Monitoring of predictions at terminal device
A timer-based mechanism for monitoring predictive RRM models in wireless networks addresses performance degradation issues by detecting failures and switching to fallback configurations, improving the accuracy and responsiveness of predictive RRM.
Patent Information
- Application Number
- PCT/EP2026/061827
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-09
- Filing Date
- 2026-04-24
- Publication Date
- 2026-11-12
AI Technical Summary
Modern wireless networks face challenges in efficiently monitoring the performance of predictive Radio Resource Management (RRM) models due to model drifts and non-stationary data distributions, leading to suboptimal decision-making in measurement and event predictions.
Implementing a method for terminal devices to monitor predictive model performance using a timer-based mechanism that detects prediction failures by comparing predicted outcomes with ground truth measurements, and switching to a fallback configuration if the model fails.
Enhances the responsiveness to model degradation, reducing false positives and ensuring timely adjustments to maintain accurate predictive RRM performance.
Smart Images

Figure EP2026061827_12112026_PF_FP_ABST
Abstract
Description
MONITORING OF PREDICTIONS AT TERMINAL DEVICEFIELD
[0001] Various example embodiments of the present disclosure relate to communication networks,and more specifically to methods, apparatus and a computer program product for monitoringpredictions at a terminal device.BACKGROUND
[0002] The dynamic and heterogeneous nature of modern wireless networks, such as 5G / 6G,Internet of Things (loT), and ultra-dense deployments, necessitates predictive Radio ResourceManagement (RRM) to address critical challenges in scalability, efficiency, and Quality of Service(QoS). Predictive RRM (e.g., RRM measurement predictions and / or event predictions) addressesthese limitations by forecasting future network states (e.g., channel quality degradation, mobilitytrajectories) using Artificial Intelligence (AI)ZMachine Learning (ML) models, non-AI / ML models orother statistical algorithms etc. 3GPP’s evolving frameworks further validate the necessity ofintegrating prediction into RRM. However, the dynamic nature of wireless environments necessitatesmonitoring of performance of predictive models, because the models may degrade due to model drifts,non-stationary data distributions, or evolving network topologies, leading to suboptimal decisions inmeasurement or event predictions. Therefore, there is a need for efficient monitoring frameworks toensure sustained efficacy in measurement prediction and event prediction.SUMMARY
[0003] According to a first aspect of the present disclosure, there is provided a method. The methodmay include receiving, by a terminal device, from a network device a configuration for monitoringpredictions at the terminal device. The configuration comprises a timer for detecting a failure of the ipredictions. The method may also include transmitting, by the terminal device, an indication of the failure to the network device.
[0004] According to a second aspect of the present disclosure, there is provided a method. The method may include transmitting, by a network device, to a terminal device a configuration for monitoring predictions at the terminal device. The configuration comprises a timer for detecting a failure of the predictions. The method may also include receiving, by the network device, an indication of the failure from the terminal device.
[0005] According to a third aspect of the disclosure there is provided an apparatus in a communication network. The apparatus may comprise a processor and a memory communicatively coupled to the processor. The memory may be adapted to store instructions which, when executed by the processor, cause the apparatus to receive from a network device a configuration for monitoring predictions at the terminal device. The configuration comprises a timer for detecting a failure of the predictions. The apparatus may also be caused to transmit an indication of the failure to the network device.
[0006] According to a fourth aspect of the disclosure, there is provided an apparatus in a communication network. The apparatus may comprise a processor and a memory communicatively coupled to the processor. The memory may be adapted to store instructions which, when executed by the processor, cause the apparatus to transmit to a terminal device a configuration for monitoring predictions at the terminal device, wherein the configuration comprises a timer for detecting a failure of the predictions. The configuration comprises a timer for detecting a failure of the predictions. The apparatus may also be caused to receive an indication of the failure from the terminal device.
[0007] According to a fifth aspect of the present disclosure, there is provided a method. The method may include receiving, by a terminal device, from a network device a configuration for monitoringpredictions at the terminal device. The method may also include transmitting, by the terminal device, to the network device a notification indicating that ground truth measurements are not available to the terminal device. The method may further include receiving, by the terminal device, from the network device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
[0008] According to a sixth aspect of the present disclosure, there is provided a method. The method may include transmitting, by a network device, to a terminal device a configuration for monitoring predictions at the terminal device. The method may also include receiving, by the network device, from the terminal device a notification indicating that ground truth measurements are not available to the terminal device. The method may further include in response to receipt of the notification, transmitting, by the network device, to the terminal device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
[0009] According to a seventh aspect of the disclosure there is provided an apparatus in a communication network. The apparatus may comprise a processor and a memory communicatively coupled to the processor. The memory may be adapted to store instructions which, when executed by the processor, cause the apparatus to receive from a network device a configuration for monitoring predictions at the terminal device. The apparatus may also be caused to transmit to the network device a notification indicating that ground truth measurements are not available to the terminal device. The apparatus may further be caused to receive from the network device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
[0010] According to an eighth aspect of the disclosure there is provided an apparatus in acommunication network. The apparatus may comprise a processor and a memory communicatively coupled to the processor. The memory may be adapted to store instructions which, when executed by the processor, cause the apparatus to transmit to a terminal device a configuration for monitoring predictions at the terminal device. The apparatus may also be caused to receive from the terminal device a notification indicating that ground truth measurements are not available to the terminal device. The apparatus may further be caused to in response to receipt of the notification, transmit to the terminal device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
[0011] According to a ninth aspect of the present disclosure, there is provided a non-transitory machine-readable medium having a computer program stored thereon. The computer program, when executed by at least one processor of an apparatus, causes the apparatus to perform steps of the method according to any one of the first to eighth aspects.
[0012] According to a tenth aspect of the present disclosure, there is provided a computer program product comprising instructions. The instructions, when executed by at least one processor of an apparatus, cause the apparatus to perform steps of the method according to any one of the first to eighth aspects.
[0013] According to an eleventh aspect of the disclosure, there is provided an apparatus. The apparatus may comprise means for performing steps of the method according to any one of the first to eighth aspects.BRIEF DESCRIPTION OF DRAWINGS
[0014] Some example embodiments will now be described, by way of non-limiting examples, with reference to the accompanying drawings.
[0015] Fig. 1 illustrates an example communication network in which example embodiments of thepresent disclosure can be implemented.
[0016] Fig. 2 illustrates an example signaling flow between a terminal device and a network device for monitoring predictions according to some example embodiments of the present disclosure.
[0017] Fig. 3 illustrates a flow diagram for an example method for monitoring predictions in accordance with some example embodiments.
[0018] Fig. 4 illustrates a flow diagram for an example method for monitoring predictions in accordance with some example embodiments.
[0019] Fig. 5 illustrates an example signaling flow between a terminal device and a network device for obtaining ground truth measurements when the ground truth measurements are not available to the terminal device according to some example embodiments of the present disclosure.
[0020] Fig. 6 illustrates a flow diagram for an example method for obtaining ground truth measurements when the ground truth measurements are not available to a terminal device in accordance with some example embodiments.
[0021] Fig. 7 illustrates a flow diagram for an example method for obtaining ground truth measurements when the ground truth measurements are not available to a terminal device in accordance with some example embodiments.
[0022] Fig. 8 illustrates a simplified block diagram of an apparatus that is suitable for implementing some example embodiments of the present disclosure.
[0023] Fig. 9 illustrates a block diagram of an example embodiment of a computer readable medium in accordance with some example embodiments of the present disclosure.DETAILED DESCRIPTION
[0024] The example embodiments of the present disclosure are described in detail with reference to the accompanying drawings. It should be understood that these embodiments are discussed onlyfor the purpose of enabling those skilled persons in the art to better understand and thus implement the present disclosure, rather than suggesting any limitations on the scope of the present disclosure.
[0025] References in the specification to "one embodiment," "an embodiment," "an example embodiment," and the like indicate that the embodiment described may include a particular feature, structure, or characteristic, but it is not necessary that every embodiment includes the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the art to affect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described.
[0026] It shall be understood that although the terms "first" and "second" etc., may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and similarly, a second element could be termed a first element, without departing from the scope of example embodiments. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed terms.
[0027] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. 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", "has", "having", "includes" and / or "including", when used herein, specify the presence of stated features, elements, and / or components etc., but do not preclude the presence or addition of one or more other features, elements, components and / or combinations thereof.
[0028] Further, in accordance with example embodiments of the disclosure there is circuitry for performing operations in accordance with example embodiments of the disclosure as disclosed herein. This circuitry can include any type of circuitry including content coding circuitry, content decoding circuitry, processing circuitry, image generation circuitry, data analysis circuitry, etc. Further, this circuitry can include discrete circuitry, application-specific integrated circuitry (ASIC), and / or field-programmable gate array circuitry (FPGA), etc., as well as a processor specifically configured by software to perform the respective function, or dual-core processors with software and corresponding digital signal processors, etc. Additionally, there are provided necessary inputs to and outputs from the circuitry, the function performed by the circuitry and the interconnection (perhaps via the inputs and outputs) of the circuitry with other components that may include other circuitry in order to perform example embodiments of the disclosure as described herein.
