Method, device, equipment, storage medium and product for identifying network coverage blind area

Abstract

The embodiment of the disclosure provides a kind of network coverage blind area identification method, device, equipment, storage medium and product, it is related to network communication field.The method comprises: the trajectory similarity between the first trajectory of vehicle and the second trajectory of terminal associated with vehicle is calculated;From multiple vehicles, target vehicle with trajectory similarity less than preset threshold is screened out, the first signal attenuation feature of target vehicle is extracted, and the second signal attenuation feature of terminal is extracted;The position information reported before network interruption of multiple target vehicles is clustered to obtain at least one clustering area;According to the consistency degree between the multiple first signal attenuation features and the multiple second signal attenuation features in the clustering area, the data density of clustering area, the probability that clustering area is signal blind area is predicted;According to the prediction probability corresponding to each clustering area, the identification result of network coverage blind area is obtained, the accuracy of identifying network coverage blind area can be improved by multi-source cross verification.

Description

Technical Field

[0001] This disclosure belongs to the field of network communication technology, and in particular relates to a method, apparatus, device, storage medium and product for identifying network coverage blind spots. Background Technology

[0002] Currently, technologies for identifying signal blind spots generally rely on single data sources such as Measurement Reports (MR), signaling, crowdsourced testing, or remote sensing imagery. This makes it difficult to effectively filter out false disconnections and misjudgments caused by individual terminal issues or abnormal events. Furthermore, most solutions rely solely on static judgments based on weak coverage clustering regions, deep fading points, or signal strength thresholds, resulting in coarse identification and insufficient confidence in the boundaries of coverage holes. In addition, MR data itself lacks precise latitude and longitude coordinates, and existing clustering methods have limited positioning accuracy. Crowdsourced testing or manual verification methods are inefficient and costly, failing to achieve highly reliable automated judgments. Therefore, existing technologies are not accurate enough in identifying network signal coverage areas. Summary of the Invention

[0003] This disclosure provides a method, apparatus, device, storage medium, and product for identifying network coverage blind spots, which can improve the accuracy of identifying network coverage blind spots through multi-source cross-validation.

[0004] In a first aspect, embodiments of this disclosure provide a method for identifying network coverage blind spots, the method comprising the following steps: For each vehicle, the trajectory similarity between the vehicle's first trajectory and the second trajectory of the terminal associated with the vehicle is calculated; where the time window corresponding to the first trajectory and the second trajectory is the historical period before the terminal associated with the vehicle lost network access. Target vehicles with trajectory similarity less than a preset threshold are selected from multiple vehicles. The first signal attenuation feature of the target vehicle is extracted based on the first signal measurement data of the target vehicle in the historical period. The second signal attenuation feature of the terminal associated with the target vehicle is extracted based on the second signal measurement data of the terminal in the historical period. Cluster the location information reported by multiple target vehicles before the network outage to obtain at least one clustering region; For each cluster region, the probability that the cluster region is a signal blind zone is predicted based on the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the cluster region and the data density of the cluster region. Based on the predicted probabilities corresponding to each cluster region, the identification results of network coverage blind spots are obtained.

[0005] In one feasible implementation, the first trajectory includes a plurality of first trajectory points, the second trajectory includes a plurality of second trajectory points, and calculating the trajectory similarity between the first trajectory of the vehicle and the second trajectory of the terminal associated with the vehicle includes: Create a distance list with the first trajectory point as the row and the second trajectory point as the column, and store the Euclidean distance between each first trajectory point and each second trajectory point in the corresponding cell of the distance list; Sum the values ​​of each cell in the first row of the distance list and the first target cell in the distance list, and sum the values ​​of each cell in the first column of the distance list and the second target cell, to obtain the cumulative distances corresponding to each cell in the first row and each cell in the first column of the distance list. For each first cell in the distance list, the cumulative distance of the first cell is obtained based on the Euclidean distance stored in the first cell and the minimum cumulative distance of the cells surrounding the first cell. The first cell refers to the cell in the distance list excluding the first row and the first column. Build a cumulative matrix based on the cumulative distances corresponding to the distance list; The trajectory similarity between the vehicle's first trajectory and the second trajectory of the terminal associated with the vehicle is determined based on the cumulative matrix.

[0006] In one feasible implementation, the second signal attenuation characteristics of the terminal are extracted based on second signal measurement data of the terminal associated with the target vehicle over a historical period, including: Acquire second signal measurement data of the terminal associated with the target vehicle within a historical time period, the second signal measurement data including at least the signal reception power; Based on the range of signal received power values, configure at least two observation states for the signal received power; Obtain the state transition matrix of the received signal power; the state transition matrix is ​​used to represent the state changes of the received signal power between at least two observation states. Based on the state transition matrix, the second signal attenuation characteristic of the terminal is obtained.

[0007] In one feasible implementation, the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within a clustering region is calculated using the following formula: in, Indicates the degree of consistency; Indicates the first The first signal attenuation characteristic; This represents the average value of multiple first signal attenuation characteristics within a clustered region; This indicates the number of first signal attenuation features in the cluster region; Indicates the first A second signal attenuation characteristic; This represents the average value of multiple second signal attenuation features in the clustered region.

[0008] In one feasible implementation, the location information reported by multiple target vehicles before the network outage is clustered to obtain at least one clustering region, including: Multiple data points are generated based on the location information reported by multiple target vehicles before the network outage. For each data point, if the number of data points within the target neighborhood is greater than or equal to a preset threshold, the data point is identified as a core point. For data points within the target neighborhood of a core point, if the number of data points within the target neighborhood is greater than or equal to a preset threshold, the data point is identified as the first core point of the core point. For each core point, a clustering region corresponding to the core point is established based on the target neighborhood of the core point and the target neighborhood of the first core point corresponding to the core point.

[0009] In one feasible implementation, determining the data density of the clustering regions includes: Calculate the total number of data points in the target neighborhood of each data point within the clustering region; The data density of a cluster region is obtained by calculating the ratio of the number of data points within the cluster region to the total number of data points.

[0010] In one feasible implementation, the probability that a clustered region is a signal blind zone is predicted based on the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region, including: The probability that a clustered region is a signal blind zone is obtained by weighting and summing the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region based on the prediction weights and the data density of the clustered region.

[0011] In one feasible implementation, before obtaining the probability that the clustered region is a signal blind zone by weighted summation of the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region based on prediction weights and the data density of the clustered region, the method further includes: Collect signal blind zone labels and samples within the clustering area. The samples include data density samples and consistency samples. The predicted probability of a sample is obtained by weighting the data density sample and the consistency sample according to the candidate weights. The candidate weights are adjusted based on the difference between the signal blind zone label and the predicted probability to obtain the predicted weights.

[0012] Secondly, embodiments of this disclosure provide a network coverage blind spot identification device, the device comprising: The consistency verification module is used to calculate the trajectory similarity between the first trajectory of the vehicle and the second trajectory of the terminal associated with the vehicle for each vehicle; wherein, the time window corresponding to the first trajectory and the second trajectory is the historical period before the terminal associated with the vehicle lost network access; The feature extraction module is used to filter out target vehicles with trajectory similarity less than a preset threshold from multiple vehicles, extract the first signal attenuation feature of the target vehicle based on the first signal measurement data of the target vehicle in the historical period, and extract the second signal attenuation feature of the terminal based on the second signal measurement data of the terminal associated with the target vehicle in the historical period. The clustering module is used to cluster the location information reported by multiple target vehicles before the network outage, and obtain at least one clustering region. The prediction module is used to predict the probability that a clustered region is a signal blind zone for each clustered region, based on the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region. The output module is used to obtain the identification results of network coverage blind spots based on the predicted probabilities corresponding to each cluster region.

