Rapid target allocation method suitable for large-scale unmanned aerial vehicle group and related equipment

Abstract

The invention discloses a rapid target allocation method suitable for a large-scale unmanned aerial vehicle cluster and related equipment, and relates to the technical field of unmanned aerial vehicle cluster control. By introducing the auction algorithm and utilizing a parallel bidding mechanism of the auction algorithm, a plurality of unmanned aerial vehicles independently execute target searching and bidding decisions each time, the parallel characteristic of the auction algorithm is fully played, the matching convergence speed is greatly increased, the matching time is greatly shortened, and rapid target matching of a large-scale unmanned aerial vehicle group is realized; in addition, through a cyclic matching mode, the solution speed is increased, the distribution quality is improved, and the final matching result and the optimal matching result are kept basically consistent.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm control technology, and in particular to a rapid target allocation method and related equipment suitable for large-scale UAV swarms. Background Technology

[0002] In recent years, drone light shows have gained widespread application in large-scale celebrations, commercial promotions, and cultural tourism due to their dynamic programmability, spatial flexibility, and stunning visual effects. Drone light shows require the safe, efficient, and coordinated movement of a large number of drones in three-dimensional space to accurately achieve desired patterns. The core technology lies in solving the path planning problem for large-scale drone swarms. Target allocation, as the first step in drone swarm path planning, directly impacts the overall efficiency of path planning and the smoothness of the performance through its efficiency and quality.

[0003] Currently, the target assignment methods used in drone performances generally employ the Hungarian algorithm. While this algorithm can achieve globally optimal task-drone matching results, it requires repeated matrix row and column operations, resulting in a time complexity of O(n³). The algorithm's complexity increases cubically with the number of drones and cannot be parallelized. When the scale of a drone performance increases (e.g., more than 5000 drones), the matching time of this target assignment algorithm becomes excessively long, significantly increasing the overall path planning time for the drone swarm and severely restricting the real-time performance of large-scale shows.

[0004] In summary, existing target allocation methods used in drone performances suffer from excessively long matching times when dealing with target allocation tasks for large-scale drone swarms, which severely impacts the overall path planning efficiency. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a fast target allocation method and related equipment suitable for large-scale drone swarms, so as to solve the technical problem that the target allocation method used in the existing drone performance has too long matching time when dealing with target allocation tasks of large-scale drone swarms.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The first aspect of this invention provides a fast target allocation method suitable for large-scale UAV swarms, comprising the following steps: S10, constructing a mathematical model of the UAV swarm target matching problem based on an auction algorithm; S20, initializing the set of UAVs to be allocated U, the set of target point costs P, the allocation set A, the incremental parameter ε, the incremental scaling ratio k, and the task reward of the target point; S30, solving the mathematical model based on the auction algorithm to perform bidding matching for each UAV, and updating the set of UAVs to be allocated U, the set of target point costs P, and the allocation set A based on the matching results; S40, determining whether the incremental parameter ε is greater than 1, if yes, proceeding to step S50; if no, returning the allocation set A; S50, updating the incremental parameter ε = ε / k, retaining the current cost of each target point, and resetting the set of UAVs to be allocated U and the allocation set A; S60, solving the mathematical model based on the auction algorithm to perform bidding matching for each UAV, updating the set of UAVs to be allocated U, the set of target point costs P, and the allocation set A based on the matching results, and jumping to step S40.

[0007] A second aspect of the present invention provides a rapid target allocation system suitable for large-scale UAV swarms, comprising at least one processor and at least one memory communicatively connected to the processor, wherein the memory stores program instructions that, when executed by the processor, implement the aforementioned rapid target allocation method suitable for large-scale UAV swarms.

[0008] A third aspect of the present invention provides a computer-readable storage medium having internally stored program instructions that, when executed by a processor, implement the aforementioned rapid target allocation method applicable to large-scale UAV swarms.

[0009] The beneficial technical effects of this invention are as follows: by introducing an auction algorithm and utilizing its parallel bidding mechanism, multiple UAVs can independently perform target search and bidding decisions each time, giving full play to the parallel characteristics of the auction algorithm, greatly accelerating the matching convergence speed, significantly reducing the matching time, and realizing rapid target matching of large-scale UAV swarms; in addition, by using a cyclic matching method, the solution speed is accelerated while the allocation quality is improved, so that the final matching result is basically consistent with the optimal matching result. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the rapid target allocation method for large-scale UAV swarms according to the present invention. Figure 2 This is a schematic diagram illustrating the results of 11,800 comparative tests. Figure 3 This is a diagram illustrating the ratios of the results from 11,800 comparative tests. Figure 4This is a diagram illustrating the relative error of the comparative test results from 11,800 sorties. Detailed Implementation

[0011] To enable those skilled in the art to more clearly understand the purpose, technical solution, and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0012] like Figure 1 As shown, in one embodiment of the present invention, the rapid target allocation method suitable for large-scale UAV swarms mainly includes steps S10 to S60: S10. Construct a mathematical model for the target matching problem of unmanned aerial vehicle (UAV) swarms based on the auction algorithm.

