Multi-mode terminal autonomous network selection method, system, medium and terminal based on lightweight machine learning model

Abstract

This application provides a method, system, medium, and terminal for autonomous network selection by multi-mode terminals based on a lightweight machine learning model. The method includes: collecting physical measurement data and service experience data of multi-mode terminals based on a preset low-power sampling mechanism and preprocessing them; extracting features from the preprocessed physical measurement data and service experience data to obtain scene coverage features, time period features, network features, and service experience features to construct a multi-dimensional feature set; inputting the constructed multi-dimensional feature set into a lightweight machine learning model for offline incremental training under low-load periods to obtain a cell experience scoring model; and, under a preset inference triggering scenario, calling the trained cell experience scoring model and selecting the cell to camp on based on the experience scoring information generated by the cell experience scoring model. This application can adapt to the personalized usage scenarios of different users, improve the actual service experience of users, and at the same time, reduce the difficulty of implementing the solution.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to the field of RedCap / LTE multi-mode terminal mobility management optimization compliant with 3GPP protocol specifications. Specifically, it relates to a method, system, medium, and terminal for multi-mode terminals to autonomously select networks based on a lightweight machine learning model. Background Technology

[0002] With the large-scale deployment of 5G RedCap (Reduced Capability) technology, hybrid networking of RedCap and LTE (Long Term Evolution) has become a typical network architecture for IoT terminals. In this scenario, cell dwell and mobility management for RedCap / LTE multi-mode IoT terminals become critical aspects affecting user service experience. Multi-mode terminals need to perform operations such as cell selection, prioritization, dwell decision, and mobility management between RedCap and LTE cells to ensure service continuity and quality of service under different network coverage conditions.

[0003] Currently, mobility management for all multi-mode terminals follows the 3GPP (Third Generation Partnership Project) standard protocol, employing a standard static threshold network selection scheme to screen, prioritize, and decide which candidate cells to serve. This scheme, as the native default scheme for terminals, constitutes the industry's common foundation. Specifically, the cell selection and reselection process for LTE follows the 3GPP TS 36.304 specification, while the cell selection and reselection process for 5G NR RedCap follows the 3GPP TS 38.304 specification. In detail, the static threshold network selection scheme prioritizes candidate cells based on pre-set physical layer measurement thresholds such as RSRP (Reference Signal Received Power) and SINR (Signal to Interference plus Noise Ratio), and defaults to enabling the RedCap network-priority serving strategy. However, this scheme only determines serving and reselection based on physical signal quality and cannot consider cell interference levels, actual service performance, network load status, or user usage scenarios.

[0004] To overcome the shortcomings of the aforementioned general basic solutions, the industry has developed terminal-fixed parameter adaptive optimization solutions and cloud-end-network collaborative AI (Artificial Intelligence) network selection solutions. Among them, the terminal-fixed parameter adaptive optimization solution is based on extensive field testing experience, manually adjusting fixed parameters such as RedCap, LTE cell camping threshold, reselection hysteresis, and handover bias to compensate for and optimize parameters for common weak coverage and interference scenarios. However, all parameters in this solution are fixed configurations in firmware, which take effect uniformly across the entire network and cannot be dynamically adjusted according to individual user scenarios and spatiotemporal movement patterns. The cloud-based, end-to-end collaborative AI (Artificial Intelligence) network selection solution relies heavily on the operator's cloud-based big data platform to collect network measurement data and service experience data from all terminals across the network. It then trains a general network selection optimization model using a deep learning model in the cloud and distributes the optimization parameters to the terminals, thereby achieving unified intelligent network selection optimization across the entire network. However, this solution depends on data reporting from network-side devices, cloud computing power support, and end-to-end collaborative signaling interaction. It requires the cooperation of operator network upgrades, has a high deployment threshold, is difficult to implement, and cannot be adapted to users' specific usage scenarios.

[0005] Therefore, it is necessary to provide a multi-mode terminal autonomous network selection method, system, medium, and terminal based on a lightweight machine learning model to solve the above-mentioned problems in the prior art. Summary of the Invention

[0006] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a method, system, medium and terminal for autonomous network selection of multi-mode terminals based on a lightweight machine learning model, in order to solve the technical problems of poor user experience, inability to adapt to personalized user scenarios and high deployment costs of existing RedCap / LTE multi-mode IoT terminals in mixed networking scenarios.