[0013] Thirdly, this disclosure provides a network coverage blind spot identification device, the device comprising: a processor, and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement any of the network coverage blind spot identification methods in the above embodiments.

[0014] Fourthly, this disclosure provides a computer storage medium storing computer program instructions, which, when executed by a processor, implement any of the network coverage blind spot identification methods described above.

[0015] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements any of the network coverage blind spot identification methods described in the above embodiments.

[0016] The method, apparatus, device, storage medium, and product for identifying network coverage blind spots in this disclosure improve the accuracy of identifying network coverage blind spots through multi-source cross-validation. First, this disclosure utilizes the trajectory similarity between the terminal's trajectory and the vehicle's trajectory during the period before the terminal loses network access to eliminate signal abrupt interference caused by scenarios such as the terminal leaving the vehicle, thereby avoiding false blind spot reports caused by changes in terminal usage behavior. Furthermore, when the trajectories are consistent, the terminal signal attenuation process and the vehicle signal attenuation process during the period before network loss are extracted. Considering that inconsistent signal attenuation indicates a malfunction or performance change in the vehicle or terminal, the consistency of signal attenuation effectively eliminates single-end attenuation caused by antenna aging or software failures, thus filtering out more reliable signal coverage blind spots. In addition, this disclosure clusters discrete location points of the vehicle before network loss, using the density of the clustered regions to adaptively identify irregularly shaped coverage holes, effectively eliminating outlier locations and avoiding the waste of computational resources caused by random events such as temporary large vehicle obstruction, enabling more accurate determination of the signal coverage blind spot outline. By combining signal attenuation consistency and cluster density to jointly identify signal blind spots, the accuracy of identifying network signal coverage blind spots is improved. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments of this disclosure will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for identifying network coverage blind spots provided in an embodiment of this disclosure; Figure 2 This is a schematic diagram illustrating the formation process of multiple clustered regions provided in an embodiment of this disclosure; Figure 3 This is a schematic diagram of the overall process of a method for identifying network coverage blind spots provided in an embodiment of this disclosure; Figure 4 This is a schematic diagram of the structure of a network coverage blind spot identification device provided in an embodiment of this disclosure; Figure 5 This is a schematic diagram of the structure of a network coverage blind spot identification device provided in an embodiment of this disclosure. Detailed Implementation

[0019] The features and exemplary embodiments of various aspects of this disclosure will now be described in detail. To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description, in conjunction with the accompanying drawings and specific embodiments, will provide a further detailed description. It should be understood that the specific embodiments described herein are intended only to explain this disclosure and not to limit it. For those skilled in the art, this disclosure can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this disclosure by illustrating examples.

[0020] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0021] The acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0022] In the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, they do not mean that the applicant has used or necessarily used the solution.

[0023] In communication network technology, identifying areas without network coverage is a challenge for network optimization. Currently, known techniques for identifying coverage gaps, hidden blind spots, and other areas without network coverage mainly utilize single data sources such as drive tests (DT), call quality tests (CQT), mapping, signaling / call statistics, and over-the-top crowdsourcing tests (OTT), and employ geographic rendering and threshold determination processing. The system automatically collects air interface metrics such as Reference Signal Receiving Power (RSRP) and Signal to Interference plus Noise Ratio (SINR) through professional drive test equipment / auxiliary test units (ATUs), and verifies them by manual fixed-point dialing tests. It utilizes MR data reported by terminals (e.g., RSRP below -120dBm), access failures, handover failures, dropped calls, abnormal location updates, and zero traffic characteristics in core network signaling, combined with information from the Global Positioning System (GPS) / BeiDou Positioning and Geographic Information System (GIS), and overlays user disconnection, lag, and no-service reports from third-party OTT applications. Then, through spatial clustering, cluster regionalization analysis, and coverage boundary fitting, it identifies areas with no coverage where positioning signals are completely missing and network access is impossible. This is further verified by user complaint work orders, base station work participation, and propagation model simulation.

[0024] More specifically, the existing technologies mainly include the following: First, weak coverage clusters are screened based on MR data, and then coverage hole clusters are further identified and their locations determined. Second, user terminal signaling data is used to obtain location and movement speed, high-speed mobile user data is eliminated, and deep fading points are identified by calculating the relative attenuation value of medium- and low-speed user signals. Wireless network coverage blind spots are determined based on the continuous distribution of deep fading points. Third, user measurement reports and map information are merged to form a measurement map. Clustered areas are divided according to scenarios and base station deployments, and clustering rules are set. Weak coverage clusters are first screened, and then weak coverage areas are formed through clustering by association. Fourth, historical user signaling and measurement report data are merged to extract key signaling timestamps and corresponding measurement locations after the terminal reattaches to the network. Fifth, based on regional digital elevation model (DEM) grid terrain data, the spatial attenuation weight and total attenuation value of the target point are calculated, and the connectivity is determined point by point to statistically form coverage blind spots. By analyzing the roaming behavior characteristics of terminals between two access points, it is determined whether there are coverage gaps in the roaming path, and coverage holes between wireless access points (APs) are identified using common characteristics of multi-terminal roaming. A crowdsourcing testing model is adopted, issuing blind spot detection tasks to user terminals and storing the tasks and data on the blockchain. Coverage blind spots are determined based on the detection data reported by the terminals. By statistically analyzing the uplink received power distribution within the target cell cluster, the proportion of low-power uplink signals is calculated, and a threshold is set based on the uplink voice packet loss rate to identify cells with weak uplink coverage. Weak coverage cells are screened based on minimal drive test data, and weak coverage areas are formed through clustering and aggregation processing, thereby locating network optimization and planning points. By integrating remote sensing imagery, terrain reflectivity, 5G signal strength, and user behavior data, artificial intelligence (AI) is used to predict signal attenuation risk areas and blind spots, dynamically scheduling relay node switching and parameter optimization. However, the lack of matching verification with actual offline trajectories still poses a risk of false alarms.

[0025] To address the problems of existing technologies, this disclosure provides a method, apparatus, device, storage medium, and product for identifying network coverage blind spots. This disclosure considers that mobile phones and other terminals have no data to report before venturing into areas without coverage, making them unable to actively locate themselves and only able to passively wait for user complaints to provide specific locations without coverage. In contrast, vehicle-mounted systems have powerful hardware and strong signal reception capabilities, enabling them to report location information further into areas without coverage. Multi-source cross-verification is achieved based on vehicle data and terminal data. By combining vehicle network data and terminal data, the similarity of vehicle and person trajectories within the same time axis is calculated to ensure that both types of data point to the same area without coverage, thereby eliminating invalid or conflicting data. Based on this, signal fast attenuation features are extracted, and clustering is used to aggregate the vehicle's offline location with high confidence. This solves the problem that traditional methods cannot distinguish between true coverage holes and temporary signal fluctuations or neighbor cell configuration errors, achieving accurate and automated determination of areas without coverage, reducing the workload of manual on-site verification, and providing a reliable basis for real-time coverage replenishment or network optimization.

[0026] This method for identifying network coverage blind spots can be applied to scenarios where network signal coverage areas are identified or verified, and the subject executing this method can be any electronic device.

[0027] The method for identifying network coverage blind spots provided in the embodiments of this disclosure will be described below.

[0028] Figure 1 A flowchart illustrating a method for identifying network coverage blind spots according to an embodiment of this disclosure is shown. Figure 1 As shown, the method may include the following steps: S110. For each vehicle, calculate the trajectory similarity between the vehicle's first trajectory and the second trajectory of the terminal associated with the vehicle; wherein, the time window corresponding to the first trajectory and the second trajectory is the historical time period before the terminal associated with the vehicle lost network access; S120. Select target vehicles with trajectory similarity less than a preset threshold from multiple vehicles, extract the first signal attenuation feature of the target vehicle based on the first signal measurement data of the target vehicle in the historical period, and extract the second signal attenuation feature of the terminal associated with the target vehicle based on the second signal measurement data of the terminal in the historical period. S130. Cluster the location information reported by multiple target vehicles before the network disconnection to obtain at least one clustering region; S140. For each cluster region, based on the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the cluster region and the data density of the cluster region, predict the probability that the cluster region is a signal blind zone. S150. Based on the predicted probabilities corresponding to each cluster region, the identification results of network coverage blind spots are obtained.

