Ai / ML enhancement for inter-frequency and spatial domain measurement prediction
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2026-02-02
- Publication Date
- 2026-08-13
Smart Images
Figure IB2026050945_13082026_PF_FP_ABST
Abstract
Description
[0001] AI / ML enhancement for inter-frequency and spatial domain measurement prediction
[0002] Field of the Invention
[0003] The present invention relates to wireless communication systems, and more particularly to methods and systems for enhanced artificial intelligence / machine learning (AI / ML) based mobility measurement prediction procedures that improve the reliability of inter-frequency and spatial domain measurements in 5G-Advanced networks.
[0004] Background of the Invention
[0005] Wireless communication systems have evolved rapidly over the past decades, with each generation bringing significant improvements in speed, capacity, and functionality. As networks transition to 5G and beyond, there is an increasing focus on enhancing mobility management to support seamless connectivity for users across different cells and frequency bands.
[0006] Radio resource management (RRM) plays a crucial role in optimizing network performance and user experience in mobile networks. A key aspect of RRM is mobility measurement, which involves the user equipment (UE) performing various signal measurements on serving and neighbouring cells to support handover decisions and cell reselection. These measurements typically include parameters like reference signal received power (RSRP), reference signal received quality (RSRQ), and signal-to-interference-plus-noise ratio (SINR).
[0007] In current systems, UEs are often configured to perform periodic inter-frequency and inter-radio access technology (RAT) measurements. However, these measurements can be resource-intensive, consuming significant battery power and processing capabilities of the UE. There is a need to balance the requirements for timely and accurate mobility measurements with the desire to conserve UE resources.
[0008] Recent advancements in artificial intelligence and machine learning (AI / ML) have opened up new possibilities for optimizing various aspects ofwireless networks. These techniques have shown promise in areas such as traffic prediction, resource allocation, and network optimization. There is growing interest in leveraging AI / ML capabilities to enhance mobility management and measurement procedures, including the potential to reduce reliance on frequent explicit radio measurements performed by the UE.
[0009] Predictive mobility techniques aim to anticipate user movement patterns and optimize network parameters accordingly. By leveraging historical data and contextual information, it may be possible to reduce the frequency of measurements while still maintaining reliable mobility support. However, implementing such predictive schemes presents several challenges, including model accuracy, adaptability to changing conditions, and integration with existing network procedures, particularly when predictions are applied across different frequency bands or spatial domains.
[0010] As networks become increasingly heterogeneous and complex, with a mix of macro cells, small cells, and new spectrum bands, effective mobility management becomes more challenging. There is a need for enhanced techniques that can intelligently manage inter-frequency and spatial domain measurements to support seamless mobility across diverse network deployments, without compromising measurement reliability.
[0011] Balancing measurement accuracy and resource efficiency remains an ongoing challenge in mobile networks. While frequent and comprehensive measurements can provide a more precise view of radio conditions, they come at the cost of increased signalling overhead and UE power consumption. Conversely, reducing measurement frequency or scope may lead to outdated information and suboptimal mobility decisions, particularly if prediction reliability is not sufficiently ensured.
[0012] As 5G networks continue to evolve towards 5G-Advanced and beyond, there are opportunities to revisit and enhance mobility measurement procedures to better balance measurement efficiency and reliability. Leveraging emerging technologies like AI / ML, along with theincreased processing capabilities of modem UEs, could potentially lead to more efficient and effective mobility management solutions, provided that prediction reliability can be adequately ensured.
[0013] Objective of the Invention
[0014] The principal objective of the present invention is to improve the reliability and accuracy of AI / ML-based measurement predictions for interfrequency and spatial domain mobility management by leveraging temporal domain prediction accuracy as a reliability-based enabling criterion.
[0015] Another objective of the present invention is to introduce a real-time evaluation mechanism where the network continuously assesses the accuracy of intra-frequency temporal domain predictions prior to enabling AI / ML models to predict inter-frequency and spatial domain measurements.
[0016] Another objective of the present invention is to ensure robust network performance by implementing a fallback mechanism that allows the UE to revert to legacy measurement procedures if AI / ML-based predictions fall below a predefined accuracy threshold.
[0017] Another objective of the present invention is to create a continuous feedback loop between the UE and the network, allowing AI / ML models to dynamically adjust based on observed prediction accuracy and environmental variations, ensuring continuous optimization of mobility predictions.
[0018] Another objective of the present invention is to minimize unnecessary direct radio measurements by optimizing the frequency and spatial measurement process, leading to lower UE power consumption and reduced network signalling overhead while maintaining accurate handover (HO) performance.
[0019] A further objective of the present invention is to enable scalable AI / ML-based mobility prediction across different network scenarios, including varying mobility speeds, environmental conditions, and frequencybands, ensuring adaptability for both FR1 and FR2 in 5G-Advanced networks.
[0020] Summary of the Invention
[0021] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.
