Distributed deployment method and system for three-dimensional wireless communication network of unmanned aerial vehicle

Abstract

The invention provides a distributed deployment method and system for an unmanned aerial vehicle three-dimensional wireless communication network, and relates to the field of unmanned aerial vehicle communication networks, and the method comprises the steps: obtaining the geographic information of an inspection region, constructing a three-dimensional terrain matrix, and carrying out the self-adaptive resolution voxelization; based on a voxelization model and unmanned aerial vehicle communication parameters, considering path loss and obstacle penetration loss, and constructing a three-dimensional communication quality potential field; designing an anti-collision repulsion potential field and a communication blind area gravitational potential field, and obtaining a total potential field gradient of unmanned aerial vehicle inspection; based on the real-time information, updating the three-dimensional position of each unmanned aerial vehicle in a distributed manner by adopting a gradient descent method; determining the responsibility coverage area of each unmanned aerial vehicle according to the updated position, and cooperatively determining the beam direction and the beam width; and defining a network coverage efficiency and communication quality balance index, and dynamically adjusting a learning rate to complete real-time online optimization. According to the invention, autonomous deployment and optimization of the unmanned aerial vehicle communication network can be realized, and communication coverage efficiency and quality balance are improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) communication networks, specifically to a distributed deployment method and system for a three-dimensional wireless communication network for UAVs. Background Technology

[0002] With the rapid development of drone technology, drones are increasingly being used in fields such as power line inspection, emergency communication, and disaster relief. In complex terrain environments, building an efficient and stable three-dimensional wireless communication network for drones has become a key technology for ensuring collaborative operations of drone swarms. Currently, drone communication network deployment mainly adopts methods based on preset waypoints or fixed coverage patterns, which are difficult to adapt to the demands of complex and ever-changing three-dimensional environments.

[0003] In the area of ​​UAV inspection terrain modeling, existing technologies typically use uniform grids or simplified models to represent terrain features. For example, prior art with publication number CN120445226A discloses a UAV flight path planning method for hydropower projects. This method generates a 3D terrain mesh model corresponding to the target operation area, performs voxelization to obtain an initial voxel map, and determines candidate waypoints based on the initial voxel map. Prior art with publication number CN119047778A proposes an equipment layout method for UAV countermeasures equipment. This method reconstructs 3D buildings in the target area, determines the passable area, performs spatial voxelization, and then traverses the voxelized passable area using UAV flight paths. However, these methods typically use fixed-resolution voxel partitioning, making it difficult to adaptively adjust for terrain complexity and key areas, resulting in wasted computational resources or insufficient accuracy.

[0004] In the area of ​​UAV cooperative obstacle avoidance and communication, prior art (CN120652993A) discloses a multi-UAV cooperative obstacle avoidance control method based on a hierarchical intelligent mechanism of pigeon flock leadership. This method draws on the intelligent decision-making process of pigeon flocks facing predators, where higher-level leaders lead lower-level leaders to avoid collisions. It designs a leader obstacle avoidance and trajectory tracking switching controller and establishes a collision avoidance control protocol based on logarithmic potential field gradients. Prior art (CN120445223A) proposes a UAV heterogeneous swarm cooperative navigation method for complex terrain. This method dynamically reconstructs the radiation direction of the airborne antenna of selected nodes based on the distribution of communication attenuation areas, establishing directional communication links around terrain obstacles. While these methods have made some progress in obstacle avoidance and communication, most adopt a centralized control architecture, relying on a central node for decision-making, resulting in poor system robustness and insufficient consideration of the balance between communication quality and network coverage optimization. However, existing UAV communication network deployment methods generally suffer from the following problems: terrain modeling is coarse, mostly using fixed-resolution grids or voxels, which cannot adaptively adjust according to terrain complexity and the importance of key areas, while ignoring the impact of obstacle penetration loss and terrain gradient changes on communication quality; there is a lack of effective anti-collision and anti-interference coordination mechanisms between UAVs, especially in high-density deployment scenarios, where mutual interference between UAVs is prominent, affecting overall network performance; beam control and network coverage lack real-time autonomous optimization capabilities, making it difficult to adjust communication resource allocation according to dynamically changing environments and mission requirements, leading to uneven coverage or fluctuating communication quality; most methods adopt a centralized computing architecture, relying on a central node for decision-making and control, and once the central node fails, the entire system faces the risk of collapse, exhibiting poor robustness.

[0005] Therefore, there is an urgent need for a method for deploying UAV communication networks capable of distributed autonomous decision-making, adapting to three-dimensional terrain, and optimizing coverage and balance in real time, in order to improve the communication support capabilities and system reliability of UAV inspection operations. The information disclosed in the background section is only for enhancing the understanding of the background of this disclosure and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a distributed deployment method and system for a three-dimensional wireless communication network for unmanned aerial vehicles (UAVs) to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A distributed deployment method for a UAV three-dimensional wireless communication network includes the following steps: S1: Obtain geographic information of the UAV inspection area, construct a three-dimensional terrain matrix of the inspection area, perform adaptive resolution voxelization on the three-dimensional terrain matrix, divide it into voxels of unequal size, and the set of voxels constitutes the voxelized model of the inspection area. S2: Based on the voxelized model of the inspection area and the communication parameters of the UAV, the communication quality index between each voxel and the UAV is obtained considering the loss, and a three-dimensional communication quality potential field is constructed for each voxel.