[0029] It should be noted that the trajectory and signal attenuation features disclosed herein are extracted within the historical period before the terminal loses network access, in order to promptly capture the timing and characteristics of network changes in response to changes in the terminal. The clustering regions, on the other hand, are extracted within the historical period before the vehicle loses network access, used to accurately determine the approximate area of ​​the signal blind spot, while avoiding processing excessive amounts of irrelevant location data.

[0030] For example, the historical time period could be 3 minutes before the terminal went offline to avoid excessive computation. Being offline typically refers to a state where the device cannot connect to a mobile or wireless network.

[0031] Here, if none of the multiple vehicles have a target vehicle with a trajectory similarity less than a preset threshold, it indicates that the network disconnection triggered by the terminal is not real, that is, a false network disconnection situation has been identified.

[0032] Here, signal attenuation characteristics can refer to the signal strength drop coefficient of 4G / 5G, the handover from 5G to 4G, Evolved Packet System Fallback (EPSFB), and interoperability during redirection. The data density of a cluster region can be the density distribution value of each data point within the cluster region.

[0033] Thus, the network coverage blind spot identification method, apparatus, device, storage medium, and product of this disclosure can improve the accuracy of identifying network coverage blind spots through dual-source cross-validation. First, this disclosure utilizes the trajectory similarity between the terminal trajectory and the vehicle trajectory during the period before the terminal loses network access to eliminate signal abrupt interference caused by scenarios such as the terminal leaving the vehicle, thereby avoiding false blind spot reports caused by changes in terminal usage behavior. Furthermore, when the trajectories are consistent, the terminal signal attenuation process and the vehicle signal attenuation process during the period before network loss are extracted. Considering that inconsistent signal attenuation indicates a malfunction or performance change in the vehicle or terminal, the consistency of signal attenuation effectively eliminates single-end attenuation caused by antenna aging or software failure, thereby filtering out more reliable signal coverage blind spots. In addition, this disclosure also clusters discrete location points of the vehicle before network loss, using the density of the clustered regions to adaptively identify irregularly shaped coverage holes, and effectively eliminates outlier location points, avoiding the waste of computational resources caused by random events such as temporary large vehicle obstruction, and enabling more accurate determination of the outline of the signal coverage blind spot. By combining signal attenuation consistency and cluster density to jointly identify signal blind spots, the accuracy of identifying network signal coverage blind spots is improved.

[0034] The above steps are explained in detail below: Before S110, vehicle data and terminal data need to be collected to facilitate subsequent data processing.

[0035] In a specific example, vehicle-to-everything (V2X) OTT data and terminal MR data are acquired. From the V2X OTT data, data showing individual vehicles whose signal strength gradually declines to an extremely low point and for at least 5 minutes there is no subsequent data reporting is selected and entered into the database. During the V2X OTT data collection period, MR data showing mobile phone users' signal strength gradually declines to offline (including fading after 5G to 4G interoperation) is also entered into the database.

[0036] Regarding S110, considering that the acquisition cycles of vehicle data and terminal data may be completely inconsistent, this disclosure uses the distance accumulation between the first trajectory and the second trajectory to stretch or compress certain data points of the first trajectory or the second trajectory, so as to non-linearly stretch and align the two time series of the vehicle and the terminal on the time axis, thereby avoiding trajectory similarity errors caused by time asynchrony, rate inconsistency, etc.

[0037] In one optional implementation, the first trajectory includes a plurality of first trajectory points, the second trajectory includes a plurality of second trajectory points, and calculating the trajectory similarity between the first trajectory of the vehicle and the second trajectory of the terminal associated with the vehicle includes: Create a distance list with the first trajectory point as the row and the second trajectory point as the column, and store the Euclidean distance between each first trajectory point and each second trajectory point in the corresponding cell of the distance list; Sum the values ​​of each cell in the first row of the distance list and the first target cell in the distance list, and sum the values ​​of each cell in the first column of the distance list and the second target cell, to obtain the cumulative distances corresponding to each cell in the first row and each cell in the first column of the distance list. For each first cell in the distance list, the cumulative distance of the first cell is obtained based on the Euclidean distance stored in the first cell and the minimum cumulative distance of the cells surrounding the first cell. The first cell refers to the cell in the distance list excluding the first row and the first column. Build a cumulative matrix based on the cumulative distances corresponding to the distance list; The trajectory similarity between the vehicle's first trajectory and the second trajectory of the terminal associated with the vehicle is determined based on the cumulative matrix.

[0038] Here, the distance list can also be organized with the first trajectory point as the column and the second trajectory point as the row. The distances here can also be other variables besides Euclidean distance that can determine the similarity between the first and second trajectory points.

[0039] Here, the first target cell refers to all cells to the left of the cells in the first row; the second target cell refers to all cells above the cells in the first column. By calculating the cumulative distance of the cells in the first row and the first column, we can use them as the boundaries to build a cumulative matrix. We then explore below and to the right of the boundaries using a distance list, flexibly determining the surrounding cells of each first cell, until we obtain the cumulative distance of the last cell in the lower right corner of the distance list. This is used to determine the trajectory similarity.

[0040] In a specific example, a first trajectory is obtained from MR data acquired from the vehicle network, and a second trajectory is obtained from MR data acquired from the user's mobile terminal. Then, the Dynamic Time Warping (DTW) algorithm is used to calculate the trajectory similarity between the first and second trajectories at different sampling rates. Table 1 shows a summary table of first trajectory points for the first trajectory provided in one embodiment of this disclosure. As shown in Table 1, the first trajectory includes 24 first trajectory points, each recording its time, latitude and longitude location, and signal strength. Similarly, Table 2 shows a summary table of second trajectory points for the second trajectory provided in one embodiment of this disclosure. As shown in Table 2, the second trajectory includes 20 second trajectory points, each also recording its time, latitude and longitude location, and signal strength.

[0041] Table 1 Summary of the first trajectory points of the first trajectory Table 2 Summary Table of Second Trajectory Points for the Second Trajectory Next, a distance list is established based on the first and second trajectory points. First, the distance between the first and second trajectory points is calculated. Here, the latitude and longitude are treated as planar coordinates using the principle of small-range approximation, and the Euclidean distance is used as shown in equation (1).

[0042] (1) in, express and The Euclidean distance between them is expressed in degrees; Indicates the second trajectory point, Indicates the first trajectory point; Indicates the longitude of the second trajectory point; This represents the latitude of the first trajectory point.

[0043] It should be noted that the Euclidean distance in an approximate sphere... Therefore, 1 meter ≈ 1 / 111000 ≈ 9.009 × 10 -6 Spend.

[0044] After obtaining the Euclidean distances between all first trajectory points and each second trajectory point, the Euclidean distances are filled into the corresponding cells of the distance list. Table 3 shows a schematic table of the distance list provided in an embodiment of this disclosure. As shown in Table 3, the distance list is arranged with the first trajectory points as rows and the second trajectory points as columns. A1 represents the first second trajectory point, B1 represents the first first trajectory point, and the cell where A1 and B1 intersect stores the Euclidean distance between A1 and B1.

[0045] Table 3 Distance List Illustration After obtaining the distance list, it is necessary to determine the cumulative distance corresponding to each cell, and build a cumulative matrix based on this to obtain the final trajectory similarity.

