Cell access optimization method and system, computer equipment and storage medium

Abstract

The invention provides a cell access optimization method and system, computer equipment and a storage medium, and relates to the technical field of communication, and the method comprises the steps: predicting the residence probability of a user in a corresponding candidate cell in each preset time period in a plurality of preset time periods in the future; obtaining a load cost value of each candidate cell; obtaining an access score of each candidate cell based on the residence probability and the load cost of each candidate cell, and outputting an access score sequence of the candidate cells; and screening out at least one selectable target cell from the candidate cells based on a preset screening rule, and screening out the selectable target cell with the highest access score based on the access score sequence of the candidate cells as a final target cell to be accessed by the user. According to the technical scheme provided by the invention, dynamic optimization of the access decision is realized through deep fusion of user behavior prediction and cell load perception, and continuity and stability of user experience are effectively guaranteed while the utilization rate of network resources is improved.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a cell access optimization method, a cell access optimization system, a computer device, and a computer-readable storage medium. Background Technology

[0002] With the rapid deployment of 5G (5th Generation Mobile Communication Technology) networks and the gradual advancement of the future 6G (6th Generation Mobile Communication Technology) communication architecture, cellular networks are exhibiting characteristics such as high cell density, frequent handovers, and coexistence of heterogeneous networks (macro cells, micro cells, and hotspot cells). Against this backdrop, user access optimization and cell selection strategies have become key technologies for ensuring user quality of experience (QoE) and network resource utilization.

[0003] Traditional cell access control strategies typically make decisions based on the following factors: radio signal strength (such as RSRP (Reference Signal Receiving Power) and RSSI (Received Signal Strength Indication)); channel quality indicators (such as SINR (Signal to Interference plus Noise Ratio) and CQI); current cell load (number of users and PRB (Physical Resource Block) utilization); and historical access experience rules (such as load balancing factor bias).

[0004] While these strategies can achieve load balancing and reasonable resource scheduling to a certain extent, they face the following challenges as user behavior becomes increasingly complex and network service scenarios diversify: users' movement trajectories have obvious periodicity and individuality, and static strategies cannot adapt to the dynamic evolution of user behavior; users access a cell briefly and then quickly leave, resulting in frequent handovers (Ping-Pong effect) that seriously affect user experience; most access algorithms only focus on a single perspective (load or channel quality), ignoring the deep coupling relationship between user spatial behavior and cell resources. Summary of the Invention

[0005] This invention was developed to at least partially address the technical problems of existing cell access strategies, such as their inability to adapt to the evolution of dynamic user behavior, their neglect of the coupling relationship between user spatial behavior and cell resources, and their impact on user experience.

[0006] According to one aspect of the present invention, a cell access optimization method is provided, comprising: Based on the user's historical trajectory and current location, predict the user's dwell probability in the corresponding candidate cell in each of the multiple preset time periods in the future, and output the dwell probability distribution matrix of the candidate cell; Based on the current load of each candidate cell and the historical load sequence of multiple past times, the load cost value of each candidate cell is obtained, and the load cost set of the candidate cells is output. The access score of each candidate cell is obtained based on the retention probability distribution matrix and load cost set of the candidate cells, and the access score sequence of the candidate cells is output; and, Based on preset filtering rules, at least one optional target cell is selected from each candidate cell to form an optional target cell set. Then, based on the access score sequence of the candidate cells, the optional target cell with the highest access score is selected from the optional target cell set as the final target cell for the user to access.

[0007] Optionally, the prediction of the user's dwell probability in the corresponding candidate cell during each of multiple preset time periods based on the user's historical trajectory and current location includes: The user's current location data is input into the trained user trajectory prediction model, which outputs the predicted trajectory points of the user at multiple future times. The user trajectory prediction model is pre-trained based on the user's historical trajectory data set. Spatial matching is performed between the predicted trajectory points of the user at multiple future times and the spatial distribution set of the cell boundary to obtain the cells corresponding to the user at multiple future times as candidate cells; The multiple future moments are sequentially mapped to multiple consecutive and non-overlapping preset time periods in chronological order. Based on the candidate cells corresponding to the user in each of the multiple preset time periods in the future, the user's dwell probability in the corresponding candidate cell in each preset time period in the multiple preset time periods is obtained.

[0008] Optionally, the user historical trajectory data set is defined as: ,in Let represent the i-th historical trajectory point of the user, where i ranges from 1 to m, and m represents the total number of historical trajectory points of the user. ,in For three-dimensional spatial position coordinates, For timestamps, The unique identifier for the cell the user was connected to at that time. For the user's motion status data; And / or, The set of spatial distributions of the cell boundary is defined as follows: ,in This represents the spatial attribute data of the j-th cell within a preset area, where j ranges from 1 to n, and n represents the total number of cells within the preset area. ,in Let j be the unique identifier of the j-th cell within the preset area. Let j be the physical cell ID of the j-th cell within the preset area. This refers to the cell coverage boundary area data of the j-th cell within a preset area. The coordinates of the center point or reference point of the j-th cell within the preset area. These are the antenna engineering parameters for the j-th cell within the preset area.