[0029] In accordance with example embodiments as disclosed in this disclosure, the "circuitry" provided can include at least one or more or all of the following:(a) hardware-only circuit implementations (such as implementations in analog, digital and / or quantum circuitry);(b) combinations of hardware circuit(s) and software, such as (as applicable):(i) a combination of analog, digital and / or quantum hardware circuit(s) with software / firmware; and(ii) any or all portions of hardware processor(s) (including digital and / or quantum processor(s)) with software, and memory(ies) that work together to cause an apparatus, such as a mobile device, computing device, or server, to perform various functions, such as functions or operations in accordance with example embodiments of the disclosure as disclosed herein); and(c) any or all portions of hardware circuit(s), such as a microprocessor(s) processor(s) and / or quantumprocessor(s), that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation."
[0030] This definition of circuitry applies to all uses of this term in this disclosure, including in any claims. As a further example, as used in this disclosure, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in server, a cellular network device, or other computing or network device.
[0031] As used herein, the term "communication network" refers to a network following any suitable communication standards, such as New Radio (NR), Long Term Evolution (LTE), LTE-Advanced (LTE-A), Wideband Code Division Multiple Access (WCDMA), High-Speed Packet Access (HSPA), Narrow Band Internet of Things (NB-loT), an Enhanced Machine type communication (eMTC) and so on. Furthermore, the communications between a terminal device and a network device in the communication network may be performed according to any suitable generation communication protocols, including, but not limited to, the first generation (1G), the second generation (2G), 2.5G, 2.75G, the third generation (3G), the fourth generation (4G), 4.5G, the fifth generation (5G), the sixth generation (6G) communication protocols, and / or any other protocols either currently known or to be developed in the future. Embodiments of the present disclosure may be applied in various communication systems. Given the rapid development in communications, there will of course also be future type communication technologies and systems with which the present disclosure may be embodied. It should not be seen as limiting the scope of the present disclosure to only the aforementioned system.
[0032] As used herein, the term "network device" refers to a node in a communication network via which a terminal device accesses the network and receives services therefrom. The network device may refer to a base station (BS) or an access point (AP), for example, a node B (NodeB or NB), an evolved NodeB (eNodeB or eNB), a NR next generation NodeB (also referred to as a gNB), a relay, a low power node such as a femto, a pico, and so forth, depending on the applied terminology and technology. The base station may comprise several distributed network units, such as a central unit (CU), one or more distributed units (DUs), one or more remote radio heads (RRHs) or remote radio units (RRUs). The number and functions of these distributed units depend on the selected split RAN architecture. In the following, the terms "network device", "base station (BS)", "NodeB or NB", "eNodeB or eNB" and "gNB" etc., may be used interchangeably.
[0033] As used herein, the term "terminal device" refers to any end device that may be capable of wireless communication. By way of example rather than limitation, a terminal device may also be referred to as a communication device, user equipment (UE), a Subscriber Station (SS), a Portable Subscriber Station, a Mobile Station (MS), or an Access Terminal (AT). The terminal device may include, but not limited to, a mobile phone, a cellular phone, a smart phone, voice over IP (VoIP) phones, wireless local loop phones, a tablet, a wearable terminal device, a personal digital assistant (PDA), portable computers, desktop computer, image capture terminal devices such as digital cameras, gaming terminal devices, music storage and playback appliances, vehicle-mounted wireless terminal devices, wireless endpoints, mobile stations, laptop-embedded equipment (LEE), laptopmounted equipment (LME), USB dongles, smart devices, wireless customer-premises equipment (CPE), an Internet of Things (loT) device, a watch or other wearable, a head-mounted display (HMD), a vehicle, a drone, a medical device and applications (e.g., remote surgery), an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automatedprocessing chain contexts), a consumer electronics device, a device operating on commercial and / orindustrial wireless networks, and the like. In the following, the terms "terminal device", "communicationdevice", "terminal", "user equipment" and "UE" etc., may be used interchangeably.
[0034] As used herein, the term "prediction" refers to methodologies, for example, AI / ML-based (e.g.,neural networks, reinforcement learning) or non-AI / ML-based (e.g., rule-based thresholds, statisticalextrapolation or based on other predictive model), that forecast network conditions or events, servingpurposes analogous to existing measurement and reporting features as specified in 3GPPspecifications TS 38.300 and TS 38.331. The prediction features share the same objectives with manyexisting features as specified in 3GPP specifications TS 38.300 and TS 38.331. However, they mayhave distinct requirements. Take the measurement prediction feature configured by the Network (NW)as an example. This is a new feature aimed at enhancing the performance of wireless communicationsystems. In this context, the User Equipment (UE) can utilize the existing L1 (Layer 1) / L3 (Layer 3)reporting mechanisms to report these measurement predictions. It should be noted that compared to theactual measurements reported using the same signaling framework, the reporting of measurementpredictions has different requirements. Specifically, the reporting of measurement predictions may have morespecific or stricter specifications in terms of time accuracy, data format, transmission frequency, andprocessing complexity. This is because measurement predictions involve estimating future radioenvironments. Ensuring the accuracy and timeliness of predictions is crucial for networks to make reasonableresource allocation and optimization decisions, thereby improving the efficiency of the entire wirelesscommunication system and user experience.
[0035] As used herein, the term "ground truth" refers to actual / real measurements, or measurementsmade based on the legacy measurement framework as specified in 3GPP specifications TS 38.300 and TS38.331 , which can be used by the UE 110 to compare with predictions, e.g., in order to determine whether iothe predictions are in sync (consistent) with them, or whether the predicted measurement event occurs or not.
[0036] As used herein, the term "AI / ML model" refers to various types of an artificial intelligence or machine learning-based computational structure as well as associated features, functionalities or configurations, including but not limited to neural networks (e.g., convolutional neural networks (CNNs), recurrent neural networks (RNNs), graph neural networks (GNNs), transformers, or long short-term memory (LSTM) networks), decision trees, reinforcement learning (RL) agents, federated learning frameworks, or hybrid architectures. These models can be trained using supervised, unsupervised, semi-supervised, or online learning methodologies to perform tasks such as RRM measurement prediction, measurement event prediction, or network parameter optimization etc., as defined in 3GPP standards or specifications. The AI / ML model may incorporate feature extraction layers, activation functions (e.g., ReLU, sigmoid), attention mechanisms, or regularization techniques (e.g., dropout, L1 / L2 normalization) tailored for wireless network data, such as time-series channel state information (CSI), beamforming metrics, or mobility traces. It may also leverage 3GPP-compliant interfaces to dynamically adapt its weights, hyperparameters, or topology based on network feedback. This definition encompasses both centralized and distributed implementations, including edge-trained models, federated learning configurations, and other types of variants. In the following, the terms "Al model", "ML model", "AI / ML model", "AI / ML feature", "AI / ML functionality" and "AI / ML configuration" etc., may be used interchangeably where the context permits to describe aspects of artificial intelligence or machine learning systems. While these terms may carry distinct technical meanings in other contexts, the example embodiments of the present disclosure are applicable to implementations involving any such component, model, feature, functionality, configuration, or capability.
[0037] As used herein, the term "slot" refers to a configurable unit of resource allocation in time,frequency, or spatial domains, as defined in 3GPP standards or specifications. A time slot can correspond to a discrete time interval comprising one or more orthogonal frequency-division multiplexing (OFDM) symbols, configurable based on subcarrier spacing (e.g., 15 kHz, 30 kHz). A frequency slot can denote a contiguous or non-contiguous set of subcarriers or resource blocks (RBs) within a carrier bandwidth. A spatial slot can refer to a beamforming or spatial layer allocation, such as a precoded transmission layer in multi-antenna systems. Slots may be dynamically adjusted in duration, bandwidth, or spatial granularity to align with network conditions (e.g., via Radio Resource Control (RRC) signaling). For instance, time-frequency slots can be aggregated in carrier aggregation (CA) scenarios, while spatial slots are mapped to beam indices in millimeter-wave (mmWave) deployments. This multi-dimensional slot definition enables flexible resource partitioning, interference mitigation, and QoS-aware scheduling across heterogeneous 3GPP-compliant networks.
[0038] It is noted that these terms as used in this disclosure are used only for ease of description and differentiation among nodes, devices or networks etc. With the development of the technology, other terms with the similar / same meanings may also be used. In the following description and claims, unless defined otherwise, all technical terms used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this disclosure belongs.
[0039] In modern wireless communication systems, RRM plays a critical role in optimizing network performance, ensuring efficient spectrum utilization, and maintaining QoS for diverse user demands. The dynamic and heterogeneous nature of modern wireless networks, such as 5G / 6G, loT, and ultra-dense deployments, necessitates predictive RRM to address critical challenges in scalability, efficiency, and QoS. Predictive RRM (e.g., RRM measurement predictions and / or event predictions) addresses these limitations by forecasting future network states (e.g., channel quality degradation, mobility trajectories) using AI / ML models, non-AI / ML models or other statistical algorithms etc. 3GPP’sevolving frameworks further validate the necessity of integrating prediction into RRM. However, the dynamic nature of wireless environments necessitates monitoring of performance of predictive models, because the models may degrade due to model drifts, non-stationary data distributions, or evolving network topologies, leading to suboptimal decisions in measurement or event predictions.
