Methods and devices for enhancing user terminal mobility
By combining historical and real-time signal information from user terminals, a neural network model is used to predict future path nodes, optimize mobility operations, and solve the problems of handover failure and communication interruption caused by mobility decision-making in existing technologies, thereby improving communication continuity and stability.
Patent Information
- Application Number
- CN202611320863.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-27
- Publication Date
- 2026-11-17
AI Technical Summary
Existing mobility decision-making methods rely on real-time signal information and fixed thresholds, which can easily lead to handover failures, wireless link reconstruction, or communication interruptions, affecting user experience.
By combining historical and real-time communication signal information from user terminals, a neural network model is used to predict future path nodes and signal information, thereby optimizing measurement and handover operations in advance and improving communication continuity and stability.
It effectively reduces network drop rate and latency, improves the communication continuity and stability of user terminals in high-speed mobile scenarios, and enhances user experience.
Smart Images

Figure CN123123989A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of mobile communication technology, and in particular to a method and apparatus for enhancing the mobility of user terminals. Background Technology
[0002] With the development of fifth-generation mobile communication technology (5G) and future mobile communication systems, high-speed mobile scenarios such as high-speed rail and vehicle-to-everything (V2X) are becoming increasingly popular, placing higher demands on the timeliness and reliability of mobility management.
[0003] In existing technologies, user terminals monitor signal information such as Reference Signal Received Power (RSRP), Reference Signal Receiving Quality (RSRQ), and Signal-to-Interference-plus-Noise Ratio (SINR) of the serving cell (i.e., the stationary cell) and neighboring cells based on measurement rules configured by the network. When the triggering conditions of a specific measurement event are met, a measurement report is reported to the network side. The network side then decides whether to configure inter-frequency measurement, inter-system measurement, cell handover, or other mobility operations based on this measurement report.
[0004] However, existing mobility decision-making methods, which trigger measurement events by real-time measured signal information and fixed thresholds, are prone to handover failures, wireless link reconstruction, or communication interruptions, resulting in a poor user experience. Summary of the Invention
[0005] This application provides a method and device for enhancing user terminal mobility. By combining historical communication signal information with real-time communication signal information, it predicts future path nodes and / or future signal information, and accordingly triggers and optimizes mobility operations such as measurement, reporting, handover, or cell search in advance, thereby improving communication continuity and stability, reducing network drop rate and latency, and thus improving user experience.
[0006] In a first aspect, this application provides a method for enhancing the mobility of a user terminal, the method comprising:
[0007] Acquire historical and real-time communication signal information of the user terminal. The historical communication signal information includes the historical path node sequence and the historical signal information corresponding to each historical path node in the historical path node sequence. The real-time communication signal information includes the current path node and the corresponding current signal information.
[0008] Based on historical and real-time communication signal information, at least one future path node and / or future signal information is determined, including the future signal information of the current path node and / or the future signal information of each future path node.
[0009] Based on at least one future path node and / or future signal information, determine the mobility operation of the user terminal. The mobility operation includes at least one of measuring neighboring cell signals, reporting measurement events to the network side, handing over a cell, reselecting a cell, performing cell search, cell selection, radio link reconstruction, and radio link establishment.
[0010] In one possible implementation, the real-time communication signal information also includes at least one preceding path node of the current path node;
[0011] Based on historical and real-time communication signal information, determine at least one future path node and / or future signal information, including:
[0012] Construct the current movement trajectory based on the current path node and at least one preceding path node;
[0013] The historical path node sequence is matched with the current movement trajectory based on the path node identifier to extract at least one target historical path segment that matches the current movement trajectory from the historical path node sequence.
[0014] Based on at least one historical path node that appears consecutively after the last path node in the target historical path segment, determine at least one future path node and / or future signal information based on at least one historical path node that appears consecutively after the last path node in the target historical path segment and the corresponding historical signal information.
[0015] In one possible implementation, determining at least one future path node and / or future signal information based on historical communication signal information and real-time communication signal information includes:
[0016] Construct a historical spatiotemporal feature sequence based on the historical path node sequence and the historical signal information corresponding to each historical path node;
[0017] The prediction model is trained using historical spatiotemporal feature sequences. The prediction model is a machine learning model built on a neural network.
[0018] Construct the current spatiotemporal feature sequence based on the current path node and the current signal information of the current path node;
[0019] Input the current spatiotemporal feature sequence into the prediction model;
[0020] The predictive model outputs predicted values of at least one future path node's identifier and / or future signal information.
[0021] In one possible implementation, the prediction model is a multi-model fusion structure, including at least one path prediction sub-model and at least one signal information prediction sub-model.
[0022] The path prediction sub-model is used to predict the identifier of at least one future path node based on the current spatiotemporal feature sequence;
[0023] The signal information prediction sub-model is used to predict the future signal information of each future path node based on the current spatiotemporal feature sequence and the future path node identifier output by the path prediction sub-model, and / or to predict the future signal information of the current path node based on the current spatiotemporal feature sequence.
[0024] In one possible implementation, the method further includes:
[0025] During the application of the prediction model, the actual path nodes traversed by the user terminal and the corresponding actual signal information are collected in real time.
[0026] Error analysis is performed between the actual path nodes and the corresponding predicted future path nodes, and / or error analysis is performed between the actual signal information and the predicted values of the corresponding predicted future signal information to generate model analysis results.
[0027] When the model analysis results meet the preset retraining trigger conditions, the prediction model is incrementally trained or retrained based on the actual path nodes traversed and the corresponding actual signal information and / or historical communication signal information.
[0028] In one possible implementation, determining the user terminal's mobility operation based on at least one future path node and / or future signal information includes:
[0029] Obtain the current signal measurement value of the currently camped cell of the current path node;
[0030] When the future signal information of the current camping cell of the current path node indicates that a first communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a first preset signal quality threshold, perform co-frequency measurement to obtain the current signal measurement value of the neighboring cell at the same frequency point; and / or, when the future signal information of the current camping cell of the current path node indicates that a second communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a second preset signal quality threshold, report a measurement event to the network side to trigger inter-frequency or inter-system measurement configuration, or directly perform inter-frequency / inter-system measurement; and / or, based on the future path node and / or the future signal information of the future path node, set measurement priorities for multiple test frequencies and obtain the current signal measurement value of the neighboring cell of at least one of the multiple test frequencies;
[0031] When the future signal information of the current camped cell at the current path node indicates that a third communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than the third preset signal quality threshold, the measurement event used to trigger handover or cell reselection is reported to the network side.
[0032] Wherein, the first preset signal quality threshold is greater than or equal to the second preset signal quality threshold, and the second preset signal quality threshold is greater than the third preset signal quality threshold; the reporting of measurement events used to trigger inter-frequency or inter-system measurement configurations, and the reporting of measurement events used for handover, respectively include: directly triggering the reporting based on the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, or after correcting the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, triggering the reporting based on the corrected value;
[0033] After a Radio Link Failure (RLF) event or an Out of Coverage (OOC) event occurs, the cell search priority is determined based on future path nodes and / or the future signal information corresponding to the future path nodes.
[0034] In one possible implementation, before acquiring the historical communication signal information and real-time communication signal information of the user terminal, the method further includes:
[0035] Preprocessing and feature alignment are performed on the historical path node sequence in the communication signal records of the user terminal within a preset time window;
[0036] Calculate the path similarity index between each historical path node sequence;
[0037] Based on a preset path similarity threshold, historical path node sequences with path similarity indices within the preset path similarity threshold range are grouped into the same category to extract commonly used mobile paths; wherein, the preset path similarity threshold includes a fixed threshold or an adaptive threshold, and the adaptive threshold is dynamically learned through calculation or machine learning models based on the distribution of currently recorded historical paths;
[0038] Commonly used movement paths and their corresponding signal information are stored as historical communication signal information.
[0039] In one possible implementation, the method further includes:
[0040] Based on commonly used mobile paths, historical signal information corresponding to each historical path node is extracted from historical communication signal information, summary information corresponding to the historical path node is generated, and the summary information is associated with and stored with commonly used mobile paths.
[0041] Based on summary information and / or future signal information, report measurement events, trigger measurement procedures, select and execute cell handover or reselection, and determine the priority of measurement order and / or cell search order;
[0042] The summary information includes at least one of the following: signal quality statistics of each historical path node of each commonly used mobile path, the number and / or probability of RLF events, the number and / or probability of OOC events, the types of poor cell events and their corresponding number and / or probability, the number and / or probability of entering each node, and the average dwell time; the summary information of the same physical cell in the historical path nodes of different commonly used mobile paths is stored independently.
[0043] The signal quality statistics include: for each historical path node of each commonly used mobile path, calculating and recording the average of the previous preset number of times each type of measurement value was stationed at the node, and performing sliding window averaging or infinite impulse response (IIR) filtering on the average value corresponding to multiple stationings; and calculating and recording the number and / or probability of each measurement value being higher than each preset threshold each time stationed at the node, and performing sliding averaging or IIR filtering on the number and / or probability corresponding to multiple stationings.
[0044] The measured values include the Reference Received Power (RSRP) and the Signal-to-Interference-plus-Noise Ratio (SINR); poor cell events include: events where the number of times the cell's measured value is higher than a preset threshold is lower than a preset number threshold during a single dwell period; events where the probability of the cell's measured value being higher than a preset threshold is lower than a preset probability threshold; events where the dwell time of the cell is lower than a preset duration threshold; events where the cell's signal quality parameters are higher than a preset quality threshold and the resource allocation is lower than a preset allocation threshold; and events where network registration is rejected.
[0045] In one possible implementation, the method further includes:
[0046] When performing cell search, cell selection, radio link reconstruction, or radio link establishment, if one or more cells are detected on the same radio access technology (RAT) and frequency, then from the one or more cells, candidate cells corresponding to consecutive nodes after a path segment that matches a common mobile path are selected, and / or candidate cells corresponding to predicted future path nodes are selected.
[0047] Based on the Physical Cell Identifier (PCI) and summary information and / or future signal information of each candidate cell, assign a trial camping priority or access priority to the detected multiple cells;
[0048] If the summary information and / or future signal information corresponding to any candidate cell indicate that the probability of a bad cell event occurring in the candidate cell is greater than a first preset threshold, then a second preset threshold that is positively correlated with the probability of a bad cell event occurring is determined.
[0049] Generate a uniformly distributed random number in the range of 0 to 1, and prohibit camping or access to the candidate cell when the random number is less than a second preset threshold; wherein, the second preset threshold is positively correlated with the probability of a poor cell event, so that the probability of the corresponding candidate cell being prohibited from camping or accessing is positively correlated with the probability of a poor cell event.
[0050] If the currently camped or accessed cell detects a poor congestion cell event, or if the corresponding future signal information indicates that the probability of a poor congestion cell event in the current cell is greater than a third preset threshold, then the signal measurement values of the currently camped or accessed cell and / or the detected neighboring cells will be corrected.
[0051] Based on the corrected signal measurement values, cell reselection or cell handover is triggered to actively escape the current cell.
[0052] In one possible implementation, before acquiring the historical communication signal information and real-time communication signal information of the user terminal, the method further includes:
[0053] Set up a circular buffer of length N, where N is a positive integer, and use the circular buffer to temporarily store the online acquired communication signal samples;
[0054] The collected communication signal samples are written sequentially into a circular buffer in chronological order, and when the amount of data written exceeds N, the new data overwrites the earliest written data.
[0055] When a preset event is detected, a random number generator is started to generate an integer offset M, where M is greater than or equal to n and less than or equal to N-1-m, and n and m are preset non-negative integer buffer values;
[0056] After detecting a preset event, continue to collect and write M communication signal samples to the circular buffer, and backtrack from the current writing position to extract N consecutive historical samples from the circular buffer, and save the extracted N historical samples to the persistent storage medium.
[0057] The preset events include at least one of the following: RLF event, OOC event, or measurement report trigger event.
[0058] In one possible implementation, before acquiring the historical communication signal information and real-time communication signal information of the user terminal, the method further includes:
[0059] Obtain the complete historical movement path of the user terminal, which contains multiple path nodes arranged in chronological order;
[0060] Identify the target path node among multiple path nodes that has experienced communication quality anomalies;
[0061] Determine the position of the target path node in the complete historical movement path, and extract a local path segment containing the target path node, a first preset number of path nodes before the target path node, and a second preset number of path nodes after the target path node.
[0062] Local path segments and their corresponding signal information are stored as historical communication signal information.
[0063] Secondly, this application provides an apparatus for enhancing the mobility of a user terminal, the apparatus comprising:
[0064] The acquisition module is used to acquire historical communication signal information and real-time communication signal information of the user terminal. The historical communication signal information includes the historical path node sequence and the historical signal information corresponding to each historical path node in the historical path node sequence. The real-time communication signal information includes the current path node and the corresponding current signal information.
[0065] The prediction module is used to determine at least one future path node and / or future signal information based on historical communication signal information and real-time communication signal information. The future signal information includes the future signal information of the current path node and / or the future signal information of each future path node.
[0066] The determination module is used to determine the mobility operation of the user terminal based on at least one future path node and / or future signal information. The mobility operation includes at least one of measuring neighboring cell signals, reporting measurement events to the network side, handing over a cell, reselecting a cell, performing cell search, cell selection, radio link reconstruction, and radio link establishment.
[0067] In one possible implementation, the real-time communication signal information also includes at least one preceding path node of the current path node. The prediction module includes: a current trajectory construction module, a matching module, and a node determination module.
[0068] The current trajectory construction module is used to construct the current movement trajectory based on the current path node and at least one preceding path node;
[0069] The matching module is used to perform trajectory matching based on path node identifiers between the historical path node sequence and the current movement trajectory, so as to extract at least one target historical path segment that matches the current movement trajectory from the historical path node sequence.
[0070] The node determination module is used to determine at least one future path node based on at least one historical path node that appears consecutively after the last path node in the target historical path segment; and / or to determine at least one future path node and / or future signal information based on at least one historical path node that appears consecutively after the last path node in the target historical path segment and the corresponding historical signal information.
[0071] In one possible implementation, the prediction module further includes: a historical sequence construction module, a training module, a current sequence construction module, an input model module, and a result output module.
[0072] The historical sequence construction module is used to construct historical spatiotemporal feature sequences based on the historical path node sequence and the historical signal information corresponding to each historical path node.
[0073] The training module is used to train a prediction model using historical spatiotemporal feature sequences. The prediction model is a machine learning model built on a neural network.
[0074] The current sequence construction module is used to construct the current spatiotemporal feature sequence based on the current path node and the current signal information of the current path node;
[0075] The input model module is used to input the current spatiotemporal feature sequence into the prediction model;
[0076] The results output module is used to output predicted values of at least one future path node's identifier and / or future signal information through the prediction model.
[0077] In one possible implementation, the prediction module has a multi-model fusion structure, including at least one path prediction sub-model and at least one signal information prediction sub-model.
[0078] The path prediction sub-model is used to predict the identifier of at least one future path node based on the current spatiotemporal feature sequence;
[0079] The signal information prediction sub-model is used to predict the future signal information of each future path node based on the current spatiotemporal feature sequence and the future path node identifier output by the path prediction sub-model, and / or to predict the future signal information of the current path node based on the current spatiotemporal feature sequence.
[0080] In one possible implementation, the device further includes: an acquisition module, an error analysis module, and a training module.
[0081] The acquisition module is used to collect the actual path nodes and corresponding actual signal information actually traversed by the user terminal in real time during the application of the prediction model.
[0082] The error analysis module is used to perform error analysis between the actual path nodes and the corresponding predicted future path nodes, and / or to perform error analysis between the actual signal information and the predicted values of the corresponding predicted future signal information, and generate model analysis results.
[0083] The training module is used to incrementally train or retrain the prediction model based on the actual path nodes and corresponding actual signal information and / or historical communication signal information when the model analysis results meet the preset retraining trigger conditions.
[0084] In one possible implementation, the determining module includes: a current signal measurement value acquisition module, a first operation module, a second operation module, and a third operation module;
[0085] The current signal measurement value acquisition module is used to acquire the current signal measurement value of the currently camped cell of the current path node;
[0086] The first operation module is configured to perform co-frequency measurement and obtain the current signal measurement value of neighboring cells at the same frequency point when the future signal information of the current camping cell of the current path node indicates that a first communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a first preset signal quality threshold; and / or, when the future signal information of the current camping cell of the current path node indicates that a second communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a second preset signal quality threshold, report a measurement event for triggering inter-frequency or inter-system measurement configuration to the network side, or directly perform inter-frequency / inter-system measurement; and / or, based on the future path node and / or the future signal information of the future path node, set measurement priorities for multiple test frequency points and obtain the current signal measurement value of neighboring cells of at least one of the multiple test frequency points;
[0087] The second operation module is used to report a measurement event for triggering handover or to perform cell reselection to the network side when the future signal information of the current camp cell of the current path node indicates that a third communication abnormality event will occur and / or the signal quality parameter in the future signal information is lower than a third preset signal quality threshold.
[0088] Wherein, the first preset signal quality threshold is greater than or equal to the second preset signal quality threshold, and the second preset signal quality threshold is greater than the third preset signal quality threshold; the reporting of measurement events used to trigger inter-frequency or inter-system measurement configurations, and the reporting of measurement events used for handover, respectively include: directly triggering the reporting based on the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, or after correcting the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, triggering the reporting based on the corrected value;
[0089] The third operation module is used to determine the cell search priority based on future path nodes and / or the future signal information corresponding to the future path nodes after a radio link failure (RLF) event or an out-of-coverage (OOC) event occurs.
[0090] In one possible implementation, the device further includes: an alignment module, a similarity index calculation module, a path extraction module, and a path storage module.
[0091] The alignment module is used to preprocess and align the historical path node sequence in the communication signal records of the user terminal within a preset time window.
[0092] The similarity index calculation module is used to calculate the path similarity index between each historical path node sequence;
[0093] The path extraction module is used to group historical path node sequences with path similarity indices within the preset path similarity threshold range into the same category to extract commonly used mobile paths. The preset path similarity threshold includes a fixed threshold or an adaptive threshold. The adaptive threshold is dynamically learned by calculation or machine learning model based on the distribution of currently recorded historical paths.
[0094] The path storage module is used to store commonly used mobile paths and their corresponding signal information as historical communication signal information.
[0095] In one possible implementation, the device further includes: a summary information extraction module and a sequence determination module;
[0096] The summary information extraction module is used to extract historical signal information corresponding to each historical path node from historical communication signal information based on commonly used mobile paths, generate summary information corresponding to the historical path nodes, and store the summary information in association with commonly used mobile paths.
[0097] The sequence determination module is used to report measurement events, trigger measurement procedures, select and execute cell handover or reselection based on summary information and / or future signal information, and determine the priority of measurement sequence and / or cell search sequence.
[0098] The summary information includes at least one of the following: signal quality statistics of each historical path node of each commonly used mobile path, the number and / or probability of RLF events, the number and / or probability of OOC events, the types of poor cell events and their corresponding number and / or probability, the number and / or probability of entering each node, and the average dwell time; the summary information of the same physical cell in the historical path nodes of different commonly used mobile paths is stored independently.
[0099] The signal quality statistics include: for each historical path node of each commonly used mobile path, calculating and recording the average of the previous preset number of times each type of measurement value was stationed at the node, and performing sliding window averaging or infinite impulse response (IIR) filtering on the average value corresponding to multiple stationings; and calculating and recording the number and / or probability of each measurement value being higher than each preset threshold each time stationed at the node, and performing sliding averaging or IIR filtering on the number and / or probability corresponding to multiple stationings.
[0100] The measured values include the Reference Received Power (RSRP) and the Signal-to-Interference-plus-Noise Ratio (SINR); poor cell events include: events where the number of times the cell's measured value is higher than a preset threshold is lower than a preset number threshold during a single dwell period; events where the probability of the cell's measured value being higher than a preset threshold is lower than a preset probability threshold; events where the dwell time of the cell is lower than a preset duration threshold; events where the cell's signal quality parameters are higher than a preset quality threshold and the resource allocation is lower than a preset allocation threshold; and events where network registration is rejected.
[0101] In one possible implementation, the device further includes: a filtering module, a priority allocation module, a threshold determination module, a prohibition module, a correction module, and an avoidance module.
[0102] The filtering module is used to filter out, when performing cell search, cell selection, radio link reconstruction or radio link establishment, candidate cells corresponding to consecutive nodes after a path segment that matches a common mobile path and / or candidate cells corresponding to predicted future path nodes, if one or more cells are detected on the same radio access technology (RAT) and frequency point when one or more cells are detected.
[0103] The priority allocation module is used to allocate a trial camping priority or access priority to multiple detected cells based on the Physical Cell Identifier (PCI) and summary information and / or future signal information of each candidate cell.
[0104] The threshold determination module is used to determine a second preset threshold that is positively correlated with the probability of a poor cell event occurring if the summary information and / or future signal information corresponding to any candidate cell indicates that the probability of a poor cell event occurring in the candidate cell is greater than a first preset threshold.
[0105] The prohibition module is used to generate a uniformly distributed random number in the range of 0 to 1, and prohibit the initiation of camping or access to the candidate cell when the random number is less than a second preset threshold; wherein, the second preset threshold is positively correlated with the probability of a bad cell event, so that the probability of the corresponding candidate cell being prohibited from camping or accessing is positively correlated with the probability of a bad cell event.
[0106] The correction module is used to correct the signal measurement values of the currently camped or accessed cell and / or the detected neighboring cells if the cell currently camped or accessed detects a poor congestion cell event, or if the corresponding future signal information indicates that the probability of a poor congestion cell event in the current cell is greater than a third preset threshold.