[0009] Optionally, the load cost set of the candidate cells is defined as: ,in Let $\frac{j}{j}$ represent the load cost value of the $j-th candidate cell, where $j$ ranges from 1 to $J$, and $J$ represents the total number of candidate cells. ,in Let be the current load value of the j-th candidate cell. Let be the historical load fluctuation index of the j-th candidate cell. , To adjust the weights.

[0010] Optionally, the historical load fluctuation index of the j-th candidate cell The following formula is used to calculate: ; in, Let be the historical load standard deviation of the j-th cell. Let j be the historical load sequence of the j-th cell, and ,in, The current time is represented by 1, and T′ represents the past T′ preset times. Let J be the historical average load of the j-th cell; , To adjust the weights.

[0011] Optionally, the access score sequence of the candidate cells is defined as: ,in This represents the access score of the j-th candidate cell, where j ranges from 1 to J, and J represents the total number of candidate cells. The following formula is used to calculate: ; Where 1 to T represent T preset time periods in the future; This represents the probability that a user will remain in the j-th candidate cell during the t-th preset time period in the future; This represents the load cost value of the j-th candidate cell.

[0012] Optionally, the set of selectable target cells is obtained by filtering under the following constraints: ; in, Represents the set of selectable target cells; This represents the j-th candidate cell; This represents the average RSRP value of the j-th candidate cell; This represents the average SINR value of the j-th candidate cell; The first threshold factor; It is the second threshold factor; And / or, The final target cell for the user to access is determined using the following formula: ; in, Indicates the final target cell that the user is to access; This represents the access score of the j-th candidate cell; This represents the set of selectable target cells.

[0013] According to another aspect of the present invention, a cell access optimization system is provided, comprising: The cell dwelling probability prediction module is set to predict the dwelling probability of a user in a corresponding candidate cell during multiple preset time periods in the future, based on the user's historical trajectory and current location, and output the dwelling probability distribution matrix of the candidate cells. The cell load cost calculation module is configured to obtain the load cost value of each candidate cell based on the current load of each candidate cell and the historical load sequence of multiple past times, and output the load cost set of the candidate cells. The cell access score calculation module is configured to obtain the access score of each candidate cell based on the residency probability distribution matrix and load cost set of the candidate cells, and output the access score sequence of the candidate cells; and, The cell filtering module is configured to select at least one optional target cell from each candidate cell based on preset filtering rules and form an optional target cell set. Then, based on the access score sequence of the candidate cells, the optional target cell with the highest access score is selected from the optional target cell set as the final target cell for the user to access.

[0014] According to another aspect of the present invention, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the aforementioned cell access optimization method.

[0015] According to another aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor performs the aforementioned cell access optimization method.

[0016] The technical solution provided by this invention may include the following beneficial effects: The cell access optimization method and system provided by this invention constructs a dynamic correlation model between user behavior intent and network resource status by introducing user historical trajectory and current location information, fundamentally breaking through the limitations of traditional strategies that only focus on single signal strength or load. By accurately predicting users' future dwell behavior and integrating real-time cell load for joint optimization, this method achieves personalized and dynamic adjustment of access decisions, effectively solving the adaptation problem between user behavior evolution and network resource scheduling, significantly improving the service continuity and stability of users during mobility, avoiding frequent switching and experience fluctuations caused by rigid strategies, and thus effectively improving user quality perception while ensuring efficient use of network resources.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description

[0018] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0019] Figure 1 This is a flowchart illustrating a cell access optimization method provided in an embodiment of the present invention. Figure 2 A flowchart illustrating another cell access optimization method provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a cell access optimization system provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the specific implementation methods of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific implementation methods described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a set order or sequence; furthermore, in the absence of conflict, the embodiments and features in the embodiments of this invention can be arbitrarily combined with each other. In the following description, the use of suffixes such as "module," "component," or "unit" to represent elements is only for the convenience of the description of this invention and has no inherent meaning. Therefore, "module," "component," or "unit" can be used interchangeably.

[0022] In related technologies, cell access strategies suffer from technical problems such as inability to adapt to dynamic network behavior evolution, neglect of the coupling relationship between user spatial behavior and cell resources, and impact on user experience. To address these issues, this invention provides an intelligent access optimization scheme that integrates user behavior prediction and network state awareness, achieving proactive and intelligent network-side scheduling. A detailed description is provided below through specific embodiments.

[0023] Figure 1 This is a flowchart illustrating a cell access optimization method provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps S101 to S104.

[0024] S101. Based on the user's historical trajectory and current location, predict the user's dwell probability in the corresponding candidate cell during each of the multiple preset time periods in the future, and output the dwell probability distribution matrix of the candidate cell.

