An artificial intelligence-based low-altitude communication test optimization method, device and equipment

Abstract

The application relates to an artificial intelligence-based low-altitude communication test optimization method, device and equipment. Relates to the network test technical field, the method comprises the following steps: acquiring relevant data of a test area; wherein the relevant data is used for representing influencing factors of low-altitude communication; based on the relevant data, the test area is divided in three-dimensional space and marked with occlusion, to generate an adaptive test track; a plurality of unmanned aerial vehicles are controlled to fly cooperatively according to the adaptive test track, to determine communication test data and environment sensing data; wherein the communication test data and the environment sensing data are obtained when the unmanned aerial vehicles fly cooperatively; based on the communication test data and the environment sensing data, an abnormal root cause result of a low-altitude communication network is determined; and based on the abnormal root cause result of the low-altitude communication network, the low-altitude communication is tested and optimized.

Description

Technical Field

[0001] This invention relates to the field of network testing technology, and specifically to an artificial intelligence-based method, apparatus, and equipment for optimizing low-altitude communication testing. Background Technology

[0002] With the rapid development of the low-altitude economy, low-altitude applications based on mobile communication networks, such as drone logistics, aerial inspection, and low-altitude transportation, are becoming increasingly widespread. Due to the openness, minimal obstruction, and long signal propagation distance of low-altitude areas, coupled with the large-scale deployment of 5G networks, low-altitude scenarios are highly susceptible to problems such as signal interference, uneven coverage, chaotic handover, and large speed fluctuations, severely impacting the stability and reliability of low-altitude communication. In related technologies, drones fly along a fixed altitude and a fixed circular or rectangular trajectory to collect basic network indicators such as data rate. These data are then combined with on-site photos for post-event analysis to determine the root cause of the anomaly.

[0003] However, flying at a fixed altitude and with a fixed trajectory can easily miss key test areas, such as areas with weak coverage, leading to inaccurate test results and making it impossible to accurately determine the root cause of the anomaly.

[0004] Therefore, there is an urgent need for a method that can accurately determine the root cause of anomalies. Summary of the Invention

[0005] The technical problem that this invention aims to solve is that flying at a fixed altitude and with a fixed shape trajectory can easily miss key test areas, such as weakly covered areas, resulting in inaccurate test results and making it impossible to accurately determine the root cause of the anomaly.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: A method for optimizing low-altitude communication testing based on artificial intelligence, the method comprising: acquiring relevant data of the test area; wherein, the relevant data is used to characterize the influencing factors of low-altitude communication; based on the relevant data, dividing the test area into three-dimensional spaces and marking occlusions to generate an adaptive test trajectory; controlling multiple UAVs to perform coordinated flight according to the adaptive test trajectory to determine communication test data and environmental perception data; wherein, the communication test data and environmental perception data are obtained when the UAVs perform coordinated flight; based on the communication test data and environmental perception data, determining the root cause results of anomalies in the low-altitude communication network; and optimizing low-altitude communication testing based on the root cause results of anomalies in the low-altitude communication network.

[0007] The beneficial effects of this invention are as follows: By acquiring relevant data characterizing factors affecting low-altitude communication, and using this data to perform three-dimensional spatial division and occlusion marking of the test area, this invention breaks free from the constraints of fixed heights and fixed circular or rectangular trajectories. It can identify and cover key test areas such as weakly covered areas, avoiding omissions of critical regions. Furthermore, by controlling multiple UAVs to fly collaboratively along a generated adaptive test trajectory, and simultaneously acquiring communication test data and environmental perception data, the invention enriches the dimensions of data acquisition, improving the comprehensiveness and accuracy of the test data. Then, based on comprehensive and accurate communication test data and environmental perception data, the root causes of anomalies are determined, and test optimization is performed. This effectively solves the problem of inaccurate test results leading to the inability to accurately determine the root causes of anomalies in related technologies.

[0008] Based on the above technical solution, the present invention can be further improved as follows.

[0009] Furthermore, the relevant data includes terrain and feature data and base station configuration parameters. Based on this data, the test area is divided into three-dimensional spaces and marked with occlusion markers to generate an adaptive test trajectory. This includes: dividing the test area into a three-dimensional voxel grid based on terrain and feature data and base station configuration parameters; marking the occlusion status of the three-dimensional voxel grid based on simulated base station signal propagation paths and terrain and feature data to obtain the occlusion status of the three-dimensional voxel grid; wherein, the simulated base station signal propagation path is the propagation path from the base station under test to the center of the three-dimensional voxel grid; determining the signal propagation attenuation value of the three-dimensional voxel grid based on the occlusion status of the three-dimensional voxel grid and the base station signal propagation path; and generating an adaptive test trajectory based on the occlusion status of the three-dimensional voxel grid, the signal propagation attenuation value of the three-dimensional voxel grid, and preset multi-UAV cooperative flight constraints.

[0010] The beneficial effect of adopting the above-mentioned further scheme is that by using terrain and feature data and base station configuration parameters, the test area is finely divided into three-dimensional voxel grids. Then, by combining the simulated signal propagation path from the base station under test to the center of the voxel grid with terrain and feature data, the occlusion status of each three-dimensional voxel grid is accurately marked. Then, based on the grid occlusion status and signal propagation path, the signal propagation attenuation value of each grid is quantified. Finally, by combining the voxel grid occlusion status, signal propagation attenuation value and multi-UAV collaborative flight constraints, an adaptive test trajectory is generated. This can autonomously plan the trajectory in accordance with the actual low-altitude terrain and base station signal propagation characteristics, effectively avoiding the defects of regular trajectories that easily miss key test areas such as weak coverage, severe occlusion, and abnormal signal attenuation. At the same time, it takes into account the flight constraints of multiple UAVs, so that the generated test trajectory not only conforms to the low-altitude communication wave propagation law, but also has flight feasibility, thereby improving the accuracy of the root cause of anomalies.

[0011] Furthermore, based on the occlusion state of the 3D voxel mesh, the signal propagation attenuation value of the 3D voxel mesh, and the preset multi-UAV cooperative flight constraints, an adaptive test trajectory is generated. This includes: generating an initial test trajectory based on the occlusion state of the 3D voxel mesh, the signal propagation attenuation value of the 3D voxel mesh, and the preset multi-UAV cooperative flight constraints; wherein the initial test trajectory is composed of multiple sampling points connected together; if there are multi-UAV collisions and / or overlapping sampling points in the initial test trajectory, the initial test trajectory is adjusted to obtain the target test trajectory; based on the target test trajectory, the data missing rate when the UAV flies along the initial test trajectory is determined; if the data missing rate is not greater than the missing rate threshold, the target test trajectory is determined as the adaptive test trajectory; if the data missing rate is greater than the missing rate threshold, the step of adjusting the initial test trajectory is performed until the data missing rate is not greater than the missing rate threshold.

[0012] The beneficial effect of adopting the above-mentioned further scheme is that it generates an initial test trajectory consisting of multiple sampling points by using the occlusion state of the three-dimensional voxel mesh, the signal propagation attenuation value, and the constraints of multi-UAV cooperative flight. Then, it optimizes and adjusts the trajectory to obtain the target test trajectory to address the problems of multi-UAV collision and sampling point overlap in the initial trajectory. It further calculates the data missing rate of UAVs flying along the trajectory and uses the missing rate threshold as the judgment criterion to iteratively optimize the trajectory until the data missing rate meets the requirements and the final adaptive test trajectory is determined. This can constrain the trajectory quality from the dimension of data acquisition integrity and avoid the problems of test data missing and uneven sampling distribution caused by unreasonable trajectory layout.

[0013] Furthermore, multiple UAVs are controlled to fly collaboratively according to an adaptive test trajectory to determine communication test data and environmental perception data. This includes: during the collaborative flight of multiple UAVs according to the adaptive test trajectory, collecting basic performance data, air interface signaling data, base station data, UAV status data, and environmental point cloud data of the test area at sampling points; wherein, environmental perception data is determined based on the environmental point cloud data of the test area; based on the UAV status data, error compensation is performed on the basic performance data to obtain target performance data; and communication test data is determined based on the target performance data, air interface signaling data, base station data, and UAV status data.

[0014] The beneficial effect of adopting the above-mentioned further scheme is that, during the collaborative flight of multiple UAVs along the adaptive test trajectory, basic performance data, air interface signaling data, base station data, UAV status data, and environmental point cloud data of the test area are collected simultaneously. Environmental perception data is obtained from the environmental point cloud data. At the same time, error compensation is performed on the basic performance data using UAV status data to obtain accurate target performance data. Then, the target performance data, air interface signaling data, base station data, and UAV status data are integrated to comprehensively construct communication test data. This can solve the limitations of traditional methods that only collect basic indicators such as single rate and rely on on-site photos for simple environmental judgment. By compensating for UAV flight status errors, the test data distortion caused by fluctuations in motion attitude, altitude, and speed is eliminated.

[0015] Furthermore, the UAV status data includes flight speed, attitude angle, and flight altitude. Based on the UAV status data, error compensation is performed on the basic performance data to obtain the target performance data, including: based on flight speed and a preset uniform flight speed, speed compensation model is used to compensate the signal power in the basic performance data for speed; based on attitude angle and a preset attitude angle, attitude compensation model is used to compensate the signal power for pitch, roll, and yaw attitude; based on flight altitude and a preset flight altitude, altitude fluctuation compensation model is used to compensate the signal power for altitude, thus obtaining the compensated target performance data.

[0016] The beneficial effect of adopting the above-mentioned further solution is to simultaneously correct the measurement deviation introduced by the UAV's motion state from three dimensions: flight speed, flight attitude, and flight altitude. It systematically eliminates the problem of communication signal power acquisition distortion caused by UAV variable speed flight, attitude deflection, and altitude fluctuations, and makes up for the failure to consider UAV motion disturbances, thereby improving the accuracy of the root cause of anomalies.

[0017] Furthermore, based on communication test data and environmental perception data, the root causes of anomalies in the low-altitude communication network are determined, including: multi-dimensional data fusion of communication test data and environmental perception data to obtain feature fusion data; identification of abnormal regions and their spatial locations from the feature fusion data according to a preset signal quality judgment threshold; wherein, abnormal regions include weak coverage areas, interference areas, abnormal handover areas, and areas with insufficient data rates; based on the feature fusion data of the abnormal regions and their spatial locations, a machine learning fusion diagnostic model is used to determine the root causes of anomalies in the low-altitude communication network; wherein, the root causes of anomalies include at least one of the following: environmental obstruction, unreasonable base station configuration, communication interference, insufficient base station resources, beam coverage deviation, and signal propagation characteristic attenuation.

[0018] The beneficial effect of adopting the above-mentioned further solution is that by performing multi-dimensional fusion processing of communication test data and environmental perception data, feature fusion data is constructed. Then, based on the preset signal quality judgment threshold, various abnormal areas such as weak coverage areas, interference areas, abnormal handover areas, and insufficient rate areas are accurately identified from the feature fusion data. At the same time, the spatial location corresponding to the abnormal area is locked. Then, by combining the feature fusion data and spatial location information of the abnormal area, the machine learning fusion diagnostic model is used to make intelligent judgments and output the root cause results of the anomalies, including environmental obstruction, unreasonable base station configuration, communication interference, insufficient base station resources, beam coverage deviation, and signal propagation characteristic attenuation. This achieves accurate positioning of abnormal areas in low-altitude communication networks and intelligent and refined judgment of fault root causes, effectively improving the accuracy of anomaly root causes.