[0022] According to an aspect of the present invention, a method for enhancing mobility measurement prediction in a wireless communication system is provided. The method includes configuring, by a network, a user equipment (UE) with an accuracy threshold parameter for intra-frequency temporal domain predictions. The method further includes receiving, from the UE, measurement data and prediction accuracy for intra-frequency temporal domain predictions. The method also includes evaluating whether the prediction accuracy exceeds the accuracy threshold parameter. In response to determining that the prediction accuracy exceeds the accuracy threshold parameter, the method includes instructing the UE to use a preconfigured measurement configuration for inter-frequency or spatial domain predictions.
[0023] According to other aspects of the present invention, the method may include one or more of the following features. The method may further include continuously monitoring the prediction accuracy for intra-frequency temporal domain predictions and instructing the UE to revert to a legacy measurement procedure if the prediction accuracy falls below a fallback threshold. The fallback threshold may be lower than the accuracy threshold parameter. The pre-configured measurement configuration may comprise instructions for the UE to perform cyclic measurements of serving and neighbouring cells at predetermined intervals. The cyclic measurements may comprise measuring a serving cell and predicting measurements for neighbouring cells at a first time instant, measuring a first neighbouring celland predicting measurements for the serving cell and other neighbouring cells at a second time instant, and measuring a second neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a third time instant. The pre-configured measurement configuration may prioritize measurements of the serving cell over measurements of neighbouring cells. Prioritizing measurements of the serving cell may comprise instructing the UE to perform two measurements of the serving cell for every one measurement of a neighbouring cell.
[0024] According to another aspect of the present invention, a user equipment (UE) for enhancing mobility measurement prediction in a wireless communication system is provided. The UE includes a processor and a memory storing instructions that, when executed by the processor, cause the UE to perform measurements for a serving cell, conduct intrafrequency temporal domain predictions, evaluate accuracy of the intrafrequency temporal domain predictions, and based on the accuracy exceeding a predefined threshold, perform inter-frequency or spatial domain predictions using a pre-configured measurement configuration.
[0025] According to other aspects of the present invention, the UE may include one or more of the following features. The instructions may further cause the UE to continuously monitor the accuracy of the intra-frequency temporal domain predictions and revert to a legacy measurement procedure if the accuracy falls below a fallback threshold. The fallback threshold may be lower than the predefined threshold. The pre-configured measurement configuration may comprise instructions for the UE to perform cyclic measurements of serving and neighbouring cells at predetermined intervals. The cyclic measurements may comprise measuring the serving cell and predicting measurements for neighbouring cells at a first time instant, measuring a first neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a second time instant, and measuring a second neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a third time instant. Thepre-configured measurement configuration may prioritize measurements of the serving cell over measurements of neighbouring cells. Prioritizing measurements of the serving cell may comprise performing two measurements of the serving cell for every one measurement of a neighbouring cell.
[0026] According to another aspect of the present invention, a system for enhancing mobility measurement prediction in a wireless communication system is provided. The system includes a base station configured to set an accuracy threshold parameter for intra-frequency temporal domain predictions, receive measurement data and prediction accuracy from a user equipment (UE), and instruct the UE to use a pre-configured measurement configuration for inter-frequency or spatial domain predictions when the prediction accuracy exceeds the accuracy threshold parameter. The system also includes the UE configured to perform measurements and predictions based on instructions from the base station.
[0027] According to other aspects of the present invention, the system may include one or more of the following features. The base station may be further configured to continuously monitor the prediction accuracy for intra-frequency temporal domain predictions and instruct the UE to revert to a legacy measurement procedure if the prediction accuracy falls below a fallback threshold. The fallback threshold may be lower than the accuracy threshold parameter. The pre-configured measurement configuration may comprise instructions for the UE to perform cyclic measurements of serving and neighbouring cells at predetermined intervals. The cyclic measurements may comprise measuring a serving cell and predicting measurements for neighbouring cells at a first time instant, measuring a first neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a second time instant, and measuring a second neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a third time instant. The pre-configured measurement configuration may prioritize measurements of the serving cellover measurements of neighbouring cells by instructing the UE to perform two measurements of the serving cell for every one measurement of a neighbouring cell.
[0028] The foregoing general description of the illustrative embodiments and the following detailed description thereof are merely exemplary aspects of the teachings of this disclosure and are not restrictive.
[0029] Brief description of the drawings
[0030] The figures described below depict various aspects of the system and methods disclosed herein. It should be understood that each figure depicts an embodiment of a particular aspect of the disclosed system and methods, and that each of the figures is intended to accord with a possible embodiment thereof. Further, wherever possible, the following description refers to the reference numerals included in the following figures, in which features depicted in multiple figures are designated with consistent reference numerals.
[0031] FIG. 1 illustrates a flowchart for AI / ML-based mobility prediction (100), in accordance with aspects of the present invention.