[0025] S102. Based on the current load of each candidate cell and the historical load sequence of multiple past times, obtain the load cost value of each candidate cell and output the load cost set of the candidate cells.

[0026] S103. Based on the dwell probability distribution matrix and load cost set of the candidate cells, obtain the access score of each candidate cell and output the access score sequence of the candidate cells.

[0027] S104. Based on preset filtering rules, at least one optional target cell is selected from each candidate cell to form an optional target cell set. Then, based on the access score sequence of the candidate cells, the optional target cell with the highest access score is selected from the optional target cell set as the final target cell for the user to access.

[0028] In this embodiment, a deep coupling model between user behavior and network resources is constructed by fusing user historical trajectory and current location information. On the one hand, user trajectory prediction technology is used to dynamically capture user movement patterns, solving the problem that traditional static strategies cannot adapt to the dynamic evolution of user behavior. On the other hand, user dwell probability and cell load cost are jointly optimized, making up for the shortcomings of existing technologies that ignore the coupling relationship between user spatial behavior and cell resources. Ultimately, the target cell with the highest matching degree is selected for the user, effectively improving the access stability of the user during movement and significantly improving user quality perception and network resource utilization.

[0029] In one specific embodiment, step S101 specifically includes the following steps S1011 to S1014.

[0030] S1011. Input the user's current location data into the trained user trajectory prediction model and output the predicted trajectory points of the user at multiple future times. The user trajectory prediction model is pre-trained based on the user's historical trajectory data set.

[0031] In this step, the user historical trajectory data set refers to the sequence of spatiotemporal location information of a user or device over a past period, used to model their movement patterns, behavioral habits, path preferences, etc. The user historical trajectory data set includes multiple historical trajectory points of the user, where each historical trajectory point is a multi-dimensional vector including user coordinates, movement status, and connected cell data. Specifically, the data originates from periodic reports from the user terminal via the GPS positioning module, and is provided by the network side (such as gNB / eNB) through MR (measurement report) and the positioning platform.

[0032] The user's historical trajectory data set is used to train the user trajectory prediction model. The model is trained by learning the spatiotemporal movement patterns of users in different time periods in the past. The user trajectory prediction model preferably adopts an improved time-aware LSTM (Long Short-Term Memory) model.

[0033] S1012. Spatial matching is performed between the predicted trajectory points of the user at multiple future time points and the spatial distribution set of the cell boundary to obtain the cells corresponding to the user at multiple future time points as candidate cells.

[0034] In this step, the cell boundary spatial distribution set refers to the spatial geometric coverage area of ​​all cells in the network. It can be represented as a set containing multiple polygonal or polyhedral structures, used to determine which cell a user's location or trajectory point belongs to. The cell boundary spatial distribution set includes spatial attribute data of multiple cells, where the spatial attribute data of each cell is a multi-dimensional vector, including cell coordinates, coverage area, antenna engineering parameters, and other related data.

[0035] S1013. The multiple future moments are sequentially mapped to multiple consecutive and non-overlapping preset time periods in chronological order, and the user's dwell probability in the corresponding candidate cell in each preset time period is obtained based on the candidate cell corresponding to the user in each preset time period in the multiple future time periods.

[0036] S1014. Based on the user's dwell probability in the corresponding candidate cell during each of the multiple preset time periods in the future, output the dwell probability distribution matrix of the candidate cell.

[0037] In this step, the residence probability distribution matrix D is a J×T matrix, where each element... This represents the probability that a user will reside in the j-th candidate cell within the t-th preset time period in the future, where J represents the total number of candidate cells and T represents the total number of preset time periods in the future.

[0038] In this embodiment, the user's current location is input into a trained prediction model, which outputs trajectory points for multiple future time points. This is then spatially matched and mapped over time periods to generate a dwell probability distribution. This step achieves a precise conversion from the user's real-time location to future behavioral intentions, providing fine-grained behavioral prediction for subsequent access decisions and enhancing the foresight and personalization of the decision-making process.

[0039] In one specific implementation, in step S1011, the user historical trajectory data set is defined as: ,in Let represent the i-th historical trajectory point of the user, where i ranges from 1 to m, and m represents the total number of historical trajectory points of the user. ,in For three-dimensional spatial position coordinates, For timestamps, The unique identifier for the cell the user was connected to at that time. This is the user's motion status data.

[0040] In this embodiment, by structurally defining the user's historical trajectory data set, multi-dimensional information such as location coordinates, timestamps, connected cells, and movement status is clarified, providing high-quality input data for the trajectory prediction model and ensuring the accuracy and completeness of user movement pattern extraction.