[0008] S3: Design a repulsive potential field for collision prevention and interference resistance, as well as a gravitational potential field for communication blind zone requirements. Based on the above three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, obtain the gradient of the total potential field for UAV inspection. S4: Based on the real-time inspection information of each UAV, calculate the gradient of the total potential field of the UAV inspection at its location, and update the three-dimensional position of each UAV in a distributed manner using the gradient descent method. S5: Based on the updated 3D positions of each UAV, determine the coverage area of ​​each UAV, and accordingly determine the beam direction and beam width. S6: Define the network coverage performance and communication quality balance index, and dynamically adjust the learning rate in S4 based on the above index to complete the real-time online optimization of UAV inspection.

[0009] Furthermore, the 3D terrain matrix is ​​adaptively voxelized at resolution, with the specific dynamic voxel size function as follows: ; in, For spatial points The side length of the voxel cube at that location. Based on resolution, These are the terrain gradient weighting coefficients. For spatial points The gradient vector of terrain height change. This is the weighting coefficient for the distance between power towers. For spatial points The normalized distance to the nearest power tower is calculated; the area is divided into sets of voxels of unequal size, which form a voxelized model of the inspection area. Specifically, the division is accomplished by executing a spatial partitioning algorithm, using the smallest voxel size as the initial partition, and then applying this algorithm to each candidate voxel. Calculate its center point place If the voxel size is larger than Then, octree subdivision is performed, and the process is iterated until all voxels meet the size constraints and the subdivision is completed. Finally, a voxel set is generated, and each voxel contains attributes: center coordinates, size, obstacle identifier, and communication requirement weight. The voxel set constitutes the voxelized model of the inspection area.

[0010] Furthermore, based on the voxelized model of the inspection area and the UAV communication parameters, the communication quality index between each voxel and the UAV is obtained considering losses; the UAV communication parameters include: UAV transmission power and UAV position; the losses include: path loss and obstacle penetration loss; wherein, the path loss calculation formula is as follows: ; in, voxels With drones Euclidean distance between them For path loss, For reference distance Path loss, This is the path loss index. The standard deviation of shadow fading is given; the formula for calculating obstacle penetration loss is as follows: ; in, For obstacle penetration loss, For the first Loss coefficient per unit thickness of obstacles For the ray to pass through the first The thickness of the obstacle-like object This represents the total number of obstacles.

[0011] Furthermore, based on the voxelized model of the inspection area and the UAV communication parameters, the communication quality index between each voxel and the UAV is obtained, considering losses. The UAV communication parameters include: UAV transmission power and UAV position. The losses include: path loss and obstacle penetration loss. The communication quality index between each voxel and the UAV is obtained, and the calculation formula for the communication quality index is as follows:

[0012] in, voxels With drones Inter-communication quality index, For the drone's transmission power, voxels With drones Euclidean distance between them For path loss, The cumulative value of obstacle penetration loss is used to construct a three-dimensional communication quality potential field for each voxel based on the obtained communication quality index, as shown in the following formula: ; in, voxels The communication quality potential field value at that location, This is the collection of all drones in the current network. is the potential field attenuation coefficient.

[0013] Furthermore, the anti-collision and anti-interference repulsion potential field and the gravitational potential field required for communication blind zones are designed. The formula for the anti-collision and anti-interference repulsion potential field is as follows: ; in, For drones Affected by drones Repulsive force, To repel potential amplitude, For drones With drones Three-dimensional position vector, This is the repulsion range coefficient. For the safe interval distance; the gravitational potential field formula required for the communication blind zone is as follows: ; in, voxels The appeal of drones For the attractive potential coefficient, voxels Communication demand weight, voxels by drone The degree of coverage; its calculation formula is as follows: ; in, This is the steepness coefficient. This is the communication quality threshold.

[0014] Furthermore, based on the three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, the gradient of the total potential field for UAV inspection is obtained: for each UAV The gradient is calculated based on local information, and the specific formula is as follows: ; in, For drones In position The gradient vector of the total potential field at that location. For drones The neighborhood group, For drones A set of voxels within the range of perception.

[0015] Furthermore, the gradient descent method is used to update the 3D position of each UAV in a distributed manner. The specific formula is as follows: ; in, To determine the three-dimensional position of the drone at the next moment, For drones The position at the next iteration. For learning rate, To prevent dividing by zero from resulting in a small positive number.