[0007] To achieve the above and other related objectives, the first aspect of this application provides a method for autonomous network selection by a multi-mode terminal based on a lightweight machine learning model, comprising: collecting physical measurement data and service experience data of a multi-mode terminal based on a preset low-power sampling mechanism, and preprocessing the collected physical measurement data and service experience data; extracting features from the preprocessed physical measurement data and service experience data to obtain scene coverage features, time period features, network features, and service experience features to construct a multi-dimensional feature set; inputting the constructed multi-dimensional feature set into a lightweight machine learning model for offline incremental training under low-load periods to train a cell experience scoring model; the cell experience scoring model is used to output experience scoring information for each candidate cell; under a preset inference triggering scenario, calling the trained cell experience scoring model, and selecting the cell to reside in based on the experience scoring information generated by the cell experience scoring model.

[0008] In some embodiments of the first aspect of this application, the preset low-power sampling mechanism is as follows: if the multi-mode terminal is detected to be in an idle state, a low-frequency sampling mechanism is used to collect physical measurement data and service experience data; if the network status of the multi-mode terminal is detected to change or to be in a service operation state, a high-frequency sampling mechanism is used to collect physical measurement data and service experience data.

[0009] In some embodiments of the first aspect of this application, the physical measurement data includes: time information, location information, reference signal received power, and signal-to-interference-plus-noise ratio; the service experience data includes: block error rate, uplink and downlink rates, and end-to-end latency data.

[0010] In some embodiments of the first aspect of this application, the preprocessing step further includes: converting the location information into a grid code based on the Geohash algorithm to aggregate physical measurement data and business experience data in the same spatial dimension.

[0011] In some embodiments of the first aspect of this application, the low-load period includes: charging period, screen-off period, and terminal idle period.

[0012] In some embodiments of the first aspect of this application, the preset inference triggering scenarios include: initial cell selection scenario, cell reselection scenario, and mobility handover scenario.

[0013] In some embodiments of the first aspect of this application, the method further includes: continuously collecting newly added physical measurement data and service experience data, and periodically iterating and training the trained cell experience scoring model based on the newly added physical measurement data and service experience data, so as to optimize the cell experience scoring model and network selection parameters.

[0014] To achieve the above and other related objectives, a second aspect of this application provides a multi-mode terminal autonomous network selection system based on a lightweight machine learning model, comprising: a data acquisition module, used to acquire physical measurement data and service experience data of the multi-mode terminal based on a preset low-power sampling mechanism, and to preprocess the acquired physical measurement data and service experience data; a feature set construction module, used to extract features from the preprocessed physical measurement data and service experience data to obtain scene coverage features, time period features, network features, and service experience features, so as to construct a multi-dimensional feature set; a model training module, used to input the constructed multi-dimensional feature set into a lightweight machine learning model for offline incremental training under low load periods, so as to train a cell experience scoring model; the cell experience scoring model is used to output the experience scoring information of each candidate cell; and a model inference module, used to call the trained cell experience scoring model to generate the experience scoring information of each candidate cell under a preset inference trigger scenario, and to select the cell to be camped based on the generated experience scoring information.

[0015] To achieve the above and other related objectives, a third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method.

[0016] To achieve the above and other related objectives, a fourth aspect of this application provides an electronic terminal, including a memory, a processor, and a computer program stored in the memory; the processor executes the computer program to implement the method.

[0017] As described above, the multi-mode terminal autonomous network selection method, system, medium, and terminal based on a lightweight machine learning model of this application have the following beneficial effects:

[0018] Physical measurement data and service experience data of multi-mode terminals are collected based on a preset low-power sampling mechanism, and the collected physical measurement data and service experience data are preprocessed. Then, feature extraction is performed on the preprocessed physical measurement data and service experience data to extract scene coverage features, time period features, network features, and service experience features, and a multi-dimensional feature set is constructed based on this. Then, the constructed multi-dimensional feature set is input into a lightweight machine learning model for offline incremental training under low load periods, thereby training a cell experience scoring model for outputting experience scoring information for each candidate cell. Finally, under a preset inference trigger scenario, the trained cell experience scoring model is invoked to perform inference, and the cell to be camped is selected based on the experience scoring information generated by the cell experience scoring model. This can adapt to the personalized usage scenarios of different users, improve the actual service experience of users, and at the same time reduce the difficulty of implementation and save deployment costs. Attached Figure Description

[0019] Figure 1 The diagram shown is a flowchart illustrating a multi-mode terminal autonomous network selection method based on a lightweight machine learning model in one embodiment of this application.