[0019] Furthermore, multi-dimensional data fusion of communication test data and environmental perception data is performed to obtain feature fusion data, including: constructing a three-dimensional environmental point cloud model of the test area based on environmental perception data; quantifying the signal attenuation value corresponding to different occlusions in the three-dimensional environmental point cloud model to obtain an environmental occlusion-signal attenuation correlation model; performing three-dimensional localization and tracking of effective interference sources based on interference signal characteristics in the communication test data to distinguish interference types; establishing an interference source-signal quality correlation model based on interference types; and performing multi-dimensional spatiotemporal fusion of communication test data, three-dimensional environmental point cloud model, environmental occlusion-signal attenuation correlation model, and interference source-signal quality correlation model to obtain feature fusion data.

[0020] The beneficial effect of adopting the above-mentioned further scheme is that by constructing a three-dimensional environment model, quantifying the impact of occlusion attenuation, locating and tracking interference sources and establishing a correlation model, the deep integration of communication data with three-dimensional environment features, occlusion attenuation features and interference impact features is achieved. This not only makes up for the shortcomings of single data dimensions and the inability to correlate environment and interference factors, but also makes the data more correlated and complete through spatiotemporal alignment fusion, significantly improving the information richness and accuracy of feature fusion data.

[0021] Furthermore, based on the root cause results of anomalies in the low-altitude communication network, the low-altitude communication is tested and optimized, including: determining the supplementary testing area and the corresponding supplementary testing 3D trajectory from the test area according to the root cause results of anomalies in the low-altitude communication network; controlling multiple UAVs to perform collaborative supplementary testing along the supplementary testing 3D trajectory to obtain supplementary testing data; integrating the supplementary testing data into feature fusion data for secondary root cause verification to obtain the target anomaly root cause result; generating the optimization result corresponding to the target anomaly root cause result based on the target anomaly root cause result; and testing and optimizing the low-altitude communication based on the optimization result.

[0022] The beneficial effect of adopting the above-mentioned further scheme is to accurately determine the supplementary measurement area through the anomaly root cause results of the low-altitude communication network, and plan the corresponding supplementary measurement three-dimensional trajectory for the supplementary measurement area to ensure that the supplementary measurement direction conforms to the characteristics of the anomaly area; then, control multiple UAVs to fly collaboratively along the supplementary measurement three-dimensional trajectory, synchronously collect supplementary measurement data, and then integrate the supplementary measurement data into the previous feature fusion data to complete the secondary verification of the anomaly root cause, further calibrating the accuracy of the root cause judgment; finally, based on the target anomaly root cause after secondary verification, a targeted optimization scheme is generated to achieve precise optimization and adjustment of low-altitude communication.

[0023] Furthermore, this invention provides an artificial intelligence-based low-altitude communication test optimization device, which includes: an acquisition module for acquiring relevant data of a test area; wherein the relevant data is used to characterize the influencing factors of low-altitude communication; a processing module for performing three-dimensional spatial division and occlusion marking of the test area based on the relevant data, and generating an adaptive test trajectory; a first determination module for controlling multiple UAVs to perform coordinated flight according to the adaptive test trajectory to determine communication test data and environmental perception data; wherein the communication test data and environmental perception data are obtained when the UAVs perform coordinated flight; a second determination module for determining the root cause results of anomalies in the low-altitude communication network based on the communication test data and environmental perception data; and an optimization module for performing test optimization of low-altitude communication based on the root cause results of anomalies in the low-altitude communication network.

[0024] Furthermore, this application provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the artificial intelligence-based low-altitude communication test optimization method described in the first aspect or any corresponding embodiment above.

[0025] Furthermore, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the artificial intelligence-based low-altitude communication test optimization method described in the first aspect or any corresponding embodiment above.

[0026] Furthermore, this application provides a computer program product, including computer instructions for causing a computer to execute the artificial intelligence-based low-altitude communication test optimization method described in the first aspect or any corresponding embodiment above. Attached Figure Description

[0027] Figure 1 A flowchart illustrating the AI-based low-altitude communication testing optimization method provided by this invention; Figure 2 A flowchart illustrating another AI-based low-altitude communication testing optimization method provided by the present invention; Figure 3 A flowchart illustrating another AI-based low-altitude communication testing optimization method provided by the present invention; Figure 4 A flowchart illustrating another AI-based low-altitude communication testing optimization method provided by the present invention; Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0028] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0029] With the rapid development of the low-altitude economy, low-altitude applications based on mobile communication networks, such as drone logistics, aerial inspection, and low-altitude transportation, are becoming increasingly widespread. However, due to the openness, minimal obstruction, and long signal propagation distance of low-altitude areas, coupled with the large-scale deployment of 5G networks, low-altitude scenarios are highly susceptible to problems such as signal interference, uneven coverage, chaotic handover, and large speed fluctuations, severely impacting the stability and reliability of low-altitude communication. In related technologies, drones fly along a fixed altitude and a fixed circular or rectangular trajectory to collect basic network indicators such as data rate. These data are then combined with on-site photos for post-event analysis to determine the root cause of the anomaly.

[0030] However, the relevant technology has the following drawbacks: The test trajectory is singular and fixed: it only uses a circular horizontal path with the base station as the center, without considering the spatial distribution of three-dimensional obstructions such as terrain, buildings, and vegetation, and does not combine the three-dimensional coverage characteristics of 5G beamforming to plan the trajectory. The test path has a low degree of matching with the actual low-altitude signal propagation environment, and the sampling points cannot cover key areas such as weak signal areas and interference areas. Data collection dimensions are missing: Only basic performance data such as speed, latency, and jitter are collected, while core data such as 5G air interface signaling, beamforming status, interference source characteristics, and base station resource allocation are not collected, making it impossible to accurately diagnose the root causes of poor signal coverage; Low testing efficiency and sampling density: The test is performed using a single drone, which limits the spatial coverage of a single batch of tests, results in insufficient sampling point density, and cannot simulate the 5G network resource competition and interference scenarios when multiple drones fly simultaneously. The test results are out of touch with the actual application scenarios. Track adjustment relies on manual intervention and lacks adaptability: on-demand adjustment is merely manual fine-tuning within the original fixed height / path range, without an intelligent adaptive adjustment mechanism based on real-time test data. The adjustment accuracy is low, the efficiency is low, and it is impossible to accurately supplement the test in areas with abnormal signals. The environmental analysis lacks quantification: the environment is only recorded by taking pictures of the environment in a plane at the characteristic location, which cannot build a three-dimensional environmental model, and cannot quantify the degree of attenuation of 5G signal propagation by the height, volume and material of the obstruction. The correlation analysis between environmental factors and signal quality is missing. Test errors that do not consider the drone's motion state: It is assumed that the drone completes the test at a constant speed, and the impact of the drone's speed changes, attitude deflection (pitch, roll, yaw) and altitude fluctuations on signal reception during actual flight is not considered, which leads to systematic errors in the test data; It only supports post-test optimization and lacks real-time support: test data is only centrally analyzed and optimization plans are formulated after all tests are completed. There are no real-time signal coverage warnings or temporary optimization suggestions during the test process, making it impossible to optimize while testing. The efficiency of connecting testing and optimization is low. The authenticity of test data is not guaranteed: there is no anti-tampering mechanism in the process of collecting, transmitting and storing test data. The data is easily distorted due to equipment failure and human operation, which makes the basis for subsequent optimization plan formulation unreliable.

[0031] Based on this, the present invention provides an artificial intelligence-based method for optimizing low-altitude communication testing. This method acquires relevant data characterizing factors influencing low-altitude communication, and uses this data to perform three-dimensional spatial division and occlusion marking of the test area. This breaks free from the constraints of fixed heights and fixed circular or rectangular trajectories, enabling the identification and coverage of key test areas such as weakly covered areas, avoiding omissions of critical regions. Furthermore, by controlling multiple UAVs to fly collaboratively along a generated adaptive test trajectory, simultaneously acquiring communication test data and environmental perception data, the data acquisition dimensions are enriched, improving the comprehensiveness and accuracy of the test data. Then, based on comprehensive and accurate communication test data and environmental perception data, the root causes of anomalies are determined and test optimization is performed, effectively solving the problem that inaccurate test results in some related technologies prevent accurate determination of the root causes of anomalies.

[0032] According to an embodiment of the present invention, an embodiment of a low-altitude communication test optimization method based on artificial intelligence is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0033] This embodiment provides an artificial intelligence-based method for optimizing low-altitude communication testing, which can be used in computer equipment such as computers and servers. Figure 1 This is a flowchart of an artificial intelligence-based low-altitude communication testing optimization method according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain relevant data for the test area; wherein, the relevant data is used to characterize the influencing factors of low-altitude communication.

[0034] The test area can be understood as the airspace and ground-related areas where low-altitude communication signal coverage and quality testing are required. Specifically, the test area can include urban buildings, vegetation, open spaces, base station coverage airspace, and airspace where drones can fly.

[0035] Relevant data can be understood as data on all factors influencing low-altitude communication. This relevant data can include terrain and feature data, base station configuration parameters, airspace constraint data, etc.

[0036] Terrain and feature data can include building outlines, heights, vegetation distribution, mountain obstructions, road network topography, etc. Base station configuration parameters can include the location of the 5G / base station under test, antenna height, azimuth, downtilt angle, transmit power, operating frequency band (Sub-6G / millimeter wave), beam configuration, etc. Airspace constraint data can include no-fly zones, height-restricted zones, obstacle distribution, controlled airspace boundaries, etc.

[0037] Specifically, the ground control center can collect all data that could affect the propagation, coverage, and interference of low-altitude communications.

[0038] As an example, terrain and feature outlines, building heights, and vegetation distribution data are imported from a geographic information platform; base station location, frequency band, antenna parameters, and transmit power configuration are synchronously obtained from a base station network management system; and airspace control platform is accessed to obtain no-fly zones, height-restricted zones, and obstacle coordinate constraint data.

[0039] Step S102: Based on relevant data, the test area is divided into three-dimensional spaces and marked with occlusion markers to generate an adaptive test trajectory.

[0040] An adaptive test trajectory can be understood as a three-dimensional flight trajectory that is automatically planned and dynamically adjusted based on occlusion distribution, signal attenuation, and UAV flight constraints. The adaptive test trajectory can be multiple polylines or a straight line, etc., without specific limitations.

[0041] Specifically, the test area is divided into three-dimensional grids using relevant data; signal obstruction is then judged and marked by combining terrain features and base station locations; and a three-dimensional adaptive test trajectory adapted to the current environment is automatically generated by taking into account obstruction, signal attenuation, and drone flight safety constraints.

[0042] As an example, a three-dimensional voxel grid is divided according to different test accuracies, with multiple resolution levels of 5m, 10m, or 20m selected as needed; a ray tracing algorithm is used to simulate the signal propagation path from the base station to each grid, determine the type of occlusion, and label the attenuation; an improved A* algorithm is used to plan multi-drone partitioned trajectories to avoid obstructions and maintain a safe distance for the drones; trajectories are planned in multiple altitude layers according to low altitude, low-medium altitude, medium-high altitude, and high altitude; after the initial trajectory is generated, it undergoes collaborative verification and signal loss rate verification, and iterative fine-tuning to obtain the final adaptive trajectory.

