CSI-inference, wherein S is a positive scaling factor. In another example of implementation, 0PU-CSI-PM-report=

[0356] @

[0357]

[0358] PU -CSI -inference d” OPu_cieif-a.

[0359] Case 3: the UE 22 is configured to report CSI-inference and CSI-PM via separate CSI reports.

[0360] Case 3a: separate CSI reports triggered by separate DCIs

[0361] FIG. 18 shows an example, where a CSI-PM report is triggered together with a CSI-inference report using two separate DCI signaling. Both DCIs are sent before the first CSI-RS resource in the aperiodic CSI-RS burst.

[0362] FIG. 19 shows another example where a CSI-PM report is triggered together with a CSI-inference report using two separate DCI signaling. Different from the example shown in FIG. 18, the DCI for triggering the CSI-PM report is sent after the last CSI-RS resource in the aperiodic CSI-RS burst.

[0363] In both examples of FIGS. 18 and 19, the UE 22 is scheduled to report CSI-PM report after the CSI-inference report. This may correspond to the following designs of the performance metric estimator, which requires Al CSI prediction inference output to generate an estimated intermediate KPI:

[0364] • Use the Al CSI prediction inference output and the Performance metric estimator model / algorithm, together with one or more network node configured conditions if needed, to generate a CSI-PM report, or

[0365] • Use the Al CSI prediction model input sample, the Al CSI prediction inference output, and the performance metric estimator model / algorithm, together with one or more network node configured conditions if needed, to generate a CSI-PM report.

[0366] For another design option of performance metric estimator, which only requires the CSI measurements on the K CSI-RS resources to estimate the performance metric of the corresponding inference output, the UE 22 may be able to generate the CSI-PM report faster than generating the CSI-inference report. In this case, the UE 22 may be scheduled to report CSI-PM report before the CSI-inference report, or on the same UL slot as for the CSI-inference report.

[0367] FIG. 20 shows an example where a CSI-PM report is triggered together with a CSI-inference report using a single DCI. In this example, the CSI-inference report and theCSI-PM report are scheduled to be fed back on different UL slots. As discussed before, depending on the performance metric estimator design, the UE 22 may be able to generate the estimated performance metric before or after it generates the CSI-inference output. Hence, the CSI-PM report may be configured to be fed back before or after the CSI-inference report.

[0368] In some embodiments, for CSI predicting using UE-side AI / ML model, the UE 22 is configured to report CSI-inference and CSI-PM via separate CSI reports based on separate CSI report configurations.

[0369] • In some embodiments, the CSI-PM report configuration points to a CMR configuration that contains a set of aperiodic CSI-RS burst consisting of K CSI-RS resources, with K>1. In some embodiments, the CSI-PM report configuration points to the corresponding CSI-inference report configuration.

[0370] • In some embodiments, the CSI-PM report is scheduled to be transmitted before the CSI-inference report in time. In some embodiments, the CSI-PM report is scheduled to be transmitted after the CSI-inference report in time. In some embodiments, the CSI-PM report is scheduled to be transmitted on the same UL slot as for the CSI-inference report.

[0371] o In some embodiments, the UE 22 has indicated to the network node 16, e.g., via capability reporting, whether it is able to send the CSI-PM report before it sends the CSI-inference report.

[0372] The PUs required for CSI-inference report and CSI-PM report may be counted separately. Since at least the occupied PU(s) for measuring the K CSI-RS resources may be reused for both CSI-inference and CSI-PM report, the occupied PU(s), used for measurement, may not be counted twice.

[0373] For the CSI-inference report, 0PU-CSI-inference, PUs may be occupied from the first symbol after the PDCCH triggering the CSI-inference report, until the last symbol of the scheduled PUSCH carrying the CSI-inference report.

[0374] In some embodiments, for CSI prediction using UE-side AI / ML model, when the UE 22 is triggered for a CSI-PM report together with a corresponding CSI-inference report, the number of occupied PUs for this CSI-PM report, denoted as 0PU-CSI-PM, is defined a

[0375]

[0376] s 0PU-CSi_PM=0PU-CSi_PM-repOrt, where 0PU-CSi_PM-repOrtis the number of occupied PUs for computing the CSI-PM report using the obtained CSI measurements. In some embodiments, 0PU-CSI-PM-report= ■ K + S2• N4wherein and S2are nonnegative scaling factor values the UE reports as part of UE capability reporting, K is thenumber of CSI-RS resources configured in the CMR configuration, and 1V4is the number of prediction time instances.