[0107] The avoidance module is used to trigger cell reselection or cell handover based on the corrected signal measurement values in order to actively escape the current cell.
[0108] In one possible implementation, the device further includes: a storage setting module, a data writing module, an offset generation module, and a persistent storage module.
[0109] The storage setting module is used to set a circular buffer of length N, where N is a positive integer. The circular buffer is used to temporarily store online acquired communication signal samples.
[0110] The data writing module is used to write the collected communication signal samples sequentially into the circular buffer in chronological order, and when the amount of data written exceeds N, the new data overwrites the earliest written data;
[0111] The offset generation module is used to start a random number generator to generate an integer offset M when a preset event is detected, where M is greater than or equal to n and less than or equal to N-1-m, and n and m are preset non-negative integer buffer values;
[0112] The persistent storage module is used to continue to collect and write M communication signal samples to the circular buffer after detecting a preset event, and backtrack from the current writing position to extract N consecutive historical samples in the circular buffer, and save the extracted N historical samples to the persistent storage medium.
[0113] The preset events include at least one of the following: RLF event, OOC event, or measurement report trigger event.
[0114] In one possible implementation, the device further includes: a historical path acquisition module, an anomaly identification module, a local segment extraction module, and a local segment storage module.
[0115] The historical path acquisition module is used to acquire the complete historical movement path of the user terminal. The complete historical movement path contains multiple path nodes arranged in chronological order.
[0116] The anomaly detection module is used to identify target path nodes among multiple path nodes that have communication quality anomalies;
[0117] The local fragment extraction module is used to determine the position of the target path node in the complete historical movement path, and extract a local path fragment containing the target path node, a first preset number of path nodes before the target path node, and a second preset number of path nodes after the target path node.
[0118] The local segment storage module is used to store local path segments and their corresponding signal information as historical communication signal information.
[0119] Thirdly, this application provides an electronic device, including: a processor, and a memory communicatively connected to the processor.
[0120] The memory stores the instructions that the computer executes.
[0121] The processor executes computer execution instructions stored in memory to implement a method for enhancing the mobility of a user terminal, as described in the first aspect of the invention.
[0122] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement a method for enhancing the mobility of a user terminal according to the first aspect of the invention.
[0123] Fifthly, this application provides a computer program product, including a computer program, which, when executed by a processor, is used to implement a method for enhancing the mobility of a user terminal according to the first aspect of the invention.
[0124] In a sixth aspect, this application provides a chip including at least one processor for executing program instructions to implement a method for enhancing the mobility of a user terminal according to the first aspect of the invention.
[0125] Based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods.
[0126] This application provides a method and apparatus for enhancing user terminal mobility. By acquiring historical and real-time communication signal information of the user terminal, and jointly predicting at least one future path node and / or future signal information based on these two data points, the method can anticipate network coverage changes experienced by the user terminal. Since the historical path node sequence and the corresponding historical signal information reflect the actual movement trajectory and signal environment evolution of the user terminal, combining this with current information can effectively improve the accuracy of future signal prediction. Based on this, the user terminal can proactively trigger and optimize mobility operations such as measuring neighboring cell signals, reporting measurement events, handover, cell reselection, cell search, cell selection, radio link reconstruction, or radio link establishment. This avoids handover failures, connection interruptions, reconstruction delays, or link establishment delays caused by delayed decision-making or poor measurement and network search order, significantly improving connection continuity, handover success rate, and user experience during mobility. Attached Figure Description
[0127] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0128] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0129] Figure 1 This is a schematic diagram illustrating an application scenario of a method for enhancing user terminal mobility provided in an embodiment of this application.
[0130] Figure 2 A flowchart illustrating a method for enhancing user terminal mobility provided in an embodiment of this application;
[0131] Figure 3 A schematic diagram of a circular buffer structure provided in an embodiment of this application;
[0132] Figure 4 A schematic diagram of a device for enhancing user terminal mobility provided in an embodiment of this application;
[0133] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0134] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.
[0135] In the embodiments of this application, the terms "first" and "second" are used to distinguish identical or similar items with substantially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" do not necessarily imply difference. It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate that something is being used as an example, illustration, or description. Any embodiment or design scheme described as "exemplary" or "for example" in this application should not be construed as being better or more advantageous than other embodiments or design schemes. Specifically, the use of "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner. In the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more.
[0136] It should be noted that the phrase "at...time" in the embodiments of this application can refer to the instant at which a certain situation occurs, or to a period of time after the occurrence of a certain situation; the embodiments of this application do not specifically limit this. Furthermore, the method for enhancing user terminal mobility provided in the embodiments of this application is merely an example; a method for enhancing user terminal mobility may also include more or less content.
[0137] With the rapid deployment of 5G and future mobile communication systems, high-speed or high-dynamic mobile scenarios such as high-speed rail, urban rail transit, vehicle-to-everything (V2X) communication, and drone communication are becoming increasingly common. In these scenarios, user equipment (UE) traverses multiple cells at speeds of tens or even hundreds of kilometers per hour, posing unprecedented challenges to mobility management in order to maintain communication continuity and ensure user experience.
[0138] In existing technologies, mobility management primarily relies on event-triggered mechanisms. Specifically, user terminals monitor signal information such as RSRP, RSRQ, and SINR of the serving cell and neighboring cells according to measurement rules configured on the network side. When preset threshold conditions are met (such as the serving cell signal being below a threshold, or the neighboring cell signal being better than the serving cell by a certain offset) and this condition is maintained for a certain period of time, the user terminal in connected state can report a measurement report to the network side. Based on this measurement report, the network side can trigger measurement and handover. In idle state, the user terminal can autonomously perform cell reselection based on system broadcast information.
[0139] However, this mechanism has significant limitations: First, it relies entirely on the real-time signal quality within the current or short-term window, lacking the ability to predict user movement trajectories and signal change trends. If neighbor cell measurement is not initiated in a timely manner or the measurement sequence is poor, the handover may fail due to untimely measurement or a severely degraded radio link when the handover command is issued, ultimately leading to Radio Link Failure (RLF) and network drop. Second, to avoid frequent ping-pong handovers between two cells, a relatively long hysteresis or trigger time is usually set, resulting in delayed handover or reselection decisions when the signal deteriorates rapidly (e.g., entering tunnels or areas obstructed by tall buildings), thus triggering RLF or Out-of-Case (OOC) and causing network drop. Third, due to the lack of prediction of movement trajectories, poor network search sequence after a network drop can lead to long or even very long delays in restoring communication. These problems are particularly prominent in complex propagation environments (such as underground parking garages, dense urban areas, subway tunnels, and along high-speed railway lines), severely impacting communication continuity and user experience.
[0140] To address the aforementioned issues, this application proposes a method for enhancing user terminal mobility. By combining historical and real-time communication signal information of the user terminal, and based on the user terminal's historical movement behavior and signal evolution patterns, future path nodes and / or future signal information are predicted. The prediction results are then used to determine the user terminal's mobility operations, thereby triggering and optimizing mobility operations such as measurement, reporting, handover, or network search in a timely or proactive manner. This transforms passive response into proactive prediction. Furthermore, by optimizing the measurement and network search order, suitable cells are identified more quickly, preparing for mobility management operations such as handover, reselection, or reconstruction. This improves the communication continuity and stability of the user terminal during movement, reduces network drop rates and latency, and ultimately enhances the user experience.
[0141] Specifically, the system first acquires the user terminal's historical communication signal information (including historical path node sequences and their corresponding historical signal information) and real-time communication signal information (including current path nodes and their corresponding current signal information). Based on this, at least one future path node (next-hop cell) and / or future signal information (such as predicted RSRP, SINR, and whether a weak signal event has occurred) can be determined using a neural network model (a pre-trained prediction model) or a lightweight path matching algorithm.
[0142] Furthermore, based on the prediction results, intelligent decisions are made and corresponding mobility operations are executed. For example, when it is predicted that the signal of the current cell will deteriorate rapidly, measurements of the same frequency, different frequency, or different RAT (i.e., different systems) are triggered in advance. The measurement priority is dynamically adjusted by combining the predicted set of future path nodes with the network-configured measurement candidate set, prioritizing measurements of high-probability target frequencies. If the traditional handover event threshold is not reached, the measurement value is corrected based on the prediction confidence level, and events such as A3 or B1 are reported in advance. Here, A3 refers to an event where the neighboring cell signal quality is better than the serving cell by a certain offset, and B1 refers to an event where the neighboring cell signal quality of the different RAT is better than a threshold. Alternatively, if the traditional reselection conditions are not met, the measurement value is corrected based on the prediction confidence level, and reselection is performed in advance. When it is predicted that an RLF or OOC will occur, or when an RLF or OOC has already occurred, the target frequency information determined by the future path nodes is used to optimize the network search order during reconstruction, link establishment, or cell selection, significantly reducing latency. For path nodes that have frequently experienced poor cell events in the past (such as persistently poor signal quality or congestion), a probabilistic avoidance strategy is implemented to improve the communication experience.
[0143] The method presented in this application can be implemented independently on the user terminal side without requiring network-side modifications and is transparent to the network (i.e., the network side does not need to be aware of or participate in the process). It also supports dual-mode deployment: in high-performance terminals, a neural network model (a pre-trained prediction model) can be used for high-precision prediction; in resource-constrained user terminals, a low-overhead, commonly used movement path extraction and real-time matching mechanism can be employed, balancing performance and power consumption. Furthermore, the proposed temporary circular storage scheme and local path fragment storage scheme effectively reduce the pressure on user terminal storage resources from historical data storage.
[0144] The above-mentioned technical means can improve the mobile foresight, handover success rate and communication stability of user terminals in high-speed and complex scenarios, effectively reduce network drop rate and latency, and thus comprehensively improve user experience.
[0145] It should be noted that, in the embodiments of this application, the camping cell or access cell both refer to the cell currently selected by the user terminal (UE) for communication interaction, that is, the serving cell or primary cell defined in the 3GPP protocol.
[0146] In this application, the currently camped cell is the serving cell of the user terminal, regardless of whether the user terminal is in the idle mode (listening to signals) or the connected mode (transmitting service data). In this application, the user terminal camping on a certain cell means that the cell is the current serving cell of the user terminal, and the user terminal can be in the connected mode or the idle mode.
[0147] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.
[0148] To better understand the solutions of the embodiments of this application, an application scenario involved in the embodiments of this application will be introduced below.
[0149] For specific application scenarios of this application, please refer to [link / reference needed]. Figure 1 . Figure 1 This is a schematic diagram illustrating an application scenario for a method to enhance user terminal mobility provided in an embodiment of this application. It should be noted that... Figure 1 The examples shown are merely application scenarios that can be applied to the embodiments of this application, in order to help those skilled in the art understand the technical content of this application, but do not mean that the embodiments of this application cannot be used in other devices, systems, environments or scenarios.
[0150] like Figure 1 As shown, the application scenario includes: user terminal 101 and base station 102.
[0151] The user terminal 101 can be a smartphone, tablet, vehicle communication device, wearable device, or other mobile terminal supporting cellular communication. The base station 102 can be an evolved Node B (eNodeB) of a Fourth Generation (4G) Long-Term Evolution (LTE) system, a Next Generation Node B (gNodeB) of a Fifth Generation (5G) New Radio (NR) system, a 4G and 5G co-located base station, or other access nodes in future communication systems, used to provide wireless access services to the user terminal 101.
[0152] In actual deployment, user terminal 101 will pass through the coverage areas of multiple cells (such as base station 102 and its adjacent base stations) during its movement. User terminal 101 can continuously collect and store signal information and corresponding cell identification information during communication in each cell to form historical communication information.
[0153] The signal information collected by user terminal 101 may include RSRP, SINR, etc. The corresponding cell identification information may include Radio Access Technology (RAT), Absolute Radio Frequency Channel Number (ARFCN), Physical Cell Identity (PCI), Cell Identity (Cell ID), etc.
[0154] When user terminal 101 is in a current serving cell (i.e., the current camping cell, the current path node), it can combine its real-time communication signal information with historical communication signal information and use the method of this application to predict at least one future path node and / or future signal information that it may enter. The future signal information includes the future signal information of the current path node and / or the future signal information of each future path node. Based on this, it can perform and / or optimize mobility operations such as neighbor cell measurement, reporting measurement events, cell handover, cell reselection, reconstruction, link establishment, and cell selection in advance, thereby improving connection stability and improving user experience.
[0155] Specifically, the method of this application can be applied to complex wireless environments such as high-speed rail private networks, dense urban micro base station deployments, and underground rail transit systems, and is also applicable to other communication scenarios that require improved user terminal mobility.
[0156] Figure 2 This is a flowchart illustrating a method for enhancing user terminal mobility provided in an embodiment of this application. The execution entity of this embodiment can be... Figure 1 The user terminal 101 in the illustrated embodiment can also be other terminal devices with wireless communication capabilities, and this embodiment does not impose any particular restrictions on this. Figure 2 As shown, the method includes:
[0157] S101. Obtain historical and real-time communication signal information of the user terminal.
[0158] The historical communication signal information includes the historical path node sequence and the historical signal information corresponding to each historical path node in the historical path node sequence, while the real-time communication signal information includes the current path node and the corresponding current signal information.
[0159] In this step, the user terminal can be a smartphone, tablet, vehicle communication device, wearable device, or other single-mode, dual-mode, or even multi-mode mobile terminal that supports 4G LTE, 5G NR, or future communication technologies (such as 6G).
[0160] Specifically, during daily use, user terminals can continuously monitor and record signal information data of the surrounding wireless environment. This recorded data forms two types of information within the terminal: one is historical communication signal information obtained from signal information data accumulated over a period of time, and the other is real-time communication signal information obtained from currently occurring real-time signal information data.
[0161] Historical communication signal information refers to the network environment records experienced by the user terminal during several past moves. It contains a sequence of historical path nodes arranged in chronological order. Path nodes refer to the serving cells of the user terminal during its movement. Each path node can be identified by multiple network identifiers, such as: Tracking Area Code (TAC), RAT (e.g., 4G LTE or 5G New Radio), Frequency Band, ARFCN, and Physical Cell Identifier (PCI).
[0162] In addition, each historical path node can also be associated with corresponding historical signal information. For example, the sequence of RSRP, RSRQ, and SINR measurements and their statistical values (such as mean and variance) of the cell sampled over time reflect the dynamic evolution and statistical information of the signal strength and quality of the serving cell when the user terminal moves within the cell. Another example is the sequence of RSRP, RSRQ, and SINR measurements of several neighboring cells sampled over time, which reflects the dynamic fluctuation of the signal strength and quality of each neighboring cell as the terminal moves with its relative spatial position. Yet another example is the number of downlink duplicate packets dropped at the RLC layer of the serving cell, which indirectly reflects the uplink quality of the user terminal.
[0163] In addition, historical communication signal information can also include various communication anomalies, such as radio link failure (RLF) events, out-of-coverage (OOC) events, registration rejection events, and events that are judged as "poor cells" due to persistently weak signals and extremely short dwell times in the cell.
[0164] Specifically, historical communication signal information includes not only the aforementioned sequence of signal measurement values and abnormal events, but also basic time-related information, such as the time of entering the cell (staying or accessing), the time of leaving the cell (the reason for leaving could be handover, reselection, or RLF, etc.), and the duration of stay in the cell (i.e., the time interval from staying in the cell to the occurrence of handover, reselection, or RLF).
[0165] Furthermore, historical communication signal information may also include base station / network configuration information, such as the configuration of timers like T310 and T304, network configuration information triggered by A2 event reports at different thresholds (for example, an A2 event at one threshold triggers network configuration inter-system measurement, an A2 event at another threshold triggers network configuration inter-frequency measurement, and an A2 event at yet another threshold triggers network configuration blind redirection, etc.), neighbor cell information, etc.
[0166] Historical communication signal information can be obtained directly from the collected data of communication protocol layers such as the physical layer, RLC layer, RRC layer, and Non-Access Stratum (NAS) layer, or through processes such as calculation, condition judgment, and deduction.
[0167] During movement, the user terminal continuously collects current communication data to form real-time communication signal information. This real-time communication signal information includes the currently camped path node (e.g., RAT=5G NR, frequency band=n41, frequency point=50400, PCI=512), and the sequence of signal quality parameters (such as the RSRP, RSRQ, and SINR measurement values of the serving cell) formed by continuous sampling over time under the current path node; in addition, it also includes the sequence of signal quality parameters of several neighboring cells detected after triggering neighbor cell measurements, which are continuously sampled over time. As the terminal moves, this real-time data is continuously updated and, under specific conditions (such as idle periods or after the task ends), is accumulated and aggregated, and then transformed into the aforementioned historical communication signal information used for prediction or model training.
[0168] Similarly, real-time communication signal information can also include various communication quality events, such as weak signal events, extremely weak signal events, congestion events, RLF events, etc.
[0169] Furthermore, real-time communication signal information can also include time-related information, such as the time of entering the cell (staying or accessing the cell), the time of leaving the cell (leaving can be handover, reselection, or RLF, etc.), and the duration of stay in the cell (e.g., the duration from staying in the cell to handover, reselection, or RLF).
[0170] Furthermore, real-time communication signal information may also include base station / network configuration information, such as the configuration of timers like T310 and T304, network configuration information triggered by A2 event reporting at different thresholds (for example, an A2 event at one threshold triggers network configuration inter-system measurement, an A2 event at another threshold triggers network configuration inter-frequency measurement, and an A2 event at yet another threshold triggers network configuration blind redirection, etc.), neighbor cell information, etc.
[0171] These communication signal information can be obtained directly from the wireless communication protocol layer data collected by the user terminal or through processes such as calculation, condition judgment, and deduction, without relying on external positioning methods such as the Global Navigation Satellite System (GNSS). Therefore, they are suitable for scenarios where GNSS signals are unavailable, such as underground garages, tunnels, and indoor spaces.
[0172] S102. Based on historical communication signal information and real-time communication signal information, determine at least one future path node and / or future signal information.
[0173] The future signal information includes the future signal information of the current path node and / or the future signal information of each future path node.
[0174] After obtaining the aforementioned historical and real-time communication signal information, the user terminal can infer the network area it may enter next based on its own behavior patterns.
[0175] Specifically, user terminals can use trajectory matching, model prediction, or hybrid modes to determine future path nodes and / or future signal information.
[0176] Trajectory matching refers to the user terminal constructing the current real-time path node sequence (e.g., including the current path node and at least one preceding path node) into the current movement trajectory; and performing trajectory matching between the current movement trajectory and the locally stored historical path node sequence based on path node identifiers (such as Cell ID, PCI, etc.).
[0177] Furthermore, since the Physical Cell Identifier (PCI) is locally unique and can be reused between different base stations, in practical applications, the path node identifier can be a globally unique identifier jointly determined by Radio Access Technology (RAT), Frequency Point and Tracking Area Code (TAC), and Public Land Mobile Network (PLMN).
[0178] Specifically, the user terminal can extract one or more target historical path segments that match the current movement trajectory from the historical path node sequence; and determine at least one future path node based on at least one historical path node that appears consecutively after the last path node in the target historical path segment.
[0179] In this approach, the user terminal can directly utilize the signal information recorded in the historical path, or combine it with the current signal trend, to deduce the future signal information corresponding to the current node and future path nodes. For example, if historical records show that after entering a certain NR cell along a specific preceding trajectory, there were multiple instances of sudden signal drops followed by persistently poor or deteriorating signal, eventually leading to an RLF event (e.g., entering an underground tunnel or enclosed basement from an open area), followed by NR reconstruction failure, NR frequency scanning failure, and then switching to LTE frequency scanning, eventually connecting to the LTE cell with a long delay, then it is predicted that when entering the cell again along the same trajectory in the future, there is a high probability that a similar process of sudden signal drops followed by persistently poor or deteriorating signal will occur, eventually leading to a connection to the LTE cell. Based on this prediction, after entering the cell (e.g., when the NR cell outside the tunnel is the current serving cell), the user terminal can trigger LTE measurement (i.e., inter-system measurement), optimize the measurement order (prioritize measuring the LTE cell inside the tunnel), and report the measurement event that triggers handover in advance, thereby making a smooth handover to the LTE cell inside the tunnel highly probable. Even if the handover fails and an RLF occurs, the terminal can use this prior information to optimize the network search order, quickly find and connect to the LTE cell inside the tunnel, and significantly shorten the communication interruption latency.
[0180] In other words, the user terminal can compare its current real-time path (e.g., the three most recently visited cells) with commonly used movement trajectories stored locally. If the current movement pattern is found to be highly similar to a historical path (e.g., a daily commute route), future path nodes can be extracted from that historical path, i.e., one or more cells that may be entered next. Each future path node can also be composed of parameters such as RAT, frequency point, and PCI, and can include the probability of that future path node appearing in the historical path node sequence.
[0181] Furthermore, the user terminal can also combine historical signal information to predict the future signal information of these future path nodes, and / or predict the future signal information of the current path node.
[0182] For example, if the signal of the current cell has repeatedly shown a rapid drop in RSRP to below -110dBm after entering the tunnel in the past, accompanied by a high probability (e.g., 80%) of RLF events, it can be predicted that the signal of the current cell will also deteriorate rapidly this time, and there is a high possibility of network drop.
[0183] Model prediction refers to the process whereby the user terminal first constructs a historical spatiotemporal feature sequence based on the historical path node sequence and the historical signal information corresponding to each historical path node. Subsequently, this historical spatiotemporal feature sequence is used to train the prediction model. This prediction model is a machine learning model based on neural networks, such as Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), or Transformer Neural Network (Transformer). Through training, the prediction model can learn and master the evolution patterns of historical signals in the temporal and spatial dimensions.
