A multi-task dynamic allocation method for drone swarm

Through the dynamic task allocation method of drone cluster, the response radius and sub-team mechanism are used to optimize the task insertion position, solving the problem of low task allocation efficiency in dynamic environments, achieving efficient task reallocation and resource utilization, and enhancing the drone The collaborative execution capability of the cluster.

CN115202402BActive Publication Date: 2024-11-19GUANGZHOU HUM INTELLIGENT TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211001316.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-19
Publication Date
2024-11-19
Estimated Expiration
2042-08-19

AI Technical Summary

Technical Problem

The task allocation algorithm of existing drone clusters is difficult to deal with multiple emergencies in a dynamic environment, the response efficiency is low and the computing resources are wasted. The original algorithm ignores the initial plan of task scheduling, resulting in unoptimized task allocation.

Method used

A dynamic allocation method for multi-tasking in the drone cluster is proposed. Through the drone, the response radius is determined based on the dynamic tasks, the drone sub-team is formed, the task information is shared, the task list is updated, and the task reassignment is carried out under the distributed architecture to optimize task insertion. Position for efficiency.

Benefits of technology

It realizes rapid response and task reallocation in a dynamic environment, improves the task execution efficiency of the drone cluster, reduces waste of computing resources, adapts to complex and changeable task scenarios, and enhances the timeliness of collaborative execution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115202402B_ABST
    Figure CN115202402B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of unmanned aerial vehicles, and in particular relates to a method for dynamically allocating multiple tasks of an unmanned aerial vehicle cluster, comprising: a unmanned aerial vehicle determines the response radius of a task according to a dynamically emerging new task; unmanned aerial vehicles within the response range of the task form a unmanned aerial vehicle sub-team for the task, and the unmanned aerial vehicle sub-team competes for the task; the current unmanned aerial vehicle forms a new task set with all tasks that need to be reallocated, and the new task set is the dynamically emerging new task and the unfinished tasks of each unmanned aerial vehicle in the unmanned aerial vehicle sub-team; the unmanned aerial vehicles share task information and update the local task list, and share the updated task list with other unmanned aerial vehicles, and complete the task allocation when the task list of the unmanned aerial vehicle is no longer updated due to the shared information; the present invention realizes dynamic task allocation of an unmanned aerial vehicle cluster, can adapt to complex and changeable task scenarios, enhances the timeliness of the coordinated execution of tasks by the unmanned aerial vehicle cluster, and effectively solves the problem of emergencies occurring during the task execution process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicles, and in particular relates to a method for dynamically allocating multiple tasks of an unmanned aerial vehicle cluster. Background Art

[0002] Nowadays, drone technology is becoming more and more mature, and drone-related research has made great breakthroughs in theoretical exploration and practical application. Drone swarms can work together to replace people to complete tasks that are beyond the limits of human endurance. The task allocation problem is to assign a set of tasks to drones without conflict or violation of constraints, which is an important part of the successful operation and efficient collaboration of multi-drone systems.

[0003] Most existing studies have solved the static version of the multi-UAV system task allocation problem, in which all tasks are known at the beginning and the environment will not change. But in fact, the environment is a dynamic event that is constantly changing. That is, in a dynamic environment, the original task allocation scheme may be interrupted by many unexpected events during the execution of the task, which may cause some tasks to fail to be successfully completed. Therefore, it is necessary to consider the problem of task reallocation in a dynamic environment.

[0004] However, there are many difficulties in task allocation in dynamic environments that need to be solved. First, the proposed method should be able to cope with various dynamic events. However, most existing algorithms can only handle a few dynamic events, usually by re-running the algorithm, ignoring the differences between different dynamic events. Second, since the dynamic battlefield changes rapidly, the algorithm needs to respond to different dynamic events as quickly as possible, which poses a higher challenge to the efficiency of the algorithm. Although re-running can obtain optimal or suboptimal solutions, it requires more computing time and is not suitable for dynamic environments. In addition, re-running always ignores the originally planned scheduling, resulting in a waste of computing resources. Summary of the invention

[0005] In order to respond to dynamic events that occur during the execution of UAV tasks and improve the efficiency of solving dynamic events, the present invention proposes a multi-task dynamic allocation method for UAV clusters.