[0032] FIG. 2 depicts an inter-frequency measurement prediction (200), as per one embodiment of the present invention.
[0033] FIG. 3(a) shows an inter-frequency measurement prediction with equal weightage for serving and neighbouring cells (300), in accordance with aspects of the present invention.
[0034] FIG. 3(b) illustrates an inter-frequency measurement prediction with higher emphasis on the serving cell (300), according to one embodiment of the present invention.
[0035] FIG. 4(a) depicts a spatial domain measurement prediction with equal weightage for serving and neighbouring cells (400), as per one embodiment of the present invention.FIG. 4(b) presents a spatial domain measurement prediction with higher emphasis on the serving cell (400), in accordance with one embodiment of the present invention.
[0036] FIG. 5 illustrates the message flow for switching AI / ML-based mobility prediction with network signalling (500), as per one embodiment of the present invention.
[0037] FIG. 6 illustrates the message flow diagram for the switching of AI / ML with reduced network signalling (600), depicting the process by which the User Equipment (UE) and gNB interact to manage AI / ML-based measurement predictions.
[0038] FIG. 7 is a block diagram illustrating an example schematic hardware configuration of a network node (700), in accordance with one embodiment of the present invention.
[0039] Persons skilled in the art will appreciate that elements in the figures are illustrated for simplicity and clarity and may have not been drawn to scale. For example, the dimensions of some of the elements in the figure may be exaggerated relative to other elements to help to improve understanding of various exemplary embodiments of the present disclosure.
[0040] Throughout the drawings, it should be noted that like reference numbers are used to depict the same or similar elements, features, and structures.
[0041] Detailed Description of the Invention
[0042] The following description with reference to the accompanying drawings is provided to assist in a comprehensive understanding of exemplary embodiments of the invention as defined by the claims and their equivalents. It includes various specific details to assist in that understanding but these are to be regarded as merely exemplary.
[0043] Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Inaddition, descriptions of well-known functions and constructions are omitted for clarity and conciseness.
[0044] The terms and words used in the following description and claims are not limited to the bibliographical meanings but are merely used by the inventor to enable a clear and consistent understanding of the invention. Accordingly, it should be apparent to those skilled in the art that the following description of exemplary embodiments of the present invention are provided for illustration purpose only and not for the purpose of limiting the invention as defined by the appended claims and their equivalents.
[0045] It is to be understood that the singular forms “a,” “an,” and “the” include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to “a component surface” includes reference to one or more of such surfaces.
[0046] By the term “substantially” it is meant that the recited characteristic, parameter, or value need not be achieved exactly, but that deviations or variations, including for example, tolerances, measurement error, measurement accuracy limitations and other factors known to those of skill in the art, may occur in amounts that do not preclude the effect the characteristic is intended to provide.
[0047] As used herein, prediction accuracy refers to a measure of deviation between predicted and measured radio parameters, which may be determined based on absolute or relative error of metrics including, but not limited to, RSRP, RSRQ, SINR, beam quality indicators, or model-reported confidence values.
[0048] Figures discussed below, and the various embodiments used to describe the principles of the present disclosure in this patent document are by way of illustration only and should not be construed in any way that would limit the scope of the disclosure. Those skilled in the art will understand that the principles of the present disclosure may be implemented in any suitably arranged system. The terms used to describe various embodiments are exemplary. It should be understood that these are provided to merely aidthe understanding of the description, and that their use and definitions, in no way limit the scope of the invention. Terms first, second, and the like are used to differentiate between objects having the same terminology and are in no way intended to represent a chronological order, unless where explicitly stated otherwise. A set is defined as a non-empty set including at least one element.
[0049] The invention introduces an AI / ML-based mobility measurement prediction mechanism in which prediction for inter-frequency and spatial domains is enabled only when prediction reliability in the intra-frequency temporal domain is sufficient. The system first evaluates the accuracy of temporal domain measurement prediction for the serving cell and uses this accuracy as a control input to determine whether AI / ML-based prediction can be safely applied to additional frequency bands or spatial directions. When prediction reliability is high, the system selectively reduces direct measurements and relies on predicted values, and when prediction reliability degrades, the system increases direct measurements or reverts to legacy measurement procedures. By coupling prediction activation and fallback behaviour to observed prediction accuracy, the invention improves the robustness of mobility measurement prediction while reducing unnecessary measurement overhead and preserving reliable mobility performance. In particular, when inter-frequency or spatial domain predictions are derived from unreliable intra-frequency temporal predictions, prediction errors propagate across domains and amplify over time, leading to degraded handover decisions and unstable mobility performance.
[0050] FIG. 1 illustrates the flowchart for AI / ML-based mobility prediction (100), detailing the step-by-step process that the network and User Equipment (UE) follow to optimize inter-frequency and intra-frequency measurements for mobility management.