[0013] The original target matching problem of drone swarms can usually be modeled as the following mathematical problem: ; in, Let X be the reward obtained by drone i when it reaches target point j. Matrix X represents the allocation matrix, consisting of the elements of the decision variables. composition, The target point j is assigned to UAV i. and This represents a one-to-one correspondence between target points and drones. The optimal matching result is obtained when the total allocation benefit of the objective function is maximized.

[0014] Based on the duality principle and the optimality principle, the above mathematical problem can be transformed into: ; and For the new variables that need to be solved after the transformation, where This represents the current cost of target point j. Let represent the revenue gained by drone i. By observing the dual problem, it is found that a feasible solution to this mathematical problem must satisfy the following constraints: , equivalent to Therefore, based on the dual problem, the above mathematical model can be further optimized as follows: ; From this point on, the original UAV swarm target matching problem is optimized into a mathematical model based on the auction algorithm, where P is the set of target point costs, including the current cost of all target points; This represents the mission benefit of target point j for drone i; Let represent the current cost of target point j, and n be the number of drones.

[0015] By solving the mathematical model based on the auction algorithm, the target matching result of the drone swarm can be determined. This mathematical model based on the auction algorithm can be understood as achieving an expected equilibrium, i.e., obtaining the optimal solution, when any drone i obtains the target position j that yields the highest benefit to itself at a lower price.

[0016] S20. Initialize the set of drones to be assigned U, the set of target point costs P, the assignment set A, the incremental parameter ε, the incremental scaling ratio k, and the task benefits of the target points.

[0017] Step S20 is the initialization step, and the specific process includes: S21. Obtain drone information, including drone ID and drone location information.

[0018] Each drone contains two pieces of information: ID information and location information (where the ID information refers to the number of each drone in the drone swarm), and these two pieces of information are stored using a structure D. The ID information representing drone i, (This represents the location information of drone i).

[0019] S22. Construct and initialize the set of drones to be assigned, U. The initialized set of drones to be assigned includes the ID information of all drones.

[0020] The set of drones to be assigned, U, constructed in this step, is used to store the drone IDs of all unassigned target points. The initialized set of drones to be assigned, U, contains the IDs of all drones (the total number of drones is n), that is: .

[0021] S23. Construct and initialize the target point cost set P. After initialization, the cost value of all target points in the target point cost set P is 0.

[0022] The target point cost set P constructed in this step is used to store the current cost of all target points: , in This represents the current cost of target point j. The initial cost set P of the target point is set to 0, i.e., at the initial time step. .

[0023] S24. Construct and initialize the allocation set A. The initialized allocation set A is an empty set.

[0024] The allocation set A constructed in this step is used to store the current target allocation. The initialized allocation set A is empty, that is: .

[0025] S25. Set the incremental parameter ε.

[0026] In this step, the incremental parameter ε is initially set to ε > n / 10, where n is the number of drones.

[0027] S26. Set the incremental scaling ratio k, k > 1.

[0028] S27. Initialize the task rewards for each target point.

[0029] In this step, the task reward for each target point is calculated using the following formula: , in, This represents the mission benefit of target point j for drone i. This represents the location information of drone i. This represents the location information of the target point j.

[0030] S30. Solve the mathematical model based on the auction algorithm using the auction algorithm to match each UAV in the bidding process, and update the set of UAVs to be allocated, the set of target point costs, and the allocation set A based on the matching results.

[0031] The bidding process for each drone involves: S31. Select P drones from the set of drones to be assigned, U; S32. Use the auction algorithm to perform bidding matching on the selected P drones in parallel, and update the set of drones to be allocated U, the set of target point costs P, and the allocation set A based on the matching results.

[0032] In step S32, the selected P drones perform operations in parallel: S321. Find the target point j that yields the highest benefit to drone i, and calculate the highest benefit: , in, This represents the mission benefit of target point j for drone i. This represents the current cost of target point j.

[0033] S322. Find the target point m with the second highest benefit for drone i, and calculate the second highest benefit: , in, This represents the mission benefit of target point m for drone i. This represents the current cost of the target point m.