[0020] Figure 2 The diagram shown illustrates the working principle of a multi-mode terminal autonomous network selection method based on a lightweight machine learning model in one embodiment of this application.

[0021] Figure 3 The diagram shown is a block diagram of a multi-mode terminal autonomous network selection system based on a lightweight machine learning model, according to an embodiment of this application.

[0022] Figure 4 The diagram shown is a structural schematic of an electronic terminal according to an embodiment of this application. Detailed Implementation

[0023] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0024] In the embodiments of this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially 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 the terms "first" and "second" do not necessarily imply that they are different.

[0025] It should be noted that, in the embodiments of this application, the words "exemplary" or "for example" indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0026] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0027] Before providing a further detailed description of the present invention, the nouns and terms used in the embodiments of the present invention are explained, and the nouns and terms used in the embodiments of the present invention are subject to the following interpretations:

[0028] <1> Lightweight Machine Learning Model (LMLM) refers to a machine learning method designed specifically for resource-constrained environments that significantly reduces computational, memory, power consumption, and storage overhead compared to traditional deep learning models, while maintaining usability.

[0029] <2> GPS (Global Positioning System) is a radio navigation and positioning system based on artificial Earth satellites that can provide accurate geographic location, speed, and time information at any location in the world and in near-Earth space.

[0030] <3> Reference Signal Received Power (RSRP) is a key parameter in LTE networks that represents the strength of wireless signals and is one of the physical layer measurement requirements. It is the average signal power received on all REs (resource particles) carrying the reference signal within a certain symbol. It is also used in 5G NR mobile communication networks to characterize the power strength of the cell signal received by the terminal.

[0031] <4> Signal to Interference plus Noise Ratio (SNR): This refers to the ratio of the strength of the received useful signal to the strength of the received interference signal (noise and interference). It is a core indicator used to characterize the quality of the signal received by the terminal and reflects the degree to which the signal is affected by interference and noise.

[0032] <5> Block Error Rate (BLER) is an important indicator for measuring link quality in wireless communication. It represents the proportion of error blocks in the initial transmission block.

[0033] <6> Uplink and downlink speeds refer to the speed at which data is uploaded from a user device to the network and downloaded from the network to the user device.

[0034] <7> End-to-End Delay: refers to the total time it takes for a data packet to travel from the source node to the destination node. It reflects the overall latency performance of network transmission and is an important indicator for measuring network quality.

[0035] <8> Geohash algorithm: It is an algorithm that encodes the latitude and longitude of a geographic location into a string. By mapping two-dimensional space to one-dimensional string, it can efficiently perform fuzzy queries and nearest-neighbor searches of geographic locations.

[0036] <9> Gradient Boosting Decision Tree (GBDT) model: a commonly used ensemble learning algorithm that builds multiple weak learners (usually decision trees) and sums their results to form a powerful predictive model.

[0037] <10> LambdaRank algorithm: It is an algorithm for learning to rank (LTR) and is widely used in recommendation systems and information retrieval tasks. Its core idea is based on the RankNet algorithm and optimizes the ranking by introducing the Lambda gradient (λ value) to minimize the loss function.

[0038] In heterogeneous hybrid networking scenarios for RedCap / LTE multi-mode terminals, the traditional network selection mechanism based on 3GPP standard protocols has significant limitations in terms of user experience. Existing terminals typically rely on fixed static thresholds for cell network selection, reselection, and mobility decisions. The decision-making logic depends solely on physical signal indicators such as RSRP and SINR, without effectively linking core experience indicators such as actual service rate, transmission delay, network interference level, and cell load status. This leads to a discrepancy between the terminal's network dwell strategy and the actual user experience, ultimately resulting in three typical experience degradation scenarios: (1) Although the terminal legally dwells in a RedCap weak coverage cell based on the static threshold, the actual service experience is significantly worse than that of an accessible LTE cell, i.e., a mismatch occurs between network selection and service requirements; (2) The terminal continuously dwells in a RedCap cell with strong signal but severe interference. Due to the lack of comprehensive perception of interference intensity and service rate in the network selection decision mechanism, the rate and transmission delay deteriorate; (3) In complex environments where building indoor distribution systems and outdoor macro base stations overlap, the terminal mistakenly dwells in a cell with weak coverage due to the limitations of traditional measurement mechanisms, leading to handover delays or service interruptions. To address the shortcomings of static threshold network selection schemes, existing optimization solutions still have significant limitations. Terminal-fixed parameter adaptive optimization schemes can only perform static configurations for general scenarios, making it difficult to adapt to the personalized usage patterns, business habits, and local network characteristics of different users. Furthermore, cloud-end-network collaborative AI network selection schemes require deep reliance on network architecture modifications on the operator side, resulting in high deployment barriers and long cycles. The terminal side cannot independently complete capability building and policy activation. Neither of these solutions fundamentally solves the problem of personalized experience degradation in existing network scenarios. Therefore, this application provides a multi-mode terminal autonomous network selection method, system, medium, and terminal based on a lightweight machine learning model. It adaptively selects the network based on a multi-dimensional feature set unique to each user, thereby adapting to the personalized usage scenarios of different users, improving the actual business experience, lowering the deployment barrier, and saving implementation costs.