[0043] Step S103: Control multiple UAVs to fly collaboratively according to an adaptive test trajectory to determine communication test data and environmental perception data; wherein, the communication test data and environmental perception data are obtained when the UAVs fly collaboratively.

[0044] Communication test data can be understood as the actual test data of low-altitude networks collected by the UAV's onboard communication module. This data may include signal strength, signal-to-noise ratio, throughput, latency, packet loss rate, handover events, and access success rate.

[0045] Environmental perception data can be understood as the surrounding environment data collected by UAV-borne radar, visual sensors, and point cloud sensors. This environmental perception data can include 3D point clouds, obstacle location outlines, interference source locations, and real-world information on airspace occlusion.

[0046] Specifically, the ground control center sends out trajectories to multiple UAV clusters, and the multiple UAVs fly in coordination according to preset time sequences, zones, and altitude levels; during the flight, they simultaneously and passively collect network communication quality data and actively sense obstacles and interference information in the surrounding environment.

[0047] Step S104: Based on communication test data and environmental perception data, determine the root cause of the anomalies in the low-altitude communication network.

[0048] The root cause of anomalies can be understood as the fundamental reason for the location anomaly. These root causes can include network issues such as weak coverage, signal interference, insufficient data rate, abnormal handover, local blind spots, and millimeter-wave line-of-sight interruptions.

[0049] Specifically, after determining the communication test data and environmental perception data, the root cause of the anomaly can be further determined.

[0050] As an example, the preprocessed communication test data is compared with a normal network benchmark model. If the RSRP of a sampling point in a certain area is ≤-110dBm (Sub-6G band) or ≤-100dBm (millimeter wave band), and the downlink rate is ≤100Mbps and the uplink rate is ≤20Mbps, with a duration ≥3s, it is judged as weak coverage and rate failure anomaly. The three-dimensional coordinate range of the abnormal area is marked, and the environmental perception data (3D point cloud, obstacle outline) corresponding to the abnormal area is retrieved. Combined with the marked 3D voxel grid occlusion status, if it is found that the grid corresponding to the abnormal area is a non-line-of-sight grid or a weak line-of-sight grid, and the 3D point cloud data shows that there are tall buildings (height >30m) in the area, then... If the area has dense vegetation (height > 5m) and the straight-line propagation path between the area and the base station being tested is blocked by the aforementioned obstacles, the root cause is determined to be "excessive signal attenuation due to terrain and ground features, leading to weak coverage and substandard speed." If the abnormal area is a line-of-sight grid with no obvious obstruction, but the base station configuration parameters show that the transmission power of the corresponding base station in this area does not reach the preset threshold and the antenna downtilt angle is unreasonable (deviating from the abnormal area), the root cause is determined to be improper base station parameter configuration, and the signal coverage area does not cover the abnormal area. If the abnormal area is an open suburban area with no obstruction and normal base station parameters, but there is long-distance propagation (beyond the base station coverage area), the root cause is determined to be insufficient base station coverage, and the signal propagation distance is too far, resulting in excessive attenuation.

[0051] In one possible implementation, the communication test data and environmental perception data collected by the UAV collaborative flight are preprocessed to remove invalid data and calibration deviation data.

[0052] Specifically, for the collected uplink and downlink rates, latency, packet loss rates, and other data, a sliding window filtering algorithm is used to remove sudden outliers (such as instantaneous signal interruptions caused by the drone's instantaneous attitude shift), and a small amount of missing data is supplemented using linear interpolation. Simultaneously, in conjunction with drone status data, the communication test data is calibrated to correct distortions caused by drone flight attitude fluctuations and positioning deviations. Noise reduction and deduplication are performed on the 3D point cloud data, obstacle contour data, and interference source feature data collected by the drone. A point cloud stitching algorithm is used to integrate environmental data collected by multiple drones to generate a complete 3D environmental model of the test area. Feature extraction is performed on the interference source feature data to distinguish between internal interference and external device interference.

[0053] Step S105: Based on the root cause results of the low-altitude communication network anomalies, test and optimize the low-altitude communication.

[0054] Based on the identified root causes of anomalies in the low-altitude communication network, testing and optimization of the low-altitude communication system were conducted.

[0055] As an example, for weak coverage caused by occlusion, the supplementary 3D trajectory of detour or elevation is replanned, and local densification retesting is performed.

[0056] As an example, specific test tracks were added around the interference source to target the root cause of the interference and pinpoint the interference range and its impact.

[0057] As an example, to address the root cause of unreasonable base station parameters, the base station beam tilt angle and coverage direction were adjusted, and then resampling tests were conducted to verify the results.

[0058] In a specific implementation scenario, for the preliminary preparation stage of this invention, basic information about the test area, the base station under test, and the 5G frequency band is collected, and differentiated test parameters are formulated for different frequency bands and test area characteristics. The specific implementation is as follows: The ground control center collected two-dimensional topographic data and surface feature data of the test area through satellite remote sensing and topographic mapping systems. The two-dimensional topographic data included elevation, road, and water system information, while the surface feature data included the planar location, height, and number of layers of buildings, vegetation, and bridges. The data acquisition accuracy was set to 0.5m. The ground control center obtained base station configuration parameters from the tested 5G base station through the base station-side data interaction unit. These parameters included base station location, antenna height, antenna azimuth, downtilt angle, transmit power, beamforming capability, operating frequency band, and cell coverage area. The operating frequency band was divided into Sub-6GHz and millimeter-wave bands. The ground control center determined the corresponding frequency band characteristic parameters based on the operating frequency band of the tested 5G base station: for the Sub-6GHz band, it recorded the diffraction attenuation coefficient and penetration attenuation coefficient; for the millimeter-wave band, it recorded the line-of-sight propagation threshold, rain attenuation coefficient, and horizontal and vertical beamwidths. The ground control center sets the test accuracy levels according to actual testing needs. The accuracy levels are divided into Level 1, Level 2, and Level 3. Level 1 accuracy uses dense sampling, with a sampling point spacing of no more than 5 meters, suitable for urban core areas and areas with high concentrations of low-altitude aircraft. Level 2 accuracy uses conventional sampling, with a sampling point spacing of 10-15 meters, suitable for general urban areas. Level 3 accuracy uses sparse sampling, with a sampling point spacing of 20-30 meters, suitable for suburban areas and open areas. Simultaneously, the ground control center uniformly sets the UAV flight constraint parameters, including a maximum flight altitude not exceeding 120 meters, a minimum flight altitude not less than 3 meters, a maximum flight speed not exceeding 10 m / s, and attitude deflection thresholds: pitch and roll angles not exceeding 15°, and yaw angle not exceeding 30°.

[0059] The ground control center configures the operating parameters of the UAV's onboard test module differently for the two test types: Sub-6G frequency band and millimeter wave frequency band. For the Sub-6G frequency band, the wideband receiving module is enabled, the sampling frequency is set to 10Hz, and the signal detection sensitivity is -120dBm. For the millimeter wave frequency band, the narrow beam tracking receiving module is enabled, the sampling frequency is set to 50Hz, the signal detection sensitivity is -100dBm, and the beam alignment calibration function is enabled to achieve real-time alignment between the UAV's receiving antenna and the base station beam.

[0060] Based on the selected test accuracy level and test area, the ground control center configures the collaborative working parameters of the multi-UAV test cluster, including the number of UAVs participating in the test being 3 to 6, the test zone range corresponding to each UAV, the collaborative communication frequency between UAVs being 5.8 GHz, and limiting the data synchronization time error of the sampling points to no more than 10 ms.

[0061] The ground control center pre-configures a unified data acquisition list, dividing the collected data into four main categories: basic performance data, air interface signaling data, base station-side data, and UAV status data. Basic performance data includes RSRP, SINR, uplink / downlink rates, end-to-end latency, jitter, packet loss rate, handover success rate, and retransmission count. Air interface signaling data includes PCI and RRC connection status, PDSCH / PUSCH resource allocation, CSI-RS beamforming reference signal, and the frequency, power, and modulation characteristics of interference signals. Base station-side data includes base station transmit power, beamforming matrix, cell load rate, RB resource block occupancy rate, and interference source location data. UAV status data includes real-time latitude and longitude and altitude position, flight speed, acceleration, pitch / roll / yaw attitude angles, battery level, and test module operating status, with a positioning accuracy of no less than 0.1m.

[0062] The ground control center synchronously configures the three-dimensional point cloud scanning parameters: the urban area uses a point cloud scanning resolution of 5cm, the suburban area uses a point cloud scanning resolution of 10cm, the fixed scanning frequency is 2Hz, and the scanning angle is set to 360° horizontally and 120° vertically.

[0063] After completing the basic information collection and differentiated test parameter configuration, the process proceeds to step 2, the three-dimensional adaptive test trajectory planning process for multi-UAV collaboration, which provides prerequisite constraints for subsequent UAV collaborative flight sampling, communication, and environmental perception data collection.

[0064] The AI-based low-altitude communication testing optimization method provided by this invention acquires relevant data characterizing factors affecting low-altitude communication. Based on this data, it performs three-dimensional spatial division and occlusion marking of the test area, breaking free from the constraints of fixed heights and fixed circular or rectangular trajectories. This method can identify and cover key test areas such as weakly covered areas, avoiding omissions of critical regions. Furthermore, by controlling multiple UAVs to fly collaboratively along a generated adaptive test trajectory, it simultaneously acquires communication test data and environmental perception data, enriching the dimensions of data collection and improving the comprehensiveness and accuracy of the test data. Then, based on comprehensive and accurate communication test data and environmental perception data, it determines the root causes of anomalies and optimizes the test, effectively solving the problem of inaccurate test results leading to inaccurate identification of root causes in related technologies.

[0065] Based on the relevant data in this embodiment, including terrain and feature data and base station configuration parameters, in step S102, the test area is divided into three-dimensional spaces and marked with occlusion based on the relevant data, and an adaptive test trajectory is generated as follows: Figure 2 Implementation of steps S2021 to S2024: Step S2021: Based on terrain and feature data and base station configuration parameters, the test area is divided into a three-dimensional voxel grid.

[0066] Topographic and feature data can be understood as data on the topographic features and distribution of surface objects in the test area. Specifically, it can include topographic information such as elevation, roads, and water systems in the test area, as well as information on the planar location, height, outline, and distribution density of surface objects such as buildings, vegetation, bridges, and mountains. Base station configuration parameters can be understood as the core operating parameters of the 5G base station under test. Specifically, they may include the base station's physical location (latitude and longitude, altitude), antenna height, antenna azimuth angle, antenna downtilt angle, transmit power, operating frequency band (Sub-6G / millimeter wave), beamforming capability, cell coverage, etc., which are used to simulate signal propagation paths and calculate signal attenuation values.

[0067] Specifically, based on the distribution characteristics of terrain features and the coverage of base stations, the test area (three-dimensional spatial domain) is divided into a uniform and regular three-dimensional voxel grid.

