A method and system for decision-making in UAV graph matching tasks

By constructing a two-layer graph structure and a path-following optimization method, the problems of high-dimensional matrix calculation and dynamic perturbation in multi-UAV cooperative configuration tasks are solved, and efficient and stable UAV task allocation is achieved.

CN123114033APending Publication Date: 2026-11-17SICHUAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611343646.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-01
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional UAV task allocation methods have large computational scale and high storage requirements when dealing with multi-UAV cooperative configuration tasks, and their reconstruction capability under dynamic disturbances is insufficient, resulting in suboptimal matching results or failure to meet physical constraints.

Method used

The state and target mission configuration of the UAV swarm are modeled as a two-layer graph structure. By using the node affinity and edge affinity matrices, and through a continuous-discrete joint optimization method of factor decomposition and path following, the integer solutions that satisfy one-to-one assignment are gradually approximated, thus avoiding the direct construction of a high-dimensional global matrix.

Benefits of technology

It reduces computational and storage complexity, improves the feasibility and stability of matching results, and enables rapid reconstruction of high-fit topology allocation schemes under external perturbations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114033A_ABST
    Figure CN123114033A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for UAV graph matching task decision-making, belonging to the field of UAV swarm collaborative task decision-making technology. The current state and target task configuration of a multi-UAV swarm are constructed as a UAV resource capability graph and a task requirement graph, respectively. Node affinity matrices and edge affinity matrices are calculated, forming a two-layer graph structure for graph matching solutions. An objective function for multi-UAV graph matching task decision-making is constructed based on this two-layer graph structure, and factor decomposition is used to reconstruct the objective function into an equivalent trace form, thus constructing the multi-UAV graph matching task decision problem. A continuous-discrete joint optimization method based on path following is used to solve the multi-UAV graph matching task decision problem, obtaining the final UAV-task matching matrix. This method solves the problems of high dimensionality of the global affinity matrix, large rounding errors after continuous relaxation, and insufficient reconstruction capability under dynamic perturbations in existing technologies.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of UAV swarm collaborative task decision-making technology, and particularly relates to a UAV graph matching task decision-making method and system. Background Technology

[0002] With the rapid development of UAV swarm technology, the collaborative execution of complex tasks by multiple UAVs has become an important direction for the intelligent application of unmanned systems. In tasks such as passive positioning, cooperative reconnaissance, target tracking, area coverage, and communication relay, UAVs not only need to be assigned to corresponding task points, but also need to meet specific cooperative configuration requirements in space. For example, in passive positioning tasks based on time difference of arrival, the relative positional relationship between different UAVs directly affects the positioning geometric accuracy; in wide-area cooperative reconnaissance tasks, the distance between UAVs, the field of view coverage relationship, and the communication link relationship affect the overall mission effectiveness.

[0003] Traditional UAV task allocation methods typically model the task allocation problem as a linear allocation problem. Their objective function mainly considers the matching cost between a single UAV and a single task point, such as flight distance, remaining energy, arrival time, or task rewards. These methods can be solved using the Hungarian algorithm, auction algorithm, or contract net algorithm, and have advantages such as simple model and fast computation speed. However, when the task requires UAVs to maintain a specific cooperative topology relationship, simply considering the one-to-one matching cost between "UAVs and task points" is insufficient. It is also necessary to simultaneously consider whether the relative relationship between "UAVs" is consistent with the desired configuration relationship between "task points."

[0004] By introducing pairwise constraints, the multi-UAV task matching problem can be transformed into a quadratic allocation problem. A quadratic allocation problem means that while determining the matching relationship between a single UAV and a single task point, it is also necessary to evaluate whether the relationship between two UAVs matches the relationship between two task points. Therefore, this problem includes not only node matching costs but also edge matching costs, enabling a more accurate description of multi-UAV cooperative configuration tasks. However, the quadratic allocation problem is NP-hard; as the number of UAVs and task points increases, the computational scale grows rapidly, causing traditional methods to face significant difficulties under conditions of limited onboard computing power.

[0005] Existing technologies have the following main shortcomings when solving the quadratic assignment problem of multiple UAVs with pairwise constraints: 1. The curse of dimensionality and computational bottlenecks. For items containing drones and The task point matching problem, if directly constructing a global affinity matrix, requires modeling all possible "UAV-task point" matching pairs and the combinations between matching pairs, and its matrix dimension typically reaches [number missing]. As the number of drones or mission points increases, the storage and computational requirements of this matrix will increase rapidly. For drone swarms with limited onboard computing resources, explicitly calculating and storing such high-dimensional matrices can easily lead to excessive memory consumption and increased computational latency, making it difficult to meet the needs of real-time mission decision-making.

[0006] 2. Accuracy loss exists between continuous relaxation and discrete assignment. Existing graph matching or spectral matching methods typically first relax the original integer-constrained problem into a continuous optimization problem, obtaining a continuous matching result, and then converting it into an integer solution that satisfies the one-to-one assignment constraint through post-processing methods such as rounding or projection. Since the continuous optimization step and the discretization step are separate, the rounding process may destroy the optimality of the original optimization objective, leading to suboptimal matching in the output result, or even producing a solution that does not satisfy the physical assignment constraint.

[0007] 3. Insufficient adaptability to disturbances and node failures. In actual UAV swarm missions, the position, speed, remaining energy, communication status, and local link relationships of UAVs may be affected by wind disturbances, communication interference, or single-unit failures. When the swarm topology changes, traditional task allocation methods often require reconstructing the complete cost matrix and solving it again, making it difficult to complete dynamic reconstruction in a short time. Summary of the Invention

[0008] To address the problems of high dimensionality of the global affinity matrix, large rounding errors after continuous relaxation, and insufficient reconstruction capability under dynamic perturbations in existing technologies, this invention provides a UAV graph matching task decision-making method and system. This method models the current state and target task configuration of a multi-UAV swarm as attribute graphs, and simultaneously describes the single-UAV task matching relationship and multi-UAV collaborative configuration relationship through node affinity and edge affinity. Furthermore, it utilizes the graph structure association matrix and feature affinity matrix to factorize the global affinity matrix, avoiding the direct construction of a high-dimensional global matrix. Finally, it employs a path-following continuous-discrete joint optimization strategy to gradually converge the matching results from the continuous relaxation domain to the integer domain satisfying the one-to-one assignment constraint, thereby obtaining the final UAV-task matching matrix.