[0006] S1. The drone determines the response radius of the task according to the dynamically emerging new task;

[0007] S2. The drones within the response range of the task form a drone sub-team for the task, and the drone sub-team competes for the task;

[0008] S3, the current drone forms a new task set by all tasks that need to be reallocated. The new task set is the dynamically emerging new tasks and the unfinished tasks of each drone in the drone sub-team;

[0009] S4. The drones share task information and update the local task list, and share the updated task list with other drones. When the drone's task list is no longer updated due to the shared information, the task allocation is completed. The process of updating the local task list includes the following steps:

[0010] 401. The benefits of the current candidate task at different insertion positions in the current drone calculation task set;

[0011] 402. Select the position where the marginal benefit is the largest as the insertion point of the current candidate task;

[0012] 403. Determine whether the marginal benefit of inserting the current candidate task in step 402 is greater than the benefit of the winner of the task in the local winner list; if not, return to 401 and calculate the next candidate task;

[0013] 404. If it is greater than, then add the task to the task timing package of the current UAV, and determine whether the task timing package of the current UAV exceeds the upper limit. If it exceeds, then share the task information with other UAVs, otherwise return to step 401 to calculate the next candidate task;

[0014] 405. When the number of tasks in the task sequence package of the current drone reaches the upper limit, the current drone shares the task allocation information with other drones and updates the local task information. If the local task information of a drone is inconsistent with the shared information, the bundle is updated and 401 is executed; if they are consistent, the task allocation result is output.

[0015] Furthermore, the response radius of the task includes:

[0016]

[0017] Among them, r u Indicates the response radius of the drone; X max Indicates the maximum distance in the generated map space; λ is a conversion parameter; R max Represents the maximum value among all tasks, R u Indicates the value of the newly added task; l max Indicates the latest start time of all tasks, l u Indicates the latest start time of the new task, t ψ Indicates the current moment; ω 1 To balance the weight of the static value on the response radius, ω 2 The weight for balancing the impact of the cut-off time on the response radius.

[0018] Furthermore, the process of the drone calculating the position where a task is inserted into the drone's task timing package with the greatest benefit includes:

[0019]

[0020] Constraints:

[0021]

[0022]

[0023]

[0024] Among them, N V is the number of drones, all drones form a drone set N M is the number of tasks, all tasks form a task set c jk Indicates drone U j Execute Task M k The income at that time; X jk is a binary variable, X jk =1 means UAV U j Execute Task M k , otherwise it means that the drone U j Unexecuted task M k ; L T Indicates the maximum capacity of the mission that the drone can perform.

[0025] Furthermore, UAV j Execute Task M k The profit at time c jk It is expressed as:

[0026]

[0027] Constraints:

[0028] in, Indicates that the kth task M k Insert into task list p j The task list formed at the nth position in Indicates the list of tasks that the drone performs Rewards received; Represents the list of tasks executed by the drone p j Rewards received; jk [b i ] indicates execution of task list b i of income;

[0029] Indicates execution in task list b i The benefit of adding a new task b to the end of the new task list.

[0030] Furthermore, the kth task M k Insert into task list P j After reaching the nth position in the list, the drone executes the task list P j Rewards It is expressed as:

[0031]

[0032] Among them, R 0 For task M k The basic value of R is the task M k The static value of j is the discount factor, Indicates execution of task M k Time, ET k Represents task M k The earliest start time, Indicates that in the task list p j Insert task M at position n k marginal cost at time.

[0033] Furthermore, in the task list p j Insert task M at position n k Marginal cost at It is expressed as:

[0034]

[0035] in, Indicates execution in task list p i Insert task M at the nth position in k The distance required for all tasks in the task list after that; D(p i ) indicates the execution of task list p i The range required for all tasks in j Indicates fuel consumption.