[0039] To facilitate understanding of the embodiments of this application, in conjunction with Figure 1 Detailed explanation. Figure 1 The illustration shows a flowchart of a multi-mode terminal autonomous network selection method based on a lightweight machine learning model in an embodiment of the present invention. Figure 2 This illustration demonstrates the working principle of a multi-mode terminal autonomous network selection method based on a lightweight machine learning model, as described in this embodiment of the invention. The multi-mode terminal autonomous network selection method based on a lightweight machine learning model in this embodiment is applied to a multi-mode terminal; that is, the method is executed entirely locally and in a closed loop on the multi-mode terminal, without any data interaction or collaboration dependency with the cloud or network side. The method includes the following steps:

[0040] Step S11: Collect physical measurement data and service experience data of the multi-mode terminal based on the preset low-power sampling mechanism, and preprocess the collected physical measurement data and service experience data.

[0041] In some embodiments of this application, the preset low-power sampling mechanism is as follows: if the multi-mode terminal is detected to be in an idle state, a low-frequency sampling mechanism is used to collect physical measurement data and service experience data; if the network status of the multi-mode terminal is detected to change or to be in a service running state, a high-frequency sampling mechanism is used to collect physical measurement data and service experience data.

[0042] In some embodiments of this application, the physical measurement data includes: time information, location information, reference signal received power, and signal-to-interference-plus-noise ratio; the service experience data includes: block error rate, uplink and downlink rates, and end-to-end latency data.

[0043] Specifically, under normal standby and service operation states, the multi-mode terminal silently collects local multi-dimensional data, including but not limited to: physical measurement data such as time information, GPS location information, and cell RSRP / SINR; and service experience data including but not limited to block error rate (BLER), uplink and downlink rates, and end-to-end latency data. The multi-mode terminal has a built-in preset low-power sampling mechanism that dynamically adjusts the sampling frequency based on the terminal's current operating state. Specifically, when the multi-mode terminal detects that it is in an idle state, it uses a low-frequency sampling mechanism for data collection, for example, collecting physical measurement data every 300 seconds and service experience data every 1800 seconds, thereby reducing terminal power consumption and storage overhead. When the multi-mode terminal detects a change in network status or is in service operation state, it switches to a high-frequency sampling mechanism for data collection, for example, collecting physical measurement data every 1000 milliseconds and service experience data every 500 milliseconds. Within the first 5 seconds after a network status change, it further performs continuous collection at 200 millisecond intervals to ensure the data integrity of critical network events and service processes. It should be noted that the aforementioned low-frequency and high-frequency data are merely illustrative examples and can be flexibly adjusted according to the terminal's computing power and business scenarios. All collected data undergoes preprocessing locally on the terminal, including data cleaning, outlier removal, and format standardization, thereby ensuring the quality of the input data. Furthermore, all preprocessed physical measurement data and business experience data are encrypted and stored locally on the terminal only, without uploading to the cloud or any external nodes. This not only ensures user privacy and security but also avoids the power consumption generated by data transmission.

[0044] In some embodiments of this application, the preprocessing step is followed by: converting the location information into a grid code based on the Geohash algorithm to aggregate physical measurement data and business experience data in the same spatial dimension.

[0045] Specifically, in order to improve the efficiency of data aggregation under the same spatial dimension, the multi-mode terminal converts the preprocessed location information into grid code based on the Geohash algorithm. Through this grid code, continuous geographical locations are mapped to a specific standardized spatial grid, thereby aggregating physical measurement data and business experience data within the same grid, reducing the dimensional redundancy of location information and enhancing the generalization ability of spatial features, so as to realize data aggregation in scenarios such as the same building, park, and fixed commuting route.

[0046] Step S12: Extract features from the preprocessed physical measurement data and business experience data to obtain scene coverage features, time period features, network features and business experience features, in order to construct a multi-dimensional feature set.