[0009] The technical solution of this invention is as follows: On one hand, the present invention provides a method for decision-making in UAV map matching tasks, comprising the following steps: The current state and target task configuration of the multi-UAV swarm are constructed as UAV resource capability graph and task requirement graph, respectively. The node affinity matrix and edge affinity matrix are calculated, which together form a two-layer graph structure for graph matching solution.

[0010] The objective function for multi-UAV graph matching task decision-making is constructed based on a two-layer graph structure, and factor decomposition is used to reconstruct the objective function for multi-UAV graph matching task decision-making into an equivalent trace form, thereby constructing the multi-UAV graph matching task decision-making problem.

[0011] The decision-making problem of multi-UAV graph matching task is solved by using a continuous-discrete joint optimization method based on path following, and the final matching matrix between UAVs and tasks is obtained.

[0012] Furthermore, the current state and target task configuration of the multi-UAV swarm are respectively constructed into a UAV resource capability graph and a task requirement graph, and the node affinity matrix and edge affinity matrix are calculated to jointly form a two-layer graph structure for graph matching solution, specifically including: Let the number of currently available drones be... The number of task points to be assigned is The current state of multiple drone clusters is constructed into a drone resource capability graph. ,in, Represents a set of drone nodes; This represents the set of edges between drone nodes. This is the feature matrix of UAV nodes. The edge feature matrix of the UAV. This is the association matrix between drone nodes and edges.

[0013] The UAV node feature matrix Used to describe the status of each drone. The dimension of the UAV node feature vector is represented by the dimensionality of the UAV node feature matrix. The line represents the first Node feature vectors of a drone .

[0014] The set of edges between the drone nodes Used to describe spatial adjacency, communication link, or cooperative observation relationships between drones, including elements Indicates the first drones and the first The drone side is composed of drones. This indicates the drone's number; when two drones can communicate, conduct collaborative observations, or meet a preset distance threshold, a drone edge is considered to exist between them; otherwise, no drone edge exists between them.

[0015] The UAV side feature matrix Used to describe the relative relationships between drones Indicates the number of drone edges. Let represent the dimension of the edge eigenvector; for the edge eigenvectors derived from the eigenvectors of the drones and the first The drone side composed of drones Its edge eigenvectors are represented as .

[0016] The association matrix between the UAV nodes and edges Used to represent the connection relationship between drone nodes and drone edges, if the first... The drone side and the first If a drone is connected to a corresponding drone node, then the elements in the association matrix between the drone node and the edge are... ,otherwise, , This indicates the number on the side of the drone.

[0017] At the same time, the target task configuration is constructed into a task requirement diagram. ,in, Represents a set of task nodes; This represents the set of edges between task nodes. The feature matrix of task nodes. The task edge feature matrix, This is the association matrix between task nodes and edges.

[0018] The set of edges between task nodes Used to describe the desired spatial configuration relationships, cooperative relationships, and topological relationships between task points, where elements... Indicates the first The task node corresponding to the task point and the first task point The task edges are formed by the task nodes corresponding to each task point. Indicates the task point number.

[0019] The task node feature matrix Used to describe the required attributes for each task point. The dimension of the task node feature vector is represented by the dimensionality of the task node feature matrix. The line represents the first Task node feature vector of each task point .

[0020] The task side feature matrix Used to describe the expected collaborative relationships between task points in a target task configuration. Indicates the number of task edges. The dimension of the task edge feature vector is represented by the number of features derived from the first feature vector. The first task point and the first The task edge is composed of task points. Its task edge feature vector is represented as .

[0021] The association matrix between task nodes and edges In the middle, if the first Task edge and the first If task nodes are connected, then the element Otherwise, element , This indicates the number of the task edge.

[0022] The node affinity matrix Used to describe the The drone carried out the first The degree of fit for each task point is represented by its elements. ,in, The distance weighting coefficient for node features. As a feasibility mask, when the first The drone meets the first When considering the capabilities, range, power, or payload requirements of each mission point. ;otherwise, .

[0023] Edge affinity matrix This describes the degree of similarity between the actual relative relationships between UAVs and the expected configuration relationships between mission points, for UAV edges. and the task side The corresponding element in the edge affinity matrix is ​​represented as ,in, Indicates the edge of the drone. Indicates the task edge, , where is the distance weight coefficient for edge features.

[0024] Furthermore, the objective function for multi-UAV graph matching task decision-making is constructed based on a two-layer graph structure, and factor decomposition is used to reconstruct the objective function for multi-UAV graph matching task decision-making into an equivalent trace form, thereby constructing the multi-UAV graph matching task decision-making problem, specifically as follows: The objective function for the multi-UAV map matching task decision is: (3); in, The representation represents the matching matrix. The corresponding total score for multi-UAV image matching, the matching matrix is: Matching elements in the matrix Indicates the first The drone was assigned to the first One task point, This indicates that the two do not match; This represents the total node matching score. Indicates the total edge matching score; The edge matching weight coefficient.

[0025] Based on the UAV resource capability graph, the starting point association matrix and the ending point association matrix of the UAV edges are defined as follows: , For the first drone side If the starting point is the first If a drone is deployed, then the elements in the starting correlation matrix... Otherwise, the element is 0; for the first drone side If the endpoint is the first If a drone is deployed, then the elements in the endpoint correlation matrix will be... Otherwise, the element is 0.

[0026] For the task requirement graph, the starting point and ending point association matrices of the task edges are defined as follows: and For the first If the starting point is the first task edge... For each task point, the elements in the starting point correlation matrix... Otherwise, the element is 0; for the first If the endpoint of the task edge is the first one... For each task point, then Otherwise, the element is 0.

[0027] The decision problem for multi-UAV graph matching task is expressed as: (9); in, Represents the trace operation of a matrix. Indicates transpose. and They represent dimensions as follows: and A completely uniform vector.