[0036] Furthermore, the local mission information of the drone includes mission bundle, mission timing package, winner list, winner payoff set, timestamp and execution time set, where:

[0037] The task bundle is the set of tasks that the current drone competes for. The order of tasks in the bundle is arranged according to the order in which the tasks are added. The size of the bundle is the maximum number of tasks that the drone can perform;

[0038] The mission sequence package is a collection of UAV execution tasks, and the order in the package is arranged according to the priority of task execution;

[0039] The winner list is a list of drones with the highest profit in each task. If the element in this list is 0, it means that no drone has won the bid for this task yet;

[0040] The winner's profit set is the highest profit of each task. If the element in this list is -1, it means that no drone has won the bid for this task.

[0041] The timestamp is the moment when two drones receive information from each other;

[0042] The execution time set is the execution time when the UAV arrives at the mission area to perform the mission according to the mission timing set.

[0043] Further, the update of the list of winners of drones, that is, when drones U j Get Task M k When task M k Will be inserted into the UAV j The task position that currently increases the marginal benefit the most in the path. If the benefit of this task position is greater than the benefit of the winner of this task, then UAV U j Add the task to your task bundle, update the bundle and add the task M k The information is shared with other drones.

[0044] Furthermore, if the drone U a Received UAV b Shared about task M c When the information of UAV a Perform updates on the following information:

[0045] When ac <y bc When the UAV a Update local information and replace y bc Assign to y ac , put z bc Assign to z ac ;

[0046] When ac >y bc When the UAV a Do not update local information, y ac 、z ac remain unchanged;

[0047] When ac =y bc When the timestamps in the tasks conflict or are inconsistent, y ac Set to -1, z ac Set to 0;

[0048] Among them, y bcIndicates drone U b The winner set of task M c The maximum benefit value of y ac Indicates drone U a The winner set of task M c The maximum benefit value of z bc Indicates drone U b Perform task M on the winner set c The most profitable drone; ac Indicates drone U a Perform task M on the winner set c The most profitable drone.

[0049] The present invention enables drones to quickly respond to sudden dynamic environments and reallocate tasks under a distributed architecture. By proposing the response radius of drones, the drone sub-team for task allocation is selected, and the tasks are screened and selected to participate in the reallocation, which greatly increases the operating efficiency of the algorithm and can better adapt to dynamic environments that require rapid response. The task allocation method of the present invention achieves dynamic task allocation of drone clusters with the highest overall benefit and the shortest completion time. In the face of sudden dynamic situations, it can adapt to complex and changeable task scenarios, enhance the timeliness of drone clusters' collaborative execution of tasks, and effectively solve the problem of emergencies during task execution. In addition, the algorithm does not ignore the originally planned scheduling, thereby reducing the waste of computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of the process used in the present invention;

[0051] Figure 2 An initial allocation diagram for the method adopted by the present invention;

[0052] Figure 3 A task allocation diagram at any time of the method adopted by the present invention;

[0053] Figure 4 A schematic diagram of a dynamic sub-team using the method of the present invention;

[0054] Figure 5 This is a task redistribution diagram of the method adopted by the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] The present invention proposes a method for dynamically allocating multiple tasks of a drone cluster. Figure 1 , mainly including the following steps:

[0057] S1. The drone determines the response radius of the task according to the dynamically emerging new task;

[0058] S2: Unmanned aircraft within the response range of the task form a drone sub-team for the task, and the drone sub-team competes for the task;

[0059] S3, the current drone forms a new task set by all tasks that need to be reallocated. The new task set is the dynamically emerging new tasks and the unfinished tasks of each drone in the drone sub-team;

[0060] S4. The drones share task information and update the local task list, and share the updated task list with other drones. When the drone's task list is no longer updated due to the shared information, the task allocation is completed. The process of updating the local task list includes the following steps:

[0061] 401. The benefits of the current candidate task at different insertion positions in the current drone calculation task set;

[0062] 402. Select the position where the marginal benefit is the largest as the insertion point of the current candidate task;

[0063] 403. Determine whether the marginal benefit of inserting the current candidate task in step 402 is greater than the benefit of the winner of the task in the local winner list; if not, return to 401 and calculate the next candidate task;

[0064] 404. If it is greater than, then add the task to the task timing package of the current UAV, and determine whether the task timing package of the current UAV exceeds the upper limit. If it exceeds, then share the task information with other UAVs, otherwise return to step 401 to calculate the next candidate task;

[0065] 405. When the number of tasks in the task sequence package of the current drone reaches the upper limit, the current drone shares the task allocation information with other drones and updates the local task information. If the local task information of a drone is inconsistent with the shared information, the bundle is updated and 401 is executed; if they are consistent, the task allocation result is output.

[0066] In this implementation, the following operations are mainly performed when performing dynamic task allocation of drone clusters:

[0067] Taking into account multiple constraints such as resource consumption, time sensitivity, task completion effect, heterogeneity of UAVs and tasks, an optimization model for task allocation was established;

[0068] Considering the task allocation model under multiple constraints, the design of the benefit function with high overall task completion benefits and fast dynamic event response speed as the optimization goals;

[0069] Use the consensus bundling algorithm (CBBA) to solve the initial allocation of tasks;

[0070] This embodiment also optimizes CBBA so that it can quickly respond to sudden dynamic environments and reallocate tasks under a distributed architecture.

[0071] In this embodiment, N is first randomly generated V = 5 heterogeneous drones and N M = 15 heterogeneous tasks (U j represents the jth drone, M k represents the kth task), the information of the drone includes its drone number (id), drone type (type), geographic location coordinates (x, y, z), flight speed (v), and fuel consumption rate (f); the information of the target task includes its task number (id), task type (type), geographic location coordinates (x, y, z), task value (R), task time window constraint (ET k ≤t k ≤LT k ). The flight speed v of the drone is set to 1m / s, the fuel consumption rate is 3, the value of the initial task is set to 100, the size of the task area is set to 100m×100m, and the randomly generated drone information and task information are shown in Tables 1 and 2.

[0072] Table 1 UAV information table

[0073] Drone Number Position-x coordinate (m) Position-y coordinate (m) 0 24.7149 82.5224 1 73.6149 88.3472 2 35.6994 44.4896 3 65.2416 12.8390 4 33.5278 98.5542

[0074] Table 2 Task information table

[0075] Task Number Position-x coordinate (m) Position-y coordinate (m) Time window 0 84.7805 92.3331 366.3379-381.3379 1 71.8327 37.6225 129.4821-144.4821 2 18.4526 26.5405 211.8747-226.8747 3 16.8289 30.6515 81.6287-96.6287 4 10.8252 86.9999 119.0905-134.0905 5 46.3167 48.1102 21.4519-36.4519 6 73.3712 5.5017 153.8478-168.8478 7 33.4065 78.1626 173.2852-188.2852 8 61.0642 64.2355 409.8607-424.8607 9 91.5519 61.3491 5.4063-20.4063 10 25.4309 16.9998 431.6825-446.6825 11 28.2539 46.3951 157.2171-172.2171 12 81.4646 39.9064 80.4074-95.4074 13 73.3181 35.6978 424.7443-439.7443 14 52.5859 70.4896 383.2183-398.2183

[0076] N V Heterogeneous UAVs are denoted as a set N M Heterogeneous tasks are denoted as a set U j represents the jth drone, M kRepresents the kth task, establishes a task allocation model, and obtains the number of drones U j Execute Task M k The maximum profit c jk , the process includes:

[0077]

[0078] Constraints:

[0079]

[0080]

[0081]