[0068] As an example, the three-dimensional boundary of the test area is determined based on terrain and feature data (horizontally covering the coverage area of ​​the base station cell, and vertically covering the flight altitude of the drone from 3 to 120 meters). Then, combined with the cell coverage area and operating frequency band characteristics in the base station configuration parameters, the grid division accuracy is determined (the higher the frequency band and the smaller the coverage area, the higher the grid accuracy). Finally, according to the preset accuracy, the test area is divided into multiple cubic voxel grids, and each grid is bound to a unique identifier and three-dimensional coordinates to ensure that the grid coverage has no blind spots and no overlap, while associating the corresponding terrain and feature data.

[0069] As an example, based on the cell coverage range in the base station configuration parameters, the test area is divided into multiple base station coverage zones (each zone corresponds to one base station under test). Then, for each zone, combined with the terrain and landform data of that zone (such as whether there are mountains or rivers in the zone), different grid division methods are adopted: zones with mountain obstruction are divided with a precision of 5m×5m×3m to focus on covering the airspace around the mountains; zones with rivers are divided with a precision of 10m×10m×5m to avoid dangerous airspace above the rivers; after the division is completed, a unique zone identifier is assigned to the grid of each zone.

[0070] Step S2022: Based on the simulated base station signal propagation path and terrain data, the occlusion state of the three-dimensional voxel grid is marked to obtain the occlusion state of the three-dimensional voxel grid; wherein, the simulated base station signal propagation path is the propagation path from the base station under test to the center of the three-dimensional voxel grid.

[0071] The simulated base station signal propagation path can be understood as the propagation path from the base station under test to the center of the three-dimensional voxel grid. Specifically, the simulated base station signal propagation path is obtained by simulating the propagation path of signals transmitted from the base station under test to various locations within the test area (with the core being the center of the three-dimensional voxel grid).

[0072] The occlusion state of a 3D voxel grid can be understood as a marker of the obstruction of signal propagation for each 3D voxel grid. 3D voxel grids can be divided into three types: line-of-sight grids (signal propagation path is unobstructed, signal can directly reach the grid center), non-line-of-sight grids (signal propagation path is completely obstructed by heavy obstacles such as buildings or mountains), and weak line-of-sight grids (signal propagation path is partially obstructed by thin vegetation or low obstacles, or the signal is weak due to long-distance propagation).

[0073] Specifically, by simulating the signal propagation path and combining it with the occlusion characteristics of terrain features, the signal propagation obstruction status of each grid is determined, and the occlusion status is classified and marked.

[0074] As an example, based on the base station configuration parameters (base station location, antenna parameters, transmit power), the signal propagation path from the base station under test to the center of each three-dimensional voxel grid is simulated (the starting point, ending point, and propagation direction of the path are clearly defined); then, the terrain and feature data corresponding to each grid are retrieved to analyze whether there are obstacles on the signal propagation path, the type of obstacles, and their thickness; finally, according to the occlusion situation, each grid is marked as a line-of-sight grid, a non-line-of-sight grid, or a weak line-of-sight grid, while recording the type of occlusion (such as buildings and vegetation).

[0075] As an example, a 3D point cloud model of the test area is constructed using 3D point cloud information from terrain and feature data to accurately reconstruct the 3D outline of obstacles. Then, based on base station configuration parameters, a signal propagation path (represented as a line segment) from the base station to the center of each grid is generated. Through a path collision detection algorithm, it is determined whether there is a collision between the propagation path line segment and the obstacle in the 3D point cloud model: if there is no collision, it is marked as a line-of-sight grid; if there is a complete collision, it is marked as a non-line-of-sight grid; and if there is a partial collision, it is marked as a weak line-of-sight grid. At the same time, the coordinates of the collision point and the obstacle type are recorded to achieve accurate marking of the occlusion state.

[0076] Step S2023: Based on the occlusion state of the three-dimensional voxel grid and the base station signal propagation path, determine the signal propagation attenuation value of the three-dimensional voxel grid.

[0077] The signal propagation attenuation value of a three-dimensional voxel grid can be understood as the signal strength loss caused by factors such as propagation distance, terrain and object obstruction, and frequency band characteristics during the process of the signal propagating from the base station under test to the center of the three-dimensional voxel grid.

[0078] Specifically, by combining the grid occlusion status, the length of the signal propagation path, and the environment, the signal attenuation value at the center of each grid is calculated, and the signal strength within the grid is quantified.

[0079] As an example, the occlusion status and corresponding signal propagation path parameters (propagation distance, type and thickness of obstructions on the path) of each grid are extracted; then, combined with the operating frequency band in the base station configuration parameters, the corresponding attenuation calculation model is used to calculate the total attenuation value of the signal propagating from the base station to the center of the grid; among them, only distance attenuation is calculated for line-of-sight grids, while penetration attenuation and diffraction attenuation of obstructions need to be calculated for non-line-of-sight grids, and partial occlusion attenuation and long-distance attenuation need to be calculated for weak line-of-sight grids.

[0080] As an example, combining the transmit power and antenna gain in the base station configuration parameters, and the propagation environment (such as densely populated urban areas and suburbs) in the terrain and landform data, the corresponding propagation path loss model is selected: the COST231-Hata model is used in densely populated urban areas, and the Okumura-Hata model is used in suburbs; the path loss (i.e., attenuation value) of the signal from the base station to the grid center is calculated through the model, and the calculation results are corrected by combining the occlusion status in step S2022: the occlusion attenuation is increased by 10~20dB on the basis of the path loss for non-line-of-sight grids, the occlusion attenuation is increased by 5~10dB for weak line-of-sight grids, and no correction is made for line-of-sight grids; finally, the accurate signal propagation attenuation value of each grid is obtained.

[0081] Step S2024: Based on the occlusion state of the three-dimensional voxel mesh, the signal propagation attenuation value of the three-dimensional voxel mesh, and the preset multi-UAV cooperative flight constraints, an adaptive test trajectory is generated.

[0082] Preset multi-drone cooperative flight constraints can be understood as pre-defined flight rules to avoid collisions between multiple drones, ensure flight safety and the validity of test data. These rules may include the maximum and minimum flight altitude of drones, flight speed, attitude deflection threshold, minimum safe distance between drones, zoned flight range, and sampling point spacing.

[0083] Specifically, by taking into account three core conditions (occlusion status, attenuation value, and flight constraints), a three-dimensional trajectory is planned that is suitable for multi-UAV collaborative flight and can efficiently collect effective test data.

[0084] As an example, based on the grid occlusion status and signal propagation attenuation value, the test priority is determined: priority is given to covering non-line-of-sight grids, weak line-of-sight grids, and weak signal grids with attenuation values ​​≥20dB, followed by coverage of line-of-sight grids as a supplement; then, combined with preset multi-UAV cooperative flight constraints (safe distance, flight altitude, partition range, etc.), a path planning algorithm is used to allocate test partitions for multiple UAVs and plan the initial flight trajectory of each UAV; finally, the initial trajectory is verified (collision risk is checked, sampling point coverage is checked), unreasonable parts are fine-tuned, and finally an adaptive test trajectory is generated.

[0085] Specifically, step S2024 above includes the following steps: Step a1: Based on the occlusion state of the three-dimensional voxel mesh, the signal propagation attenuation value of the three-dimensional voxel mesh, and the preset multi-UAV cooperative flight constraints, an initial test trajectory is generated; wherein, the initial test trajectory is composed of multiple sampling points connected together.

[0086] Specifically, based on the occlusion state and signal propagation attenuation value of the three-dimensional voxel grid, the priority of sampling point placement is determined: dense sampling points are placed in non-line-of-sight grids, weak line-of-sight grids, and signal-weak grids with attenuation values ​​≥20dB, while sparse sampling points are placed in line-of-sight grids, with the spacing between sampling points matching the test accuracy level; then, combined with the preset multi-UAV cooperative flight constraints, a path planning algorithm is used to assign a dedicated test partition to each UAV; finally, within each partition, the placed sampling points are connected by a smooth curve to generate the initial test trajectory for each UAV.

[0087] Step a2: If there are multiple UAV collisions and / or overlapping sampling points in the initial test trajectory, adjust the initial test trajectory to obtain the target test trajectory.

[0088] Specifically, a simulation is performed on the initial trajectory cluster of multiple UAVs, simulating the synchronous flight of UAVs along the initial trajectory to check for collision risks and overlapping sampling points. If there are no collisions or overlapping sampling points, the initial test trajectory is directly used as the target test trajectory. If one or both of these problems exist, targeted adjustments are made: for collision risks, the trajectory and flight altitude of one or more UAVs are adjusted, or the flight sequence of the UAVs is adjusted (staggered flight) to ensure that the distance between UAVs always meets safety constraints; for overlapping sampling points, duplicate sampling points are deleted, or the sampling point position of one UAV is adjusted to ensure that the distance between sampling points meets the test accuracy requirements. After the adjustments are completed, the target test trajectory is obtained.

[0089] Step a3: Based on the target test trajectory, determine the data missing rate when the UAV flies along the initial test trajectory.

[0090] Based on the occlusion state and signal propagation attenuation value of the three-dimensional voxel grid, combined with the sampling point distribution of the target test trajectory, a signal coverage simulation algorithm is used to simulate the sampling process when the UAV flies along the target test trajectory. For each sampling point, it is determined whether the corresponding three-dimensional voxel grid can collect a valid signal. The number of sampling points that cannot collect valid data is counted, and the ratio of the number of sampling points to the total number of sampling points is calculated, which is the data missing rate.

[0091] Step a4: If the data missing rate is not greater than the missing rate threshold, the target test trajectory is determined as the adaptive test trajectory.

[0092] If the data missing rate is less than or equal to the missing rate threshold (e.g., first-level precision ≤ 5%), it means that the coverage effectiveness of the target test trajectory meets the test requirements and can ensure the collection of complete and effective test data. At this time, the target test trajectory is officially determined as the adaptive test trajectory, and the trajectory parameters are recorded synchronously.

[0093] Step a5: If the data missing rate is greater than the missing rate threshold, perform the step of adjusting the initial test trajectory until the data missing rate is no greater than the missing rate threshold.

[0094] If the data missing rate is greater than the missing rate threshold, it indicates that the target test trajectory has coverage blind spots or the sampling point layout is unreasonable. It is necessary to return to step a2 and make secondary adjustments to the initial test trajectory: focus on increasing sampling points in areas with high data missing rates (such as non-line-of-sight high attenuation grids), adjust the trajectory direction to cover blind spots, or optimize the sampling point distribution density; after the adjustment, a new target test trajectory is obtained, and step a3 is executed again to recalculate the data missing rate; repeat the above "adjustment-verification" process until the data missing rate is less than or equal to the missing rate threshold, and finally determine the target test trajectory that meets the requirements as the adaptive test trajectory to form a closed-loop optimization.

