Communication network coverage quality monitoring method and system fused with edge computing

Abstract

The invention relates to a communication network coverage quality monitoring method fused with edge computing. The method comprises the following steps: deploying a plurality of edge computing nodes in a target monitoring area, and collecting multi-modal data from user equipment, a base station and an environment sensor in real time; performing preprocessing and feature extraction on the local multi-modal data by each edge computing node to generate a local coverage quality map, and constructing a global coverage quality map through inter-node collaboration; inputting the global atlas into a pre-trained deep learning anomaly detection model, identifying spatial distribution and severity levels of a coverage blind area, a weak coverage area and an interference area, and predicting evolution trends of the coverage blind area, the weak coverage area and the interference area; a targeted strategy including parameter adjustment, resource scheduling or network topology optimization is generated based on the identification and prediction result, and is issued to a corresponding network element for execution; and after execution, re-collecting a data evaluation effect, and feeding back to the model to realize incremental learning. According to the invention, the real-time performance of coverage quality monitoring is effectively improved, so that the quality of network coverage is improved.

Description

Technical Field

[0001] This invention relates to the technical field of communication networks, and in particular to a method and system for monitoring the coverage quality of communication networks that integrates edge computing. Background Technology

[0002] With the widespread deployment of 5G networks and the application of the Internet of Things (IoT), accurately and efficiently monitoring and optimizing network coverage quality in complex environments such as densely populated urban areas and large venues is crucial for ensuring user experience. Current technologies primarily rely on traditional drive-testing and fixed-point testing methods. These methods require professionals to travel to the site with equipment to collect data, which is not only time-consuming, labor-intensive, and costly, but also limits the spatial and temporal accuracy of the collected data. This makes it difficult to comprehensively and in real-time reflect dynamically changing network conditions, leading to delayed optimization responses and an inability to meet the need for rapid response to network faults or quality degradation. Consequently, network coverage quality is reduced, and the user's online experience is compromised, necessitating improvements. Summary of the Invention

[0003] To improve the quality of network coverage, this application provides a communication network coverage quality monitoring method and system that integrates edge computing.

[0004] Firstly, the above-mentioned inventive objective of this application is achieved through the following technical solution: A method for monitoring the coverage quality of a communication network integrating edge computing, the method comprising the following steps: Multiple edge computing nodes are deployed within the target monitoring area. These edge computing nodes collect multimodal data from user equipment, base stations, and environmental sensors in real time. The multimodal data includes reference signal received power and quality on the user side, load and channel state information on the base station side, and wireless channel impulse response characteristics on the environmental side. Each edge computing node preprocesses and extracts features from the collected local multimodal data, and generates a local coverage quality map based on this. Data is synchronized and fused through a secure communication link between edge nodes to construct a global coverage quality map. The global coverage quality map is input into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas, and interference areas, and to predict their evolution trend. Based on the anomaly identification and prediction results, targeted optimization strategies, including parameter adjustment, resource scheduling, or network topology optimization, are generated and distributed to the corresponding network elements for execution. After the optimization strategy is implemented, network performance data is collected again to evaluate the optimization effect, and the effect data is fed back to the anomaly detection model for incremental learning and adaptive optimization of the model.

[0005] By adopting the above technical solutions, a complete closed loop from perception, analysis, decision-making to optimization is constructed, transforming network operation and maintenance from a traditional model that relies on manual experience and has a delayed response to a new model that is highly automated, intelligent, and predictive. Its core value lies in the fact that through the deep integration and intelligent analysis of multi-source data, it achieves accurate location and trend prediction of network coverage problems, thereby enabling proactive optimization, greatly improving the ability to manage complex and dynamic network environments, ultimately ensuring the end-user's business experience, and reducing operating costs.

[0006] In a preferred embodiment, this application can be further configured as follows: In the step of deploying multiple edge computing nodes within the target monitoring area, and collecting multimodal data from user equipment, base stations, and environmental sensors in real time through the edge computing nodes, the collaborative collection of the multimodal data specifically includes the following steps: The edge computing node uses a built-in protocol parsing module to periodically collect MR measurement reports from user equipment and performance management counter data from the base station. It invokes a dedicated driver deployed on the edge computing node to control the connected environmental sensor array to collect multipath delay spread and Doppler frequency shift characteristics of the wireless channel; Based on unified timestamps and geographic information labels, the data from different sources are spatiotemporally aligned to form a structured data stream with spatiotemporal context.