[0051] In one embodiment, the process begins with Step 105, where the network configures the UE with the necessary measurement configuration and AI / ML model parameters for intra-frequency temporal prediction. In thisstep, the network also provides the accuracy threshold (AAT) that is used to determine whether the AI / ML model is permitted to be used for interfrequency temporal predictions. Additionally, the network defines the fallback threshold (Afaiiback), which specifies when to revert to legacy measurement procedures if the predictions are inaccurate. The network also configures the measurement pattern for inter-frequency measurements that the UE will follow upon meeting the accuracy threshold.
[0052] In Step 110, the UE performs intra-frequency measurement predictions using the AI / ML model and calculates the prediction accuracy for the intra-frequency measurements. This step is crucial in determining whether the AI / ML model is reliable enough to be applied to inter-frequency measurements.
[0053] Following this, Step 115 includes comparing the intra-frequency prediction accuracy against the predefined AAT threshold. If the prediction accuracy is below AAT, the process moves to Step 140. Here, the system checks if the accuracy is less than the A tailback threshold (< A fallback). If this is the case, the process moves to Step 135, where the UE falls back to the legacy measurement procedure. On the other hand, if the prediction accuracy exceeds AAT, the process proceeds to Step 120.
[0054] In Step 120, if the intra-frequency prediction accuracy is satisfactory, the UE starts using the AI / ML model to predict inter-frequency temporal measurements. The UE applies the pre-configured observation window and measurement pattern for inter-frequency measurements, such as those for neighbouring cells or frequency bands. This allows the UE to predict the signal quality and other important metrics for frequencies beyond the serving cell, optimizing mobility management.
[0055] Once the inter-frequency predictions are in place, Step 125 involves the UE calculating both intra-frequency and inter-frequency measurement prediction accuracies. This ensures that the overall prediction system, including both intra-frequency and inter-frequency measurements, is operating with sufficient accuracy.In Step 130, the system evaluates whether the prediction accuracy for both domains (inter-frequency and intra-frequency) has fallen below the A tailback threshold. If the accuracy is still above this threshold, the process loops back to Step 125 for ongoing accuracy evaluation and refinement. If the accuracy is deemed insufficient, the process proceeds to Step 135.
[0056] In Step 135, if the prediction accuracy drops below the A fallback threshold, the UE falls back to legacy measurement procedures based on pre-configured measurement configuration. Alternatively, the UE indicates to the network that the prediction accuracy is below the A fallback the network instructs the UE to revert to traditional non-AI / ML based measurement methods, this ensures that mobility management continues reliably, even when AI / ML predictions are no longer accurate.
[0057] Finally, in Step 140, if the intra-frequency measurement prediction accuracy falls below the A fallback threshold, the process returns to Step 110 to re-evaluate the intra-frequency prediction accuracy, ensuring the system adapts and adjusts as needed to maintain prediction reliability.
[0058] FIG. 2 illustrates an inter-frequency measurement prediction (200), focusing on the use of AI / ML-based mobility prediction to optimize the prediction of radio measurements in neighbouring frequency bands. This inter-frequency measurement prediction approach represents a baseline AI / ML-based mobility prediction technique as discussed in existing study items and does not incorporate any reliability-based gating or activation mechanism. This method aims to reduce the number of actual measurements the User Equipment (UE) needs to perform, thereby improving energy efficiency, reducing signalling, and minimizing the network overhead related to transmitting reference signals (RSs).
[0059] In this embodiment, frequency f1 corresponds to the serving cell, which the UE directly measures. Meanwhile, frequencies f2 and f3 represent the neighbouring cells. The UE performs measurements at the serving frequency (f1), and based on the measured data, the AI / ML model predicts the measurements for the neighbouring cells (frequencies f2 and f3). Thisapproach significantly reduces the energy consumption of the UE by minimizing the need for direct measurements, relying on the model to predict measurements for the neighbouring frequencies instead.
[0060] However, this method does have several limitations. One of the primary challenges is the complexity of beamforming in spatial domain predictions. The UE’s position and orientation relative to the base station can greatly influence the accuracy of the model’s predictions. Variations in the UE’s antenna configurations and its gain across different spatial directions add another layer of complexity to the prediction process. Furthermore, if the UE does not measure all potential beams (due to limited antenna configurations or environmental obstructions), the AI / ML model may struggle to generate accurate predictions, particularly in dynamic environments with high mobility or changing obstacles. These factors can undermine the effectiveness of the prediction method for neighbouring cells (f2 and f3).
[0061] Another limitation arises from the unavailability of actual measurement values for frequencies f2 and f3. Since the AI / ML model is used to predict these measurements, it depends entirely on the model’s accuracy. If the model is not sufficiently trained or if the conditions in the network are significantly different from those encountered during training, the predictions may be inaccurate. Any errors in the model could lead to substantial discrepancies between the actual and predicted measurements for neighbouring cells, which could affect network decisions, such as handovers.
[0062] FIG. 3 illustrates an enhanced approach for inter-frequency measurement prediction, aiming to improve prediction accuracy and optimize mobility management in wireless networks (300).