[0095] In one possible implementation scenario, the ground control center leads the grid division. The spatial resolution of the grid is strictly determined according to the test accuracy level, specifically: Level 1 accuracy is 5m×5m×3m (horizontal×horizontal×height), Level 2 accuracy is 10m×10m×5m, and Level 3 accuracy is 20m×20m×10m. Each voxel grid is a basic sampling unit, ensuring that the grid accuracy matches the test requirements. Specifically, the three-dimensional boundary of the test area is determined based on terrain and feature data (horizontally covering the base station cell coverage area, and vertically covering the UAV's flight altitude of 3~120m). Then, combined with the cell coverage area and operating frequency band characteristics in the base station configuration parameters, the grid resolution is checked for suitability. According to the above accuracy standards, the test area is divided into multiple cubic voxel grids, each bound to a unique identifier and three-dimensional coordinates, ensuring no blind spots or overlaps in grid coverage, while simultaneously associating with the corresponding terrain and feature data.

[0096] The ground control center uses a ray tracing algorithm to simulate the 5G signal propagation path from the base station under test to the center of each three-dimensional voxel grid. Based on the simulated propagation path and terrain data, it determines the occlusion type of each grid and clearly marks it into three states: line-of-sight grid (no occlusion, signal can directly reach the grid center), non-line-of-sight grid (obvious occlusions such as buildings and vegetation, signal cannot propagate directly), and weak line-of-sight grid (thin occlusions such as thin vegetation and low obstacles, or weak signal due to long-distance propagation). While marking the occlusion status, the ground control center simultaneously calculates the signal propagation attenuation value of each voxel grid, taking into account factors such as occlusion status, propagation distance, and frequency band characteristics.

[0097] The occlusion status (line-of-sight / non-line-of-sight / weak line-of-sight), signal propagation path length, and the type and thickness of obstructions along the path are extracted for each grid. Simultaneously, the operating frequency band (Sub-6G / millimeter wave) in the base station configuration parameters is combined to determine the core parameters for attenuation calculation. Different calculation methods are adopted for different occlusion statuses and frequency bands: line-of-sight grids only calculate distance attenuation; non-line-of-sight grids additionally calculate penetration attenuation and diffraction attenuation from obstructions; weak line-of-sight grids calculate partial occlusion attenuation and long-distance attenuation; millimeter wave bands require additional consideration of attenuation differences caused by beam propagation characteristics. The calculated signal propagation attenuation values ​​are bound to the grid's unique identifier and occlusion status, forming a correspondence between grid coordinates, occlusion status, and attenuation values.

[0098] Based on the occlusion state of the voxel grid and the signal propagation attenuation value, the ground control center clearly delineates the core test path and auxiliary test path. Using an improved A* algorithm, it plans partitioned 3D initial trajectories for the multi-UAV test cluster. The trajectory is a continuous spatial polyline, with each polyline being a sampling point. These sampling points coincide with the center of the voxel grid, ensuring a one-to-one correspondence between sampling points and the grid. The planning process strictly adheres to the following constraints to guarantee flight safety and test effectiveness: Obstacle avoidance constraints: The trajectory avoids the tops and edges of obstructions such as buildings and vegetation. The minimum horizontal distance between the drone's flight trajectory and obstructions is ≥3m, and the minimum vertical distance is ≥2m. Frequency band constraints: The initial trajectory of the millimeter wave band is planned only within the line-of-sight grid, and the trajectory direction is consistent with the propagation direction of the base station beam to ensure line-of-sight signal propagation and avoid significant attenuation of millimeter wave signals due to blockage. Cooperative constraints: The trajectories of multiple drones do not overlap, and the minimum safe distance between drones is ≥10m to avoid drone collisions; Altitude-layered constraints: Altitude-layered trajectories are planned for each UAV, divided into low-altitude (3~5m), mid-low-altitude (5~35m), mid-high-altitude (35~50m), and high-altitude (50~120m), matching the altitude division of existing technologies; at the same time, trajectory connection points are set at each altitude layer to achieve smooth take-off and landing of UAVs at different altitude layers, with the speed of take-off and landing ≤5m / s and attitude deflection ≤10°, ensuring flight stability.

[0099] The ground control center sends the planned 3D initial trajectory for each zone to a multi-UAV test cluster. Each UAV performs trajectory collaborative verification through a self-organizing network, and detects the collision risk and sampling point overlap of its own trajectory with the trajectories of other UAVs in real time. If a collision risk or sampling point overlap is found, the UAV immediately sends an adjustment request to the ground control center. The ground control center performs local fine-tuning of the trajectory in the problem area (such as adjusting the trajectory direction, fine-tuning the sampling point position, and optimizing the flight sequence) to ensure that the trajectory meets the collaborative constraints.

[0100] The ground control center performs signal coverage simulation verification on the optimized initial trajectory. A ray tracing algorithm is used to simulate the signal reception quality of the UAV flying along the trajectory, with a focus on calculating the signal data loss rate at the simulated sampling points. If the simulated sampling point signal data loss rate is ≤5%, the trajectory verification passes, and the final 3D adaptive test trajectory is generated. If the loss rate is >5%, the system returns to the initial trajectory planning stage, readjusting path planning parameters (such as increasing sampling point density and optimizing trajectory direction) until verification passes. The ground control center sends the final 3D adaptive test trajectory, partition range, and collaborative parameters (flight timing, safety distance, etc.) to each UAV in the multi-UAV test cluster. The UAVs store the trajectory data in their local storage modules.

[0101] The AI-based low-altitude communication testing optimization method provided by this invention refines the test area into a three-dimensional voxel grid by using terrain and ground feature data and base station configuration parameters. Then, combining the simulated signal propagation path from the base station to the center of the voxel grid with terrain and ground feature data, the occlusion status of each three-dimensional voxel grid is accurately marked. Based on the grid occlusion status and signal propagation path, the signal propagation attenuation value of each grid is quantified. Finally, an adaptive test trajectory is generated by comprehensively considering the voxel grid occlusion status, signal propagation attenuation value, and multi-UAV collaborative flight constraints. This method can autonomously plan trajectories that conform to the actual low-altitude terrain and base station signal propagation characteristics, effectively avoiding the shortcomings of regular trajectories that easily miss key test areas such as weak coverage, severe occlusion, and abnormal signal attenuation. Simultaneously, it takes into account the flight constraints of multiple UAVs, ensuring that the generated test trajectory not only conforms to the low-altitude communication wave propagation laws but also possesses flight feasibility, thereby improving the accuracy of anomaly root causes.

[0102] Furthermore, by using the occlusion state of the three-dimensional voxel mesh, the signal propagation attenuation value, and the constraints of multi-UAV cooperative flight, an initial test trajectory consisting of multiple sampling points is generated. Then, to address the issues of multi-UAV collisions and overlapping sampling points in the initial trajectory, the trajectory is optimized and adjusted to obtain the target test trajectory. The data missing rate of the UAV flying along the trajectory is further calculated, and the trajectory is iteratively optimized using the missing rate threshold as the judgment criterion until the data missing rate meets the requirements. The final adaptive test trajectory is then determined. This approach can constrain the trajectory quality from the perspective of data acquisition integrity, avoiding problems such as missing test data and uneven sampling distribution caused by unreasonable trajectory layout.

[0103] Based on step S103 of this embodiment, controlling multiple UAVs to fly collaboratively according to an adaptive test trajectory is used to determine communication test data and environmental perception data, such as... Figure 3 The implementation of steps S3031 to S3034 in the above steps: Step S3031: During the process of controlling multiple UAVs to fly collaboratively according to the adaptive test trajectory, basic performance data of sampling points, air interface signaling data, base station side data, UAV status data, and environmental point cloud data of the test area are collected.

[0104] The sampling points of the adaptive test trajectory can be understood as corresponding to the center point of the three-dimensional voxel grid. They are the core nodes for the UAV to collect various test data, and their distribution density matches the test accuracy level (Level 1 accuracy ≤ 5m, Level 2 accuracy ≤ 8m, Level 3 accuracy ≤ 10m).

[0105] Basic performance data can be understood as data directly collected by the UAV's onboard testing module. This data may include reference signal received power, signal-to-interference-plus-noise ratio, jitter, packet loss rate, and handover success rate.

[0106] Air interface signaling data can be understood as the control signals and data exchanged between the drone and the tested 5G base station via the air interface (wireless interface). This air interface signaling data may include physical cell identifiers, interference signal characteristics, etc.

[0107] Base station-side data can be understood as the base station operational status data obtained by the ground control center from the tested 5G base station through a dedicated interface. This data may include base station transmit power, beamforming matrix, cell load rate, etc.

[0108] Drone status data can be understood as real-time data collected on the drone's operational status during flight. This data may include battery level, onboard test module status, positioning deviation, acceleration, and speed.

[0109] Environmental point cloud data can be understood as discrete point data that reflects the three-dimensional contours of terrain and features in a test area, collected by an airborne 3D point cloud scanning module on a UAV. This environmental point cloud data can include the three-dimensional coordinates, contours, density, and reflectance intensity of surface obstacles such as buildings, vegetation, mountains, and bridges.

[0110] Specifically, during the process of controlling multiple drones to fly collaboratively according to an adaptive test trajectory, basic performance data, air interface signaling data, base station data, drone status data, and environmental point cloud data of the test area can be collected through set sensors.

[0111] As an example, the ground control center issues adaptive test trajectories and cooperative flight commands, controlling multiple UAVs to conduct cooperative flights according to preset zones and flight sequences, strictly adhering to the constraints of multi-UAV cooperative flight. Each UAV pauses for 0.5 to 1 second after arriving at the sampling point to ensure stable data acquisition. The airborne test module synchronously collects basic performance data and air interface signaling data at preset sampling frequencies (10Hz in the Sub-6G band and 50Hz in the millimeter-wave band). The ground control center obtains data from the base station side in real time through a dedicated interface and correlates it with the data collected by the UAVs according to the timestamp. The UAV's airborne positioning module and attitude sensor collect their own status data in real time, and the airborne 3D point cloud scanning module collects environmental point cloud data at a frequency of 2Hz and a 360° horizontal scanning angle.

[0112] Step S3032: Determine environmental perception data based on the environmental point cloud data of the test area.

[0113] Environmental point cloud data is preprocessed and features are extracted to transform the raw point cloud data into structured environmental perception data that can be used for subsequent analysis.

[0114] As an example, a Gaussian filtering algorithm is used to remove point cloud noise (such as invalid point clouds caused by birds and dust). The point cloud data collected by multiple UAVs is stitched together by an iterative nearest-point algorithm to generate a complete 3D point cloud model of the test area, eliminating the deviation of point clouds collected by multiple UAVs. Obstacles are identified and classified on the 3D point cloud model, and the 3D contours, positions, heights and distribution densities of obstacles such as buildings, vegetation and mountains are extracted. The occlusion level of each area is marked. Combined with the reflection intensity of the point cloud data, the location of interference sources is identified (strong reflection signals correspond to possible interference devices). Information such as obstacle distribution, occlusion level, terrain undulation, and preliminary location of interference sources are integrated to generate environmental perception data.

[0115] Step S3033: Based on the UAV status data, perform error compensation on the basic performance data to obtain the target performance data.

[0116] After determining the drone's status data, error compensation can be applied to the basic performance data based on the drone's status data to obtain the target performance data.