[0007] In a preferred example, this application can be further configured as follows: Preprocessing and feature extraction of the collected local multimodal data at each edge computing node, generating a local coverage quality map based on this, and synchronizing and fusing data through a secure communication link between edge nodes to construct a global coverage quality map. The specific steps for constructing the distributed quality map include: Each edge computing node uses the Kriging spatial interpolation algorithm to convert discrete local measurement data into a continuous spatialized quality layer; Edge nodes exchange their respective local quality layer summary information based on digital signatures and secure tunnels; By using a federated learning framework, a global graph generation model is collaboratively trained while protecting the privacy of the original data. This model synthesizes a high-precision global coverage quality graph by aggregating the model parameter updates of each node rather than the original data.

[0008] In a preferred example, this application can be further configured such that the step of inputting the global coverage quality map into a pre-trained deep learning anomaly detection model, identifying the spatial distribution and severity level of coverage blind spots, weak coverage areas, and interference areas, and predicting their evolution trend includes the following steps. The deep learning anomaly detection model adopts a hybrid architecture combining graph convolutional networks and long short-term memory networks to simultaneously capture the spatial dependence and temporal dynamics of coverage quality. The model output includes the geographical boundaries of the anomalous area, the anomaly type, the confidence score, and the probability of the anomaly worsening within a specific future time period; Based on the pre-set Shapleyga interpretive analysis framework, the model's judgment criteria are attributed to causes, key network elements or environmental factors that lead to anomalies are located, and anomaly tracing is achieved.

[0009] In a preferred embodiment, this application can be further configured to: generate targeted optimization strategies, including parameter adjustment, resource scheduling, or network topology optimization, based on the anomaly identification and prediction results, and distribute the strategies to the corresponding network elements for execution. This includes the following steps: Construct a policy generator based on the combination of a rule engine and reinforcement learning, where the rule engine handles known and typical abnormal scenarios, and the reinforcement learning component explores the optimal policy for complex and novel anomalies; The generated optimization strategies include adjusting the antenna tilt angle and transmit power of specific base stations, modifying individual cell offsets, dynamically allocating spectrum resources, or guiding user equipment to perform inter-frequency / inter-system handover. The policy is converted into a network configuration command that can be issued through the northbound interface, and the success of the command execution is verified.

[0010] In a preferred example, this application can be further configured as follows: after the optimization strategy is executed, network performance data is re-collected, the optimization effect is evaluated, and the effect data is fed back to the anomaly detection model for incremental learning and adaptive optimization of the model. This step includes the following steps: Define a set of key performance indicators to quantify the optimization effect, including overall coverage compliance rate, switching success rate, and user-perceived speed. The KPI data before and after the strategy execution, the strategy itself, and its effect labels are used as a new training sample and stored in the experience replay pool. Periodically trigger the incremental learning process of the model, sample data from the experience replay pool to fine-tune the anomaly detection model and policy generator, and achieve continuous improvement in system performance.

[0011] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions: A communication network coverage quality monitoring device integrating edge computing, the device comprising: a multimodal data acquisition unit, used to deploy multiple edge computing nodes within a target monitoring area, and to acquire multimodal data from user equipment, base stations and environmental sensors in real time through the edge computing nodes; The local coverage quality map generation unit is used to preprocess and extract features from the collected local multimodal data at each edge computing node, and generate a local coverage quality map based on it. Data synchronization and fusion are carried out through a secure communication link between edge nodes to construct a global coverage quality map. The evolution trend prediction unit is used to input the global coverage quality map into a pre-trained deep learning anomaly detection model, identify the spatial distribution and severity level of coverage blind spots, weak coverage areas and interference areas, and predict their evolution trends. The optimization strategy generation unit is used to generate targeted optimization strategies, including parameter adjustment, resource scheduling or network topology optimization, based on the anomaly identification and prediction results, and to distribute the strategies to the corresponding network elements for execution. The optimization effect evaluation unit is used to re-collect network performance data after the optimization strategy is executed, evaluate the optimization effect, and feed the effect data back to the anomaly detection model for incremental learning and adaptive optimization of the model.

[0012] Thirdly, the above-mentioned objectives of this application are achieved through the following technical solutions: An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described communication network coverage quality monitoring method for converged edge computing.

[0013] Fourthly, the above-mentioned objectives of this application are achieved through the following technical solutions: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described communication network coverage quality monitoring method for converged edge computing. Attached Figure Description

[0014] Figure 1 This is a flowchart of a communication network coverage quality monitoring method integrating edge computing according to an embodiment of this application; Figure 2 This is a schematic diagram of a communication network coverage quality monitoring device integrating edge computing according to one embodiment of this application; Figure 3 This is a schematic diagram of an electronic device according to an embodiment of this application.