[0046] First, the cumulative matrix is ​​defined as the cumulative cost matrix. And its dimensions are 20 rows × 24 columns, where A represents the row and B represents the column. Representing a subsequence and The minimum cumulative distance. The boundary condition for the cumulative matrix is... First, calculate the first row and first column, and then calculate the cumulative distance for the remaining cells step by step.

[0047] here, The first line The first column is .

[0048] For a cell in the first row, the cumulative distance is the sum of the Euclidean distances between that cell and all cells to its left. That is, j = 2..24, and ; ; ;……; .

[0049] For a cell in the first column, the cumulative distance is the sum of the Euclidean distances between that cell and all cells above it. That is, i = 2..20, and ; ; ;……; .

[0050] For cells that are not in the first row and first column, i.e., the first cell, the corresponding cumulative distance is calculated sequentially. That is, for the first cell where i=2..20 and j=2..24, the cumulative distance is calculated using equation (2): (2) in, Indicates the first Liede The cumulative distance of the first cell in the row, from That is, the cumulative distance of the cells surrounding the first cell.

[0051] For example, when the first cell is hour, ; ,therefore .

[0052] When the first cell is M[2][3], , ,therefore .

[0053] Similarly, M

[20]

[24] = Finally, all the values ​​of M are used to obtain the cumulative matrix, and the last cell is then processed. The cumulative distance is used as the distance similarity. In other words, This translates to approximately 60 meters. If this distance is less than a preset distance threshold, it indicates that the two trajectories are highly similar.

[0054] Thus, this disclosure transforms the problem of comparing spatiotemporal trajectories into a regular matrix operation by constructing a list of Euclidean distances with the first trajectory point as the row and the second trajectory point as the column. This makes the data structured, facilitating subsequent calculations and memory addressing, and fully preserves the initial differences in spatial dimensions, avoiding information loss due to feature compression. In the initialization phase of the cumulative distance, physical constraints are established using the cumulative distance of each cell in the first row and each cell in the first column of the distance list. This allows each first cell to flexibly bend in different directions towards surrounding cells under physical constraints. This ensures that even when faced with local time misalignments caused by inconsistent reporting frequencies, signal delays, or GPS drift in urban canyons, the similarity in macroscopic path morphology between the vehicle and the terminal can still be accurately captured, enhancing anti-interference capabilities. Therefore, the cumulative distance at the endpoint essentially represents the minimum total cost required to optimally stretch the two trajectories along the time axis, and can more accurately represent trajectory similarity.

[0055] Involving S120, when the trajectories of the vehicle and the terminal are consistent, the dynamic process of signal attenuation of the vehicle and the terminal is extracted separately, so as to use the signal attenuation characteristics to determine the degree of consistency of signal changes.

[0056] In one optional implementation, the second signal attenuation characteristics of the terminal are extracted based on second signal measurement data of the terminal associated with the target vehicle over a historical period, including: Acquire second signal measurement data of the terminal associated with the target vehicle within a historical time period, the second signal measurement data including at least the signal reception power; configuring at least two observation states for the signal reception power according to the value range of the signal reception power; acquiring a state transition matrix of the signal reception power; wherein the state transition matrix is used to represent the state change of the signal reception power between the at least two observation states; obtaining a second signal attenuation feature of the terminal based on the state transition matrix.

[0057] Here, the present disclosure may further acquire first signal measurement data of a target vehicle according to the above method, where the first signal measurement data at least includes signal reception power, and obtain a first signal attenuation feature of the target vehicle according to the state transition matrix thereof.

[0058] It should be noted that, according to the value range of the signal reception power, the present disclosure divides the signal reception power into different observation states.

[0059] In a specific example, a signal fast fading identification model is established to perform signal fast fading feature identification on Internet of Vehicles OTT data and mobile phone MR data respectively, so as to improve the accuracy of identifying uncovered areas. The fast fading identification may be implemented by using an existing method, such as a Hidden Markov Model (HMM) parameter estimation model. The HMM is used to describe a time series process containing implicit unknown states, and estimates the hidden internal state sequence by observing the external performance sequence.

[0060] The extraction of the signal attenuation feature of the terminal is taken as an example for description below, and the signal attenuation feature of the vehicle may also be extracted according to the following process. First, reference signal reception power RSRP in the wireless measurement report MR is divided into three discrete observation states: when RSRP > -85dBm, the observation state is good; when -95dBm < RSRP ≤ -85dBm, the observation state is available; when -105dBm < RSRP ≤ -95dBm, the observation state is fast fading, that is, a state of rapid signal attenuation. Then based on this, an observation state set is established, as shown in Equation (3): (3) wherein, respectively represent the observation states of good, available and fast fading; represents the observation state set.

[0061] Then, a hidden state set is established, and each observation state corresponds to one hidden state, so as to determine the real state of the channel quality to be estimated, as shown in Equation (4): (4) wherein, respectively represent the corresponding hidden states; represents the hidden state set.

[0062] Then, based on the triplet an HMM model is established, in which is a state transition matrix, is a probability matrix, is an initial probability, and a second signal attenuation feature is directly obtained based on the HMM model. The initial probability is as shown in Equation (5), the state transition matrix is as shown in Equation (6), and the emission probability matrix is as shown in Equation (7): (5) wherein, represents that the hidden state at the first moment is ; is the observation value at the first moment; represents the hidden state initial probability; represents the number of occurrences of samples where the hidden state at the first moment is exactly ; represents the total number of occurrences of samples where the first observation value is .

[0063] (6) wherein, represents the state transition probability of transitioning from the hidden state to the hidden state , and all state transition probabilities constitute the state transition matrix; represents the hidden state variable at the moment .

[0064] (7) wherein, represents the estimated probability of observing the observation state under the condition that the hidden state is .

[0065] In addition, a rule-based multi-dimensional threshold combination combined with time window sliding statistics may also be used. For example, a rule that the signal drop rate > X dB / s, the duration < Y seconds, and redirection signaling is accompanied may also approximate the fast fading process, so as to obtain the signal attenuation feature. A traditional machine learning classifier may also be adopted, such as a random forest or eXtreme Gradient Boosting (XGBoost). The input features include RSRP variance, handover attempt times, redirection flag, etc., the output indicates whether it is a coverage hole, and the signal attenuation feature is determined based on the input data of the coverage hole.

[0066] Thus, by discretizing the terminal signal receiving power into states and introducing a state transition matrix to describe the changes in the signal on the time axis, the originally discrete and abrupt measurement values ​​are transformed into a state transition process with time memory. This can accurately capture the probability path of the signal transition between different states, thereby effectively filtering out random glitches caused by fast fading or transient interference. Furthermore, by analyzing the relative changes in the probability values ​​of each item in the state transition matrix, it is possible to distinguish between environmental fading and equipment fading from a mechanistic perspective, thereby obtaining dynamic change characteristics with more and more accurate information.

[0067] Involving S130, considering that the vehicle positioning is more accurate and the signal reception is more stable, the location information reported by the target vehicle last time before the network was disconnected is used for clustering. Location clustering is achieved based on the reachability between data points to determine the approximate outline of the signal blind zone.

[0068] In one optional implementation, the location information reported by multiple target vehicles before the network outage is clustered to obtain at least one clustering region, including: Multiple data points are generated based on the location information reported by multiple target vehicles before the network outage. For each data point, if the number of data points within the target neighborhood is greater than or equal to a preset threshold, the data point is identified as a core point. For data points within the target neighborhood of a core point, if the number of data points within the target neighborhood is greater than or equal to a preset threshold, the data point is identified as the first core point of the core point. For each core point, a clustering region corresponding to the core point is established based on the target neighborhood of the core point and the target neighborhood of the first core point corresponding to the core point.

[0069] Here, the target neighborhood refers to the neighborhood of a data point. It is a local area centered on the data point, with a preset neighborhood radius, and covering the minimum number of data points. The core point can be determined through the target neighborhood.