[0184] During the model application phase, the user terminal can construct a current spatiotemporal feature sequence based on the current path node and its current signal information. This current spatiotemporal feature sequence is then input into the pre-trained prediction model. Finally, the prediction model outputs at least one identifier for a future path node (e.g., predicted cell identifier, frequency point, etc.) and / or predicted values for future signal information (e.g., predicted signal strength, signal quality level, etc.).
[0185] In this approach, the prediction model can be a multi-model fusion structure, such as containing a path prediction sub-model and a signal information prediction sub-model. The path prediction sub-model predicts future node identifiers, while the signal information prediction sub-model predicts future signal information; its input, in addition to current signal information, may include the future node identifiers output by the path prediction sub-model. This multi-model fusion structure decouples the complex joint prediction task into a cascaded task of first predicting the location and then predicting the signal based on the known location, thereby effectively reducing the difficulty of model training.
[0186] Hybrid mode refers to a user terminal combining trajectory matching and model prediction. First, trajectory matching is performed. Based on this, if the similarity of the matched path is below a preset threshold (i.e., the current path is not a commonly used path), the model prediction mechanism can be triggered. Furthermore, even if a historical path segment is matched, if the user is currently near a historically anomalous communication node (such as a node that has experienced RLF or poor cell events), the model prediction mechanism can still be triggered. In other words, to conserve the user terminal's computing resources and battery consumption, a more refined signal information prediction mechanism can be activated only when the user is identified as being near a path node with a higher probability of communication anomalies. This on-demand activated hybrid prediction mechanism effectively solves the problem of terminal computing power overload and rapid battery consumption caused by continuously running high-precision neural network model prediction, achieving a balance between enhanced mobility and terminal battery life.
[0187] In another implementation, the user terminal can first use lightweight trajectory matching for coarse screening to determine a set of candidate future path nodes; then, it can use model prediction to sort or correct the signal quality of the nodes in the candidate future path node set. This architecture of coarse screening followed by fine sorting can reduce the computational cost of the model while ensuring prediction accuracy.
[0188] S103. Determine the mobility operation of the user terminal based on at least one future path node and / or future signal information.
[0189] In this application embodiment, mobility operations include at least one of measuring neighboring cell signals, reporting measurement events to the network side, handing over cells, reselecting cells, performing cell search, cell selection, radio link reconstruction, and radio link establishment.
[0190] Based on the determined future path nodes and / or future signal information, the user terminal can further determine mobility operations. Regardless of the prediction or matching method used, it can ultimately be used to guide mobility operations. That is, based on the prediction of future paths and signal conditions, the user terminal can adjust and optimize its mobility strategy in advance, thereby improving communication continuity and user experience.
[0191] Mobility operations include, but are not limited to, measuring neighboring cell signals, reporting measurement events to the network side, handing over cells, reselecting cells, performing cell search, cell selection, radio link reconstruction, and radio link establishment.
[0192] Specifically, in connected mode, user terminals can optimize neighbor cell measurement strategies in the following ways:
[0193] Firstly, pre-triggered co-frequency measurement. If the signal strength of the currently serving cell is predicted to weaken before it reaches the co-frequency measurement threshold, the user terminal can start co-frequency measurement in advance.
[0194] Secondly, the measurement value can be corrected based on prediction to trigger the A2 event in advance. If a weak signal event is predicted to occur before the serving cell's signal meets the conditions for triggering a traditional A2 event (i.e., the serving cell signal is below a threshold for a certain period), the user terminal can immediately correct the current cell's measurement value (e.g., by adding a negative bias to the actual measurement value). The A2 event is reported after the corrected measurement value meets the A2 event conditions; alternatively, the current measurement value can be directly corrected to meet the threshold for reporting the A2 event, thus immediately triggering the A2 event report. In this way, the user terminal can trigger neighbor cell measurements configured on the network side for different frequencies or systems (e.g., if the current cell is 5G, triggering measurements of 4G neighbor cells), and prioritize measuring frequencies predicted to be future path nodes.
[0195] Third, a threshold bias strategy for historically abnormal nodes. When a historical path segment is matched and the current node has historically frequently experienced abnormal events such as RLF, in addition to starting synchronous measurement in advance, the user terminal can also add a positive bias to the trigger threshold of some A2 events (for example, adding a +3dB bias to the A2 event threshold with an RSRP threshold of -110dBm to -80dBm). When the measured value meets the A2 event after the correction threshold, the measured value is subtracted from the positive bias and reported to the network side. Since A2 events with excessively high thresholds typically do not trigger inter-frequency or inter-system measurements, while A2 events with excessively low thresholds are likely to trigger blind redirection by the network side, by appropriately increasing the threshold of A2 events within a certain range, A2 events that are likely to trigger inter-frequency or inter-system measurements by the network side can be reported in advance without triggering blind redirection. Alternatively, the user terminal can also more precisely add a positive bias to the threshold of only A2 events that will trigger inter-frequency and / or inter-system measurements, based on the network side configurations triggered by each A2 event recorded in historical information. This allows the terminal to measure inter-frequency or inter-system neighboring cells in a timely manner and prioritize the measurement of the frequency points corresponding to the matched future path nodes, thus making full preparations for smooth handover.
[0196] In idle state, neighbor cell signal measurement can also include: if a historical path segment is matched and the current node has historically experienced frequent out-of-coverage (OOC) events, or if the signal of the currently camped cell is predicted to weaken, then neighbor cell measurement can be initiated in advance. Furthermore, the user terminal can dynamically adjust the priority of cell measurement based on predicted or matched future path nodes, thereby accelerating the cell reselection process.
[0197] In connected mode, user terminals can optimize the handover process in the following ways:
[0198] Firstly, the measurement values are corrected based on predictions to trigger the switching event in advance.
[0199] If the signal strength of the current cell and / or neighboring cells has not yet reached the triggering conditions of a traditional handover triggering measurement event (such as an A3 or B1 event), and if it is predicted that an extremely weak signal event will occur in the current cell, and the current measurement value or predicted signal strength of the future path node (a neighboring cell) is relatively good, the user terminal can reasonably correct the current measurement value of the current cell and / or future path node (e.g., by adding an offset) to make it meet the event triggering conditions, thereby reporting to the network side in advance.
[0200] Secondly, the delay time can be shortened based on historical anomaly records or predictions. If a historical path segment is matched and the current node has historically experienced RLF events frequently, or if an extremely weak signal event is predicted to occur in the current cell, the user terminal can shorten the time-to-trigger (TTT) delay of the measurement event that triggers the handover, thereby prompting the network side to issue the handover command in advance.
[0201] Third, the target handover cell is selected based on prediction or path matching results. When multiple possible future path nodes exist, the user terminal can combine measurement values and prediction results or historical information to select the optimal target cell. For example, when the current measurement values of multiple neighboring cells are similar, the terminal can prioritize the cell with the lowest probability of historically occurring bad cell events (such as congestion, network registration rejection, etc.) as the target cell.
[0202] In idle state, the user terminal can also optimize the cell reselection process. Even if the signal of the currently camped cell and / or neighboring cells has not yet met the traditional reselection conditions, the terminal can combine the current measurement value with the predicted future signal value to perform the cell reselection operation in advance. Alternatively, if a historical path segment is matched, and the current node has historically experienced OOC events frequently, the user terminal can shorten the cell reselection hysteresis time and perform the cell reselection operation in advance. Furthermore, it can combine measurement values and prediction results or historical information to select the optimal target cell. For example, the user terminal can prioritize cells with good measurement values, good prediction quality, and a low probability of historically occurring poor cell events as target cells.
[0203] When an RLF or OOC event occurs (which may include both early triggering and strict network configuration condition triggering), the user terminal can optimize cell search in the radio link reconstruction, link establishment, or cell selection process in the following ways:
[0204] First, target frequency points are prioritized based on prediction or path matching results. During reconstruction, link establishment, or cell selection after reconstruction failure, user terminals can prioritize searching for cells on the Radio Access Technology (RAT) and frequency points corresponding to the predicted or matched future path nodes, thereby accelerating the reconstruction, link establishment, or cell selection process and significantly shortening recovery latency.
[0205] Secondly, the search priority is dynamically adjusted based on historical or predictive information. The user terminal can further combine historical signal information or predicted future signal information of future path nodes to prioritize the RAT and frequency points corresponding to each future path node, and strictly perform cell search according to this priority order.
[0206] Third, intelligent camping decisions are made by combining real-time physical layer signal characteristics with historical information. During wireless link reconstruction, establishment, or cell selection, when multiple cells are detected on a target RAT and frequency point, relying solely on the correlation peaks of the Primary Synchronization Signal (PSS) and / or Secondary Synchronization Signal (SSS) for selection can easily lead to inaccurate detection due to noise, interference, or multipath effects, resulting in long latency issues. Therefore, the user terminal can combine the predicted or path-matched Physical Cell Identifier (PCI) of the next possible path node, the corresponding future signal information or summary information, and the PSS and SSS correlation detection results of each detected cell to prioritize or comprehensively score each cell. Through this multi-dimensional joint evaluation, the terminal can quickly focus on a potential target cell with good communication quality that it has successfully camped on before, for example, when the PSS or SSS peaks of multiple cells are similar. This effectively improves the limitations of underlying physical signal detection, accelerates the reconstruction, link establishment, or cell selection process, and enhances the communication stability and user experience of terminals in scenarios with weak coverage, strong interference, or high-speed movement.
[0207] Fourth, a probabilistic avoidance strategy for poor cells based on historical anomaly probabilities or predicted confidence levels. User terminals maintain communication statistics for historical path nodes. During cell selection, radio link reconstruction, or radio link establishment, user terminals can retrieve records of poor cell events on historical paths for each candidate cell (i.e., predicted or matched next possible path nodes) and calculate the historical probability of poor cell events such as signal quality issues, dwell time, network registration rejection, and congestion. If the historical probability of a candidate cell's poor cell occurrence or the confidence level of its predicted poor cell event exceeds a preset avoidance threshold (e.g., 80%), even if the cell's current priority or overall score is high, the user terminal will lower the cell's dwell priority or directly remove it from the candidate list (i.e., probabilistic avoidance). The purpose of using probabilistic avoidance is also to achieve dynamic adaptation and fault tolerance in the algorithm. Specifically, for poor cells that frequently appear in historical statistics, they are not absolutely blocked, but rather a certain probability is retained for tentative dwelling or access. This mechanism effectively addresses dynamic changes in the network environment (for example, after network optimization, the congestion situation of cells that were previously frequently congested may have significantly improved). By preserving the opportunity for trial and error, it can promptly detect and adapt to updates in the network status, thereby avoiding misjudgments of cells that have restored communication quality due to the lag in historical data. Through this strategy of avoiding disadvantages, user terminals can avoid blindly camping on or accessing cells that appear to have good signals but are actually unable to communicate effectively, thus significantly improving service continuity after reconstruction, link establishment, or cell selection, and effectively adapting to network changes.
[0208] In this way, user terminals no longer passively wait for the signal to deteriorate to a critical point before taking action. Instead, based on their understanding of their own movement patterns and signal evolution patterns, they proactively, intelligently, and efficiently manage connection maintenance during the movement process.
[0209] This embodiment provides a method for enhancing the mobility of a user terminal, comprising: first, acquiring historical communication signal information and real-time communication signal information of the user terminal, wherein the historical communication signal information includes a sequence of historical path nodes and historical signal information corresponding to each historical path node in the sequence, and the real-time communication signal information includes the current path node and its corresponding current signal information; then, determining at least one future path node and / or future signal information based on the historical communication signal information and the real-time communication signal information, wherein the future signal information includes the future signal information of the current path node and / or the future signal information of each future path node; finally, determining a mobility operation of the user terminal based on at least one future path node and / or future signal information, wherein the mobility operation includes at least one of measuring neighboring cell signals, reporting measurement events to the network side, handing over a cell, reselecting a cell, performing cell search, cell selection, radio link reconstruction, and radio link establishment.
[0210] The following technical effects are achieved: By acquiring historical and real-time communication signal information from user terminals, and jointly determining at least one future path node and / or future signal information based on these two data points, the network coverage change trend experienced by user terminals can be predicted in advance. Since the historical path node sequence and the corresponding historical signal information for each historical path node in the sequence profoundly reflect the actual movement trajectory and signal environment evolution of the user terminal, combined with the current state, future path nodes and future signal information can be effectively predicted. Based on this, user terminals can proactively trigger and optimize mobility operations such as measuring neighboring cell signals, reporting measurement events, handover, cell reselection, cell search, cell selection, radio link reconstruction, or radio link establishment, rather than passively responding only when the signal deteriorates below a fixed threshold and remains below it for a fixed period. This effectively avoids problems such as handover failure, reselection failure, connection interruption, and reconstruction or link establishment delays caused by decision lag or poor measurement and network search order, significantly improving connection continuity and user experience during mobility.
[0211] In one possible implementation, the real-time communication signal information further includes at least one preceding path node of the current path node. Determining at least one future path node and / or future signal information based on historical communication signal information and real-time communication signal information includes: constructing a current movement trajectory based on the current path node and at least one preceding path node; performing trajectory matching based on path node identifiers between the historical path node sequence and the current movement trajectory to extract at least one target historical path segment matching the current movement trajectory from the historical path node sequence; determining at least one future path node based on at least one historical path node appearing consecutively after the last path node in the target historical path segment; and / or, determining at least one future path node and / or future signal information based on at least one historical path node appearing consecutively after the last path node in the target historical path segment and the corresponding historical signal information.
[0212] Specifically, the user terminal can operate in typical urban scenarios, such as commuting daily by subway or bus along fixed routes. The user terminal continuously collects and temporarily stores various communication signal information related data. The real-time communication signal information formed by these communication signal information related data includes: the physical identifier of the current path node (the current camped cell, i.e., the current serving cell) (e.g., using a globally unique identifier composed of RAT, frequency point, PCI, TAC, and PLMN), the physical identifier of neighboring cells, signal quality related information of the serving cell and neighboring cells (such as RSRP, RSRQ, SINR), the time of entering and leaving the current node, various event records (such as whether an RLF event occurred, whether A2 or A3 measurement events were triggered), and network configuration information (e.g., network configuration information triggered by A2 event reporting at different thresholds).
[0213] Meanwhile, the user terminal locally stores historical communication signal information accumulated within a preset time period (such as the past 30 days). This historical communication signal information may include a sequence of historical path nodes recorded in chronological order, as well as historical signal information corresponding to each historical path node (such as signal quality-related information, whether an RLF event has occurred, dwell time, whether a poor cell event has occurred, network configuration information, etc.).
[0214] When a user terminal is currently residing in a certain cell (denoted as "current path node", such as cell A), it can obtain the globally unique identifier of that node and read at least one of its preceding path nodes (such as preceding nodes O and P), thereby constructing a current movement trajectory that reflects the current movement direction and location. For example, the trajectory sequence is "P, O, A".
[0215] The user terminal can perform trajectory matching based on path node identifiers between the current movement trajectory and the stored historical path node sequence.
[0216] It should be noted that the matching process here does not rely on GNSS geographic coordinates, but rather on logical matching based on cell topology relationships. Considering phenomena such as ping-pong handover in real-world network environments (e.g., repeated handovers between adjacent cells T and S at their boundaries), before trajectory comparison, the current movement trajectory can be preprocessed by de-jittering or deduplication to remove redundant nodes that are continuously repeated or alternate repeatedly within a short period, thus restoring an effective trajectory sequence that reflects the true macroscopic direction of movement.
[0217] Subsequently, using methods such as sliding window, longest common subsequence algorithm, or lightweight clustering, segments of valid trajectory sequences that match cell identifiers (such as RAT, frequency point, PCI, etc.) in historical records and whose macroscopic change order and direction are consistent are retrieved. Specifically, the matching mechanism can include the following two scenarios:
[0218] First, a perfect match. If there is a continuous sequence of changes in the historical records that is completely consistent with the current trajectory (for example, if the current trajectory is "P, O, A" and the historical records show "P, O, A, B" or "P, O, A, C"), then the match is directly determined to be successful.
[0219] Secondly, incomplete matching (fault-tolerant matching). Considering that actual movement trajectories may deviate (e.g., multiple cells may exist at the same geographical location, and multiple visits may result in residing in different cells), the matching process allows setting a matching threshold. If the current trajectory differs from the historical sequence in some nodes, but the overall similarity reaches the preset threshold, a successful match can still be determined. For example, if the current trajectory is "P, T, S, O, A", while the historical record contains "P, M, S, O, A", although the intermediate nodes differ (T and M are different), the core path nodes highly overlap, and if the matching threshold requirement is met, a successful match can still be determined.
[0220] Similarly, the above-mentioned mechanism of full matching and partial matching is also applicable when determining "frequently used movement paths" to ensure that the identification of frequently used paths is both accurate and has sufficient stability.
[0221] After a successful match, the target historical path segment (such as "P, O, A" in the historical record) can be extracted, and the subsequent node that immediately follows this segment in the historical sequence, such as node B (or node C), can be obtained. Node B (or node C) can be identified as the predicted future path node. Next, the historical record corresponding to node A in the historical path segment "P, O, A" is retrieved (e.g., RLF event occurs in all three records), and it is predicted that node A will also experience an RLF event this time (if no action is taken). The signal quality related information of node B in the historical record corresponding to the historical path segment "P, O, A, B" is retrieved (e.g., three records are -95dBm, -100dBm, and -97dBm respectively), and its weighted average (e.g., -97dBm) is calculated as the predicted future signal information.
[0222] In other words, after a successful match, the user terminal can extract a target historical path segment (e.g., "P, O, A") that highly matches the current movement trajectory from the historical path. Based on this segment, the user terminal can further obtain historical communication signal records of the current path node (e.g., node A) and subsequent potential path nodes (e.g., node B or C), and determine prediction information from the following two dimensions:
[0223] First, assess the abnormal risk of the current node. The user terminal can retrieve the historical performance of the current path node A in the target historical path segment. For example, if the historical record shows that node A has experienced multiple RLF events in the "P, O, A" segment, it can be determined that the current node A has a high risk of connection interruption, thereby triggering the corresponding optimization strategy in advance.
[0224] Secondly, predicting the signal quality of future nodes. The user terminal can detect at least one historical path node (such as node B or C) that appears consecutively after the target historical path segment and obtain its corresponding historical signal information. The process of determining future signal information includes statistical processing or feature labeling of historical data: for example, obtaining the mean, median, and variance of historical SINR and RSRP (Reference Signal Received Power), as well as event labels (such as the "Poor Cell Event Occurred" label). It should be noted that for cases where there are multiple historical samples of the same future node (for example, node B appeared three times on different dates, and the mean RSRP of the first 5 occurrences after camping was -95dBm, -100dBm, and -97dBm, respectively; node C appeared three times on different dates, two of which were identified as "Poor Cell Event Occurred"), the user terminal can use various strategies to generate predicted values: including calculating the statistical mean and median, or using a time decay mechanism to give higher weight to recent data to calculate a weighted average or infinite impulse response (IIR) filter value; for labeled data, the number of occurrences and probability can also be calculated. Furthermore, if the historical trajectory forks after the current node (e.g., "P, O, A, B" and "P, O, A, C" exist in the historical records), the user terminal can also output multiple possible future nodes and their probabilities of occurrence (e.g., the probability of heading to node B is 50%, and the probability of heading to node C is 50%), and combine them with the corresponding historical signal information as a comprehensive prediction result.
[0225] Based on the above predictions, user terminals can perform a series of operations to enhance mobility in advance.
[0226] For example, if node A's historical records in the historical path segment "P, O, A" show frequent occurrences of RLF events, a positive bias value (such as 3dB) can be added to the threshold of A2 events with a threshold greater than a certain threshold (such as -110dBm) to report A2 events in advance and trigger the network side to configure inter-frequency or inter-system measurements.
[0227] If the network is configured to measure several frequency points at different frequencies, and the predicted frequencies of the next nodes B and C are also among these frequencies, then the measurement priority of the frequencies of B and C can be increased. If the predicted signal quality of node B is better than that of node C, and B and C belong to different frequency points, then the priority of the frequency of B can be set higher than that of C.
[0228] If the signal of node A is currently degraded, while the measurement of node B is relatively good, but the conditions for triggering event A5 have not yet been met, and the signal of node B is predicted to be good in the future, then a corrected measurement report (such as artificially increasing the neighboring cell RSRP) should be proactively reported to prompt the network side to issue a handover command in advance. Alternatively, if the signal of A is currently very poor, while the measurement of node B is relatively good, but the conditions for cell reselection have not yet been met, and the signal of node B is predicted to be good in the future, then cell reselection can be initiated in advance. Furthermore, if the measurement values of A and B have already met the conditions for event A5 or cell reselection, and the signal of A is already very poor, but the hysteresis time has not yet elapsed, the hysteresis time can be shortened before reporting event A5 or cell reselection.
[0229] If an RLF or OOC event occurs, the user terminal can abandon the traditional method of "first searching historical frequency points and then performing a full-band blind scan." Instead, it can prioritize cell search on the {RAT, frequency point} corresponding to the predicted future path node set {B, C} to quickly lock onto the target. If the signal quality of the predicted node B is better than that of node C, and B and C belong to different frequency points, the frequency point priority of B can be set higher than that of C. If B and C belong to the same RAT and frequency point, and cell search detects B and C on this frequency point, the PSS / SSS related peak values can be combined with the predicted signal information for joint scoring. If other cells that do not belong to the predicted future nodes are also detected, these other cells are scored only using the PSS / SSS related peak values, and priority is given to attempting to camp, rebuild, or establish a link on nodes with higher scores. If the next node C frequently experiences "bad cell events" (such as congestion) in the past, even if the synchronization signal of node C is detected (PSS / SSS peak values meet the standard), the user terminal will be prohibited from attempting to camp on that cell with a certain probability. In other words, this solution breaks away from the limitations of the traditional mechanism that "determines camping solely based on signal strength detected at the physical layer." By introducing probabilistic avoidance based on historical service data, it proactively filters out cells that "have strong signals but cannot actually provide effective service." When a user terminal is prohibited from camping, it will attempt to camp on other cells on the same frequency band according to priority, or switch to the next frequency band to search for cells to attempt reconstruction, link establishment, or cell selection.
