Unmanned aerial vehicle cluster task planning method and system

Abstract

The invention provides an unmanned aerial vehicle cluster task planning method and system, and the method comprises the steps: randomly creating an initial task planning scheme for an unmanned aerial vehicle cluster, and carrying out the circulation of the initial task planning scheme for multiple times in a single iteration mode, thereby obtaining an unmanned aerial vehicle cluster task planning scheme, the single iteration comprises the following steps of: updating the initial task planning scheme or the single iteration ending task planning scheme for the first time based on prior knowledge; screening a local scheme from all the schemes updated for the first time; guiding three times of updating based on experience; and selecting a single iteration ending task planning scheme or an unmanned aerial vehicle cluster task planning scheme from all schemes updated for three times. The system comprises a creation unit, a primary iteration unit, a secondary iteration unit and an Nth iteration unit. According to the method, the problems of large planning scale and complex decision in unmanned aerial vehicle cluster task planning are solved, a task planning scheme is guided and updated by utilizing priori knowledge and experience based on the characteristic of mutual relation among multiple sub-problems, and the task execution benefit of the unmanned aerial vehicle is improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm mission planning, and more particularly to a method and system for UAV swarm mission planning. Background Technology

[0002] In the field of UAV swarm mission planning, with the modularization of airborne resources and the development of wide-speed-domain UAV platforms, the pre-mission planning problem for UAV swarms now involves not only major planning sub-problems such as mission allocation, timing scheduling, and trajectory planning, but also new types of sub-problems such as airborne resource allocation and flight speed-domain switching. Furthermore, these sub-problems are interconnected and influence each other. In addition, the mission planning process involves numerous constraints such as time windows, aircraft dynamics, environmental limitations, resource allocation, and priorities, resulting in a large-scale and highly complex task planning process with coupled sub-problems. Therefore, there is an urgent need to design efficient optimization methods for efficient optimization of mission planning in large-scale and complex decision spaces.

[0003] The solutions for task planning in heterogeneous UAV swarms can be broadly categorized into two types: exact solutions and heuristic solutions. While exact solutions can yield the optimal solution to the optimization problem, their solution time increases dramatically with the size of multiple subproblems, making it difficult to improve optimization efficiency. Heuristic solutions, although capable of providing approximate optimal solutions within a finite timeframe, cannot theoretically estimate the deviation between the generated solution and the optimal solution. Currently, numerous studies have validated the effectiveness of heuristic solutions in optimization problems.

[0004] Among them, the heuristic solution method, when providing an approximate optimal solution, will increase the difficulty of optimization if the time window of the task execution is considered in the heterogeneous UAV swarm task planning. The heuristic solution method cannot meet the performance requirements of heterogeneous UAV swarm task planning, especially in the case of large and medium-scale task planning. Even if the planning result generated by the heuristic solution method is greatly improved compared with the initial solution, the coupling effect of multiple subproblems of heterogeneous UAV swarms will cause the heuristic solution method to optimize to a local solution too early, resulting in a certain deviation between the local solution and the optimal solution. There is still a lot of room for improvement in the optimization direction. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for planning unmanned aerial vehicle (UAV) swarm missions, which solves the problems of large planning scale and complex decision-making in UAV swarm mission planning. Based on the characteristics of the interrelationship between multiple sub-problems, it uses prior knowledge and experience to guide the updating of mission planning schemes, thereby improving the benefits of UAV mission execution.

[0006] The technical solution for achieving the objective of this invention is as follows:

[0007] On one hand, the present invention provides a method for planning unmanned aerial vehicle (UAV) swarm missions, including:

[0008] Step 1: Randomly create an initial mission planning scheme for the drone swarm;

[0009] Step 2: The initial task planning scheme is iterated multiple times using a single iteration method to obtain a UAV swarm task planning scheme. The single iteration includes: first, updating the first planning scheme based on prior knowledge; second, selecting local schemes from all the schemes updated in the first iteration for a second update to obtain a second planning scheme; then, updating the second planning scheme a third time based on experience to obtain a third planning scheme; and finally, selecting a fourth planning scheme from all the schemes in the third planning scheme. The first planning scheme is either the initial task planning scheme or the single-iteration-ending task planning scheme obtained in the previous iteration; the fourth planning scheme is either the single-iteration-ending task planning scheme or the UAV swarm task planning scheme.