[0063] In FIG. 3(a), the method involves measuring both the serving cell (f1) and the neighbouring cells (f2 and f3) at configured measurement intervals. The key feature of this approach is that the measurement configuration assigns equal measurement priority to the serving cell and the neighbouringcells, ensuring that predictions for all frequencies are treated with equal importance.
[0064] At time t1 , the UE measures the serving cell (f1 ) and uses the AI / ML model to predict the measurements for the neighbouring frequencies f2 and f3. At the next time interval (t2), the UE switches to measure frequency f2 and predicts the measurements for frequencies f1 and f3. In time interval t3, the UE measures frequency f3 and predicts the measurements for frequencies f1 and f2. This process continues cyclically, with the UE alternately measuring different frequencies and predicting the remaining frequencies in accordance with the configured measurement pattern. The cyclic measurement configuration is applied only after the intra-frequency temporal domain prediction accuracy satisfies the predefined accuracy threshold.
[0065] The primary advantage of this method is that it allows the UE to validate the accuracy of the AI / ML model’s predictions by comparing them to actual measurements taken in real-time. This cyclic validation helps the network and the UE refine and improve the accuracy of the predictions, making the overall mobility management more reliable. This approach helps reduce energy consumption, minimize network signalling overhead, and enhance the overall efficiency of the system, all while ensuring the accuracy of inter-frequency predictions.
[0066] In FIG. 3(b), a similar method is used, but with a slight modification that prioritizes the measurements of the serving cell (f1) over the neighbouring cells (f2 and f3). In this case, the UE takes two measurements of the serving cell (f1) before switching to measure the neighbouring frequencies. This ensures that the reliability of the serving cell’s measurement is maintained, which is critical for accurate prediction and mobility management.
[0067] The UE first measures the serving cell (f1) twice, ensuring that the prediction for the serving cell is highly reliable. After this, the UE switches to measure the neighbouring frequencies, using the AI / ML model to predict themeasurements for the neighbouring cells (f2 and f3). By prioritizing the serving cell’s measurements, this approach ensures that the serving cell prediction accuracy is not compromised, which is especially important when the serving cell is subject to more dynamic conditions (e.g., mobility or fluctuating network conditions).
[0068] This method allows the network to adjust the weighting of the measurements according to the priorities. For example, the network may decide to place more importance on the accuracy of the serving cell measurement if it is critical to ensure seamless handover or if the serving cell experiences frequent changes in signal quality. This prioritization helps to balance serving cell prediction accuracy with the need to gather data from neighbouring cells, thereby improving mobility management while optimizing UE energy consumption and network overhead.
[0069] FIG. 4 illustrates the enhanced approach for spatial domain AI / ML-based measurement prediction (400). The spatial domain prediction configuration is enabled only when the intra-frequency temporal domain prediction accuracy exceeds the predefined accuracy threshold. This method is focused on improving measurement accuracy and optimizing mobility management through spatial domain predictions, similar to the approaches for inter-frequency measurements previously discussed. Specifically, FIG. 4(a) and FIG. 4(b) illustrate two variations of spatial domain prediction that balance the weightage given to the serving cell and the other spatial directions.
[0070] In FIG. 4(a), the UE applies an equal weightage approach to both the serving cell and the other spatial directions in the spatial domain measurement prediction. This means that when the UE is predicting measurements for spatial domains (such as different beams or antennas), equal importance is placed on both the serving cell’s measurements and the measurements of other spatial directions.
[0071] At each measurement interval, the UE utilizes the AI / ML model to predict spatial domain measurements based on both the serving cell andspatial directions (spatial direction 1 and spatial direction 2). The AI / ML model relies on the data provided by the measurements of these cells to predict how the spatial domain (e.g., beam direction, signal strength) will behave across the spatial directions.
[0072] The key advantage of this method is that it allows for a balanced approach to prediction, where no single cell is prioritized over others. By equally considering both serving and other spatial directions, the method ensures that spatial domain measurements for all cells are taken into account, improving the overall prediction accuracy for mobility decisions such as handover and beam management. This equal weightage strategy allows for more comprehensive and reliable predictions, especially when the network needs to optimize resources based on multiple measurements in the spatial domain.
[0073] In FIG. 4(b), the approach places higher emphasis on the serving cell in the spatial domain measurement prediction. This means that the UE prioritizes measurements for the serving cell over other spatial directions (spatial direction 1 and spatial direction 2) to ensure that the serving cell’s prediction is more accurate and reliable. The idea is that the serving cell’s data is more critical in making immediate mobility decisions, such as maintaining an active connection or optimizing beam alignment.
[0074] In this approach, the UE might measure the serving cell multiple times (e.g., two measurements of serving cell) before shifting focus to other spatial directions. This prioritization ensures that the accuracy of the serving cell’s prediction remains high, even when network conditions change dynamically. Once the serving cell’s measurements are reliably predicted, the UE can then use the AI / ML model to predict the measurements for other spatial directions, allowing for more accurate beamforming and handover decisions.
