Network evolution calculation method for relay communication guarantee of unmanned aerial vehicle

Abstract

The invention relates to the technical field of communication guarantee, and discloses a network evolution calculation method for relay communication guarantee of an unmanned aerial vehicle, which overcomes the defects of the existing method in the aspects of calculation complexity, usability and coverage rate, and can better meet the communication requirements of each ground communication terminal. According to the method, a center is selected and clusters are divided based on a clustering optimization algorithm, a new center of each cluster is obtained by using a gradient descent method, the steps are repeated until the algorithm converges, the unmanned aerial vehicle is deployed to the position of the center point, and an optimal deployment scheme of the unmanned aerial vehicle in the time slice is obtained. When an actual unmanned aerial vehicle relay communication guarantee task is executed, the optimal deployment scheme of the unmanned aerial vehicle is solved by utilizing the model in different time slices, and then the scheme that the sum of the moving distances of the unmanned aerial vehicle between adjacent time slices is minimum is solved through a bipartite graph optimal matching algorithm of polynomial time complexity. Therefore, the problem of evolution of the communication support network of the unmanned aerial vehicle is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of communication guarantee, and particularly relates to a technical method for air relay communication guarantee based on a communication drone. Background Art

[0002] In recent years, the drone technology has been developing towards maturity and is increasingly widely used in fields such as photography, detection, communication, and military. Drones have the characteristics of high flexibility, strong controllability, wide coverage, and long endurance time, and are the best devices for carrying wireless communication base stations to implement emergency communication guarantee. In emergency situations such as earthquake relief, communication guarantee is a key lifeline, which not only affects the rescue efficiency but also directly relates to the life safety of the affected people. After natural disasters such as earthquakes and floods occur, the communication infrastructure in the disaster area is often severely damaged, resulting in communication interruption, blocking the connection between the disaster area and the outside world, making it difficult for rescue teams to obtain accurate information about the disaster area in a timely manner, and thus affecting the development of rescue operations. Using drones to carry communication base station equipment such as 4G or 5G is a powerful support for relay communication guarantee in large-scale disaster sites. Therefore, establishing a stable and reliable drone-based relay communication guarantee system is crucial for scenarios such as earthquake relief and military communication.

[0003] During the implementation of drone relay communication guarantee, due to the continuous changes in the disaster environment, such as the movement of rescue teams, damage to drones caused by bad weather, and the movement of affected people, the communication requirements may change. Therefore, a communication network evolution calculation method that can dynamically evaluate the deployment changes of drones is needed to maximize the communication requirements of the disaster site. This method can quickly recalculate the optimal deployment position of the drone according to the real-time changes, ensuring the stability of the communication network and the communication quality, thereby improving the rescue efficiency and effect. Therefore, for a ground mobile terminal to access the communication network through a relay drone, how to deploy and adjust the position and trajectory of the drone to best meet the communication requirements of each terminal under the conditions of dynamic changes in the terminal position, requirements, and resources is an important issue with wide and important applications in real life.

[0004] This problem can be regarded as a facility location problem. Existing UAV communication coverage methods can be divided into two categories: static deployment and dynamic deployment. Static deployment usually only needs to determine the optimal communication position of the UAV. After finding the optimal position, the number of UAVs is minimized to ensure that at least one ground device is within the communication range of the UAV. The advantage of this method is rapid deployment and simple implementation, but its static nature has limited practicality in actual scenarios. Dynamic deployment is more challenging. As the environment changes, the deployment position of the UAV needs to be continuously adjusted, which makes it more in line with the actual scenario requirements while adding more optimization constraints, such as UAV power limitations and the impact of terrain on communication quality. Therefore, it is difficult to accurately model the communication scenario. Common methods for dynamic deployment include heuristic algorithms, reinforcement learning, virtual force algorithms, etc. Heuristic algorithms face problems such as difficulty in obtaining global information, high computational complexity, and low coverage rate as the cluster size increases. Reinforcement learning algorithms face problems such as high computational cost and difficulty in accurately modeling the communication scenario. Although the virtual force algorithm is distributed, it is difficult to design the virtual force weight coefficient and has high computational complexity. Therefore, existing methods have not achieved a good balance in terms of computational complexity, ease of use, and coverage rate, and all have certain deficiencies. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: to propose a network evolution calculation method for UAV relay communication guarantee, to make up for the deficiencies of existing methods in terms of computational complexity, ease of use, and coverage rate, so as to better meet the communication needs of each ground terminal.