[0034] S323, UAV i bids for target point j: , in, This represents the current cost of target point j. To find the highest profit, In search of the second-highest return, The set incremental parameter.

[0035] S324. Assign target point j to drone i and store the assignment in assignment set A. Remove drone from drone set U to be assigned.

[0036] In step S324, when assigning target point j to drone i, if target point j has already been assigned to another drone, the previous assignment is removed, and the drone previously assigned to target point j is moved from the assignment set A to the drone set U to be assigned.

[0037] S325, Update the cost of target point j: , in, The bid made by drone i for target point j.

[0038] Step S32 involves simultaneously performing bidding and matching on P drones. For steps S321 to S325, from step S323 where drone i bids for target point j to step S324 where target point j is allocated to drone i, target point j is locked using a global lock to ensure that data does not conflict during parallel execution.

[0039] After all P drones to be selected have completed the bidding and matching process in steps S321 to S325, it is determined whether the set U of drones to be assigned is an empty set. If it is, the iteration ends; otherwise, it jumps to step S31 to continue the iteration, that is, steps S31 to S32 are repeated until the set U of drones to be assigned is an empty set.

[0040] S40. Determine if the incremental parameter ε is greater than 1. If yes, proceed to step S50; otherwise, return to the allocation set A.

[0041] It should be noted that, based on the auction algorithm principle, after executing step S30, the mathematical model based on the auction algorithm reaches an equilibrium point. At this point, the allocation result is not the optimal allocation. Therefore, this invention proposes to repeatedly execute the bidding matching through a cyclic matching method so that the allocation result continuously approaches the optimal allocation.

[0042] This invention controls the iterative loop by using an incremental parameter ε, dynamically decaying the incremental parameter during each matching iteration. ,when The loop terminates when the time is right, ensuring that the final matching result is basically consistent with the optimal matching result.

[0043] S50, Update the incremental parameter ε=ε / k, retain the current cost of each target point, and reset the set of drones to be assigned U and the assignment set A; (1) Update the incremental parameter ε = ε / k; (2) Retain the current cost of each target point, that is, retain the updated target point cost set P; (3) Reset the set of drones to be assigned, U, and reassign all drones to the set of drones to be assigned, U; (4) Reset the allocation set A, so that .

[0044] S60. Solve the mathematical model based on the auction algorithm using the auction algorithm to match each UAV in the bidding process. Update the set of UAVs to be assigned, the set of target point costs, and the set of assignments, A, based on the matching results, and then proceed to step S40.

[0045] After updating the incremental parameter ε and resetting the set of drones to be assigned U and the assignment set A, the auction algorithm is used again to solve the mathematical model based on the auction algorithm to match each drone in the bidding process.

[0046] The specific process of matching each drone in this step is the same as the specific process of matching each drone in step S30.

[0047] In step S60, the various drones are matched through a bidding process, which includes: S61. Select P drones from the set of drones to be assigned, U; S62. Use the auction algorithm to perform bidding matching on the selected P drones in parallel, and update the set of drones to be allocated U, the set of target point costs P, and the allocation set A based on the matching results.

[0048] In step S62, the selected P drones perform operations in parallel: S621. Find the target point j that yields the highest benefit to drone i, and calculate the highest benefit: , in, This represents the mission benefit of target point j for drone i. This represents the current cost of target point j.

[0049] S622. Find the target point m with the second highest benefit for drone i, and calculate the second highest benefit: , in, This represents the mission benefit of target point m for drone i. This represents the current cost of the target point m.

[0050] S623, UAV i bids for target point j: , in, This represents the current cost of target point j. To find the highest profit, In search of the second-highest return, The set incremental parameter.

[0051] S624. Assign target point j to drone i and store the assignment in assignment set A. Remove drone from drone set U to be assigned.

[0052] In step S624, when assigning target point j to drone i, if target point j has already been assigned to another drone, the previous assignment is removed, and the drone previously assigned to target point j is moved from the assignment set A to the drone set U to be assigned.

[0053] S625, Update the cost of target point j: , in, The bid made by drone i for target point j.

[0054] Step S62 involves simultaneously performing bidding and matching on P drones. For steps S621 to S625, from step S623 (where drone i bids for target point j) to step S624 (where target point j is allocated to drone i), target point j is locked using a global lock to ensure that no data conflicts occur during parallel execution.