[0028] Furthermore, the method of continuous-discrete joint optimization based on path following is used to solve the multi-UAV graph matching task decision problem, and the final UAV-task matching matrix is ​​obtained, specifically as follows: First, the original integer feasible field is represented as Relax the original integer feasible region into a continuous feasible region. When the number of drones equals the number of task points and a one-to-one match is required, the continuous feasible region can be written in the form of a double random matrix constraint: .

[0029] Constructing from path parameters Adjusted interpolation objective function: , (10); in, Describes the interpolation objective function. Indicates path parameters, This represents the initial relaxation objective function that facilitates continuous optimization. Let represent the objective function used to guide the solution to converge to the discrete integer domain. Specifically, take , ,in, It is the identity matrix. To adjust the parameters, , This represents the symmetric edge affinity matrix and node affinity vector obtained by factorization. ,vector .

[0030] Substituting the initial relaxation objective function, which facilitates continuous optimization, and the objective function used to guide the solution to converge to the discrete integer domain, into the interpolation objective function, we obtain: (11); remember ,but: (12); Path parameters during the solution process Incrementing gradually from 0 to 1: , ,in, Indicates the number of the path following stage. This represents the total number of stages in the path following process. Represents path parameters In the The values ​​of each path following stage, each This corresponds to a continuous optimization subproblem: ,exist At this point, an initial solution is first obtained in the continuous feasible region, and then the solution is gradually increased. The solution from the previous stage is used as the initial value for the next stage, so that the matching matrix gradually moves along the continuous path towards the original integer feasible region.

[0031] For each path parameter Perform inner-layer iterative solution: Let the current iterative solution be... Its vectorized form is The interpolation objective function with respect to The gradient is , Indicates path parameter taking The quadratic coefficient matrix of the time-interpolation objective function is used to rearrange the gradient into matrix form: ,in, express The reverse operation, Indicates the first s The gradient matrix of the interpolation objective function with respect to the matching matrix during the next inner iteration has the same dimension as the matching matrix. X same.

[0032] Then solve the following linear allocator problem: ,in, Represents the inner product of matrices. This represents the candidate matching matrix in the linear allocation subproblem. Indicates the first s The matching matrix obtained during the next inner iteration is obtained by solving the linear allocator subproblem. Then, the update direction is Next, in updating the direction Perform a one-dimensional search to determine the optimal step size: , Indicates the candidate step size, Indicates the first s The optimal step size is obtained from the innermost iteration.

[0033] make Then the step size optimization problem is about Maximizing a quadratic function in one variable: Therefore, if The candidate step size is then taken as and truncate it to the interval Inside: ,like Then compare them separately. and The interpolation objective function value is calculated, and a step size with a larger interpolation objective function value is selected. The matching matrix is ​​then updated. . Stop the current process when any of the following conditions are met. The inner iteration below: ,or or reaching the maximum number of iterations. ,in, and These represent the threshold for changes in the objective function and the threshold for changes in the matching matrix, respectively. Represents the Frobenius norm, current path parameter The optimal solution is denoted as ,when When that happens, use it as the next path parameter. Initial values ​​for the following solutions: Repeat the above process until... Finally, in Output matching matrix .

[0034] When the elements in the output matching matrix are not 0 or 1, the final UAV-mission matching result is obtained according to the following rules: ,in, Indicates the first The drone was eventually assigned to the first One task point, This indicates the elements of the output matching matrix; if the number of task points is less than the number of drones, the unselected drones remain idle or are used as backup drones.

[0035] On the other hand, the present invention also provides an unmanned aerial vehicle (UAV) map matching task decision system for implementing an UAV map matching task decision method, comprising: The two-layer graph structure construction module is used to construct the current state and target task configuration of a multi-UAV cluster into a UAV resource capability graph and a task requirement graph, respectively, and calculate the node affinity matrix and edge affinity matrix to jointly form a two-layer graph structure for graph matching solution.

[0036] The decision problem construction module is used to construct the objective function for multi-UAV graph matching task decision based on a two-layer graph structure, and to reconstruct the objective function for multi-UAV graph matching task decision into an equivalent trace form using factor decomposition, thereby constructing the multi-UAV graph matching task decision problem.

[0037] The solution module uses a path-following-based continuous-discrete joint optimization method to solve the multi-UAV graph matching task decision problem, and obtains the final UAV-task matching matrix.

[0038] Thirdly, this application proposes an electronic device comprising: one or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform the aforementioned UAV map matching task decision method.

[0039] Fourthly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the aforementioned UAV graph matching task decision method.

[0040] Fifthly, this application proposes a computer program product, including a computer program or instructions that, when executed by a processor, implement the aforementioned UAV map matching task decision method.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Computational and memory friendly: This invention eliminates the need for computationally and spatially expensive pairwise affinity matrices. By using matrix factorization, the UAV network structure and characteristic parameters are decoupled, significantly reducing the algorithm's space and time complexity, making it highly suitable for use in computationally limited airborne flight platform clusters.

[0042] 2. Improve the feasibility of discrete matching results: Unlike traditional algorithms that first perform continuous optimization and then obtain discrete solutions through independent rounding, this invention introduces a path following optimization strategy based on the factorized graph matching objective function. This strategy allows the matching matrix to gradually approach the discrete matching domain that satisfies the allocation constraints from the continuous relaxation domain. This reduces the loss of matching accuracy caused by independent rounding post-processing and improves the feasibility and stability of UAV-mission matching results.

[0043] 3. Strong anti-interference and dynamic reconstruction capabilities: When the cluster faces external disturbances (such as edge noise caused by drone yaw) or the failure of some nodes, this graph model and optimization framework have strong robustness and can quickly converge to generate a new topology allocation scheme with high fitting degree. Attached Figure Description

[0044] Figure 1 This is a flowchart of a UAV graph matching task decision-making method according to an embodiment of the present invention.

[0045] Figure 2 This is a task scenario diagram in an embodiment of the present invention.

[0046] Figure 3 This is a comparison chart of matching accuracy under different perturbations in an embodiment of the present invention.