[0117] As an example, this study analyzes the correlation between UAV status data and basic performance data to identify error sources. For instance, attitude angle fluctuations leading to unstable signal reception, sudden changes in flight speed causing untimely sampling, and positioning deviations resulting in incorrect sampling point correspondences all contribute to distortion of basic performance data. A multi-dimensional error compensation algorithm is employed to correct different types of distortion. Specifically, linear correction is performed based on attitude angle deviations to compensate for signal strength deviations; linear interpolation algorithms are used to fill data gaps and correct rate and latency data; and the correspondence between sampling points and the 3D voxel grid is adjusted in conjunction with positioning data to correct signal attenuation-related performance indicators, thus obtaining the target performance data.

[0118] Specifically, step S3033 above includes the following steps: Step b1: Based on the flight speed and the preset constant flight speed, the signal power in the basic performance data is compensated for using a speed compensation model.

[0119] Specifically, the flight speed (including horizontal speed and vertical speed) is extracted from the UAV status data. A preset constant flight speed is retrieved, and the speed difference between the two is calculated. Based on the test frequency band (Sub-6G / millimeter wave), the Sub-6G frequency band corrects for signal power fluctuations caused by horizontal speed deviation, while the millimeter wave frequency band adds additional compensation weight for vertical speed deviation. Because millimeter wave signals are more sensitive to changes in propagation distance, the speed difference and the original signal power from the basic performance data are input into the speed compensation model. Based on the sign and magnitude of the speed difference, the corresponding signal power correction value is output. If the speed difference is positive, the signal power will be lower due to untimely sampling, and positive compensation is performed; if the speed difference is negative, the signal power will be higher, and reverse compensation is performed.

[0120] As an example, the following formula can be used to perform speed compensation on the signal power in the basic performance data: ;in, The signal power after speed compensation. To measure the signal power, For speed compensation coefficient, This refers to the actual flight speed of the drone. The preset constant speed is used.

[0121] Step b2: Based on the attitude angle and the preset attitude angle, the attitude compensation model is used to perform pitch attitude compensation, roll attitude compensation and yaw attitude compensation on the signal power.

[0122] Real-time pitch, roll, and yaw attitude angles are extracted from the UAV status data. Preset attitude angles are retrieved, and the deviation values ​​of the three types of attitude angles are calculated respectively. The attitude compensation model comprises three sub-models. These three sub-models correspond to three types of attitude compensation, with the sub-model parameters adapted to the attitude angle deviation range. Specifically, linear compensation is used when the attitude angle deviation is ≤5°; nonlinear compensation is used when the deviation is >5° and ≤10° to avoid over-correction.

[0123] The pitch attitude compensation sub-model corrects the signal reception angle offset caused by the drone's vertical pitch based on the pitch angle deviation, thus compensating for signal power deviation. The roll attitude compensation sub-model corrects the signal reception angle offset caused by the drone's horizontal tilt based on the roll angle deviation, thus balancing signal power. The yaw attitude compensation sub-model corrects the signal reception angle offset caused by the drone's heading deviation based on the yaw angle deviation, especially suitable for millimeter-wave frequency bands, ensuring that the signal reception direction is consistent with the base station beam direction.

[0124] As an example, pitch attitude compensation, roll attitude compensation, and yaw attitude compensation can be performed on signal power using the following formulas: ;in, The signal power after attitude compensation. For pitch / roll / yaw compensation coefficients, This represents the actual pitch / roll / yaw angle.

[0125] in, All were calibrated experimentally. The value is -0.5dB / °. The value is -0.4dB / °. The value is -0.1 dB / °.

[0126] Step b3: Based on the flight altitude and the preset flight altitude, the signal power is compensated for altitude using an altitude fluctuation compensation model to obtain the compensated target performance data.

[0127] The real-time flight altitude is extracted from the UAV status data, and a preset flight altitude (i.e., the center height of the voxel grid corresponding to the sampling point) is retrieved to calculate the altitude difference. Simultaneously, the signal propagation attenuation value of the grid corresponding to the sampling point in step S2023 is retrieved as a compensation auxiliary parameter. The altitude difference, signal propagation attenuation value, and the attitude-compensated signal power output in step b2 are input into the altitude fluctuation compensation model. The altitude fluctuation compensation model calculates the signal power correction value based on the magnitude of the altitude difference and the signal propagation attenuation value (the higher the altitude, the greater the attenuation value, and the lower the signal power). If the real-time flight altitude is higher than the preset altitude, the signal power is lower due to the increased propagation distance, and the model performs positive compensation; if it is lower than the preset altitude, the signal power is higher, and the model performs reverse compensation. The compensation magnitude is positively correlated with the attenuation value.

[0128] In one possible implementation, the signal power height compensation can be determined using the following formula: ;in, The signal power after high fluctuation compensation, The height compensation factor (calibrated experimentally, with a value of 0.1 dB / m) is used. This represents the actual altitude of the drone. For the planned height.

[0129] Step S3034: Based on target performance data, air interface signaling data, base station side data, and UAV status data, determine communication test data.

[0130] After determining the target performance data, air interface signaling data, base station side data, and UAV status data, the target performance data, air interface signaling data, base station side data, and UAV status data are integrated to obtain communication test data.

[0131] The artificial intelligence-based low-altitude communication testing optimization method provided by this invention simultaneously collects basic performance data, air interface signaling data, base station data, UAV status data, and environmental point cloud data of the test area during the collaborative flight of multiple UAVs along an adaptive test trajectory. Environmental perception data is obtained from the environmental point cloud data, and UAV status data is used to perform error compensation on the basic performance data to obtain accurate target performance data. Then, the target performance data, air interface signaling data, base station data, and UAV status data are integrated to comprehensively construct communication test data. This method can solve the limitations of traditional methods that only collect basic indicators such as single rate and rely on on-site photos for simple environmental judgment. By compensating for UAV flight status errors, it eliminates the problem of test data distortion caused by fluctuations in motion attitude, altitude, and speed.

[0132] Furthermore, by simultaneously correcting the measurement deviations introduced by the UAV's motion state from three dimensions—flight speed, flight attitude, and flight altitude—the system systematically eliminates the distortion in communication signal power acquisition caused by the UAV's variable speed flight, attitude deflection, and altitude fluctuations, thus compensating for the lack of consideration of UAV motion disturbances and improving the accuracy of anomaly root causes.

[0133] Based on the communication test data and environmental perception data in step S104 of this embodiment, the root cause of the low-altitude communication network anomaly is determined as follows: Figure 4 The implementation of steps S4041 to S4043 is as follows: Step S4041: Multi-dimensional data fusion is performed on communication test data and environmental perception data to obtain feature fusion data.

[0134] Specifically, core features are extracted from communication test data and environmental perception data. That is, core features (target performance data, key air interface signaling parameters, core indicators of the base station, and key parameters of UAV status) are extracted from communication test data, and core features (obstacle distribution, occlusion level, interference source location, and terrain undulation parameters) are extracted from environmental perception data. Then, a feature-level fusion algorithm is used to associate and integrate the core features of the two types of data to form feature-fused data.

[0135] As an example, multiple UAVs are controlled to fly collaboratively along an adaptive test trajectory (zonal and altitude-layered). Each UAV is equipped with an edge computing module, employing a local preprocessing and cloud synchronization mode: After the UAVs collect five types of data at the sampling point (voxel grid center), the edge computing module filters and compresses the data in real time to reduce data transmission bandwidth pressure; a 5.8GHz collaborative communication frequency is used to synchronously transmit the preprocessed data to the ground control center, while the UAVs reserve local cache to prevent data transmission interruption; base station-side data is obtained by the ground control center in real time through the northbound interface, and associated with the data collected by the UAVs at millisecond-level timestamps to ensure synchronization; for millimeter-wave band testing, the airborne narrow-beam tracking receiver module is activated to synchronously collect beam alignment deviation data, which is incorporated into the UAV status data to improve the accuracy of subsequent error compensation; at the same time, the UAV altitude layering information and climb / fall speed are collected in real time to ensure matching with trajectory constraints.

[0136] Specifically, step S4041 above may include the following steps: Step c1: Based on the environmental perception data, construct a three-dimensional environmental point cloud model of the test area.

[0137] A 3D environment point cloud model can be understood as a 3D stereoscopic model of the test area generated based on environmental perception data after preprocessing, stitching, and rendering.

[0138] Specifically, preprocessed environmental point cloud data, obstacle classification information, and terrain undulation parameters are extracted from environmental perception data. A preliminary 3D point cloud model is retrieved and stitched together. The model is optimized using a 3D point cloud rendering algorithm to supplement obstacle outline details and terrain textures, and to eliminate point cloud stitching deviations. The model's spatial coordinates are calibrated by combining the positioning information in the UAV status data and superimposed and compared with the voxel mesh model in step S2022 to finally generate a complete 3D environmental point cloud model.

[0139] Step c2: Based on the materials of different occluders in the 3D environment point cloud model, quantify the signal attenuation value corresponding to the material to obtain the environment occlusion-signal attenuation correlation model.

[0140] The environmental occlusion-signal attenuation correlation model can be understood as an algorithm model built based on the correspondence between the material of the occlusion object and the signal attenuation value.

[0141] Specifically, based on the point cloud reflection intensity and obstacle contour features of the 3D environmental point cloud model, combined with a pre-set material sample library, the specific materials of various occlusions are identified; typical samples of each material are selected, and the power values ​​before and after signal penetration are collected through an airborne test module to calculate the attenuation value and establish a quantification table of attenuation values ​​under different thicknesses and frequency bands; a linear regression algorithm is used to construct an environmental occlusion-signal attenuation correlation model, with the input being the occlusion material, thickness, and test frequency band, and the output being the attenuation value.

[0142] Step c3: Based on the characteristics of interference signals in the communication test data, perform three-dimensional localization and tracking of effective interference sources and distinguish the types of interference.

[0143] Interference signal features are extracted from air interface signaling data and basic performance data of communication test data. Unaffected clutter signals are eliminated, and effective interference signals are screened and their corresponding sampling point coordinates and timestamps are marked. A multi-UAV cooperative positioning algorithm is adopted. Combining the differences in interference signal intensity collected by multiple UAVs and the UAVs' own three-dimensional coordinates, the three-dimensional coordinates of the effective interference source are calculated by triangulation. The interference source is tracked in real time by combining a three-dimensional environmental point cloud model to avoid the influence of obstructions. The trajectory of the moving interference source is recorded. External electromagnetic interference and co-frequency interference are distinguished based on interference signal features and interference source location.

[0144] Step c4: Based on the type of interference, establish an interference source-signal quality correlation model.

[0145] The interference source-signal quality correlation model can be understood as an algorithm model built based on the type of interference and the strength of the interference source.

[0146] Specifically, the interference type, interference signal power, and distance between the interference source and the sampling point of the effective interference source are extracted, along with the signal quality index of the corresponding sampling point in the communication test data, and a one-to-one correspondence is established. The correlation between interference signal power, interference distance, and signal quality index under different interference types is quantified. An interference source-signal quality correlation model is constructed using the support vector machine algorithm, with the input being interference type, interference signal power, and interference distance, and the output being the predicted value of the signal quality index.

[0147] Step c5 involves multi-dimensional spatiotemporal fusion of communication test data, 3D environmental point cloud model, environmental occlusion-signal attenuation correlation model, and interference source-signal quality correlation model to obtain feature fusion data.