[0040] For AI / ML or non-AI / ML prediction performance monitoring, an AI / ML-enabled or non-AI / ML feature configuration parameter to configure the length of the sliding model monitoring window to calculate the performance metrics of prediction was proposed. It specifically provides an example of how to use this in measurement prediction. However, it is based on calculating a sliding average, which introduces, depending on the length of the sliding window, delay in the monitored performance metrics. This can be an issue if a very reaction to any degradation is needed. Furthermore, Key Performance Indicators (KPI) that measure performance of AI / ML or non-AI / ML models / functionalities / configurations (e.g., accuracy, F1 score, Reference Signal Received Power (RSRP) difference etc.) have a statistical nature, i.e., many prediction samples with their samples are required to compute such KPIs. The number of required samples depends on the specific use case. For use cases, where fast response to any performance degradation of prediction is required, an efficient method for monitoring is required.
[0041] In this regard, it is proposed a novel monitoring method for predictions, to enhance a faster response to UE-side model performance degradations. In use cases where the ground truth measurements of the predictions can be obtained based on network configuration, for example in time domain RRM measurement prediction and / or measurement event prediction, the UE monitors for prediction errors for a certain number of slots. As an example, if a configured number of subsequent errors (a prediction drift) are observed (e.g., within a configured number of slots or for a duration of time or within a configured number of slots within a certain duration of time), the UE starts a timer.When the timer is running, if a configurable number of (subsequent) correct predictions are received (either within a configured number of slots or for a duration of time or within a number of slots within a certain duration of time), which indicates that there is no longer a drift in the predictive model, the UE cancels or resets the timer. In another example, when the timer is running, the UE monitors during the duration of a further timer for a certain number of in-sync (consistent) instances (i.e., instances where there is no drift in the predictive model) and then stops the timer if such instances are detected. Otherwise, if the timer expires (e.g., the configurable number of correct predictions are not received during the running of the timer), the UE indicates that the predictive model / functionality / configuration has failed and takes actions accordingly.
[0042] The term "error" as used herein may be, for example, a regression prediction error higher than a specified threshold (e.g., the predicted measurement and real measurement deviate from each other by more than x decibel (dB) (which can refer to a unit used to express the ratio between two values) for a certain number of slots, or the average deviations between predicted and real measurements are more than x dB) or a misclassification (e.g., UE predicts that there will be an event in the next x millisecond (ms) (which can be a configured amount of time), when there actually is no event in the next x ms). The term "correct prediction" as used herein may be, for example, a regression prediction error lower than a specified threshold (e.g., the predicted measurement and real measurement deviate from each other by less than x dB for a certain (configured) number of slots, or the average deviations between predicted and real measurements are less than x dB) or a correct classification (e.g., UE predicts that there will be an event in the next x ms, when there actually is no event in the next x ms). The terms "drift" or "prediction drift" as used herein or "invalidity" or "prediction invalidity" may be, for example, a configured number of prediction errors made by the predictive model (e.g., predictions which are inconsistent or out of sync with actual measurements) to trigger the timer, e.g., as a condition forstarting the timer. Therefore, the terms "drift", "prediction drift", "invalidity" or "prediction invalidity" may be used interchangeably. Further, the term "failure" as used herein may be, for example, a real degradation / malfunction / breakdown of the predictive model, e.g., determined when the timer expires after a drift is detected.
[0043] In the following, the example embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.
[0044] Fig.1 illustrates an example communication network 100 in which example embodiments of the present disclosure can be implemented. As shown in FIG. 1, the communication network 100 may include a terminal device 110, which may also be referred to as a UE. The communication network 100 may further include a network device 120, which may also be referred to as a gNB respectively. The terminal device 110 and the network device 120 can be in communication with each other. To transmit data and / or control information, the terminal device 110 may perform communications with the network device 120. A link from the network device 120 to the terminal device 110 is referred to as a downlink (DL), while a link from the terminal device 110 to the network device 120 is referred to as an uplink (UL). Although Fig. 1 only shows one terminal device 110 and one network device 120, it should be noted that a number of terminal devices and network devices, as well as other components / devices / functions, can be provided in the communication network 100. As an example, the terminal device 110 and the network device 120 may be in communication with one or more further devices not shown in Fig. 1.
[0045] In some example embodiments, the terminal device 110 can host a predictive model 130 (which can be an AI / ML or non-AI / ML model) configured to perform RRM measurement prediction and / or measurement event prediction, as defined in 3GPP specifications TS 38.300, TS 38.331 and possibly 38.321. The predictive model 130 can be trained on historical and real-time RRM data (e.g.,Reference Signal Received Power (RSRP), Channel State Information (CSI), or interference levels) and the trained predictive model 130 can generate predictive outputs about RRM measurement (such as values of RSRP, Signal-to-lnterference-plus-Noise Ratio (SINR), Received Signal Strength Indicator (RSSI) and Channel Quality Indicator (CQI) etc.) and / or measurement event (such as any of events A1-A6 and B1-B2 as defined in 3GPP specifications).
[0046] In some example embodiments, the terminal device 110 can transmit these predictions to the network device 120 via standardized signaling protocols, such as RRC measurement reports or Medium Access Control (MAC) control elements (CEs). The predictions may include time-stamped metadata (e.g., confidence intervals, prediction horizons) to enable network device 120 to dynamically adjust RRM parameters (e.g., scheduling grants, beamforming configurations, or mobility settings).
[0047] In some example embodiments, in the communication network 100, the predictive model 130 at the terminal device 110 may exhibit a performance degradation or a prediction drift (e.g., persistent inaccuracies in RRM measurement prediction or event prediction) due to dynamic network fluctuations, non-stationary data distributions, or abrupt topology changes in the communication network. To mitigate such risks, the terminal device 110 can be configured to monitor the performance of the predictive model 130, for example, by comparing its predictions against real-time RRM measurements (e.g., values of RSRP, SINR, RSSI and CQI etc.). In some examples, the terminal device 110 may compare the predictions with the actual measurements to determine whether there is a prediction drift in the predictive model 130 (e.g., a configured number of prediction errors for a certain number of slots, where the prediction errors can refer to predictions which are not synchronized (not consistent) with real measurements, i.e., out-of-sync predictions). Alternatively, the terminal device 110 may calculate the variance of the predictions to determine whether there is a prediction drift in the predictive model 130.
[0048] However, in some cases, the predictive model 130 may resume to a satisfactory performance it should have after a prediction drift for a very short period, which indicates that the predictive model 130 may not have failed. Therefore, a mechanism is needed to consider this situation to reduce false positives.
[0049] In some example embodiments, the terminal device 110 can be configured to distinguish temporary prediction anomalies from sustained / real performance degradation. As an example, the terminal device 110 can be configured to start a timer in response to detecting a prediction drift in the predictive model 130. If the drift remains in the predictive model 130 during a period of the timer (e.g., causing the timer to expire), the terminal device 110 can indicate a failure of the predictive model 130 to the network device 120 and take actions accordingly. If there is no longer a prediction drift in the predictive model 130 (i.e., the predictions from the predictive model 130 come back to in sync (consistent) with real measurements) during the running of the timer, the terminal device 110 can reset or cancel (stop or release) the timer.
[0050] It should be noted that the above-mentioned prediction may be performed by any AI / ML-based or non-AI / ML manner / model or in other ways. In the following, the examples embodiments for monitoring predictions in accordance with the present disclosure will be described in detail.
[0051] Fig. 2 illustrates an example signaling flow 200 between a terminal device 110 and a network device 120 for monitoring predictions according to some example embodiments of the present disclosure. For the purpose of discussion, the signaling flow 200 will be discussed with reference to Fig. 1. The network device 120 (hereinafter referred to as "NW") may comprise any 3GPP-compliant infrastructure component involved in radio access, resource management, or network control, for example, Centralized Units (CUs) and Distributed Units (DUs) such as Split CU (S-CU), Distributed CU (D-CU), or Open RAN (O-RAN) Alliance-defined Open DU (O-DU) and Open CU (O-CU) indisaggregated RAN architectures. As an example, communications can be transmitted between the terminal device 110 (hereinafter referred to as "UE") and the S-CU (as part of the NW 120), with the S-DC (also as part of the NW 120) acting as a forwarding entity, i.e., the S-DC serves as an intermediary for data transmission.
[0052] At step 210, the NW 120 requests the capabilities from the UE 110. In some examples, this is achieved through standardized RRC signaling. Specifically, the NW 120 can transmit a U ECapabi lityE nq uiry message to the UE 110 via a dedicated signaling radio bearer (SRB1 or SRB2), requesting its capabilities, e.g., predictive capabilities.
[0053] At step 220, UE 110 reports the predictive capabilities to the network device 120. As an example, UE 110 can indicate support for measurement prediction. In another example, UE 110 can indicate support for measurement event prediction. Independent of the use-case, UE 110 can further indicate support for UE-side monitoring of the predictive model.
[0054] At step 230, the NW 120 sends configuration information to the UE 110. The information may comprise at least one of: i) a prediction configuration (or inference configuration) for the measurement and / or event prediction; ii) a prediction drift configuration for the measurement and / or event prediction; iii) a timer associated with prediction drift; and iv) a configuration to determine that predictions are no longer in drift. In some examples, the configuration information can be sent via a RRC message / signaling.