[0070] It should be noted that the first core point here is indeed a core point, belonging to the core points determined during the outward expansion associated with the aforementioned core points. By checking whether each data point in the neighborhood of the first core point is also the first core point, other core points within the neighborhood of the first core point that can become core points are selected. At this point, the core point and the first core point are reachable, and the core point and the data points in the target neighborhood of the first core point are also reachable, indicating that the core point and the first core point are in the same cluster region. By expanding outward in this manner, a second core point within the first core point, a third core point within the second core point, and so on, can be found, thereby determining a cluster region.

[0071] In a specific example, the density-based spatial clustering of applications with noise (DBSCAN) algorithm is used to cluster the latitude and longitude of the target vehicle's last reported data before it left the network, with two preset global parameters: neighborhood radius. And the minimum number of points, MinPts.

[0072] For example, due to the high accuracy of vehicle-mounted positioning and the strong signal reception, it is recommended to set the neighborhood radius. The distance is set to 5 meters, and the minimum number of points (MinPts) is set to 10 to narrow the clustering range and improve the accuracy of identifying uncovered areas and the efficiency of data processing.

[0073] For any data point Its target neighborhood can be defined as... Centered on, with radius The set of all points within the hypersphere is given by equation (8): (8) in, Representing data points and data points The distance between them is measured using Euclidean distance, but other distance metrics can be selected based on the characteristics of the data. For point sets; To remove Beyond the point Other points within.

[0074] After obtaining the target neighborhood, the properties of each data point within the target neighborhood are determined based on the number of data points within the neighborhood and their relationship with other core points. This includes core points (other core points associated with the first core point, such as the first core point), boundary points, and noise points. Core points are used to expand outwards to explore other core points associated with and reachable from the first core point, forming the main content of the clustering region. Boundary points constitute the boundaries of the clustering region, while noise points are data points that need to be removed, regardless of the parameter used in constructing the clustering region.

[0075] Specifically, if ,but This is the core point. The neighborhood contains at least MinPts points, including the core point itself. And if... ,but Landing at a certain core point Within the neighborhood, This is a boundary point. It is neither a core point nor any core point. Within the neighborhood, it is a noise point.

[0076] To facilitate the description of the reachability of core points, this disclosure utilizes Directly Density-Reachable to describe the reachability relationship between core points and data points within their target neighborhood. That is, if... It is the core point, and Then point With point Density reachability. Density-reachable is used to describe the reachability relationships between multiple core points. If a data point chain exists... ,in , ,and from Density reaches directly, then from Density reachability. Density-connectedness describes the reachability relationship between two data points that share at least one density-connected core point; that is, if a point... , making and All from If the density is achievable, then and Density connected.

[0077] Any two points within the same cluster are density-connected; in other words, all points within each cluster region are data points that are density-connected. Figure 2 A schematic diagram illustrating the formation process of multiple clustered regions provided in one embodiment of this disclosure is shown. Figure 2 As shown, white dots represent data points that have not yet been clustered, blue dots form one cluster region, red dots form another, and green dots are forming a third, but not yet, cluster region. Each green region is the target region. Cluster regions are formed by progressively identifying other data points reachable from the core points. Data points within each cluster region are density-connected, while data points within different cluster regions are not reachable.

[0078] It should be noted that other clustering methods can also be used. For example, the K-means method can be combined with geographical constraints, or a density-based Ordering Points To Identify the Clustering Structure (OPTICS) algorithm can be used, or a Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm can be directly used to achieve similar spatial aggregation of offline points. Alternatively, time series clustering can be introduced, using the signal sequence of each terminal N seconds before offline as input, and Dynamic Time Warping (DTW) similarity clustering can be used to identify common spatial behavior patterns, which can also locate signal blind spots.

[0079] Thus, by strictly defining the core points through a preset quantity threshold, and merging and expanding the target neighborhoods of the core points and the first core point, the clustering region can adapt to any complex shape in the real geographical environment by relying on the correlation between the neighborhoods of the core points. In addition, points that cannot meet the density requirements are excluded from the core point cluster during the clustering process, avoiding the misjudgment of random events as permanent blind spots and improving the accuracy of the clustering region boundaries.

[0080] Before S140, it is necessary to determine the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustering region, as well as the data density of data points within the clustering region, in order to predict the probability that the clustering region is a signal blind zone.

[0081] In an optional implementation, the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within a clustering region is calculated using the following formula (9): (9) in, Indicates the degree of consistency; Indicates the first The first signal attenuation characteristic; This represents the average value of multiple first signal attenuation characteristics within a clustered region; This indicates the number of first signal attenuation features in the cluster region; Indicates the first A second signal attenuation characteristic; This represents the average value of multiple second signal attenuation features in the clustered region.

[0082] Here, the state transition matrix can be used as a signal attenuation feature.

[0083] In a specific example, the Pearson Correlation Coefficient (PCC) algorithm is used to determine the degree of consistency. The state transition matrix of the terminal is then calculated using formula (9). Expand into vectors , and the vector of the state transition matrix of the target vehicle Calculate the correlation coefficient .when A value greater than or equal to 0.8 indicates a single fast decay mode, with both signals exhibiting highly similar fast decay characteristics, and are more significantly affected by signal dead zones; when A value less than 0.8 indicates diverse fast decay patterns, potentially due to network issues, requiring specific analysis for each case. Therefore, the fast decay characteristics of the two signals are not similar. This indicates a high degree of consistency in the rapid decay pattern.

[0084] Thus, by calculating the synergy of the deviations of the signal attenuation at both ends from their respective mean values, it can be determined that both devices, which are independent in terms of RF hardware, are affected by signal blind zones, rather than individual factors such as terminal antenna failure, holding loss, or software failure, thereby improving the reliability of blind zone identification. Furthermore, when r is positive and large, it indicates that the signal attenuation of the vehicle and the terminal exhibits a synchronous increase or decrease, reflecting that the wireless signal has suffered continuous and stable physical attenuation within the geographical area, effectively distinguishing between instantaneous, occasional fluctuations and persistent coverage holes.

[0085] In one optional implementation, determining the data density of the clustering regions includes: Calculate the total number of data points in the target neighborhood of each data point within the clustering region; The data density of a cluster region is obtained by calculating the ratio of the number of data points within the cluster region to the total number of data points.

[0086] In a specific example, the data density of a cluster region is determined using the number of core points and cluster regions in DBSCAN. Here, cluster regions can also be clusters, so the data density can also be determined by the average density of the clusters, and the average density of the clusters = the average density per point. The number of points contained within the neighborhood. The specific formula is shown in equation (10): (10) in, It represents a specific cluster, i.e., a clustering region; This represents any data point belonging to cluster C; This represents the total number of data points in cluster C, where || represents the number of elements in the set. Point The set of all points contained in the target neighborhood; It represents the number of points in the set; it represents the number of points. How many points are within a radius ε around the perimeter? Represents the sum of all points in cluster C. Perform summation.

[0087] Thus, this disclosure uses the target neighborhood of each data point as the basic statistical unit. By accumulating the number of neighbors of all data points within their neighborhood radius, a comprehensive measure of the overall data abundance of the clustered region is formed, effectively avoiding density misjudgments caused by irregular geometric shapes or area differences in the clustered region. Furthermore, the total neighborhood count is normalized relative to the total number of data points using ratio calculation, eliminating scale interference caused by differences in the absolute number of data points within the clustered region, facilitating the output of highly reliable blind zone identification results.

[0088] Regarding S140, at the edge of the no-coverage area, the signal attenuates sharply to the point of disconnection. This attenuation pattern differs significantly from the smooth or slow attenuation pattern in normal coverage areas. Therefore, the signal obtained by comparing it with the normal pattern is... The values ​​will be relatively low. In areas without coverage, multiple users will disconnect from the network in the same area, and the disconnection event points will form dense clusters, resulting in higher data density values. After obtaining the consistency between the data density of the clustered regions and the signal attenuation characteristics, the data variation characteristics are used to jointly predict signal blind spots.