[0006] The technical solution adopted by the present invention to solve the above technical problem is:

[0007] A network evolution calculation method for UAV relay communication guarantee, comprising the following steps:

[0008] S1. Input relevant parameters of relay UAVs and ground communication terminals;

[0009] S2. Based on the clustering optimization method, solve the optimal deployment plan of relay UAVs within a time slice; specifically, all terminals select the center point closest to them to divide into clusters; use the gradient descent method to obtain the new center points of each cluster. After obtaining the center points of all clusters, it is the optimal deployment position. Deploy the UAVs to the obtained optimal deployment position to obtain the optimal deployment plan of the UAVs within the time slice;

[0010] S3. Solve the optimal matching scheme of relay UAVs between adjacent time slices; solve the optimal deployment positions of UAVs within different time slices; connect the UAV nodes between adjacent time slices to obtain a bipartite graph, and then use the optimal matching algorithm of the bipartite graph to solve the optimal matching scheme between adjacent time slices;

[0011] S4. Based on the optimal matching scheme between adjacent time slices, obtain the evolution scheme of the UAV communication network, and use the obtained optimal evolution scheme for the actual UAV relay communication guarantee task.

[0012] The clustering optimization method is implemented through the following steps:

[0013] S21: Solve the distance between terminal i and UAV a according to the Euclidean distance formula to obtain the distance set D = {D1, D2... D i ... D M}, where D i = {d i1 , d i2 ... d ia ... d iN}, D M represents the Mth terminal, and d iN represents the distance between terminal i and the Nth UAV;

[0014] S22: All terminals select the nearest central point and divide into cluster classes: Based on the distance set D calculated in step S21, for terminal i, select the UAV a with the smallest distance d ia as the central UAV of terminal i. The connection relationship between the terminal and the UAV is represented by a two-dimensional connection matrix A of size M×N. Initialize the connection matrix A = 0; update the connection matrix A, set the corresponding matrix element A ia = 1, and for set A ib = 0; traverse the a-th column of the connection matrix A, and those satisfying A ia = 1 are divided into the same cluster C a , and obtain the clustering situation of all terminals {C a | a = 1,..., N};

[0015] S23: Select a metric function, and use the gradient descent algorithm for the corresponding metric function to obtain the new central points of each cluster; for each terminal connected to UAV a, denote it as the same class C a , for each point in each cluster C a , its cost function is:

[0016]

[0017] where f(·) is the metric function, is the position of terminal i, is the coordinate of relay UAV a, is the communication distance threshold; represents the cost function; for all terminals within each cluster C a first solve the cost function gradient of the cluster, and then solve the minimum value along the direction of gradient descent, repeating the iteration until convergence or reaching the specified number of iterations. At this time, is the new center point position of cluster C a ;

[0018] S24: Calculate the distances {d ia} between each terminal and the UAV and update the connection matrix A, re-partition the cluster classes, and calculate the new center point position on this basis. Repeat the above steps until the center point positions converge. At this time, the center point positions are the optimal deployment positions of the UAVs.

[0019] During the iteration process, follow the following formula:

[0020]

[0021] where x a (k) and y a (k) respectively represent the abscissa and ordinate of the center of the a-th class at the k-th iteration, α is the learning rate, represents taking the partial derivative with respect to x a , represents taking the partial derivative with respect to y a .

