Mobile communication management and control method and system based on service area division

Abstract

The application discloses a mobile communication management and control method and system based on service area division, and particularly relates to the technical field of source network load storage, wherein by identifying the synchronous mobile characteristics of user groups in real time and predicting signaling peaks, the fine grids in the gathering area are actively combined into large-scale service areas, and the unified degradation strategy is used to suppress cross-area signaling, only the stable resident users are gradually restored to fine management and control, and when the extreme overload occurs, the strategy execution is decentralized to the base station asynchronous backhaul, after the load is restored, the original strategy is recovered in batches and complete event logs are output, the dynamic balance between the area precision and the network stability is realized, and the precise differentiated service is ensured while avoiding the service interruption caused by the signaling impact.

Description

Technical Field

[0001] This invention relates to the field of mobile data communication technology, and more specifically, to a mobile communication management and control method and system based on service area division. Background Technology

[0002] In mobile data services, to meet the differentiated needs of various locations such as campuses and venues, networks need to perform precise control based on users' real-time locations. A common practice is to divide the coverage area into finely divided geographical grids, with each grid corresponding to an independent bandwidth or access policy. When a user's location changes and crosses the grid boundary, the network must immediately refresh the location and trigger the control plane to reconfigure the policy, causing signaling overhead to increase sharply with the granularity of the grid.

[0003] A more pronounced contradiction lies in the fact that when people move synchronously, a large number of users cross grid boundaries almost simultaneously, generating a massive surge in policy update requests that creates signaling spikes. However, this signaling overload prevents policies from taking effect promptly, causing user services to become stagnant. Furthermore, even when some users stabilize after their movement stops, the existing mechanism lacks the means to differentiate them, treating them alongside continuously moving users and further straining signaling resources. When the core network's processing capacity reaches its limit, network-wide control is lost, and while base stations possess local measurement capabilities, these are not fully utilized. These issues make it difficult for existing methods to achieve a balance between regional control accuracy and stable network operation. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a mobile communication management and control method and system based on service area division to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The mobile communication management and control method based on service area division includes the following steps:

[0007] The network coverage area is divided into multiple fine grids, and differentiated service strategies are configured for each grid.

[0008] Based on the user density within each grid, the frequency of cross-grid handover, and the core network signaling processing latency, the spatiotemporal clustering characteristics of user groups are identified, and the timing and intensity of signaling spikes are predicted.

[0009] When it is predicted that the signaling spike will cause the signaling load to exceed the processing capacity, adjacent fine grids within the spatiotemporal aggregation range are merged into a temporary large-scale service area, and a lenient policy is uniformly applied to users within the large-scale service area.

[0010] Within the large-scale service area, the movement trajectory of users between each fine grid is continuously tracked. Only for users who have stayed in the same fine grid for a longer period than a preset threshold, the differentiated strategy of applying fine grids is restored, while the relaxed strategy is maintained for other users.

[0011] If the signaling load is still trending towards overload, the policy matching and execution functions will be delegated to the base station, which will complete grid matching and policy implementation based on local measurement information and asynchronously transmit the execution records back to the core network.

[0012] Once the signaling load falls back to a safe threshold and remains stable, the large-scale service area is deactivated in batches, and the differentiated strategies of all fine grids are restored.

[0013] In a preferred embodiment, dividing the network coverage area into multiple fine grids and configuring differentiated service strategies for each grid specifically includes:

[0014] The coverage area is divided into square grids of equal size according to the preset side length as fine grids. User dwell records and business traffic data of each grid are collected in the historical period and statistically analyzed by time period. The business demand index of each grid under each time period is calculated. The business demand index is the weighted sum of the normalized value of the average number of users, the normalized value of the average business throughput, and the scene weight, each multiplied by the corresponding weighting coefficient.

[0015] Based on the business demand index, grids whose business demand index is higher than the first threshold for multiple consecutive statistical periods within the same time period are configured as high-security strategies, grids whose business demand index is lower than the second threshold for multiple consecutive statistical periods are configured as low-security strategies, and the remaining grids are configured as normal-security strategies.

[0016] In a preferred embodiment, the high-guarantee strategy includes setting guaranteed uplink and downlink rates for users within the grid and allowing access to all services; the low-guarantee strategy includes limiting the available bandwidth for video and file download services; and the normal-guarantee strategy includes not setting guaranteed uplink and downlink rates, not imposing additional restrictions on service access types, and scheduling according to the user's subscribed standard package rate.

[0017] In a preferred embodiment, identifying the spatiotemporal clustering characteristics of user groups and predicting the timing and intensity of signaling spikes specifically includes:

[0018] Using a preset statistical period as the time window, the system collects the real-time number of users in each grid, the total number of cross-grid handovers, and the average latency of signaling processing related to core network and policy updates.

[0019] Based on changes in the number of users, the grid is marked as inflow type and outflow type. Spatially adjacent inflow type and outflow type grids that meet the matching conditions of the number of users are paired as migration grid pairs. The migration grid pairs that have intersections are connected to form a connected set. When the total number of grids in the connected set exceeds the minimum cluster size, it is determined to be a spatiotemporal cluster area.

[0020] For the spatiotemporal clustering area, the signaling impact intensity is estimated by weighting and summing the total change in the number of users in the area, the total number of cross-grid handovers, and the increment of signaling processing delay relative to the non-clustering delay benchmark.

[0021] If the estimated value of the current time window exceeds the warning threshold, a signaling spike is predicted to occur in the next time window, and the impact intensity is determined by the ratio of the current estimated value to the maximum estimated value of historical spike events.

[0022] In a preferred embodiment, merging adjacent fine grids within the spatiotemporal aggregation range into a temporary large-scale service area, and uniformly applying a lenient strategy to users within the large-scale service area, specifically includes:

[0023] The impact intensity is obtained, and the occupancy ratio and total capacity of the signaling queue are updated in combination with the current strategy to calculate the signaling load estimate for the next time window.

[0024] When the estimated value exceeds the product of the total queue capacity and the safety carrying capacity coefficient, grid merging is triggered, and adjacent grids in and around the spatiotemporal cluster area that meet the migration association conditions are added to the grid set to be merged; the migration association conditions are that the number of bidirectional cross-grid switching between adjacent grids and grids in the core set in the current time window exceeds the preset migration association threshold.

[0025] The set of grids to be merged is merged into a temporary large-scale service area. Fine grid boundaries are no longer retained in the large-scale service area. A relaxed policy is uniformly configured for the entire area, and admission control with suppression duration is adopted for boundary switching events.

[0026] In a preferred embodiment, the restoration of the differentiated strategy applying the fine grid, while maintaining the lenient strategy for other users, specifically includes:

[0027] Record the original fine grid where each user is located at the time of merging as the reference grid, retain the original grid boundary data, maintain a continuous dwell timer for each user, and accumulate the timer when the user's current grid is consistent with the reference grid; otherwise, update the reference grid and clear it to zero.

[0028] Compare the continuous residence duration with the dynamically adjusted stable residence threshold;

[0029] Users who exceed the stable residency threshold are marked as stable users, and their corresponding differentiated policies are put into the waiting queue. Idle resources in the signaling queue are restored in batches according to the policy, with priority given to those with longer residency times.

[0030] In a preferred embodiment, delegating the policy matching and execution function to the base station specifically includes:

[0031] After existing users have resumed differentiated policies, if the occupancy rate of the policy update signaling queue continues to rise and exceeds the severe overload threshold, the core network will identify the base station whose service range covers a large-scale service area as the target base station.

[0032] The decentralization acceptance index is calculated based on the target base station's CPU utilization, available memory, and the number of effective policy entries. Base stations that meet the decentralization acceptance index are included in the decentralization set.