[0015] Icon labels: 1. Multimodal data acquisition unit; 2. Local coverage quality map generation unit; 3. Evolution trend prediction unit; 4. Optimization strategy generation unit; 5. Optimization effect evaluation unit. Detailed Implementation

[0016] The present application will be further described in detail below with reference to the accompanying drawings.

[0017] In one embodiment, such as Figure 1 As shown, this application discloses a method for monitoring the coverage quality of communication networks that integrates edge computing, specifically including the following steps: S10: Multiple edge computing nodes are deployed within the target monitoring area to collect multimodal data from user equipment, base stations, and environmental sensors in real time through the edge computing nodes; The multimodal data includes the reference signal received power and quality on the user side, the load and channel state information on the base station side, and the wireless channel impulse response characteristics on the environmental side.

[0018] S20: Each edge computing node preprocesses and extracts features from the collected local multimodal data, and generates a local coverage quality map based on this. Data is synchronized and fused through a secure communication link between edge nodes to construct a global coverage quality map. S30: Input the global coverage quality map into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas and interference areas, and predict their evolution trend; S40: Based on the anomaly identification and prediction results, generate targeted optimization strategies including parameter adjustment, resource scheduling, or network topology optimization, and distribute the strategies to the corresponding network elements for execution; S50: After the optimization strategy is executed, network performance data is collected again, the optimization effect is evaluated, and the effect data is fed back to the anomaly detection model for incremental learning and adaptive optimization of the model.

[0019] Regarding steps S10-S50, the embodiment of this application is as follows: When optimizing the network of a large smart stadium, the system first collects uplink data of video streams from spectators' mobile phones (such as reference signal received power RSRP), wireless resource utilization of small base stations in the stadium, and wireless channel impulse response characteristics captured by specially deployed radio frequency sensors in real time through edge computing nodes deployed in various areas of the stadium in step S10. These multimodal data together constitute the basis of network status awareness.

[0020] In step S20: Each edge node independently processes the data for its own floor, using a spatial interpolation algorithm to generate a local coverage quality heatmap from discrete signal strength measurements. Subsequently, these local heatmaps are fused through secure inter-edge node communication links, ultimately synthesizing a three-dimensional global coverage quality map covering the entire venue at the central node, thereby accurately locating potential weak coverage points in areas such as grandstand corners or entrance passages.

[0021] In step S30, the system inputs the aforementioned global map into a pre-trained deep learning model (e.g., a hybrid model combining graph convolutional networks and long short-term memory networks) for analysis. This model can not only identify existing coverage blind spots, but also predict, based on historical data, which areas will face the risk of service quality degradation due to sudden capacity demands during peak periods such as after concerts.

[0022] In step S40, the system automatically generates and executes targeted optimization strategies. For example, for a predicted high-load area during the scattering phase, the system can automatically adjust the beamforming parameters of nearby distributed antenna systems and temporarily allocate additional spectrum resources to the cell to cope with the instantaneous traffic surge.

[0023] In step S50, the effect evaluation is immediately initiated, key performance indicators are re-collected, and the optimization effect is fed back to the deep learning model for incremental learning, so that the model can make more accurate and faster decisions when dealing with similar scenarios in the future.

[0024] Overall, this solution constructs a complete closed loop from perception, analysis, decision-making to optimization, transforming network operation and maintenance from a traditional model that relies on manual experience and has a delayed response to a highly automated, intelligent, and predictive new model. Its core value lies in achieving accurate location and trend prediction of network coverage problems through deep fusion and intelligent analysis of multi-source data, thereby enabling proactive optimization, greatly improving the ability to manage complex and dynamic network environments, ultimately ensuring the end-user's business experience, and reducing operating costs.

[0025] In step S10: Deploying multiple edge computing nodes within the target monitoring area, and using these edge computing nodes to collect multimodal data from user equipment, base stations, and environmental sensors in real time, the collaborative collection of this multimodal data specifically includes the following steps: S11: Periodically collect MR measurement reports from user equipment and performance management counter data from base stations through the protocol parsing module built into the edge computing node; Specifically, in this embodiment, an edge computing node with a built-in protocol parsing module is deployed beneath the catwalk on each floor of a large intelligent stadium. This module periodically collects MR measurement reports from the Radio Resource Management Protocol (RRM) via the base station's southbound interface, obtaining the reference signal received power and signal quality reported by spectator terminals. Simultaneously, it reads cell load, scheduling success rate, and number of wireless link failures from the base station's performance management counter. This module supports parallel parsing of multiple protocols, uniformly converting heterogeneous data into a standard format and caching it in local memory. This avoids original packet fragments occupying network bandwidth for extended periods, significantly reducing backhaul pressure and improving the real-time performance and consistency of data acquisition.