[0022] Furthermore, the nodes in the bipartite graph are each UAV, the weight of the edge is the distance between UAVs between time slices, and the optimal matching of the bipartite graph is the minimum sum of the moving distances of the UAVs. The optimal matching algorithm of the bipartite graph includes the Hungarian method, the Kuhn - Munkras algorithm or the polynomial time complexity algorithm.

[0023] The beneficial effects of the present invention are:

[0024] 1) Low computing resource requirements and longer UAV hovering time. The clustering optimization method is used to solve the optimal deployment position of the UAV, and the gradient descent method is used to find the optimal solution, thereby reducing the computational cost.

[0025] 2) High computing efficiency of the evolution model. Through the optimal matching algorithm of the bipartite graph, the matching scheme with the minimum sum of the moving distances of UAVs between different time slices can be solved within polynomial time, greatly improving the computing efficiency of the evolution model.

[0026] 3) Wide application scenarios. The proposed method can not only be used for communication guarantee in tactical battlefields, but also has great application potential in fields such as earthquake relief and daily emergency relay communication.

[0027] 4) The evolved network has strong robustness. Considering the network changes brought about by the failure of UAVs or the addition of new UAVs, it improves the satisfaction rate and communication quality of terminal communication requirements, and enhances the robustness and resilience of the UAV relay communication network. Description of the Drawings

[0028] Figure 1 It is the overall flowchart of the network evolution calculation method for UAV relay communication guarantee in the present invention;

[0029] Figure 2 It is the flowchart of solving the optimal deployment model within a time slice based on the clustering optimization algorithm in the present invention;

[0030] Figure 3 It is the schematic diagram of the optimal clustering deployment of UAVs in the present invention;

[0031] Figure 4 It is the schematic diagram of the optimal deployment of UAVs under the time slice in the present invention and the matching between UAVs in the slice. Detailed Embodiment

[0032] The present invention aims to propose a network evolution calculation method for UAV relay communication guarantee, making up for the deficiencies of existing methods in terms of computational complexity, ease of use, and coverage rate, so as to better meet the communication requirements of each ground terminal. Its overall implementation process is as Figure 1 shown, including the following steps:

[0033] S1. Input the relevant parameters of relay UAVs and ground communication terminals;

[0034] S2. Based on the clustering optimization method, solve the optimal deployment plan of relay UAVs within a certain time slice;

[0035] S3. Solve the optimal matching plan of relay UAVs between adjacent time slices;

[0036] S4. Obtain the evolution plan of the relay communication guarantee network.

[0037] Embodiment:

[0038] Taking the number M of ground terminals and the number N of aerial relay UAVs as examples, the present invention scheme will be elaborated in detail as follows. The specific implementation process is as follows:

[0039] Step 1. First, input the relevant parameters of relay UAVs and ground communication terminals, and initialize the basic parameters of the algorithm: Input the relevant parameters of M ground communication terminals and N relay UAVs. Specifically, input the positions of all communication terminals (denoted as i, where i = 1, 2... M) within the task area and the communication distance threshold Number N of Relay UAVs and Communication Distance Threshold Wherein The coordinates of relay UAV a are denoted as The set minimum altitude of the UAVs is H min km, and the maximum altitude is H max km, and there is such a relationship that 0 ≤ z i < H min ≤ z a ≤ H max . The connection relationship between the terminal and the UAVs is represented by a two-dimensional connection matrix A of size M×N. Initialize the connection matrix A = 0, the gradient threshold ε = 10 -6 , and the UAV position change threshold δ = 0.005 km.