[0230] The entire process does not require GNSS positioning, but only relies on the communication layer information that the user terminal can obtain, making it applicable to scenarios without satellite signals, such as underground tunnels and indoor shopping malls.
[0231] By introducing preceding path nodes to construct the current movement trajectory and matching it with historical paths, this method can predict the cell a user terminal will enter and its signal quality relative to the current cell with low computational complexity and storage overhead. Compared to traditional mobility mechanisms, this method achieves forward-looking decision-making and can dynamically optimize measurement and network search priorities. On the one hand, it effectively reduces connection interruptions caused by untimely measurements, poor measurement order, or untimely handover or reselection; on the other hand, after a connection interruption, the terminal can prioritize attempting reconstruction, link establishment, or cell selection on the Radio Access Technology (RAT) and frequency point where the predicted path node is located, thereby significantly reducing recovery latency. Especially in repetitive movement scenarios such as commuting and logistics transportation (particularly in weak coverage scenarios such as tunnels and underground parking lots), this solution can significantly improve handover and reselection success rates, reduce the occurrence of RLF and OOC, and even if they occur, can significantly shorten link recovery time and effectively avoid known "bad cells." Ultimately, without increasing network-side modification costs, it significantly enhances terminal mobility performance and user experience.
[0232] In one possible implementation, determining at least one future path node and / or future signal information based on historical communication signal information and real-time communication signal information includes: constructing a historical spatiotemporal feature sequence based on the historical path node sequence and the historical signal information corresponding to each historical path node; training a prediction model using the historical spatiotemporal feature sequence, wherein the prediction model is a machine learning model built based on a neural network; constructing a current spatiotemporal feature sequence based on the current path node and the current signal information of the current path node; inputting the current spatiotemporal feature sequence into the prediction model; and outputting the predicted value of the identifier of at least one future path node and / or the predicted value of the future signal information through the prediction model.
[0233] In this embodiment, the user terminal uses historical and real-time communication signal information and a pre-trained prediction model (a machine learning model based on a neural network) to predict future path nodes and future signal information, thereby achieving enhanced intelligent mobility. The entire process can include three stages: feature construction, model inference, and decision execution. The feature construction stage includes two sub-stages: historical spatiotemporal feature sequence construction and model training.
[0234] First, during the feature construction phase, the user terminal can continuously collect and locally store a series of communication signal information from its various communication layers (such as the RRC layer and the physical layer). This communication signal information includes: the path node identifier of each camped cell (e.g., composed of the Radio Access Technology (RAT), frequency point, and physical cell identifier PCI, TAC, PLMN), the corresponding signal information (such as the RSRP, SINR, mobility events, and RSRP and SINR of the current camped cell), and timestamps, etc.
[0235] To train the prediction model, the user terminal can extract historical path node sequences and corresponding historical signal information to construct a historical spatiotemporal feature sequence. This sequence is entirely based on information from each communication layer and does not rely on GNSS positioning information. Each item in the sequence can include node identifiers of the stationary cell and neighboring cells, as well as normalized RSRP and SINR values for the stationary cell and neighboring cells, to comprehensively depict the user terminal's historical movement trajectory and signal state evolution in time and space. Subsequently, the constructed historical spatiotemporal feature sequence can be used to train a machine learning model based on a neural network to obtain a pre-trained prediction model.
[0236] During the online inference phase after model training, when the user terminal is currently camped on a cell (i.e., the serving cell, referred to as the current path node), a current spatiotemporal feature sequence can be constructed based on the current path node and its current signal information. This current spatiotemporal feature sequence also does not rely on GNSS positioning information; its feature terms correspond to historical spatiotemporal feature sequences and are used to characterize the terminal's current real-time signal state and location attributes. Subsequently, the constructed current spatiotemporal feature sequence can be input into a pre-trained prediction model, which outputs the identifier of at least one future path node and / or predicted values of future signal information.
[0237] In one specific implementation, the prediction model can employ an LSTM (Long Short-Term Memory) model to jointly predict future path nodes and signal quality based on the current spatiotemporal feature sequence.
[0238] The prediction model can be a two-layer stacked Long Short-Term Memory (LSTM) network. The hidden layer configuration can be flexibly set to 16, 32, 64, 128, etc., following the empirical principle that the number of hidden units in the upper layer is less than or equal to that in the lower layer. This model learns the node and signal change characteristics of the user terminal on daily routes (such as commuting routes, subway lines). The model input can be a current spatiotemporal feature sequence of length L (L can be 10, 15, 20, etc.), where the feature vector at each time step contains multiple dimensions, such as the node identifiers of the current and neighboring cells, the normalized RSRP value (mapped to the [0, 1] interval), and the normalized SINR value. The model output can be the prediction results for the next K (e.g., 2) time steps, including the predicted RSRP value or prediction range classification and the predicted SINR value or prediction range classification for the current node; the PCI and its confidence level, the predicted RSRP value or prediction range classification, and the predicted SINR value or prediction range classification for each future node.
[0239] Furthermore, the prediction model can support a multi-step prediction mechanism, capable of outputting prediction results for the next K time steps (K≥1). In practical applications, the setting of the prediction step size K is a comprehensive trade-off between prediction accuracy and computational resource consumption, and can be flexibly set according to requirements.
[0240] The prediction model can also employ a hybrid architecture, such as LSTM and Graph Neural Network (GNN). On one hand, a "neighboring cell map" can be constructed based on historical path node sequences, the serving cell of each node, and measured neighboring cells. A Graph Convolutional Network (GCN) is then used to extract the spatial topology relationships of the cells. On the other hand, based on the nodes experienced by the user terminal and the historical temporal sequence of changes in various measurements, LSTM is used to learn the user terminal's movement patterns and signal changes. During real-time inference, a current spatiotemporal feature sequence can be constructed based on the current path node and its current signal information. The time-series hidden state output by the LSTM is fused with the spatial topology features output by the GCN, and the fused features are input into a fully connected layer to output the predicted value. To improve prediction accuracy, the current spatiotemporal feature sequence can also include at least one preceding path node of the current path node. This hybrid architecture, through the full coupling of spatiotemporal features, significantly improves prediction accuracy in complex wireless networks.
[0241] The prediction model can be fully trained offline. The training data comes from communication information continuously collected and stored locally by user terminals during daily use.
[0242] Specifically, an "event-triggered sampling" strategy can be adopted. Whenever a critical communication anomaly such as an RLF event or an OOC event is detected, the communication information from T seconds before the event to S seconds after accessing the next node is automatically saved. Then, the saved communication information is used to construct a historical spatiotemporal feature sequence as input for training samples. For example, for an RLF event, the path node sequence and corresponding signal information from 30 seconds before the event to 3 seconds after accessing the next node can be extracted to construct a historical spatiotemporal feature sequence. The corresponding training labels (ground truth) are the actual measurement values of the serving cell and neighboring cells before the event, and / or the event itself, and / or the actual physical cell identifier of the first cell (i.e., the next node) that was successfully accessed after the event, as well as the RSRP and SINR values of that cell and neighboring cells measured after camping / accessing that cell. In some implementations, linear interpolation or other data completion methods can be used to handle cases where, for example, there are no measurement values. Those skilled in the art can flexibly select appropriate anomaly handling strategies according to the actual data distribution and training requirements.
[0243] The loss function of a prediction model can be designed as a joint optimization objective for multiple tasks. For example,
[0244] L=L PCI +aL RSRP +bL SINR
[0245] Where L is the global total loss, representing the comprehensive error of the prediction model on the current training samples, used for backpropagation to update model parameters, and the optimization objective is to minimize this comprehensive error. PCI For the PCI classification task, the loss term can be Cross-Entropy Loss, used to supervise the prediction accuracy of the prediction model for the PCI of future path nodes; L RSRP For the RSRP prediction task, the loss term is L. SINR For the loss term of the SINR prediction task; L RSRP and L SINR Both can use Mean Absolute Error (MAE) loss or a custom classification loss function. Coefficients a=0.5 and b=0.5 are used to balance the proportion of loss for each task and the gradient scale during backpropagation. Coefficients a and b can also be dynamically adjusted for adaptive learning.
[0246] The training process can use the Adam optimizer, and the initial learning rate can be set to 1×10. -3 The batch size can be adjusted according to the amount of data. To prevent overfitting and ensure the convergence quality of the prediction model, a multi-dimensional training termination decision mechanism can be constructed.
[0247] Specifically, a separate validation set (e.g., 15% of the total data) can be used to monitor model performance in real time. Key evaluation metrics include: the probability that the true PCI equals the model-predicted PCI, or the proportion of true PCIs appearing in the top three of the model's prediction confidence (PCI Top-3 hit rate), the MAE of RSRP (target less than 5dB), and the MAE of SINR (target less than 3dB). If the overall performance of the validation set does not significantly improve within five consecutive training epochs (e.g., improvement less than 0.5%), training is automatically terminated. A maximum training epoch limit of 50 epochs can be set to avoid unnecessary computation. Training is considered complete when the prediction model reaches a preset performance threshold on the validation set (e.g., PCI hit rate ≥ 85% and RSRP MAE ≤ 2.8dB).
[0248] After training is complete, the prediction model can be further optimized through quantization compression and operator fusion to reduce memory usage and inference time, and finally safely deployed to user terminals for online real-time prediction.
[0249] The above training process ensures the accuracy, stability, and deployment feasibility of the prediction model in real, complex wireless environments.
[0250] If, during actual training, it is found that the model has difficulty converging due to reasons such as rapid changes in wireless signals, the above multi-task model can be split into two or more, each responsible for predicting the physical identifier of future nodes and the signal information, respectively, and trained using their respective loss functions.
[0251] During the online inference phase, the prediction model can receive a pre-constructed sequence of current spatiotemporal features as input and directly output the prediction results for one or more (e.g., two) future path nodes. The output includes the prediction confidence level of the PCI for each future cell, as well as the corresponding predicted RSRP and SINR values. For example, the prediction model can predict the next cell's PCI to be 201 (confidence level 85%), with an RSRP of -96dBm and a SINR of 5dB. Specifically, online inference can be triggered only near nodes with historical communication anomalies (e.g., cells that have experienced RLF events), meaning prediction is only initiated at nodes where mobility improvement is most needed, thus balancing mobility enhancement with the computational resources and power consumption of user terminals.
[0252] Based on the above predictions, user terminals can perform a series of forward-looking mobility operations to achieve intelligent mobility enhancement. Specifically, these include:
[0253] First, when poor signal quality is predicted for the current cell, the user terminal can initiate intra-frequency measurement in advance or report an A2 event to trigger network-side configuration of inter-frequency or inter-system measurement. If the frequency of the predicted future path node belongs to a test frequency assigned by the network side, the user terminal can increase the measurement priority of that frequency. In addition, based on the predicted future signal information, differentiated measurement priorities can be assigned to each test frequency, prioritizing the measurement of frequencies where cells with predicted better future signal quality are located.
[0254] Secondly, if the signal of the current cell is already poor, or if it is predicted that the signal of the current cell will deteriorate drastically, and the measured signal quality of the future path node is still acceptable (although it has not yet met the measurement event for traditional handover triggering, the predicted signal is good), the user terminal can proactively report the corrected measurement report to the network side in order to prompt the network side to issue a handover command in advance.
[0255] Finally, if a user terminal unfortunately experiences a radio link failure (RLF), cell search can be performed first within the predicted future path node set (RAT, frequency point). This method greatly reduces the need for blind frequency scanning, thereby significantly shortening the connection recovery latency.
[0256] By constructing a current spatiotemporal feature sequence that integrates location (such as physical identifiers and signal quality changes of the current cell and neighboring cells), signal quality (such as RSRP and SINR), and time information, and utilizing a pre-trained prediction model, this implementation achieves high-precision, low-latency prediction of future path nodes and signal conditions. Based on the prediction results, user terminals can proactively and intelligently optimize a series of mobility operations. Compared to traditional mobility management mechanisms that rely on fixed thresholds and hysteresis times, this solution can proactively and intelligently adjust measurement, reporting, handover, reselection, reconstruction, link establishment, and cell selection strategies, effectively avoiding connection interruptions and prolonged network drops caused by untimely measurement or reporting, or poor measurement or search order. Ultimately, without relying on external positioning systems (such as GNSS), it significantly improves connection continuity, handover success rate, and overall communication experience for users in high-speed mobile or complex wireless environments; simultaneously, through model compression and selective inference, it can effectively control the computational and power consumption overhead of user terminals.
[0257] Furthermore, in one possible implementation, the mobility operations that the user terminal should perform can also be directly predicted using a prediction model. Considering that directly predicting specific operations is highly dependent on the model and requires collecting richer data for training, those skilled in the art can flexibly choose the aforementioned "prediction node and signal" scheme, or adopt the "direct prediction operation" scheme, or combine both to achieve the optimal technical effect, depending on the needs of the actual application scenario.
[0258] In one alternative implementation, to further improve prediction accuracy, the current spatiotemporal feature sequence can be expanded to include more preceding path nodes that contain the current path node. By introducing preceding node information over a longer time span, the terminal's movement trend can be more fully characterized, thereby effectively assisting in the identification of potential target cells.
[0259] In one possible implementation, the prediction model is a multi-model fusion structure, including at least one path prediction sub-model and at least one signal information prediction sub-model; the path prediction sub-model is used to predict the identifier of at least one future path node based on the current spatiotemporal feature sequence; the signal information prediction sub-model is used to predict the predicted value of the future signal information of each future path node based on the current spatiotemporal feature sequence and the future path node identifier output by the path prediction sub-model, and / or predict the predicted value of the future signal information of the current path node based on the current spatiotemporal feature sequence.
[0260] Specifically, the prediction model proposed in this application embodiment can be composed of two working sub-models: a path prediction sub-model and a signal information prediction sub-model.
[0261] First, the path prediction sub-model takes the current spatiotemporal feature sequence as input, which includes, but is not limited to, the current path node and L (e.g., 5) preceding node sequences, and the signal information corresponding to each node (such as serving cell and neighboring cell measurements, serving cell dwell time, communication anomaly event markers, etc.). This path prediction sub-model can be based on a neural network architecture (e.g., a Transformer or LSTM encoder and decoder), and outputs an identifier for at least one future path node. Each future path node identifier can be a cell's physical identifier (e.g., RAT, frequency point, PCI, etc.), used to characterize the network topology location most likely to be traversed by the user. This sub-model predicts current node changes by learning historical node (cell) change patterns.
[0262] Subsequently, the signal information prediction sub-model can use the current spatiotemporal feature sequence used by the path prediction sub-model, or construct different current spatiotemporal feature sequences, and take these current spatiotemporal feature sequences and the future path node identifiers output by the path prediction sub-model as input. The current spatiotemporal feature sequence used by the path prediction sub-model can include more preceding nodes, more fully characterizing the terminal's movement trend, and can have a larger time step; while the current spatiotemporal feature sequence used by the signal information prediction sub-model can only include the current path node and its signal information (such as measurements of the serving cell and neighboring cells) to predict the signal trend in the near future, and can have a smaller time step. The signal information prediction sub-model can predict the predicted value of the signal information of the current path node at future times, and can also predict the predicted value of the future signal information of each future path node.
[0263] This signal information prediction sub-model can utilize an LSTM architecture, or it can use a multilayer perceptron (MLP) regression layer to perform nonlinear mapping on the temporal hidden state to perform regression prediction on the signal trend in the short term, thereby obtaining the future signal information prediction values of the current path node and the future path node.
[0264] This embodiment first outputs the cell identifiers (such as PCI) that may be passed through in the future from the path prediction sub-model, and then inputs the result into the signal information prediction sub-model. This decouples the prediction of future path nodes and the prediction of future signal information, effectively reducing the difficulty of model training and the overall complexity of the model.
[0265] In one possible implementation, the method further includes: during the application of the prediction model, real-time acquisition of the actual path nodes traversed by the user terminal and the corresponding actual signal information; performing error analysis between the actual path nodes and the corresponding predicted future path nodes, and / or performing error analysis between the actual signal information and the predicted values of the corresponding predicted future signal information, to generate model analysis results; when the model analysis results meet the preset retraining trigger conditions, incremental training or retraining of the prediction model based on the actual path nodes traversed and the corresponding actual signal information and / or historical communication signal information.
[0266] During the model deployment and application phase, real-time data on actual signal information from user terminals is continuously collected, including: the physical identifier of the cell where the user terminal is actually stationed or accessing (such as RAT, frequency point, PCI, etc.), the actual RSRP and SINR values of the serving cell and neighboring cells, communication anomalies, and various poor cell events. This actual signal information is then aligned with the predicted values of future path nodes and future signal information previously predicted by the model. This alignment can be achieved through timestamp matching or sliding time window mapping. For example, for the RSRP and SINR of the serving cell, the predicted value at prediction time t+k can be paired with the measured value collected at actual time t+k to calculate the prediction error (e.g., if the predicted RSRP of the serving cell is -100dBm and the actual value is -105dBm, then the RSRP MAE is 5dB). For example, for predicted future nodes, the physical identifier of the predicted future node can be compared with the physical identifier of the actual next node. If they match, the prediction is correct; otherwise, the prediction is incorrect (e.g., if the predicted next node is LTE, frequency 38400, PCI=123, but the actual access is LTE, frequency 38000, PCI=456, then the prediction is incorrect). If the path node is correct but the RSRP predicted value is classified as (-100dBm to -95dBm), while the actual value is -102dBm, then the RSRP predicted value is classified incorrectly.
[0267] Based on the above-mentioned method for calculating prediction error, model analysis results can be generated through multi-sample statistics, including indicators such as the accuracy of future path nodes, the MAE of RSRP / SINR, and the proportion of abnormal predictions for various events. When preset retraining trigger conditions are met, for example, if any of the following conditions are met: the average MAE of recent actual samples exceeds a threshold (e.g., 3dB), or the path prediction accuracy is lower than a threshold (e.g., 85%), the model update process can be automatically triggered at an appropriate time (e.g., at night).
[0268] Specifically, the model update process can also be triggered based on a comparison between the actual communication quality after model deployment and historical communication quality. If a decline in communication quality is detected in a certain area, such as an increase in the probability of RLF (e.g., the historical probability is 50%, but the probability after model deployment is 60%) or an increase in network drop duration (e.g., the historical average network drop latency is 10 seconds, but the average network drop latency after model deployment increases to 20 seconds), the model update process can be triggered.
[0269] Specifically, if the model's performance is poor, that is, the model has a high accuracy and small error on the validation set, but a low accuracy and large error in actual applications, or the actual communication quality has declined, the recently accumulated test samples, that is, the test samples collected after the last model training (including the actual path nodes actually traversed by the user terminal and their corresponding actual signal information), can be used as new training data to incrementally train the prediction model.
[0270] Furthermore, to break free from reliance on model predictions and obtain more comprehensive network environment data, an active exploration mechanism can be employed. After triggering the model update process, the use of the prediction model for prediction and decision-making (based on prediction results) is paused for a preset period (e.g., two days). During this pause, the terminal device reverts to conventional mobility-related strategies based on the traditional physical layer and RRC layer (such as reporting measurement events, handover and reselection, and network search strategies after RLF), and continues to collect communication data in real time. The exploration samples collected during this pause are used as new training data for incremental training of the prediction model.
[0271] If the poor performance of the model persists, a full retraining can be initiated.
[0272] The full retraining can be performed locally on the user terminal, or the user terminal can upload the new test data to the cloud / edge server. The uploaded data can be anonymized or federated learning frameworks can be used to complete the collaborative model update without exposing the user's original trajectory and identity information. The cloud / edge server uses historical data and new test data to jointly construct a training set for training and then sends the updated model parameters to the user terminal.
[0273] By further introducing a closed-loop feedback mechanism, the model error is dynamically analyzed by comparing the prediction results with the actual terminal data, and changes in communication quality are detected. When the prediction performance fails to meet the requirements or the communication quality deteriorates, incremental training or full retraining is automatically triggered. This not only keeps the prediction error stably within a preset range, but also endows the prediction model with the ability to continuously adapt to network changes and environmental disturbances, achieving self-evolution without manual intervention, effectively improving user experience and network efficiency.
[0274] In one possible implementation, determining the user terminal's mobility operation based on at least one future path node and / or future signal information includes: acquiring the current signal measurement value of the currently camped cell of the current path node; when the future signal information of the currently camped cell of the current path node indicates that a first communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a first preset signal quality threshold, performing co-frequency measurement to acquire the current signal measurement value of neighboring cells at the same frequency point; and / or, when the future signal information of the currently camped cell of the current path node indicates that a second communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a second preset signal quality threshold, reporting a measurement event to the network side to trigger inter-frequency or inter-system measurement configuration, or directly performing inter-frequency / inter-system measurement; and / or, based on the future path node and / or the future signal information of the future path node, setting measurement priorities for multiple test frequencies, and acquiring the current signal measurement value of neighboring cells of at least one of the multiple test frequencies. When the future signal information of the currently camped cell at the current path node indicates that a third communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a third preset signal quality threshold, a measurement event for triggering handover or cell reselection is reported to the network side. Specifically, the first preset signal quality threshold is greater than or equal to the second preset signal quality threshold, and the second preset signal quality threshold is greater than the third preset signal quality threshold. The reported measurement events for triggering inter-frequency or inter-system measurement configurations and for handover include: reporting directly based on the current signal measurement value of the currently camped cell and / or the current signal measurement value of neighboring cells, or reporting based on the corrected value after correcting the current signal measurement value of the currently camped cell and / or the current signal measurement value of neighboring cells. After a Radio Link Failure (RLF) event or an Out-of-Coverage (OOC) event occurs, the cell search priority is determined based on the future path node and / or the future signal information corresponding to the future path node.