[0148] All data and models are spatiotemporally aligned, with the time dimension unified to millisecond-level timestamps and the spatial dimension unified to a coordinate system. The core features of communication test data, obstacle features of 3D environmental point cloud models, attenuation values ​​of environmental occlusion-signal attenuation correlation models, and predicted values ​​of interference source-signal quality correlation models are integrated, with each sampling point corresponding to a complete set of fused data. Duplicate and invalid data are removed, key features are strengthened through weight allocation, redundant features are suppressed, and finally feature fusion data is generated.

[0149] In one possible implementation, a 3D modeling server receives environmental point cloud data collected by multiple UAVs. A point cloud registration algorithm is used to fuse the point cloud data from multiple UAVs, eliminating point cloud overlap and deviation, and generating a preliminary 3D point cloud model of the test area. Point cloud filtering is applied to the preliminary point cloud model to remove noise and outliers. Then, a point cloud segmentation algorithm (region growing algorithm) is used to classify the features in the preliminary 3D point cloud model into categories such as buildings, vegetation, roads, and water systems, and material attributes (concrete, glass, trees, metal, etc.) are assigned to each category. Finally, spatial coordinate information is added to the segmented 3D point cloud model to generate a high-precision 3D environmental point cloud model of the test area. The spatial resolution of the 3D environmental point cloud model is consistent with the preset point cloud scanning resolution, allowing for a direct display of the spatial location, height, volume, and material of all obstructions within the test area.

[0150] The ground control center spatially correlates the high-precision 3D environmental point cloud model with the precise full-dimensional data set, mapping the signal quality data of each sampling point to the corresponding spatial location of the model; through the signal attenuation quantization algorithm, it calculates the attenuation value of 5G signal for different obstructions and different materials: concrete, glass, trees, and metal; at the same time, it calculates the correlation between the obstruction angle, obstruction distance and signal attenuation value, and establishes an environmental obstruction-signal attenuation correlation model.

[0151] The interference source diagnosis module of the ground control center extracts the characteristics of interference signals from the collected air interface signaling data. If the power of the detected interference signal is ≥-90dBm and the attenuation of the SINR of the 5G signal is ≥10dB, it is determined to be a valid interference source. A multi-UAV cooperative positioning algorithm is used to dynamically locate the valid interference source. The real-time three-dimensional position of the interference source is calculated by using the time difference and angle difference of the same interference signal received by multiple UAVs, with a positioning accuracy of ≤10m. If the interference source is a mobile interference source, the ground control center tracks the movement trajectory of the interference source in real time, calculates its movement speed and direction, and displays it in real time on the monitoring interface. At the same time, it sends instructions to the multi-UAV test cluster to allow the UAVs to follow and sample the movement trajectory of the interference source and continuously collect interference signal characteristic data. The valid interference sources are classified and labeled into co-frequency interference sources (5G co-frequency band), adjacent frequency interference sources (5G adjacent frequency band), and spurious interference sources (other frequency bands). The interference intensity and interference range of each interference source are recorded, and an interference source-signal quality correlation model is established.

[0152] Using the 3D coordinates of sampling points and test timestamps as the core, this method performs multi-dimensional spatiotemporal fusion of communication test data, high-precision 3D environmental point cloud models, environmental occlusion-signal attenuation correlation models, and interference source-signal quality correlation models. The time dimension is unified to millisecond-level timestamps to ensure that test data, interference source locations, and attenuation values ​​of obstructions at the same time point are accurately correlated. The spatial dimension uses a unified coordinate system to accurately overlay the 3D coordinates of sampling points, interference sources, and obstruction locations. Duplicate and invalid data are removed, and key features (such as signal quality, attenuation value, and interference type) are strengthened through weight allocation, ultimately generating feature fusion data.

[0153] Step S4042: Based on a preset signal quality judgment threshold, identify abnormal regions and their spatial locations from the feature fusion data; wherein, abnormal regions include weak coverage regions, interference regions, handover abnormal regions, and insufficient rate regions.

[0154] The preset signal quality judgment threshold can be understood as a pre-set threshold. Specifically, the preset signal quality judgment threshold is retrieved according to the test accuracy level, test frequency band, and scenario. For example, weak coverage area: reference signal received power ≤ -120dBm; interference area: signal-to-interference-plus-noise ratio ≤ 3dB; abnormal handover area: handover success rate ≤ 90%; insufficient rate area: uplink rate ≤ 100Mbps, downlink rate ≤ 1Gbps.

[0155] Specifically, the feature fusion data is traversed, and the core performance indicators of each sampling point are compared with the preset thresholds for the corresponding anomaly type to determine whether the sampling point belongs to an anomalous sampling point. Cluster analysis is performed on consecutive anomalous sampling points to merge them into an anomalous region, thus clarifying the type of the anomalous region. Based on the three-dimensional coordinates (latitude, longitude, and altitude) of the sampling points within the anomalous region, combined with voxel grid distribution and trajectory height layering information, the spatial location range of the anomalous region is determined. The boundary coordinates, coverage area, and altitude layer of the anomalous region are marked, and associated with the corresponding voxel grid and base station coverage area to accurately locate the specific position of the anomalous region.

[0156] Step S4043: Based on the feature fusion data of the abnormal area and the spatial location of the abnormal area, a machine learning fusion diagnostic model is used to determine the root cause of the abnormality in the low-altitude communication network; wherein, the root cause of the abnormality includes at least one of the following: environmental obstruction, unreasonable base station configuration, communication interference, insufficient base station resources, beam coverage deviation, and signal propagation characteristic attenuation.

[0157] Extract feature fusion data of abnormal areas, combine the spatial location of abnormal areas (such as whether they are in areas blocked by tall buildings or at the edge of base stations), analyze the correlation between various features in the feature fusion data and the root causes of the anomalies, output the root cause results of the anomalies, and identify one or more root causes corresponding to the abnormal areas.

[0158] The AI-based low-altitude communication testing optimization method provided by this invention performs multi-dimensional fusion processing of communication test data and environmental perception data to construct feature fusion data. Then, based on a preset signal quality judgment threshold, it accurately identifies various abnormal areas such as weak coverage areas, interference areas, abnormal handover areas, and insufficient rate areas from the feature fusion data, and simultaneously locks the spatial location corresponding to the abnormal areas. Then, combining the feature fusion data and spatial location information of the abnormal areas, it uses a machine learning fusion diagnostic model to intelligently judge and output the root cause results of the anomalies, including environmental obstruction, unreasonable base station configuration, communication interference, insufficient base station resources, beam coverage deviation, and signal propagation characteristic attenuation. This achieves accurate positioning of abnormal areas in low-altitude communication networks and intelligent and refined identification of fault root causes, effectively improving the accuracy of anomaly root causes.

[0159] Furthermore, by constructing a three-dimensional environment model, quantifying the impact of occlusion attenuation, locating and tracking interference sources, and establishing a correlation model, deep integration of communication data with three-dimensional environment features, occlusion attenuation features, and interference impact features was achieved. This not only made up for the shortcomings of single data dimensions and the inability to correlate environment and interference factors, but also made the data more correlated and complete through spatiotemporal alignment fusion, significantly improving the information richness and accuracy of feature fusion data.

[0160] In one possible implementation, in order to enable accurate low-altitude communication, step S105 may include the following steps: Step d1: Based on the root cause results of the low-altitude communication network anomalies, determine the supplementary test area and the corresponding supplementary test 3D trajectory from the test area.

[0161] The supplementary testing area can be understood as an abnormal or suspected abnormal area that, based on the abnormal root cause results, is determined by analysis to have an unclear root cause or insufficient data support.

[0162] The supplementary 3D trajectory can be understood as a drone flight trajectory specifically planned for the supplementary testing area and adapted to the supplementary testing requirements.

[0163] Specifically, the output of the abnormal root cause results is retrieved, and combined with feature fusion data and a high-precision 3D environmental point cloud model, the root cause supporting data for each abnormal region is analyzed one by one to determine whether it is sufficient. If the root cause is ambiguous, the data is missing, or the abnormal boundary is ambiguous, the region is marked as a supplementary test region. Based on the spatial characteristics of the supplementary test region (such as the area blocked by high-rise buildings, the area around the interference source, and the area at the edge of the base station coverage) and the root cause type (such as the obstruction type, the interference type, and the base station configuration type), a supplementary 3D trajectory is planned. The trajectory sampling point density is higher than that of the adaptive test trajectory mentioned above. After covering the root cause-related points (such as the edge of the obstruction, the area near the interference source, and the base station beam coverage boundary) and verifying the overlap of sampling points (to avoid duplicate sampling), the final supplementary 3D trajectory is determined.

[0164] Step d2: Control multiple UAVs to perform collaborative supplementary measurements along the supplementary 3D trajectory and acquire supplementary measurement data.

[0165] Specifically, the ground control center issues supplementary three-dimensional trajectory and collaborative supplementary measurement instructions, controlling multiple UAVs to fly collaboratively according to the supplementary three-dimensional trajectory; at a higher sampling frequency, it collects basic performance data and air interface signaling data of the supplementary measurement area; and simultaneously collects environmental perception data and UAV status data (flight altitude, attitude, speed) to obtain supplementary measurement data.

[0166] Step d3 involves integrating the supplementary test data into the feature fusion data for secondary root cause verification to obtain the target anomaly root cause results.

[0167] The standardized supplementary test data is integrated into the feature fusion data to supplement key information such as signal quality data, interference signal details, and obstruction information in the supplementary test area. The core features in the feature fusion data (such as attenuation value and interference intensity) are updated. The updated feature fusion data and the spatial location of the abnormal area are then input into the machine learning fusion diagnostic model to perform root cause analysis again and obtain the root cause results of the target anomaly.

[0168] Step d4: Based on the root cause results of the target anomaly, generate the optimization results corresponding to the root cause results of the target anomaly.

[0169] For different types of target anomaly root cause results, generate corresponding optimization results.

[0170] For example, environmental obstruction issues include: optimizing drone flight paths (avoiding highly obstructed areas) and adding signal relay equipment; communication interference issues include: shielding invalid interference sources, adjusting base station frequency bands, and optimizing interference suppression algorithms; base station configuration issues include: adjusting base station transmit power, beamforming parameters, and cell coverage; base station resource issues include: expanding base station RB resource blocks and balancing cell load; beam coverage issues include: adjusting base station beam direction, optimizing beam coverage, and adapting to drone flight paths; and signal propagation attenuation issues include: adopting signal enhancement technology and optimizing frequency band selection (avoiding high attenuation frequency bands).

[0171] Step d5: Based on the optimization results, test and optimize low-altitude communication.

[0172] Following the implementation steps and parameter adjustment ranges in the optimization results, optimize the low-altitude communication network (e.g., adjust base station parameters, add signal relay equipment, shield interference sources, and optimize UAV flight trajectories). After optimization, control multiple UAVs to conduct collaborative testing again along the original test trajectory and the supplementary test area trajectory, and collect test data. Compare the data from the second test with the feature fusion data and supplementary test data before optimization to verify the effectiveness of the optimization measures. If the optimization meets the standards, the test optimization is completed. If the optimization does not meet the standards, analyze the reasons for the failure, adjust the optimization results, re-implement the optimization, and re-test until the optimization meets the standards, forming a complete test optimization closed loop.