[0010] Based on one aspect, in one embodiment of the present invention, step one, randomly creating an initial task planning scheme for the drone swarm, includes:

[0011] Obtain parameter information of the drone swarm; wherein, the parameter information includes drone information, mission information, and onboard resource information;

[0012] The task planning information for each drone in the drone cluster is randomly assigned based on the parameter information; the task planning information includes the execution task information, onboard resource information, task timing priority, and drone flight speed for each drone.

[0013] The initial task planning scheme is obtained by summarizing the task planning information.

[0014] Based on one aspect, in one embodiment of the present invention, step two, which involves iterating the initial task planning scheme multiple times using a single-iteration method to obtain a UAV swarm task planning scheme, includes:

[0015] The initial task planning scheme is iterated once to obtain all task planning schemes updated in one iteration, and the task planning scheme with the highest target value is selected as the input for the next single iteration;

[0016] Repeat the single iteration and calculate the probability of each task planning scheme being accepted, and obtain the UAV swarm task planning scheme based on the probability.

[0017] Based on one aspect, in one embodiment of the present invention, the single iteration specifically refers to:

[0018] First, the probability of the first planning scheme is calculated based on the prior knowledge of the drone swarm. Based on the probability, the task planning scheme is replanned to obtain an updated task planning scheme. The prior knowledge includes the assigned tasks, the assigned resources, and the flight speed.

[0019] Secondly, all schemes of the one-time update task planning scheme are obtained and grouped. The target value of all schemes is calculated according to the group. The group with the highest target value is selected as a local scheme to form the second planning scheme.

[0020] Then, based on experience, the probability of the second planning scheme is calculated, and the task planning scheme is re-planned based on the probability to obtain the third planning scheme;

[0021] Finally, all schemes of the third planning scheme are grouped into groups of two, and their target values ​​are calculated for each group. Based on the target values, a group of schemes is selected from the groups, and the optimal and second-best schemes of the group are determined. The optimal or second-best scheme is then selected as the fourth planning scheme.

[0022] Based on one aspect, in one embodiment of the present invention, calculating the probability of the first planning scheme based on prior knowledge of the UAV swarm includes: calculating an optimization term based on prior knowledge of the UAV swarm, and calculating a probability term corresponding to the optimization term; wherein,

[0023] The optimization terms include the mission benefit range ratio, mission benefit segment length ratio, average mission benefit weight ratio, and time window penalty; the probability terms include mission redistribution probability, mission timing scheduling probability, airborne resource redistribution probability, and UAV flight speed switching probability.

[0024] Based on one aspect, in one embodiment of the present invention, the expression for the mission benefit-range ratio is:

[0025] Q h ( T h , S h , G ( h )) = R ( T h , S h ) L [ G ( h )]

[0026] In the formula, Indicates drone Mission benefit-to-range ratio Indicates drone The set of assigned tasks Indicates drone The set of loading resources, Indicates drone Flight range, Indicates drone Total task revenue, L [ G ( h )] Indicates drone The total flight distance;

[0027] The expression for the mission benefit segment length ratio is:

[0028]

[0029] In the formula, Indicates task Mission benefit segment length ratio Indicates drone The execution of the first One task, express The prerequisite tasks. Indicates drone The set of loading resources, Indicates the number of drones, Indicates drone Carry resources Execute the task The rewards obtained from the task Indicates drone Carry resources Execute the task The rewards obtained from the task;

[0030] The expression for the average task revenue weight ratio is:

[0031]

[0032] In the formula, Representing resources Average task reward weight ratio Indicates drone , Indicates drone The set of assigned tasks Indicates drone The loaded first One resource, Indicates the first The resource number, Indicates drone Carry resources Mission rewards obtained along the flight path Representing resources The weight;

[0033] The expression for the time window penalty is:

[0034]

[0035] In the formula, Indicates drone Execute the task Time window penalty Indicates drone To perform the task, Indicates the execution of a task The earliest expected start time, Indicates the execution of a task The actual time of execution Indicates task The latest expected end time;

[0036] The expression for the task redistribution probability is:

[0037]