[0275] The embodiments of this application can realize intelligent mobility management operations based on the predicted future path nodes and / or future signal information, including intelligent measurement reporting control, intelligent measurement / reselection triggering, intelligent measurement priority setting, intelligent reconstruction / network search control, etc.
[0276] Specifically, the first communication anomaly event refers to a weak signal event, that is, the signal of the current path node is poor and remains so for a certain period of time. The user terminal can start the neighbor cell measurement of the current serving cell frequency in advance so as to obtain the measurement results of the co-frequency neighbor cells in a timely manner.
[0277] The second communication anomaly event refers to a weak signal event, i.e., a further deterioration of the signal that persists for a certain period of time. In connected mode, an A2 event (serving cell quality below a threshold) can be reported to the network side in advance to trigger inter-frequency / inter-system measurements configured by the network side. In idle mode, inter-frequency / inter-system measurements can be started directly in advance to obtain measurement results promptly. Furthermore, regardless of whether inter-frequency / inter-system measurements are triggered in advance or strictly according to network configuration, the terminal can set measurement priorities for multiple frequency points to be measured based on predicted information of future path nodes, prioritizing the measurement of higher-priority frequencies to improve measurement efficiency.
[0278] The third communication anomaly event refers to the extremely weak signal event and the RLF / OOC event warning, which means that the signal deteriorates further and remains so for a period of time. In the connected state, the user terminal can report measurement events (such as A3, A5, B1 or B2, etc.) in advance to trigger handover and to trigger cell handover; in the idle state, it can perform cell reselection in advance.
[0279] In addition, if a Radio Link Failure (RLF) event or Out of Coverage (OOC) event is triggered in advance, or if an RLF event or OOC event is triggered strictly according to network configuration conditions, the user terminal will no longer search blindly, but will determine the cell search priority based on the future path nodes and / or the future signal information corresponding to the future path nodes, and will prioritize searching for cells on the RAT and frequency points corresponding to the future path nodes, thereby speeding up the connection recovery speed.
[0280] The first preset signal quality threshold is greater than or equal to the second preset signal quality threshold, and the second preset signal quality threshold is greater than the third preset signal quality threshold. For example, the first preset signal quality threshold can be set to RSRP = -80dBm, the second preset signal quality threshold can be set to RSRP = -90dBm, and the third preset signal quality threshold can be set to RSRP = -110dBm. Specifically, when the predicted RSRP in future signal information is lower than the first preset signal quality threshold (i.e., lower than -80dBm), the triggering condition for the first communication anomaly event is determined to be met, and intra-frequency measurement is triggered in advance. When the predicted RSRP is further lower than the second preset signal quality threshold (i.e., lower than -90dBm), the triggering condition for the second communication anomaly event is determined to be met, and A2 event reporting or inter-frequency / inter-system measurement is triggered in advance. When the predicted RSRP is further lower than the third preset signal quality threshold (i.e., lower than -110dBm), the triggering condition for the third communication anomaly event is determined to be met, and handover-related measurement event reporting or cell reselection is triggered in advance.
[0281] In addition, in practical applications, SINR or a combination of RSRP and SINR can be used to determine the above signal quality thresholds in order to more comprehensively assess future signal conditions.
[0282] In one possible implementation, to prompt the network side to issue measurement or handover commands in advance, or to perform measurement or cell reselection in advance, the user terminal can actively correct the signal measurement values of the currently camped cell and / or neighboring cells. Specific correction methods include, but are not limited to, the following three:
[0283] Firstly, weighted superposition involves superimposing the predicted future signal information (such as RSRP, RSRQ, or SINR prediction values) of the current stationed cell and / or future path nodes onto the current signal measurement values of the current stationed cell and / or neighboring cells according to preset weights.
[0284] Secondly, bias compensation, based on the predicted future signal information, adds a bias value to the current signal measurement value of the current cell and / or neighboring cells starting from the current moment.
[0285] Third, directly replace the trigger value by replacing the current signal measurement value of the currently camped cell and / or neighboring cells with a preset value that meets the triggering conditions of a specific network configuration measurement event. For example, replace it with a value that meets the triggering conditions of measurement events such as A2, A3, A5, and B1, thereby prompting the network side to issue measurement configuration or handover commands in advance; or replace it with a value that meets the triggering conditions of inter-frequency, inter-system measurement or cell reselection, thereby initiating the inter-frequency, inter-system measurement or cell reselection process in advance.
[0286] In one possible implementation, the user terminal can utilize predicted future signal information to trigger the measurement and handover process in advance, before link deterioration leads to communication interruption. Specifically, embodiments of this application propose an intelligent measurement reporting control mechanism.
[0287] In connected mode, the user terminal can fuse the predicted future signal information of the current cell and / or future path nodes with the real-time measured current signal information, and can adopt the following two reporting strategies:
[0288] Firstly, the measurement value reporting is corrected by using predicted future signal information to correct the current measurement value, thereby meeting the measurement event triggering conditions configured by the network side in advance (such as A2 / A3 / A5 / B1, etc.), prompting the network side to configure inter-frequency / inter-system measurements or issue switching commands in advance.
[0289] Secondly, direct advance reporting means that when the current measurement result has reached the event threshold but has not yet met the time to tirgger, if it is predicted that the signal of the currently camped cell will continue to deteriorate in the future, the user terminal can directly shorten the time to tirgger based on the current signal measurement value to report in advance (for example, shorten the time to tirgger to half of the original set value to report in advance, or shorten it to 0 to report immediately), without introducing measurement value correction.
[0290] Corrected measurement reporting refers to the process where, when the real-time signal measurement values (such as RSRP, RSRQ, or SINR) of the currently camped cell and / or neighboring cells have not yet reached the trigger threshold of a measurement event (such as A2, A3, A5, or B2 event) configured by the network side, the user terminal corrects the current measurement value using a preset correction strategy based on the current measurement value and predicted future signal information. When the corrected measurement value reaches the trigger threshold configured by the network side and meets the hysteresis time requirement, the user terminal reports the event to the network side. This allows for advance reporting of A2 events to promptly trigger inter-frequency and inter-system measurements in preparation for handover, or advance reporting of handover triggering measurement events before the signal quality of the currently camped cell deteriorates to the point of being unable to transmit or receive normally, thereby triggering the handover process and improving the handover success rate. Specific application scenarios include, but are not limited to:
[0291] First, if the measurement value of the currently camped cell has not yet met the A2 event threshold, but the prediction result shows that a weak cell event will occur in the currently camped cell, the user terminal can generate a corrected measurement value according to a preset correction strategy (for example, adding a negative bias to the measurement value from the current moment, or introducing a prediction confidence factor to weight the measurement value). When the corrected measurement value meets the A2 event threshold, it is reported to the network side to trigger inter-frequency or inter-system measurement.
[0292] Secondly, if the measurement values of the currently camped cell and / or neighboring cells have not yet reached the trigger threshold of the measurement events (such as A3, A5 or B2 events) configured on the network side, but it is predicted that the signal quality of the currently camped cell will deteriorate rapidly and significantly, and the future signal information of the target neighboring cell (future path node) is good (such as predicted RSRP and SINR are higher than the preset standard), the user terminal can generate the corrected measurement results according to the preset correction strategy and report them to the network side to trigger the handover process in advance.
[0293] Furthermore, in practical applications, the aforementioned corrected measurement reporting mechanism can also be combined with mechanisms to shorten latency to further improve the response speed of mobility management.
[0294] Intelligent measurement / reselection triggering refers to the ability of a user terminal, in idle state, to combine the predicted future signal information of the current cell and / or neighboring cells with the real-time measured current signal information to actively correct the measurement value and / or shorten the hysteresis time, thereby triggering inter-frequency measurement, inter-system measurement or cell reselection in advance.
[0295] Prediction-based measurement priority setting refers to the process where, after receiving measurement configurations (containing multiple frequency points to be measured) from the network side, the user terminal can combine predicted future path nodes and / or future signal information of those future path nodes to set measurement priorities for these frequency points. Specifically, for frequency points to be measured that match future path nodes, the user terminal can set a higher measurement priority and measure them first. For frequency points to be measured that do not match future path nodes, a lower measurement priority is set and the measurement is delayed. Through this mechanism, the user terminal can prioritize measuring high-priority frequency points, thereby discovering suitable cells more quickly and preparing for subsequent handover procedures.
[0296] Specifically, the above-mentioned measurement priority settings can be achieved through priority measurement mechanisms, delayed measurement mechanisms, dynamic iterative adjustments, and joint determination mechanisms.
[0297] The priority measurement mechanism refers to the user terminal's ability to determine the intersection between multiple network-configured frequency points to be measured and predicted future path nodes. For the frequency points to be measured within this intersection, the user terminal can prioritize their measurement. Furthermore, the user terminal can further assign measurement priorities to the frequency points to be measured within this intersection based on the predicted future signal quality of the future path nodes (for example, frequency points with better predicted signal quality have higher measurement priority, while frequency points with poor predicted signal quality can be assigned a lower priority than those outside the intersection).
[0298] The delayed measurement mechanism refers to the user terminal setting a lower measurement priority for frequency points that do not belong to the intersection of multiple frequency points to be measured in the network configuration, thereby delaying the measurement of them.
[0299] Dynamic iterative adjustment refers to the process where, during actual measurement, the user terminal can perform the first round of measurements according to the aforementioned priority order. When subsequent rounds of measurement are required, the user terminal can dynamically adjust the measurement priority of the frequency points to be measured based on the actual measurement results of the previous round, and continue iterative measurements according to the adjusted priority order.
[0300] The joint determination mechanism refers to the ability of a user terminal to jointly determine the measurement priority with the measurement priority configured on the network side when setting the measurement priority in the idle state, so as to comprehensively consider the scheduling intention of the network side and the prediction result of the terminal side.
[0301] Intelligent reconstruction, link establishment, cell selection, and network search refer to the process by which a user terminal can determine cell search priority based on the predicted future path node and / or the corresponding future signal information when a Radio Link Failure (RLF) event, an Out-of-Coverage (OOC) event, or an RLF / OOC event is triggered prematurely (e.g., if the predicted future path node is an LTE cell, but the currently camped NR cell does not have inter-system measurement configured and the real-time measured signal is already very poor, then the RLF event can be triggered prematurely before the network-configured RLF event triggering conditions are met). Specific priority determination mechanisms include constructing a priority search set, joint priority ranking, and avoiding poorly performing cells.
[0302] The construction of a priority search set refers to the user terminal prioritizing the Radio Access Technology (RAT) and frequency points corresponding to predicted future path nodes as its primary search targets, thus constructing a dedicated priority search set. When an RLF or OOC event occurs, the user terminal can first perform cell search on the targets (RAT, frequency points) in the priority search set, instead of directly performing the traditional historical frequency point search plus full-band blind search, thereby significantly shortening network search latency and accelerating the reconstruction, link establishment, or cell selection process.
[0303] Furthermore, the user terminal can further prioritize targets in the search set based on the future signal information corresponding to the predicted future path nodes (e.g., prioritizing future path nodes with better predicted signal quality corresponding to (RAT, frequency points)) to further shorten network search latency and restore connection as quickly as possible. In practical applications, the terminal can also combine historical frequently used path information (including historical summary information) to assist in correcting search priorities and actively avoid poor cells during the search process.
[0304] Joint priority ranking refers to the process by which a user terminal, when multiple cells are detected on a specific Radio Access Technology (RAT) and frequency point, can comprehensively score and rank these detected cells to determine the priority for attempting to camp or access the network. Specifically, the user terminal can calculate a weighted score for each detected cell by combining the real-time peak values of the physical layer synchronization signals (such as PSS and / or SSS) of each cell with the future signal information corresponding to predicted future path nodes. For example, if a detected cell belongs to a predicted future path node, its predicted future signal information is combined with the weighted score of the PSS / SSS peak values; if a detected cell does not belong to a predicted future path node, the score is based solely on the PSS / SSS peak values. This comprehensive scoring mechanism effectively mitigates the inaccuracy in PSS / SSS peak detection caused by noise, interference, and other factors, thereby more accurately determining the priority order for attempting to camp or access the network. In practical applications, the user terminal can further refine the score by combining historical summary information or real-time measurement results before RLF / OOC events occur.
[0305] Avoiding poorly performing cells refers to the process where, if prediction results or historical summary information indicate a high probability of a detected cell experiencing a communication anomaly (such as a high probability of congestion or persistently poor signal quality during the dwell period), the user terminal can place it in a low priority list for dwelling or access, or perform probabilistic avoidance. For example, the user terminal can multiply the dwell or access priority score of the candidate cell by an attenuation coefficient (such as 1-p) based on the predicted probability p of the anomaly event, causing it to naturally sink to the bottom of the list in the comprehensive scoring ranking. Alternatively, when the probability p exceeds a preset high confidence threshold (such as 0.8), the user terminal can probabilistically remove the cell directly from the priority list, thereby avoiding invalid attempts and further shortening the connection recovery latency.
[0306] This application's embodiments, by introducing a predictive model-based intelligent mobility management mechanism, achieve a technological leap from passive response to proactive prediction, significantly improving communication continuity and energy efficiency. Specifically, through prediction-based intelligent measurement triggering, measurement priority setting, intelligent measurement reporting control (such as correcting measurement value reporting), and intelligent reselection triggering, user terminals can fully utilize the prediction results of future nodes and future signal quality, promptly trigger measurements, and dynamically optimize the measurement sequence to quickly find suitable neighboring cells. Furthermore, before the serving cell signal deteriorates to the point of being unable to transmit or receive, handover or reselection procedures can be triggered in a timely manner, effectively improving the success rate of handover / reselection and significantly reducing the occurrence of radio link failures or coverage events. In addition, after a link failure or coverage event occurs, through intelligent reconstruction, link establishment, cell selection, and network search mechanisms (such as constructing a priority search set, joint priority ranking, and avoiding poor cells), user terminals can directly search for cells in high-priority (RAT, frequency point) combinations based on predictive information, locking in high-quality cells and proactively avoiding predicted poor cells. This mechanism significantly reduces the latency of network search for rebuilding, establishing links, and selecting cells, thereby providing users with a seamless and stable communication experience in various complex mobile scenarios.
[0307] In one possible implementation, before acquiring the historical and real-time communication signal information of the user terminal, the method further includes: preprocessing and feature alignment of the historical path node sequences in the communication signal records of the user terminal within a preset time window; calculating the path similarity index between each historical path node sequence; grouping historical path node sequences whose path similarity index falls within the preset path similarity threshold range into the same category based on a preset path similarity threshold, so as to extract commonly used mobile paths; wherein, the preset path similarity threshold includes a fixed threshold or an adaptive threshold, and the adaptive threshold is dynamically learned by calculation or machine learning model based on the distribution of currently recorded historical paths; and storing the commonly used mobile paths and their corresponding signal information as historical communication signal information.
[0308] In this embodiment, before acquiring historical and real-time communication signal information of the user terminal, the user terminal can continuously collect and store communication signal data during daily operation. Subsequently, within a preset time window (e.g., the most recent 7 days or 30 days), cluster analysis is performed on this communication signal data to extract its commonly used movement paths.
[0309] Specifically, the user terminal can continuously collect and store its communication signal data during its movement. This data includes, but is not limited to: path node information (such as TAC, RAT, frequency point, PCI, Cell ID, etc.), signal quality information (such as RSRP, RSRQ, SINR), time-related information (such as timestamps of user terminal camping, access, handover, reselection, and the occurrence of RLF / OOC events), and communication abnormal events (such as RLF events, OOC events, network registration rejection events, etc.).
[0310] The data collected can come from communication protocol layers such as the Physical Layer, Radio Link Control (RLC) Layer, RRC Layer, and NAS Layer. It does not rely on GNSS positioning information, so it has strong scene adaptability and can be widely used in various complex environments such as indoors, underground, and tunnels.
[0311] Subsequently, the user terminal can perform offline clustering processing on the path node sequences in the aforementioned communication signal data during idle periods (such as at night or while charging). Clustering algorithms can include, but are not limited to, sequence alignment, Dynamic Time Warping (DTW), Density-Based Spatial Clustering of Applications with Noise (DBSCAN), trajectory clustering based on graph neural networks, and various traditional or machine learning classification algorithms to group similar path node sequences into one class.
[0312] For example, for a sequence of path nodes composed of discrete cell identifiers, the terminal can use algorithms such as DTW, edit distance, or Hamming distance to calculate path similarity, and group historical path node sequences whose path similarity meets a preset threshold into the same category, thereby forming several candidate path clusters. This path similarity threshold can be a pre-set fixed value or an adaptive threshold (i.e., dynamically calculated based on the currently recorded historical path distribution or obtained through machine learning). Furthermore, the terminal can count the number of historical path sequences contained in each candidate path cluster. If a path cluster contains a number of historical path sequences greater than or equal to a preset threshold (e.g., 2), then that path cluster is determined to be a frequently used mobile path.
[0313] Each frequently used mobile path can be composed of ordered path nodes, where each path node can be represented as {RAT, frequency point, PCI} or {TAC, eNB ID, Cell ID}, etc. The communication signal data corresponding to each path node, along with the results obtained from the communication signal data through calculation, judgment, and derivation (such as the determination of poor cell events), can be stored as its corresponding signal information. The frequently used mobile paths and their corresponding signal information constitute the historical communication signal information stored in the user terminal's local database in this embodiment. Alternatively, for each path node in each frequently used path, historical signal quality statistics (such as the mean and variance of RSRP and SINR, dwell time, etc.) and the probability of communication anomalies (such as RLF probability, poor cell event probability, etc.) can be calculated for each path node in each frequently used path. The historical signal quality statistics and anomaly probability information corresponding to each path node in each frequently used mobile path together constitute the summary information of the frequently used mobile path and are stored together in the user terminal's local database. The commonly used movement paths and their summary information can also constitute the historical communication signal information in the embodiments of this application.
[0314] Specifically, a mechanism can be employed that retains only critical local path segments containing communication quality anomalies and their corresponding communication signal information. Specifically, path nodes in the collected communication signal data that experienced communication quality anomalies (e.g., RLF, OOC, or poor cell events) are identified; local path segments containing the path node, along with X (e.g., 8) path nodes preceding it and Y (e.g., 1) path nodes following it are extracted; these path segments are clustered to obtain frequently used mobile paths, and the frequently used mobile paths and their corresponding signal information or summary information are associated and saved, thus constituting the historical communication signal information in this embodiment. This method reduces data storage overhead, retains the critical path segments whose communication quality most needs improvement, and ensures that the retained data includes key features such as the direction of movement.
[0315] Through the above processing, a lightweight, personalized knowledge base of commonly used mobile routes can be constructed. This knowledge base forms the core of historical communication signal information and can be used for subsequent real-time route matching and determining corresponding optimization strategies after matching, or as a high-quality sample source for training predictive models. In addition, the knowledge base also has a dynamic update mechanism: for example, when a commonly used mobile route has not been used for a long time (e.g., it has not appeared for more than 100 days), or when new routes appear frequently, the system can automatically delete low-frequency old routes and add emerging high-frequency routes. Especially when local database storage space is limited, eviction strategies such as Least Recently Used (LRU) can be adopted to prioritize the cleaning up of low-value route records.
[0316] It should be noted that the preset threshold for the number of historical path sequences required to be included in a path cluster classified as a frequently used mobile path can be set to 1. This means that even if a path appears infrequently within a preset time window, it can still be included in the frequently used mobile path category as long as the number of its corresponding historical path sequences meets the preset threshold. This method is particularly suitable for scenarios where only critical local path segments containing communication quality anomalies are stored, facilitating a more comprehensive record of path segments experiencing communication quality anomalies. For example, with a preset time window of 7 days, a user terminal might only traverse a certain path once every 7 days, but this path contains nodes that frequently experience RLF events. By setting the preset threshold to 1, such low-frequency but high-risk paths within the preset time window can be prevented from being ignored. Furthermore, combined with the aforementioned dynamic update mechanism, this ensures that the frequently used mobile path knowledge base covers abnormal and risky paths while maintaining low data storage overhead for the user terminal.
[0317] The signal information or summary information corresponding to commonly used mobile routes is also updated regularly, for example, once a day, based on the communication signal data continuously collected by user terminals. This dynamic update mechanism ensures the timeliness and representativeness of the commonly used mobile route knowledge base.
[0318] This implementation optimizes mobility operations by pre-identifying users' regular movement behaviors, eliminating the need for long-term, real-time model prediction and significantly reducing the computational and power consumption overhead of user terminals. By focusing on frequently used movement paths, subsequent path matching, strategy optimization, and / or signal prediction can better align with users' actual movement patterns, especially in repetitive scenarios such as commuting and daily activities. Furthermore, by combining historical anomaly statistics at path nodes, when a user terminal re-enters the vicinity of a path node where anomalies have occurred, it can proactively trigger warnings and implement optimization strategies (such as avoiding poor cells, early measurement, and optimizing measurement sequence). It can also proactively trigger the model for real-time prediction to further optimize mobility management. Finally, this solution only retains frequently used paths and their summary information, rather than storing the original full logs, effectively alleviating the storage pressure on user terminals.