[0047] Figure 4 This is a comparison chart of computation time under different cluster sizes in the embodiments of the present invention. Detailed Implementation

[0048] Example 1: A UAV graph matching task decision-making method, such as Figure 1 As shown, it includes the following steps: Step 1: Construct the current state and target task configuration of the multi-UAV cluster into a UAV resource capability graph and a task requirement graph, respectively, and calculate the node affinity matrix and edge affinity matrix to form a two-layer graph structure for graph matching solution.

[0049] Let the number of currently available drones be... The number of task points to be assigned is The current state of multiple drone clusters is constructed into a drone resource capability graph. ,in, Represents a set of drone nodes. For the first The drone node corresponding to each drone The drone's serial number; This represents the set of edges between drone nodes. This is the feature matrix of UAV nodes. The edge feature matrix of the UAV. This is the association matrix between drone nodes and edges.

[0050] The UAV node feature matrix Used to describe the status of each drone. The dimension of the UAV's node feature vector is represented by the number of node feature terms contained in the node feature vector of each UAV, and the dimensionality of the UAV node feature matrix. The line represents the first The node feature vector of a drone can be represented as: ,in, Indicates the first The spatial coordinates of the drone Indicates the first The remaining battery power of the drone, Indicates the first The maximum mission radius or remaining range of the drone. Indicates the first The mission capability type or payload capacity of the UAV; depending on the specific mission requirements, the node feature items in the node feature vector may also include information such as speed, communication status, sensor type and number of available payloads.

[0051] The set of edges between the drone nodes Used to describe spatial adjacency, communication link, or cooperative observation relationships between drones, including elements Indicates the first drones and the first The drone side is composed of drones. This indicates the drone's number; when two drones can communicate, conduct collaborative observations, or meet a preset distance threshold, a drone edge is considered to exist between them; otherwise, no drone edge exists between them.

[0052] The UAV side feature matrix Used to describe the relative relationships between drones Indicates the number of drone edges. Let represent the dimension of the edge eigenvector; for the edge eigenvectors derived from the eigenvectors of the drones and the first The drone side composed of drones Its edge eigenvectors can be represented as ,in, Indicates the first The drone and the first The Euclidean distance between the drones Indicates the first The drone and the first The relative azimuth angles or relative geometric relationships between drones. Indicates the first The drone and the first Does the drone have spatial adjacency, communication link, or collaborative observation relationship with other drones?

[0053] The association matrix between the UAV nodes and edges Used to represent the connection relationship between drone nodes and drone edges, if the first... The drone side and the first If a drone is connected to a corresponding drone node, then the elements in the association matrix between the drone node and the edge are... ,otherwise, , This indicates the number on the side of the drone.

[0054] At the same time, the target task configuration is constructed into a task requirement diagram. ,in, Represents a set of task nodes. Indicates the first The task node corresponding to each task point. Indicates the task point number; This represents the set of edges between task nodes. The feature matrix of task nodes. The task edge feature matrix, This is the association matrix between task nodes and edges.

[0055] The set of edges between task nodes Used to describe the desired spatial configuration relationships, cooperative relationships, and topological relationships between task points, where elements... Indicates the first The task node corresponding to the task point and the first task point The task edges are formed by the task nodes corresponding to each task point. Indicates the task point number.

[0056] The task node feature matrix Used to describe the required attributes for each task point. The dimension of the task node feature vector represents the number of task requirement feature terms contained in each task node feature vector, and the dimensionality of the task node feature matrix is... The line represents the first The feature vector of each task node can be represented as: ,in, Indicates the first The expected spatial location of each task point This indicates the type of task capability required for this task point. This indicates the task priority or task intensity requirements. Depending on the task scenario, task requirement features may also include information such as expected arrival time, observation accuracy requirements, and payload requirements.

[0057] The task side feature matrix Used to describe the expected collaborative relationships between task points in a target task configuration. Indicates the number of task edges. The dimension of the task edge feature vector is represented by the number of features derived from the first feature vector. The first task point and the first The task edge is composed of task points. Its task edge feature vector can be represented as ,in, This represents the expected distance between two task points. This indicates the expected relative positional relationship between two task points. This indicates whether there is a collaborative relationship between two task points.

[0058] The association matrix between task nodes and edges Definition method and The same, if the first Task edge and the first If task nodes are connected, then the element Otherwise, element , This indicates the number of the task edge.

[0059] After constructing the UAV resource capability graph and mission requirement graph, the node affinity matrix and edge affinity matrix are calculated respectively. Node affinity matrix Used to describe the The drone carried out the first The degree of fit for each task point can be represented by the following elements: ,in, The distance weighting coefficient for node features. As a feasibility mask, when the first The drone meets the first When considering the capabilities, range, power, or payload requirements of each mission point. ;otherwise, By using a feasibility mask, drone mission matching relationships that do not meet mission requirements can be directly excluded.

[0060] Edge affinity matrix This describes the degree of similarity between the actual relative relationships between UAVs and the expected configuration relationships between mission points. For UAV edges... and the task side The corresponding element in the edge affinity matrix can be represented as ,in, This represents the edge feature distance weight coefficient. A higher edge affinity indicates a stronger edge affinity. , The closer the actual geometric relationship between the two drones is to the first... , The desired configuration relationship between two task points.

[0061] Through the above processing, the matching relationship between a single UAV and a single task point can be represented as node affinity, and the cooperative configuration matching relationship between multiple UAVs can be represented as edge affinity, thus providing input for subsequent secondary graph matching optimization.

[0062] Step 2: Construct the objective function for multi-UAV graph matching task decision based on the two-layer graph structure, and use factor decomposition to reconstruct the objective function for multi-UAV graph matching task decision into an equivalent trace form, thereby constructing the multi-UAV graph matching task decision problem.

[0063] Let the matching matrix be Matching elements in the matrix Indicates the first The drone was assigned to the first One task point, This indicates a mismatch. If all task points need to be assigned, and each drone executes at most one task point's corresponding task, then the matching matrix satisfies... , ,in, and They represent dimensions as follows: and A completely one vector, This indicates transpose. When the number of drones is the same as the number of mission points and a one-to-one match is required, the above formula can be further written as: , .