[0089] In one optional implementation, the probability that a clustered region is a signal blind zone is predicted based on the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region, including: The probability that a clustered region is a signal blind zone is obtained by weighting and summing the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region based on the prediction weights and the data density of the clustered region.

[0090] The prediction weights here can be obtained by training the training data.

[0091] In a specific example, for a clustered region, its Pearson correlation coefficient and data density At this point, normalization is performed on each of them to obtain normalized values. and As in equations (11) and (12): (11) (12).

[0092] Then, the normalized values ​​are weighted and summed using the prediction weights, and then expressed in a probabilistic manner, as shown in equations (13) and (14): (13) (14) in, Represents a linear combination; This indicates the probability that a clustered region is a signal blind zone; This indicates the prediction weight.

[0093] Furthermore, this disclosure also sets a preset probability threshold, which can be, for example, 0.6. In cases where the clustered region is determined to have an uncovered area, measures can be taken to address this issue. The clustered regions are directly covered and supplemented with points; while for Clustering regions, if the Pearson correlation coefficient Low, and data density The signal strength is high, and on-site signal screening can be conducted manually in the clustered area to identify any blind spots.

[0094] Thus, this disclosure fuses the consistency of signal attenuation at both ends within a clustered region and data density as two independent and complementary dimensions. The probability of this region being a signal blind zone is calculated according to preset weights. This probability reflects both the stability of the attenuation pattern and the confidence level of the sample size, jointly constraining the signal blind zone from both qualitative and quantitative perspectives, resulting in greater accuracy. Furthermore, the prediction weights can be dynamically adjusted according to actual operational needs, enabling the algorithm to adaptively adapt to network characteristics in different geographical regions and time periods, thus improving the solution's adaptability to various deployment environments.

[0095] This disclosure uses samples to determine the prediction weights.

[0096] In an optional implementation, before weighted summation of the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustering region and the data density of the clustering region based on prediction weights to obtain the probability that the clustering region is a signal blind zone, the method further includes: Collect signal blind zone labels and samples within the clustering area. The samples include data density samples and consistency samples. The predicted probability of a sample is obtained by weighting the data density sample and the consistency sample according to the candidate weights. The candidate weights are adjusted based on the difference between the signal blind zone label and the predicted probability to obtain the predicted weights.

[0097] Here, the signal blind zone label is used to indicate whether the clustered region is a signal blind zone, or the probability of it being a signal blind zone. The consistency sample is a sample representing the degree of consistency between the first and second signal attenuation features within the clustered region, and the data density sample is a sample representing the data density of the clustered region.

[0098] In a specific example, binary classification is used to determine the prediction weights. When the clustered region is a signal blind zone, the signal blind zone label is 1, and when the clustered region is a normal region, the signal blind zone label is 0. A model is then built based on formula (15): (15) in, These are the prediction weights obtained during training; The data density representing the clustering region, , indicating the clustering region; Indicates the degree of consistency; Indicates a signal blind spot label.

[0099] Table 4 shows a sample illustration provided in one embodiment of this disclosure. As shown in Table 4, signal blind zone labels and sample values ​​for 10 clustered regions were obtained through field testing.

[0100] Table 4 Sample Illustration Table 5 shows a schematic table of sample normalization results provided in one embodiment of this disclosure. As shown in Table 5, one iteration of the gradient descent method is used, but multiple iterations are required in practical applications. The sample values ​​are normalized to bring them to a similar scale.

[0101] Table 5. Schematic diagram of sample normalization results Then the prediction weights are initialized, so that Simultaneously define the loss function and gradient. Let r′ be... , with D′ as For a single sample Predicted probability ,in , Learning rate is taken The gradient is as shown in equation (16): (16) in, Indicates the first The gradient in the next iteration.

[0102] Then, gradient update iterations are performed. First, the gradient update for each sample is calculated. , and error Since the initial weights are 0, , And calculate the gradient using equations (17) and (18): (17) (18) in, This represents the estimated value of y, i.e., the predicted probability.

[0103] Table 6 shows a schematic table of sample iterative update results provided in one embodiment of this disclosure. As shown in Table 6, samples with y=0 and an error of +0.5 are considered as positive samples, including cluster regions 1, 3, 5, 7, and 9. Negative samples are cluster regions with y=1 and an error of -0.5, including 2, 4, 6, 8, and 10.

[0104] Table 6 Sample Iterative Update Results Then, the updated gradient is determined using the iterative update values ​​of the positive and negative samples, as shown in equation (19): (19) The sum of r′ for positive samples is 1.0 + 0.9125 + 0.7875 + 0.6625 + 0.95 = 4.3125. The total gradient is (0.5 * 4.3125) + (-0.5 * 1.0) = 2.15625 - 0.5 = 1.65625. Dividing by 10 gives 0.165625. The sum of D′ for positive samples is 0.122 + 0 + 0.089 + 0.211 + 0.067 = 0.489. The sum of r′ for negative samples is 0.2125 + 0.325 + 0.0875 + 0 + 0.375 = 1.0. The total gradient is (0.5 * 4.3125) + (-0.5 * 1.0) = 2.15625 - 0.5 = 1.65625. Dividing by 10 gives 0.165625. The sum of D′ for negative samples is 0.833 + 0.878 + 0.700 + 1.000 + 0.633 = 4.044. Therefore, the updated total gradient is (0.5 * 0.489) + (-0.5 * 4.044) = 0.2445 - 2.022 = -1.7775. Dividing by 10 gives -0.17775.

[0105] Update the candidate weights using the updated total gradient. , , After one iteration, Negative, It is positive, which is in line with expectations. The smaller the value, the more likely it is to have no coverage. The larger the value, the more likely it is to result in no coverage. Actual training requires multiple iterations until convergence, and the final training yields the following prediction weights: .

[0106] Thus, this disclosure collects historical samples labeled with signal blind zone tags, uses candidate weights to weight and sum the data density and consistency features in the samples, and uses the difference between the predicted probability and the true label as a feedback signal to iteratively correct the weight coefficients of each feature. This transforms the determination of weights from manual preset to data-driven, ensuring that the final predicted weights objectively reflect the true contribution of each feature to blind zone discrimination in real-world scenarios, fundamentally improving generalization and adaptability. Since the causes of blind zones vary significantly across different geographical regions, signal consistency often dominates in urban canyons, while in remote suburbs, sparse traffic makes data density the decisive factor for confidence. Sample feedback correction can automatically capture this regional specificity, which is beneficial for improving the accuracy of signal blind zone prediction.

[0107] In another specific example Figure 3 This is a schematic diagram of the overall flow of a method for identifying network coverage blind spots provided in an embodiment of this disclosure. Figure 3 As shown, the main signal blind spot identification method includes steps S1-S4. First, the data source is a combination of vehicle-to-everything (V2X) OTT data and mobile terminal MR data. Latitude and longitude are backfilled into the MR data through cross-network joint positioning or fingerprint database matching. Then, S1 extracts and filters data features, comparing the similarity between the first and second trajectories of the preprocessed data. The similarity of the vehicle and person trajectories within the same time axis before disconnection is calculated to confirm whether the two types of data match and point to the same uncovered area. If the trajectories do not match, S2-1 directly discards the data set and does not proceed to the next step; if the trajectories are similar, S21 and S22 are performed simultaneously. After trajectory matching, a large AI model is used to calculate the signal attenuation characteristics of the mobile MR data and the vehicle-to-everything (V2X) OTT data, including the consistency before and after entering the uncovered area, such as the sharp drop in 4G / 5G signal strength (signal fast attenuation model matching) and 5G to 4G interoperability, improving the confidence of the disconnection location clustering. After the mobile terminal disconnects from the network, the vehicle-to-everything (V2X) system is used to obtain the offline location deeper into the uncovered area, and the DBSCAN clustering algorithm is used to obtain the clustering areas of these offline locations. S3-2 identifies high-confidence coverage areas only when both fast fading consistency and obvious clustering are simultaneously satisfied, and determines the identification result of signal blind spots according to steps S140 and S150. Otherwise, S3-1 requires manual on-site verification to confirm the actual existence of coverage. If the verification confirms coverage, or if it is not a coverage issue, other network optimization measures are implemented. After S3-1 and S3-2, S4 is performed to supplement the coverage points of the network.