[0033] The core network pushes local policy execution packages, including grid boundaries, policy mappings, lists of recovered users, and threshold parameters, to base stations that can be delegated.

[0034] In a preferred embodiment, the differentiated strategy of gradually deactivating the large-scale service area and restoring all fine-scale raster cells specifically includes:

[0035] Once the estimated signaling load falls below the safety threshold and remains stable for a preset period, the process is deactivated.

[0036] All fine grids to be de-identified within the large-scale service area are sorted in descending order of the total number of times they are crossed during the aggregation period. Grid boundaries and differentiated policies are restored in batches. The policy instructions restored in each batch are distributed and inserted into the signaling queues of subsequent cycles. The number of insertions in each cycle does not exceed the preset proportion of the estimated free slots in that cycle.

[0037] In a preferred embodiment, after each batch of releases, the signaling queue occupancy ratio is monitored. If the signaling queue occupancy ratio exceeds the safe load factor, the release is paused, and frequently switched restored grids are rolled back to the relaxed strategy until the load stabilizes and continues.

[0038] If the policy execution function has been previously delegated to the base station, the function will be reclaimed after the first round of deregulation is completed and the signaling queue stabilizes. The base station will upload a policy status snapshot and non-returned records and clear the local engine. The core network will then resume centralized control accordingly.

[0039] Record the grid recovery time, number of affected users, and signaling queue changes throughout the entire removal process, and merge them with the base station execution records to form an event chain log from aggregation trigger to removal completion.

[0040] In a preferred embodiment, the mobile communication management and control system based on service area division includes a grid policy initialization module, a spike prediction module, an aggregation suppression module, a hierarchical recovery module, a local takeover module, and a batch rollback module.

[0041] The grid policy initialization module is used to divide the network coverage area into multiple fine grids and configure differentiated service policies for each grid.

[0042] The peak prediction module is used to identify the spatiotemporal clustering characteristics of user groups based on the user density within each grid, the frequency of cross-grid handover, and the core network signaling processing latency, and to predict the timing and intensity of signaling peaks.

[0043] The aggregation suppression module is used to merge adjacent fine grids within the spatiotemporal aggregation range into a temporary large-scale service area when it is predicted that the signaling spike will cause the signaling load to exceed the processing capacity, and to uniformly apply a lenient policy to users within the large-scale service area.

[0044] The graded recovery module is used to continuously track the user's movement trajectory between each fine grid within the large-scale service area. Only for users who have stayed in the same fine grid for a longer period than a preset threshold, the differentiated strategy of applying the fine grid is restored, while the relaxed strategy is maintained for other users.

[0045] The local takeover module is used to decentralize the policy matching and execution functions to the base station if the signaling load is still trending towards overload. The base station then completes grid matching and policy implementation based on local measurement information and asynchronously transmits the execution records back to the core network.

[0046] The batch rollback module is used to gradually release the large-scale service area and restore the differentiated strategies of all fine grids after the signaling load falls back to a safe threshold and remains stable.

[0047] The technical effects and advantages of this invention are as follows:

[0048] 1. This invention addresses the problem of massive signaling caused by the synchronous movement of people in fine-grid division of mobile data services, leading to platform overload and policy stagnation. By identifying the synchronous movement characteristics of user groups in real time and predicting signaling spikes, it proactively merges fine-grids within clustered areas into large-scale service areas and unifies the degradation strategy to suppress cross-area signaling. Fine-grained control is gradually restored only for stable users. In case of extreme overload, the policy execution is delegated to the base station for asynchronous backhaul. After the load stabilizes, the original policy is restored in batches and a complete event log is output. This achieves a dynamic balance between regional accuracy and network stability, ensuring accurate and differentiated services while avoiding service interruptions caused by signaling surges.

[0049] 2. This invention identifies the spatiotemporal aggregation characteristics formed by the synchronous movement of crowds by real-time monitoring of user density, handover frequency and signaling delay of each grid and predicts signaling spikes. Then, before overload occurs, it actively merges several adjacent fine grids into a temporary large-scale service area and applies a lenient policy of only guaranteeing basic communication to users in the area. This eliminates a large amount of policy update signaling caused by users frequently crossing the boundaries of fine grids from the source.

[0050] 3. This invention continuously tracks user movement trajectories within the aggregation area and gradually restores differentiated strategies only to users who have remained stably in the same fine grid for more than the dynamic adjustment threshold. This ensures that the service experience of inactive users is not continuously degraded and prevents secondary impacts caused by centralized restoration strategies when the signaling queue is still high. This achieves a graded restoration effect where stable users exit the aggregation in an orderly manner and mobile users continue to be controlled.

[0051] 4. When the signaling load continues to deteriorate, this invention decentralizes the policy matching and execution functions to the base station, and autonomously completes grid matching and service quality parameter adjustment using local measurement information of the base station. The entire process does not trigger core network policy update signaling, ensuring that user data service is not interrupted under the most severe overload conditions. At the same time, the core network maintains the final visibility of the global state by asynchronously batch transmitting execution records back.

[0052] 5. After the signaling load stabilizes, this invention restores the fine boundaries and differentiated strategies of the grid in batches according to the frequency of the grid being crossed during the aggregation. The strategy instructions for each batch of restoration are distributed and inserted into multiple statistical periods and gradually issued, avoiding the load rebound caused by centralized restoration. Finally, the complete event chain log from triggering aggregation to batch release is output. Attached Figure Description

[0053] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0054] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention;

[0055] Figure 2 This is a module connection diagram of the system in Embodiment 2 of the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Example 1: Figure 1The present invention provides a mobile communication management and control method based on service area division, comprising the following steps:

[0058] The network coverage area is divided into multiple fine grids, and differentiated service strategies are configured for each grid.

[0059] Based on the user density within each grid, the frequency of cross-grid handover, and the core network signaling processing latency, the spatiotemporal clustering characteristics of user groups are identified, and the timing and intensity of signaling spikes are predicted.

[0060] When a signaling spike is predicted to cause the signaling load to exceed the processing capacity, adjacent fine grids within the spatiotemporal aggregation range are merged into a temporary large-scale service area, and a relaxed policy to ensure basic communication is uniformly applied to users within the service area, so as to reduce cross-area signaling from the source.

[0061] Within a large service area, continuously track the user's movement trajectory between various fine grids. Only for users who have stayed in the same fine grid for more than a preset threshold, restore the application of the differentiated strategy for that fine grid, and maintain a lenient strategy for other users.

[0062] If the signaling load is still trending towards overload, the policy matching and execution functions will be delegated to the base station, which will then complete grid matching and policy implementation based on local measurement information and asynchronously transmit the execution records back to the core network.

[0063] Once the signaling load drops to a safe threshold and remains stable, large-scale service areas are deactivated in batches, all fine-grid differentiated policies are restored, and grid change and policy activation records are output to maintain service continuity.

[0064] In this embodiment of the invention, the network coverage area is divided into multiple fine grids, and a differentiated service strategy is configured for each grid, specifically including:

[0065] Obtain an electronic map of the network coverage area and the location of base stations, and divide the coverage area into equal-sized square grids according to preset side lengths, with each grid serving as a fine grid.

[0066] The preset side length is determined by: calculating the minimum coverage radius of each base station in the coverage area, taking the larger of half of the minimum coverage radius and the preset lower limit as the grid side length, so that the maximum number of users in a single grid does not exceed an independently manageable upper limit during busy hours. This upper limit is jointly determined by the minimum signaling processing interval required for single-user policy issuance and the policy effective delay requirement within the grid.