[0041] In one specific implementation, in step S1012, the set of spatial distributions at the cell boundary is defined as: ,in This represents the spatial attribute data of the j-th cell within a preset area, where j ranges from 1 to n, and n represents the total number of cells within the preset area. ,in Let j be the unique identifier of the j-th cell within the preset area. Let j be the physical cell ID of the j-th cell within the preset area. This refers to the cell coverage boundary area data of the j-th cell within a preset area. The coordinates of the center point or reference point of the j-th cell within the preset area. These are the antenna engineering parameters for the j-th cell within the preset area.

[0042] Cell coverage boundary area data refers to the geometric information used to define the geographic coverage area of ​​a cell, determining which cell a user's location belongs to. Cell coverage boundary area data can be obtained by generating a cell coverage heatmap from MR (Measurement Report) data. Specifically, it can be a closed area formed by connecting a series of latitude and longitude coordinate points, representing the coverage boundary of the cell.

[0043] In this embodiment, by structurally defining the spatial distribution set of cell boundaries, cell identifiers, coverage boundaries, center points, and antenna parameters are uniformly encapsulated, providing accurate geometric and engineering references for subsequent spatial matching and candidate cell selection, thereby improving the accuracy of determining the relationship between user location and cell affiliation.

[0044] In one specific implementation, in step S102, the load cost set of candidate cells is defined as: ,in Let $\frac{j}{j}$ represent the load cost value of the $j-th candidate cell, where $j$ ranges from 1 to $J$, and $J$ represents the total number of candidate cells. ,in Let be the current load value of the j-th candidate cell. Let be the historical load fluctuation index of the j-th candidate cell. , To adjust the weights.

[0045] Specifically, the current load value of a candidate cell = the current PRB utilization rate of the candidate cell × the current number of users accessing the candidate cell.

[0046] In this embodiment, by defining a load cost set and introducing a weighted calculation of the current load and historical load fluctuation index, a comprehensive quantification of the resource stress and dynamic change trend of the cell is achieved. This provides a reliable basis for access decisions that reflects the real-time status of the network and helps to avoid accessing overloaded or drastically fluctuating cells.

[0047] In one specific implementation, the historical load fluctuation index of the j-th candidate cell The following formula is used to calculate: ; in, Let be the historical load standard deviation of the j-th cell. Let j be the historical load sequence of the j-th cell, and ,in, The current time is represented by 1, and T′ represents the past T′ preset times. Let J be the historical average load of the j-th cell; , To adjust the weights.

[0048] In this embodiment, a quantitative formula for the load fluctuation index is constructed by introducing the historical load standard deviation and the deviation of the current load from the mean. This index can effectively capture the load fluctuation characteristics of a cell, assisting access decisions in avoiding cells with drastic load fluctuations, thereby improving the service stability after user access.

[0049] In one specific implementation, in step S103, the access score sequence of the candidate cell is defined as: ,in This represents the access score of the j-th candidate cell, where j ranges from 1 to J, and J represents the total number of candidate cells. The following formula is used to calculate: ; Where 1 to T represent T preset time periods in the future; This represents the probability that a user will remain in the j-th candidate cell during the t-th preset time period in the future; This represents the load cost value of the j-th candidate cell.

[0050] In this embodiment, an access score function is constructed by weighted fusion of user dwell probability distribution and cell load cost. This score comprehensively reflects the degree of matching between user behavior intent and network resource status. A higher score indicates that the cell has good resource carrying capacity while meeting user dwell needs, providing a scientific ranking basis for access optimization.

[0051] In one specific implementation, in step S104, the selectable target cell set is obtained by filtering under the following constraints: ; in, Represents the set of selectable target cells; This represents the j-th candidate cell; This represents the average RSRP value of the j-th candidate cell; This represents the average SINR value of the j-th candidate cell; The first threshold factor; This is the second threshold factor.

[0052] In this embodiment, by introducing a dual-threshold screening mechanism of RSRP and SINR, cells with substandard signal quality are eliminated before access, ensuring basic communication quality after user access. This step serves as a pre-access guarantee, effectively avoiding the risk of handover failure or dropped calls due to poor signal strength.

[0053] In one specific implementation, in step S104, the final target cell for the user to access is selected using the following formula: ; in, Indicates the final target cell that the user is to access; This represents the access score of the j-th candidate cell; This represents the set of selectable target cells.

[0054] In this embodiment, by selecting the target cell with the highest access score from the set of cells that meet the signal quality standards, a triple collaborative optimization of user behavior intent, network resource status, and signal quality is achieved. The final selected target cell can maximize user dwell stability and network resource utilization efficiency while ensuring basic communication quality.