[0016] Furthermore, based on the updated 3D positions of each UAV, the coverage area of ​​each UAV is determined, and the beam pointing and beamwidth are collaboratively determined accordingly. The formula for determining the service area of ​​a UAV is as follows: ; in, For drones Area of ​​responsibility It is a set of voxels. voxels The three-dimensional position vector; the beam pointing is determined by calculating the weighted centroid direction of the responsibility area, as shown in the following formula: ; in, The beamwidth is determined by the unit direction vector pointing to the center of the main lobe of the beam; the beamwidth is determined based on the spatial distribution of the area of ​​responsibility, and the area of ​​responsibility is calculated first. The average terrain elevation within the area is calculated using the following formula: ; Tool The specific formula for determining the beamwidth is as follows: ; in, The main lobe width of the beam. for and Horizontal distance, For drones The three-dimensional ordinate, Area of ​​responsibility The average terrain elevation within.

[0017] Furthermore, an index for the balance between network coverage effectiveness and communication quality is defined, and the formula for calculating the balance between network coverage effectiveness and communication quality is as follows: ; ; in, For network coverage efficiency, To ensure communication quality balance, The total number of drones, For drones Responsibility coverage area; the learning rate in S4 is dynamically adjusted based on the balance between network coverage effectiveness and communication quality, using the following formula: ; in, The learning rate for the next time step. The initial baseline value for the learning rate, This is the network coverage performance adjustment coefficient. To improve network coverage performance, For the first Network coverage performance at the next iteration time This is the communication quality balance adjustment coefficient. To achieve the desired communication quality balance, No. The consistency of communication quality at each iteration.

[0018] This invention also provides a distributed deployment system for a UAV three-dimensional wireless communication network, the distributed deployment system for the UAV three-dimensional wireless communication network being used to execute the above-described distributed deployment method for the UAV three-dimensional wireless communication network, comprising: Terrain Voxelization Module: Used to acquire geographic information of the UAV inspection area, construct a three-dimensional terrain matrix of the inspection area, perform adaptive resolution voxelization on the three-dimensional terrain matrix, divide it into voxels of unequal size, and the set of voxels constitutes the voxelization model of the inspection area. The communication quality potential field construction module is used to obtain the communication quality index between each voxel and the UAV based on the voxelized model of the inspection area and the communication parameters of the UAV, taking into account the loss, and to construct the three-dimensional communication quality potential field of each voxel.

[0019] Total potential field gradient module: used to design the repulsive potential field for collision prevention and interference resistance, as well as the gravitational potential field for communication blind zone requirements. Based on the above three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, the total potential field gradient of the UAV inspection is obtained. Distributed position optimization module: Based on the real-time inspection information of each UAV, it calculates the gradient of the total potential field of the UAV inspection at its location, and updates the three-dimensional position of each UAV in a distributed manner using the gradient descent method. Beam coordination control module: used to determine the coverage area of ​​each UAV based on the updated 3D position of each UAV, and to coordinately determine the beam direction and beam width accordingly; Online optimization module: used to define the balance index between network coverage performance and communication quality, and dynamically adjust the learning rate in S4 based on the above index to complete the real-time online optimization of UAV inspection.