[0319] In one possible implementation, the method further includes: extracting historical signal information corresponding to each historical path node from historical communication signal information based on commonly used mobile paths, generating summary information corresponding to the historical path nodes, and storing the summary information in association with commonly used mobile paths.
[0320] Based on summary information and / or future signal information, report measurement events, trigger measurement procedures, select and execute cell handover or reselection, and determine the priority of measurement order and / or cell search order;
[0321] The summary information includes at least one of the following: signal quality statistics of each historical path node of each commonly used mobile path, the number and / or probability of RLF events, the number and / or probability of OOC events, the types of poor cell events and their corresponding number and / or probability, the number and / or probability of entering each node, and the average dwell time; the summary information of the same physical cell in the historical path nodes of different commonly used mobile paths is stored independently.
[0322] The signal quality statistics include: for each historical path node of each commonly used mobile path, calculating and recording the average of the previous preset number of times each type of measurement value was stationed at the node, and performing sliding window averaging or infinite impulse response (IIR) filtering on the average value corresponding to multiple stationings; and calculating and recording the number and / or probability of each measurement value being higher than each preset threshold each time stationed at the node, and performing sliding averaging or IIR filtering on the number and / or probability corresponding to multiple stationings.
[0323] The measured values include the Reference Received Power (RSRP) and the Signal-to-Interference-plus-Noise Ratio (SINR); poor cell events include: events where the number of times the cell's measured value is higher than a preset threshold is lower than a preset number threshold during a single dwell period; events where the probability of the cell's measured value being higher than a preset threshold is lower than a preset probability threshold; events where the dwell time of the cell is lower than a preset duration threshold; events where the cell's signal quality parameters are higher than a preset quality threshold and the resource allocation is lower than a preset allocation threshold; and events where network registration is rejected.
[0324] Specifically, in order to overcome the technical problem that traditional statistical methods based on cell identifiers (such as PLMN, TAC, RAT, frequency point, PCI, etc.) cannot distinguish the user's direction of movement, thus leading to inaccurate signal quality assessment, this embodiment proposes an independent summary information generation and storage mechanism based on path nodes.
[0325] First, after extracting frequently used mobile paths, the user terminal can further process each historical path node in each frequently used mobile path to generate corresponding summary information. The key improvement in this embodiment is that the summary information of historical path nodes belonging to different frequently used mobile paths within the same physical cell is stored independently. That is, the user terminal no longer merges historical data simply based on the physical cell identifier, but performs data merging processing according to the path context, thereby truly reflecting the communication status in a specific mobile direction.
[0326] Secondly, regarding the calculation and generation of summary information, the user terminal can extract historical signal information corresponding to each commonly used mobile path based on frequently used mobile paths, and use this historical signal information to calculate and generate summary information for each historical path node of the frequently used mobile path. The calculation of signal quality statistics can take, but is not limited to, the following methods: For each historical path node of each frequently used mobile path, calculate and record the average value of each type of measurement (such as RSRP, SNR) for each stay at that node over the previous preset number of times, and apply a sliding window average or infinite impulse response (IIR) filter to the average value corresponding to multiple stays; simultaneously, calculate and record the number and / or probability of each measurement value exceeding each preset threshold for each stay at that node, and apply a sliding average / IIR filter to the number and / or probability corresponding to multiple stays. Furthermore, the summary information may also include the types and corresponding number and / or probability of RLF events, OOC events, and poor cell events, the number and / or probability of entering each node, and the average stay duration, etc.
[0327] Finally, regarding the determination of poor cell events, poor cell events may include at least one of the following: during a single stay, the number of times the cell's measurement value is higher than a preset threshold is less than a preset number threshold; the probability of the cell's measurement value being higher than a preset threshold is less than a preset probability threshold; the stay duration of the cell is less than a preset duration threshold; the cell's signal quality parameters are higher than a preset quality threshold, and the resource allocation is less than a preset allocation threshold; and network registration is rejected.
[0328] For example, for a physical cell C, when it serves as a historical path node on an "east-to-west commuting route," the user terminal can create a separate summary information record for it; while when physical cell C serves as a historical path node on a "west-to-east commuting route," the user terminal can create another separate summary information record. These two records do not affect each other and can be stored separately in the user terminal's local database.
[0329] For each historical path node (i.e., the aforementioned independently stored unit) of each frequently used mobile path, the user terminal can extract the communication signal information of the node when it has passed through (i.e., stayed) the path the most recently (e.g., the last 7 times), and calculate and generate summary information.
[0330] Furthermore, the generated summary information may include at least one of the following: signal quality statistics of each historical path node of each commonly used mobile path, the number and / or probability of RLF events, the number and / or probability of OOC events, the types of poor cell events and their corresponding number and / or probability, the number and / or probability of entering each node, and the average dwell time, etc.
[0331] Specifically, the process of generating signal quality statistics includes: for each historical path node of each commonly used mobile path, extracting the signal quality information (such as RSRP and SINR) of the node when it has passed through (i.e. stayed) the path the most recently by a preset number of times, and calculating the corresponding statistical indicators.
[0332] Taking cell C in the "commuting route from east to west" as an example, the user terminal can only retrieve the RSRP and SINR data from historical visits to cell C along this specific route for statistical analysis. Specifically, the user terminal can calculate and record the average RSRP and SINR values for each preset number of visits to cell C. Subsequently, to smooth fluctuations across multiple visits and obtain a more stable long-term trend, the user terminal can apply infinite impulse response (IIR) filtering or a sliding window averaging to the average RSRP and SINR values calculated from multiple visits. Simultaneously, the user terminal can also count and record the number of times and / or the probability that each measurement value exceeds a preset good threshold during each visit, and similarly apply a sliding average or IIR filtering to these counts and / or probabilities.
[0333] Furthermore, user terminals can also identify and determine poor cell events based on historical data. The specific determination logic includes: for each historical path node on each frequently used mobile path, extracting historical signal information from the node's most recent preset number of visits (i.e., stays) on that path; if the measured value during a stay at that node exceeds a preset threshold in terms of frequency or probability, but falls below the corresponding preset frequency or probability threshold, then that node is determined to have experienced a poor signal quality cell event. Alternatively, if the stay duration at that node is less than a preset duration threshold, then that node is determined to have experienced a poor stay duration cell event. If the signal quality parameters during a stay at that node are higher than a preset quality threshold, but the resource allocation is lower than a preset allocation threshold, then that node is determined to have experienced a poor congestion cell event. If network registration is rejected during a stay at that node, then that node is determined to have experienced a poor registration rejection cell event.
[0334] After completing the above determination, the user terminal can record the types of poor cell events identified and their corresponding frequency and / or probability, along with signal quality statistics, the frequency and / or probability of RLF events, the frequency and / or probability of OOC events, the frequency and / or probability of entering each node, and the average dwell time, as a summary information of the node under that path.
[0335] To support a sliding window-based dynamic update mechanism, the number and / or probability of poor cell events and the number and / or probability of RLF events can be calculated based on the event occurrence sequence within a preset time window. Specifically, the user terminal can record the historical occurrence status when passing through the node a preset number of times (e.g., 5 times) and store this historical occurrence status in the form of an event occurrence sequence (e.g., represented by [1, 0, 1, 1, 1], where 1 represents an event and 0 represents no event). When the user terminal passes through the node again, it can update the event occurrence sequence by appending new status values and removing the historical status value at the beginning of the sequence, and calculate the latest event occurrence count and / or probability accordingly.
[0336] Taking cell C in the "commuting route from west to east" as an example, the user terminal can call up independent historical data under this specific route for calculation. Through this independent statistical mechanism based on the path context, the user terminal can accurately capture directional characteristics such as "good signal and stable communication when passing through from east to west, while signal drops suddenly and communication is easily interrupted when passing through from west to east", thus providing accurate predictive basis for subsequent mobility management.
[0337] For example, cell C is an NR cell located outside a tunnel exit in a user's east-to-west commuting path. When the user terminal leaves the tunnel and camps in cell C, the signal is good, and after continuing to move, the user terminal can successfully hand over to the next node, maintaining stable communication. However, when the user terminal passes through cell C and enters the tunnel from west to east, the tunnel is covered by LTE, and the signal in cell C often drops abruptly. At this time, the user terminal may fail to handover because the signal in cell C has deteriorated to the point of being unable to transmit or receive normally by the time the measurement event is reported or the handover command is received. After maintaining this extremely poor signal condition for a period of time, the user terminal will trigger an RLF (Remote Frequency Response) event. Traditionally, during the reconstruction process after an RLF, the user terminal usually first attempts to search for historical frequency points and perform a full-band frequency scan under the NR standard. Since there is no NR coverage inside the tunnel, the user terminal will only switch to searching for historical frequency points and performing a full-band frequency scan after the NR frequency scan fails. Communication is only restored after a successful search for an LTE cell in the tunnel and the establishment of a connection. This process results in a long communication interruption time, severely impacting the user experience. Even when passing through the same community (C), users experience drastically different communication experiences due to their different directions of movement.
[0338] Finally, the user terminal can retrieve summary information of the current path node and the possible next path node based on the currently matched common mobility paths. When performing mobility operations (such as measuring neighboring cells, cell reselection, handover, RLF post-reconstruction, or OOC post-cell selection), the user terminal can optimize its strategies based on the aforementioned summary information. Specifically, optimization operations may include: reporting measurement events in advance, triggering measurement procedures, selecting and executing cell handover or reselection, and determining the priority of measurement order and / or cell search order, thereby effectively avoiding communication interruptions or shortening communication interruption time, enhancing mobility and improving user experience.
[0339] Specifically, the optimization strategies based on path matching and summary information mentioned above can include the following scenarios:
[0340] Firstly, measurement events can be reported in advance. For example, when the summary information indicates that the current node has a high probability of experiencing an RLF event historically, the following optimization measures can be taken for user terminals in the connected state: On the one hand, the user terminal can perform co-frequency measurements in advance; on the other hand, the user terminal can add a positive bias to the trigger threshold of some A2 events configured by the network (for example, raising the trigger threshold for A2 events within a certain threshold range). When the measured value meets the corrected threshold, the measured value is subtracted from the positive bias and reported to the network side. Since A2 events with excessively low thresholds are very likely to trigger blind redirection, by adding a positive bias to the trigger threshold of some A2 events, it is possible to report A2 events that are likely to trigger inter-frequency or inter-system measurements configured by the network side in advance, while effectively avoiding blind redirection. This gives the user terminal more measurement time and significantly increases the probability of timely measurement of suitable neighboring cells. Furthermore, for nodes with a high probability of experiencing RLF events, the threshold for A2 events that trigger inter-frequency and inter-system measurements configured by the network side can be extracted from the node's historical information and added to the summary information. This allows the user terminal to more accurately apply a positive bias only to the thresholds of A2 events that trigger inter-frequency and inter-system measurements. In addition, the user terminal can shorten the latency of network-configured handover-triggered measurement events (such as A3 and B1 events), increasing the probability of successful handover by triggering the handover process earlier, thereby reducing the probability of RLF events.
[0341] Secondly, the measurement process and cell reselection can be triggered. For example, when the summary information indicates that the current node has a high probability of experiencing an OOC event historically, a positive bias can be added to the network-configured measurement trigger threshold for user terminals in an idle state. This allows for earlier triggering of same-frequency and inter-frequency / inter-system measurements. Simultaneously, the user terminal can shorten the network-configured cell reselection trigger hysteresis time. Through these dual adjustments to thresholds and time, user terminals can initiate measurements and cell reselection earlier, effectively reducing the probability of OOC events.
[0342] Third, the priority setting of the measurement order: after receiving the measurement configuration (including the target measurement set A) sent by the network side, the user terminal can allocate the measurement priority in a hierarchical manner by combining path matching and summary information.
[0343] First, the user terminal can compare the target measurement set A with the sets of all possible next path nodes (set B). For the intersection of set A and set B (i.e., the frequency points where the path matches), the user terminal can prioritize measurement. For the non-intersection parts, the user terminal can delay measurement. Through this basic path matching strategy, the user terminal can quickly measure the frequency points where there is a high probability of a suitable cell, thus preparing for subsequent handover or reselection.
[0344] Secondly, after completing the basic comparison described above, the user terminal can further refine the priority allocation based on the summary information of each possible next-path node. On the one hand, for next-path nodes with better signal quality statistics, the user terminal can assign them a higher measurement priority. On the other hand, if the summary information indicates that a possible next-path node has a high probability of experiencing a poor cell event, and in the intersection set, there are no other nodes with a low probability of experiencing a poor cell event corresponding to that node (RAT, frequency point), the user terminal can also lower the measurement priority of that frequency point, or even place it after non-intersection frequency points for measurement.
[0345] In actual measurement, the user terminal can perform the first round of measurements according to the above priority order. When subsequent rounds of measurements are required, the user terminal can dynamically adjust the measurement priority of the frequency points to be measured based on the actual measurement results of the previous round, and continue iterative measurements according to the adjusted priority order.
[0346] Fourth, when selecting a target cell for handover or reselection, if multiple neighboring cells that meet the handover or reselection thresholds are measured, the user terminal can combine the measurement value and the next possible node for path matching and its summary information to prioritize the cell with good measurement value and low probability of occurrence of bad cell events as the target cell, thereby effectively avoiding historical bad cells.
[0347] Fifth, regarding the priority setting of cell search order, when an RLF or OOC event occurs, or when an RLF or OOC event is triggered prematurely, the user terminal can allocate search priorities hierarchically based on path matching and summary information. Premature triggering of an RLF or OOC event means that, before the network-configured RLF or OOC event triggering conditions are met, the user terminal predicts an impending communication quality anomaly in the current cell based on path matching and summary information, thus triggering the RLF or OOC event in advance to escape the current cell. For example, a user terminal is traveling along a west-to-east commuting route and is stationed in NR cell C before the tunnel entrance. Cell C does not have inter-system measurement configured, and the tunnel has LTE coverage. When the user terminal enters the tunnel, the signal in cell C deteriorates rapidly, but because cell C does not have LTE measurement configured, it cannot switch to an LTE cell within the tunnel. If the RLF event triggering conditions are strictly followed according to the network configuration, it will usually result in a long connection interruption delay, affecting the user experience. Therefore, the user terminal can trigger the RLF event in advance based on path matching and summary information to escape cell C.
[0348] First, the user terminal can prioritize searching for cells among one or more (RAT, frequency points) corresponding to each possible next path node set, attempting cell reconstruction or cell selection. For example, when the user terminal matches a "west-to-east commuting path," if an RLF event occurs, the user terminal can directly skip the traditional NR historical frequency point search and full-band frequency scanning process, prioritizing cell search on the LTE frequency point corresponding to the next possible node, thereby significantly reducing network search latency and improving user experience.
[0349] Secondly, the user terminal can also assign more granular search priorities to one or more (RATs, frequency points) based on the corresponding summary information. Specifically, the user terminal can assign higher priority to (RATs, frequency points) with better signal quality statistics and a lower probability of poor cell events. Furthermore, when multiple candidate cells are detected under a certain (RAT, frequency point), the user terminal can combine the detected PSS / SSS peak values with the cell's summary information to perform joint priority ranking and probabilistic avoidance of poor cells. For example, it can prioritize cells with higher PSS / SSS peak values, summary information indicating good signal quality statistics, and a lower probability of poor cell events, and prioritize attempting to camp or access (rebuild) these cells, thereby effectively avoiding poor cells.
[0350] Finally, the user terminal can also trigger model predictions in a timely manner based on the summary information of the matched path. For example, when the summary information shows that the current node has frequently experienced RLF events historically, the user terminal can proactively run a prediction model on that node for real-time prediction. In subsequent execution, the user terminal's mobility operations can be comprehensively optimized by combining the aforementioned summary information and predicted future signal information, thereby achieving better mobility management results.
[0351] In addition, the summary information can be updated regularly (e.g., once a day) as the user terminal continuously collects communication signal data, and the timeliness and accuracy can be ensured through a dynamic update mechanism.
[0352] This application's embodiments resolve the technical problem of statistical distortion or invalidity caused by mixing data from different movement directions in traditional statistical methods by refining the statistical granularity of summary information from physical cells to historical path nodes. Specifically, by independently storing summary information of the same physical cell under different commonly used movement paths (e.g., distinguishing statistical characteristics of east-to-west and west-to-east travel), user terminals can capture the impact of movement direction on signal changes and user experience. When reporting measurement events, performing measurements, reselecting cells, and determining measurement or network search priorities, user terminals only call specific summary information of each node on the currently matched path, thereby more accurately predicting future node and signal information, effectively avoiding misjudgments caused by directional differences, and significantly improving the accuracy of mobility management and communication experience.
[0353] In one possible implementation, the method further includes: when performing cell search, cell selection, radio link reconstruction, or radio link establishment, if one or more cells are detected on the same Radio Access Technology (RAT) and frequency point, then selecting candidate cells from the one or more cells: candidate cells corresponding to consecutive nodes after a path segment matching a common mobile path, and / or candidate cells corresponding to predicted future path nodes. Based on the Physical Cell Identifier (PCI) and summary information and / or future signal information of each candidate cell, assigning an attempt to camp or an access priority to the detected multiple cells. If the summary information and / or future signal information corresponding to any candidate cell indicates that the probability of a poor cell event occurring in that candidate cell is greater than a first preset threshold, then determining a second preset threshold that is positively correlated with the probability of a poor cell event occurring. Generating a uniformly distributed random number in the interval 0 to 1, and prohibiting camping or access to the candidate cell when the random number is less than the second preset threshold; wherein the second preset threshold is positively correlated with the probability of a poor cell event occurring, such that the probability of the corresponding candidate cell being prohibited from camping or accessing is positively correlated with the probability of a poor cell event occurring. If the currently camped or accessed cell detects a poor congestion cell event, or if its corresponding future signal information indicates that the probability of a poor congestion cell event occurring in the current cell is greater than a third preset threshold, then the signal measurement values of the currently camped or accessed cell and / or the detected neighboring cells are corrected. Based on the corrected signal measurement values, cell reselection or cell handover is triggered to actively escape the current cell.
[0354] Specifically, when a user terminal experiences an RLF or OOC event at a certain path node and needs to perform radio link reconstruction, radio link establishment, or cell selection, this embodiment proposes a comprehensive processing strategy that combines real-time signals, historical summary information, and / or predicted future information for scenarios where one or more cells are detected on the same radio access technology (RAT) and frequency point.
[0355] First, the user terminal can filter candidate cells from one or more detected cells, selecting those corresponding to consecutive nodes after path segments matching common mobile paths, and / or candidate cells corresponding to predicted future path nodes. Then, by combining the Physical Cell Identifier (PCI), summary information, future signal information, and real-time detected PSS / SSS related peak values and pre-RLF / OOC measurement information of each candidate cell, the user terminal assigns a priority for attempting camping or access to the detected cells. For example, if a detected cell's PCI matches a predicted future path node, and its PSS / SSS peak value is high while the summary information indicates good signal quality, the user terminal will assign it a higher priority. For cells not belonging to the predicted nodes, the user terminal can evaluate them solely based on the PSS / SSS related peak value and pre-RLF / OOC measurement values. This mechanism effectively improves the long latency problem caused by inaccurate PSS / SSS detection in environments with noise, interference, or complex multipath, which leads to ineffective attempts at MIB, SIB readings, and random access in weak cells, resulting in significant time wasted on these attempts.
[0356] Secondly, to avoid user terminals accessing cells with poor historical performance, this embodiment introduces a probabilistic avoidance strategy. If the summary information and / or future signal information corresponding to any candidate cell indicate that the probability of a poor cell event (such as signal quality, dwell time, congestion, or network registration rejection) is greater than a first preset threshold, the user terminal will determine a second preset threshold that is positively correlated with this probability. The user terminal generates a uniformly distributed random number in the range of 0 to 1. If this random number is less than the second preset threshold, then initiating dwell or access to the candidate cell is prohibited. Since the second preset threshold is positively correlated with the probability of poor cell events, the probability of a candidate cell being prohibited from dwelling or accessing is also positively correlated with its historical risk. This effectively improves the situation where user terminals spend a lot of time on poor cells performing invalid attempts to read MIBs and SIBs and perform random access, and where effective communication is impossible even after accessing a cell with good signal (i.e., after successful reconstruction or link establishment) due to network congestion or network registration rejection.
[0357] Then, to address the communication failure caused by network congestion, the user terminal will also implement an active escape strategy. If the currently camped or accessed cell detects a poorly congested cell event, or if its corresponding future signal information indicates that the probability of a poorly congested cell event is greater than a third preset threshold, it means that although the current cell has a good signal, its actual available resources are scarce. At this time, the user terminal can correct the signal measurement values of the current cell and / or detected neighboring cells. For example, a negative offset can be applied to the measurement value of the current cell to reduce its value, or a positive offset can be applied to the measurement value of high-quality neighboring cells to increase its value. Based on the corrected signal measurement values, the user terminal can more quickly meet the triggering conditions for cell reselection or handover, thereby actively escaping the current congested cell.
[0358] Finally, in the aforementioned priority allocation and avoidance decisions, the poor cell events and their probabilities are based on statistics from the user terminal at the data statistics level, which analyze the frequency and probability of various poor cell events at each historical path node of each commonly used mobile path. Specifically, for congestion-related poor cell events, the user terminal can also statistically analyze their frequency and probability by time period to accurately reflect the significant differences in congestion characteristics of the same cell at different times, providing reliable data support for the aforementioned avoidance mechanism.