[0026] S12: Call the dedicated driver deployed on the edge computing node to control the connected environmental sensor array to collect the multipath delay spread and Doppler frequency shift characteristics of the wireless channel; At the edge node suspended in the central ceiling of the stadium, a dedicated driver program controls the surrounding array of environmental sensors: RF fingerprint sensors are activated to emit narrow-pulse detection signals, capturing the multipath delay spread distribution of the wireless channel in the stands; simultaneously, Doppler frequency shift measurement units are activated to detect minute frequency shifts caused by spectator movement. The collected channel characteristic data is directly processed within the edge node using Fast Fourier Transform (FFT) calculations and feature extraction to generate a list of multipath components of the channel impulse response. The entire process eliminates the need to upload the original sampled waveforms, effectively reducing data transmission volume and improving environmental perception accuracy.

[0027] S13: Based on unified timestamps and geographic information labels, perform spatiotemporal alignment on the data from different sources to form a structured data stream with spatiotemporal context; Specifically, upon receiving MR reports, base station counter data, and channel characteristic data, edge nodes immediately invoke a unified timestamp service (based on the PTP protocol) to add nanosecond-level synchronization tags to the three types of data. Simultaneously, combining the geographic grid coding of the electronic map, each data point is appended with its geographical coordinates indicating the viewing area or channel it resides in. Through spatiotemporal alignment, the originally isolated heterogeneous data streams are organized into a structured data stream with "time-space-modality" as the key, and stored in a local time-series database. This structured data stream allows subsequent coverage quality map construction to be directly retrieved using spatiotemporal key values, eliminating misjudgments caused by time deviations or location ambiguities, and significantly improving map accuracy and the availability of model training data.

[0028] In step S20: Each edge computing node preprocesses and extracts features from the collected local multimodal data, generates a local coverage quality map based on this, and synchronizes and fuses the data through a secure communication link between edge nodes to construct a global coverage quality map. The specific steps for constructing the distributed quality map include: S21: Each edge computing node uses the Kriging spatial interpolation algorithm to convert discrete local measurement data into a continuous spatialized quality layer; Specifically, at the edge nodes on the third floor of the smart stadium, after collecting reference signal received power values ​​from discrete locations such as seats in the stands, corridor corners, and restroom entrances, the system uses the Kriging spatial interpolation algorithm to fit the discrete measurement points into a continuous surface based on the signal covariance function with distance, generating a signal quality layer covering the entire three-story space. For example, the signal strength of areas in the back rows of the stands that were not directly measured can be estimated using weighted interpolation of surrounding sampling points, forming a color-gradient visual quality map. This method extends sparse sampling to full-field coverage estimation, reducing sensor deployment density while ensuring the layer is spatially smooth and physically meaningful, providing a unified benchmark for subsequent fusion.

[0029] S22: Edge nodes exchange their respective local quality layer summary information based on digital signatures and secure tunnels; Specifically, after generating their local quality layers, each edge node first hashes the layer and attaches a digital signature to it. Then, it sends the hash information to the central aggregation node through a pre-established TLS secure tunnel. For example, after the hashes sent by the three layer nodes are verified as having a trustworthy source, the central node compares the differences between the hashes from each layer and only requests encrypted layer fragments of areas with significant differences, rather than the entire original data. This mechanism implements authentication and data integrity protection at the transport layer, preventing layers from being tampered with or eavesdropped on during the exchange process. Simultaneously, hash comparison reduces invalid data transmission and improves fusion efficiency.

[0030] S23: Using a federated learning framework, a global graph generation model is collaboratively trained while protecting the privacy of the original data. This model synthesizes a high-precision global coverage quality graph by aggregating the model parameter updates of each node rather than the original data. Specifically, in this embodiment, a central aggregation node initiates the federated learning framework: each floor's edge node trains a lightweight map generation sub-model locally using its own data, uploading only the model gradient updates (rather than the original measurement data) to the center; the central node performs a weighted average of the gradients from each floor, aggregating the global map generation model parameters, and then distributes them to each node to update its local model. For example, the dense crowd features of the grandstand on the third floor and the open features of the entrance on the first floor are learned separately in the local model. The resulting global model can simultaneously grasp the signal propagation patterns of both high-density and low-density scenes, and the accuracy of the final synthesized global coverage quality map is significantly higher than that generated independently by a single node. This mode achieves cross-domain knowledge sharing while protecting the data privacy of each floor, improving the global consistency and generalization ability of the map.