[0020] Compared with existing technologies, the advantages of this invention are as follows: Through adaptive terrain modeling and dynamic voxelization technology, the voxel resolution is automatically adjusted according to terrain gradients and the locations of key facilities, improving modeling accuracy and computational efficiency in complex terrains; a distributed autonomous decision-making mechanism is adopted, where each UAV relies only on local information for position updates, eliminating the need for central node coordination, resulting in strong system robustness and adaptability to communication interruption scenarios; multiple potential fields, including communication quality, collision avoidance, and coverage blind spots, are integrated to achieve safe and efficient coverage, balancing communication quality and network security; beam pointing and width are dynamically adjusted according to the area of ​​responsibility, improving spectrum utilization efficiency and coverage quality; and the learning rate is dynamically optimized based on coverage effectiveness and balance, enabling real-time optimization of network performance and adapting to changes in network status. Compared with traditional methods, this invention provides a more accurate, robust, and efficient UAV communication network deployment solution in complex terrain environments. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a block diagram of the system module structure of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0023] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0024] Example: Please see Figure 1 The present invention provides a technical solution: A distributed deployment method for a UAV three-dimensional wireless communication network includes the following steps: S1: Obtain geographic information of the UAV inspection area, construct a three-dimensional terrain matrix of the inspection area, perform adaptive resolution voxelization on the three-dimensional terrain matrix, divide it into voxels of unequal size, and the set of voxels constitutes the voxelized model of the inspection area. In this embodiment, a digital elevation model of the inspection area, the latitude and longitude coordinates of the power towers, and the building distribution vector data are obtained through a geographic information system. The raw data undergoes coordinate system 1, noise filtering, and resolution normalization to construct a three-dimensional terrain matrix. Data sources include satellite remote sensing, airborne lidar scanning, and existing digital topographic maps. To balance modeling accuracy and computational complexity, a dynamic voxel size function is defined to adaptively voxelize the three-dimensional terrain matrix. The specific dynamic voxel size function is as follows: ; in, For spatial points The side length of the voxel cube at that location. Based on resolution, These are the terrain gradient weighting coefficients. For spatial points The gradient vector of terrain height change. This is the weighting coefficient for the distance between power towers. For spatial points The normalized distance to the nearest power tower is used; the area is divided into sets of voxels of unequal size, which together form a voxelized model of the inspection area; the formula design considers 1 to ensure the basic resolution. ; To measure the degree of terrain undulation, the greater the gradient, the steeper the terrain. The smaller the size, the finer the voxel, allowing for accurate modeling of shadow areas and changes in viewing distance. Control terrain sensitivity and increase This will make the algorithm more sensitive to terrain and generate more small voxels in undulating areas; spatial point The normalized distance to the nearest power tower is used because power towers are key communication nodes or key inspection targets. The channel environment near them is complex and requires detailed modeling. This distance is usually normalized. Increase the area of ​​influence of the control tower This will allow for higher resolution usage in a larger area around the control tower. In this embodiment, The reason for choosing 10 meters is that it is close to the minimum safe distance for drones, balances calculation accuracy and efficiency, and meets the basic requirements of power line inspection for the accuracy around the tower. ; The partitioning is accomplished by executing a spatial partitioning algorithm, using the smallest voxel size as the initial partition, and then partitioning each candidate voxel. Calculate its center point place If the voxel size is larger than Then, octree subdivision is performed, and the process is iterated until all voxels meet the size constraints and the partitioning is completed. Finally, a voxel set is generated, and each voxel contains attributes: center coordinates, size, obstacle identifier, and communication requirement weight. The voxel set constitutes the voxelized model of the inspection area. Traditional methods use fixed-resolution grids, resulting in computational redundancy in flat areas and insufficient accuracy in complex regions. This method dynamically adjusts voxel sizes based on terrain gradients and operational requirements, achieving an optimal balance between accuracy and efficiency. It employs octree adaptive subdivision, avoiding uniform high-resolution partitioning of the entire space and significantly reducing computational and storage requirements.

[0025] S2: Based on the voxelized model of the inspection area and the communication parameters of the UAV, the communication quality index between each voxel and the UAV is obtained considering the loss, and a three-dimensional communication quality potential field is constructed for each voxel. In this embodiment, based on the voxelized model of the inspection area and the UAV communication parameters, the communication quality index between each voxel and the UAV is obtained considering losses. The UAV communication parameters include: UAV transmission power and UAV position. The losses include: path loss and obstacle penetration loss. The path loss calculation formula is as follows: ; in, voxels With drones Euclidean distance between them For path loss, For reference distance Path loss, This is the path loss index. The standard deviation of shadow fading is given; a classic log-distance model for shadow fading is used, where the first term is the baseline loss and the second term is the deterministic attenuation caused by distance. The third item For random variables that follow a Gaussian distribution, simulate random fluctuations caused by obstructions.

[0026] The formula for calculating obstacle penetration loss is as follows: ; in, For obstacle penetration loss, For the first Loss coefficient per unit thickness of obstacles For the ray to pass through the first The thickness of the obstacle-like object The total number of obstacles. A simplified model using ray tracing. Along the line connecting the voxel and the drone, the product of the thickness of each type of obstacle traversed and its unit loss coefficient is accumulated. Determined by the electromagnetic properties of obstacles, in this embodiment, based on the voxelized model of the inspection area and the UAV communication parameters, the communication quality index between each voxel and the UAV is obtained, taking into account losses. The UAV communication parameters include: UAV transmission power and UAV position; the losses include: path loss and obstacle penetration loss; the communication quality index between each voxel and the UAV is obtained, and the formula for calculating the communication quality index is as follows: ; in, voxels With drones Inter-communication quality index, For the drone's transmission power, voxels With drones Euclidean distance between them For path loss, This represents the cumulative value of obstacle penetration loss. A larger value indicates a stronger received signal and better communication quality. A three-dimensional communication quality potential field is constructed for each voxel based on the obtained communication quality index, as shown in the following formula: ; in, voxels The communication quality potential field value at that location, This is the collection of all drones in the current network. is the potential field attenuation coefficient. The calculations demonstrate that the communication potential of each voxel is determined by the drone that best serves it. This is the distance attenuation factor. Even if a drone's signal is very strong, its influence on the voxel should attenuate if it is far away. Parameter Control the decay rate, The smaller the value, the faster the decay, meaning the drone only has a significant impact on the potential field of a very small area nearby, prompting drones to be distributed more dispersedly. A negative sign is used because, in terms of communication quality, the higher the better; when converted to potential energy, the lower the better.

[0027] Current UAV deployment methods often simplify communication quality as a negative correlation function of distance, or use a simple logarithmic distance path loss model, completely ignoring obstacle occlusion, a decisive factor affecting communication in complex terrain; mapping discrete voxel-UAV communication quality to a continuous potential field is a more practical approach. The potential field forms a topographic map in space: areas with good communication quality are potential energy troughs, and drones tend to move towards these areas; communication blind spots or weak areas are potential energy peaks, providing a natural mathematical framework for subsequent gradient-based optimization.