[0064] The objective of multi-UAV mission matching includes two parts: node matching score and edge matching score. The node matching score is used to evaluate the degree of fit between a single UAV and a single mission point, and can be expressed as: (1); in, This represents the total node matching score. Represents the trace operation of a matrix.

[0065] Edge matching score is used to evaluate whether the actual cooperative relationship between UAVs is consistent with the expected configuration relationship between task points. Let the edge matching score be the score when the first UAV is the correct score. The drone was matched to the first The task point, and the first The drone was matched to the first The edge matching score for each task point is: Therefore, the total edge matching score can be expressed as: (2); in, Indicates the total edge matching score; Indicates the edge of the drone. Indicates the task edge.

[0066] Combining the total node matching score and the total edge matching score, the objective function for decision-making in multi-UAV graph matching tasks is obtained: (3); in, The representation represents the matching matrix. The corresponding total score for multi-drone map matching, The edge matching weight coefficient is used to adjust the relative importance of single-machine task adaptation and multi-machine cooperative configuration relationships in the overall objective. To avoid directly constructing a high-dimensional global affinity matrix, this invention further uses the node-edge association relationship to adjust the total edge matching score in the objective function of multi-UAV graph matching task decision-making. Perform factorization.

[0067] Based on the UAV resource capability graph, the starting point association matrix and the ending point association matrix of the UAV edges are defined as follows: , For the first drone side If the starting point is the first If a drone is deployed, then the elements in the starting correlation matrix... Otherwise, the element is 0; for the first drone side If the endpoint is the first If a drone is deployed, then the elements in the endpoint correlation matrix will be... Otherwise, the element is 0.

[0068] Similarly, for the task requirement graph, the starting point and ending point incidence matrices of the task edges are defined as follows: and For the first If the starting point is the first task edge... For each task point, the elements in the starting point correlation matrix... Otherwise, the element is 0; for the first If the endpoint of the task edge is the first one... For each task point, then Otherwise, the element is 0.

[0069] Based on the above correlation matrix, two edge matching induction matrices can be constructed. , ,in, Used to represent the matching relationship between the starting point of the UAV and the starting point of the mission. This is used to represent the matching relationship between the endpoint of the UAV edge and the endpoint of the mission edge. For UAV edges... and the task side Edge matching induced matrix The corresponding element in the middle is Edge matching induced matrix The corresponding element in the middle is Therefore, the element-wise product of the two can represent the combination relationship when the nodes at both ends of the edge are matched at the same time.

[0070] Therefore, the total edge matching score can be reconstructed into the following equivalent trace form: (4); in, This represents the element-wise product of Hadamard.

[0071] Will and After unfolding, we have: (5); Therefore, the objective function for the original multi-UAV image matching task decision can be written as: (6); Let vector ,in, This represents the vectorization operation of stacking columns. Based on the matrix vectorization identity... , can be obtained , ,in , To match the induced matrix for edges, , .make Then the total edge matching score can be equivalently represented as Thus, the factorized global edge affinity matrix can be obtained. To maintain the symmetry of the quadratic form matrix, we can take... , This represents the symmetric edge affinity matrix obtained by factorization.

[0072] Since for any vector All Therefore, the total edge matching score can be written in standard quadratic form: (7); Furthermore, let the node affinity vector The objective function for multi-UAV map matching task decision-making can be written as: (8); Therefore, the decision-making problem for multi-UAV image matching tasks can be expressed as: (9); Through the above factorization reconstruction, this invention does not require explicit computation and storage dimensions. Instead of a global affinity matrix, it uses a smaller node affinity matrix. Edge affinity matrix and graph structure correlation matrix ( , , , The objective function is calculated. This reduces the storage overhead and computational complexity of the quadratic graph matching problem as the number of drones and task points increases, and improves the real-time performance of multi-drone swarm task decision-making under conditions of limited onboard computing power.

[0073] Step 3: Solve the multi-UAV graph matching task decision problem using the continuous-discrete joint optimization method based on path following, and obtain the final UAV-task matching matrix.

[0074] First, the original integer feasible field is represented as To facilitate continuous optimization, the original integer feasible region is relaxed to a continuous feasible region. When the number of drones equals the number of task points and a one-to-one match is required, the continuous feasible region can be written in the form of a double random matrix constraint: .

[0075] To enable continuous optimization results to gradually converge to discrete integer solutions, a method is constructed using path parameters. Adjusted interpolation objective function: , (10); in, Describes the interpolation objective function. Indicates path parameters, This represents the initial relaxation objective function that facilitates continuous optimization. Let represent the objective function used to guide the solution to converge to the discrete integer domain. Specifically, can be taken as... , ,in, It is the identity matrix. To adjust the parameters.

[0076] For integer solutions that satisfy the one-to-one assignment constraint, Since is a constant, the optimal solutions of the two objective functions above in the integer feasible region are consistent with the original quadratic graph matching objective. Wherein, Used to obtain a relatively stable initial optimization direction in a continuous feasible region. This is used to enhance the objective function's preference for discrete vertex solutions, allowing the solution to gradually approximate the desired value. Matching matrix. The one-to-one allocation constraint is that the number of drones is equal to the number of task points, each drone performs exactly one task, and each task point is assigned exactly one drone.

[0077] Substituting the two objective functions above into the interpolation objective function, we get: (11); remember ,but: (12); Path parameters during the solution process Incrementing gradually from 0 to 1: , .in, Indicates the number of the path following stage. This represents the total number of stages in the path following process. Represents path parameters In the The values ​​of each path following stage, each This corresponds to a continuous optimization subproblem: ,exist At this point, an initial solution is first obtained in the continuous feasible region, and then the solution is gradually increased. The solution from the previous stage is used as the initial value for the next stage, so that the matching matrix gradually moves along the continuous path towards the original integer feasible region.

[0078] For each fixed In this embodiment, a modified Frank-Wolfe method is preferably used for iterative solution.