[0055] In some example embodiments where the NW 120 sends the prediction configuration to the UE 110, the NW 120 can also provide the reporting configuration that the UE 110 can use to report the predicted measurements and / or the measurement event. In one example, the NW 120 can configure whether the UE 110 should report predicted L1, L3 measurements or both. In another example, NW 120 can configure whether the UE 110 can perform direct or indirect measurement eventprediction and report the corresponding outcomes as defined in TR 38.744. The NW 120 can configure what the trigger for the measurement event prediction should be (e.g., when A3 entry condition is met; in another example, the NW 120 can configure periodic prediction). As part of this configuration, the NW 120 can provide new timer and thresholds (e.g., Time-to-Trigger (TTT) and / or hysteresis) that UE 110 can use to perform inference.
[0056] In some example embodiments where the NW 120 sends the prediction drift configuration to the UE 110, NW 120 can provide the information needed for the UE 110 to detect a drift in the predictive model. In one example, the UE 110 can be configured to compare outcomes of the configured predictive model or feature or functionality to outcomes of a procedure where no predictive model or feature or functionality is configured. This can be the legacy procedures as specified in 3GPP specifications TS 38.300 and TS 38.331. As an example, for the measurement prediction use-case, the NW 120 can configure the number of slots for which the UE 110 should make measurements and predictions at the same time (needed to determine whether there are any drifts in the predictive model / functionality / configuration in the UE 110). For instance, in the case of Ax (e.g., any of A1-A6) event prediction feature, the UE 110 may perform measurements and predictions of Ax event as specified in 3GPP specifications TS 38.300 and TS 38.331. In case of measurement prediction, the UE may measure all the configured slots as specified in 3GPP specifications TS 38.300 and TS 38.331.
[0057] The dedicated monitoring resource set(s) / Synchronization Signal (SS) Block(s) can be configured by the network to measure and report (optionally) the ground truth corresponding to predictions. In addition, NW 120 may consider a period of time for which the UE 110 should monitor these slots (e.g., UE 110 monitors whether predictions have drifted over x slots in a period of y ms (guarded / defined by a timer)). For example, NW 120 can provide the system frame numbers (SFNs) for UE 110 to start measurements and align the prediction simultaneously. In another embodiment,for the measurement event prediction use-case, NW 120 can provide similar configuration, i.e., number of slots based on which the UE 110 makes predictions of whether there will be an event in the next x ms (in one embodiment, this can be the TTT. Together with the slots, in another embodiment, NW 120 can also provide a metric that the UE 110 should monitor either within a slot, or within any configured prediction window. In this case, the UE 110 may be asked to monitor the consistency of the prediction during the configured window. As an example, UE 110 can be asked to check that the predictions for a configured measurement event (e.g., A3 event) are consistent for a certain duration and / or number of slots (e.g., UE 110 can monitor if starting from time t and / or slot s until some end time or slot the UE 110 would always predict with accuracy (or any other metric of confidence such as F1 score, precision, recall etc., or any other system-level metric, e.g., theoretical PDCP failure rate) that the event A3 will always happen. This would mean that predictions are consistent and have not drifted). On the other hand, if the UE 110 observes that within the configured time or slot intervals the UE 110 is less confident in the predictions (than a pre-determined threshold given in the prediction drift configuration) or the UE 110 observes inconsistent predictions, the UE 110 may then be able to determine that an drift in the predictive model has happened. In another embodiment, the UE 110 can also be configured to monitor a number of slots within a certain duration of time.
[0058] In some example embodiments where the NW 120 sends the timer to the UE 110, the timer can be given as part of the specific prediction configuration (e.g., can be different for measurement prediction and measurement event prediction) or can be the same for all prediction configurations (e.g., in case indirect measurement event prediction is configured). In the latter case, the timer may be given outside the specific measurement and measurement event prediction configurations. The timer as used herein can be a time-based trigger configured to measure a predefined duration for a specific event. Upon expiration, the timer may initiate actions. In another example, the functions of atimer may also be performed by a counter.
[0059] In some example embodiments where the NW 120 sends to the UE 110 the configuration to determine that predictions are no longer in drift, this configuration is used to determine when the predictions from the predictive model come back to sync (e.g., the predictions are consistent with the actual measurements made by the UE 110, or the predicted measurement event actually occurs). Similar approach as described earlier can be followed, with the difference that now UE 110 checks that the predictions become stable for a certain number of slots and / or time (e.g., UE consistently observes that for x slots and y ms, the predictions are within a configured tolerance level, such as over 95% accurate. In other cases, this can be a tolerance on F1 score, precision, recall, mean squared error (MSE), mean absolute error (MAE) etc.).
[0060] At step 240, the UE 110 reports predictions of RRM measurements and / or measurement events (as configured) to the NW 120, after receipt of the configuration from the NW 120. In some examples, the UE 110 may report different predictions to the NW 120 at different times or at the same time. In some examples, the UE 110 may report the predictions periodically and / or in response to an indication from the NW 120.
[0061] At step 250, the UE 110, which is configured with UE-side monitoring of UE-sided model, monitors a drift in predictions by the predictive model.
[0062] In some example embodiments, the NW 120 will schedule Reference Signal (RS) for the UE to measure and check whether the predictions from the predictive model are synced (consistent) with real measurements. In one example, these can be given as part of a dedicated monitoring configuration towards the UE 110. In another example, UE 110 can use the same RS given in the measurement configuration (either L1 or L3 configuration) and compare with the predictions. In some examples, the periodicity of monitoring can be configured via RRC. In other examples, the periodicityof monitoring can be left to UE implementation, e.g., the periodicity of monitoring can be selected by the UE itself. In the case where it is left to UE implementation, the UE 110 indicates to the NW 120 via dedicated signaling when problems are observed with performance of model / functionality / configuration (e.g., a failure of the predictive model is detected). In one example, such indication can be a dedicated lower layer signaling (e.g., MAC CE). In another example, such indication can be piggybacked into a measurement report. In yet another example, such indication can be sent through a RRC configuration message, a MAC-level message or UE Assistance Information (UAI).
[0063] In some example embodiments, the drift in the model is detected if a sequence of predictions is not in sync with real measurements (e.g., at time t, the UE 110 predicts there will be A3 event in the next 5 seconds. Then, at time t+1 , the UE 110 predicts there will not be an A3 event in the next 5 seconds. Then, at time t+2, the UE 110 again predicts that there will be an A3 event in the next 5 seconds and so on). In some examples, the drift in the model is detected if a configured number (e.g., Nerr) of subsequent prediction errors from the predictive model are observed either within a configured number of slots or for a duration of time or within a number of slots within a certain duration of time. In one example, the number of slots and / or a period of time for which the UE 110 checks to determine the drift is configurable by the NW 120 through a timer / counter configured via RRC.
[0064] At step 260, the UE 110 starts a timer (configured via RRC) when a drift in the predictive model is detected. In some example embodiments, if a configured number (e.g., Nerr) of subsequent prediction errors from the predictive model are observed (which may indicate that the drift is detected), the UE 100 starts a timer. In one example, when the timer is started, the UE 110 is configured to monitor a configured number of correct predictions from the predictive model for a configured number of slots during the running of this timer.
[0065] In response to the timer started at step 260 expiring (e.g., "OPTION 1" as shown in Fig. 2), the UE 110 indicates a failure of the predictive model to the NW 120 at step 270 and / or falls back to another configuration at step 280. In one example, another configuration can be indicated as "fallback configuration". In such a case, the UE 110 switches to the indicated configuration and continues using it.
[0066] In some example embodiments, if a configurable number (e.g., Ncorr) of correct predictions from the predictive model are not received during the running of this timer, causing the timer to expire, the UE 110 indicates that the predictive model has failed and takes actions accordingly. In other words, upon timer expiry, the UE 110 understands that the configured predictive model cannot be supported. In some embodiments, if configured, the UE 110 switches to fallback configuration (e.g., given by the NW 120 via RRC). As an example, such fallback configuration can be legacy way of operating as specified in 3GPP specifications TS 38.300 and TS 38.331 . For instance, in the case of Ax (e.g., any of A1-A6) event prediction feature, the UE 110 may perform evaluations of Ax event as specified in 3GPP specifications TS 38.300 and TS 38.331. In case of measurement prediction, the UE may measure all the configured slots as specified in 3GPP specifications TS 38.300 and TS 38.331. The term "legacy" or "fallback" configuration as used herein refers to pre-defined measurement / reporting / prediction procedures compliant with 3GPP specifications TS 38.300, TS 38.331 and possibly 38.321. As an example, when in a legacy / fallback configuration, the UE 110 may measure and report measurements for all configured slots (even if within the prediction window), instead of performing measurement in some slots and performing only prediction in other time slots.
[0067] In one embodiment, the fallback configuration can also be an AI / ML-enabled configuration, the fallback configuration can also be a non-AI / ML configuration. In another embodiment, the fallback configuration can be the legacy A3 event configuration. In yet another embodiment, the UE 110 sendsa message to the NW 120 informing about the failure of the configured AI / ML-enabled or non-AI / ML feature / functionality (e.g., measurement event prediction). In still another embodiment, the UE 110 falls back to a legacy configuration, and sends RRC Reconfiguration complete message (or any other signaling such as Downlink Control Information (DCI), MAC CE etc.) to the NW 120 indicating the cause of failure at step 270. In another example, the UE 110 is explicitly configured to fall back to the legacy configuration (e.g., in case of measurement prediction, the UE 110 measures all the configured RS and sends measurement report to the NW at step 270. In the case of measurement event prediction, the UE 110 monitors for event prediction using legacy Ax (e.g., A1, A2...) event prediction mechanism).