[0028] S3: Design a repulsive potential field for collision prevention and interference resistance, as well as a gravitational potential field for communication blind zone requirements. Based on the above three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, obtain the gradient of the total potential field for UAV inspection. In this embodiment, a collision-resistant and interference-resistant repulsive potential field and a gravitational potential field required for communication blind spots are designed. The formula for the collision-resistant and interference-resistant repulsive potential field is as follows: ; in, For drones Affected by drones Repulsive force, To repel potential amplitude, For drones With drones Three-dimensional position vector, This is the repulsion range coefficient. This is a safe distance; the system uses an exponential decay model to simulate interference / collision risk. When the distance is less than the safe distance... It generates a repulsive force, which increases exponentially with decreasing distance, ensuring that a large repulsive force is generated at very close distances to avoid collisions. Control the intensity of the repulsive force; Control the range of repulsive force. The smaller the value, the more concentrated the repulsive force becomes over very short distances. The formula for the gravitational potential field required in communication blind zones is as follows: ; in, voxels The appeal of drones For the attractive potential coefficient, voxels Communication demand weight, voxels by drone The degree of coverage; That's the key. The closer the value is to 1, the better the voxel is covered, and the gravitational pull approaches 0; conversely, if... If the weight is very small, then the gravitational force is very large. Allow for higher attraction priority to important areas, such as power towers; The calculation formula is as follows: ; in, This is the steepness coefficient. This is the communication quality threshold.

[0029] In this embodiment, the gradient of the total potential field for UAV inspection is obtained based on the three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field: for each UAV The gradient is calculated based on local information, and the specific formula is as follows: ; in, For drones In position The gradient vector of the total potential field at that location. For drones The neighborhood group, For drones A set of voxels within the perception range. Three gradient vectors are synthesized in three-dimensional space, and the final direction determines the direction of the resultant force for the UAV's next movement.

[0030] Existing methods often treat coverage, collision avoidance, and interference resistance separately, using sequential or weighted summation methods, which can easily lead to local optima or target conflicts; this step addresses the network coverage target. and Collision resistance and interference resistance, through The simulation unifies the process to the overall potential field. The drone only needs to follow the negative gradient of this overall potential field to simultaneously achieve all targets. This involves both repulsive and gravitational potential fields. The gravitational potential field actively attracts the drone to cover blind spots; the repulsive potential field generates a repulsive force when the drone gets too close, preventing collisions and interference with other drones operating at the same frequency. This mechanism is intuitive and has clear physical meaning.

[0031] S4: Based on the real-time inspection information of each UAV, calculate the gradient of the total potential field of the UAV inspection at its location, and update the three-dimensional position of each UAV in a distributed manner using the gradient descent method. In this embodiment, the gradient descent method is used to update the 3D position of each UAV in a distributed manner. The specific formula is as follows: ; in, To determine the three-dimensional position of the drone at the next moment, For drones The position at the next iteration. For learning rate, To prevent dividing by zero from resulting in a small positive number. It is the standard form of gradient descent, moving along the direction of the negative gradient, where This is a key improvement. It ensures that the update step size is... Regardless of the current gradient, the step size is fixed to prevent "overshooting". It is a very small positive number in this embodiment. To prevent division by zero error at zero gradient; learning rate These are the core control parameters. Too large a value will cause system oscillations and prevent convergence; too small a value will result in slow convergence, which is precisely why step six requires dynamic adjustment. The reason.

[0032] Compared to traditional centralized control or preset trajectories, centralized control requires transmitting all information back to a central node, resulting in single points of failure and communication bottlenecks, while preset trajectories cannot adapt to dynamic environments. In this step, each UAV relies only on local information: its own position, neighboring positions, and the potential field within its perception range; it can make independent decisions, exhibiting strong robustness and scalability. The system can autonomously reorganize itself when a new UAV joins or some UAVs fail. Furthermore, the gradient descent method, a classic optimization strategy, ensures that the system evolves towards lower potential energy under reasonable potential field design, theoretically guaranteeing convergence. The gradient is normalized in the update formula, ensuring that the displacement of each UAV step is determined by the learning rate. The step size is determined by the gradient direction, while the displacement direction is determined by the gradient direction. This avoids oscillations caused by excessively large step sizes in steep potential regions, thus improving the algorithm's stability. S5: Based on the updated 3D positions of each UAV, determine the coverage area of ​​each UAV, and accordingly determine the beam direction and beam width. In this embodiment, the coverage area of ​​each UAV is determined based on the updated three-dimensional position of each UAV, and the beam pointing and beam width are determined accordingly. The formula for determining the service area of ​​each UAV is as follows: ; in, For drones Area of ​​responsibility It is a set of voxels. voxels The three-dimensional position vector is used; the region is divided according to the nearest distance principle. Secondly, only those regions whose communication quality with the responsible UAV is above a threshold are retained. The voxels ensure that the area of ​​responsibility is a practically communicable area. The beam pointing is determined by calculating the weighted centroid direction of the area of ​​responsibility, as shown in the following formula: ; in, The unit direction vector points towards the center of the main lobe of the beam; the weighted centroid of all voxels in the area of ​​responsibility is calculated relative to the UAV, and then normalized. Weights This allows the drone's beam to be more biased towards important areas. The beamwidth is determined based on the spatial distribution of the area of ​​responsibility; the area of ​​responsibility is first calculated. The average terrain elevation within the area is calculated using the following formula: ; Tool The specific formula for determining the beamwidth is as follows: ; in, The main lobe width of the beam. for and Horizontal distance, For drones The three-dimensional ordinate, Area of ​​responsibility The average terrain height within the area. In this embodiment, the three-dimensional coverage problem is simplified to a two-dimensional sector coverage; if the area of ​​responsibility is very large, i.e., the numerator is large, or the drone flies very low, i.e., the denominator is small, the calculated average terrain height within the area is calculated. The amplitude can be very large, even exceeding the physical limits of the antenna array, at which point amplitude limiting is necessary.