[0031] In step S30: Inputting the global coverage quality map into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas, and interference areas, and predicting their evolution trend, the steps include the following: S31: The deep learning anomaly detection model adopts a hybrid architecture combining graph convolutional networks and long short-term memory networks to simultaneously capture the spatial dependence and temporal dynamics of coverage quality. Specifically, after the global coverage quality map of a large intelligent stadium is constructed, the system inputs this map into a pre-trained deep learning anomaly detection model. The model adopts a hybrid architecture of graph convolutional networks and long short-term memory networks: the graph convolutional network models geographical areas such as stadium stands, passageways, and entrances / exits as graph nodes, and the edge weights between nodes represent the spatial correlation of signal propagation, effectively capturing the mutual influence of coverage quality in adjacent areas; the long short-term memory network processes the map sequence of multiple consecutive time periods along the time axis, learning dynamic patterns such as crowd flow and event breaks. For example, during the climax of a concert, the model uses graph convolution to identify the weak coverage clusters in the stands behind the main stage due to structural obstruction, while simultaneously using the long short-term memory network to predict that the outflow of people along the passageways when the concert ends will further deteriorate the signal quality in that area, thus accurately achieving dual perception of spatial dependence and temporal dynamics.

[0032] S32: The model output includes the geographical boundaries of the anomalous area, the anomaly type, the confidence score, and the probability of the anomaly worsening in a specific future period. Specifically, for detected abnormal areas, the model outputs structured judgment results: the geographical boundary is precisely delineated using polygon coordinates on an electronic map to define the coverage blind zone; the anomaly type is labeled as "weak coverage" or "high interference"; the confidence score reflects the reliability of the judgment; and the probability of the anomaly worsening within a specific future time period (such as 15 minutes after the concert ends) is given. For example, the system outputs "Section 5 of the east stand is a weak coverage area with a high confidence level, and the probability of significant deterioration is predicted during the concert's closing time." This information directly drives the generation of subsequent optimization strategies, enabling maintenance personnel to accurately locate and pre-deploy resources without manual investigation, improving response efficiency and targeted handling.

[0033] S33: Based on the pre-set Shapleyga interpretive analysis framework, the judgment criteria of the model are attributed to the cause, the key network elements or environmental factors that cause the anomaly are located, and the anomaly is traced back to its source. In this embodiment, to enhance model interpretability, the system introduces the Shapleyka interpretive analysis framework to attribute the causes of anomaly determinations. For example, regarding the weak coverage anomaly on the east grandstand, SHAP analysis shows that this determination is mainly attributed to three factors: insufficient transmission power of adjacent base stations, obstruction by the grandstand's steel structure, and higher-than-expected user density in the current area, with base station power contributing the most. Through this attribution, maintenance personnel can clearly prioritize adjusting base station power rather than expanding hardware, achieving precise policy implementation. This framework transforms the model's "black box" decisions into understandable causal relationships, facilitating the tracing of root causes, improving system credibility, and providing data-driven optimization directions for subsequent network planning.

[0034] In summary, by employing a hybrid architecture combining graph convolutional networks and long short-term memory networks, S31 enables the deep learning anomaly detection model to simultaneously capture the spatial dependence and temporal dynamics of coverage quality. This allows for the accurate identification of weak coverage clusters caused by structural occlusion in the complex three-dimensional space of a large intelligent stadium, and the prediction of signal quality degradation trends that may result from the outflow of crowds after the event, achieving a leap from static detection to dynamic prediction. S32 further structures the detection results into a complete information package containing geographical boundaries, anomaly types, confidence scores, and the probability of future deterioration. For example, it directly outputs "Section 5 of the east grandstand is a weak coverage area, and the probability of significant deterioration during the departure period is predicted," enabling maintenance personnel to quickly locate and deploy resources in advance without manual investigation, greatly improving response efficiency and targeted handling. S33 introduces the Shapleyga interpretive analysis framework, transforming the model's "black box" decisions into traceable causal relationships. It clearly identifies weak coverage primarily as being caused by factors such as insufficient base station power, steel structure obstruction, or unexpectedly high user density. This enables maintenance personnel to implement precise measures and prioritize adjustments to key network elements, enhancing system credibility and auditability while providing data-driven optimization directions for subsequent network planning. In this application, S31–S33, through the synergistic effect of deep detection, accurate prediction, and interpretable attribution, upgrades the location of coverage anomalies from vague perception to precise quantification, and from passive response to proactive prediction, significantly improving the intelligence level and decision-making transparency of network maintenance in densely populated areas.