[0047] Feature engineering is performed on preprocessed physical measurement data and service experience data, and deep feature extraction is conducted within the same spatial grid to construct a unique data profile for each user. Specifically, time-series analysis is used to extract time-period features to identify users' network usage patterns at different times; spatial analysis is used to extract scene coverage features to characterize the network environment characteristics of different geographical locations; signal quality statistics are used to extract network features to reflect the physical layer connectivity quality of candidate cells; and service experience features are extracted based on service layer index aggregation to characterize the network service level in actual service processes. This constructs a multi-dimensional feature set with a mapping relationship of "location-time-network quality-service experience" to characterize the spatiotemporal usage patterns and scene network experience characteristics of individual users, providing accurate samples for model training.

[0048] Step S13: Input the constructed multidimensional feature set into a lightweight machine learning model for offline incremental training during low-load periods to train a cell experience scoring model; the cell experience scoring model is used to output the experience scoring information of each candidate cell.

[0049] In some embodiments of this application, the low-load period includes: charging period, screen-off period, and terminal idle period.

[0050] Specifically, the constructed multi-dimensional feature set is input into a lightweight machine learning model that can be trained by the terminal's computing power. This lightweight machine learning model uses a lightweight network structure suitable for terminal deployment, such as a lightweight neural network combining a gradient boosting decision tree model and the LambdaRank algorithm, abandoning high-computing-power deep learning models, thereby reducing computing resource consumption and model storage overhead. The terminal-side adaptability of the gradient boosting decision tree model and LambdaRank algorithm combination is reflected in: the number of parameters of a single model can be controlled at the level of hundreds of KB, the computing power cost of a single round of incremental training is less than 5% of the terminal's peak computing power, and the latency of a single inference is less than 10ms, which will not affect the normal operation of the terminal and is fully adapted to the limited computing power resources of IoT terminals. Multi-mode terminals perform offline incremental training on the terminal side during low-load periods, including charging periods, screen-off periods, and terminal idle periods. During incremental training, the model only fine-tunes the existing network parameters based on newly added sample data, iteratively optimizes the model parameters, and gradually generates a community experience scoring model adapted to individual user habits, without the need for full data retraining, significantly reducing terminal computing power overhead. The trained cell experience scoring model is a personalized model adapted to the spatiotemporal usage habits of a single user. It can integrate multi-dimensional information such as location, time, network quality, and service experience to output experience scores for each candidate cell. At the same time, the multi-mode terminal dynamically optimizes mobility parameters such as RedCap / LTE cell camping threshold and reselection hysteresis parameters based on model feedback during training, so that the network selection strategy matches the current network environment and user behavior.

[0051] Step S14: Under the preset reasoning trigger scenario, call the trained cell experience scoring model, and select the cell to stay in based on the experience scoring information generated by the cell experience scoring model.

[0052] In some embodiments of this application, the preset inference triggering scenarios include: initial cell selection scenario, cell reselection scenario, and mobility handover scenario.

[0053] Specifically, when a multi-mode terminal enters a preset inference trigger scenario, such as the initial cell selection scenario, cell reselection scenario, or mobility handover scenario, it invokes the cell experience scoring model trained locally on the terminal for local real-time inference. During the inference process, the multi-mode terminal inputs the real-time multi-dimensional feature set of the current candidate cells into the cell experience scoring model to predict the actual rate, latency, stability, and other service experience scores of each candidate cell, replacing the traditional single RSRP / SINR ranking logic within the candidate cell range that complies with the mandatory requirements of the 3GPP protocol. Finally, based on the generated experience score information, the multi-mode terminal autonomously determines the optimal cell to camp on, prioritizing the cell with the best experience and actively avoiding inferior camping scenarios such as RedCap weak field, high interference strong signal, and overlapping weak coverage cells, thereby prioritizing the user's service experience.

[0054] By completing data collection, feature extraction, model training, inference decision-making, and model iteration entirely on the terminal, without the need for external data interaction or network modifications, and completely independent of cloud and operator network-side collaboration capabilities, it can be deployed independently and adapted to all multi-mode terminal scenarios without terminal-network collaboration conditions. This constitutes a purely terminal-side, non-collaborative, localized, and autonomous network selection architecture. Through data aggregation and spatiotemporal feature extraction, it mines the usage patterns of users at fixed locations and time periods, constructs a location-time-network experience mapping relationship exclusive to each user, and achieves personalized network selection decisions for each individual. This differs from traditional fixed rules and universal models that are universal across the entire network, thus constructing a personalized modeling mechanism based on user spatiotemporal patterns.