[0067] Collect user dwell records and traffic data for each grid within a historical period, statistically analyzed by time period. Calculate the business demand index for each grid for each time period. According to the following formula, we can obtain: ,in, The normalized value of the average number of users per grid cell during this period is obtained by dividing the average number of online users per grid cell during this period by the maximum average number of online users per grid cell during the same period. The normalized value of average traffic throughput is obtained by dividing the total downlink traffic of the grid in this time period by the maximum value of the total downlink traffic of all grids in the same time period. The scene weights are preset based on the main location types within the grid. For example, teaching buildings are assigned 0.8, dormitories 0.5, playgrounds 0.2, and roads 0.1. The higher the value, the higher the natural requirements for business assurance of that location. , , These are the corresponding weighting coefficients, and the sum of the three is 1;

[0068] The weighting coefficients are determined as follows: a batch of raster samples with known business assurance priority labels are selected, and the weighting coefficients for each sample raster are calculated for each time period. , , Substituting the values ​​into the business demand index formula, and using known priority labels as monitoring signals, logistic regression is used to determine the factor that minimizes the error between the business demand index and the priority ranking. , , The possible values ​​include:

[0069] Select a batch of grid samples with known business guarantee priority labels. The total number of samples should be no less than 20% of the product of the total number of grids in the coverage area and the time period, and should contain at least 500 grids to ensure that all types of scenarios are fully covered.

[0070] For the U and T values ​​of each sample, a sliding window spanning one statistical period before and after the original normalized values ​​is used to smooth the mean, eliminating spikes caused by instantaneous user movement or sudden traffic surges. For the current statistical period, the original normalized values ​​of three consecutive periods—the current period and one period before and after it—are taken, and their arithmetic mean is calculated as the smoothed value for the current period. If the current period is the first or last period and lacks data from the preceding or following period, the missing portion is filled with the value of the current period.

[0071] If no user stays in a grid during a continuous statistical period, then U and T for that period are both set to 0, and S still takes the preset scene weight of that grid.

[0072] The known method for obtaining service assurance priority tags is as follows: For each grid, if the percentage of times a user-reported network experience score is lower than the preset acceptable line exceeds the third threshold, or the abnormal session interruption rate recorded by the network side exceeds the fourth threshold within the historical statistical period of that grid, then that grid is marked as "requiring high assurance" for that period, and the tag value is... Otherwise, it is marked as "not requiring high protection", and the tag value is... .

[0073] Weighting coefficients , , The initial value is set to To satisfy the constraint that the sum is 1;

[0074] For each sample grid Smooth it , and original Substitute the values ​​into a linear combination: ;

[0075] The Sigmoid function is used to map the linear combination to a probability estimate that the raster belongs to the "high-security" category. : ;

[0076] Cross-entropy is used as the loss function to measure the difference between the probability estimate and the true label. The overall loss is the mean of all samples. In the formula, N is the total number of samples participating in the fitting. For the sample The tag value, It is the natural logarithm;

[0077] The gradient descent method is used to iteratively update the weight coefficients. In each iteration, the gradient of the loss function with respect to each weight coefficient is calculated, and the weight coefficients are updated along the negative gradient direction with a preset learning rate. The weight coefficients are then shifted towards the direction that satisfies... The plane projection is truncated to a weight coefficient of not less than 0.05 to ensure that each factor contributes.

[0078] The iteration stops if either of the following two conditions is met:

[0079] The relative change in the loss function value between two consecutive iterations Less than the convergence threshold And it has been iterated at least 5 times;

[0080] The total number of iterations reached the maximum of 500 iterations;

[0081] The coefficients at the stopping point or the historical best are taken as the final weighting coefficients. , , .

[0082] In one embodiment disclosed in this application, the third threshold is set to 30%, the fourth threshold is set to 5%, the gradient descent learning rate is set to 0.01, the maximum number of iterations is 500, and the convergence threshold is set to... Taking a grid sample in a campus setting as an example: In teaching building area A during weekday daytime hours, the average number of users (normalized value U=0.9), the business throughput (normalized value T=0.8), and the scene weight S=0.8 are all recorded. Historical low-experience ratings due to video stuttering and other issues account for 40%, exceeding the third threshold, and are therefore marked as "requiring high-level support." In road area B, during the same time period, U=0.2, T=0.15, S=0.1, and the low experience rating accounted for only 5%, which did not exceed the threshold. Therefore, it was marked as "High protection not required" and tagged with... Substituting the above two types of samples and N data points of the same type into the loss function for iterative solution, the loss converges after approximately 200 iterations, yielding the result. Use this set of coefficients to calculate the business demand index for region A. Region B It can be seen that the index of the high-demand grid is significantly higher than that of the low-demand grid, which is consistent with the label judgment.

[0083] Based on the business demand index, grids whose business demand index exceeds a preset first threshold for three or more consecutive statistical periods within the same time period are configured with a high guarantee policy, including setting guaranteed uplink and downlink rates for users in the grid and allowing access to all services; grids whose index falls below a preset second threshold for three or more consecutive statistical periods are configured with a low guarantee policy, including limiting the available bandwidth for video and file download services; the remaining grids are configured with a normal guarantee policy, including not setting guaranteed uplink and downlink rates, not imposing additional restrictions on service access types, and scheduling according to the user's subscribed standard package rate.

[0084] In one embodiment disclosed in this application, the statistical period is 15 minutes, the first threshold is 0.7, the second threshold is 0.3, and the weighting coefficients are obtained after fitting. , , Taking a smart campus scenario as an example, during weekday daytime hours, the teaching building has a high average number of users and the data traffic is mainly for downloading teaching videos and courseware. The calculated D value is usually maintained at around 0.85 for several consecutive 15-minute periods, and it is classified into the high-security grid to ensure low latency for teaching and student interaction. The adjacent campus road grid is mostly used by pedestrians who pass by quickly. The user dwell time is short and the traffic demand is sparse. The D value is mostly below 0.2 and also meets the continuity condition, falling into the low-security range. Only basic communication is maintained, and resources are reserved for areas that really need security.

[0085] In this embodiment of the invention, based on the user density within each grid, the frequency of cross-grid handover, and the core network signaling processing latency, the spatiotemporal clustering characteristics of user groups are identified, and the timing and intensity of signaling spikes are predicted. Specifically, this includes:

[0086] Using a preset statistical period as the time window, at the end of each time window, the real-time number of users in each grid, the total number of cross-grid handovers that occurred within that time window, and the average time delay of signaling processing related to policy updates in the core network are collected.

[0087] For each grid cell, calculate the change in the number of users between the current time window and the previous time window. Grid cells with an increase in the number of users and a change exceeding a preset threshold are marked as inflow grid cells, and grid cells with a decrease in the number of users and an absolute change exceeding a threshold are marked as outflow grid cells.

[0088] Starting with each outflow raster, search for the inflow raster that is spatially adjacent to it. When the difference between the number of users reduced by the outflow raster and the number of users increased by the inflow raster does not exceed the preset matching tolerance, the outflow raster and the inflow raster are paired as a migration raster pair.

[0089] Multiple migrating grid pairs with the same grid intersection are concatenated into a connected set. When the total number of grids in the connected set exceeds the preset minimum cluster size, it is determined to be a spatiotemporal cluster area covered by a synchronous moving group.

[0090] For each spatiotemporal cluster, the estimated signaling impact intensity of that region within the current time window is calculated using the following formula: ,in, This is the sum of the absolute values ​​of the changes in the number of users in all grid cells within the area; This represents the total number of cross-grid switching operations within this area; This represents the increment of the average time delay for core network signaling processing within the time window relative to the baseline value for time delay without clustered events. , , The preset weighting coefficients are 1 in total.