[0079] Let the current iterative solution be Its vectorized form is The interpolation objective function with respect to The gradient is , Indicates path parameter taking The coefficient matrix of the quadratic term of the interpolation objective function is given by the aforementioned... Substitution We obtain the gradients. Rearrange the gradients into matrix form: ,in, express The reverse operation, Indicates the first s The gradient matrix of the interpolation objective function with respect to the matching matrix during the next inner iteration has the same dimension as the matching matrix. X same.

[0080] Then solve the following linear allocator problem: ,in, Represents the inner product of matrices. This represents the candidate matching matrix in the linear allocation subproblem. Indicates the first s The matching matrix obtained during the inner iteration, since this subproblem is a linear allocation problem, can be quickly solved using the Hungarian algorithm, auction algorithm, or linear programming method. Then, the update direction is Next, in updating the direction. Perform a one-dimensional search to determine the optimal step size: , Indicates the candidate step size, Indicates the first s The optimal step size is obtained from the innermost iteration.

[0081] make Then the step size optimization problem is about Maximizing a quadratic function in one variable: Therefore, if The candidate step size can then be taken as and truncate it to the interval Inside: ,like Then compare them separately. and The interpolation objective function value is calculated, and a step size with a larger interpolation objective function value is selected. The matching matrix is ​​then updated. . Stop the current process when any of the following conditions are met. The inner iteration below: ,or or reaching the maximum number of iterations. .in, and These represent the threshold for changes in the objective function and the threshold for changes in the matching matrix, respectively. Represents the Frobenius norm. Current path parameters. The optimal solution is denoted as .when When that happens, use it as the next path parameter. Initial values ​​for the following solutions: Repeat the above process until... Ultimately, in Output matching matrix .

[0082] Because the objective function gradually increases its preference for discrete vertex solutions during path following, and the search direction of each linear allocation subproblem is given by a matrix satisfying the allocation constraints, the final output... Within the set tolerance range Integer constraints. When numerical errors exist, i.e., when elements in the output matching matrix are not 0 or 1, the final UAV-task matching result can be obtained according to the following rules: ,in, Indicates the first The drone was eventually assigned to the first One task point, This represents the elements of the output matching matrix. If the number of task points is less than the number of drones, the unselected drones remain idle or are used as backup drones.

[0083] In summary, this step uses path parameters. By unifying continuous relaxation optimization and discrete integer matching into a single optimization process, the accuracy loss caused by the sequential solution and independent rounding in traditional methods is avoided. Furthermore, since the objective function is calculated using the factorized trace form in step two, there is no need to explicitly construct a dimensionless function. The global affinity matrix enables multi-UAV map matching decision-making with low memory usage.

[0084] The technical solution of this invention mainly addresses the problems of high-dimensional matrix calculation and discrete allocation in multi-UAV cooperative configuration task matching. On one hand, to address the issues of excessively large global affinity matrix dimension and high explicit construction and storage costs in traditional quadratic allocation problems, this invention models the UAV swarm state and target task configuration as attribute graphs respectively, and constructs a factorized objective function using node-edge association matrices, node affinity matrices, and edge affinity matrices. This avoids directly forming a large-scale global affinity matrix, thereby reducing computational complexity and memory consumption. On the other hand, to address the problem that continuous relaxation optimization results cannot directly satisfy the one-to-one UAV allocation constraint, this invention introduces a path-following optimization strategy based on the factorized graph matching objective function. This allows the matching matrix to gradually approach the discrete matching domain from the continuous relaxation domain, reducing the matching accuracy loss caused by independent rounding post-processing and improving the feasibility and stability of the final UAV-to-task matching result.

[0085] Simulation settings: Let the number of currently available drones be...n The number of task points to be assigned is m Before the mission begins, the UAV swarm is deployed in a dispersed state, with each UAV having different positions, battery levels, remaining range, and payload capabilities. The mission arm provides several task points to be executed and imposes configuration constraints on the distance, relative orientation, and cooperative relationships between these task points. The goal of this invention is to ensure that, while satisfying individual UAV capability constraints and task point allocation constraints, the actual cooperative topology formed between the UAVs is as close as possible to the target mission configuration.

[0086] like Figure 2 As shown, the left side is the UAV resource capability map, and the right side is the mission requirement map. UAV nodes represent available UAVs, and mission nodes represent mission points to be executed. The solid lines in the map represent the communication, collaborative observation, or distance threshold-based relationships between UAVs, as well as the desired configuration relationships between mission points. The dashed arrows represent the matching relationships between UAVs and mission points obtained through graph matching mission decisions. External disturbances can manifest as position shifts, heading changes, communication link fluctuations, or a decline in the status of some nodes caused by wind fields.

[0087] To verify the matching stability of the method of this invention under perturbation conditions, different perturbation intensities were set for comparison. Perturbation intensity can be understood as the comprehensive perturbation ratio of the UAV node position, edge distance, relative orientation, or link state relative to the baseline state. This embodiment selects seven perturbation levels: 0%, 5%, 10%, 15%, 20%, 25%, and 30%. The comparison methods include the factorized graph matching path-following method of this invention, the traditional spectral matching method, and the traditional allocation method that primarily considers node costs.

[0088] Figure 3 The matching accuracy of the three methods under different perturbation intensities is compared. It can be seen that as the perturbation intensity increases from 0% to 30%, the matching accuracy of all three methods decreases, indicating that external perturbations change the state of UAV nodes and edge relationships, thus increasing the difficulty of task matching. In contrast, the accuracy of the method proposed in this invention decreases more slowly, maintaining high matching stability even under strong perturbation conditions.

[0089] To verify the computational efficiency of the method of this invention under UAV swarms of different sizes, the number of UAVs was set to 10, 20, 30, 40, 50, and 60. The number of task points was kept consistent with the number of UAVs to simulate one-to-one matching cooperative configuration tasks. Comparison methods included the factorization method of this invention, the method of explicitly constructing the global affinity matrix, and the traditional quadratic allocation method.