[0075] This method allows the network to configure priorities based on mobility requirements, ensuring that the serving cell’s prediction is not compromised. For example, if the serving cell experiences frequenthandovers or high mobility, it may require more frequent updates and prioritization. By giving the serving cell higher weight in spatial domain predictions, this method improves overall prediction reliability while optimizing the use of neighbouring cell data for more informed decisions.
[0076] For both FIG. 4(a) and FIG. 4(b), the network ensures the reliability of the spatial domain predictions through a series of actions:
[0077] First, the network evaluates or requests the UE to report the accuracy of its intra-frequency temporal domain predictions (which relate to predictions for the serving cell). This is crucial because the temporal domain predictions directly influence the accuracy of spatial domain measurements.
[0078] If the intra-frequency temporal prediction accuracy exceeds a predefined threshold (AAT), the network configures the UE to use the preconfigured measurement setup for spatial domain predictions. This setup would utilize the AI / ML model for both serving and other spatial directions (as per FIG. 4(a) or FIG. 4(b)).
[0079] Once configured, the UE follows the setup to measure the serving and other spatial directions, using the AI / ML model to predict spatial domain measurements such as beamforming and antenna gain. The UE then reports these predictions back to the network, where they are used for further optimization and mobility management decisions.
[0080] To ensure ongoing accuracy, the network or UE continuously monitors the spatial domain prediction accuracy. If the accuracy starts to decline, the system can choose to increase the frequency of measurements or, if necessary, fallback to traditional measurement procedures to maintain reliability.
[0081] If the prediction accuracy drops below the established threshold (AAT), the network can instruct the UE to fallback to legacy measurement procedures, or the UE can automatically switch to these procedures based on predefined settings. This fallback mechanism ensures that network reliability is maintained, preventing inaccurate AI / ML-based predictions from compromising system performance.FIG. 5 illustrates the message flow diagram for the switching of AI / ML with network signalling (500), depicting the process by which the User Equipment (UE) and gNodeB (gNB) interact to manage AI / ML-based measurement predictions.
[0082] In one embodiment, the process begins when the UE enters a connected mode state (505) and awaits configuration instructions from the gNB. The gNB sends an RRC Reconfiguration message to the UE, which includes the necessary measurement configuration and the accuracy threshold (AAT). Upon receiving this configuration, the UE acknowledges by sending an RRC Reconfiguration Complete message back to the gNB.
[0083] Next, the UE performs the configured measurements (510). Once the measurements are completed, the UE evaluates the accuracy of the predictions. If the accuracy is greater than or equal to AAT, the UE sets the accuracy flag to 1 (515). The UE then generates a measurement report that contains the predicted values and the accuracy indication and sends this report to the gNB.
[0084] Upon receiving the report, the gNB checks the accuracy flag (520). If the accuracy flag is set to 1, the gNB may respond by either sending an updated measurement configuration or requesting the UE to continue using the pre-configured measurement configuration associated with the enhanced measurement approach.
[0085] The UE then performs measurements and predictions according to the updated configuration (525). After completing these measurements, the UE evaluates the accuracy again. If the accuracy falls below the fallback threshold (Afaiiback), the UE resets the accuracy flag (530) and sends a measurement report with the predicted values and the accuracy indication to the gNB.
[0086] At step 535, if the accuracy flag is reset, the gNB may either sends an updated measurement configuration or requests a fallback to legacy procedures. The UE then either updates the measurement configuration or falls back to legacy measurement procedures (540). Finally, the UE sendsanother measurement report containing the predicted values and the accuracy indication.
[0087] FIG. 6 illustrates the message flow diagram for the switching of AI / ML with reduced network signalling (600), depicting the process by which the User Equipment (UE) and gNB interact to manage AI / ML-based measurement predictions.
[0088] In one embodiment, the process begins when the UE enters RRC connected mode (605) and receives an RRC Reconfiguration message from the gNB. This message contains the measurement configuration, including accuracy thresholds (AAT and AFaiiback), as well as the AI / ML model parameters for intra-frequency, inter-frequency, and spatial domain measurements. Upon receiving this configuration, the UE acknowledges the reconfiguration by sending an RRC Reconfiguration Complete message back to the gNB.
[0089] The UE then performs the configured intra-frequency temporal predictions (610) and evaluates the prediction accuracy. If the accuracy meets or exceeds AAT, the UE starts using the AI / ML models for interfrequency or spatial domain predictions in accordance with the preconfigured measurement configuration (615). The UE sends a measurement report with the predicted values to the gNB.
[0090] The UE performs another accuracy evaluation. If the accuracy falls below AFaiiback, the UE resets the accuracy flag to indicate degraded prediction reliability (620). The UE then sends an RRC reconfiguration request to the gNB, which includes either updated measurement configurations or a request to fallback to legacy procedures. At step 625, if the accuracy indication is present, the gNB sends the updated measurement configuration or fallback request back to the UE.