[0068] In some example embodiments, the UE 110 can inform the NW 120 whether it was a configuration failure, e.g., the UE 110 could not even apply the given configuration. An example may be the case where the NW 120 configures longer prediction window (PW) for measurement prediction than what the UE 110 can support. Another example may be when the NW 120 configures A5 event prediction for the UE 110 capable of only predicting an A3 event.
[0069] In other example embodiments, the UE 110 can inform the NW 120 about the relevant performance metrics that were observed during the monitoring procedure. For instance, the UE 110 can report the configured KPI (e.g., measurement event prediction accuracy or measurement prediction MAE or the number of predictions that deviate beyond a certain tolerance level) for the configured monitoring slots and / or time. In another embodiment, the UE 110 can report failure information related to its predictive configuration (e.g., observed sequence of out-of-sync intervals, cells for which the predictions were not good (e.g., not meeting the RAN4 requirements), new applicable conditions and / or functionality information, data collection request, measurement configuration request. In another embodiment, such information can also be sent through UAI (if configured). In yet another embodiment, this information is conveyed through a new RRC / MAC-levelmessage.
[0070] In additional example embodiments, upon detecting failure (for the measurement event prediction), the UE 110 sends a measurement report to the NW 120. UE 110 can also provide more information about the cause of failure. In one example, together with the measurements report, UE 110 provides information on the confidence estimates for the predictions of the predictive model. In another embodiment, the UE 110 provides information (e.g., an indicator) on whether the predictions of the predictive model are compliant with the RAN4 requirements (or any other system-level KPI that might have been configured by the NW 120). In another embodiment, the UE 110 learns the number of slots and / or amount of time and / or accuracy (or any other configured system-level KPI) thresholds needed to be observed to determine that the predictive model has drifted and / or the configured functionality / feature is no longer applicable / useful. This may be indicated through the UE 110 capabilities. In yet another embodiment, the UE 110 can learn to also determine the number of slots and / or amount of time and / or accuracy thresholds needed to be observed to determine that the predictive model is no longer drifting and / or the configured functionality / feature is still applicable / useful.
[0071] In additional example embodiments, the legacy configuration is also given to the UE 110 (e.g., UE 110 is configured both with A3 event prediction and legacy A3 event reporting). In one embodiment, upon detection of failure of the prediction model (or upon detection that the configuration is no longer applicable), UE 110 deactivates the configuration and continues using the legacy procedure. In one example implementation, UE 110 sends RRCReconfigurationComplete to the NW 120 upon stopping the usage of the prediction configuration. As part of this message. In another embodiment, the legacy configuration is not given to the UE 110 (e.g., UE 110 is configured only with A3 event prediction. This involves a similar behavior as above, except that now the UE 110 does notmonitor for legacy A3 event.
[0072] In response to determining that there is no longer a drift in the predictive model (e.g., "OPTION 2" as shown in Fig. 2), at step 290, the UE 110 resets / stop / release the timer started at step 260.
[0073] In some example embodiments, when the timer is running, if a configurable number (e.g., Ncorr) of (subsequent) correct predictions are received (either within a configured number of slots or for a duration of time or within a number of slots within a certain duration of time), the UE cancels / stops or resets or release the timer. I n other word, if the UE 110 observes that predictions from the predictive model come back to in sync with real measurements (e.g., the model gives consistent estimates of whether an event will or will not happen, for a certain number of slots; or the configured KPI is consistently met for the configured amount of slots / time), the UE 110 resets or stops timer / counter associated with determining the failure of the predictive model. In one embodiment, the number of slots that the UE 110 needs to observe consistent predictions from the predictive model is configurable by the NW 120. As an example, parameters can be given together with the RRCReconfigurationComplete (or any other signaling medium such as DCI, MAC CE etc.) for the UE 110 to determine the criteria based on which the estimates are consistent. For instance, the UE 110 observes that estimates are consistent for a certain number of slots, time (e.g., through timer) or confidence estimates for the predictions remain above a certain threshold for long number of slots and / or large amount of time.
[0074] In one embodiment, when determining that there is no longer a drift in the predictive model, the UE 110 reports nothing to the NW 120 and continues using the configuration for which failure was not observed. In another embodiment, at step 295, the UE 110 can send an indication to the NW 120 informing about the observed drifts. The granularity of this indication can vary. For example, for measurement prediction, at step 295, the UE 110 can send legacy measurement report to the NW 120(in this case, detection of drifts in the performance of configured predictive model / functional ity / feature can serve as an event that triggers measurement report sent to the NW 120) containing an indication of the observed drift; the UE 110 can send legacy measurement report to the NW 120 containing detailed information on the observed drift (e.g., number of slots, time, performance KPIs etc.); and the UE 110 can send dedicated signaling to inform the NW 120 about the observed drift and the state of its variables (e.g., UAI, DCI, MAC CE etc.). In yet another embodiment, at step 295, the UE 110 can send a measurement and / or prediction report to implicitly the NW 120 about the observed prediction drift. In still another embodiment, at step 295, the UE 110 can explicitly inform the NW 120 about the observed prediction drift.
[0075] For the above steps, as an example, consider the measurement prediction use-case and the measurement / prediction patterns "MMMMPPPP" (which means that among 8 slots, the UE 110 performs measurement in the first four slots and prediction in the last four slots). In one implementation, for the interval where the UE 110 is doing measurements, the UE 110 can be configured to also perform prediction for a certain number of slots. If the UE 110 observes that the predictions performed during this monitoring window are of good quality and no failure is detected, the UE 110 performs only predictions in the configured PW. Otherwise, if UE 110 detects a failure of predictive model, UE 110 falls back to either to another configured predictive functionality or to legacy behavior (UE 110 reports measures also in the PW). In another embodiment, UE 110 includes an indication of fallback in the measurement report. In yet another embodiment, the UE 110 sends dedicated signaling to the NW 120 to indicate the fallback (e.g., MAC CE or RRC Reconfiguration message). In one example, such indication can be included in an RRCReconfigurationComplete message sent from the UE 110 to the NW 120.
[0076] In additional example embodiments, for the timers / counters as described herein, their valuescan be selected and configured (via RRC, MAC CE or DCI) by the NW depending on the use-case. In one embodiment, the timers / counters can have different values for different use-cases. For instance, when configuring indirect event prediction (as defined in TR 38.744), the NW may want to configure smaller value for the counters for the RRM prediction part, and bigger values for the event prediction part. The reason for this is because the prediction of a configured event is derived based on the predicted measurements. Therefore, if the predicted measurements have drifted, it is highly likely that the event predictions will also not be correct. However, there may be cases where the predictions are correctly predicted, however the event information is not correct (e.g., UE estimates the time when the event is expected to happen incorrectly). In another embodiment, the timers / counters can be same for the different use-cases (e.g., RRM prediction, event prediction, RLF prediction etc.). An example may be the case where the reporting periodicity for the predicted events is high (UE is configured to make large number of predictions and therefore the drifts would be visible more quickly). In yet another embodiment, the timers / counters can have large values when the number of predictions that UE is configured to report is high (in this case, NW may want to avoid triggering actions related to failure of the predictive model too soon and therefore configures larger values for the corresponding counters and timers).
[0077] Fig. 2 is provided as one example of a method according to certain embodiments. Other examples are possible as one or more procedures discussed above or illustrated in Fig. 2 may be skipped or may be performed in a different order, or additional procedures may be added. The steps of Fig. 2 represent computer code instructions stored in the RAM memory of the device, which when executed by the central processing unit (CPU), carry out the functions of an example embodiment of the disclosure. Alternately, some or all of the steps in the procedure of the flow diagram may be embodied as hardware program logic included in programmed logic arrays of sequential and / orcombinatorial logic circuits and / or state machine logic implementing some or all of the steps performed by embodiments of the disclosure. The steps may be carried out in another order than shown and individual steps may be combined or separated into component steps.
[0078] Fig. 3 illustrates a flow diagram for an example method 300 for monitoring predictions in accordance with some example embodiments. The method 300 can be implemented at a terminal device, e.g. the terminal device 110 as discussed above. In some example embodiments, the method 300 may further include one or more steps that are performed at the terminal device 110 as described above with respect to Fig. 2. It would also be understood that details of some steps in the method 300 have been discussed above with respect to Fig. 2 and the method 300 will be described here in a simple manner. The steps of the flow diagram may represent computer code instructions stored in the RAM and / or ROM memory of the terminal device 110, which when executed by the central processing units (CPU), carry out the functions of the example embodiments of the present disclosure. The steps may be carried out in another order than shown and individual steps may be combined or separated into component steps.
[0079] According to some example embodiments, at 310, a terminal device may receive from a network device a configuration for monitoring predictions at the terminal device, wherein the configuration may comprise a timer for detecting a failure of the predictions.