[0055] In some embodiments of this application, the method further includes: continuously collecting newly added physical measurement data and service experience data, and periodically iterating and training the trained cell experience scoring model based on the newly added physical measurement data and service experience data, so as to optimize the cell experience scoring model and network selection parameters.

[0056] By adopting a lightweight machine learning model adapted to the terminal's computing power, coupled with off-peak offline training, low-power data collection, and incremental model iteration mechanisms, intelligent network selection optimization is achieved with extremely low overhead without interfering with the normal operation of the terminal's business. This addresses the core pain point of traditional algorithms, where "optimal signal does not equal optimal experience," and realizes a low-overhead, lightweight model closed-loop optimization mechanism.

[0057] Specifically, the multi-mode terminal continuously collects physical measurement data and service experience data from newly added scenarios and network environments during operation, constantly expanding the local sample dataset. Based on the newly added data, it performs periodic incremental iterative training on the cell experience scoring model, adaptively adapting to network environment fine-tuning and changes in user scenarios. Thus, through continuous data accumulation and model optimization, the cell experience scoring model and network selection parameters are adaptively updated, ensuring the long-term effectiveness of the model and the accuracy of network selection decisions, achieving long-term stable personalized network selection optimization results.

[0058] The multi-mode terminal autonomous network selection method based on a lightweight machine learning model provided in this application overcomes the shortcomings of existing technologies, which rely on network collaboration, fixed parameters, and lack personalized adaptation. It provides a purely terminal-side localized, low-overhead, and adaptive AI autonomous network selection method. Specifically, the terminal silently collects multi-dimensional network and scenario data locally; then, it completes data classification and accumulation through spatiotemporal feature extraction; next, it conducts lightweight incremental offline training during low-load periods of the terminal to generate a user-specific cell experience scoring model and dynamic network selection threshold; finally, it performs real-time inference through the terminal's local model, using service experience prediction as the core to complete cell optimal camping and mobility decisions. At the same time, it continuously iterates and updates the model to form a local closed-loop adaptive optimization system. This allows the application to eliminate the need for any collaborative capabilities between the cloud and the network side. Instead, it mines user spatiotemporal usage patterns based on the terminal's local lightweight model to construct personalized network experience mapping relationships, replacing the traditional single signal decision logic. This achieves intelligent network selection centered on actual service experience, accurately solving various network selection experience degradation problems under RedCap / LTE hybrid networking, and effectively improving core experience indicators such as terminal data service rate and latency without affecting the normal operation of the terminal.

[0059] It should be noted that all network selection decisions involved in this application are executed within the 3GPP protocol framework, strictly following the mandatory rules issued by the network side, such as the cell prohibition list, frequency band priority, and access level restrictions. Only the candidate cells allowed by the protocol are optimized for experience-oriented priority ranking. This does not violate the mandatory requirements of core specifications such as 3GPP TS 36.304 and TS 38.304, and can be directly deployed on existing multi-mode terminals.

[0060] Figure 3 This is a schematic block diagram of a multi-mode terminal autonomous network selection system based on a lightweight machine learning model, provided in an embodiment of this application. Figure 3 As shown, a multi-mode terminal autonomous network selection system 300 based on a lightweight machine learning model is deployed locally on the multi-mode terminal, requiring no cloud or network-side collaboration. The system includes:

[0061] The data acquisition module 301 is used to acquire physical measurement data and service experience data of the multi-mode terminal based on a preset low-power sampling mechanism, and to preprocess the acquired physical measurement data and service experience data.

[0062] The feature set construction module 302 is used to extract features from the preprocessed physical measurement data and business experience data to obtain scene coverage features, time period features, network features and business experience features, so as to construct a multi-dimensional feature set;

[0063] The model training module 303 is used to input the constructed multidimensional feature set into the lightweight machine learning model for offline incremental training under low load periods, so as to train the cell experience scoring model; the cell experience scoring model is used to output the experience scoring information of each candidate cell;

[0064] The model inference module 304 is used to call the trained cell experience scoring model to generate experience scoring information for each candidate cell under a preset inference triggering scenario, and select the cell to stay based on the generated experience scoring information.

[0065] It should be understood that the specific process of each module performing the above-mentioned steps has been described in detail in the above method embodiments, and will not be repeated here for the sake of brevity.