[0040] Step 2: Based on the clustering optimization method, solve the optimal deployment scheme of the relay UAVs within a time slice; as Figure 2 shown, the operation of solving the optimal deployment scheme includes calculating the distance from the terminal to each UAV, all terminals select the nearest central point and divide into clusters according to this, and use the gradient descent algorithm to obtain the new central points of each cluster for the selected metric function; after obtaining the central points of all clusters, it is the optimal deployment position; the specific steps are as follows:

[0041] Step 21: Solve the distance between terminal i and UAV a according to the Euclidean distance formula:

[0042]

[0043] Get the distance set D = {D1, D2 …… D i …… D M}, where D i = {d i1 , d i2 …… d ia …… d iN};

[0044] Step 22: All terminals select the nearest central point and divide into clusters: Based on the distance set D calculated in Step 21, for terminal i, select the UAV a with the minimum distance d ia , which is the central UAV of terminal i, update the connection matrix A, set the corresponding matrix element A ia = 1, and for set A ib = 0. Traverse the a-th column of the connection matrix A, and those satisfying A ia = 1 are divided into the same cluster C a , and obtain the clustering situation of all terminals {C a |a = 1, …, N}.

[0045] Step 23: Select a metric function and use the gradient descent algorithm to find the new center points of each cluster for the corresponding metric function; for each terminal connected to the drone a, they are denoted as belonging to the same cluster C a ; Different from the usual k-means clustering algorithm, the goal here is not to select the geometric center of the terminal positions within a certain cluster. Since the height of the drone is higher than all terminals, the centers of each cluster found by algorithms such as k-means weighted average are not on the height plane where the drone is located, and the projections of these points on the drone plane are usually not the optimal solutions (center points) of the cost function.

[0046] For each point in the cluster C a , its cost function is:

[0047]

[0048] where f(·) is the metric function. In this embodiment, f(x)=x is selected as the metric function. However, in fact, different function forms can be substituted to meet different real-world requirements. For example, when maximizing the power intensity, f(x)=-x is taken. -2 . For a simple case, such as when f(x)=x, at the same height, the convexity of the surface remains unchanged and there is a unique minimum value, which can be quickly obtained by methods such as gradient descent. For a case like f(x)=-x -2 , the surface does not have a single minimum point. However, for all the projections of the terminals within the same cluster C a on this height plane, there is still a single extreme point inside the polygon formed, which can also be quickly obtained by methods such as gradient descent. Usually, the relay drones are either as close as possible or as far as possible from the terminals. So all drones are at the same height z a =H, and H is determined by the specific properties of the metric function f(x). Its value can be the lower limit H min or the upper limit H max of the height set for the drone. Therefore, the height of the drone does not need to be iteratively updated.

[0049] During the iteration process, the following formula is followed:

[0050]

[0051] where x a (k) and y a (k) respectively represent the abscissa and ordinate of the center of the a-th class at the k-th iteration, α is the learning rate, represents the partial derivative of x a , represents the partial derivative of y a .

[0052] In this embodiment, f(x) = x is taken as an example of the metric function for illustration; for each cluster C a For all the terminals within it, first solve the gradient of the cost function of this type, and then solve the minimum value along the direction of gradient descent. Since in this example z a = H = H min , the iteration formula can be specifically in the following form:

[0053]

[0054] where α is the learning rate, which is taken as 0.01 in this embodiment. Repeat the iteration according to the formula until convergence (the gradient is less than the set threshold ε) or reach the specified number of iterations. At this time, is the new center point position of cluster C a .

[0055] Step 24: Based on Steps 21 and 23, calculate the distances {d ia} between each terminal and the UAV and update the connection matrix A, re-partition the cluster classes, and on this basis, calculate the new center point positions. Repeat the above steps until the center point positions converge (the change value of the positions is less than the set threshold δ). At this time, each center point position is the optimal deployment position of the UAV.

[0056] Step 3: Based on the optimal deployment position of the relay UAV obtained by solving in Step 2, deploy the UAV to this position to obtain the optimal deployment plan within the time slice, as Figure 3 shown.

[0057] Step 4: Solve the optimal matching plan of the relay UAV between adjacent time slices. The operation of solving the optimal matching plan includes solving the optimal deployment position of the UAV within different time slices, connecting the UAV nodes between adjacent time slices to obtain a bipartite graph, and using the optimal matching algorithm of the bipartite graph to solve the optimal matching plan between adjacent time slices:

[0058] After obtaining the optimal deployment plan through Steps 2 and 3, this plan can be used for the evolution task of the actual UAV relay communication guarantee network. Solving the optimal deployment under the time slice means determining the deployment position of the UAV based on the positions of the ground communication terminals at a certain moment. At different moments, due to terminal movement, UAV damage or new UAV addition, the terminal positions and UAV resources may be different. Therefore, re-solving can ensure the communication quality or efficiency to the greatest extent.