[0091] The preset weighting coefficients are determined as follows: extract multiple confirmed peak event time windows that triggered signaling overload from historical records, and extract the time windows within each event time window. , , The three indicators are calculated, and their average values ​​are used across all peak events. The average values ​​are then normalized using the ratio of the reciprocals of these average values ​​as the base. , , The values ​​of are chosen such that the three indicators contribute equally to E in typical peak events;

[0092] If the estimated signaling impact intensity of the current time window exceeds the warning threshold, the next time window will be predicted as the period when the signaling peak occurs, and the impact intensity of the signaling peak will be the ratio of the estimated signaling impact intensity of the current time window to the estimated maximum signaling impact intensity recorded in the historical peak events.

[0093] In one embodiment disclosed in this application, the preset statistical period is 1 minute, the number threshold is 20% of the average number of users per grid during busy hours, the matching tolerance is 1.5 times the number threshold, the minimum cluster size is 5 grids, the latency benchmark value for no clustering events is the average signaling processing latency of 30 consecutive statistical periods during the off-peak hours of the same day, and the warning threshold is 0.6. Taking a scenario during a break in a school campus as an example, if the number of users in the grids of several adjacent classrooms drops sharply by 30 to 40 within one minute, they are marked as outflow grids. At the same time, the number of users in the adjacent corridor grids increases by about 35, and they are marked as inflow grids. The difference between the number of users decreasing in each classroom and the number increasing in the corridor is within the matching tolerance range, forming multiple migration grid pairs. These migration grid pairs are connected in series with 7 grids because they share a corridor grid, which exceeds the minimum aggregation size of 5. The system determines that a synchronously moving group is migrating from the corridor towards the playground. Within this time window, ΔN is the sum of the absolute values ​​of the changes in the number of users in all grids in the set, H is the total number of cross-grid switching in the set, and ΔD is the difference between the current latency and the baseline value at dawn. After the three terms are weighted by preset weights, E=0.65 is calculated, which exceeds the warning threshold of 0.6. The system predicts that a signaling spike will occur in the next minute and determines the impact intensity based on the ratio of the historical maximum spike value to the current E value, triggering subsequent grid aggregation and strategy degradation measures in advance.

[0094] In this embodiment of the invention, when a predicted signaling spike will cause the signaling load to exceed the processing capacity, adjacent fine grids within the spatiotemporal aggregation range are merged into a temporary large-scale service area, and a relaxed policy to ensure basic communication is uniformly applied to users within this service area, so as to reduce cross-area signaling from the source. Specifically, this includes:

[0095] To obtain the predicted impact intensity of signaling spikes, and combined with the real-time queue depth of current core network policy update signaling, calculate the estimated signaling load for the next time window using the following formula: ,in, Update the ratio of the number of pending messages in the signaling queue to the total capacity of the queue for the current strategy; Impact strength; This represents the total capacity of the queue. The preset adjustment coefficient, Take 0.3, The value is set to 0.7 to make the estimated value more reflect the contribution of the upcoming impact; the estimated signaling load is compared with the safety threshold obtained by multiplying the total queue capacity by the preset safety bearing coefficient. When the estimated signaling load exceeds the safety threshold, it is determined that the signaling load will exceed the processing capacity, triggering grid merging; the safety bearing coefficient is set to 0.8 to reserve processing margin before the queue is close to overflow.

[0096] All fine grids within the spatiotemporal cluster are designated as the core set to be merged. For a fine grid located outside the spatiotemporal cluster but spatially adjacent to any grid in the core set, if the number of bidirectional cross-grid switching between the adjacent grid and a grid in the core set within the current time window exceeds a preset migration association threshold, the adjacent grid is added to the set of grids to be merged. The set of grids to be merged is then merged into a temporary large-scale service area. Fine grid boundaries are no longer retained within the large-scale service area. User movement within the area is considered as activity within the same service area, and no cross-area signaling is generated.

[0097] A lenient policy is configured for the large-scale service area, which includes: removing bandwidth restrictions on video and file download services; setting the uplink and downlink guaranteed rates for all users in the area to the minimum guaranteed values ​​that only meet the needs of instant messaging and web browsing; suspending the fine-grained service differentiation in the area; and matching all user data requests in the area through the same general rule, so that differentiated policies will no longer be reissued due to user movement.

[0098] For switching events between the boundary of a large-scale service area and unmerged fine grids on the periphery, a single-boundary access strategy with a suppression duration is adopted: when a user enters the large-scale service area from the outside, a relaxed policy is issued and a timer is started. Within the preset suppression duration, even if the user crosses the boundary again to exit and then enters, no new policy update signaling is triggered; when a user moves out of the area and stays in the same fine grid on the periphery for more than the suppression duration, the original policy corresponding to that periphery grid is restored; the suppression duration is an integer multiple of the statistical period.

[0099] In one embodiment disclosed in this application, the security bearing factor is 0.8, the minimum guaranteed value is 1 Mbps for downlink and 256 kbps for uplink, the migration association threshold is three times the average cross-grid handover rate within a single grid statistical period, and the suppression duration is three statistical periods, i.e., three minutes. Taking a campus break scenario as an example, the current queue occupancy ratio... Impact strength Total queue capacity Substituting into the formula, we get The queue's safe carrying capacity is , If the number does not exceed 8000, the system determines that the overload risk is within a controllable range and will not trigger a merger. If a gathering or assembly activity overlaps with the break time, and the impact intensity rises to 0.90, then... If the number of students exceeds the safe carrying capacity limit of 8000, the system determines that an overload risk has been established. It immediately merges the grids on the migration chain, such as classrooms, corridors, and stairwells, along with the grids of adjacent restrooms that frequently switch between them, into a temporary large-scale service area. The boundaries of the fine grids inside disappear, and students in the area no longer generate cross-area signaling during the entire movement process. The strategy is uniformly downgraded to the minimum guaranteed rate. The suppression mechanism at the boundary ensures that even if students linger at the edge of the service area, they will not repeatedly trigger strategy change signaling. After the signaling queue falls back to the safe level, the original fine grids and differentiated strategies are restored in batches.

[0100] In this embodiment of the invention, within a large-scale service area, the movement trajectory of users between various fine grids is continuously tracked. Only for users who have stayed in the same fine grid for a duration exceeding a preset threshold, the differentiated strategy for that fine grid is restored; for other users, a more lenient strategy is maintained. Specifically, this includes:

[0101] When merging a set of graticules into a temporary large-scale service area, the network side records the graticule identifier of the original fine-grained graticule where each user is located at the time of merging as the reference graticule for that user, and retains the geometric boundary data of the original fine-grained graticule for subsequent location determination; the large-scale service area only blocks the function of triggering policy updates by its internal boundary, and does not affect the network side's continuous calculation of the user's actual geographic coordinates based on the base station measurement report and its correspondence with the retained geometric boundary, thereby determining the original fine-grained graticule in which the user currently falls.

[0102] For each user within the large-scale service area, a continuous dwell timer is maintained. When the network side determines that the original fine grid that the user is currently falling into is consistent with its reference grid, the timer is incremented according to the measurement period; when it is determined that they are inconsistent, the reference grid is updated to the current grid, the timer is cleared and the timing restarts.

[0103] The continuous residence time is compared with a preset stable residence threshold, which is dynamically adjusted according to the following formula: ,in, The base dwell threshold is set to twice the average time it takes for a user to walk through a single fine grid cell. Update the ratio of the number of pending messages in the signaling queue to the total capacity of the queue for the current strategy; This is the load adjustment coefficient, which is positive. It automatically extends the stable dwell threshold when the signaling queue is still high, thus slowing down the policy recovery rhythm to avoid triggering secondary signaling impacts.

[0104] In one embodiment disclosed in this application, the load regulation coefficient The value is 1.5.