[0377] In some embodiments, if the CSI-PM report is an aperiodic CSI report, the OPU-CSI-PM PU(s) are occupied from the last symbol of the CSI-RS resource based on the corresponding CMR configuration, until the last symbol of the scheduled PUSCH carrying the CSI-PM report.

[0378] In some embodiments, if the CSI-PM report is an aperiodic CSI report, the Opu-csi-PM CPU(s) are occupied from the first symbol after the PDCCH triggering the CSI-PM report, until the last symbol of the scheduled PUSCH carrying the CSI-PM report.

[0379] In some embodiments, for CSI predicting using UE-side AI / ML model, if the UE 22 is configured to report CSI-inference and CSI-PM via separate CSI reports based on separate CSI report configurations, then the number of occupied PUs may depend on whether or not the CSI-PM report is scheduled before or after the CSI-inference report. In one example the number of occupied PUs for this CSI-PM report, denoted as

[0380] OPU-CSI-PM, is defined as:

[0381] • 0PU-CSi-PM=0PU-CSi-PM-repOrt + OPU-CSI-inference, where

[0382] Opu-csi-PM-report is the number of occupied PUs for computing the CSI-PM report using the obtained CSI measurements and OPU-CSI-inferenceis the number of occupied PUs for computing the CSI-inference report using the obtained CSI measurements. If the CSI-PM report is scheduled before the CSI-inference report; and / or

[0383] • 0PU-CSI_pM=0PU-CSI_pM-report, where 0PU-CSI_pM-repOrtis the number of occupied PUs for computing the CSI-PM report using the obtained CSI measurements. If the CSI-PM report is scheduled after the CSI-inference report.

[0384] Additional information on the monitoring report

[0385] As the performance metric / intermediate KPI determination is based on a metric / KPI predictor model, the performance accuracy depends on the similarity between the dataset used to train the metric / KPI predictor model and the actual measurement of the CSI-RS in the measurement window, e.g., there may be dataset distribution / characteristic mismatch between the dataset used to train the metric / KPI predictor model and the actual measurement of the CSI-RS in the measurement window (a.k.a. CSI prediction model inputs). In addition, the accuracy of the metric / KPI predictor model may also depend on the performance metric / intermediate KPI range, etc. For example, the accuracy between a first performance metric / intermediate KPI value and the second performancemetric / intermediate KPI value may be different compared to the accuracy between a second performance metric / intermediate KPI value and the third performance metric / intermediate KPI value. Further, the accuracy may also depend on other factors such as the link quality, UE speed, number of monitoring occasions, other running algorithms at the UE 22, e.g., channel estimation algorithm, other UE-specific conditions, e.g., active antennas, etc. Therefore, in some embodiments, the UE 22 may also report the confidence level of the reported performance monitoring report. In some embodiments, a single confidence level information may be reported per monitoring occasion. In some embodiments, multiple confidence level information may be reported per monitoring occasion, e.g., one for every prediction instance, one for every AI / ML model, etc.

[0386] The bitfield size to indicate the confidence level information may depend on the number of supported intermediate KPI ranges. For example, when 2, 3, or 4 ranges of confidence level being supported, the confidence level bitfield may have a size of 1, 2, or 2 bits respectively. In general, in an alternative, the bitfield size may be defined as [log2C], where C is the number of confident level ranges. Note that the report may have a confidence level that at least, above a certain threshold. The threshold may be defined, e.g., in the specification related to the conformance test or the alike. In some embodiments, such confidence may only be reported together with a CSI-PM report, and no PUs may be occupied for the calculation.

[0387] Some embodiments may include one or more of the following:

[0388] Embodiment Al . A method in a user equipment (UE) configured to communicate with a network node, the method comprising:

[0389] determining one or more performance monitoring results of a channel state information (CSI) prediction feature using an artificial intelligence and / or machine learning (AI / ML) model, the one or more performance monitoring results being determined without one or more ground-truth measurements, the one or more ground-truth measurements being one or more measurements associated with an expected UE prediction; and

[0390] transmitting a CSI performance monitoring (CSI-PM) report to the network node, the CSI-PM report including the one or more performance monitoring results.