[0035] In step S40: Based on the anomaly identification and prediction results, a targeted optimization strategy including parameter adjustment, resource scheduling, or network topology optimization is generated, and the strategy is distributed to the corresponding network elements for execution. This step includes the following steps: Construct a policy generator based on the combination of a rule engine and reinforcement learning, where the rule engine handles known and typical abnormal scenarios, and the reinforcement learning component explores the optimal policy for complex and novel anomalies; The generated optimization strategies include adjusting the antenna tilt angle and transmit power of specific base stations, modifying individual cell offsets, dynamically allocating spectrum resources, or guiding user equipment to perform inter-frequency / inter-system handover. The policy is converted into a network configuration command that can be issued through the northbound interface, and the success of the command execution is verified.

[0036] Specifically, in this embodiment, an example is given: during the closing hours of a concert in a large smart stadium, a surge in crowds on the east side of the stands leads to a risk of weak coverage. After the system's built-in rule engine identifies this typical scenario of "high probability of deterioration in weak coverage areas," it immediately triggers a preset rule: automatically increasing the transmission power of the micro base stations covering the area and decreasing the antenna tilt angle to focus the signal. This rule is based on historical concert data, has extremely low response latency, and can complete policy generation and distribution within seconds, avoiding the response delays caused by traditional manual analysis and ensuring a good user experience during peak closing times.

[0037] When a VR interactive experience area was temporarily added to the stadium, a mixed interference pattern of "high bandwidth and low latency" emerged, which was not covered by traditional rules. A reinforcement learning component was initiated to explore this issue: different combinations of "beamforming width-power-spectrum resources" were repeatedly tested in a simulation environment, recording the user experience score and overall network interference level for each combination, gradually converging to the optimal strategy of "narrow beam + millimeter-wave carrier + edge computing offloading." This mechanism overcomes the limitations of human experience, autonomously discovering the optimal parameter configuration for new scenarios and improving the system's adaptability to unknown business requirements.

[0038] To address the weak coverage on the east grandstand, the policy generator outputs a multi-dimensional optimization package: adjusting the tilt angle of specific base station antennas to cover deep blind spots, modifying individual cell offset values ​​to avoid premature handover to neighboring cells, temporarily allocating idle spectrum resources to expand hotspot cells, and simultaneously guiding some non-critical user terminals to actively switch to the venue's Wi-Fi network via system information broadcast. The policy is converted into a 3GPP-compliant configuration instruction set via the northbound interface, distributed through the base station management plane channel, and the performance counter is immediately read to confirm that the parameters have taken effect. This comprehensive policy achieves coordinated optimization of coverage, capacity, and interference, avoiding the unintended consequences of adjusting a single parameter. After the policy is distributed, the system periodically queries the actual effectiveness status of the configuration items via the base station's southbound interface, comparing the distributed values ​​with the current values. If execution fails due to hardware limitations or version incompatibility, the instruction is immediately marked as "incomplete," and alternative policies (such as adjusting neighboring cell parameters for compensation) are activated, while an alarm is generated to prompt maintenance personnel to intervene. This verification mechanism ensures that every strategy step is accurately implemented, avoiding the risk of losing control due to "disconnection immediately after deployment," forming a complete "generation-deployment-confirmation-backup" closed loop, and ensuring that the optimization process is reliable and controllable.

[0039] It should be understood that the sequence number of each step in the above embodiments 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.

[0040] In one embodiment, a communication network coverage quality monitoring device integrating edge computing is provided, which corresponds one-to-one with the communication network coverage quality monitoring method integrating edge computing in the above embodiments. For example... Figure 2 As shown, the communication network coverage quality monitoring device integrating edge computing includes a multimodal data acquisition unit 1, which is used to deploy multiple edge computing nodes in the target monitoring area and collect multimodal data from user equipment, base stations and environmental sensors in real time through the edge computing nodes; The local coverage quality map generation unit 2 is used to preprocess and extract features from the collected local multimodal data at each edge computing node, and generate a local coverage quality map based on it. Data synchronization and fusion are performed through a secure communication link between edge nodes to construct a global coverage quality map. Evolution trend prediction unit 3 is used to input the global coverage quality map into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas and interference areas, and predict their evolution trend. The optimization strategy generation unit 4 is used to generate targeted optimization strategies, including parameter adjustment, resource scheduling or network topology optimization, based on the anomaly identification and prediction results, and to distribute the strategies to the corresponding network elements for execution. The optimization effect evaluation unit 5 is used to re-collect network performance data after the optimization strategy is executed, evaluate the optimization effect, and feed the effect data back to the anomaly detection model for incremental learning and adaptive optimization of the model.