[0105] It should be noted that, The value of 1.5 is based on actual measurements and calibration of the relationship between the signaling queue emptying rate and the stable user dwell time. The determination process is as follows: During system trial operation, a signaling spike is artificially triggered, and the policy is updated to adjust the signaling queue occupancy ratio. The queue was emptied from a high level of approximately 0.7 to 0.8, after which all new policy request injections were stopped. The average time required for the queue to naturally empty from this high level to a safe threshold (e.g., below 0.3) was measured. Simultaneously, the average dwell time of users in the large-scale service area who completed cross-grid migration and re-stabilized through natural walking during this emptying period was recorded. After multiple tests, the average time for the queue to empty from 0.8 to 0.3 was approximately 4.5 minutes, while the average dwell time for users to re-stabilize was approximately 2.8 minutes. If the value is too small, then the dynamic threshold... Insufficient growth will lead to a large number of users triggering policy recovery before the queue is fully emptied, creating a secondary impact; if If the value is too large, the threshold will be too long, requiring users to wait an excessively long time before the differentiated service is restored after the migration ends, thus impacting the user experience. At that time, , The time taken is approximately 1 minute, which is basically matched with the time it takes to empty the queue. This allows sufficient time for emptying the queue without causing users to wait too long. In actual testing, it achieved a reasonable compromise between suppressing signaling shocks and restoring user experience.

[0106] When the continuous dwell time of any user exceeds the stable dwell time threshold, the user is identified as a stable user, and the differentiated policy corresponding to the original fine grid where the user dwells is added to the waiting queue. For stable users in the waiting queue, the network side does not immediately issue the policy at the time of measurement, but inserts the policy issuance instruction into the end of the policy update signaling queue and processes it as a regular signaling message, thereby reusing the queue's own congestion control mechanism.

[0107] The number of idle slots in the synchronous monitoring policy update signaling queue is taken as the sliding average of the number of idle slots in the most recent measurement periods as a stable reference value. In each measurement period, the number of policy recovery requests taken from the queue to be issued shall not exceed the preset proportion of the stable reference value, and the excess shall be carried over to the subsequent period for processing. When there are multiple stable users in the queue to be issued, they shall be sorted from longest to shortest continuous residence time, and users with longer residence time shall be given priority for recovery.

[0108] In one embodiment disclosed in this application, the measurement period is 15 seconds. Take 2 minutes. The value is set to 1.5, the preset ratio is 20%, and the sliding average window for idle slots is set to 4 measurement cycles, or 1 minute. Taking a combined campus break scenario as an example, a student stands in a certain original fine grid in the corridor, and the timer starts to accumulate. Calculate the dynamic threshold After 3.8 minutes, the student was marked as a stable user, and the high-security policy request for the corridor grid where the student resided was placed in the waiting queue. At the same time, the system detected that the policy update signaling queue had an average of 150 free slots in the last minute, so a maximum of 150 × 20% = 30 stable users could be restored in this measurement cycle. If the student was ranked high in the waiting queue, the differentiated policy would be restored in the current cycle, and the video service could be used normally. Other students who continued to move in the corridor could not reach the threshold because their timers were repeatedly reset due to the repeated changes in the baseline grid. They would continue to be subject to the lenient policy that only guaranteed instant communication. This would achieve a hierarchical recovery in which stable users exited the aggregation in an orderly manner and mobile users continued to be controlled within the large-scale service area.

[0109] In this embodiment of the invention, if the signaling load still tends to be overloaded, the policy matching and execution functions are delegated to the base station. The base station completes grid matching and policy implementation based on local measurement information and asynchronously transmits the execution records back to the core network. Specifically, this includes:

[0110] After some users in the large-scale service area have resumed the differentiated policy, the core network continuously monitors the occupancy ratio of the policy update signaling queue. If the occupancy ratio increases instead of decreasing within a set number of consecutive statistical periods, and the current value exceeds the preset severe overload threshold, it is determined that the signaling load is still trending towards overload, and the policy execution function is decentralized.

[0111] The core network uses the geographical boundary of the large-scale service area as a benchmark, queries all base stations in the network topology whose coverage areas intersect with the geographical boundary, and identifies them as the target base station set. It then requests the current local load status from each target base station, including the base station main control board CPU utilization, available memory capacity, and the number of effective policy rule entries. For each target base station, its decentralization admission index is calculated using the following formula. : ,in, For base stations The normalized value of CPU utilization is expressed as a percentage of the base station's rated CPU processing capacity. This is a normalized value for available memory capacity, expressed as a percentage of the total base station memory currently available. The normalized value of the number of effective policy rule entries is calculated as a percentage of the current number of loaded rule entries relative to the maximum entry capacity of the base station policy storage. , , These are preset non-negative weighting coefficients;

[0112] Preset non-negative weighting coefficients , , The method for determining it is as follows:

[0113] Extract all base station samples that have undertaken policy execution function decentralization tasks from historical operation and maintenance records. The total number of base station samples K extracted must be no less than 50. If insufficient, backtrack to earlier historical records until the quantity requirement is met. Read the CPU utilization rate, available memory capacity, and number of effective policy rule entries reported last time before the decentralization time from the operation and maintenance logs, and perform data preprocessing: First, invalid samples with missing CPU utilization rate or available memory capacity values ​​are removed; then, the CPU utilization rate, available memory capacity, and number of effective rule entries for all samples are normalized by maximum and minimum values, mapping the original values ​​to the [0,1] interval to obtain normalized values, including the normalized CPU utilization rate value. Normalized value of available memory Normalized value of the number of effective policy rule entries Simultaneously, based on whether the base station triggers automatic backoff or generates a performance alarm due to overload within a preset observation window after deployment, a successful acceptance label is set. If the above events do not occur, then ,otherwise ;

[0114] Weighting coefficients , , The initial value is set to be equally distributed, that is, each takes 1 / 3, and the constraint is satisfied. And all coefficients are not less than 0.1;

[0115] The three indicators for each sample are combined according to their weighting coefficients to form the decentralization acceptance index: ;

[0116] By using the Sigmoid function This is mapped to an estimated probability of successful acceptance by the base station after the transfer is completed: ;

[0117] Using cross-entropy as the loss function, the difference between the probability estimate and the true acceptance success label is measured, and the loss is averaged over all samples: In the formula, K is the total number of base station samples participating in the fitting. It is the natural logarithm;

[0118] With the goal of minimizing the loss function L, gradient descent is used to iteratively adjust... , , The learning rate is set to 0.01; after each iteration, the convergence condition is checked: when the change in loss between two adjacent iterations is less than 0.01, the convergence condition is checked. If the iteration count reaches the preset maximum of 500 iterations, the iteration stops; the set of weight coefficients with the smallest loss value at the time of stopping is selected as the final result; if the loss value is still decreasing when the maximum number of iterations is reached, the iteration continues from that point until the loss value change condition is triggered, or if it is still not met after 200 additional iterations, the iteration is forcibly stopped, and the current optimal weight is output.