[0091] The UE updates the measurement configuration or falls back to legacy measurement methods (630). The UE sends an RRC Reconfiguration Complete message and the measurement report, containing the predicted values and accuracy indication, back to the gNB.This flow enables the UE and gNB to manage A 1 / M L -based measurement predictions with reduced real-time network signalling, while adapting measurement behaviour based on prediction accuracy evaluated under pre-configured conditions. The process supports efficient mobility management by utilizing AI / ML-based predictions while maintaining network control over configuration and fallback decisions.
[0092] FIG. 7 is a block diagram illustrating an example schematic hardware configuration of a network node (700), in accordance with one embodiment of the present invention. The network node (700) may correspond to a base station, gNodeB, centralized unit, distributed unit, or any other network-side entity configured to support mobility measurement prediction and control in a wireless communication system. The network node (700) comprises a network interface (710), a processor (720), a memory (730), and a storage unit (740). The network interface (710), the processor (720), the memory (730), and the storage unit (740) are communicatively coupled to one another via one or more internal communication buses or interconnects.
[0093] The network interface (710) is configured to facilitate communication between the network node (700) and one or more user equipment (UEs) and / or other network entities. In one embodiment, the network interface (710) is configured to receive measurement reports, predicted measurement values, and prediction accuracy information associated with intra-frequency temporal domain measurements from the UE. The network interface (710) is further configured to transmit configuration information to the UE, including measurement configuration parameters, accuracy threshold parameters, fallback threshold parameters, and control information related to AI / ML-based mobility measurement prediction.
[0094] The processor (720) is configured to control the overall operation of the network node (700). The processor (720) executes instructions to process measurement information received from the UE, evaluate prediction accuracy associated with intra-frequency temporal domain measurements, and determine prediction reliability. Based on the evaluatedprediction accuracy, the processor (720) is configured to conditionally enable or restrict the use of AI / ML-based measurement prediction for interfrequency and spatial domains by instructing the UE to apply a corresponding pre-configured measurement configuration. The processor (720) is further configured to initiate fallback control by instructing the UE to revert to legacy measurement procedures when prediction accuracy falls below a predefined fallback threshold.
[0095] The memory (730) is configured to store program instructions and data that are executed or accessed by the processor (720). The memory (730) may store software modules implementing prediction accuracy evaluation, threshold comparison, and measurement control logic. The memory (730) may further store configuration data including accuracy threshold values, fallback threshold values, and measurement configuration profiles associated with AI / ML-based mobility measurement prediction.
[0096] The storage unit (740) provides non-volatile storage for persistent data and executable code. The storage unit (740) may store AI / ML model parameters, historical prediction accuracy information, predefined measurement configuration templates, and policy information used to support reliability-controlled mobility measurement prediction. The storage unit (740) may further store operational logs or statistical information for network optimization and performance analysis.
[0097] In operation, the processor (720), in cooperation with the network interface (710), the memory (730), and the storage unit (740), enables the network node (700) to implement a reliability-controlled AI / ML-based mobility measurement prediction mechanism. The hardware architecture illustrated in FIG. 7 enables the network node (700) to dynamically control UE measurement behaviour based on observed prediction accuracy, selectively enabling inter-frequency and spatial domain measurement prediction while ensuring reliable fallback to legacy measurement procedures when prediction reliability degrades.In summary, the invention provides a reliability-driven AI / ML-based mobility measurement prediction mechanism that adapts measurement behaviour based on observed prediction accuracy. By using intra-frequency temporal domain prediction accuracy as a control signal, the system dynamically determines when AI / ML-based prediction can be applied to inter-frequency and spatial domains and when direct measurements or legacy procedures should be used instead. This adaptive control ensures that prediction-based measurement reduction is applied only under reliable conditions, thereby improving the robustness of mobility management while reducing unnecessary measurement overhead and preserving stable handover and beam management performance.
[0098] While AI / ML-based mobility prediction techniques are known, existing approaches do not recognize that the reliability of inter-frequency and spatial domain predictions fundamentally depends on intra-frequency temporal domain prediction accuracy. The present invention uniquely identifies this dependency and introduces a reliability-gated activation and fallback mechanism to prevent compounded prediction errors.
[0099] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the disclosure. Accordingly, other implementations are within the scope of the following claims.
Claims
We Claim:
1. A method for enhancing mobility measurement prediction in a wireless communication system, comprising:configuring, by a network, a user equipment (UE) with an accuracy threshold parameter for intra-frequency temporal domain predictions;receiving, from the UE, measurement data and prediction accuracy for intra-frequency temporal domain predictions;evaluating whether the prediction accuracy exceeds the accuracy threshold parameter;determining that the prediction accuracy exceeds the accuracy threshold parameter, instructing the UE to use a pre-configured measurement configuration for inter-frequency or spatial domain predictions, wherein inter-frequency or spatial domain prediction is prohibited unless the accuracy threshold parameter is satisfied; and monitoring the prediction accuracy of intra-frequency temporal domain predictions and adjusting the pre-configured measurement configuration based on prediction reliability.