[0359] Specifically, after selecting candidate cells, a camping or access priority is assigned to each detected cell. Specifically, the user terminal can combine real-time detected PSS / SSS related peak values, pre-RLF / OOC measurement information, and the Physical Cell Identifier (PCI) of each candidate cell and its corresponding summary information and / or future signal information to calculate a comprehensive camping / access priority score for each detected cell. This score is positively correlated with the signal quality in the detected PSS / SSS related peak values, pre-RLF / OOC measurement values, summary information, or future signal information, for example, S=f(g1(C PSS C SSS g2(SNR) meas RSRP meas ), g3(SNR history RSRP history ), g4(SNR prediction RSRP prediction )), where C PSS and C SSS SNR represents the normalized correlation peak of PSS and SSS. meas and RSRP meas Represents the measured values of SNR and RSRP before RLF / OOC, SNR history and RSRP history The statistical values of SNR and RSRP in the summary information are represented by SNR.prediction and RSRP prediction The values of f() and g represent the predicted SNR and RSRP values in future signal information. n (), n=1 to 4, positively correlated (e.g., a weighted average function), the value of g1() and C PSS C SSS Positive correlation (e.g., a weighted average function), g n (), where n=2 to 4, the values of are positively correlated with their corresponding SNR and RSRP (for example, they can be defined as (SNR+3+RSRP+110) / 153). In particular, if there are cells among the detected cells that do not belong to the aforementioned candidate cells (i.e., not on the matching or predicted path), then for these cells, the score is calculated only based on their PSS / SSS related peak values and the measurements before RLF / OOC.
[0360] Secondly, to avoid user terminals from residing in or accessing cells with poor historical performance, this embodiment introduces a probabilistic avoidance strategy. Before triggering this strategy, it is necessary to clarify the specific definition of the poor cell event in this embodiment.
[0361] Specifically, in this embodiment, the poor cell events defined by the user terminal include four categories: poor cell events related to signal quality, poor cell events related to dwell time, poor cell events related to congestion, and poor cell events related to network registration rejection.
[0362] Among them, poor signal quality cell events refer to events in which the number of times and / or the probability of each measurement value being higher than each preset threshold during the cell stay is lower than each preset number of times and / or probability threshold (e.g., the probability of RSRP>-110dBm and SINR>0dB is less than 5%).
[0363] Poor cell dwell time events refer to situations where a user terminal's dwell time in a cell is less than a preset threshold (e.g., less than 2 seconds), indicating that although it can briefly access the network, it cannot maintain a stable connection.
[0364] Congestion-related poor cell events refer to events where the signal is good, but the cell's resource allocation is below a preset threshold due to network congestion (e.g., low probability of responding to scheduling requests, large amount of data waiting to be sent).
[0365] Network registration rejection is a type of poor cell event that refers to an event in which a user terminal, after successfully accessing the network, attempts to register but is rejected.
[0366] The frequency and probability of various poor cell events are statistically analyzed at each node along historical / frequently used travel paths. In particular, for congestion-related poor cell events, the frequency and probability can be statistically analyzed by time period to accurately reflect the significant differences in congestion characteristics of the same cell at different times.
[0367] After defining a poor cell event, if the future signaling information or summary information of any candidate cell indicates the comprehensive probability (or confidence level, denoted as P) of a poor cell event occurring, then... bad If the value is greater than the first preset threshold (e.g., T1=0.7), the user terminal will trigger a probabilistic avoidance mechanism.
[0368] Regarding the overall probability (P) bad To determine the probability of occurrence of various poor cell events, the user terminal can calculate it by comprehensively considering the occurrence probabilities of the aforementioned events. Specifically, the user terminal can use a comprehensive function P. bad =f(P quality P duration P rejection P busy (time) is used for calculation. Where P quality P duration P rejection P represents the probability of occurrence of poor cell events in the categories of signal quality, dwell time, and network registration rejection, respectively; busy (time) represents the probability of a congestion-related poor cell event occurring within the time period corresponding to the current time. The value of this comprehensive function f is positively correlated with the probability of occurrence of each type of poor cell event. In practical applications, this function f can be a maximum value function (i.e., f() = max()) or a weighted average function, thus comprehensively and accurately reflecting the overall risk level of candidate cells.
[0369] After determining the trigger probability avoidance, the user terminal can determine the overall probability (P0) of a poor cell event occurring in the candidate cell. bad ), determine a second preset threshold (T) that is positively correlated with it. avoid Specifically, the user terminal can directly set T. avoid equals P bad Or set to P bad linear functions (e.g., T) avoid =0.75×P bad Alternatively, other nonlinear functions can be used for mapping.
[0370] Subsequently, the user terminal can activate the random number generator to generate a random number R that follows a uniform distribution in the interval [0, 1], and then use this random number R and a second preset threshold T to... avoid The comparison results will be used to implement the corresponding avoidance strategy.
[0371] If R <T avoidIf the random number is less than the second preset threshold, it is determined that an avoidance mechanism has been triggered. In this case, the user terminal will be prohibited from initiating a stay or access request for the candidate cell. In practice, the user terminal can set the priority of the physical cell identifier (PCI) of the candidate cell to the lowest in the attempt stay list, or directly delete it from the list, thereby avoiding subsequent stay, RRC reconstruction, or connection establishment attempts.
[0372] If R≥T avoid If the candidate cell is selected, the attempt is allowed. In this case, the user terminal will retain the normal camping or access priority for that candidate cell.
[0373] Furthermore, to achieve more refined avoidance control, the user terminal can set independent first and second preset thresholds for each of the aforementioned poor cell events (such as signal quality issues, congestion, and registration rejection), and generate random numbers to independently determine whether avoidance is triggered. Under this multi-dimensional independent avoidance mechanism, as long as the determination result of any type of poor cell event triggers avoidance, the user terminal will perform an operation to prohibit camping or access to the candidate cell.
[0374] Through the above probabilistic avoidance mechanism, the probability of a candidate cell being prohibited from camping or accessing is reduced to the combined probability of a poor cell event (P0). bad There is a positive correlation between the probability of a candidate cell experiencing a poor cell event and the probability of it being prohibited from camping or accessing the cell. This effectively improves the terminal's ineffective attempts on high-risk cells. At the same time, it also retains a certain probability of attempting the attempt, and with the update mechanism of historical communication signal information and summary information, it can be updated and adapted in a timely manner when network conditions change.
[0375] The following is an illustration using a specific application scenario: In a high-speed mobile scenario (such as a subway tunnel entrance), a user terminal experiences a Radio Link Failure (RLF) event at path node C (NR cell). During the subsequent cell search for radio link reconstruction, the user terminal detects three co-frequency cells at the same frequency point F (e.g., frequency point 38400 in the LTE B39 band), with Physical Cell Identifiers (PCIs) of 101, 99, and 300, respectively. The PSS / SSS related peak values of these three detected cells are similar, and prior to the RLF event, the NR cell network side of node C did not have cross-system (i.e., LTE) measurements configured, resulting in the user terminal lacking pre-RLF measurement values for these three cells. In this case, the user terminal can combine the signal quality statistics in the aforementioned summary information with the real-time detected PSS / SSS related peak values to calculate the comprehensive priority score for the three cells, and determine whether to trigger probabilistic avoidance based on the probability statistics of poor cell events in the summary information. This can improve the long delay problem caused by judging the camping / access priority solely based on the PSS / SSS peak values, as well as the situation where effective communication is not possible even after accessing a cell with good signal (i.e., after successful reconstruction or link establishment) due to network congestion or network registration rejection.
[0376] The query results for the historical summary information of each next possible path node of node C in the matched common movement path are as follows:
[0377] For a cell with PCI=101, its signal quality statistics are relatively good, with only congestion-type poor cell events occurring. Furthermore, the probability of this event occurring in the current time period is 0.25. Therefore, the overall probability P of a poor cell event is calculated. bad =0.25.
[0378] For a cell with PCI=99, its signal quality statistics are good, and it has only experienced congestion-related poor cell events. However, the probability of this event occurring in the current time period is 0.8. Therefore, P is calculated. bad =0.8.
[0379] For a cell with PCI=300, its signal quality statistics are poor. The probability of a poor signal quality cell event is 1, and the probability of a poor dwell time cell event is 0.75. Based on this, P is calculated as follows: bad =1.
[0380] After comprehensively considering the peak values and signal quality statistics of PSS / SSS, the calculated priority scores for staying or accessing the station, from highest to lowest, are: PCI=99, PCI=101, PCI=300.
[0381] In the probabilistic avoidance determination stage, assuming the first preset threshold T1 = 0.7, and setting the second preset threshold T... avoid =0.75×P bad For a cell with PCI=99, the user terminal generates a random number R1=0.58. Since 0.58 < 0.75 × 0.8 (i.e., R1...),... <T avoid User terminals will be prohibited from attempting to camp on or access cells with PCI=99, thereby avoiding congestion risks. For cells with PCI=300, the user terminal generates a random number R2=0.95. Since 0.95>0.75×1 (i.e., R2≥T), avoid The user terminal will be allowed to attempt to camp on a cell with PCI=300.
[0382] Ultimately, the user terminal can attempt to camp or access the device according to the revised priority order (i.e., first try cells with PCI=101, then try cells with PCI=300), or select only the K cells with the highest priority (e.g., when K=1, only select cells with PCI=101). If there are no cells to try after the revision, or if the attempt to camp or access the device fails, the user terminal will move on to the next cell (RAT, frequency point) for cell search.
[0383] When a user terminal detects one or more cells on the same Radio Access Technology (RAT) and frequency, during the screening phase, the user terminal can extract consecutive nodes after the end of the segment that match the historical path segment (i.e., each possible next path node) as the next key monitoring targets. Alternatively, the user terminal can also select each possible next path node predicted by the predictive model as the key monitoring targets.
[0384] During the cell priority allocation phase, priorities can be allocated according to the following rules: high priority is given to candidate cells that are potential next path nodes in the matching or prediction and whose summary or prediction information shows good signal quality and a low probability of poor cell events; medium priority is given to cells that are not potential path nodes in the matching or prediction but have good signal quality; and the lowest priority is given to candidate cells whose summary or prediction information shows poor signal quality and / or a high probability of poor cell events, or they can be probabilistically avoided through the aforementioned random number mechanism.
[0385] Furthermore, this embodiment also provides a mechanism for proactively escaping congested cells. When a user terminal detects or predicts that the probability of a poorly congested cell event occurring in the currently camped or accessed cell is greater than a third preset threshold (for example, the third preset threshold can be configured to 0.85 to characterize a higher congestion risk confidence level), the user terminal will correct the signal measurement values of the currently camped or accessed cell and the detected neighboring cells. Specifically, the user terminal can reduce the RSRP or SINR measurement value of the currently camped or accessed cell and / or increase the RSRP or SINR measurement value of high-quality neighboring cells to meet the handover threshold. Subsequently, the user terminal can proactively trigger cell reselection or cell handover based on the corrected signal measurement values, thereby escaping the current congested cell and ensuring the user's communication experience.
[0386] This embodiment prioritizes one or more cells detected at the same frequency point by comprehensively utilizing historical summary information and / or predicted future signal information, as well as real-time PSS / SSS related peak values and measurements before RLF or OOC events. Simultaneously, it introduces a multi-dimensional definition of poor cells, including congestion, network registration rejection, dwell time, and signal quality, and employs a random number probability avoidance mechanism. This effectively improves the long latency issues in reconstruction, link establishment, or cell selection caused by inaccurate PSS / SSS detection in multi-cell environments on the same frequency, and addresses situations where effective communication is impossible even after accessing a cell with good signal (i.e., after successful reconstruction or link establishment) due to network congestion or network registration rejection. By combining real-time and historical communication signal information from the user terminal, this embodiment significantly reduces the latency of reconstruction, link establishment, or cell selection, thereby improving the user's communication experience.
[0387] In one possible implementation, before acquiring historical and real-time communication signal information of the user terminal, the method further includes: setting a circular buffer of length N, where N is a positive integer, and the circular buffer is used to temporarily store online acquired communication signal samples. The acquired communication signal samples are sequentially written into the circular buffer in chronological order, and when the amount of data written exceeds N, the earliest written data is overwritten with new data. When a preset event is detected, a random number generator is started to generate an integer offset M, where M is greater than or equal to n and less than or equal to N-1-m, and n and m are preset non-negative integer buffer values. After the preset event is detected, M more communication signal samples are acquired and written to the circular buffer, and N consecutive historical samples are extracted from the circular buffer from the current writing position, and the extracted N historical samples are saved to a persistent storage medium. The preset event includes at least one of an RLF event, an OOC event, or a measurement report trigger event.
[0388] This application provides an event-triggered data acquisition method based on a circular buffer, which is used to efficiently capture context data before and after key communication events before acquiring historical communication signal information of a user terminal.
[0389] Figure 3 A schematic diagram of a circular buffer structure provided in an embodiment of this application is shown below. Figure 3 As shown, the solid circle represents a circular buffer of capacity N, and the dashed arrows indicate the direction of data writing and reading (e.g., clockwise). It should be noted that the structural diagram is used to more clearly illustrate the operation of the circular buffer, and the circular shape is used to represent the circular buffer without losing its fundamental principle.
[0390] Specifically, the user terminal can dynamically maintain a circular buffer of length N during operation, where N is a positive integer (e.g., N=256, 4096, etc., which can be flexibly configured according to requirements and the user terminal's memory resources). This circular buffer is used to temporarily store communication signal samples collected online at certain time steps (e.g., 200 milliseconds). Each sample corresponds to a time step (t, t+1, t+2, t+3, t+4, ..., t+M, ..., t+N-1) and contains a structure used to store the communication signal data collected within the time window corresponding to that time step. For example, this structure may include: path node information (such as RAT, frequency point, PCI, Cell ID, TAC); signal quality information (such as RSRP, SINR, RSRQ, downlink duplicate packet drop count), specifically including information for the serving cell and neighboring cells; event flags (such as handover, RLF event, OOC event, A2 event reporting, etc.); time-related information (such as the end time of the time window corresponding to the time step, the time of event occurrence, the time of entering and leaving the serving cell, etc.); resource allocation information (such as scheduling request response probability, the amount of data waiting to be sent, etc.); and network configuration information (such as measurement configuration information, etc.).
[0391] Communication signal samples can be written to the circular buffer in chronological order. New samples are written to the current write pointer position. When the amount of data written exceeds the buffer capacity N, the new data can automatically overwrite the earliest written data (i.e., adopting a first-in-first-out overwrite strategy), thereby ensuring that the buffer always stores the most recent N consecutive samples and avoiding infinite growth that could lead to memory overflow.
[0392] When the user terminal detects a preset event (such as an RLF event, an OOC event, or a measurement report trigger event), the following data saving process can be initiated immediately: First, a random number generator can be started to generate an integer offset M, which satisfies: n ≤ M ≤ N-1-m. Here, n and m are preset non-negative integer buffer values (e.g., n=16, m=32), used to ensure that at least m preceding samples are retained before the event occurs and at least n subsequent samples are retained after the event occurs, preventing insufficient data prediction due to an excessively large random offset and truncating crucial posterior data due to an excessively small random offset M. In other words, setting the upper limit ensures that at least m samples are available for prediction before the event occurs, and setting the lower limit n ensures that at least n subsequent samples can still be collected after the event occurs, preventing truncating crucial post-event data due to an excessively small random offset M.
[0393] Then, the user terminal continues to collect and write M communication signal samples. For example... Figure 3 As indicated by the middle arrow M, after the event trigger point, the user terminal does not stop data acquisition, but continues to write M communication signal samples into the circular buffer. Once these M data points have been written, the data block containing the event can be extracted and saved to the storage medium.
[0394] It is worth noting that throughout the entire extraction and saving process, the circular buffer remains continuously operational. That is, at time t+M+1, the buffer continues to sequentially write subsequent sampled data, without interrupting real-time acquisition due to data extraction operations. This mechanism not only ensures that events are in random positions within the final saved data block, preventing the model from learning fixed temporal location information, but also guarantees the continuity and real-time nature of data acquisition.
[0395] After writing M samples, we can start from the current write pointer position (i.e., Figure 3 Starting from the end of the middle arrow M, extract N consecutive historical samples from the circular buffer along the time backtracking direction (e.g., counterclockwise or reverse reading) and save them to persistent storage. If the read range crosses the logical boundary of the buffer (i.e., a wraparound occurs), the data at the end and beginning of the buffer can be automatically concatenated to ensure that the N consecutive historical samples are completely extracted and saved to persistent storage.
[0396] In other words, because the buffer is circular, the extracted N historical samples may cross the physical boundaries of the buffer (i.e., traversing back from the current position to the head of the buffer and then looping back to the tail). This data can be automatically concatenated to ensure that a complete N-point data block containing the preset event and its preceding and following context is extracted. Subsequently, the extracted N historical samples can be saved to persistent storage media (such as Flash memory).
[0397] This embodiment introduces a random offset M, making the timing of events in the saved data block randomly variable (fluctuating between n and N-1-m). This mechanism avoids the subsequent training of the prediction model from overfitting to a fixed time offset pattern (e.g., preventing the prediction model from only learning the rule that events always occur at the 256th sample), thereby improving the generalization ability of the prediction model in practical applications. Simultaneously, only local data triggered by key events is persisted, avoiding full log storage and significantly saving storage space. The saved event context data can be used for offline training of the prediction model to predict RLF events, OOC events, measurement report trigger events, and also for predicting signal quality and weak signal events. The user terminal can continuously store a portion of the communication signals used for extracting common movement paths and summary information, such as path node information, time-related information, and event information. The overall communication signal record can utilize this circular buffer scheme, storing only local data before and after key events for prediction model training. By combining commonly used mobile paths and their corresponding summary information, user terminals can trigger prediction models to make predictions only near historical critical communication anomaly points in commonly used mobile paths (such as path nodes in commonly used mobile paths where summary information indicates that RLF events or OOC events have occurred). This method can not only save storage, but also reduce the computational and power consumption overhead of prediction.
[0398] For example, during a user's daily commute, the user terminal frequently experiences Recurrent Leaks (RLFs) at subway tunnel entrances. After each RLF trigger, M samples can be generated (e.g., M=56 in a given instance). After writing M samples following the RLF event, N (e.g., 1024) historical samples containing the RLF event are extracted. This data segment contains contextual information before and after the RLF event (e.g., changes and drops in RSRP and SINR before the RLF, and characteristics of neighboring cell signals). Due to the randomness of M, the timing distribution of RLF events within this data block is not fixed, providing high-quality and randomly distributed training samples for the prediction model.
[0399] This embodiment implements a lightweight, event-driven strategy for acquiring and storing communication signal data, which can save storage space and provide a reliable data foundation for model training.
[0400] In one possible implementation, before acquiring the historical communication signal information and real-time communication signal information of the user terminal, the method further includes: acquiring the complete historical movement path of the user terminal, the complete historical movement path containing multiple path nodes arranged in chronological order; identifying the target path node among the multiple path nodes where communication quality anomalies occur; determining the position of the target path node in the complete historical movement path, and extracting a local path segment containing the target path node, as well as a first preset number of path nodes before the target path node and a second preset number of path nodes after the target path node; and storing the local path segment and its corresponding signal information as historical communication signal information.
[0401] Specifically, to save storage space on user terminals, this embodiment proposes a local path storage mechanism. First, the complete historical mobility path of the user terminal is obtained, which includes multiple path nodes arranged in chronological order. Next, the target path node experiencing communication quality anomalies (e.g., a node experiencing an RLF event, an OOC event, or a poor cell event) is identified among the multiple path nodes.
[0402] Subsequently, the position of the target path node in the complete historical movement path is determined, and a local path segment containing the target path node, a first preset number of path nodes preceding the target path node, and a second preset number of path nodes following the target path node is extracted. The first and second preset numbers can be fixed values or dynamically adjusted values configured according to actual needs. The first preset number of nodes is mainly used to determine the movement direction of the user terminal before the node with abnormal communication quality, while the second preset number of nodes is mainly used to record subsequent movement directions (future path nodes).
[0403] For example, a user terminal records a raw movement trajectory (or historical movement record) from node 1 to node 100. Real-time monitoring or post-event analysis reveals that communication quality anomalies (such as RLF events, OOC events, or poor cell events) only occurred at nodes 50 and 56, while communication remained stable at other nodes.
[0404] At this point, to reduce storage overhead, it is not necessary to store the complete information of nodes 1 to 100. Instead, a local path extraction mechanism is adopted. Specifically, the first preset quantity is set to 7 and the second preset quantity is set to 1, so that only the local segment centered on the abnormal node can be extracted and stored.
[0405] For node 50, cut the segment extending forward by 7 nodes (node 43) to extending backward by 1 node (node 51) (i.e., node 43 to 51).
[0406] For node 56, the segment extending forward by 7 nodes (node 49) to the segment extending backward by 1 node (node 57) is truncated (i.e., nodes 49 to 57). Combining the above two segments, only the local path segment from node 43 to node 57 and its corresponding communication signal information need to be persistently stored. As for the remaining nodes (such as nodes 1 to 42 and nodes 58 to 100) and their related information, they are either discarded or not stored.
[0407] This mechanism retains only critical local path segments containing communication quality anomalies. These filtered local path segments and their corresponding communication signal information can then be used to generate frequently used paths, build summary information, or train models. This method reduces data storage overhead, retains the critical path segments whose communication quality most needs improvement, and ensures that the retained data includes key features such as movement direction, thereby effectively improving mobility management and enhancing user experience.
[0408] This application embodiment can divide an electronic device or main control device into functional modules according to the above method examples. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional module. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division; in actual implementation, there may be other division methods.
[0409] Figure 4 This is a schematic diagram of a device for enhancing user terminal mobility, provided as an embodiment of this application. Figure 4 As shown, the device includes: an acquisition module 410, a prediction module 420, and a determination module 430.