[0391] Embodiment A2. The method of Embodiment Al, wherein the method further includes:

[0392] receiving from the network node a CSI prediction performance monitoring report configuration, the one or more performance monitoring results being based on the CSIprediction performance monitoring report configuration.

[0393] Embodiment A3. The method of any one of Embodiments Al and A2, wherein the method further includes:

[0394] determining the CSI-PM report with or without a CSI-inference report.

[0395] Embodiment A4. The method of Embodiment A3, wherein the method further includes:

[0396] determining a quantity of occupied CSI processing units (CPUs) and a CPU occupancy time required for the CSI-PM report with or without the CSI-inference report.

[0397] Embodiment A5. The method of any one of Embodiments A1-A4, wherein the CSI prediction feature includes one or more of predicted precoding matrix indicators (PMI), one or more predicted raw channels, one or more predicted rank indicators (RIs), one or more predicted reference signal receive power (RSRP) values, one or more predicted channel quality indicators (CQIs), one or more predicted CSI reference signal (RS) resource indicator (CRIs), one or more predicted top-K strongest beams, and one or more predicted top-K cells.

[0398] Embodiment A6. The method of any one of Embodiments A1-A5, wherein the AI / ML model uses a model input sample for CSI prediction including multiple CSI-RS measurements spread over time on a set of K aperiodic CSI-RS resources.

[0399] Embodiment A7. The method of any one of Embodiments A1-A6, wherein the AI / ML model is configured to generate a model output sample for CSI prediction including predicted CSI for one or more prediction time instances.

[0400] Embodiment A8. The method of any one of Embodiments A1-A7, wherein the method further includes:

[0401] training the AI / ML model using a ground-truth measurement obtained using measured CSI for corresponding one or more prediction time instances.

[0402] Embodiment A9. The method of Embodiment A8, wherein the method further includes:

[0403] performing performance monitoring using the ground-truth measurement.

[0404] Embodiment A10. The method of any one of Embodiments A1-A9, wherein the method further includes one or both of:

[0405] determining a training data sample for a prediction of the AI / ML model, the training data sample including a model input sample and a corresponding ground-truth measurement; and

[0406] determining an estimated performance metric for an associated predictioninference output using a performance metric estimator, the estimated performance metric including an intermediate key performance indicator.

[0407] Embodiment All. The method of any one of Embodiments A1-A10, wherein: the UE is configured to report CSI-inference and CSI-PM via separate CSI reports or jointly via a single CSI report; or

[0408] the UE is configured to report CSI-PM without reporting an associated CSI-inference report.

[0409] Embodiment A12. The method of any one of Embodiments Al-All, wherein the one or more ground-truth measurements include one or more of ground-truth measurements for AI-CSI prediction model training, ground-truth measurements for AI-CSI prediction monitoring, ground-truth measurements for a performance metric.

[0410] Embodiment Bl. A user equipment configured to communicate with a network node, the UE being associated with one or more features corresponding to one or more of Embodiments Al -Al 1 and / or the UE being configured to, and / or comprising a radio interface and / or processing circuitry configured to perform one or more steps corresponding to one or more of Embodiments A1-A12.

[0411] Embodiment Cl. A method in a network node configured to communicate with a user equipment (UE), the method comprising:

[0412] transmitting a channel state information (CSI) prediction performance monitoring report configuration to the UE;

[0413] receiving a CSI performance monitoring (CSI-PM) report from the UE, the CSI-PM report including one or more performance monitoring results of a CSI prediction feature associated with an artificial intelligence and / or machine learning (AI / ML) model, the one or more performance monitoring results being determined by the UE without one or more ground-truth measurements, the one or more ground-truth measurements being one or more measurements associated with an expected UE prediction; and performing one or more action based on the CSI-PM report.

[0414] Embodiment C2. The method of Embodiment Cl, the network node being associated with one or more features corresponding or complementary to one or more of Embodiments A2-A11 and / or the network node being configured to, and / or comprising a radio interface and / or processing circuitry configured to perform one or more steps corresponding to or complementary to one or more of Embodiments A2-A12.

[0415] Embodiment DI . A network node configured to communicate with a user equipment (UE), the network node being associated with one or more featurescorresponding to one or both of Embodiments Cl and C2 and / or the network node being configured to, and / or comprising a radio interface and / or processing circuitry configured to perform one or more steps corresponding to one or both of Embodiments Cl and C2.