2. The method as claimed in claim 1 , further comprising:instructing the UE to revert to a legacy measurement procedure if the prediction accuracy falls below a fallback threshold.
3. The method as claimed in claim 2, wherein the fallback threshold is lower than the accuracy threshold parameter.
4. The method as claimed in claim 1, wherein the pre-configured measurement configuration comprises instructions for the UE to perform cyclic measurements of serving and neighbouring cells at predetermined intervals.
5. The method as claimed in claim 4, wherein the cyclic measurements comprise:measuring a serving cell and predicting measurements for neighbouring cells at a first time instant;measuring a first neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a second time instant; and measuring a second neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a third time instant.
6. A user equipment (UE) for enhancing mobility measurement prediction in a wireless communication system, comprising:a processor; anda memory storing instructions that, when executed by the processor, cause the UE to:perform measurements for a serving cell;conduct intra-frequency temporal domain predictions;evaluate the accuracy of intra-frequency temporal domain predictions;based on the accuracy exceeding a predefined threshold, perform inter-frequency or spatial domain predictions using a pre-configured measurement configuration; andadaptively adjust measurement behaviour based on evaluated prediction accuracy.
7. The UE as claimed in claim 6, wherein the instructions further cause the UE to:continuously monitor the accuracy of the intra-frequency temporal domain predictions; andrevert to a legacy measurement procedure if the accuracy falls below a fallback threshold.
8. The UE as claimed in claim 7, wherein the fallback threshold is lower than the predefined threshold.
9. The UE as claimed in claim 6, wherein the pre-configured measurement configuration comprises instructions for the UE to perform cyclic measurements of serving and neighbouring cells at predetermined intervals.
10. The UE as claimed in claim 9, wherein the cyclic measurements comprise:measuring the serving cell and predicting measurements for neighbouring cells at a first time instant;measuring a first neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a second time instant; and measuring a second neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a third time instant.
11. A system for enhancing mobility measurement prediction in a wireless communication system, comprising:a base station configured to:set an accuracy threshold parameter for intra-frequency temporal domain predictions;receive measurement data and prediction accuracy from a user equipment (UE);instruct the UE to use a pre-configured measurement configuration for inter-frequency or spatial domain predictions when the prediction accuracy exceeds the accuracy threshold parameter; anddynamically adjust prediction control parameters based on prediction accuracy.a user equipment (UE) configured to:perform measurements and predictions based on instructions from the base station;continuously report accuracy metrics; andrevert to legacy measurement procedures if prediction accuracy falls below a fallback threshold.
12. The system as claimed in claim 11 , wherein the base station is further configured to:continuously monitor the prediction accuracy for intra-frequency temporal domain predictions; andinstruct the UE to revert to a legacy measurement procedure if the prediction accuracy falls below a fallback threshold.
13. The system as claimed in claim 12, wherein the fallback threshold is lower than the accuracy threshold parameter.
14. The system as claimed in claim 11, wherein the pre-configured measurement configuration comprises instructions for the UE to perform cyclic measurements of serving and neighbouring cells at predetermined intervals.
15. The system as claimed in claim 14, wherein the cyclic measurements comprise:measuring a serving cell and predicting measurements for neighbouring cells at a first time instant;measuring a first neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a second time instant; and measuring a second neighbouring cell and predicting measurements for the serving cell and other neighbouring cells at a third time instant.
16. A network node in a wireless communication system, the network node comprising:a processor; anda memory storing instructions that, when executed by the processor, cause the network node to:configure a user equipment (UE) with an artificial intelligence / machine learning (AI / ML) model for mobility measurement prediction and with an accuracy threshold parameter for intra-frequency temporal domain prediction;receive, from the UE, measurement data and prediction accuracy information associated with intra-frequency temporal domain predictions;evaluate the prediction accuracy relative to the accuracy threshold parameter;conditionally enable inter-frequency or spatial domain mobility measurement prediction by instructing the UE to apply a preconfigured measurement configuration only when the evaluated prediction accuracy satisfies the accuracy threshold parameter; prevent execution of inter-frequency or spatial domain prediction at the UE when the evaluated prediction accuracy does not satisfy the accuracy threshold parameter;monitor ongoing prediction accuracy during application of inter-frequency or spatial domain prediction; andinstruct the UE to revert to a legacy measurement procedure when the prediction accuracy falls below a fallback threshold lower than the accuracy threshold parameter,wherein the accuracy of intra-frequency temporal domain prediction is used as a control gate for enabling or disabling AI / ML -based prediction across inter-frequency and spatial domains.