[0059] When the granularity of the time slice is fine enough, it can be considered that the optimal deployment is achieved at any moment. However, limited by the solving speed or other factors, the time interval of the slice cannot be arbitrarily small. Therefore, the time slice needs to be set to a reasonable size. At T oWithin the time slice range from \(T_0 = 0s\) to \(T_1 = 60s\), with a slice granularity of \(\Delta T = 5s\), recalculate every \(\Delta T\) through step 2 to obtain the optimal deployment positions of the drones within different time slices.

[0060] Step 42: As Figure 4 shown, according to the optimal deployment positions of the drones within time slice \(T\) and time slice \((T + \Delta T)\) obtained in step 41, use the optimal matching algorithm of the bipartite graph to solve for the optimal matching that minimizes the sum of the moving distances of all drones. In this embodiment, the Hungarian method is selected for the solution:

[0061] Connect the drone nodes between different time slices with edges to form a bipartite graph, and set the weight of the edge as the Euclidean distance between the corresponding two nodes; first create the connection matrix \(\Gamma\) of the bipartite graph, where the rows represent the drone nodes within time slice \(T\), and the columns represent the drone nodes within time slice \((T + \Delta T)\). It should be noted that when using the Hungarian method to solve, when the number of nodes within time slice \(T\) and time slice \((T + \Delta T)\) is inconsistent, some virtual nodes need to be added in the connection matrix \(\Gamma\) to make the connection matrix \(\Gamma\) a square matrix, and the virtual nodes are connected to the nodes in the other time slice with a same and relatively large weight. The specific steps of the Hungarian method include: (1) Subtract the minimum value of each row of the connection matrix \(\Gamma\); (2) Then subtract the minimum value of each column; (3) Cover the 0 elements in the matrix \(\Gamma\) with the minimum number of straight lines (rows or columns of the matrix). If this minimum number is \(N\), the algorithm ends; (4) Otherwise, for all uncovered elements, find the minimum value among them, then subtract this minimum value from all uncovered elements, add this minimum value to all covered elements that are covered by rows and columns twice, and then repeat step (3). When the above operations end, use the maximum flow algorithm to find the optimal matching.

[0062] Step 5: After obtaining the optimal matching scheme between adjacent time slices through step 4, it can be applied to the calculation of the evolution scheme of the actual drone relay communication guarantee network to implement communication network guarantee.

[0063] Although the present invention has been described herein with reference to the embodiments of the present invention, the above embodiments are only the preferred embodiments of the present invention, and the embodiments of the present invention are not limited by the above embodiments. It should be understood that those skilled in the art can design many other modifications and embodiments, and these modifications and embodiments will fall within the scope of the principles and spirit disclosed in this application.

Claims

1. A network evolution calculation method for ensuring UAV relay communication, characterized in that It includes the following steps: S1. Obtain relevant parameters of the relay UAV and the ground communication terminal; S2. Based on the clustering optimization method, solve the optimal deployment scheme of the relay UAV within a time slice; specifically, all terminals select the center point closest to them to divide the cluster classes; use the gradient descent method to obtain the new center points of each cluster. After obtaining the center points of all cluster classes, they are the optimal deployment positions. Deploy the UAVs to the obtained optimal deployment positions to obtain the optimal deployment scheme of the UAVs within the time slice; S3. Solve the optimal matching scheme of the relay UAVs between adjacent time slices; solve the optimal deployment positions of the UAVs in different time slices; connect the UAV nodes between adjacent time slices to obtain a bipartite graph, and then use the optimal matching algorithm of the bipartite graph to solve the optimal matching scheme between adjacent time slices; S4. Based on the optimal matching scheme between adjacent time slices, obtain the evolution scheme of the UAV communication network, and use the obtained optimal evolution scheme for the actual UAV relay communication guarantee task.