[0055] The cell access optimization method provided in this invention deeply integrates user behavior trajectories and network load status to construct a behavior-load coupling model, achieving personalized and dynamic cell access decisions. First, a time-aware LSTM model is used to predict the probability distribution of user dwell time in each candidate cell in the future, accurately capturing user movement patterns and behavioral intentions. Second, by combining real-time cell load and its historical fluctuation characteristics, the load cost of each cell is quantified, comprehensively evaluating the network resource status. Then, by constructing a coupling score function, user dwell probability and cell load cost are jointly optimized to generate an access optimization score, achieving deep matching between user behavior intentions and network resource status. Finally, the target cell with the highest score is selected from the set of cells with acceptable signal quality to complete access. This method effectively solves the problems of traditional access strategies failing to adapt to dynamic user behavior and neglecting the coupling relationship between user spatial behavior and cell resources. It significantly improves the accuracy of user access decisions, the efficiency of network resource utilization, and user experience, avoids the ping-pong effect caused by frequent handovers, and ensures service continuity and stability during user mobility.

[0056] Figure 2 This is a flowchart illustrating another cell access optimization method provided in an embodiment of the present invention. Figure 2 As shown, the method includes the following steps S201 to S204.

[0057] S201. User trajectory modeling and dwell probability prediction.

[0058] The purpose of this step is to construct a model of the probability distribution of a user's stay in each candidate cell over a future preset time period.

[0059] The input for this step is: User's current location data (GPS or network location) P0; User historical trajectory data set: ; Set of spatial distributions of community boundaries: .

[0060] "User historical trajectory data set" refers to the spatiotemporal location information sequence of a user or device over a period of time in the past, used to model their movement patterns, behavioral habits, path preferences, etc.

[0061] Each historical trajectory point p i It is a multidimensional vector, which can be defined as: .

[0062] The meanings of each parameter are shown in Table 1 below: Table 1

[0063] The above data comes from the terminal's periodic reporting via GPS positioning module and is provided by the network side (such as gNB / eNB) via TMR (Measurement Report) and positioning platform.

[0064] The purpose of this data is to train a time-aware LSTM model, extract user movement patterns in different time periods, and then predict the probability distribution of user stay in various candidate cells in multiple preset time periods in the future. It is the core input of the behavioral load coupling modeling step.

[0065] The “cell boundary spatial distribution set” refers to the spatial geometric coverage of all cells in the network. It can be represented as a set containing multiple polygonal or polyhedral structures and is used to determine which cell a user’s location or trajectory point belongs to.

[0066] Each cell c j The spatial attribute data is a multidimensional vector, which can be defined as: .

[0067] The meanings of each parameter are shown in Table 2 below: Table 2

[0068] Cell coverage boundary area data Polygon j The coverage heatmap of the cell can be obtained by reverse-engineering MR data.

[0069] Base station MR data (Measurement Report) is wireless environment measurement data collected and reported by user terminals (UEs) or base stations (eNodeBs) in wireless communication networks. It is primarily used for network optimization, coverage assessment, and fault diagnosis. MR data specifically includes: Serving cell metrics such as Reference Received Power (RSRP), Reference Received Quality (RSRQ), Signal-to-Interference-plus-Noise Ratio (SINR), and Timing Advance (Tadv); Neighbor cell information such as the RSRP, RSRQ, and Physical Cell Identifier (PCI) of neighboring cells; and Terminal status such as Transmit Power Headroom (PHR) and Buffer Status Report (BSR).

[0070] There are two ways to collect MR data: Terminal reporting: The terminal actively reports the measurement results to the base station on a periodic basis or triggered by events (such as handover failure or signal change), and then transmits them to the core network or data processing platform; Base station side collection: The base station generates measurement data, such as uplink interference power (RIP) and antenna angle of arrival (AOA), through the physical layer and radio resource management process.

[0071] Then, using the user's historical trajectory data set, an improved Time-Aware LSTM (TA-LSTM) model is employed to learn the user's spatiotemporal movement patterns.

[0072] In the user trajectory modeling stage, each historical trajectory point p of the user can be... i Mapped to specific cell c j Furthermore, in the subsequent dwell probability prediction stage, it can be determined which candidate cell each future trajectory point of a user belongs to.

[0073] In this step, the dwell probability prediction model outputs: For each candidate cell c j , j∈[1,J], predict the user's location in candidate cell c during the t-th preset time period in the future. j The probability of residence within Pr j (t), which in turn generates the residence probability distribution matrix D, a J×T matrix, and ,in .

[0074] Specifically, the relationship between the two data sets P and C is as follows: User historical trajectory data set P → Train LSTM → Predict user trajectory points at multiple future time points → Spatial matching with cell boundary spatial distribution set C → Obtain the dwell probability Pr of each candidate cell + user in each preset time period in the future. j (t).

[0075] For example, a user's activity patterns over the past day might include the following: p1=(22.548, 114.085, 0, 10:00, CellID=1001); p2=(22.549, 114.086, 0, 10:05, CellID=1001); p3=(22.555, 114.092, 0, 10:10, CellID=1002); ... ... The community boundaries are distributed as follows: c1:CellID=1001, covering Polygon A (red line area). c2:CellID=1002, covering Polygon B (blue line area). ... ... By analyzing p1~p m Sequence modeling can predict that a user is more likely to be in candidate cell c1 in the next 15 minutes.