[0066] It should also be understood that the module division in the embodiments of this application is illustrative and only represents a logical functional division; in actual implementation, there may be other division methods. Furthermore, the functional modules in the various embodiments of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0067] Figure 4 This is a schematic block diagram of the electronic terminal provided in an embodiment of this application. Figure 4 As shown, the electronic terminal 400 includes at least one processor 401, a memory 402, at least one network interface 403, and a user interface 405. The various components in the electronic terminal 400 are coupled together via a bus system 404. It is understood that the bus system 404 is used to implement communication between these components. In addition to a data bus, the bus system 404 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 4 The general will label all buses as bus systems.

[0068] The user interface 405 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0069] It is understood that memory 402 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable categories of memory.

[0070] In this embodiment of the invention, the memory 402 is used to store various types of data to support the operation of the electronic terminal 400. Examples of this data include: any executable program for operation on the electronic terminal 400, such as the operating system 4021 and application programs 4022; the operating system 4021 includes various system programs, such as the framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. The application program 4022 may include various applications, such as a media player, browser, etc., for implementing various application services. The methods provided in this embodiment of the invention can be included in the application program 4022.

[0071] The methods disclosed in the above embodiments of the present invention can be applied to processor 401, or implemented by processor 401. Processor 401 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 401 or by instructions in the form of software. The processor 401 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 401 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. General-purpose processor 401 may be a microprocessor or any conventional processor, etc. The steps of the accessory optimization method provided in the embodiments of the present invention can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in a memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.

[0072] In an exemplary embodiment, the electronic terminal 400 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to execute the aforementioned method.

[0073] According to the method provided in the embodiments of this application, this application also provides a computer-readable storage medium storing program code, which, when executed on a computer, causes the computer to perform... Figures 1 to 2 The method of any of the embodiments shown.

[0074] As used in this specification, the terms "component," "module," "system," etc., are used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. As illustrated, applications running on computing devices and computing devices can both be components. One or more components may reside in a process and / or an execution thread, and components may be located on a single computer and / or distributed among two or more computers. Furthermore, these components can be executed from various computer-readable media on which various data structures are stored. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0075] Those skilled in the art will recognize that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software 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 implementations should not be considered beyond the scope of this application.

[0076] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0077] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0078] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0080] In the above embodiments, the functions of each functional unit can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. A computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0081] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0082] In summary, the following technical problems exist with the existing network selection solutions: (1) The static threshold network selection solution has a single network selection decision logic, which is based solely on the physical signal. This results in the core problem of "signal meeting the standard but poor experience". It cannot adapt to the complex interference and overlapping coverage scenarios of RedCap / LTE hybrid networking, and it leads to frequent occurrences of camping in weak coverage, high interference RedCap cells, and incorrect camping in overlapping coverage scenarios; (2) The fixed parameters of the terminal fixed parameter adaptive optimization solution have no adaptive capability. They can only solve common scenario problems in the industry and cannot adapt to the personalized spatiotemporal usage patterns and fixed camping scenarios of different users. This results in limited optimization effect and difficulty in eradicating individual problems. (3) The cloud-end-network collaborative AI network selection solution is highly dependent on the operator's network transformation and cloud computing power support. The deployment threshold is high and the implementation conditions are limited. It cannot achieve independent optimization of the terminal. At the same time, the model trained in the cloud is a general model for the whole network and cannot be adapted to the exclusive use scenario of a single user. The personalized optimization effect is poor. This application provides a multi-mode terminal autonomous network selection method, system, medium and terminal based on a lightweight machine learning model. It focuses on RedCap / LTE multi-mode terminal cell network selection and mobility management optimization. It collects physical measurement data and service experience data of multi-mode terminals based on a preset low power sampling mechanism, and analyzes the collected physical measurement data and service experience data. The data is preprocessed; then the preprocessed physical measurement data and service experience data are feature extracted to extract scene coverage features, time period features, network features and service experience features, and a multi-dimensional feature set is constructed accordingly; then, the constructed multi-dimensional feature set is input into a lightweight machine learning model for offline incremental training under low load periods, so as to train a cell experience scoring model for outputting experience scoring information of each candidate cell; finally, under the preset reasoning trigger scenario, the trained cell experience scoring model is called to perform reasoning, and the cell to be stationed is selected according to the experience scoring information generated by the cell experience scoring model, which has the following beneficial effects: (1) This application breaks through the single physical information The limitations of the judgment are addressed by taking actual business experience as the core basis for network selection, actively identifying and avoiding poor camping scenarios such as RedCap weak field, strong signal high interference, and overlapping coverage weak cell, solving the problem of "legal camping but poor experience" in traditional solutions, and greatly improving terminal data rate and latency stability, adapting to complex RedCap / LTE hybrid networking scenarios; (2) This application abandons the optimization logic of fixed unified parameters, has adaptive and personalized optimization capabilities, and can dynamically iterate network selection parameters and cell scoring rules according to the spatiotemporal usage patterns of a single user, accurately adapting to the personalized scenarios of different users such as home, office, and commuting, with more accurate optimization effect, wider coverage, and long-term adaptation to network and scenario changes;(3) This application does not rely on operator network upgrades, cloud computing power, or data transmission, making deployment extremely easy and practical. It also abandons the universal model used across the entire network, instead modeling for individual user scenarios, resulting in superior personalized optimization. Furthermore, all data is stored and processed locally, eliminating the risk of data privacy leaks. Terminal autonomy is enhanced, and computing power and power consumption are reduced. Therefore, this application effectively overcomes the various shortcomings of existing technologies and possesses high industrial application value.