[0033] Compared to traditional omnidirectional antennas or fixed beams: Omnidirectional antennas waste energy and cause significant interference. Fixed beams cannot adapt to dynamically changing network topologies and coverage requirements. In this step, the beam parameters—pointing and width—are not preset but dynamically calculated based on the actual coverage area of ​​the UAV, achieving closed-loop coordination between the network topology and the physical layer beam. This improves energy efficiency and reduces interference: the beam is precisely pointed to the centroid of its voxel set, concentrating energy in that area, thus improving signal strength for users and reducing interference in other directions. The beamwidth formula considers the relative relationship between the UAV's altitude and the average terrain of the coverage area, ensuring the beam covers its designated area, rather than simply pointing horizontally.

[0034] S6: Define the network coverage performance and communication quality balance index, and dynamically adjust the learning rate in S4 based on the above index to complete the real-time online optimization of UAV inspection.

[0035] In this embodiment, a network coverage performance and communication quality balance index is defined, and the calculation formulas for network coverage performance and communication quality balance are as follows: ; ; in, For network coverage efficiency, To ensure communication quality balance, The total number of drones, For drones Coverage area: measures the proportion of all voxels that are effectively covered. It is a sigmoid function that determines the continuous communication quality. Soft decisions have a coverage range of 0 to 1. The closer to 1, the better the network coverage. (Communication quality balance) This is an application of the Jain Fairness Index to communication quality, with values ​​ranging from... If the sum of the total communication quality provided by all drones for their respective voxels is exactly equal, then =1, absolutely fair. If one drone serves all the high-quality users, while other drones provide poor service, then... near This is extremely unfair. This indicator prevents the strong from becoming stronger and promotes a more balanced allocation of drone network resources. The learning rate in S4 is dynamically adjusted based on the balance between network coverage effectiveness and communication quality, using the following formula: ; in, The learning rate for the next time step. The initial baseline value for the learning rate, This is the network coverage performance adjustment coefficient. To improve network coverage performance, For the first Network coverage performance at the next iteration time This is the communication quality balance adjustment coefficient. To achieve the desired communication quality balance, No. The communication quality balance at each iteration. If the current performance is below the target, increase the learning rate. To accelerate improvement; if the current performance is higher than the target (negative deviation), the learning rate is slightly reduced. , These are the proportional gains for coverage and equalization bias, respectively. They determine which performance metric the system is more sensitive to.

[0036] This step compares with the fixed learning rate strategy: A fixed learning rate cannot adapt to the needs of different stages of the optimization process: large step sizes are needed for rapid exploration in the early stages, while small step sizes are needed for fine-tuning in the later stages; it also affects the overall network performance metrics. , As a feedback signal, the parameters of the underlying optimization algorithm are dynamically adjusted. This forms a complete closed loop of perception-decision-execution-evaluation. When the current performance deviates from the target, the learning rate is automatically increased to accelerate convergence; when it approaches the target, the learning rate is automatically decreased to avoid oscillations near the optimal solution. and The two coefficients essentially allow system designers to set the relative importance of two indicators: network coverage effectiveness and communication quality balance, and achieve the optimal trade-off under this weight by automatically adjusting the learning rate.