[0076] S202. Quantification of cell load status.

[0077] The purpose of this step is to dynamically sense the load intensity and changing trend of each candidate cell.

[0078] The input for this step is: Current load of candidate cells: L j (t0), defined as: current PRB utilization rate × current number of access users. This product is used to comprehensively characterize the resource occupancy of the cell. Historical load sequence of candidate cells over the past T′ time steps: L j ={L j (t0-1), L j (t0-2),..., L j (t0-T′)}.

[0079] First, define the historical load fluctuation index for candidate cells: ; Where: std(L j ): Historical load standard deviation, measures volatility; mean(L j ): Historical load average; α, β: Adjustment coefficients (empirical settings or optimization solutions).

[0080] Redefine the candidate cell load cost function: ; γ and δ are adjustment weights to ensure that the volatility of the current load and the historical load are taken into account.

[0081] The output of this step is the set of load costs for candidate cells: .

[0082] S203. Construct a user behavior load coupling scoring function.

[0083] The purpose of this step is to calculate the access score for each candidate cell by comprehensively considering the user retention probability and the cell load cost.

[0084] The inputs for this step are: residency probability distribution matrix: D; candidate cell load cost set: Cost.

[0085] First, construct the user behavior load coupling scoring function: ; The numerator represents the user's presence in candidate cell c within a preset time period t. j The probability of possible dwell time, the denominator represents the access candidate cell c. j The higher the network cost score, the better for candidate cell c. j The higher the match between user behavior and network load.

[0086] The output of this step is: Candidate cell access score sequence: Score={Score1, Score2, ..., Score} J}

[0087] S204. Access Decision and Policy Execution.

[0088] The purpose of this step is to select the optimal access cell and connect to it.

[0089] The input for this step is: Candidate cell access score sequence: Score; Candidate cell RSRP sequence: RSRP={r1, r2, ...,r J}, where r j The average RSRP of candidate cell j; Candidate cell SINR sequence: SINR={s1, s2, ...,s J}, where s j The average SINR of candidate cell j.

[0090] Specifically, threshold factors τ1 (signal strength limit) and τ2 (SINR limit) are set to construct a set of selectable target cells C′: .

[0091] Select from .

[0092] The output of this step is: Output target cell number And user access to the target cell is completed through RRC (Radio Resource Control) signaling. .

[0093] The cell access optimization method provided in this invention optimizes cell access based on the dynamic coupling of user behavior trajectory and network load. Specifically, it constructs an improved time-aware LSTM model to predict the probability distribution of user dwell time in each candidate cell, and defines a coupling score function based on the real-time load and fluctuation of each candidate cell to dynamically select the optimal access cell. This achieves deep integration of user behavior intent and network resource status, effectively improving the accuracy of access decisions, the efficiency of network resource utilization, and user experience.

[0094] Figure 3 This is a schematic diagram of the structure of a cell access optimization system provided in an embodiment of the present invention. Figure 3 As shown, the system includes: a cell dwell probability prediction module 301, a cell load cost calculation module 302, a cell access score calculation module 303, and an access cell screening module 304.

[0095] The cell dwell probability prediction module 301 is configured to predict the dwell probability of the user in the corresponding candidate cell during multiple preset time periods based on the user's historical trajectory and current location, and output the dwell probability distribution matrix of the candidate cells; the cell load cost calculation module 302 is configured to obtain the load cost value of each candidate cell based on the current load of each candidate cell and the historical load sequence of multiple past times, and output the load cost set of the candidate cells; the cell access score calculation module 303 is configured to obtain the access score of each candidate cell based on the dwell probability distribution matrix and the load cost set of the candidate cells, and output the access score sequence of the candidate cells; the access cell filtering module 304 is configured to filter at least one optional target cell from each candidate cell based on preset filtering rules to form an optional target cell set, and then filter the optional target cell with the highest access score from the optional target cell set based on the access score sequence of the candidate cells as the final target cell for the user to access.

[0096] In one specific implementation, the cell dwell probability prediction module 301 includes: a user trajectory prediction unit, a candidate cell determination unit, and a dwell probability prediction unit.

[0097] The user trajectory prediction unit is configured to input the user's current location data into a trained user trajectory prediction model and output the predicted trajectory points of the user at multiple future time points. The user trajectory prediction model is pre-trained based on the user's historical trajectory data set. The candidate cell determination unit is configured to spatially match the predicted trajectory points of the user at multiple future time points with the spatial distribution set of cell boundaries to obtain the cells corresponding to the user at each of the multiple future time points as candidate cells. The dwell probability prediction unit is configured to sequentially map the multiple future time points to multiple consecutive and non-overlapping preset time periods in chronological order, and based on the candidate cells corresponding to the user at each of the multiple preset time periods, derive the dwell probability of the user at each preset time period in the corresponding candidate cell.