[0119] In one embodiment disclosed in this application, the observation window is set to 15 minutes. In the maximum-minimum-maximum value normalization, the minimum CPU utilization rate is set to 0%, and the maximum is set to 100%. The minimum available memory capacity is set to 0 MB, and the maximum is set to the total memory configured for the base station. The minimum number of effective rule entries is set to 0, and the maximum is set to the maximum entry capacity of the base station policy storage, 1000 entries. The gradient descent learning rate is set to 0.01. Taking historical data from a campus scenario as an example, 63 base station deployment samples were extracted from the operation and maintenance records of the past 3 months. After removing 3 samples with missing data, 60 valid samples were retained, meeting the minimum sample size requirement. After normalization, sample 1... , , ,Label Sample 2 , , ,Label Sample 3 , , ,Label Substituting all 60 samples into the cross-entropy loss function, gradient descent iterations were started with initial weights. On the 187th iteration, the difference between two consecutive loss values ​​first became less than [a certain value]. Stop iterating and obtain the final weights. , , Backtesting the three typical samples using this set of weights yielded the following results: Base station 1 acceptance index A = 0.4 × 0.55 + 0.4 × 0.6 − 0.2 × 0.2 = 0.42, with a mapping probability of 0.60, consistent with the success label; Base station 2 acceptance index A = 0.4 × 0.2 + 0.4 × 0.25 − 0.2 × 0.7 = 0.04, with a mapping probability of 0.51, consistent with the failure label; Base station 3 acceptance index A = 0.4 × 0.45 + 0.4 × 0.5 − 0.2 × 0.35 = 0.31, with a mapping probability of 0.58, meeting expectations. Therefore, this set of weights can be directly used to calculate the acceptance index in subsequent decentralization decisions, eliminating the need for repeated training.

[0120] When the delegation acceptance index is greater than the acceptance threshold, the base station is included in the delegation set; if none of the target base stations meet the requirements, the delegation is abandoned and a warning is issued.

[0121] For base stations included in the decentralization set, the core network pushes a local policy execution package to them. The execution package includes: the geometric boundary and policy mapping table of the original fine grid, the current large-scale service area boundary and relaxed policy parameters, the list of users whose differentiated policies have been restored and their grid identifiers, the dynamic adjustment parameters of the stable stay threshold, and the queuing processing rules for users to be restored.

[0122] After receiving the data, the base station loads it into its local policy engine. Simultaneously, using the user's uplink probe reference signal received by the base station, it calculates the user's azimuth and distance relative to the base station antenna based on the signal angle of arrival and latency. This information, combined with the base station's own location, is then converted into the user's geographical coordinates.

[0123] The base station receives uplink detection reference signals sent by user equipment through a planar antenna array configured on the top of the antenna tower. The planar antenna array adopts an 8-element uniform circular array configuration. The 8 antenna elements are equally spaced on a circle with a radius of half the carrier wavelength. The arc spacing between adjacent elements corresponds to half the carrier wavelength. The array plane is installed parallel to the ground, and the center of the circle is located on the central axis of the antenna tower. Each element is connected to the same baseband processing unit through radio frequency cables of equal length to ensure phase consistency between channels.

[0124] The baseband processing unit extracts the phase value of each signal relative to the center of the array from the eight received signals, and fits the eight phase values ​​to a single incoming wave direction according to the circular array manifold vector to obtain the azimuth angle of the uplink detection reference signal arriving at the antenna array. The azimuth angle is measured clockwise with geographic true north as the reference, and the angle measurement accuracy is ±2 degrees under the condition of direct signal path and ±5 degrees in a typical urban environment with multipath reflection.

[0125] To suppress the interference of multipath effects on angle measurement, the baseband processing unit uses the channel estimation results of the uplink sounding reference signal on multiple consecutive subcarriers in the frequency domain. It distinguishes the propagation paths with different arrival times by the difference in phase slope between the subcarriers, retains the path with the earliest arrival time and signal strength higher than the preset noise threshold as the direct path, and only uses the phase information of the direct path to participate in the azimuth angle fitting, discarding the other delayed multipath components. When the uplink sounding reference signal is blocked and there is no identifiable direct path, the angle measurement is determined to be invalid, and the user's position is reverted to the previous valid positioning result or the default strategy of the large-scale service area.

[0126] The base station simultaneously extracts the time advance from the difference between the reception and transmission times of the uplink detection reference signal. The quantization accuracy of the time advance depends on the resolution of the baseband sampling clock. In this embodiment, the sampling clock frequency is 122.88MHz, and the minimum quantization step size of the time advance is approximately 8.14 nanoseconds, corresponding to a single-way distance measurement resolution of approximately 1.22 meters. The base station multiplies the time advance by the propagation speed of electromagnetic waves in air and divides it by two to obtain the radial distance between the user equipment and the center of the antenna array. The measurement accuracy of this distance is ±1.5 meters under direct line-of-sight conditions.

[0127] The base station obtains the geographical coordinates of the antenna array center from its own engineering parameter configuration, extends the radial distance forward along the direction line indicated by the azimuth angle to obtain the projection point of the user equipment on the horizontal plane, orthogonally decomposes the horizontal offset of the projection point relative to the antenna center in the latitude and longitude direction, and superimposes it on the geographical coordinates of the antenna center to obtain the user's current estimated geographical coordinates.

[0128] In one embodiment disclosed in this application, the carrier frequency is 3.5 GHz, which corresponds to a half wavelength of about 4.3 cm, and the radius of the circular array is 4.3 cm; the preset noise threshold is one-tenth of the peak power of the signal. Taking a single measurement as an example, the geographical coordinates of the base station antenna center are 31.2304 degrees north latitude and 121.4737 degrees east longitude. The measured azimuth is 35 degrees north of east, and the corresponding radial distance for the time advance is 156 meters. In an open scene without multipath, the azimuth measurement accuracy is ±2 degrees, and the distance measurement accuracy is ±1.5 meters. The calculated user coordinates are 31.2315 degrees north latitude and 121.4746 degrees east longitude, with a positioning error ellipse major axis of approximately 3.5 meters. In the same scene, when there is reflection from a building's glass curtain wall, the reflected path is identified from the channel estimation of 8 subcarriers as arriving approximately 15 nanoseconds later than the direct path and its signal strength is only one-quarter of that of the direct path. This reflected path is discarded, and only the phase of the direct path is used for fitting. The final azimuth calculation result is consistent with the scenario without multipath, and no additional angular deviation is caused by reflection.

[0129] By comparing the geographic coordinates with the locally stored fine-grained raster boundaries one by one, the containment of points within the polygon is determined, thus deriving the raster identifier where the user is currently located:

[0130] The base station reads the vertex coordinate sequence of the polygons that make up each grid from the locally stored fine-grained grid boundary data; for each grid polygon, it uses the ray casting method to determine whether the user's estimated geographic coordinates are located inside it.

[0131] A horizontal ray is drawn from the user's estimated geographic coordinates, with its direction fixed to due east. The number of valid intersections between this ray and each edge of the raster polygon is counted. To avoid ambiguity in the counting caused by the ray passing through vertices or coinciding with edges, the following rules are used for determination:

[0132] Rule 1: When a ray passes through a vertex of a polygon, the vertex is counted as a valid intersection only if its two adjacent sides are located on the top and bottom sides of the ray, respectively; if the two adjacent sides are located on the same side of the ray, they are not counted as an intersection.

[0133] Rule 2: When a ray completely coincides with an edge of a polygon, that edge is not counted as an intersection point; only the two endpoints of that edge are judged according to the vertex rules mentioned above.

[0134] The number of valid intersections is counted according to the above rules. If the number is odd, the user is determined to be inside the grid; if the number is even, the user is determined to be outside the grid. When all grids are traversed and a grid is determined to contain the point, traversal stops and the grid's identifier is used as the grid identifier of the user's current location. If the user's estimated geographic coordinates do not fall inside any fine-grained grid, the broad strategy of falling back to the large-scale service area continues to apply.

[0135] When a user's movement causes a change in the grid identifier, the base station performs a policy lookup locally: for users who are in a large service area but are not in the recovered list, a lenient policy is applied; for users in the recovered list whose grid matches, a corresponding differentiated policy is applied; and for the rest, the default policy is applied.