[0083] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for multi-mode terminal autonomous network selection based on lightweight machine learning model, characterized in that, The method comprises: collecting physical measurement data and service experience data of a multi-mode terminal based on a preset low-power sampling mechanism, and preprocessing the collected physical measurement data and service experience data; extracting features from the preprocessed physical measurement data and service experience data to obtain scene coverage features, time period features, network features and service experience features, so as to construct a multi-dimensional feature set; inputting the constructed multi-dimensional feature set into a lightweight machine learning model to perform offline incremental training in a low-load period, so as to train a cell experience scoring model; the cell experience scoring model is used to output experience score information of each candidate cell; in a preset inference triggering scenario, calling the trained cell experience scoring model, and selecting a camping cell according to the experience score information generated by the cell experience scoring model.

2. The method of claim 1, wherein the method further comprises: The preset low-power sampling mechanism is: if it is detected that the multi-mode terminal is in an idle state, a low-frequency sampling mechanism is used to collect physical measurement data and service experience data; if it is detected that the network state of the multi-mode terminal changes or the multi-mode terminal is in a service running state, a high-frequency sampling mechanism is used to collect physical measurement data and service experience data. 3.The method of claim 1, wherein, The physical measurement data includes: time information, location information, reference signal received power, signal to interference plus noise ratio; the service experience data includes: block error rate, uplink and downlink rate, end-to-end delay data.

4. The method of claim 3, wherein the method further comprises: After the preprocessing step, it further comprises: converting the location information into grid encoding based on the Geohash algorithm, so as to aggregate physical measurement data and service experience data in the same spatial dimension.

5. The method of claim 1, wherein the method further comprises: The low-load period includes: charging period, screen-off period, terminal idle period.

6. The method of claim 1, wherein the method further comprises: The preset inference triggering scenario includes: initial cell selection scenario, cell reselection scenario, mobility switching scenario.

7. The method of claim 1, wherein the method further comprises: Further comprising: continuously collecting newly added physical measurement data and service experience data, and periodically iteratively training the trained cell experience scoring model according to the newly added physical measurement data and service experience data, so as to optimize the cell experience scoring model and network selection parameters.

8. A multi-mode terminal autonomous network selection system based on a lightweight machine learning model, characterized in that, The method comprises: a data collection module configured to collect physical measurement data and service experience data of a multi-mode terminal based on a preset low-power sampling method, and preprocess the collected physical measurement data and service experience data; a feature set construction module configured to extract features from the preprocessed physical measurement data and service experience data to obtain scene coverage features, time cycle features, network features and service experience features, so as to construct a multi-dimensional feature set; a model training module configured to input the constructed multi-dimensional feature set into a lightweight machine learning model to perform offline incremental training in a light-load period, so as to train a cell experience scoring model; the cell experience scoring model is configured to output experience score information of each candidate cell; a model inference module configured to, in a preset inference triggering scenario, call the trained cell experience scoring model to generate experience score information of each candidate cell, and select a camping cell according to the generated experience score information.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by a processor to implement the method of any one of claims 1 to 7. The computer program is executed by a processor to implement the method of any one of claims 8 to 14.

10. An electronic terminal comprising a memory, a processor and a computer program stored on the memory, characterized in that, The processor executes the computer program to implement the method of any one of claims 1 to 7. The processor executes the computer program to implement the method of any one of claims 1 to 7.

Images