[0041] Specific limitations regarding the communication network coverage quality monitoring device for converged edge computing can be found in the limitations of the communication network coverage quality monitoring method for converged edge computing described above, and will not be repeated here. Each module in the aforementioned communication network coverage quality monitoring device for converged edge computing can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the electronic device, or stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of each module.

[0042] In one embodiment, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3As shown, this electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The database stores the database. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a communication network coverage quality monitoring method that integrates edge computing.

[0043] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Multiple edge computing nodes are deployed within the target monitoring area. These edge computing nodes collect multimodal data from user equipment, base stations, and environmental sensors in real time. The multimodal data includes reference signal received power and quality on the user side, load and channel state information on the base station side, and wireless channel impulse response characteristics on the environmental side. Each edge computing node preprocesses and extracts features from the collected local multimodal data, and generates a local coverage quality map based on this. Data is synchronized and fused through a secure communication link between edge nodes to construct a global coverage quality map. The global coverage quality map is input into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas, and interference areas, and to predict their evolution trend. Based on the anomaly identification and prediction results, targeted optimization strategies, including parameter adjustment, resource scheduling, or network topology optimization, are generated and distributed to the corresponding network elements for execution. After the optimization strategy is implemented, network performance data is collected again to evaluate the optimization effect, and the effect data is fed back to the anomaly detection model for incremental learning and adaptive optimization of the model.

[0044] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Multiple edge computing nodes are deployed within the target monitoring area. These edge computing nodes collect multimodal data from user equipment, base stations, and environmental sensors in real time. The multimodal data includes reference signal received power and quality on the user side, load and channel state information on the base station side, and wireless channel impulse response characteristics on the environmental side. Each edge computing node preprocesses and extracts features from the collected local multimodal data, and generates a local coverage quality map based on this. Data is synchronized and fused through a secure communication link between edge nodes to construct a global coverage quality map. The global coverage quality map is input into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas, and interference areas, and to predict their evolution trend. Based on the anomaly identification and prediction results, targeted optimization strategies, including parameter adjustment, resource scheduling, or network topology optimization, are generated and distributed to the corresponding network elements for execution. After the optimization strategy is implemented, network performance data is collected again to evaluate the optimization effect, and the effect data is fed back to the anomaly detection model for incremental learning and adaptive optimization of the model.

[0045] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0046] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0047] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for monitoring the coverage quality of a communication network integrating edge computing, characterized in that, The method includes the following steps: Multiple edge computing nodes are deployed within the target monitoring area. These edge computing nodes collect multimodal data from user equipment, base stations, and environmental sensors in real time. The multimodal data includes reference signal received power and quality on the user side, load and channel state information on the base station side, and wireless channel impulse response characteristics on the environmental side. Each edge computing node preprocesses and extracts features from the collected local multimodal data, and generates a local coverage quality map based on this. Data is synchronized and fused through a secure communication link between edge nodes to construct a global coverage quality map. The global coverage quality map is input into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas, and interference areas, and to predict their evolution trend. Based on the anomaly identification and prediction results, targeted optimization strategies, including parameter adjustment, resource scheduling, or network topology optimization, are generated and distributed to the corresponding network elements for execution. After the optimization strategy is implemented, network performance data is collected again to evaluate the optimization effect, and the effect data is fed back to the anomaly detection model for incremental learning and adaptive optimization of the model.

2. The communication network coverage quality monitoring method integrating edge computing according to claim 1, characterized in that, In the step of deploying multiple edge computing nodes within the target monitoring area and collecting multimodal data from user equipment, base stations, and environmental sensors in real time through these edge computing nodes, the collaborative collection of multimodal data specifically includes the following steps: The edge computing node uses a built-in protocol parsing module to periodically collect MR measurement reports from user equipment and performance management counter data from the base station. It invokes a dedicated driver deployed on the edge computing node to control the connected environmental sensor array to collect multipath delay spread and Doppler frequency shift characteristics of the wireless channel; Based on unified timestamps and geographic information labels, the data from different sources are spatiotemporally aligned to form a structured data stream with spatiotemporal context.