[0136] When implementing the policy, the base station modifies the quality of service parameters of the user's radio data bearer. For users who only need to retain the lenient policy, the base station actively lowers the quality of service level identifier of their radio bearer and sets the guaranteed rate parameter to the value corresponding to the lenient policy. For users who need to restore the differentiated policy, the base station restores the quality of service level identifier and guaranteed rate to the values ​​corresponding to the user's subscription or the grid policy. The above modifications are completed within the base station through the bearer reconfiguration process of the radio resource control message. This message is exchanged between the base station and the user via the air interface but does not require the core network to participate. During the bearer reconfiguration process, the user data plane remains connected without interruption.

[0137] The base station generates an execution record for each policy change and caches it locally. The record includes the user identifier, change time, policy version number before the change, policy version number after the change, and trigger reason. When the number of cached records reaches the batch transmission threshold or the time interval since the last transmission exceeds the preset time interval, the records are packaged and uploaded asynchronously through the management interface between the base station and the core network. This upload message is marked as a low-priority service at the transport layer and does not occupy the resources of the policy update signaling queue.

[0138] In one embodiment disclosed in this application, the severe overload threshold is set to 0.9, and the number of statistical periods for judging continuous increases is set to 3. , , The thresholds are set to 0.4, 0.4, and 0.2 respectively, the acceptance threshold is set to 0.5, the batch backhaul record threshold is set to 500 records, the time interval is set to 5 minutes, and the maximum entry capacity of the policy rule is set to 1000 records per base station. Continuing with the campus scenario, after the large-scale service area is established and partially restored, the signaling queue occupancy ratio increases from 0.6 to 0.92 and rises for three consecutive cycles, triggering decentralization; the core network selects four base stations whose serving cells cover the area, one of which has a CPU utilization rate of 35%. Memory availability is 75%. 300 rule entries have been loaded. Substituting into the formula, we get The system just reached the acceptance threshold and was selected as the execution base station. After the base station loaded the policy package, it used the uplink reference signal to autonomously calculate the user's location and complete grid matching and policy switching without initiating any policy update request to the core network. As a result, the core network signaling queue dropped significantly, and the execution record was sent back in batches every 5 minutes, thus ensuring service continuity under signaling overload.

[0139] In this embodiment of the invention, after the signaling load falls back to a safe threshold and remains stable, the large-scale service area is deactivated in batches, all fine-grid differentiated policies are restored, and grid change and policy activation records are output to maintain service continuity. Specifically, this includes:

[0140] When the estimated signaling load falls below the safety threshold and remains no higher than the safety threshold for a continuous set number of statistical periods, the signaling load is determined to have stabilized, and the release process is initiated.

[0141] All fine-grained grids to be removed within the large-scale service area are sorted in descending order of the total number of times they were crossed during the aggregation period. The total number of times they were crossed refers to the cumulative value of events in which a user moves from a grid to an adjacent grid during the duration of this aggregation round. This data is continuously extracted from base station measurement reports and collected by the network side under each grid identifier during the aggregation period. The grids are divided into multiple removal batches from front to back according to the sorting results. The number of grids in each batch does not exceed a preset proportion of the total number of grids to be removed, thereby ensuring that the grids with the highest signaling risk are removed from the aggregation first and undergo inspection.

[0142] For each batch of deactivation, perform the following operations: remove the grids in this batch from the large-scale service area boundary, restore their original fine-grained grid boundaries and corresponding differentiated service policies, and restore the normal handover triggering mechanism between the grid boundary and adjacent grids; for users currently residing in this batch of grids, instead of using a centralized distribution method, the policy restoration instructions required by each user are distributed and inserted into the policy update queues of multiple consecutive statistical periods. The number inserted in each period does not exceed the preset proportion of the estimated number of idle slots in that period, thereby avoiding a new round of signaling backlog caused by centralized distribution;

[0143] After each batch of deactivation, observe the change in the occupancy ratio of the policy update signaling queue within at least one complete statistical period. If the occupancy ratio does not exceed the safety capacity factor, continue with the next batch of deactivation. If the occupancy ratio rises again and exceeds the safety capacity factor, suspend the subsequent batches of deactivation. Select several grids with the highest number of cross-grid handovers from the restored grids, revert them to the relaxed policy state, and temporarily return them to the control scope of the large-scale service area. This allows users in these grids to apply the relaxed policy again to suppress the signaling generation. The system will be restored after the estimated signaling load falls back below the safety threshold.

[0144] If the policy matching and execution functions had been delegated to the base stations before this aggregation event, after the first round of batch removal is completed and the queue occupancy ratio continues to stabilize, a function recovery instruction will be sent to each executing base station. Upon receiving the instruction, the base station will upload a snapshot of the policy status of all current users and any execution records that have not been returned to the core network, and then clear its local policy engine. The core network will restore centralized policy control based on the uploaded status snapshot. For users whose current policies recorded in the status snapshot are inconsistent with the policies that should take effect on the core network side, the alignment will be gradually completed in subsequent statistical periods.

[0145] At the start of the release, the completion of each batch, and the completion of all releases, record the time point when each grid is restored from the lenient policy to the differentiated policy, the number of users involved, and the signaling queue occupancy ratio change curve during the process; after all user policies have been restored to the original fine grid differentiated policy, merge the above records with the execution records returned by the base station according to the timestamp to form a complete event chain log of this signaling spike event from triggering aggregation to completion of release.

[0146] Example 2: This example introduces a mobile communication management and control system based on service area division, such as... Figure 2 As shown, it includes a grid strategy initialization module, a spike prediction module, an aggregation suppression module, a hierarchical recovery module, a local takeover module, and a batch rollback module;

[0147] The grid policy initialization module is used to divide the network coverage area into multiple fine grids and configure differentiated service policies for each grid.

[0148] The peak prediction module is used to identify the spatiotemporal clustering characteristics of user groups based on the user density within each grid, the frequency of cross-grid handover, and the core network signaling processing latency, and to predict the timing and intensity of signaling peaks.

[0149] The aggregation suppression module is used to merge adjacent fine grids within the spatiotemporal aggregation range into a temporary large-scale service area when the predicted signaling spike will cause the signaling load to exceed the processing capacity, and uniformly apply a relaxed policy to ensure basic communication to users within the service area.

[0150] The tiered recovery module is used to continuously track the user's movement trajectory between various fine grids within a large-scale service area. Only for users who have stayed in the same fine grid for a longer period than a preset threshold, the differentiated strategy of applying that fine grid is restored, while the lenient strategy is maintained for other users.

[0151] The local takeover module is used to decentralize the policy matching and execution functions to the base station if the signaling load is still trending towards overload. The base station then completes grid matching and policy implementation based on local measurement information and asynchronously transmits the execution records back to the core network.

[0152] The batch rollback module is used to gradually remove large-scale service areas and restore the differentiated strategies of all fine grids after the signaling load falls back to a safe threshold and remains stable.

[0153] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0154] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are 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 or wireless (e.g., infrared, wireless, microwave, etc.) means. A 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 includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0155] It should be understood that in the various embodiments of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0156] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and method described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0157] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways.

[0158] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A mobile communication management and control method based on service area division, characterized in that: Includes the following steps: The network coverage area is divided into multiple fine grids, and differentiated service strategies are configured for each grid. Based on the user density within each grid, the frequency of cross-grid handover, and the core network signaling processing latency, the spatiotemporal clustering characteristics of user groups are identified, and the timing and intensity of signaling spikes are predicted. When it is predicted that the signaling spike will cause the signaling load to exceed the processing capacity, adjacent fine grids within the spatiotemporal aggregation range are merged into a temporary large-scale service area, and a lenient policy is uniformly applied to users within the large-scale service area. Within the large-scale service area, the movement trajectory of users between each fine grid is continuously tracked. Only for users who have stayed in the same fine grid for a longer period than a preset threshold, the differentiated strategy of applying fine grids is restored, while the relaxed strategy is maintained for other users. If the signaling load is still trending towards overload, the policy matching and execution functions will be delegated to the base station, which will complete grid matching and policy implementation based on local measurement information and asynchronously transmit the execution records back to the core network. Once the signaling load falls back to a safe threshold and remains stable, the large-scale service area is deactivated in batches, and the differentiated strategies of all fine grids are restored.