[0055] After all P drones to be selected have completed the bidding and matching process in step S62 (including steps S621 to S625), it is determined whether the set U of drones to be assigned is an empty set. If it is, the iteration ends; if not, it jumps to step S61 to continue the iteration, that is, steps S61 to S62 are repeated until the set U of drones to be assigned is an empty set.

[0056] In step S60, after the bidding and matching of each drone is completed, the process jumps to step S40 to determine whether the conditions for ending the loop have been met.

[0057] The fast target allocation method for large-scale UAV swarms provided by this invention can achieve the following beneficial technical effects compared with traditional target allocation algorithms: (1) By introducing an auction algorithm and utilizing its parallel bidding mechanism, multiple UAVs can independently perform target search and bidding decisions each time, giving full play to the parallel characteristics of the auction algorithm, greatly accelerating the matching convergence speed, significantly reducing the matching time, and realizing rapid target matching of large-scale UAV swarms. (2) By using a cyclic matching method and dynamically decaying the incremental parameter during each cyclic matching. ( (Terminate the loop when the time is right), which speeds up the solution process and improves the quality of the allocation, so that the final matching result is basically consistent with the optimal matching result; (3) Unlike traditional target assignment algorithms that only use the distance norm as the task reward, this invention constructs a new task reward: .

[0058] The advantages of this construction method are: (3.1) The minimum distance between the straight-line trajectories of each UAV to its target point is guaranteed by using the squared distance norm, based on the matching results. There exists a lower bound that reduces the pressure on subsequent path planning, namely , in, This represents the minimum distance between any two drones at their starting positions. This represents the minimum distance between any two drones at the target location; (3.2) Multiply by the square of the distance norm. ,Cooperate The cyclic matching ensures that the final allocation result is basically consistent with the optimal matching result.

[0059] To test the performance of the fast target allocation method of this invention applicable to large-scale UAV swarms, comparative tests were conducted on UAV swarms of four different sizes: 200, 1000, 7000, and 11800 UAVs (the 7000-UAV swarm included two different test cases, denoted as 7000-A and 7000-B). In each test, the initial and target positions were randomly shuffled, and the Hungarian algorithm and the fast target allocation method of this invention were executed 1000 times each. The average of the total test time was taken. The test results are as follows: It should be noted that, based on the test results in the table, when the number of test sorties is 200, the processing time of the Hungarian algorithm is shorter than that of the fast target allocation method proposed in this invention. This is because the fast target allocation method proposed in this invention requires cyclic matching, which results in additional time consumption when processing small-scale UAV swarm matching such as 200 sorties. However, when processing UAV swarm matching problems exceeding 1,000 sorties, the processing time of the fast target allocation method proposed in this invention is significantly faster.

[0060] Figure 2 It further showcased the results of 11,800 comparative tests.

[0061] In description Figure 2 Before comparing the test results, define the variable as the sum of squared distances M: ; in, This indicates the location information of drone i. This represents the location information of the matched target point j obtained based on the target assignment algorithm, where n is the total number of drones.

[0062] Figure 2 The red line represents the sum of squared distances (M1) obtained from the allocation results when the Hungarian algorithm terminates, and the blue dots represent the sum of squared distances (M2) obtained from the allocation results based on the fast target matching method proposed in this invention. Figure 3 The comparison test results of 11,800 sorties are shown. The ratio point in the figure is the ratio of M2 to M1, i.e., M2 / M1. Figure 4 The figure shows the relative error of 11,800 comparative tests, with the error points representing the ratio (M2-M1) / M1. The closer M2 / M1 is to 1 and (M2-M1) / M1 is to 0, the closer the proposed fast target matching method is to the optimal solution (the matching result obtained by the Hungarian algorithm is the optimal solution). Figure 3 and Figure 4 As can be seen, the fast target matching method proposed in this invention is basically consistent with the results obtained by the Hungarian algorithm.

[0063] The test results show that, compared with the Hungarian algorithm, the fast target allocation method proposed in this invention significantly improves the solution speed when dealing with target allocation tasks of large-scale UAV swarms, and ensures that the allocation results are basically consistent with the optimal matching results (the relative error of the final matching results is close to 0, and the average ratio is close to 1), which greatly improves the matching efficiency in the path planning process of large-scale UAV performances.

[0064] The present invention also provides a rapid target allocation system suitable for large-scale UAV swarms, comprising at least one processor and at least one memory communicatively connected to the processor, wherein the memory stores program instructions, which are executed by the processor to implement... Figure 1 The steps of the rapid target allocation method for large-scale UAV swarms shown in the embodiment.