3. The communication network coverage quality monitoring method integrating edge computing according to claim 1, characterized in that, In the step of preprocessing and extracting features from the collected local multimodal data at each edge computing node, generating a local coverage quality map based on this, and synchronizing and fusing data through a secure communication link between edge nodes to construct a global coverage quality map, the specific steps for constructing the distributed quality map include: Each edge computing node uses the Kriging spatial interpolation algorithm to convert discrete local measurement data into a continuous spatialized quality layer; Edge nodes exchange their respective local quality layer summary information based on digital signatures and secure tunnels; By using a federated learning framework, a global graph generation model is collaboratively trained while protecting the privacy of the original data. This model synthesizes a high-precision global coverage quality graph by aggregating the model parameter updates of each node rather than the original data.

4. The communication network coverage quality monitoring method integrating edge computing according to claim 3, characterized in that, The step of inputting the global coverage quality map into a pre-trained deep learning anomaly detection model to identify the spatial distribution and severity level of coverage blind spots, weak coverage areas, and interference areas, and predicting their evolution trend, includes the following steps. The deep learning anomaly detection model adopts a hybrid architecture combining graph convolutional networks and long short-term memory networks to simultaneously capture the spatial dependence and temporal dynamics of coverage quality. The model output includes the geographical boundaries of the anomalous area, the anomaly type, the confidence score, and the probability of the anomaly worsening within a specific future time period; Based on the pre-set Shapleyga interpretive analysis framework, the model's judgment criteria are attributed to causes, key network elements or environmental factors that lead to anomalies are located, and anomaly tracing is achieved.

5. The communication network coverage quality monitoring method integrating edge computing according to claim 1, characterized in that, The step of generating targeted optimization strategies, including parameter adjustment, resource scheduling, or network topology optimization, based on the anomaly identification and prediction results, and distributing the strategies to the corresponding network elements for execution, includes the following steps: Construct a policy generator based on the combination of a rule engine and reinforcement learning, where the rule engine handles known and typical abnormal scenarios, and the reinforcement learning component explores the optimal policy for complex and novel anomalies; The generated optimization strategies include adjusting the antenna tilt angle and transmit power of specific base stations, modifying individual cell offsets, dynamically allocating spectrum resources, or guiding user equipment to perform inter-frequency / inter-system handover. The policy is converted into a network configuration command that can be issued through the northbound interface, and the success of the command execution is verified.

6. The communication network coverage quality monitoring method integrating edge computing according to claim 5, characterized in that, After the optimization strategy is implemented, network performance data is re-collected, the optimization effect is evaluated, and the effect data is fed back to the anomaly detection model for incremental learning and adaptive optimization. This step includes the following steps: Define a set of key performance indicators to quantify the optimization effect, including overall coverage compliance rate, switching success rate, and user-perceived speed. The KPI data before and after the strategy execution, the strategy itself, and its effect labels are used as a new training sample and stored in the experience replay pool. Periodically trigger the incremental learning process of the model, sample data from the experience replay pool to fine-tune the anomaly detection model and policy generator, and achieve continuous improvement in system performance.

7. A communication network coverage quality monitoring device integrating edge computing, applied to the communication network coverage quality monitoring method integrating edge computing as described in any one of claims 1 to 6, characterized in that, The device includes: a multimodal data acquisition unit (1), used to deploy multiple edge computing nodes in the target monitoring area, and to acquire multimodal data from user equipment, base stations and environmental sensors in real time through the edge computing nodes; The local coverage quality map generation unit (2) is used to preprocess and extract features from the collected local multimodal data at each edge computing node, and generate a local coverage quality map based on this. Data synchronization and fusion are carried out through a secure communication link between edge nodes to construct a global coverage quality map. Evolution trend prediction unit (3) is used to input the global coverage quality map into the pre-trained deep learning anomaly detection model, identify the spatial distribution and severity level of coverage blind spots, weak coverage areas and interference areas, and predict their evolution trend. The optimization strategy generation unit (4) is used to generate targeted optimization strategies, including parameter adjustment, resource scheduling or network topology optimization, based on the anomaly identification and prediction results, and to send the strategies to the corresponding network elements for execution. The optimization effect evaluation unit (5) is used to re-collect network performance data after the optimization strategy is executed, evaluate the optimization effect, and feed the effect data back to the anomaly detection model for incremental learning and adaptive optimization of the model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of a communication network coverage quality monitoring method that integrates edge computing as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of a communication network coverage quality monitoring method that integrates edge computing as described in any one of claims 1 to 6.

Images