[0098] In one specific implementation, the user historical trajectory data set is defined as: ,in Let represent the i-th historical trajectory point of the user, where i ranges from 1 to m, and m represents the total number of historical trajectory points of the user. ,in For three-dimensional spatial position coordinates, For timestamps, The unique identifier for the cell the user was connected to at that time. This is the user's motion status data.

[0099] In one specific implementation, the set of spatial distributions of the cell boundary is defined as: ,in This represents the spatial attribute data of the j-th cell within a preset area, where j ranges from 1 to n, and n represents the total number of cells within the preset area. ,in Let j be the unique identifier of the j-th cell within the preset area. Let j be the physical cell ID of the j-th cell within the preset area. This refers to the cell coverage boundary area data of the j-th cell within a preset area. The coordinates of the center point or reference point of the j-th cell within the preset area. These are the antenna engineering parameters for the j-th cell within the preset area.

[0100] In one specific implementation, the load cost set of the candidate cells is defined as: ,in Let $\frac{j}{j}$ represent the load cost value of the $j-th candidate cell, where $j$ ranges from 1 to $J$, and $J$ represents the total number of candidate cells. ,in Let be the current load value of the j-th candidate cell. Let be the historical load fluctuation index of the j-th candidate cell. , To adjust the weights.

[0101] In one specific implementation, the historical load fluctuation index of the j-th candidate cell The following formula is used to calculate: ; in, Let be the historical load standard deviation of the j-th cell. Let j be the historical load sequence of the j-th cell, and ,in, The current time is represented by 1, and T′ represents the past T′ preset times. Let J be the historical average load of the j-th cell; , To adjust the weights.

[0102] In one specific implementation, the access score sequence of the candidate cells is defined as: ,in This represents the access score of the j-th candidate cell, where j ranges from 1 to J, and J represents the total number of candidate cells. The following formula is used to calculate: ; Where 1 to T represent T preset time periods in the future; This represents the probability that a user will remain in the j-th candidate cell during the t-th preset time period in the future; This represents the load cost value of the j-th candidate cell.

[0103] In one specific implementation, the set of selectable target cells is selected using the following constraints: ; in, Represents the set of selectable target cells; This represents the j-th candidate cell; This represents the average RSRP value of the j-th candidate cell; This represents the average SINR value of the j-th candidate cell; The first threshold factor; This is the second threshold factor.

[0104] In one specific implementation, the final target cell for the user to access is determined using the following formula: ; in, Indicates the final target cell that the user is to access; This represents the access score of the j-th candidate cell; This represents the set of selectable target cells.

[0105] The cell access optimization system provided in this invention deeply integrates user behavior trajectories and network load status to construct a behavior-load coupling model, achieving personalized and dynamic cell access decisions. First, a time-aware LSTM model is used to predict the probability distribution of user dwell time in candidate cells in the future, accurately capturing user movement patterns and behavioral intentions. Second, by combining real-time cell load and its historical fluctuation characteristics, the load cost of each cell is quantified, comprehensively assessing the network resource status. Then, by constructing a coupling score function, user dwell probability and cell load cost are jointly optimized to generate an access optimization score, achieving deep matching between user behavior intentions and network resource status. Finally, the target cell with the highest score is selected from the set of cells with acceptable signal quality to complete access. This method effectively solves the problems of traditional access strategies failing to adapt to dynamic user behavior and neglecting the coupling relationship between user spatial behavior and cell resources, significantly improving the accuracy of user access decisions, network resource utilization efficiency, and user experience, avoiding the ping-pong effect caused by frequent handovers, and ensuring service continuity and stability during user mobility.

[0106] Based on the same technical concept, embodiments of the present invention also provide a computer device, such as... Figure 4 As shown, the computer device includes a memory 401 and a processor 402. The memory 401 stores a computer program. When the processor 402 runs the computer program stored in the memory 401, the processor 402 executes the aforementioned cell access optimization method.

[0107] Based on the same technical concept, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, wherein when the computer program is executed by a processor, the processor executes the aforementioned cell access optimization method.

[0108] In summary, the cell access optimization method, system, computer equipment, and storage medium provided in this embodiment of the invention construct a behavior-load coupling model by integrating user trajectories and network load status. Specifically, the probability distribution of user dwell behavior is coupled with cell load fluctuations for calculation, dynamically generating cell access scores. This enables personalized and dynamic cell access decisions by introducing selectable target cell sets and threshold filtering mechanisms to ensure service quality. While improving user experience, it also optimizes the utilization of network resources.

[0109] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention 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; and 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 the present invention.