[0082] Among them, P j is the execution list of tasks; X jk is a binary variable, X jk =1 means the drone performs the task; L T Indicates the maximum capacity of the mission that the drone can perform. jk It represents the income obtained when the drone performs the task, and its calculation formula is defined as follows:

[0083]

[0084] in, Indicates inserting the kth task into the task list P j The nth position in Represents the drone's task list P j The total score reward that can be obtained, so c jk It can be defined as the reward of inserting task k into the execution list to obtain the maximum benefit at a certain position. In the original CBBA, the reward function S is defined as:

[0085]

[0086] Among them, R is the static value of the task, λ j is a custom discount factor, Indicates the time to execute the task, f j Indicates fuel consumption, d jk represents the Euclidean distance between the initial position of the UAV and the mission target. k Inserted into p j When the UAV is in the path, the marginal benefit value of the task at the subsequent position in the path will be affected accordingly, which is not considered in the original CBBA. As for the marginal cost, the original CBBA only calculates the static distance cost between the initial position of the UAV and the task target, that is, fj d jk However, the marginal cost is essentially related to the actual position of the UAV in the mission path. Therefore, in order to improve the accuracy of the marginal cost function, a new range restriction for the UAV is added in this method, namely:

[0087]

[0088] It means that after the new task is added, the flight distance of the task path of the drone must be less than the maximum flight distance of each drone. Therefore, we can also change the static distance cost between the initial position of the drone and the task target into the distance cost of the drone completing M k The marginal cost is calculated by the extra voyage cost required for the mission. That is:

[0089]

[0090] in, Indicates that in the task list p j Insert task M at position n k marginal cost at time.

[0091] Therefore, considering the marginal benefit and marginal cost, the marginal reward function newly defined by this method is:

[0092]

[0093] Among them, R 0 is the basic value to avoid the total score reward becoming negative, R is the value of the task, and λ j is a custom discount factor, Indicates the time to execute the task, ET k Indicates the earliest start time of the task. Represents marginal cost.

[0094] After the model is built, this embodiment further performs initial task allocation on the drone, which mainly includes two parts, namely, the bundle building stage and the conflict resolution stage.

[0095] Each drone can perform a maximum of L T Each mission also has the following data information structure:

[0096] (1) Task Bundle: where b jn is the nth target won by the drone in the auction, |b j | represents the number of tasks in the task bundle. The tasks in the task bundle are arranged in the order in which they were added to the bundle.

[0097] (2) Task timing package (Path): p j refers to the optimal execution order of the jth UAV, p jn represents the task performed by the jth UAV nth time, |p j | represents the number of tasks in the task sequence package of the jth UAV, that is, the jth UAV is assigned a task according to p j The tasks are executed in the order in which they are performed.

[0098] (3) Winners: z jk is the number of the UAV with the highest profit for the kth task recorded by the jth UAV, N M is the number of tasks to be reallocated. If no drone currently bids for a task, then z jk =0.

[0099] (4) Winner's bid: represents the set of maximum benefits during the drone mission auction, y jk represents the maximum profit of the kth task recorded by the jth drone. If no drone currently bids for the task, then y jk =-1.

[0100] (5)Time stamp: represents the set of moments when two drones receive information from each other, s ji represents the time when the jth UAV receives information from the ith UAV, N V Indicates the number of drones. This parameter is used to compare the information status of the drones themselves. If two drones share information about a certain task and the timestamps are inconsistent, the drones need to reset the values ​​of the corresponding tasks in the winner set and winner benefit set, that is, set the value of the object corresponding to the winner of the task to 0, and the value of the winner's benefit to -1.

[0101] Execution time collection (Time): t jn Indicates the time sequence set p when the jth UAV starts to perform the mission j The time of the nth task in the task sequence set p j The execution time of arriving at the mission area to execute the mission.