[0037] Please see Figure 2The present invention also provides a distributed deployment system for a UAV three-dimensional wireless communication network, the distributed deployment system for the UAV three-dimensional wireless communication network being used to execute the above-described distributed deployment method for the UAV three-dimensional wireless communication network, comprising: Terrain Voxelization Module: Used to acquire geographic information of the UAV inspection area, construct a three-dimensional terrain matrix of the inspection area, perform adaptive resolution voxelization on the three-dimensional terrain matrix, divide it into voxels of unequal size, and the set of voxels constitutes the voxelization model of the inspection area. The communication quality potential field construction module is used to obtain the communication quality index between each voxel and the UAV based on the voxelized model of the inspection area and the communication parameters of the UAV, taking into account the loss, and to construct the three-dimensional communication quality potential field of each voxel. Total potential field gradient module: used to design the repulsive potential field for collision prevention and interference resistance, as well as the gravitational potential field for communication blind zone requirements. Based on the above three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, the total potential field gradient of the UAV inspection is obtained. Distributed position optimization module: Based on the real-time inspection information of each UAV, it calculates the gradient of the total potential field of the UAV inspection at its location, and updates the three-dimensional position of each UAV in a distributed manner using the gradient descent method. Beam coordination control module: used to determine the coverage area of ​​each UAV based on the updated 3D position of each UAV, and to coordinately determine the beam direction and beam width accordingly; Online optimization module: used to define the balance index between network coverage performance and communication quality, and dynamically adjust the learning rate in S4 based on the above index to complete the real-time online optimization of UAV inspection.

[0038] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0039] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0040] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0041] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A distributed deployment method for a UAV three-dimensional wireless communication network, characterized in that, The specific steps include: S1: Obtaining geographic information of the UAV inspection area, constructing a three-dimensional terrain matrix of the inspection area, performing adaptive resolution voxelization on the three-dimensional terrain matrix, dividing it into voxels of unequal size, and the set of voxels constitutes the voxelized model of the inspection area. S2: Based on the voxelized model of the inspection area and the communication parameters of the UAV, the communication quality index between each voxel and the UAV is obtained considering the loss, and a three-dimensional communication quality potential field is constructed for each voxel. S3: Design a repulsive potential field for collision prevention and interference resistance, as well as a gravitational potential field for communication blind zone requirements. Based on the above three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, obtain the gradient of the total potential field for UAV inspection. S4: Based on the real-time inspection information of each UAV, calculate the gradient of the total potential field of the UAV inspection at its location, and update the three-dimensional position of each UAV in a distributed manner using the gradient descent method. S5: Based on the updated 3D positions of each UAV, determine the coverage area of ​​each UAV, and accordingly determine the beam direction and beam width. S6: Define the network coverage performance and communication quality balance index, and dynamically adjust the learning rate in S4 based on the above index to complete the real-time online optimization of UAV inspection.

2. The distributed deployment method of a UAV three-dimensional wireless communication network according to claim 1, characterized in that: Adaptive resolution voxelization is performed on the 3D terrain matrix, and the specific dynamic voxel size function is as follows: ； in, For spatial points The side length of the voxel cube at that location. Based on resolution, These are the terrain gradient weighting coefficients. For spatial points The gradient vector of terrain height change. This is the weighting coefficient for the distance between power towers. For spatial points The normalized distance to the nearest power tower is calculated; the area is divided into sets of voxels of unequal size, which form a voxelized model of the inspection area. Specifically, the division is accomplished by executing a spatial partitioning algorithm, using the smallest voxel size as the initial partition, and then applying this algorithm to each candidate voxel. Calculate its center point place If the voxel size is larger than Then, octree subdivision is performed, and the process is iterated until all voxels meet the size constraints and the subdivision is completed. Finally, a voxel set is generated, and each voxel contains attributes: center coordinates, size, obstacle identifier, and communication requirement weight. The voxel set constitutes the voxelized model of the inspection area.

3. The distributed deployment method of a UAV three-dimensional wireless communication network according to claim 1, characterized in that: Based on the voxelized model of the inspection area and the communication parameters of the UAV, the communication quality index between each voxel and the UAV is obtained considering the loss. The UAV communication parameters include: UAV transmission power and UAV position; the losses include: path loss and obstacle penetration loss; wherein, the path loss calculation formula is as follows: ； in, voxels With drones Euclidean distance between them For path loss, For reference distance Path loss, This is the path loss index. The standard deviation of shadow fading is given; the formula for calculating obstacle penetration loss is as follows: ； in, For obstacle penetration loss, For the first Loss coefficient per unit thickness of obstacles For the ray to pass through the first The thickness of the obstacle-like object This represents the total number of obstacles.

4. The distributed deployment method of a UAV three-dimensional wireless communication network according to claim 2, characterized in that: Based on the voxelized model of the inspection area and the communication parameters of the UAV, the communication quality index between each voxel and the UAV is obtained considering the loss. The UAV communication parameters include: UAV transmit power and UAV position; the losses include: path loss and obstacle penetration loss; the communication quality index between each voxel and the UAV is obtained, and the communication quality index is calculated using the following formula: ; in, voxels With drones Inter-communication quality index, For the drone's transmission power, voxels With drones Euclidean distance between them For path loss, The cumulative value of obstacle penetration loss is used to construct a three-dimensional communication quality potential field for each voxel based on the obtained communication quality index, as shown in the following formula: ; in, voxels The communication quality potential field value at that location, This is the collection of all drones in the current network. is the potential field attenuation coefficient.