[0108] In summary, this disclosure addresses the problem in existing technologies where false network disconnection interferes with the identification of truly uncovered areas due to differences in terminal performance such as antenna performance and software issues, as well as the challenges of low positioning accuracy of mobile phone MR data, inability to accurately depict the rapid signal drop and decay before and after network disconnection, and rapid signal decay processes such as 5G to 4G interoperation behavior. It utilizes the strong signal and high-precision positioning capabilities of vehicle-to-everything (V2X) terminals as probes and benchmarks to calibrate and verify the true state of mobile phone terminals during signal interruption. By calculating the trajectory similarity between mobile phone terminal MR data and V2X terminal OTT data before network disconnection, false network disconnection caused by terminal issues such as antenna performance, software failures, and temporary obstructions can be effectively filtered out, improving the reliability of uncovered area identification and avoiding misjudging issues such as missing neighbor cells or incorrect handover parameters as coverage holes. Simultaneously, by utilizing the high-precision positioning capabilities of the vehicle-mounted system and the last latitude and longitude reported before network disconnection, the large errors of traditional MR gridding or signaling triangulation positioning are overcome, enabling precise location of coverage holes and significantly improving the positioning accuracy of uncovered locations. By simultaneously analyzing mobile MR data and vehicle-to-everything (V2X) OTT data, compared to existing solutions that rely solely on weak coverage thresholds or deep fading points for static judgment, this approach avoids misjudging network disconnection caused by handover failures due to limited mobile signal reception capabilities, such as missing neighbor cell configurations or incorrect handover parameters. The accuracy of identifying coverage-free areas is improved by combining cluster density values ​​and fast fading coefficients. This transforms passive, ambiguous, and low-confidence blind spot clues into active, precise, and high-confidence blind spot evidence, representing a complete technical solution from anomaly detection to precise location and effective optimization.

[0109] Figure 4 This is a schematic diagram of a network coverage blind spot identification device provided in an embodiment of this disclosure. Figure 4 As shown, the network coverage blind spot identification device 200 may include a consistency verification module 210, a feature extraction module 220, a clustering module 230, a prediction module 240, and an output module 250.

[0110] The consistency verification module 210 is used to calculate the trajectory similarity between the first trajectory of the vehicle and the second trajectory of the terminal associated with the vehicle for each vehicle; wherein the time window corresponding to the first trajectory and the second trajectory is the historical period before the terminal associated with the vehicle lost network access. The feature extraction module 220 is used to filter out target vehicles with trajectory similarity less than a preset threshold from multiple vehicles, extract the first signal attenuation feature of the target vehicle based on the first signal measurement data of the target vehicle in the historical period, and extract the second signal attenuation feature of the terminal based on the second signal measurement data of the terminal associated with the target vehicle in the historical period. Clustering module 230 is used to cluster the location information reported by multiple target vehicles before the network disconnection to obtain at least one clustering region. The prediction module 240 is used to predict the probability that a clustered region is a signal blind zone for each clustered region based on the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region. The output module 250 is used to obtain the identification results of network coverage blind spots based on the predicted probabilities corresponding to each cluster region.

[0111] Optionally, the first trajectory includes multiple first trajectory points, the second trajectory includes multiple second trajectory points, and the consistency verification module 210 is further used for: Create a distance list with the first trajectory point as the row and the second trajectory point as the column, and store the Euclidean distance between each first trajectory point and each second trajectory point in the corresponding cell of the distance list; Sum the values ​​of each cell in the first row of the distance list and the first target cell in the distance list, and sum the values ​​of each cell in the first column of the distance list and the second target cell, to obtain the cumulative distances corresponding to each cell in the first row and each cell in the first column of the distance list. For each first cell in the distance list, the cumulative distance of the first cell is obtained based on the Euclidean distance stored in the first cell and the minimum cumulative distance of the cells surrounding the first cell. The first cell refers to the cell in the distance list excluding the first row and the first column. Build a cumulative matrix based on the cumulative distances corresponding to the distance list; The trajectory similarity between the vehicle's first trajectory and the second trajectory of the terminal associated with the vehicle is determined based on the cumulative matrix.

[0112] Optionally, the feature extraction module 220 is also used for: Acquire second signal measurement data of the terminal associated with the target vehicle within a historical time period, the second signal measurement data including at least the signal reception power; Based on the range of signal received power values, configure at least two observation states for the signal received power; Obtain the state transition matrix of the received signal power; the state transition matrix is ​​used to represent the state changes of the received signal power between at least two observation states. Based on the state transition matrix, the second signal attenuation characteristic of the terminal is obtained.

[0113] Optionally, the prediction module 240 is further configured to calculate the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within a clustering region using the following formula: in, Indicates the degree of consistency; Indicates the first The first signal attenuation characteristic; This represents the average value of multiple first signal attenuation characteristics within a clustered region; This indicates the number of first signal attenuation features in the cluster region; Indicates the first A second signal attenuation characteristic; This represents the average value of multiple second signal attenuation features in the clustered region.

[0114] Optionally, the clustering module 230 is also used for: Multiple data points are generated based on the location information reported by multiple target vehicles before the network outage. For each data point, if the number of data points within the target neighborhood is greater than or equal to a preset threshold, the data point is identified as a core point. For data points within the target neighborhood of a core point, if the number of data points within the target neighborhood is greater than or equal to a preset threshold, the data point is identified as the first core point of the core point. For each core point, a clustering region corresponding to the core point is established based on the target neighborhood of the core point and the target neighborhood of the first core point corresponding to the core point.

[0115] Optionally, the prediction module 240 is also used to determine the data density of the clustering regions: Calculate the total number of data points in the target neighborhood of each data point within the clustering region; The data density of a cluster region is obtained by calculating the ratio of the number of data points within the cluster region to the total number of data points.

[0116] Optionally, the prediction module 240 is also used for: The probability that a clustered region is a signal blind zone is obtained by weighting and summing the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region based on the prediction weights and the data density of the clustered region.

[0117] Optionally, before obtaining the probability that a clustered region is a signal blind zone by weighted summation of the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region based on the prediction weights, the prediction module 240 is further configured to: Collect signal blind zone labels and samples within the clustering area. The samples include data density samples and consistency samples. The predicted probability of a sample is obtained by weighting the data density sample and the consistency sample according to the candidate weights. The candidate weights are adjusted based on the difference between the signal blind zone label and the predicted probability to obtain the predicted weights.

[0118] Figure 5 A schematic diagram of the hardware structure of a network coverage blind spot identification device provided in an embodiment of this disclosure is shown.

[0119] The network coverage blind spot identification device may include a processor 301 and a memory 302 storing computer program instructions.

[0120] Specifically, the processor 301 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this disclosure.

[0121] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 302 may include removable or non-removable (or fixed) media, or memory 302 may be non-volatile solid-state memory. Memory 302 may be internal or external to the integrated gateway disaster recovery device.

[0122] In one instance, memory 302 may be read-only memory (ROM). In one instance, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0123] Memory 302 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this disclosure.