[0102] Bundle construction: Each UAV unit uses a sequential greedy strategy to bid on multiple tasks and add its own task bundle until the quantity limit is reached. j Get Task M k When task M kIt will be inserted into the task position that currently increases the marginal benefit the most in its own path. If the benefit of the task is greater than the bid of the winner in its own information, the task will be added to the bundle. Indicates that when UAV U j Get Task M k When it is inserted into the nth position, the profit is c jk ; In order for these units to converge in communication, the marginal benefits need to gradually decrease, that is:

[0103]

[0104]

[0105] Conflict resolution: After each round of bundle updates, drones share information to resolve differences in task knowledge, measure information update time via timestamps, and unify the winner list and bid list.

[0106] After each round of bundle update, the UAVs share information to resolve task cognition differences, measure information update time by timestamp, and unify the winner list and bid list. For example, in this UAV cluster, when UAV U a Receive UAVU b Shared task information M c The update requirements of information are as follows:

[0107] When ac <y bc When the UAV a Update local information and replace y bc Assign to y ac , put z bc Assign to z ac ;

[0108] When ac >y bc When the UAV a Do not update local information, y ac 、z ac remain unchanged;

[0109] When ac =y bc When the timestamps in the tasks conflict or are inconsistent, y ac Set to -1, z ac Set to 0;

[0110] Among them, y bc Indicates drone U b The winner set of task M c The maximum benefit value of y acIndicates drone U a The winner set of task M c The maximum benefit value of z bc Indicates drone U b Perform task M on the winner set c The most profitable drone; ac Indicates drone U a Perform task M on the winner set c The most profitable drone.

[0111] During the operation of the algorithm, some tasks in the package will be returned to the task pool after the conflict is resolved. After multiple communications, the winner list z of each drone is j It does not change with the bid of the winning bidder, that is, the task allocation result reaches convergence. The running results are as follows Figure 2 shown.

[0112] Next, this embodiment constructs the model dynamic information and generates dynamic events. First, when a dynamic event occurs, it is necessary to obtain the drone and task information at that moment in advance. The drone information mainly includes the location information of the drone at that moment when the dynamic event occurs, and the task information includes the list of tasks that the drone has not completed at the current moment. The method of obtaining the dynamic location information of the drone is as follows:

[0113]

[0114]

[0115]

[0116] Among them, t ψ represents the current time; t p Indicates that at t ψ The start time of the last task that was just completed; t n Indicates that at t ψ The start time of the next task to be executed; t p-end Indicates the end time of the previous task; t p-dur Indicates t ψ The duration of the previous task before time; x (t), χ y (t), χ z (t) are the positions of the drone on the X-axis, Y-axis, and Z-axis at time t, respectively.

[0117] In this embodiment, the time when the dynamic task appears is randomly set to 250 seconds, and the task number of the new task is set to 15. Therefore, the drone and task information at time 250 are shown in Tables 3 and 4, and the task allocation is shown in Table 3. Figure 3 shown.

[0118] Table 3 UAV information table

[0119]

[0120]

[0121] Table 4 Task information table

[0122] Task Number Position-x coordinate (m) Position-y coordinate (m) Time window 0 84.7805 92.3331 366.3379-381.3379 8 61.0642 64.2355 409.8607-424.8607 10 25.4309 16.9998 431.6825-446.6825 13 73.3181 35.6978 424.7443-439.7443 14 52.5859 70.4896 383.2183-398.2183 15 38.2387 88.4863 401.5486–416.5486

[0123] This embodiment uses the optimized partial reallocation algorithm to handle dynamic events where new tasks appear. When a dynamic event occurs, if a new task appears, if the original CBBA algorithm is still used to respond to the dynamic event, all current drones will be reallocated, which may cause drones that are far away from the new task target at this moment to participate in the reallocation process. The result of the allocation is that the drones at a long distance still execute the initial task allocation plan, and new tasks will not be allocated to these drones. Therefore, this causes more waste of resources and causes unnecessary drones to participate in the new task allocation.