[0416] As will be appreciated by one of skill in the art, the concepts described herein may be embodied as a method, data processing system, computer program product and / or computer storage media storing an executable computer program. Accordingly, the concepts described herein may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects all generally referred to herein as a “circuit” or “module.” Any process, step, action and / or functionality described herein may be performed by, and / or associated to, a corresponding module, which may be implemented in software and / or firmware and / or hardware. Furthermore, the disclosure may take the form of a computer program product on a tangible computer usable storage medium having computer program code embodied in the medium that may be executed by a computer. Any suitable tangible computer readable medium may be utilized including hard disks, CD-ROMs, electronic storage devices, optical storage devices, or magnetic storage devices.

[0417] Some embodiments are described herein with reference to flowchart illustrations and / or block diagrams of methods, systems and computer program products. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer (to thereby create a special purpose computer), special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0418] These computer program instructions may also be stored in a computer readable memory or storage medium that may direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instruction means which implement the function / act specified in the flowchart and / or block diagram block or blocks.

[0419] The computer program instructions may also be loaded onto a computer or otherprogrammable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0420] It is to be understood that the functions / acts noted in the blocks may occur out of the order noted in the operational illustrations. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality / acts involved. Although some of the diagrams include arrows on communication paths to show a primary direction of communication, it is to be understood that communication may occur in the opposite direction to the depicted arrows.

[0421] Computer program code for carrying out operations of the concepts described herein may be written in an object oriented programming language such as Python, Java® or C++. However, the computer program code for carrying out operations of the disclosure may also be written in conventional procedural programming languages, such as the "C" programming language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer. In the latter scenario, the remote computer may be connected to the user's computer through a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0422] Many different embodiments have been disclosed herein, in connection with the above description and the drawings. It will be understood that it would be unduly repetitious and obfuscating to literally describe and illustrate every combination and subcombination of these embodiments. Accordingly, all embodiments may be combined in any way and / or combination, and the present specification, including the drawings, shall be construed to constitute a complete written description of all combinations and subcombinations of the embodiments described herein, and of the manner and process of making and using them, and shall support claims to any such combination or subcombination.

[0423] It will be appreciated by persons skilled in the art that the embodiments described herein are not limited to what has been particularly shown and described herein above. Inaddition, unless mention was made above to the contrary, it should be noted that all of the accompanying drawings are not to scale. A variety of modifications and variations are possible in light of the above teachings without departing from the scope of the following claims.

Claims

55What is claimed is:

1. A method in a user equipment, UE (22), configured to communicate with a network node (16), the method comprising:generating (SI 14) predicted channel state information, CSI, samples via a trained artificial intelligence, Al, CSI prediction model, the Al CSI prediction model being trained based at least in part on a first training data set of samples of CSI measurements in a prediction window; andtransmitting (SI 16) a CSI performance monitoring, PM, report of CSI-PM results based on a performance metric estimator.

2. The method of Claim 1, wherein training the Al CSI prediction model includes deriving ground truth labels for a performance metric based at least in part on the predicted CSI samples, and training the performance metric estimator based at least in part on the derived ground truth labels.

3. The method of Claim 2, wherein training the performance metric estimator includes obtaining a second training data set of samples of CSI measurements in an observation window.

4. The method of Claim 2, wherein training the performance metric estimator includes obtaining a third training data set of samples of predicted CSI.

5. The method of Claim 2, wherein training the performance metric estimator includes obtaining a fourth training data set of samples of CSI measurements in an observation window and predicted CSI.

6. The method of any of Claims 1-5, further comprising receiving from the network node (16) a CSI-PM report configuration, the transmitted CSI-PM report being based at least in part on the CSI-PM report configuration.

7. The method of any of Claims 1-6, further comprising determining the CSI-PM report with or without a CSI inference report.

8. The method of any of Claims 1-7, further comprising signaling a capability56report, the capability report including a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report.9 The method of Claim 8, further comprising receiving a request for a CSI-PM report, and not computing a CSI-PM report when a number of CPUs is not available.

10. The method of any of Claims 1-9, wherein the UE (22) is configured to report CSI inference reports and CSI PM reports, separately or jointly.

11. The method of any of Claims 1-10, wherein the predicted CSI samples include at least one of a predicted precoding matrix indicator, PMI, a predicted rank indicator, RI, a predicted reference signal received power, RSRP, a predicted channel quality indicator, CQI, a predicted CSI reference signal resource indicator, CRI, a predicted strongest beam of K strongest beams, and a predicted cell of top K cells.