[0090] Figure 4The computation time of the three methods is compared under different cluster sizes. As the number of drones and task points increases, the computation time of all three methods increases, but the rate of increase differs significantly. The time increase of the factorization method in this invention is relatively gradual; the time increase of the explicit global affinity method and the traditional quadratic allocation method is faster, with a significant increase in computational burden under larger clusters.

[0091] Example 2: A UAV map matching task decision system, used to implement a UAV map matching task decision method, includes: The two-layer graph structure construction module is used to construct the current state and target task configuration of a multi-UAV cluster into a UAV resource capability graph and a task requirement graph, respectively, and calculate the node affinity matrix and edge affinity matrix to jointly form a two-layer graph structure for graph matching solution.

[0092] The decision problem construction module is used to construct the objective function for multi-UAV graph matching task decision based on a two-layer graph structure, and to reconstruct the objective function for multi-UAV graph matching task decision into an equivalent trace form using factor decomposition, thereby constructing the multi-UAV graph matching task decision problem.

[0093] The solution module uses a path-following-based continuous-discrete joint optimization method to solve the multi-UAV graph matching task decision problem, and obtains the final UAV-task matching matrix.

[0094] Example 3: This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the aforementioned UAV map matching task decision method.

[0095] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements a UAV map matching task decision-making method as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0096] The processor is used to execute all or part of the steps in the UAV map matching task decision method as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in the electronic device, as well as application-related data.

[0097] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the UAV map matching task decision method described in the above embodiments.

[0098] Example 4: This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0099] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the UAV map matching task decision method described in the various embodiments of this application.

[0100] The aforementioned storage media include: flash memory, hard disk, multimedia card, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory, random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disk, optical disk, server, APP (Application) application store, and other media capable of storing program verification codes, on which computer programs are stored. When the computer programs are executed by the processor, they can implement the various steps of the UAV map matching task decision method described above.

[0101] Example 5: This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned UAV map matching task decision method.

[0102] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0103] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0104] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of this disclosure and its equivalents, then the intent of this disclosure also includes these modifications and variations.

Claims

1. A method for making decisions on UAV map matching tasks, characterized in that, Includes the following steps: The current state and target task configuration of the multi-UAV swarm are constructed as UAV resource capability graph and task requirement graph, respectively. The node affinity matrix and edge affinity matrix are calculated to form a two-layer graph structure for graph matching solution. The objective function for multi-UAV graph matching task decision-making is constructed based on a two-layer graph structure, and factor decomposition is used to reconstruct the objective function for multi-UAV graph matching task decision-making into an equivalent trace form, thereby constructing the multi-UAV graph matching task decision-making problem. The decision-making problem of multi-UAV graph matching task is solved by using a continuous-discrete joint optimization method based on path following, and the final matching matrix between UAVs and tasks is obtained.

2. The UAV map matching task decision method according to claim 1, characterized in that, The process involves constructing the current state and target task configuration of a multi-UAV swarm into a UAV resource capability graph and a task requirement graph, respectively, and calculating the node affinity matrix and edge affinity matrix to jointly form a two-layer graph structure for graph matching solutions. Specifically, this includes: Let the number of currently available drones be... The number of task points to be assigned is The current state of multiple drone clusters is constructed into a drone resource capability graph. ,in, Represents a set of drone nodes; This represents the set of edges between drone nodes. This is the feature matrix of UAV nodes. The edge feature matrix of the UAV. This is the association matrix between drone nodes and edges; The UAV node feature matrix Used to describe the status of each drone. The dimension of the UAV node feature vector is represented by the dimensionality of the UAV node feature matrix. The line represents the first Node feature vectors of a drone ; The set of edges between the drone nodes Used to describe spatial adjacency, communication link, or cooperative observation relationships between drones, including elements Indicates the first drones and the first The drone side is composed of drones. This indicates the drone's ID number; when two drones can communicate, conduct collaborative observations, or meet a preset distance threshold, a drone edge is considered to exist between them; otherwise, no drone edge exists between them. The UAV side feature matrix Used to describe the relative relationships between drones Indicates the number of drone edges. Let represent the dimension of the edge eigenvector; for the edge eigenvectors derived from the eigenvectors of the drones and the first The drone side composed of drones Its edge eigenvectors are represented as ; The association matrix between the UAV nodes and edges Used to represent the connection relationship between drone nodes and drone edges, if the first... The drone side and the first If a drone is connected to a corresponding drone node, then the elements in the association matrix between the drone node and the edge are... ,otherwise, , Indicates the number of the drone's side; At the same time, the target task configuration is constructed into a task requirement diagram. ,in, Represents a set of task nodes; This represents the set of edges between task nodes. The feature matrix of task nodes. The task edge feature matrix, This is the association matrix between task nodes and edges; The set of edges between task nodes Used to describe the desired spatial configuration relationships, cooperative relationships, and topological relationships between task points, where elements... Indicates the first The task node corresponding to the task point and the first task point The task edges are formed by the task nodes corresponding to each task point. Indicates the task point number; The task node feature matrix Used to describe the required attributes for each task point. The dimension of the task node feature vector is represented by the dimensionality of the task node feature matrix. The line represents the first Task node feature vector of each task point ; The task side feature matrix Used to describe the expected collaborative relationships between task points in a target task configuration. Indicates the number of task edges. The dimension of the task edge feature vector is represented by the number of features derived from the first feature vector. The first task point and the first The task edge is composed of task points. Its task edge feature vector is represented as ; The association matrix between task nodes and edges In the middle, if the first Task edge and the first If task nodes are connected, then the element Otherwise, element , Indicates the number of the task edge; The node affinity matrix Used to describe the The drone carried out the first The degree of fit for each task point is represented by its elements. ,in, The distance weighting coefficient for node features. As a feasibility mask, when the first The drone meets the first When considering the capabilities, range, power, or payload requirements of each mission point. ;otherwise, ; Edge affinity matrix This describes the degree of similarity between the actual relative relationships between UAVs and the expected configuration relationships between mission points, for UAV edges. and the task side The corresponding element in the edge affinity matrix is ​​represented as ,in, Indicates the edge of the drone. Indicates the task edge, , where is the distance weight coefficient for edge features.