[0173] The AI-based low-altitude communication testing optimization method provided by this invention accurately determines the retesting area based on the root cause results of anomalies in the low-altitude communication network, and plans the corresponding three-dimensional retesting trajectory for the retesting area to ensure that the retesting direction conforms to the characteristics of the anomaly area. Subsequently, multiple UAVs are controlled to fly collaboratively along the retesting three-dimensional trajectory, synchronously collecting retesting data. The retesting data is then integrated into the previous feature fusion data to complete the secondary verification of the anomaly root cause, further calibrating the accuracy of the root cause judgment. Finally, based on the target anomaly root cause after secondary verification, a targeted optimization scheme is generated to achieve precise optimization and adjustment of low-altitude communication.

[0174] In one possible implementation, as shown in Table 1, the ground control center outputs optimization suggestions in two categories based on the root cause diagnosis results: real-time temporary optimization suggestions (which can be implemented remotely through base stations without on-site construction) and long-term precise optimization suggestions (which require on-site construction / hardware adjustment). Specific suggestions are formulated for different root cause categories.

[0175] Table 1 More specifically, the ground control center organizes the test data from the entire process into a data storage list, including: basic information, test parameters, 3D test trajectory, precise full-dimensional data sets, 3D environmental point cloud model, interference source location data, root cause diagnosis results, optimization suggestions, and optimization verification effects. The ground control center hashes and encrypts the data storage list to generate a unique hash value, and then sends the hash value and the original data to the blockchain storage node. The blockchain storage node adopts a consortium blockchain architecture, including nodes such as the tester, base station operator, and third-party testing agency, to ensure the impartiality of the storage. The blockchain storage node writes the storage data into the blockchain block to complete the tamper-proof storage and generates a storage certificate for the tester. The certificate includes the storage time, data hash value, blockchain node information, etc., which can realize the traceability and authenticity verification of the test data.

[0176] The ground control center compiles the test data, analysis results, optimization suggestions, and evidence information from the entire process to generate a smart test report for 5G network low-altitude coverage. The report includes test overview, test data, environmental analysis, root cause diagnosis, optimization suggestions, optimization effect prediction, and data evidence. The ground control center then archives the test report, visualization report, and evidence certificate in a unified manner and sends them to relevant units such as base station operators and testers, completing the entire testing and optimization process.

[0177] In some embodiments, the AI-based low-altitude communication test optimization device of the present invention can be implemented in a combination of hardware and software. As an example, the AI-based low-altitude communication test optimization device of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the AI-based low-altitude communication test optimization method of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0178] The modules described in the embodiments of this invention can be implemented in software or hardware. The names of the modules are not, in some cases, limiting the scope of the module itself.

[0179] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the aforementioned artificial intelligence-based low-altitude communication test optimization methods. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the artificial intelligence-based low-altitude communication test optimization method shown in any embodiment of the present invention by calling the computer program.

[0180] In one alternative embodiment, an electronic device is provided, such as Figure 5 As shown, Figure 5The illustrated electronic device 5000 includes a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, for example, via a bus 5002. Optionally, the electronic device 5000 may further include a transceiver 5004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 5004 is not limited to one type, and the structure of the electronic device 5000 does not constitute a limitation on the embodiments of the present invention.

[0181] Processor 5001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 5001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0182] Bus 5002 may include a path for transmitting information between the aforementioned components. Bus 5002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 5002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus 5002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0183] The memory 5003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0184] The memory 5003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 5001. The processor 5001 executes the application code stored in the memory 5003 to implement the content shown in the foregoing method embodiments.

[0185] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0186] It should be noted that, Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0187] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the aforementioned artificial intelligence-based low-altitude communication test optimization methods.

[0188] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0189] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned AI-based low-altitude communication test optimization method.

[0190] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0191] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0192] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0193] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0194] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0195] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0196] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a circuit, module, or system. Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0197] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for optimizing low-altitude communication testing based on artificial intelligence, characterized in that, The method includes: Acquire relevant data for the test area; wherein, the relevant data is used to characterize the influencing factors of low-altitude communication; Based on the relevant data, the test area is divided into three-dimensional spaces and marked with occlusion to generate an adaptive test trajectory; Multiple drones are controlled to fly collaboratively according to the adaptive test trajectory to determine communication test data and environmental perception data; wherein, the communication test data and environmental perception data are obtained when the drones fly collaboratively. Based on the communication test data and environmental perception data, the root causes of anomalies in the low-altitude communication network are determined. Based on the root cause results of the anomalies in the low-altitude communication network, the low-altitude communication was tested and optimized.

2. The artificial intelligence-based low-altitude communication test optimization method according to claim 1, characterized in that, The relevant data includes terrain and feature data and base station configuration parameters. Based on the relevant data, the test area is divided into three-dimensional spaces and marked with occlusion markers to generate an adaptive test trajectory, including: Based on the terrain and feature data and base station configuration parameters, the test area is divided into a three-dimensional voxel grid; Based on the simulated base station signal propagation path and the terrain and feature data, the occlusion state of the three-dimensional voxel grid is marked to obtain the occlusion state of the three-dimensional voxel grid; wherein, the simulated base station signal propagation path is the propagation path from the base station under test to the center of the three-dimensional voxel grid. Based on the occlusion state of the three-dimensional voxel grid and the base station signal propagation path, the signal propagation attenuation value of the three-dimensional voxel grid is determined; An adaptive test trajectory is generated based on the occlusion state of the three-dimensional voxel mesh, the signal propagation attenuation value of the three-dimensional voxel mesh, and the preset multi-UAV cooperative flight constraints.

3. The low-altitude communication test optimization method based on artificial intelligence according to claim 2, characterized in that, The adaptive test trajectory is generated based on the occlusion state of the three-dimensional voxel mesh, the signal propagation attenuation value of the three-dimensional voxel mesh, and the preset multi-UAV cooperative flight constraints, including: Based on the occlusion state of the three-dimensional voxel mesh, the signal propagation attenuation value of the three-dimensional voxel mesh, and the preset multi-UAV cooperative flight constraints, an initial test trajectory is generated; wherein, the initial test trajectory is composed of multiple sampling points connected together; If the initial test trajectory involves multiple UAV collisions and / or overlapping sampling points, the initial test trajectory is adjusted to obtain the target test trajectory. Based on the target test trajectory, determine the data missing rate when the UAV flies along the initial test trajectory; If the data missing rate is not greater than the missing rate threshold, the target test trajectory is determined as the adaptive test trajectory; If the data missing rate is greater than the missing rate threshold, the step of adjusting the initial test trajectory is performed until the data missing rate is no greater than the missing rate threshold.

4. The low-altitude communication test optimization method based on artificial intelligence according to claim 1, characterized in that, The control of multiple drones to fly collaboratively according to the adaptive test trajectory to determine communication test data and environmental perception data includes: During the process of controlling multiple UAVs to fly collaboratively according to the adaptive test trajectory, basic performance data, air interface signaling data, base station side data, UAV status data, and environmental point cloud data of the test area are collected from the sampling points. Based on the environmental point cloud data of the test area, determine the environmental perception data; Based on the UAV status data, error compensation is performed on the basic performance data to obtain the target performance data; Based on the target performance data, the air interface signaling data, the base station data, and the UAV status data, communication test data is determined.

5. The low-altitude communication test optimization method based on artificial intelligence according to claim 4, characterized in that, The UAV status data includes flight speed, attitude angle, and flight altitude. Based on the UAV status data, error compensation is performed on the basic performance data to obtain the target performance data, including: Based on the flight speed and the preset uniform flight speed, the signal power in the basic performance data is compensated for speed using a speed compensation model. Based on the attitude angle and the preset attitude angle, the attitude compensation model is used to perform pitch attitude compensation, roll attitude compensation and yaw attitude compensation on the signal power. Based on the flight altitude and the preset flight altitude, the signal power is compensated for altitude using an altitude fluctuation compensation model to obtain the compensated target performance data.

6. The low-altitude communication test optimization method based on artificial intelligence according to claim 1, characterized in that, Based on the aforementioned communication test data and environmental perception data, the root causes of anomalies in the low-altitude communication network are determined, including: The communication test data and environmental perception data are fused in multiple dimensions to obtain feature fusion data; Based on a preset signal quality judgment threshold, abnormal regions and their spatial locations are identified from the feature fusion data; wherein, the abnormal regions include weak coverage regions, interference regions, abnormal handover regions, and regions with insufficient data rate. Based on the feature fusion data and spatial location of the abnormal area, a machine learning fusion diagnostic model is used to determine the root cause of the anomaly in the low-altitude communication network. The root cause includes at least one of the following: environmental obstruction, unreasonable base station configuration, communication interference, insufficient base station resources, beam coverage deviation, and signal propagation characteristic attenuation.

7. The low-altitude communication test optimization method based on artificial intelligence according to claim 6, characterized in that, The multi-dimensional data fusion of the communication test data and environmental perception data to obtain feature fusion data includes: Based on the environmental perception data, a three-dimensional environmental point cloud model of the test area is constructed; Based on the materials of different occluders in the 3D environment point cloud model, the signal attenuation value corresponding to the material is quantified to obtain the environment occlusion-signal attenuation correlation model. Based on the characteristics of interference signals in the communication test data, the effective interference sources are located and tracked in three dimensions, and the interference types are distinguished. Based on the type of interference, establish an interference source-signal quality correlation model; The communication test data, the three-dimensional environmental point cloud model, the environmental occlusion-signal attenuation correlation model, and the interference source-signal quality correlation model are fused in a multi-dimensional spatiotemporal manner to obtain feature fusion data.

8. The low-altitude communication test optimization method based on artificial intelligence according to claim 6, characterized in that, The testing and optimization of the low-altitude communication network based on the root cause analysis of the anomalies includes: Based on the root cause results of the anomalies in the low-altitude communication network, a supplementary testing area and the corresponding three-dimensional trajectory of the supplementary testing area are determined from the test area. Control multiple UAVs to perform collaborative supplementary measurements along the supplementary 3D trajectory and acquire supplementary measurement data; The supplementary test data is integrated into the feature fusion data for secondary root cause verification to obtain the target anomaly root cause result; Based on the target anomaly root cause results, generate the optimization results corresponding to the target anomaly root cause results; Based on the optimization results, the low-altitude communication was tested and optimized.

9. A low-altitude communication testing and optimization device based on artificial intelligence, characterized in that, The device includes: The acquisition module is used to acquire relevant data of the test area; wherein, the relevant data is used to characterize the influencing factors of low-altitude communication; The processing module is used to perform three-dimensional spatial division and occlusion marking of the test area based on the relevant data, and generate an adaptive test trajectory; The first determining module is used to control multiple UAVs to fly collaboratively according to the adaptive test trajectory in order to determine communication test data and environmental perception data; wherein, the communication test data and environmental perception data are obtained when the UAVs fly collaboratively. The second determining module is used to determine the root cause of anomalies in the low-altitude communication network based on the communication test data and environmental perception data. The optimization module is used to test and optimize the low-altitude communication based on the root cause results of the anomalies in the low-altitude communication network.

10. An electronic device, characterized in that, include: A processor and a memory connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to cause the processor to perform the AI-based low-altitude communication test optimization method as described in any one of claims 1-8.

Images