[0065] The present invention also provides a computer-readable storage medium having internally stored program instructions that, when executed by a processor, implement... Figure 1 The steps of the rapid target allocation method for large-scale UAV swarms shown in the embodiment.

[0066] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any way. Those skilled in the art can make various equivalent changes and improvements based on the above embodiments, and all equivalent variations or modifications made within the scope of the claims should fall within the protection scope of the present invention.

Claims

1. A rapid target allocation method suitable for large-scale UAV swarms, characterized in that, Includes the following steps: S10. Construct a mathematical model for the target matching problem of unmanned aerial vehicle (UAV) swarms based on the auction algorithm; S20. Initialize the set of drones to be assigned U, the set of target point costs P, the assignment set A, the incremental parameter ε, the incremental scaling ratio k, and the task benefits of the target points; S30. Solve the mathematical model based on the auction algorithm to match each UAV in the bidding process, and update the set of UAVs to be allocated, the set of target point costs, and the allocation set A based on the matching results; S40. Determine if the incremental parameter ε is greater than 1. If yes, proceed to step S50; otherwise, return to the allocation set A. S50, Update the incremental parameter ε=ε / k, retain the current cost of each target point, and reset the set of drones to be assigned U and the assignment set A; S60. Solve the mathematical model based on the auction algorithm to match each UAV in the bidding process. Update the set of UAVs to be allocated, the set of target point costs, and the allocation set A based on the matching results, and then jump to step S40.

2. The rapid target allocation method for large-scale UAV swarms as described in claim 1, characterized in that, Step S30 further includes: S31. Select P drones from the set of drones to be assigned, U; S32. Use the auction algorithm to perform bidding matching on the selected P drones in parallel, and update the set of drones to be allocated U, the set of target point costs P, and the allocation set A based on the matching results; Repeat steps S31 to S32 until the set of drones to be assigned, U, is an empty set.

3. The rapid target allocation method for large-scale UAV swarms as described in claim 2, characterized in that, Step S32 further includes: S321. Find the target point j that yields the highest benefit to drone i, and calculate the highest benefit: ， in, This represents the mission benefit of target point j for drone i. This represents the current cost at target point j, and n is the number of drones. S322. Find the target point m with the second highest benefit for drone i, and calculate the second highest benefit: ， in, This represents the mission benefit of target point m for drone i. This represents the current cost of the target point m; S323, UAV i bids for target point j: ； S324. Assign target point j to drone i and store the assignment in assignment set A. Remove drone from drone set U to be assigned. S325, Update the cost of target point j: ； in, The bid made by drone i for target point j.

4. The rapid target allocation method for large-scale UAV swarms as described in claim 3, characterized in that, In step S324, when assigning target point j to drone i, if target point j has already been assigned to other drones, the drones previously assigned to target point j are moved from the assignment set A to the drone set U to be assigned.

5. The rapid target allocation method for large-scale UAV swarms as described in claim 1, characterized in that, Step S20 further includes: S21. Obtain drone information, including drone ID information and drone location information; S22. Construct and initialize the set of drones to be assigned, U. The initialized set of drones to be assigned includes the ID information of all drones. S23. Construct and initialize the target point cost set P. After initialization, the cost value of all target points in the target point cost set P is 0. S24. Construct and initialize the allocation set A. The initialized allocation set A is an empty set. S25. Set the incremental parameter ε; S26. Set the incremental scaling ratio k, k > 1; S27. Initialize the task rewards for each target point.

6. The rapid target allocation method for large-scale UAV swarms as described in claim 5, characterized in that, Step S27 calculates the task reward for each target point using the following formula: ， in, This represents the mission benefit of target point j for drone i. This represents the location information of drone i. This represents the location information of target point j, and n is the number of drones.

7. The rapid target allocation method for large-scale UAV swarms as described in claim 5, characterized in that, In step S25, the initially set incremental parameter is ε > n / 10, where n is the number of drones.

8. The rapid target allocation method for large-scale UAV swarms as described in claim 1, characterized in that, In step S10, the mathematical model based on the auction algorithm is expressed as follows: ； Where P is the set of target point costs, including the current cost of all target points; This represents the mission benefit of target point j for drone i; Let represent the current cost of target point j, and n be the number of drones.

9. A rapid target allocation system suitable for large-scale UAV swarms, characterized in that, It includes at least one processor and at least one memory communicatively connected to the processor, wherein the memory stores program instructions that, when executed by the processor, implement the rapid target allocation method for large-scale UAV swarms as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium internally stores program instructions that, when executed by a processor, implement the rapid target allocation method for large-scale UAV swarms as described in any one of claims 1 to 8.

Images