3. The UAV map matching task decision method according to claim 1, characterized in that, The objective function for multi-UAV graph matching task decision-making is constructed based on a two-layer graph structure, and factor decomposition is used to reconstruct the objective function into an equivalent trace form, thereby constructing the multi-UAV graph matching task decision problem, specifically: The objective function for the multi-UAV map matching task decision is: (3); in, The representation represents the matching matrix. The corresponding total score for multi-UAV image matching, the matching matrix is: Matching elements in the matrix Indicates the first The drone was assigned to the first One task point, This indicates that the two do not match; This represents the total node matching score. Indicates the total edge matching score; The edge matching weight coefficient; Based on the UAV resource capability graph, the starting point association matrix and the ending point association matrix of the UAV edges are defined as follows: , For the first drone side If the starting point is the first If a drone is deployed, then the elements in the starting correlation matrix... Otherwise, the element is 0; for the first drone side If the endpoint is the first If a drone is deployed, then the elements in the endpoint correlation matrix will be... Otherwise, the element is 0; For the task requirement graph, the starting point and ending point association matrices of the task edges are defined as follows: and For the first If the starting point is the first task edge... For each task point, the elements in the starting point correlation matrix... Otherwise, the element is 0; for the first If the endpoint of the task edge is the first one... For each task point, then Otherwise, the element is 0; The decision problem for multi-UAV graph matching task is expressed as: (9); in, Represents the trace operation of a matrix. Indicates transpose. and They represent dimensions as follows: and A completely uniform vector.

4. The UAV map matching task decision method according to claim 1, characterized in that, The method employs a continuous-discrete joint optimization approach based on path following to solve the decision-making problem for multi-UAV graph matching tasks, yielding the final matching matrix between UAVs and tasks. Specifically: First, the original integer feasible field is represented as Relax the original integer feasible region into a continuous feasible region. When the number of drones equals the number of task points and a one-to-one match is required, the continuous feasible region can be written in the form of a double random matrix constraint: ; Constructing from path parameters Adjusted interpolation objective function: , (10); in, Describes the interpolation objective function. Indicates path parameters, This represents the initial relaxation objective function that facilitates continuous optimization. Let represent the objective function used to guide the solution to converge to the discrete integer domain. Specifically, take , ,in, It is the identity matrix. To adjust the parameters, , This represents the symmetric edge affinity matrix and node affinity vector obtained by factorization. ,vector ; Substituting the initial relaxation objective function, which facilitates continuous optimization, and the objective function used to guide the solution to converge to the discrete integer domain, into the interpolation objective function, we obtain: (11); remember ,but: (12); Path parameters during the solution process Incrementing gradually from 0 to 1: , ,in, Indicates the number of the path following stage. This represents the total number of stages in the path following process. Represents path parameters In the The values ​​of each path following stage, each This corresponds to a continuous optimization subproblem: ,exist At this point, an initial solution is first obtained in the continuous feasible region, and then the solution is gradually increased. The solution from the previous stage is used as the initial value for the next stage, so that the matching matrix gradually moves along the continuous path towards the original integer feasible region. For each path parameter Perform inner-layer iterative solution: Let the current iterative solution be... Its vectorized form is The interpolation objective function with respect to The gradient is , Indicates path parameter taking The quadratic coefficient matrix of the time-interpolation objective function is used to rearrange the gradient into matrix form: ,in, express The reverse operation, Indicates the first s The gradient matrix of the interpolation objective function with respect to the matching matrix during the next inner iteration has the same dimension as the matching matrix. X same; Then solve the following linear allocator problem: ,in, Represents the inner product of matrices. This represents the candidate matching matrix in the linear allocation subproblem. Indicates the first s The matching matrix obtained during the next inner iteration is obtained by solving the linear allocator subproblem. Then, the update direction is Next, in updating the direction Perform a one-dimensional search to determine the optimal step size: , Indicates the candidate step size, Indicates the first s The optimal step size obtained from the next inner iteration; make Then the step size optimization problem is about Maximizing a quadratic function in one variable: Therefore, if The candidate step size is then taken as and truncate it to the interval Inside: ,like Then compare them separately. and The interpolation objective function value is calculated, and a step size with a larger interpolation objective function value is selected. The matching matrix is ​​then updated. ; Stop the current process when any of the following conditions are met. The inner iteration below: ,or or reaching the maximum number of iterations. ,in, and These represent the threshold for changes in the objective function and the threshold for changes in the matching matrix, respectively. Represents the Frobenius norm, current path parameter The optimal solution is denoted as ,when When that happens, use it as the next path parameter. Initial values ​​for the following solutions: Repeat the above process until... Finally, in Output matching matrix ; When the elements in the output matching matrix are not 0 or 1, the final UAV-mission matching result is obtained according to the following rules: ,in, Indicates the first The drone was eventually assigned to the first One task point, This indicates the elements of the output matching matrix; if the number of task points is less than the number of drones, the unselected drones remain idle or are used as backup drones.

5. A UAV map matching task decision system, used to implement the UAV map matching task decision method according to any one of claims 1-4, characterized in that, include: The two-layer graph structure construction module is used to construct the current state and target task configuration of a multi-UAV cluster into a UAV resource capability graph and a task requirement graph, respectively, and calculate the node affinity matrix and edge affinity matrix to jointly form a two-layer graph structure for graph matching solution. The decision problem construction module is used to construct the objective function for multi-UAV graph matching task decision based on a two-layer graph structure, and to reconstruct the objective function for multi-UAV graph matching task decision into an equivalent trace form using factor decomposition, thereby constructing the multi-UAV graph matching task decision problem. The solution module uses a path-following-based continuous-discrete joint optimization method to solve the multi-UAV graph matching task decision problem, and obtains the final UAV-task matching matrix.

6. An electronic device, characterized in that, include: One or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform a UAV map matching task decision method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed, cause the processor to perform a UAV graph matching task decision method according to any one of claims 1-4.

8. A computer program product, characterized in that, It includes a computer program or instructions that, when executed by a processor, implement the UAV map matching task decision method according to any one of claims 1-4.