5. The distributed deployment method of a UAV three-dimensional wireless communication network according to claim 4, characterized in that: The design incorporates a collision-resistant and interference-resistant repulsive potential field, as well as a gravitational potential field required for communication blind zones. The formula for the collision-resistant and interference-resistant repulsive potential field is as follows: ; in, For drones Affected by drones Repulsive force, To repel potential amplitude, For drones With drones Three-dimensional position vector, This is the repulsion range coefficient. For the safe interval distance; the gravitational potential field formula required for the communication blind zone is as follows: ； in, voxels The appeal of drones For the attractive potential coefficient, voxels Communication demand weight, voxels by drone The degree of coverage; its calculation formula is as follows: ; in, This is the steepness coefficient. This is the communication quality threshold.

6. The distributed deployment method of a UAV three-dimensional wireless communication network according to claim 5, characterized in that: Based on the three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, the gradient of the total potential field for UAV inspection is obtained: for each UAV The gradient is calculated based on local information, and the specific formula is as follows: ； in, For drones In position The gradient vector of the total potential field at that location. For drones The neighborhood group, For drones A set of voxels within the range of perception.

7. A distributed deployment method for a UAV three-dimensional wireless communication network according to claim 6, characterized in that: The gradient descent method is used to update the 3D position of each UAV in a distributed manner. The specific formula is as follows: ; in, To determine the three-dimensional position of the drone at the next moment, For drones The position at the next iteration. For learning rate, To prevent dividing by zero from resulting in a small positive number.

8. The distributed deployment method of a UAV three-dimensional wireless communication network according to claim 7, characterized in that: Based on the updated 3D positions of each UAV, the coverage area of ​​each UAV is determined, and the beam pointing and beamwidth are collaboratively determined accordingly. The formula for determining the service area of ​​a UAV is as follows: ； in, For drones Area of ​​responsibility It is a set of voxels. voxels The three-dimensional position vector; the beam pointing is determined by calculating the weighted centroid direction of the responsibility area, as shown in the following formula: ; in, The beamwidth is determined by the unit direction vector pointing to the center of the main lobe of the beam; the beamwidth is determined based on the spatial distribution of the area of ​​responsibility, and the area of ​​responsibility is calculated first. The average terrain elevation within the area is calculated using the following formula: ; Tool The specific formula for determining the beamwidth is as follows: ; in, The main lobe width of the beam. for and Horizontal distance, For drones The three-dimensional ordinate, Area of ​​responsibility The average terrain elevation within.

9. A distributed deployment method for a UAV three-dimensional wireless communication network according to claim 8, characterized in that: The following defines an index for the balance between network coverage effectiveness and communication quality, and the formulas for calculating this balance are as follows: ; ； in, For network coverage efficiency, For the sake of communication quality balance, The total number of drones, For drones Responsibility coverage area; the learning rate in S4 is dynamically adjusted based on the balance between network coverage effectiveness and communication quality, using the following formula: ； in, The learning rate for the next time step. The initial baseline value for the learning rate, This is the network coverage performance adjustment coefficient. To improve network coverage performance, For the first Network coverage performance at the next iteration time This is the communication quality balance adjustment coefficient. To achieve the desired communication quality balance, No. The consistency of communication quality at each iteration.

10. A distributed deployment system for a three-dimensional wireless communication network for unmanned aerial vehicles (UAVs), characterized in that, The distributed deployment system of the UAV three-dimensional wireless communication network is used to execute the distributed deployment method of the UAV three-dimensional wireless communication network according to any one of claims 1-9, including: Terrain Voxelization Module: Used to acquire geographic information of the UAV inspection area, construct a three-dimensional terrain matrix of the inspection area, perform adaptive resolution voxelization on the three-dimensional terrain matrix, divide it into voxels of unequal size, and the set of voxels constitutes the voxelization model of the inspection area. The communication quality potential field construction module is used to obtain the communication quality index between each voxel and the UAV based on the voxelized model of the inspection area and the communication parameters of the UAV, taking into account the loss, and to construct the three-dimensional communication quality potential field of each voxel. Total potential field gradient module: used to design the repulsive potential field for collision prevention and interference resistance, as well as the gravitational potential field for communication blind zone requirements. Based on the above three-dimensional communication mass potential field, repulsive potential field, and gravitational potential field, the total potential field gradient of the UAV inspection is obtained. Distributed position optimization module: Based on the real-time inspection information of each UAV, it calculates the gradient of the total potential field of the UAV inspection at its location, and updates the three-dimensional position of each UAV in a distributed manner using the gradient descent method. Beam coordination control module: used to determine the coverage area of ​​each UAV based on the updated 3D position of each UAV, and to coordinately determine the beam direction and beam width accordingly; Online optimization module: used to define the balance index between network coverage performance and communication quality, and dynamically adjust the learning rate in S4 based on the above index to complete the real-time online optimization of UAV inspection.

Images