12. The method of any of Claims 1-11, wherein first training data set of samples of CSI measurements include multiple CSI reference signal, RS, measurements spread over time on a set of K aperiodic CSI-RS resources.

13. The method of any of Claims 1-12, further comprising:receiving from the network node (16) a CSI-PM report configuration; applying the received configuration;receiving a trigger to predict and transmit a CSI-PM report based at least in part on the CSI-PM report configuration, without additional resources for measurement of ground truth labels; andtransmitting the CSI-PM report in response to the received trigger.

14. A method in a network node (16) configured to communicate with a user equipment, UE (22), the method comprising:transmitting (SI 10) to the UE (22) a channel state information performance monitoring, CSI-PM, report configuration; andreceiving (SI 12) a CSI PM report of CSI-PM results, the CSI-PM results being based at least on a performance metric estimator, the performance metric estimator being57based at least in part on ground truth labels derived from predicted CSI samples generated by an artificial intelligence, Al, CSI prediction model trained by a first training data set of CSI measurement samples.

15. The method of Claim 14, wherein the CSI-PM report configuration configures the UE (22) to determine the CSI-PM report with or without a CSI inference report.

16. The method of any of Claims 14 and 15, wherein the CSI-PM report configuration configures the UE (22) to determine a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report.

17. The method of any of Claims 14-16, wherein the CSI-PM report configuration configures the UE (22) to report CSI inference reports and CSI PM reports, separately or jointly.

18. The method of any of Claims 14-16, wherein the CSI-PM report configuration configures the UE (22) to report CSI-PM without reporting an associated CSI inference report.

19. The method of any of Claims 14-18, further comprising triggering the UE (22) to compute and transmit the CSI PM report.

20. A user equipment, UE (22), configured to communicate with a network node (16), the UE (22) including processing circuitry (50) configured to:generate predicted channel state information, CSI, samples via a trained artificial intelligence, Al, CSI prediction model, the Al CSI prediction model being trained based at least in part on a first training data set of samples of CSI measurements in a prediction window; andtransmit a CSI performance monitoring, PM, report of CSI-PM results based on a performance metric estimator.

21. The UE (22) of Claims 20, wherein the processing circuitry is configured to58receive from the network node (16) a CSI-PM report configuration, the transmitted CSI-PM report being based at least in part on the CSI-PM report configuration.

22. The UE (22) of any of Claims 20-21, wherein the processing circuitry (50) is configured to determine the CSI-PM report with or without a CSI inference report.

23. The UE (22) of any of Claims 20-22, wherein the processing circuitry (50) is configured to signal a capability report, the capability report including a quantity of CSI processing units, CPUs, and a CPU occupancy time required for the CSI PM report with or without the CSI inference report.

24. The UE (22) of any of Claims 20-23, wherein the processing circuitry (50) is configured to report CSI inference reports and CSI PM reports, separately or jointly.

25. The UE (22) of any of Claims 20-24, wherein the predicted CSI samples include at least one of a predicted precoding matrix indicator, PMI, a predicted rank indicator, RI, a predicted reference signal received power, RSRP, a predicted channel quality indicator, CQI, a predicted CSI reference signal resource indicator, CRI, a predicted strongest beam of K strongest beams, and a predicted cell of top K cells.

26. The UE (22) of any of Claims 20-25, wherein the processing circuitry (50) is further configured to:receive from the network node (16) a CSI-PM report configuration;apply the received configuration;receive a trigger to predict and transmit a CSI-PM report based at least in part on the CSI-PM report configuration, without additional resources for measurement of ground truth labels; andtransmitting the CSI-PM report in response to the received trigger.

27. A network node (16) configured to communicate with a user equipment, UE (22), the network node (16) comprising processing circuitry (36) configured to:transmit to the UE (22) a channel state information performance monitoring, CSI-PM, report configuration; andreceive a CSI PM report of CSI-PM results, the CSI-PM results being based atleast on a performance metric estimator, the performance metric estimator being based at least in part on ground truth labels derived from predicted CSI samples generated by an artificial intelligence, Al, CSI prediction model trained by a first training data set of CSI measurement samples.

28. The network node (16) of Claims 27, wherein the processing circuitry (36) is configured to trigger the UE (22) to compute and transmit the CSI PM report.

Images