Claims

1. A method for optimizing cell access, characterized in that, include: Based on the user's historical trajectory and current location, predict the user's dwell probability in the corresponding candidate cell in each of the multiple preset time periods in the future, and output the dwell probability distribution matrix of the candidate cell; Based on the current load of each candidate cell and the historical load sequence of multiple past times, the load cost value of each candidate cell is obtained, and the load cost set of the candidate cells is output. The access score of each candidate cell is obtained based on the retention probability distribution matrix and load cost set of the candidate cells, and the access score sequence of the candidate cells is output. as well as, Based on preset filtering rules, at least one optional target cell is selected from each candidate cell to form an optional target cell set. Then, based on the access score sequence of the candidate cells, the optional target cell with the highest access score is selected from the optional target cell set as the final target cell for the user to access.

2. The method according to claim 1, characterized in that, The method of predicting the user's dwell probability in a corresponding candidate cell during multiple preset time periods based on the user's historical trajectory and current location includes: The user's current location data is input into the trained user trajectory prediction model, which outputs the predicted trajectory points of the user at multiple future times. The user trajectory prediction model is pre-trained based on the user's historical trajectory data set. Spatial matching is performed between the predicted trajectory points of the user at multiple future times and the spatial distribution set of the cell boundary to obtain the cells corresponding to the user at multiple future times as candidate cells; The multiple future moments are sequentially mapped to multiple consecutive and non-overlapping preset time periods in chronological order. Based on the candidate cells corresponding to the user in each of the multiple preset time periods in the future, the user's dwell probability in the corresponding candidate cell in each preset time period in the multiple preset time periods is obtained.

3. The method according to claim 2, characterized in that, The user historical trajectory data set is defined as follows: ,in Let represent the i-th historical trajectory point of the user, where i ranges from 1 to m, and m represents the total number of historical trajectory points of the user. ,in For three-dimensional spatial position coordinates, For timestamps, The unique identifier for the cell the user was connected to at that time. For the user's motion status data; And / or, The set of spatial distributions of the cell boundary is defined as follows: ,in This represents the spatial attribute data of the j-th cell within a preset area, where j ranges from 1 to n, and n represents the total number of cells within the preset area. ,in Let j be the unique identifier of the j-th cell within the preset area. Let j be the physical cell ID of the j-th cell within the preset area. This refers to the cell coverage boundary area data of the j-th cell within a preset area. The coordinates of the center point or reference point of the j-th cell within the preset area. These are the antenna engineering parameters for the j-th cell within the preset area.

4. The method according to claim 1, characterized in that, The load cost set of the candidate cells is defined as follows: ,in Let $\frac{j}{j}$ represent the load cost value of the $j-th candidate cell, where $j$ ranges from 1 to $J$, and $J$ represents the total number of candidate cells. ,in Let be the current load value of the j-th candidate cell. Let be the historical load fluctuation index of the j-th candidate cell. , To adjust the weights.

5. The method according to claim 4, characterized in that, The historical load fluctuation index of the j-th candidate cell The following formula is used to calculate: ； in, Let be the historical load standard deviation of the j-th cell. Let j be the historical load sequence of the j-th cell, and ,in, The current time is represented by 1, and T′ represents the past T′ preset times. Let J be the historical average load of the j-th cell; , To adjust the weights.

6. The method according to claim 1, characterized in that, The access score sequence of the candidate cells is defined as follows: ,in This represents the access score of the j-th candidate cell, where j ranges from 1 to J, and J represents the total number of candidate cells. The following formula is used to calculate: ； Where 1 to T represent T preset time periods in the future; This represents the probability that a user will remain in the j-th candidate cell during the t-th preset time period in the future; This represents the load cost value of the j-th candidate cell.

7. The method according to claim 1, characterized in that, The set of selectable target cells is obtained by filtering under the following constraints: ； in, Represents the set of selectable target cells; This represents the j-th candidate cell; This represents the average RSRP value of the j-th candidate cell; This represents the average SINR value of the j-th candidate cell; The first threshold factor; It is the second threshold factor; And / or, The final target cell for the user to access is determined using the following formula: ； in, Indicates the final target cell that the user is to access; This represents the access score of the j-th candidate cell; This represents the set of selectable target cells.

8. A cell access optimization system, characterized in that, include: The cell dwelling probability prediction module is set to predict the dwelling probability of a user in a corresponding candidate cell during multiple preset time periods in the future, based on the user's historical trajectory and current location, and output the dwelling probability distribution matrix of the candidate cells. The cell load cost calculation module is configured to obtain the load cost value of each candidate cell based on the current load of each candidate cell and the historical load sequence of multiple past times, and output the load cost set of the candidate cells. The cell access score calculation module is configured to obtain the access score of each candidate cell based on the residency probability distribution matrix and load cost set of the candidate cells, and output the access score sequence of the candidate cells; and, The access cell filtering module is configured to select at least one optional target cell from each candidate cell based on preset filtering rules and form an optional target cell set. Then, based on the access score sequence of the candidate cells, the optional target cell with the highest access score is selected from the optional target cell set as the final target cell for the user to access.

9. A computer device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the cell access optimization method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, the processor performs the cell access optimization method according to any one of claims 1 to 7.

Images