[0080] According to some example embodiments, at 320, the terminal device may transmit an indication of the failure to the network device.
[0081] According to some example embodiments, the configuration may be received via an RRC message, a MAC-level message and / or DCI.
[0082] According to some example embodiments, the configuration may include a prediction drift configuration comprising the timer and information for the terminal device to detect a drift in thepredictions.
[0083] According to some example embodiments, the prediction drift configuration may further comprise a configured number of slots for which the predictions are monitored.
[0084] According to some example embodiments, for the slots, the terminal device may compare the predictions to outcomes of a procedure where no prediction is performed.
[0085] According to some example embodiments, the timer may be configured to define an amount of time in which the predictions are monitored for the configured number of slots.
[0086] According to some example embodiments, for the slots, the terminal device may detect a configured number of prediction errors for detecting the drift.
[0087] According to some example embodiments, the terminal device may start the timer in response to detecting the drift, and for the configured number of slots, detect a configured number of correct predictions during running of the timer.
[0088] According to some example embodiments, the terminal device may reset or stop or release the timer in response to detecting the configured number of correct predictions during the running of the timer.
[0089] According to some example embodiments, the terminal device may be configured to not report and continue using a current configuration for the predictions, in response to detecting the configured number of correct predictions during the running of the timer.
[0090] According to some example embodiments, the terminal device may transmit an indication to the network device informing about the detected drift in response to detecting the configured number of correct predictions during the running of the timer.
[0091] According to some example embodiments, the terminal device may transmit the indication of the failure, in response to the configured number of correct predictions not being detected during therunning of the timer such that the timer expires.
[0092] According to some example embodiments, the configuration may include a prediction configuration for measurement prediction and / or measurement event prediction, wherein the timer may be dependent on the prediction configuration, or wherein the timer may be independent of the prediction configuration.
[0093] According to some example embodiments, the indication of the failure may comprise failure information related to the predictions and performance metrics associated with the predictions.
[0094] According to some example embodiments, the indication of the failure may be transmitted through UAI, an RRC message and / or a MAC-level message.
[0095] According to some example embodiments, upon transmitting the indication of the failure, the terminal device may switch to a fallback configuration given by the network device and transmit an indication of failure cause to the network device, wherein the terminal device may be explicitly configured to switch to the fallback configuration.
[0096] According to some example embodiments, the indication of failure cause may be transmitted via an RRC Reconfiguration complete message, DCI and / or an MAC CE.
[0097] In some example embodiments, an apparatus capable of performing the method 300 (for example, the terminal device 110) may comprise means for performing the respective steps of the method 300. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module.
[0098] Fig. 4 illustrates a flow diagram for an example method 400 for monitoring predictions in accordance with some example embodiments. The method 400 can be implemented at a network device, e.g. the network device 120 as discussed above. In some example embodiments, the method 400 may further include one or more steps that are performed at the network device 120 as describedabove with respect to Fig. 2. It would also be understood that details of some steps in the method 400 have been discussed above with respect to Fig. 2 and the method 400 will be described here in a simple manner. The steps of the flow diagram represent computer code instructions stored in the RAM and / or ROM memory of the network device 120, which when executed by the central processing units (CPU), carry out the functions of the example embodiments of the present disclosure. The steps may be carried out in another order than shown and individual steps may be combined or separated into component steps.
[0099] According to some example embodiments, at 410, a network device may transmit to a terminal device a configuration for monitoring predictions at the terminal device, wherein the configuration may comprise a timer for detecting a failure of the predictions.
[0100] According to some example embodiments, at 420, the network device may receive an indication of the failure from the terminal device.
[0101] According to some example embodiments, the configuration may be transmitted via an RRC message, a MAC-level message and / or DCI.
[0102] According to some example embodiments, the configuration may include a prediction drift configuration comprising the timer and information for the terminal device to detect a drift in the predictions.
[0103] According to some example embodiments, the prediction drift configuration may further comprise a configured number of slots for which the predictions are monitored.
[0104] According to some example embodiments, for the configured number of slots, the predictions may be compared to outcomes of a procedure where no prediction is performed.
[0105] According to some example embodiments, the timer may be configured to define an amount of time in which the predictions are monitored for the configured number of slots.
[0106] According to some example embodiments, for the configured number of slots, a configured number of prediction errors are detected for detecting the drift.
[0107] According to some example embodiments, the timer may be started in response to the drift being detected, and wherein for the configured number of slots, a configured number of correct predictions from may be detected during running of the timer.
[0108] According to some example embodiments, the configuration may include a prediction configuration for measurement prediction and / or measurement event prediction, wherein the timer may be dependent on the prediction configuration, or wherein the timer may be independent of the prediction configuration.
[0109] According to some example embodiments, the network device may receive an indication of failure cause from the terminal device along with the indication of the failure.
[0110] In some example embodiments, an apparatus capable of performing the method 400 (for example, the network device 120) may comprise means for performing the respective steps of the method 400. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module.
[0111] Fig. 5 illustrates an example signaling flow 500 between a terminal device 110 (hereinafter referred to as "UE") and a network device 120 (hereinafter referred to as "NW") for obtaining ground truth measurements when the ground truth measurements are not available to the UE 110 according to some example embodiments of the present disclosure. For purpose of discussion, the signaling flow 500 will be discussed with reference to Fig. 1
[0112] The "ground truth" as used herein refers to actual / real measurements, or measurements made based on the legacy measurement framework as specified in 3GPP specifications TS 38.300 and TS 38.331, which can be used by the UE 110 to compare with predictions, e.g., in order todetermine whether the predictions are in sync (consistent) with them, or whether the predicted measurement event occurs or not. Use cases in which the ground truth measurement can be obtained include:
[0113] a) Time domain measurement prediction, where the UE 110 predicts a specific number of future measurements (for the purpose of, e.g., early measurement event prediction), and then later measures those slots corresponding to the predicted measurements. The NW 120 will need to configure the UE 110 for these measurements, known as the ground truth. These dedicated resource set(s) / SS block(s) which are configured to the UE 110 for ground truth extraction (measurement and reporting) may be considered as performance monitoring resource set(s) / SS block(s). UE 110 can then compare the ground truth measurements with the predictions.
[0114] b) Measurement event prediction, where UE 110 predicts a measurement event beforehand. If the predicted measurement event occurs (or not), UE 110 can know whether the prediction was correct or not.
[0115] c) Measurement reduction in time, spatial, or frequency domain, in which the UE 110 can decide (or be configured) to not perform / save any measurements periodically e.g., in certain slots, such that the UE 110 are not able to obtain ground truth measurements in those slots and can obtain ground truth measurements in other slots.
[0116] The measurement reduction in time, spatial, or frequency domains involves selectively limiting time, spatial or frequency scope of RRM measurements to conserve UE power, reduce signaling overhead, or prioritize critical network operations. In one example, the NW 120 configures the UE 110 via measConfig to perform and report measurements only in designated measurement slots (e.g., slots with configured SSB bursts, CSI-RS resources, or measurement gaps), while suspending measurements in other slots (e.g., reserved for AI / ML-based prediction or non-AI / MLprediction or data transmission). For instance, the UE 110 may be instructed to at least one of following modes: time-domain reduction in which the UE is configured to measure RSRP / RSRQ in periodic slots aligned with SSB transmission occasions, skipping intermediate slots; frequency-domain reduction in which the UE is configured to monitor a subset of carrier frequencies or bandwidth parts (BWPs); and spatial-domain reduction in which the UE is configured to restrict beam measurements to specific spatial layers or beam indices.
[0117] However, due to hardware constraints, interference, or misconfiguration, the UE 110 in practice may fail to execute and report measurements even in the designated measurement slots (e.g., in high-mobility scenarios or due to abrupt interference). This results in the ground truth measurements being not available to the UE 110, which negatively affects the monitoring of prediction performance. To address this, it is proposed a dynamic fallback mechanism as follows.
[0118] At step 510, the UE 110 may receive a configuration for monitoring of prediction performance from the NW 120, which may include those described with respect to Figs. 2-4. In some cases, UE 110 may find that it cannot obtain the actual measurements corresponding to these predictions and thus cannot evaluate these predictions.
[0119] At this point, at step 520, UE 110 can transmit a notification to the NW 120 indicating that the ground truth measurements are not available to the UE 110. In one example, this notification may be sent via RRC signaling, MAC CE or DCI. As an example, the notification may comprise status flag indicating measurement unavailability, configuration parameters (e.g., measurement reduction configuration used by the UE 110), time stamp when the notification was sent, and current network status of UE (e.g., RSRP, RSRQ or QoS etc.). Alternatively, the NW 120 can infer that the UE 110 may be unable to acquire actual measurements by active probing mechanisms, such as monitoring anomalies in uplink behavior or degrading reference signal quality.