[0038] In the formula, This represents the probability of task redistribution for drones. This indicates the number of drones in a drone swarm. Indicates drone Mission benefit-to-range ratio Indicates drone The set of assigned tasks Indicates drone The set of loading resources;

[0039] The expression for the task timing scheduling probability is:

[0040]

[0041] In the formula, This represents the timing probability of the drone performing the task. Indicates task Mission benefit segment length ratio Indicates drone The execution of the first One task, Represents any task , Indicates the number of tasks. Indicates the first The task number of each task. Indicates the first The task number of each task;

[0042] The expression for the airborne resource reallocation probability is:

[0043]

[0044] In the formula, This indicates that the drone carries the resource. The probability of redistribution received, Indicates the quantity of resources. Representing resources Average task reward weight ratio Indicates drone The loaded first One resource, Indicates the first The resource number, Indicates drone The resources carried;

[0045] The expression for the drone's flight speed switching probability is:

[0046]

[0047] In the formula, This indicates the probability of switching flight speeds for the drone. Indicates drone Time window penalties for all tasks This represents all the tasks of a single drone; This indicates the number of drones in a drone swarm. Indicates drone Time window penalties for all tasks Indicates drone .

[0048] Based on one aspect, in one embodiment of the present invention, the formula for calculating the target value is:

[0049]

[0050] In the formula, Indicates the target value. This represents the weighting coefficient of the total task revenue. This represents the total revenue generated from all missions performed by the drones. This represents the total flight time of all drones. This represents the weighting coefficient for total flight time. This represents the time window penalty coefficient. This represents the total time window penalty for all drone missions; where:

[0051] Total revenue from all drone missions The expression is:

[0052]

[0053] In the formula, This indicates the number of drones in a drone swarm. Indicates the number of tasks. Indicates the quantity of resources. Indicates that the drone carries resources Execute the task The rewards obtained from the task Represents 0-1 decision variables;

[0054] Total flight time of all drones The expression is:

[0055]

[0056] In the formula, This indicates the number of drones in a drone swarm. Indicates the number of tasks. This refers to a drone. Indicates another drone, Indicates the first One task, Indicates drone Execute the task and tasks The actual time of execution 0-1 decision variables;

[0057] Total time window penalties for all drone missions The expression is:

[0058]

[0059] In the formula, Indicates the number of tasks. Indicates another drone, Indicates task The earliest expected start time, Indicates the execution of a task The actual time of execution Indicates task The latest expected end time.

[0060] Based on one aspect, in one embodiment of the present invention, the probability of the second planning scheme is obtained through a UAV experience accumulation equation and experience-guided calculation, wherein the UAV experience accumulation equation includes accumulated experience in task allocation and accumulated experience in resource allocation; wherein:

[0061] The cumulative experience expression for task allocation is:

[0062]

[0063]

[0064]

[0065] In the formula, This represents the initial value for task allocation. Indicates the number of drones, Indicates drone Accumulated experience in task allocation Indicates the current iteration number. Indicates the learning rate. This indicates the number of current optimal solutions. Represents 0-1 decision variables;

[0066] The cumulative empirical expression for resource allocation is:

[0067]

[0068]

[0069]

[0070] In the formula, This represents the initial value for resource allocation. Indicates the quantity of resources. This represents accumulated experience in resource allocation. Indicates the current iteration number. Indicates the learning rate. This indicates the number of current optimal solutions. This represents a 0-1 decision variable.

[0071] On the other hand, embodiments of the present invention provide a drone swarm mission planning system. Based on the above-described drone swarm mission planning method, the drone swarm mission planning system includes:

[0072] A creation unit randomly creates an initial task planning scheme for the drone cluster;

[0073] A first iteration unit, wherein the first iteration unit iteratively updates the first planning scheme to obtain a first iteration updated task planning scheme;

[0074] A secondary iteration unit, wherein the secondary iteration unit iteratively updates the task planning scheme of the first iteration update to obtain a secondary iteration update task planning scheme;

[0075] An N-fold iteration unit is used to iteratively update the task planning scheme through N-1 iterations to obtain the UAV swarm task planning scheme; N≥3;

[0076] The iteration networks of the first iteration unit, the second iteration unit, and the Nth iteration unit are the same.

[0077] In another aspect, in one embodiment of the present invention, the iterative network includes:

[0078] The prior knowledge update module updates the first planning scheme to obtain an updated task planning scheme.