[0124] In order to solve the above problems, this embodiment proposes a concept of a drone response radius, that is, each drone has its own response radius r u When a new task appears, each drone will check whether the task is within its response radius. If it contains a new task, the drone will participate in the new task redistribution. If the drone is within the response range of the new task, the drones with new tasks in the response radius will form a new drone sub-teamΩ u To carry out the final task redistribution, the response radius r u It will change accordingly based on the value of the newly emerged task and the time window constraint. When the value of the task is greater, the multi-UAV system hopes to allow coordination by selecting a larger sub-team, so a larger response radius will be adopted; when the start time of the task is closer to the current time, the task needs to be assigned faster, so a smaller sub-team can be selected, and a smaller response radius will be adopted. The calculation formula for the response radius is as follows:

[0125]

[0126] Through calculation, it is found that r of this embodiment u =41.2631, the sub-team information of the drone is shown in Table 5, and the allocation at this time is as follows Figure 4 shown.

[0127] Table 5 UAV sub-team information table

[0128] Drone Number Position-x coordinate (m) Position-y coordinate (m) 0 39.4786 75.7334 4 33.5278 98.5542

[0129] In the original CBBA algorithm, all tasks in the task list will be allocated. Therefore, when a dynamic event occurs, if the original CBBA algorithm is still used, the tasks in the execution list of each drone after the initial allocation will be redundantly allocated again. This may also cause the drone to execute the initially allocated task list after the allocation, which will affect the allocation efficiency and cause a certain amount of resource waste. To address this problem, when new tasks appear, each drone in the drone sub-team will abandon the same number of tasks as the new tasks from its own package, and keep the allocation relationship of the remaining tasks stable. These abandoned tasks will also be those with the lowest value scores in their own task lists. Finally, these abandoned tasks will be redistributed together with the newly appeared tasks. Also, because in bundle b j When the drone is created, it adds tasks according to the sequential greedy strategy, so in the bundle b j The benefits of tasks in are decreasing in order. Then, when we need to give up tasks with the same number of lower scores when new tasks appear, we only need to add j Finally, delete the corresponding number of tasks and include them in the reset.

[0130] Through the above method, it can respond quickly to dynamic events when they occur, and finally get the following Figure 5 The allocation results are shown, where red dots represent members of the drone sub-team and blue triangles represent other drones that follow the original task allocation plan.

[0131] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-task dynamic allocation method for drone clusters, characterized in that: The specific steps include: S1. If a new dynamic task appears, the drone determines whether the task is within its response range. If so, it participates in the redistribution of the task; The response radius of the drone's mission includes: Among them, r u Indicates the mission response radius of the drone; X max Indicates the maximum distance in the generated map space; λ is a conversion parameter; R max Represents the maximum value among all tasks, R u Indicates the value of the newly added task; l max Indicates the latest start time of all tasks, l u Indicates the latest start time of the new task, t ψ represents the current moment; ω1 is the weight of the influence of the balance static value on the response radius, and ω2 is the weight of the influence of the balance deadline on the response radius; S2. The drones that respond to the new dynamic task are organized into a drone sub-team for the task, and the drone sub-team performs task redistribution; S3. The current drone forms a new task set by all tasks that need to be reallocated. The new task set is the dynamically appearing new tasks and each drone in the drone sub-team deletes the task that was added to the task bundle list the latest from the task list according to the number of new dynamic tasks. The number of deleted tasks is consistent with the number of new dynamic tasks. S4. The drones share task information and update the local task list, and share the updated task list with other drones. When the drone's task list is no longer updated due to the shared information, the task allocation is completed. The process of updating the local task list includes the following steps:

401. The benefits of the current candidate task at different insertion positions in the current drone calculation task set; 402. Selecting the position with the maximum marginal benefit as the insertion point of the current candidate task; the process of the drone calculating the position with the maximum benefit for inserting a task into the task timing package of the drone includes: Constraints: Among them, N V is the number of drones, all drones form a drone set N M is the number of tasks, all tasks form a task set c jk (p j ) indicates drone U j Execute Task M k The income at that time; X jk is a binary variable, X jk =1 means UAV U j Execute Task M k , otherwise it means that the drone U j Unexecuted task M k ; L T Indicates the maximum capacity of the mission that the drone can perform; UAV j Execute Task M k The profit at time c jk (p j ) is expressed as: Constraints: in, Indicates that the kth task M k Insert into task list p j The task list formed at the nth position in Indicates the list of tasks that the drone performs Rewards received; Represents the list of tasks executed by the drone p j Rewards received; jk [b i ] indicates execution of task list b i of income; Indicates execution in task list b i The benefit of adding a new task b to the end of the new task list; The kth task M k Insert into task list P j After reaching the nth position in the list, the drone executes the task list P j Rewards It is expressed as: Among them, R0 is task M k The basic value of R is the task M k The static value of j is the discount factor, Indicates execution of task M k Time, ET k Represents task M k The earliest start time, Indicates that in the task list p j Insert task M at position n k marginal cost at time In the task list j Insert task M at position n k Marginal cost at It is expressed as: in, Indicates execution in task list p i Insert task M at the nth position in k The distance required for all tasks in the task list after that; D(p i ) indicates the execution of task list p i The range required for all tasks in j Indicates fuel consumption; 403. Determine whether the marginal benefit of inserting the current candidate task in step 402 is greater than the benefit of the winner of the task in the local winner list; if not, return to 401 and calculate the next candidate task; 404. If it is greater than, then add the task to the task timing package of the current UAV, and determine whether the task timing package of the current UAV exceeds the upper limit. If it exceeds, then share the task information with other UAVs, otherwise return to step 401 to calculate the next candidate task; 405. When the number of tasks in the task sequence package of the current drone reaches the upper limit, the current drone shares the task allocation information with other drones and updates the local task information. If the local task information of a drone is inconsistent with the shared information, the bundle is updated and 401 is executed; if they are consistent, the task allocation result is output.

2. A method for dynamic allocation of multi-tasks of drone clusters according to claim 1, characterized in that: The local mission information of the drone includes mission bundle, mission timing package, winner list, winner payoff set, timestamp and execution time set, among which: The task bundle is the set of tasks that the current drone competes for. The order of tasks in the bundle is arranged according to the order in which the tasks are added. The size of the bundle is the maximum number of tasks that the drone can perform; The mission sequence package is a collection of UAV execution tasks, and the order in the package is arranged according to the priority of task execution; The winner list is a list of drones with the highest profit in each task. If the element in this list is 0, it means that no drone has won the bid for this task yet; The winner's profit set is the highest profit of each task. If the element in this list is -1, it means that no drone has won the bid for this task. The timestamp is the moment when two drones receive information from each other; The execution time set is the execution time when the UAV arrives at the mission area to perform the mission according to the mission timing set.

3. The method for dynamic allocation of multi-tasks of drone clusters according to claim 2 is characterized in that: Update of the drone winners list, i.e. when drone U j Get Task M k When task M k Will be inserted into the drone U j The task position that currently increases the marginal benefit the most in the path. If the benefit of this task position is greater than the benefit of the winner of this task, then UAV U j Add the task to your task bundle, update the bundle and add the task M k The information is shared with other drones.

4. A method for dynamic allocation of multi-tasks of drone clusters according to claim 3, characterized in that: If the drone U a Received UAV b Shared about task M c When the information of UAV a Perform updates on the following information: When ac <y bc When the UAV a Update local information and replace y bc Assign to y ac , put z bc Assign to z ac ; When ac >y bc When the UAV a Do not update local information, y ac 、z ac remain unchanged; When ac =y bc When the timestamps in the tasks conflict or are inconsistent, y ac Set to -1, z ac Set to 0; Among them, y bc Indicates drone U b The winner set of task M c The maximum benefit value of y ac Indicates drone U a The winner set of task M c The maximum benefit value of z bc Indicates drone U b Perform task M on the winner set c The most profitable drone; ac Indicates drone U a Perform task M on the winner set c The most profitable drone.