2. The mobile communication management and control method based on service area division according to claim 1, characterized in that: The process of dividing the network coverage area into multiple fine grids and configuring differentiated service strategies for each grid specifically includes: The coverage area is divided into square grids of equal size according to the preset side length as fine grids. User dwell records and business traffic data of each grid are collected in the historical period and statistically analyzed by time period. The business demand index of each grid under each time period is calculated. The business demand index is the weighted sum of the normalized value of the average number of users, the normalized value of the average business throughput, and the scene weight, each multiplied by the corresponding weighting coefficient. Based on the business demand index, grids whose business demand index is higher than the first threshold for multiple consecutive statistical periods within the same time period are configured as high-security strategies, grids whose business demand index is lower than the second threshold for multiple consecutive statistical periods are configured as low-security strategies, and the remaining grids are configured as normal-security strategies.

3. The mobile communication management and control method based on service area division according to claim 2, characterized in that: The high-security strategy includes setting guaranteed uplink and downlink rates for users within the grid and granting access to all services; the low-security strategy includes limiting the available bandwidth for video and file download services.

4. The mobile communication management and control method based on service area division according to claim 1, characterized in that: The method of identifying the spatiotemporal clustering characteristics of user groups and predicting the timing and intensity of signaling spikes specifically includes: Using a preset statistical period as the time window, the system collects the real-time number of users in each grid, the total number of cross-grid handovers, and the average latency of signaling processing related to core network and policy updates. Based on changes in the number of users, the grid is marked as inflow type and outflow type. Spatially adjacent inflow type and outflow type grids that meet the matching conditions of the number of users are paired as migration grid pairs. The migration grid pairs that have intersections are connected to form a connected set. When the total number of grids in the connected set exceeds the minimum cluster size, it is determined to be a spatiotemporal cluster area. For the spatiotemporal clustering area, the signaling impact intensity is estimated by weighting and summing the total change in the number of users in the area, the total number of cross-grid handovers, and the increment of signaling processing delay relative to the non-clustering delay benchmark. If the estimated value of the current time window exceeds the warning threshold, a signaling spike is predicted to occur in the next time window, and the impact intensity is determined by the ratio of the current estimated value to the maximum estimated value of historical spike events.

5. The mobile communication management and control method based on service area division according to claim 1, characterized in that: The step of merging adjacent fine grids within the spatiotemporal aggregation range into a temporary large-scale service area, and uniformly applying a lenient strategy to users within the large-scale service area, specifically includes: The impact intensity is obtained, and the occupancy ratio and total capacity of the signaling queue are updated in combination with the current strategy to calculate the signaling load estimate for the next time window. When the estimated value exceeds the product of the total queue capacity and the safety carrying capacity coefficient, grid merging is triggered, and adjacent grids in and around the spatiotemporal cluster area that meet the migration association conditions are added to the grid set to be merged; the migration association conditions are that the number of bidirectional cross-grid switching between adjacent grids and grids in the core set in the current time window exceeds the preset migration association threshold. The set of grids to be merged is merged into a temporary large-scale service area. Fine grid boundaries are no longer retained in the large-scale service area. A relaxed policy is uniformly configured for the entire area, and admission control with suppression duration is adopted for boundary switching events.

6. The mobile communication management and control method based on service area division according to claim 1, characterized in that: The recovery application uses a fine-grained grid differentiation strategy, while maintaining the lenient strategy for other users, specifically includes: Record the original fine grid where each user is located at the time of merging as the reference grid, retain the original grid boundary data, maintain a continuous dwell timer for each user, and accumulate the timer when the user's current grid is consistent with the reference grid; otherwise, update the reference grid and clear it to zero. Compare the continuous residence duration with the dynamically adjusted stable residence threshold; Users who exceed the stable residency threshold are marked as stable users, and their corresponding differentiated policies are put into the waiting queue. Idle resources in the signaling queue are restored in batches according to the policy, with priority given to those with longer residency times.

7. The mobile communication management and control method based on service area division according to claim 1, characterized in that: The decentralization of policy matching and execution functions to the base station specifically includes: After existing users have resumed differentiated policies, if the occupancy rate of the policy update signaling queue continues to rise and exceeds the severe overload threshold, the core network will identify the base station whose service range covers a large-scale service area as the target base station. The decentralization acceptance index is calculated based on the target base station's CPU utilization, available memory, and the number of effective policy entries. Base stations that meet the decentralization acceptance index are included in the decentralization set. The core network pushes local policy execution packages, including grid boundaries, policy mappings, lists of recovered users, and threshold parameters, to base stations that can be delegated.

8. The mobile communication management and control method based on service area division according to claim 1, characterized in that: The phased removal of the large-scale service area and restoration of all fine-scale raster differentiation strategies specifically includes: Once the estimated signaling load falls below the safety threshold and remains stable for a preset period, the process is deactivated. All fine grids to be de-identified within the large-scale service area are sorted in descending order of the total number of times they are crossed during the aggregation period. Grid boundaries and differentiated policies are restored in batches. The policy instructions restored in each batch are distributed and inserted into the signaling queues of subsequent cycles. The number of insertions in each cycle does not exceed the preset proportion of the estimated free slots in that cycle.

9. The mobile communication management and control method based on service area division according to claim 1, characterized in that: After each batch is deactivated, the signaling queue occupancy ratio is monitored. If the signaling queue occupancy ratio exceeds the safe load factor, deactivation is paused, and frequently switched restored grids are rolled back to the relaxed policy until the load stabilizes and continues. If the policy execution function has been previously delegated to the base station, the function will be reclaimed after the first round of deregulation is completed and the signaling queue stabilizes. The base station will upload a policy status snapshot and non-returned records and clear the local engine. The core network will then resume centralized control accordingly. Record the grid recovery time, number of affected users, and signaling queue changes throughout the entire removal process, and merge them with the base station execution records to form an event chain log from aggregation trigger to removal completion.

10. A mobile communication management and control system based on service area division, used to implement the mobile communication management and control method based on service area division as described in any one of claims 1-9, characterized in that: It includes a grid strategy initialization module, a spike prediction module, an aggregation suppression module, a hierarchical recovery module, a local takeover module, and a batch rollback module; The grid policy initialization module is used to divide the network coverage area into multiple fine grids and configure differentiated service policies for each grid. The peak prediction module is used to identify the spatiotemporal clustering characteristics of user groups based on the user density within each grid, the frequency of cross-grid handover, and the core network signaling processing latency, and to predict the timing and intensity of signaling peaks. The aggregation suppression module is used to merge adjacent fine grids within the spatiotemporal aggregation range into a temporary large-scale service area when the signaling spike is predicted to cause the signaling load to exceed the processing capacity, and to uniformly apply a lenient policy to users within the large-scale service area. The graded recovery module is used to continuously track the user's movement trajectory between each fine grid within the large-scale service area. Only for users who have stayed in the same fine grid for a longer period than a preset threshold, the differentiated strategy of applying the fine grid is restored, while the relaxed strategy is maintained for other users. The local takeover module is used to decentralize the policy matching and execution functions to the base station if the signaling load is still trending towards overload. The base station then completes grid matching and policy implementation based on local measurement information and asynchronously transmits the execution records back to the core network. The batch rollback module is used to gradually release the large-scale service area and restore the differentiated strategies of all fine grids after the signaling load falls back to a safe threshold and remains stable.

Images