[0079] The first filtering and updating module calculates and filters the first update task planning scheme to obtain the second update task planning scheme.

[0080] An experience-guided update module updates the secondary update task planning scheme to obtain a tertiary update task planning scheme.

[0081] The second filtering and updating module calculates and filters the three update task planning schemes to obtain an X-times iterative update task planning scheme, where X ≥ 1.

[0082] Compared with the prior art, the beneficial effects of the present invention are:

[0083] 1. This invention enables UAV planning to have a rich diversity of solutions by randomly creating initial task planning schemes for UAVs, thus solving the problem of difficult task planning involving multiple sub-problems such as task allocation, timing scheduling, resource allocation, and speed selection in UAV swarms.

[0084] 2. This invention extracts mission planning features based on the relationship between UAV tasks, resources, and flight speed, reflecting the efficiency of UAV mission execution. It guides the direction of UAV swarm mission planning by utilizing time-series scheduling probability, flight speed switching probability, and task and resource reallocation probability, and locally updates the UAV swarm mission planning scheme. This invention uses experience to guide the reallocation of UAV tasks and resources, improves the UAV swarm mission planning effect by using empirical solutions of UAV mission planning schemes with high mission benefits, and selects the globally optimal UAV swarm mission planning scheme based on the annealing mechanism.

[0085] 3. Based on the interrelationship among multiple sub-problems in UAV swarm mission planning, this invention utilizes prior knowledge and experience to guide the updating of mission planning schemes, thereby improving the benefits of UAV mission execution and reducing flight time and time window penalties. This invention effectively solves the problems of large mission planning scale and complex decision-making caused by multiple sub-problems. Attached Figure Description

[0086] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0087] Figure 1 A flowchart of a drone swarm mission planning method provided by the present invention;

[0088] Figure 2 A process illustration of a single iteration provided for this invention;

[0089] Figure 3 Provided by the present invention Figure 1 Flowchart for step one;

[0090] Figure 4 Provided by the present invention Figure 1 Flowchart for step two;

[0091] Figure 5 A test case table representing large, medium, and small-scale problems is provided for this invention;

[0092] Figure 6 The heterogeneous UAV parameter table provided for this invention;

[0093] Figure 7 This is a comparison chart of the average optimization performance of the small-scale algorithm provided by this invention;

[0094] Figure 8 This is a comparison chart of the average optimization performance of the medium-scale algorithm provided by this invention;

[0095] Figure 9 A comparison chart of the average optimization effect of the large-scale algorithm provided by this invention;

[0096] Figure 10 A comparison chart of the distribution of target values ​​for small-scale algorithms provided by this invention;

[0097] Figure 11 This is a comparison chart of the target value distribution of the medium-scale algorithm provided by the present invention;

[0098] Figure 12 A comparison chart of the distribution of target values ​​for large-scale algorithms provided by this invention;

[0099] Figure 13 This is a flight trajectory diagram of the initial UAV swarm task constructed in a small-scale scenario according to the present invention;

[0100] Figure 14 This is a flight trajectory diagram of a drone swarm task execution after the drone task planning method based on prior knowledge and experience guided by the present invention in a small-scale scenario.

[0101] Figure 15 A comparison chart of the average optimal target value distribution of the method provided by this invention and other methods in 30 experiments in large and medium-scale scenarios. Detailed Implementation

[0102] The present invention will now be described in detail with reference to the embodiments shown in the accompanying drawings. However, it should be noted that these embodiments are not intended to limit the present invention. Equivalent changes or substitutions in function, method, or structure made by those skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0103] Please see Figure 1 This invention provides a method for planning unmanned aerial vehicle (UAV) swarm missions, comprising the following steps:

[0104] Step 1: Randomly create an initial mission planning scheme for the drone swarm.

[0105] Please see Figure 3 The above steps in this embodiment of the invention, which randomly create an initial task planning scheme for a drone swarm, specifically include:

[0106] Step 11: Obtain the parameter information of the drone swarm; the parameter information includes drone information, mission information, and onboard resource information; specifically, drone information includes the total number of drones, maximum flight time, maximum payload, selectable flight speed, and descent time per 100 meters for each flight speed; mission information includes the number of missions to be executed, mission location information, mission approach heading angle, mission area information, and mission time window information; onboard resource information includes the number of resources, resource weight, resource type, and resource mission reward information.