[0410] The acquisition module 410 is used to acquire historical communication signal information and real-time communication signal information of the user terminal. The historical communication signal information includes a historical path node sequence and the historical signal information corresponding to each historical path node in the historical path node sequence. The real-time communication signal information includes the current path node and the corresponding current signal information.
[0411] Prediction module 420 is used to determine at least one future path node and / or future signal information based on historical communication signal information and real-time communication signal information, wherein the future signal information includes the future signal information of the current path node and / or the future signal information of each future path node;
[0412] The determination module 430 is used to determine the mobility operation of the user terminal based on at least one future path node and / or future signal information. The mobility operation includes at least one of measuring neighboring cell signals, reporting measurement events to the network side, handing over a cell, reselecting a cell, performing cell search, cell selection, radio link reconstruction, and radio link establishment.
[0413] In one possible implementation, the real-time communication signal information also includes at least one preceding path node of the current path node. The prediction module 420 includes: a current trajectory construction module, a matching module, and a node determination module.
[0414] The current trajectory construction module is used to construct the current movement trajectory based on the current path node and at least one preceding path node;
[0415] The matching module is used to perform trajectory matching based on path node identifiers between the historical path node sequence and the current movement trajectory, so as to extract at least one target historical path segment that matches the current movement trajectory from the historical path node sequence.
[0416] The node determination module is used to determine at least one future path node based on at least one historical path node that appears consecutively after the last path node in the target historical path segment; and / or to determine at least one future path node and / or future signal information based on at least one historical path node that appears consecutively after the last path node in the target historical path segment and the corresponding historical signal information.
[0417] In one possible implementation, the prediction module 420 further includes: a historical sequence construction module, a training module, a current sequence construction module, an input model module, and a result output module.
[0418] The historical sequence construction module is used to construct historical spatiotemporal feature sequences based on the historical path node sequence and the historical signal information corresponding to each historical path node.
[0419] The training module is used to train a prediction model using historical spatiotemporal feature sequences. The prediction model is a machine learning model built on a neural network.
[0420] The current sequence construction module is used to construct the current spatiotemporal feature sequence based on the current path node and the current signal information of the current path node;
[0421] The input model module is used to input the current spatiotemporal feature sequence into the prediction model;
[0422] The results output module is used to output predicted values of at least one future path node's identifier and / or future signal information through the prediction model.
[0423] In one possible implementation, the prediction module 420 has a multi-model fusion structure, including at least one path prediction sub-model and at least one signal information prediction sub-model.
[0424] The path prediction sub-model is used to predict the identifier of at least one future path node based on the current spatiotemporal feature sequence;
[0425] The signal information prediction sub-model is used to predict the future signal information of each future path node based on the current spatiotemporal feature sequence and the future path node identifier output by the path prediction sub-model, and / or to predict the future signal information of the current path node based on the current spatiotemporal feature sequence.
[0426] In one possible implementation, the device further includes: an acquisition module, an error analysis module, and a training module.
[0427] The acquisition module is used to collect the actual path nodes and corresponding actual signal information actually traversed by the user terminal in real time during the application of the prediction model.
[0428] The error analysis module is used to perform error analysis between the actual path nodes and the corresponding predicted future path nodes, and / or to perform error analysis between the actual signal information and the predicted values of the corresponding predicted future signal information, and generate model analysis results.
[0429] The training module is used to incrementally train or retrain the prediction model based on the actual path nodes and corresponding actual signal information and / or historical communication signal information when the model analysis results meet the preset retraining trigger conditions.
[0430] In one possible implementation, the determining module 430 includes: a current signal measurement value acquisition module, a first operation module, a second operation module, and a third operation module;
[0431] The current signal measurement value acquisition module is used to acquire the current signal measurement value of the currently camped cell of the current path node;
[0432] The first operation module is configured to perform co-frequency measurement and obtain the current signal measurement value of neighboring cells at the same frequency point when the future signal information of the current camping cell of the current path node indicates that a first communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a first preset signal quality threshold; and / or, when the future signal information of the current camping cell of the current path node indicates that a second communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a second preset signal quality threshold, report a measurement event for triggering inter-frequency or inter-system measurement configuration to the network side, or directly perform inter-frequency / inter-system measurement; and / or, based on the future path node and / or the future signal information of the future path node, set measurement priorities for multiple test frequency points and obtain the current signal measurement value of neighboring cells of at least one of the multiple test frequency points;
[0433] The second operation module is used to report a measurement event for triggering handover or to perform cell reselection to the network side when the future signal information of the current camp cell of the current path node indicates that a third communication abnormality event will occur and / or the signal quality parameter in the future signal information is lower than a third preset signal quality threshold.
[0434] Wherein, the first preset signal quality threshold is greater than or equal to the second preset signal quality threshold, and the second preset signal quality threshold is greater than the third preset signal quality threshold; the reporting of measurement events used to trigger inter-frequency or inter-system measurement configurations, and the reporting of measurement events used for handover, respectively include: directly triggering the reporting based on the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, or after correcting the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, triggering the reporting based on the corrected value;
[0435] The third operation module is used to determine the cell search priority based on future path nodes and / or the future signal information corresponding to the future path nodes after a radio link failure (RLF) event or an out-of-coverage (OOC) event occurs.
[0436] In one possible implementation, the device further includes: an alignment module, a similarity index calculation module, a path extraction module, and a path storage module.
[0437] The alignment module is used to preprocess and align the historical path node sequence in the communication signal records of the user terminal within a preset time window.
[0438] The similarity index calculation module is used to calculate the path similarity index between each historical path node sequence;
[0439] The path extraction module is used to group historical path node sequences with path similarity indices within the preset path similarity threshold range into the same category to extract commonly used mobile paths. The preset path similarity threshold includes a fixed threshold or an adaptive threshold. The adaptive threshold is dynamically learned by calculation or machine learning model based on the distribution of currently recorded historical paths.
[0440] The path storage module is used to store commonly used mobile paths and their corresponding signal information as historical communication signal information.
[0441] In one possible implementation, the device further includes: a summary information extraction module and a sequence determination module;
[0442] The summary information extraction module is used to extract historical signal information corresponding to each historical path node from historical communication signal information based on commonly used mobile paths, generate summary information corresponding to the historical path nodes, and store the summary information in association with commonly used mobile paths.
[0443] The sequence determination module is used to report measurement events, trigger measurement procedures, select and execute cell handover or reselection based on summary information and / or future signal information, and determine the priority of measurement sequence and / or cell search sequence.
[0444] The summary information includes at least one of the following: signal quality statistics of each historical path node of each commonly used mobile path, the number and / or probability of RLF events, the number and / or probability of OOC events, the types of poor cell events and their corresponding number and / or probability, the number and / or probability of entering each node, and the average dwell time; the summary information of the same physical cell in the historical path nodes of different commonly used mobile paths is stored independently.
[0445] The signal quality statistics include: for each historical path node of each commonly used mobile path, calculating and recording the average of the previous preset number of times each type of measurement value was stationed at the node, and performing sliding window averaging or infinite impulse response (IIR) filtering on the average value corresponding to multiple stationings; and calculating and recording the number and / or probability of each measurement value being higher than each preset threshold each time stationed at the node, and performing sliding averaging or IIR filtering on the number and / or probability corresponding to multiple stationings.
[0446] The measured values include the Reference Received Power (RSRP) and the Signal-to-Interference-plus-Noise Ratio (SINR); poor cell events include: events where the number of times the cell's measured value is higher than a preset threshold is lower than a preset number threshold during a single dwell period; events where the probability of the cell's measured value being higher than a preset threshold is lower than a preset probability threshold; events where the dwell time of the cell is lower than a preset duration threshold; events where the cell's signal quality parameters are higher than a preset quality threshold and the resource allocation is lower than a preset allocation threshold; and events where network registration is rejected.
[0447] In one possible implementation, the device further includes: a filtering module, a priority allocation module, a threshold determination module, a prohibition module, a correction module, and an avoidance module.
[0448] The filtering module is used to filter out, when performing cell search, cell selection, radio link reconstruction or radio link establishment, candidate cells corresponding to consecutive nodes after a path segment that matches a common mobile path and / or candidate cells corresponding to predicted future path nodes, if one or more cells are detected on the same radio access technology (RAT) and frequency point when one or more cells are detected.
[0449] The priority allocation module is used to allocate a trial camping priority or access priority to multiple detected cells based on the Physical Cell Identifier (PCI) and summary information and / or future signal information of each candidate cell.
[0450] The threshold determination module is used to determine a second preset threshold that is positively correlated with the probability of a poor cell event occurring if the summary information and / or future signal information corresponding to any candidate cell indicates that the probability of a poor cell event occurring in the candidate cell is greater than a first preset threshold.
[0451] The prohibition module is used to generate a uniformly distributed random number in the range of 0 to 1, and prohibit the initiation of camping or access to the candidate cell when the random number is less than a second preset threshold; wherein, the second preset threshold is positively correlated with the probability of a bad cell event, so that the probability of the corresponding candidate cell being prohibited from camping or accessing is positively correlated with the probability of a bad cell event.
[0452] The correction module is used to correct the signal measurement values of the currently camped or accessed cell and / or the detected neighboring cells if the cell currently camped or accessed detects a poor congestion cell event, or if the corresponding future signal information indicates that the probability of a poor congestion cell event in the current cell is greater than a third preset threshold.
[0453] The avoidance module is used to trigger cell reselection or cell handover based on the corrected signal measurement values in order to actively escape the current cell.
[0454] In one possible implementation, the device further includes: a storage setting module, a data writing module, an offset generation module, and a persistent storage module.
[0455] The storage setting module is used to set a circular buffer of length N, where N is a positive integer. The circular buffer is used to temporarily store online acquired communication signal samples.
[0456] The data writing module is used to write the collected communication signal samples sequentially into the circular buffer in chronological order, and when the amount of data written exceeds N, the new data overwrites the earliest written data;
[0457] The offset generation module is used to start a random number generator to generate an integer offset M when a preset event is detected, where M is greater than or equal to n and less than or equal to N-1-m, and n and m are preset non-negative integer buffer values;
[0458] The persistent storage module is used to continue to collect and write M communication signal samples to the circular buffer after detecting a preset event, and backtrack from the current writing position to extract N consecutive historical samples in the circular buffer, and save the extracted N historical samples to the persistent storage medium.
[0459] The preset events include at least one of the following: RLF event, OOC event, or measurement report trigger event.
[0460] In one possible implementation, the device further includes: a historical path acquisition module, an anomaly identification module, a local segment extraction module, and a local segment storage module.
[0461] The historical path acquisition module is used to acquire the complete historical movement path of the user terminal. The complete historical movement path contains multiple path nodes arranged in chronological order.
[0462] The anomaly detection module is used to identify target path nodes among multiple path nodes that have communication quality anomalies;
[0463] The local fragment extraction module is used to determine the position of the target path node in the complete historical movement path, and extract a local path fragment containing the target path node, a first preset number of path nodes before the target path node, and a second preset number of path nodes after the target path node.
[0464] The local segment storage module is used to store local path segments and their corresponding signal information as historical communication signal information.
[0465] This embodiment provides an apparatus for enhancing user terminal mobility, which can execute a method for enhancing user terminal mobility described in the above embodiment. Its implementation principle and technical effects are similar, and will not be repeated here.
[0466] In a specific implementation of the aforementioned device for enhancing user terminal mobility, each module can be implemented as a processor. The processor can execute computer execution instructions stored in the memory, thereby enabling the processor to execute the aforementioned method for enhancing user terminal mobility.
[0467] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device includes at least one processor 510 and a memory 520. The electronic device also includes a communication component 530. The processor 510, memory 520, and communication component 530 are connected via a bus 540.
[0468] In a specific implementation, at least one processor 510 executes computer execution instructions stored in memory 520, causing at least one processor 510 to execute a method for enhancing user terminal mobility as executed on the electronic device side as described above.
[0469] The specific implementation process of processor 510 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.
[0470] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.
[0471] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage.
[0472] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.
[0473] The above description of the functions implemented by electronic devices and main control devices has introduced the solutions provided by the embodiments of the present invention. It is understood that, in order to implement the above functions, the electronic device or main control device includes hardware structures and / or software modules corresponding to the execution of each function. By combining the units and algorithm steps of the various examples described in the embodiments of the present invention, the embodiments of the present invention can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the technical solutions of the embodiments of the present invention.
[0474] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the above-described method for enhancing user terminal mobility.
[0475] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.
[0476] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in an electronic device or a host device.
[0477] This application also provides a computer program product, which includes a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the solution provided in the above embodiments.
[0478] This application also provides a chip, which includes at least one processor for executing program instructions to perform the solutions provided in the above embodiments. A method for enhancing the mobility of a user terminal can be applied to a user terminal device, a chip within the user terminal device, or a chip module. The user terminal device can be a smartphone, an in-vehicle device, an IoT terminal, etc.
[0479] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disk, or optical disk.
[0480] The technical solutions of this application have been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it is readily understood by those skilled in the art that the scope of protection of this application is obviously not limited to these specific embodiments. The above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.
Claims
1. A method for enhancing the mobility of a user terminal, characterized in that, include: The system acquires historical communication signal information and real-time communication signal information of the user terminal. The historical communication signal information includes a historical path node sequence and historical signal information corresponding to each historical path node in the historical path node sequence. The real-time communication signal information includes the current path node and the corresponding current signal information. Based on the historical communication signal information and the real-time communication signal information, at least one future path node and / or future signal information is determined, wherein the future signal information includes the future signal information of the current path node and / or the future signal information of each future path node; Based on the at least one future path node and / or future signal information, the mobility operation of the user terminal is determined, the mobility operation including at least one of measuring neighboring cell signals, reporting measurement events to the network side, handing over a cell, reselecting a cell, performing cell search, cell selection, radio link reconstruction, and radio link establishment.
2. The method according to claim 1, characterized in that, The real-time communication signal information also includes at least one preceding path node of the current path node; The step of determining at least one future path node and / or future signal information based on the historical communication signal information and the real-time communication signal information includes: Construct the current movement trajectory based on the current path node and the at least one preceding path node; The historical path node sequence is matched with the current movement trajectory based on path node identifiers to extract at least one target historical path segment that matches the current movement trajectory from the historical path node sequence. The at least one future path node is determined based on at least one historical path node that appears consecutively after the last path node in the target historical path segment; and / or the at least one future path node and / or future signal information is determined based on at least one historical path node that appears consecutively after the last path node in the target historical path segment and the corresponding historical signal information.
3. The method according to claim 1, characterized in that, The step of determining at least one future path node and / or future signal information based on the historical communication signal information and the real-time communication signal information includes: Based on the historical path node sequence and the historical signal information corresponding to each historical path node, a historical spatiotemporal feature sequence is constructed. The prediction model is trained using the historical spatiotemporal feature sequence, and the prediction model is a machine learning model built on a neural network. Construct the current spatiotemporal feature sequence based on the current path node and the current signal information of the current path node; The current spatiotemporal feature sequence is input into the prediction model; The prediction model outputs predicted values of at least one future path node's identifier and / or future signal information.
4. The method according to claim 3, characterized in that, The prediction model is a multi-model fusion structure, including at least one path prediction sub-model and at least one signal information prediction sub-model; The path prediction sub-model is used to predict the identifier of at least one future path node based on the current spatiotemporal feature sequence; The signal information prediction sub-model is used to predict the predicted value of the future signal information of each future path node based on the current spatiotemporal feature sequence and the future path node identifier output by the path prediction sub-model, and / or to predict the predicted value of the future signal information of the current path node based on the current spatiotemporal feature sequence.
5. The method according to claim 3 or 4, characterized in that, The method further includes: During the application of the prediction model, the actual path nodes traversed by the user terminal and the corresponding actual signal information are collected in real time. Error analysis is performed between the actual path nodes and the corresponding predicted future path nodes, and / or error analysis is performed between the actual signal information and the predicted value of the corresponding predicted future signal information to generate model analysis results; When the model analysis results meet the preset retraining trigger conditions, the prediction model is incrementally trained or retrained based on the actual path nodes and corresponding actual signal information and / or the historical communication signal information.
6. The method according to claim 1, characterized in that, The step of determining the mobility operation of the user terminal based on the at least one future path node and / or future signal information includes: Obtain the current signal measurement value of the currently camped cell of the current path node; When the future signal information of the current camping cell of the current path node indicates that a first communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a first preset signal quality threshold, perform co-frequency measurement to obtain the current signal measurement value of the neighboring cell at the same frequency point; and / or, when the future signal information of the current camping cell of the current path node indicates that a second communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a second preset signal quality threshold, report a measurement event to the network side to trigger inter-frequency or inter-system measurement configuration, or directly perform inter-frequency / inter-system measurement; and / or, based on the future path node and / or the future signal information of the future path node, set measurement priorities for multiple test frequencies and obtain the current signal measurement value of the neighboring cell of at least one of the multiple test frequencies; When the future signal information of the currently camped cell at the current path node indicates that a third communication anomaly event will occur and / or the signal quality parameter in the future signal information is lower than a third preset signal quality threshold, a measurement event for triggering handover or cell reselection is reported to the network side. Wherein, the first preset signal quality threshold is greater than or equal to the second preset signal quality threshold, and the second preset signal quality threshold is greater than the third preset signal quality threshold; the reporting of measurement events for triggering inter-frequency or inter-system measurement configurations and the reporting of measurement events for handover respectively include: directly triggering the reporting based on the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, or after correcting the current signal measurement value of the current cell and / or the current signal measurement value of the neighboring cell, triggering the reporting based on the corrected value; After a Radio Link Failure (RLF) event or an Out of Coverage (OOC) event occurs, the cell search priority is determined based on the future path node and / or the future signal information corresponding to the future path node.
7. The method according to claim 1, characterized in that, Before acquiring the historical and real-time communication signal information of the user terminal, the method further includes: Preprocessing and feature alignment are performed on the historical path node sequence in the communication signal records of the user terminal within a preset time window; Calculate the path similarity index between each historical path node sequence; Based on a preset path similarity threshold, historical path node sequences whose path similarity index falls within the preset path similarity threshold range are grouped into the same category to extract commonly used mobile paths; wherein, the preset path similarity threshold includes a fixed threshold or an adaptive threshold, and the adaptive threshold is dynamically learned through calculation or machine learning models based on the currently recorded historical path distribution. The commonly used mobile paths and their corresponding signal information are stored as historical communication signal information.
8. The method according to claim 7, characterized in that, The method further includes: Based on the commonly used mobile paths, historical signal information corresponding to each historical path node is extracted from the historical communication signal information, summary information corresponding to the historical path nodes is generated, and the summary information is associated with and stored with the commonly used mobile paths; Based on the summary information and / or the future signal information, report measurement events, trigger measurement procedures, select and execute cell handover or reselection, and determine the priority of measurement order and / or cell search order; The summary information includes at least one of the following: signal quality statistics of each historical path node of each commonly used mobile path, the number and / or probability of RLF events, the number and / or probability of OOC events, the types of poor cell events and their corresponding number and / or probability, the number and / or probability of entering each node, and the average dwell time; the summary information of the same physical cell in the historical path nodes of different commonly used mobile paths is stored independently. The signal quality statistics include: for each historical path node of each commonly used mobile path, calculating and recording the average of the previous preset number of times each type of measurement value is stationed at the node, and performing sliding window averaging or infinite impulse response (IIR) filtering on the average value corresponding to multiple stationings; and calculating and recording the number and / or probability of each measurement value being higher than each preset threshold each time stationed at the node, and performing sliding averaging or IIR filtering on the number and / or probability corresponding to multiple stationings. The measured values include the Reference Received Power (RSRP) and the Signal-to-Interference-plus-Noise Ratio (SINR); the poor cell events include: during a single stay, the number of times the cell's measured value is higher than a preset threshold is less than a preset number threshold; the probability that the cell's measured value is higher than a preset threshold is less than a preset probability threshold; the stay duration of the cell is less than a preset duration threshold; the signal quality parameters of the cell are higher than a preset quality threshold and the resource allocation is less than a preset allocation threshold; and the event of network registration being rejected.
9. The method according to claim 8, characterized in that, The method further includes: When performing cell search, cell selection, radio link reconstruction, or radio link establishment, if one or more cells are detected on the same radio access technology (RAT) and frequency point, then from the one or more cells, candidate cells corresponding to consecutive nodes after the path segment that matches the common mobile path are selected, and / or candidate cells corresponding to the predicted future path nodes are selected. Based on the Physical Cell Identifier (PCI) of each candidate cell and the summary information and / or the future signal information, an attempt to camp or access priority is assigned to the detected multiple cells. If the summary information and / or the future signal information corresponding to any candidate cell indicate that the probability of a bad cell event occurring in the candidate cell is greater than a first preset threshold, then a second preset threshold that is positively correlated with the probability of a bad cell event occurring is determined. Generate a uniformly distributed random number in the range of 0 to 1, and prohibit camping or access to the candidate cell when the random number is less than the second preset threshold; wherein, the second preset threshold is positively correlated with the probability of the occurrence of the poor cell event, so that the probability of the corresponding candidate cell being prohibited from camping or access is positively correlated with the probability of the occurrence of the poor cell event; If the currently camped or accessed cell detects a poor congestion cell event, or if the corresponding future signal information indicates that the probability of a poor congestion cell event in the current cell is greater than a third preset threshold, then the signal measurement values of the currently camped or accessed cell and / or the detected neighboring cells are corrected. Based on the corrected signal measurement values, cell reselection or cell handover is triggered to actively escape the current cell.
10. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1 to 9.