[0124] The processor 301 reads and executes computer program instructions stored in the memory 302 to achieve... Figures 1 to 3The method for identifying network coverage blind spots in the illustrated embodiment.

[0125] In one example, the network coverage blind spot identification device may further include a communication interface 303 and a bus 304. For example, Figure 5 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 304 and complete communication with each other.

[0126] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this disclosure.

[0127] Bus 304 includes hardware, software, or both, that couples components of a network coverage blind spot identification device together. For example, and not as a limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in embodiments of this disclosure, this disclosure contemplates any suitable bus or interconnect.

[0128] Furthermore, in conjunction with the network coverage blind spot identification method in the above embodiments, this disclosure can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the network coverage blind spot identification methods in the above embodiments.

[0129] This application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the network coverage blind spot identification methods described in the above embodiments.

[0130] It should be clarified that this disclosure is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this disclosure is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this disclosure.

[0131] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this disclosure are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memory (ROM), flash memory, erasable read-only memory (EROM), floppy disks, compact disc read-only memory (CD-ROM), optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0132] It should also be noted that the exemplary embodiments mentioned in this disclosure describe methods or systems based on a series of steps or apparatus. However, this disclosure is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0133] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0134] The above description is merely a specific embodiment of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this disclosure is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this disclosure, and these modifications or substitutions should all be covered within the protection scope of this disclosure.

Claims

1. A method for identifying network coverage blind spots, characterized in that, The method includes the following steps: For each vehicle, the trajectory similarity between the vehicle's first trajectory and the second trajectory of the terminal associated with the vehicle is calculated; wherein, the time window corresponding to the first trajectory and the second trajectory is the historical time period before the terminal associated with the vehicle lost network access; Target vehicles with trajectory similarity less than a preset threshold are selected from multiple vehicles. First signal attenuation features of the target vehicles are extracted based on first signal measurement data of the target vehicles in the historical period. Second signal attenuation features of the terminals associated with the target vehicles are extracted based on second signal measurement data of the terminals associated with the target vehicles in the historical period. Cluster the location information reported by multiple target vehicles before the network outage to obtain at least one clustering region; For each clustered region, the probability that the clustered region is a signal blind zone is predicted based on the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region. Based on the predicted probabilities corresponding to each clustered region, the identification results of network coverage blind spots are obtained.

2. The method for identifying network coverage blind spots according to claim 1, characterized in that, The first trajectory includes multiple first trajectory points, and the second trajectory includes multiple second trajectory points. Calculating the trajectory similarity between the first trajectory of the vehicle and the second trajectory of the terminal associated with the vehicle includes: A distance list is created with the first trajectory point as the row and the second trajectory point as the column, and the Euclidean distance between each first trajectory point and each second trajectory point is stored in the corresponding cell of the distance list; The cumulative distances corresponding to each cell in the first row of the distance list and the first target cell in the distance list are summed respectively, and the cumulative distances corresponding to each cell in the first column of the distance list and the second target cell are summed respectively. For each first cell in the distance list, the cumulative distance of the first cell is obtained based on the Euclidean distance stored in the first cell and the minimum cumulative distance of the cells surrounding the first cell. The first cell represents the cell in the distance list other than the first row and the first column. Establish a cumulative matrix based on the cumulative distances corresponding to the distance list; The trajectory similarity between the first trajectory of the vehicle and the second trajectory of the terminal associated with the vehicle is determined based on the cumulative matrix.

3. The method for identifying network coverage blind spots according to claim 1, characterized in that, The extraction of the second signal attenuation feature of the terminal based on the second signal measurement data of the terminal associated with the target vehicle during the historical period includes: Acquire second signal measurement data of the terminal associated with the target vehicle during the historical period, the second signal measurement data including at least signal reception power; Based on the range of the signal received power, at least two observation states are configured for the signal received power; Obtain the state transition matrix of the received signal power; the state transition matrix is ​​used to represent the state change of the received signal power between at least two observation states; Based on the state transition matrix, the second signal attenuation characteristic of the terminal is obtained.

4. The method for identifying network coverage blind spots according to claim 1, characterized in that, The degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustering region is calculated using the following formula: in, Indicates the degree of consistency; Indicates the first The first signal attenuation characteristic; This represents the average value of multiple first signal attenuation features in the clustered region; This indicates the number of first signal attenuation features in the clustered region; Indicates the first A second signal attenuation characteristic; This represents the average value of multiple second signal attenuation features in the clustered region.

5. The method for identifying network coverage blind spots according to claim 1, characterized in that, The step of clustering the location information reported by multiple target vehicles before the network outage to obtain at least one clustering region includes: Based on the location information reported by the target vehicles before the network outage, multiple data points are generated. For each data point, if the number of data points within the target neighborhood of the data point is greater than or equal to a preset threshold, the data point is determined as a core point. For data points within the target neighborhood of the core point, if the number of data points within the target neighborhood is greater than or equal to a preset threshold, the data point is determined as the first core point of the core point. For each core point, a clustering region corresponding to the core point is established based on the target neighborhood of the core point and the target neighborhood of the first core point corresponding to the core point.

6. The method for identifying network coverage blind spots according to claim 5, characterized in that, The determination of the data density of the clustering region includes: Calculate the total number of data points in the target neighborhood of each data point within the clustered region; The data density of the clustered region is obtained by calculating the ratio of the number of data points within the clustered region to the total number of data points.

7. The method for identifying network coverage blind spots according to claim 1, characterized in that, The step of predicting the probability that a clustered region is a signal blind zone based on the consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region includes: The probability that the clustered region is a signal blind zone is obtained by weighting and summing the consistency between the plurality of first signal attenuation features and the plurality of second signal attenuation features within the clustered region based on the prediction weights and the data density of the clustered region.

8. The method for identifying network coverage blind spots according to claim 7, characterized in that, Before calculating the probability that the clustered region is a signal blind zone by weighted summation of the consistency between the plurality of first signal attenuation features and the plurality of second signal attenuation features within the clustered region based on prediction weights and the data density of the clustered region, the method further includes: Collect signal blind zone labels within the clustered region and samples within the clustered region, the samples including data density samples and consistency samples; The predicted probability of a sample is obtained by weighting and summing the data density samples and consistency samples according to the candidate weights. The candidate weights are adjusted based on the difference between the signal blind zone label and the predicted probability to obtain the predicted weights.

9. A network coverage blind spot identification device, characterized in that, The device includes: The consistency verification module is used to calculate the trajectory similarity between the first trajectory of the vehicle and the second trajectory of the terminal associated with the vehicle for each vehicle; wherein the time window corresponding to the first trajectory and the second trajectory is the historical time period before the terminal associated with the vehicle lost network access; The feature extraction module is used to filter out target vehicles with trajectory similarity less than a preset threshold from multiple vehicles, extract the first signal attenuation feature of the target vehicle based on the first signal measurement data of the target vehicle in the historical period, and extract the second signal attenuation feature of the terminal associated with the target vehicle based on the second signal measurement data of the terminal in the historical period. The clustering module is used to cluster the location information reported by multiple target vehicles before the network disconnection to obtain at least one clustering region. The prediction module is used to predict the probability that a clustered region is a signal blind zone for each clustered region, based on the degree of consistency between multiple first signal attenuation features and multiple second signal attenuation features within the clustered region and the data density of the clustered region. The output module is used to obtain the identification result of the network coverage blind spot based on the predicted probability corresponding to each clustered region.

10. A network coverage blind spot identification device, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the method for identifying network coverage blind spots as described in any one of claims 1-8.

11. A computer storage medium, characterized in that, The computer storage medium stores computer program instructions, which, when executed by a processor, implement the method for identifying network coverage blind spots as described in any one of claims 1-8.

12. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the method for identifying network coverage blind spots as described in any one of claims 1-8.

Images