2. The network evolution calculation method for UAV relay communication guarantee according to claim 1, wherein The clustering optimization method is implemented through the following steps: S21: Solve the distance between terminal i and UAV a according to the Euclidean distance formula, and obtain the distance set D = {D1, D2... D i ... D M}, where D i = {d i1 , d i2 ... d ia ... d iN}, D M represents the Mth terminal, and d iN represents the distance between terminal i and the Nth UAV; S22: All terminals select the nearest central point and divide into clusters: The connection relationship between the terminals and the drones is represented by a two-dimensional connection matrix A of size M×N. Initialize the connection matrix A = 0. Based on the distance set D calculated in step S21, for terminal i, select the drone a with the smallest distance d ia as the central drone of terminal i, update the connection matrix A, and set the corresponding matrix element A ia = 1. For set A ib = 0; traverse the a-th column of the connection matrix A, and those satisfying A ia = 1 are divided into the same cluster C a , and obtain the clustering situation {C a |a = 1, …, N} of all terminals; S23: Select a metric function, and use the gradient descent algorithm for the corresponding metric function to obtain new center points for each cluster; for each terminal connected to a drone a, denote it as the same class C a , for each cluster C a , for the points in it, its cost function is: where, f(·) is a metric function, is the position of terminal i, is the coordinate of relay UAV a, is the communication distance threshold; denotes the cost function; for all terminals within each cluster C a first solve the gradient of the cost function of this cluster, and then solve the minimum value along the direction of gradient descent, repeat the iteration until convergence or reach the specified number of iterations, at this time is the new center point position of cluster C a ; S24: Calculate the distances {d ia} between each terminal and the UAV according to the new center point positions, update the connection matrix A, re-partition the clusters, and calculate the new center point positions on this basis. Repeat the above steps until the center point positions converge. At this time, the center point positions are the optimal deployment positions of the UAVs.

3. A network evolution calculation method for UAV relay communication guarantee according to claim 2, characterized in that In the iterative process, follow the following formula: where x a (k) and y a (k) respectively represent the abscissa and ordinate of the center of the a-th class at the k-th iteration, α is the learning rate, denotes taking the partial derivative with respect to x a and denotes taking the partial derivative with respect to y a .

4. A network evolution calculation method for UAV relay communication guarantee according to claim 3, characterized in that, In the bipartite graph, the nodes are each UAV, the weight of the edge is the distance between the UAVs between time slices, and the optimal matching of the bipartite graph is the minimum sum of the moving distances of the UAVs.

5. A network evolution calculation method for drone relay communication guarantee according to claim 4, characterized in that, The optimal matching algorithm of the bipartite graph includes the Hungarian method, the Kuhn - Munkras algorithm or other algorithms with polynomial time complexity.

6. A network evolution calculation method for UAV relay communication guarantee according to claim 5, characterized in that The specific solution of the Hungarian method is carried out through the following steps: First, create the connection matrix Γ of the bipartite graph. The rows represent the UAV nodes in time slice T, and the columns represent the UAV nodes in time slice T + ΔT; when the number of nodes in time slice T and time slice T + ΔT is inconsistent, some virtual nodes need to be added in the connection matrix Γ to make the connection matrix Γ a square matrix; then perform the following steps: S31: Subtract the minimum value of each row of the connection matrix Γ; S32: Then subtract the minimum value of each column; S33: Cover the 0 elements in the matrix Γ with the minimum number of straight lines, and the straight lines are the rows or columns of the matrix; if this minimum number is N, the algorithm ends; S34: Otherwise, for all uncovered elements, find the minimum value among them, then subtract this minimum value from all uncovered elements, add this minimum value to all covered elements that are covered by rows and columns twice, and then repeat step S33. S35: When the above operations are completed, use the maximum - flow algorithm to find the optimal matching.

Images