[0120] In response to receipt of the indication from the UE 110, at step 530, the NW 120 may send a message to UE 110 instructing it to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
[0121] In some example embodiments, for use-cases where the ground truth measurements are not available to the UE 110 (e.g. measurement reduction in time, spatial, or frequency domain), the NW 120 can configure the UE 110 in a manner similar to those described above with regard to Figs. 2-4, but in addition, the NW 120 will provide configurations instructing the UE 110 to fall back to a legacy configuration in certain slots, in order to collect ground truth data and be able to compare it to predicted data to compute the configured KPIs required for monitoring (e.g., precision, recall, F1 score, hypothetical PDCP error rate or any other inter-mediate or system level KPI). In one embodiment, in the time domain measurement reduction, the NW 120 may configure the UE 110 to not perform measurement reduction in every 10thslot, so as to obtain the ground truth measurement, and then compare it to the prediction for that same slot, to calculate the prediction error. In another embodiment, the NW 120 may configure the UE 110 to not perform measurement reduction in every 3thslot. In yet another embodiment, the NW 120 may configure the UE 110 to not perform measurement reduction, i.e., zero measurement reduction, e.g., a message from the NW 120 can be detected via an RRC configuration that instructs the UE 110 to perform measurements in every slot. In one embodiment, measurement reduction configurations (e.g., designated measurement slots or ratio of measurement to prediction) can be transmitted by the NW 120 to the UE 110 via 3GPP-compliant signaling protocols, such as RRC messages or MAC CEs. The NW 120 can send different configurations of the measurement reduction to the UE 110. In one embodiment, upon receiving these configurations, the UE 110 can autonomously select an optimal configuration e.g., when ground truth measurements being not available and / or based on other real-time conditions such as battery status, channel qualityand mobility state etc. As an example, the UE 110 may switch from a sparse measurement configuration (e.g., 20% slots with measurement and 80% slots with prediction) to a dense measurement configuration (e.g., 80% slots with measurement and 20% slots with prediction).
[0122] In response to receipt of the message at step 530 from the NW 120, at step 540, the UE 110 may switch to the fallback configuration (e.g., perform and report measurements) in the configured slots, to obtain the ground truth measurements. In one example, the UE 110 may compare its predictions with the ground truth measurements to compute the configured KPIs required for monitoring of the predictions.
[0123] At step 550, the UE 110 can transmit a message indicating the measurement configuration to which it switches or falls back (e.g., in which slots measurements are performed and reported, or a configuration of 50% measurement, 80% measurement or 100% measurement (zero measurement reduction) etc.).
[0124] As a result, the proposed dynamic fallback mechanism enables the UE to be enforced to perform measurements in designated / configured slots, enables the UE to be enforced to perform measurements in designated / configured slots, in order to obtain ground truth measurements and hence calibrate its predictions by checking them against the ground truth ones.
[0125] Fig. 5 is provided as one example of a method according to certain embodiments. Other examples are possible as one or more procedures discussed above or illustrated in Fig. 5 may be skipped or may be performed in a different order, or additional procedures may be added. The steps of Fig.5 represent computer code instructions stored in the RAM memory of the device, which when executed by the central processing unit (CPU), carry out the functions of an example embodiment of the disclosure. Alternately, some or all of the steps in the procedure of the flow diagram may be embodied as hardware program logic included in programmed logic arrays of sequential and / orcombinatorial logic circuits and / or state machine logic implementing some or all of the steps performed by embodiments of the disclosure. The steps may be carried out in another order than shown and individual steps may be combined or separated into component steps.
[0126] Fig. 6 illustrates a flow diagram for an example method 600 for obtaining ground truth measurements when the ground truth measurements are not available to a terminal device in accordance with some example embodiments. The method 600 can be implemented at a terminal device, e.g. the terminal device 110 as discussed above. In some example embodiments, the method 600 may further include one or more steps that are performed at the terminal device 110 as described above with respect to Fig. 5. It would also be understood that details of some steps in the method 600 have been discussed above with respect to Fig. 5 and the method 300 will be described here in a simple manner. The steps of the flow diagram may represent computer code instructions stored in the RAM and / or ROM memory of the terminal device 110, which when executed by the central processing units (CPU), carry out the functions of the example embodiments of the present disclosure. The steps may be carried out in another order than shown and individual steps may be combined or separated into component steps.
[0127] According to some example embodiments, at 610, a terminal device may receive from a network device a configuration for monitoring predictions at the terminal device.
[0128] According to some example embodiments, at 620, the terminal device may transmit to the network device a notification indicating that ground truth measurements are not available to the terminal device.
[0129] According to some example embodiments, at 630, the terminal device may receive from the network device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
[0130] According to some example embodiments, the configuration and / or the message may be received via Radio Resource Control RRC, MAC and / or DCI
[0131] According to some example embodiments, the message may comprise information instructing the terminal device to perform and report measurements in the configured slots to obtain the ground truth measurements.
[0132] According to some example embodiments, cases where the ground truth measurements are not available to the terminal device may be caused by measurement reduction.
[0133] According to some example embodiments, the measurement reduction may comprise time domain measurement reduction, frequency domain measurement reduction or spatial domain measurement reduction.
[0134] According to some example embodiments, in cases of the measurement reduction, the message may comprise information instructing the terminal device to not perform the measurement reduction.
[0135] According to some example embodiments, wherein the configuration may comprise a timer for detecting a failure of the predictions.
[0136] According to some example embodiments, in response to receipt of the message from the network device, the terminal device may switch to the fallback configuration in the configured slots to obtain the ground truth measurements; the terminal device may detect the failure of the predictions based on the timer and the obtained ground truth measurements; and in response to the failure of the predictions being detected, the terminal device may transmit an indication of the failure to the network device.
[0137] According to some example embodiments, the configuration further may comprise a configured number of slots for which the predictions are monitored based on the ground truthmeasurements.
[0138] According to some example embodiments, the timer may be configured to define an amount of time in which the predictions are monitored for the configured number of slots based on the ground truth measurements.
[0139] According to some example embodiments, upon transmitting the indication of the failure, the terminal device may switch to the fallback configuration and transmit an indication of failure cause to the network device.
[0140] According to some example embodiments, the indication of failure cause may be transmitted via an RRC Reconfiguration complete message, DCI and / or an MAC CE.
[0141] In some example embodiments, an apparatus capable of performing the method 600 (for example, the terminal device 110) may comprise means for performing the respective steps of the method 600. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module.
[0142] Fig. 7 illustrates a flow diagram for an example method 700 for obtaining ground truth measurements when the ground truth measurements are not available to a terminal device in accordance with some example embodiments. The method 700 can be implemented at a network device, e.g. the network device 120 as discussed above. In some example embodiments, the method 700 may further include one or more steps that are performed at the network device 120 as described above with respect to Fig. 5. It would also be understood that details of some steps in the method 400 have been discussed above with respect to Fig. 5 and the method 400 will be described here in a simple manner. The steps of the flow diagram represent computer code instructions stored in the RAM and / or ROM memory of the network device 120, which when executed by the central processing units (CPU), carry out the functions of the example embodiments of the present disclosure. The steps maybe carried out in another order than shown and individual steps may be combined or separated into component steps.
[0143] According to some example embodiments, at 710, the network device may transmit to a terminal device a configuration for monitoring predictions at the terminal device.
[0144] According to some example embodiments, at 720, the network device may receive from the terminal device a notification indicating that ground truth measurements are not available to the terminal device.
[0145] According to some example embodiments, at 730, in response to receipt of the notification, the network device may transmit to the terminal device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
[0146] According to some example embodiments, the configuration and / or the message may be transmitted via RRC, MAC and / or DCI.
[0147] According to some example embodiments, the message may comprise information instructing the terminal device to perform and report measurements in the configured slots to obtain the ground truth measurements.
[0148] According to some example embodiments, cases where the ground truth measurements are not available to the terminal device may be caused by measurement reduction.
[0149] According to some example embodiments, the measurement reduction may comprise time domain measurement reduction, frequency domain measurement reduction or spatial domain measurement reduction.
[0150] According to some example embodiments, in cases for the measurement reduction, the message may comprise information instructing the terminal device to not perform the measurementreduction.
[0151] According to some example embodiments, the configuration may comprise a timer for detecting a failure of the predictions.
[0152] According to some example embodiments, the network device may receive an indication of the failure from the terminal device.
[0153] According to some example embodiments, the configuration may further comprise configured number of slots for which the predictions are monitored based on the ground truth measurements.
[0154] According to some example embodiments, the timer may be configured to define an amount of time in which the predictions are monitored for the configured number slots based on the ground truth measurements.
[0155] According to some example embodiments, the network device may receive an indication of failure cause from the terminal device along with the indication of the failure.
[0156] In some example embodiments, an apparatus capable of performing the method 700 (for example, the network device 120) may comprise means for performing the respective steps of the method 700. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module.
[0157] Fig. 8 illustrates a simplified block diagram of an apparatus 800 that is suitable for implementing some example embodiments of the present disclosure. The apparatus 800 may be provided to implement the communication device, for example, the terminal device 110, or the network device 120 as shown in Fig. 1. As shown, the apparatus 800 may include one or more processors 810, one or more memories 820 coupled to the processor 810, and one or more communication modules 840 coupled to the processor 810.
[0158] It should be understood that in some example embodiments, apparatus 800 may becomprised of an edge cloud server as a distributed computing system where the server and the radio node may be stand-alone apparatuses communicating with each other via a radio path or via a wired connection, or they may be located in a same entity communicating via a wired connection. For instance, in certain example embodiments where apparatus 800 represents a gNB, it may be configured in a central unit (CU) and distributed unit (DU) architecture that divides the gNB functionality. In such an architecture, the CU may be a logical node that includes gNB functions such as transfer of user data, mobility control, radio access network sharing, positioning, and / or session management, etc. The CU may control the operation of DU(s) over a front-haul interface. The DU may be a logical node that includes a subset of the gNB functions, depending on the functional split option. It should be noted that one of the ordinary skilled in the art would understand that apparatus 800 may include components or features not shown in Fig. 8.
[0159] The communication module 840 may be used for bidirectional communications. The communication module 840 may have at least one antenna to facilitate communication. The communication interface may represent any interface that is necessary for communication with other network elements.
[0160] The processor 810 may be of any type suitable to the local technical network and may include one or more of the following: general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs) and processors based on multicore processor architecture, as non-limiting examples. The apparatus 800 may have multiple processors, such as an application specific integrated circuit chip that is slaved in time to a clock which synchronizes the main processor.
[0161] The memory 820 may include one or more non-volatile memories and one or more volatile memories. Examples of the non-volatile memories may include, but are not limited to, a Read OnlyMemory (ROM) 824, an electrically programmable read only memory (EPROM), a flash memory, a hard disk, a compact disc (CD), a digital video disk (DVD), and other magnetic storage and / or optical storage. Examples of the volatile memories may include, but are not limited to, a random-access memory (RAM) 822 and other volatile memories that will not last in the power-down duration.
[0162] A computer program 830 includes computer executable instructions that are executed by the processor 810. The program 830 may be stored in ROM 824. The processor 810 may perform any suitable actions and processing by loading the program 830 into the RAM 822.
[0163] The embodiments of the present disclosure may be implemented by means of the program 830 so that the apparatus 800 may perform any process of the disclosure as discussed with reference to Figs. 2-7. The embodiments of the present disclosure may also be implemented by hardware or by a combination of software and hardware.
[0164] In some example embodiments, the program 830 may be tangibly contained in a computer readable medium which may be included in the apparatus 800 (such as in the memory 820) or other storage devices that are accessible by the apparatus 800. The apparatus 800 may load the program 830 from the computer readable medium to the RAM 822 for execution. The computer readable medium may include any types of tangible non-volatile storage, such as ROM, EPROM, a flash memory, a hard disk, CD, DVD, and the like.
[0165] Fig. 9 illustrates a block diagram of an example embodiment of a computer readable medium 900 in accordance with some example embodiments of the present disclosure. The computer readable medium 900 has the program 930 stored thereon. The computer readable medium 900 is depicted based on magnetic, electronic and / or optical technologies, such as magnetic disks, optical disks, semiconductor memory circuit devices and micro-SD memory cards (SD refers to the Secure Digital standard) for storing data and / or computer program code as an example computer program product,in accordance with at least one embodiment of the present disclosure. The computer readable medium 900 may be in any other form suitable for carrying or holding the program 930.
[0166] Using the description provided herein, the embodiments may be implemented as a machine, process, or article of manufacture by using standard programming and / or engineering techniques to produce programming software, firmware, hardware or any combination thereof.
[0167] Any resulting program(s), having computer readable program code, may be embodied on one or more computer readable media such as resident memory devices, smart cards or other removable memory devices, or transmitting devices, thereby making a computer program product or article of manufacture according to the embodiments. As such, the terms "article of manufacture" and "computer program product" as used herein are intended to encompass a computer program that exists permanently or temporarily on any computer readable medium.
[0168] It is to be noted that example embodiments may be implemented as circuitry, in software, hardware, application logic or a combination of software, hardware and application logic. In an example embodiment, the application logic, software or an instruction set is maintained on any computer program product or computer-readable medium. In the context of this document, a "computer-readable medium" may be any media or means that can contain, store, communicate, propagate or transport the instructions for use by or in connection with an instruction execution system, apparatus, or device, such as the radio access nodes or user equipment of the above-described example embodiments. The term "non-transitory" as used herein is a limitation of the medium itself (i.e. tangible, not a signal) as opposed to a limitation on data storage persistency (e.g. RAM vs ROM).
[0169] The example embodiments of the present disclosure also provide at least one computer program product tangibly stored on a non-transitory computer readable storage medium. The computer program product includes computer-executable instructions, such as those included inprogram modules, being executed in a device on a target real or virtual processor, to carry out the method as described above with reference to any of Figs. 2-7. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, or the like that perform particular tasks or implement particular abstract data types. The functionality of the program modules may be combined or split between program modules as desired in various embodiments. Machineexecutable instructions for program modules may be executed within a local or distributed device. In a distributed device, program modules may be located in both local and remote storage media.
[0170] Program code for carrying out methods of the example embodiments of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
[0171] More specific examples of the computer readable medium would include an electrical connection having one or more wires, a portable computer diskette, a hard disk, RAM, ROM, an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0172] Further, while operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances,multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are contained in the above discussions, these should not be construed as limitations on the scope of the present disclosure, but rather as descriptions of features that may be specific to particular embodiments. Certain features that are described in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment may also be implemented in multiple embodiments separately or in any suitable sub-combination.
[0173] Although the present disclosure has been described in languages specific to structural features and / or methodological acts, it is to be understood that the present disclosure defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.
Claims
CLAIMS1. A method, comprising:receiving, by a terminal device, from a network device a configuration for monitoring predictions at the terminal device;transmitting, by the terminal device, to the network device a notification indicating that ground truth measurements are not available to the terminal device; andreceiving, by the terminal device, from the network device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
2. The method of claim 1, wherein the configuration and / or the message is received via Radio Resource Control RRC, Medium Access Control MAC and / or Downlink Control Information DCI.
3. The method of claim 1 or 2, wherein the message comprises information instructing the terminal device to perform and report measurements in the configured slots to obtain the ground truth measurements.
4. The method of claim 1, wherein cases where the ground truth measurements are not available to the terminal device are caused by measurement reduction.
5. The method of claim 4, wherein the measurement reduction comprises time domain measurement reduction, frequency domain measurement reduction or spatial domain measurement reduction.
6. The method of claim 4, wherein in cases of the measurement reduction, the message comprises information instructing the terminal device to not perform the measurement reduction.
7. The method of claim 1 or 2, wherein the configuration comprises a timer for detecting a failure of the predictions.
8. The method of claim 7, further comprising:in response to receipt of the message from the network device, switching, by the terminal device, to the fallback configuration in the configured slots to obtain the ground truth measurements;detecting, by the terminal device, the failure of the predictions based on the timer and the obtained48ground truth measurements; andin response to the failure of the predictions being detected, transmitting, by the terminal device, an indication of the failure to the network device.
9. The method of claim 7 or 8, wherein the configuration further comprises a configured number of slots for which the predictions are monitored based on the ground truth measurements.
10. The method of claim 9, wherein the timer is configured to define an amount of time in which the predictions are monitored for the configured number of slots based on the ground truth measurements.
11. The method of claim 8, further comprising:upon transmitting the indication of the failure, switching, by the terminal device, to the fallback configuration and transmitting an indication of failure cause to the network device.
12. The method of claim 11, wherein the indication of failure cause is transmitted via an RRC Reconfiguration complete message, DCI and / or an MAC Control Element CE.
13. A method, comprising:transmitting, by a network device, to a terminal device a configuration for monitoring predictions at the terminal device; andreceiving, by the network device, from the terminal device a notification indicating that ground truth measurements are not available to the terminal device,in response to receipt of the notification, transmitting, by the network device, to the terminal device a message comprising information instructing the terminal device to switch to a fallback configuration in configured slots to obtain the ground truth measurements.
14. The method of claim 13, wherein the configuration and / or the message is transmitted via Radio Resource Control RRC, Medium Access Control MAC and / or Downlink Control Information DCI.
15. The method of claim 13 or 14, wherein the message comprises information instructing the terminal device to perform and report measurements in the configured slots to obtain the ground truth measurements.
16. The method of claim 13, wherein cases where the ground truth measurements are not available to the terminal device are caused by measurement reduction.
17. The method of claim 16, wherein the measurement reduction comprises time domain measurement reduction, frequency domain measurement reduction or spatial domain measurement reduction.
18. The method of claim 16, wherein in cases for the measurement reduction, the message comprises information instructing the terminal device to not perform the measurement reduction.
19. The method of claim 13 or 14, wherein the configuration comprises a timer for detecting a failure of the predictions.
20. The method of claim 19, further comprising:receiving, by the network device, an indication of the failure from the terminal device.
21. The method of claim 19 or 20, wherein the configuration further comprises configured number of slots for which the predictions are monitored based on the ground truth measurements.
22. The method of claim 21 , wherein the timer is configured to define an amount of time in which the predictions are monitored for the configured number slots based on the ground truth measurements.
23. The method of claim 20, further comprising:receiving, by the network device, an indication of failure cause from the terminal device along with the indication of the failure.
24. An apparatus, comprising:at least one processor; andat least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to perform a method according to any one of claims 1-12.
25. An apparatus, comprising:at least one processor; andat least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to perform a method according to any one of claims 13-23.5026. A computer program product comprising instructions which, when executed by at least one processor of an apparatus, cause the apparatus to perform the method according to any one of claims 1-23.
27. An apparatus comprising means for performing the method according to any one of claims 1-23.51