[0107] Step 12: Randomly allocate task planning information for each UAV in the UAV cluster based on parameter information. The task planning information includes the execution task information of each UAV, the onboard resource information of each UAV, the task time priority of the randomly arranged execution task information, and the execution flight speed corresponding to each execution task in the randomly allocated execution task information.

[0108] Step 13: Summarize the task planning information to obtain the initial task planning scheme.

[0109] In practical applications, step one specifically involves:

[0110] S11. Initialize the number of drones in the cluster. Cluster size Maximum number of iterations, number of empirical solutions, learning rate Total task revenue weighting coefficient Total flight time weighting coefficient and time window penalty coefficient .

[0111] S12. Obtain drone information, mission information, and airborne resource information of the drone swarm. Drone information includes the total number of drones, maximum flight time, maximum payload, selectable flight speed, and descent time per 100 meters for each flight speed. Mission information includes the number of missions to be executed, mission location information, mission approach heading angle, mission area information, and mission time window information. Airborne resource information includes the number of resources, resource weight, resource type, and resource mission benefit information.

[0112] S13. Based on the drone information and task information of the drone cluster, randomly assign tasks to each drone. In this step, a task list for each drone is generated, for example, the drone task list is represented as:

[0113] [Drone 1: Mission 1, Mission 3, Mission 6]

[0114] [Drone 2: Mission 9, Mission 7, Mission 2]

[0115] [Drone 3: Mission 5, Mission 4, Mission 8].

[0116] S14. Based on the drone information and onboard resource information of the drone cluster, randomly configure the onboard resources of each drone. In this step, a drone resource list is generated, for example, the drone resource list is represented as:

[0117] [Drone 1: Resource 2, Resource 3, Resource 5]

[0118] [Drone 2: Resources 1, 4, 7, 10]

[0119] [Drone 3: Resource 6, Resource 8, Resource 9].

[0120] S15. Based on the task information of the drone swarm, randomly arrange the time priority list of all tasks. The time priority list of all tasks formed in this step is as follows:

[0121] [Task 3, Task 9, Task 1, Task 2, Task 7, Task 5, ...].

[0122] S16. Based on the UAV information and mission information, randomly assign a flight speed for each mission during UAV execution. In this step, a list of flight speeds for all missions is generated, for example:

[0123] [Flight speed 1, flight speed 3, flight speed 2, flight speed 1, ...].

[0124] S17. Based on the current iteration count, if the current iteration count is 1, then the randomly initialized task planning result will be used as the starting point for optimization; if the current iteration count is not 1 and the maximum iteration count has not been reached, the generated... The solution is used as the starting point for optimization. A solution refers to the task planning scheme of a drone swarm (a team of drones). The optimization starting point (i.e., the search starting point) in this embodiment of the invention is... Each solution represents a solution set. Other optimization starting points correspond to a solution that includes a task list, task time sequence list, resource list, and flight speed list for all drones in a drone swarm (a team of drones), i.e., a task planning scheme.

[0125] The application example of step one of the embodiments of the present invention is as follows: creation of the initial task planning scheme (i.e., construction of the initial solution set), for example: 3 drones, 10 resources, 15 tasks, and 4 selectable flight speeds. (1) Based on the number of drones, their numbers are 0-2, and a random number is generated between 0 and 2. Next, generate a task list for all drones. (2) Create a list of 0-14, and randomly shuffle the list to obtain a task sequence list. (3) Based on the number of drones, their numbers are 0-2, and a random number is generated between 0 and 2 for repetition. Next, generate a list of loaded resources. (4) Based on the number of flight speeds, their numbers are 0-3. Use random numbers to generate numbers between 0 and 3 repeatedly. Next, generate a list of flight speeds for executing the mission. The list generated in steps (1)-(4) above is a mission planning scheme for a fleet of UAVs, i.e., a solution. Repeat steps (1)-(4). Secondary Each different task planning scheme is considered as a solution set. The solution set is regarded as the starting point of the search, and each solution corresponds to one task planning scheme.

[0126] Step 2: Iterate the initial task planning scheme multiple times to obtain the UAV swarm task planning scheme.

[0127] Please see Figure 4 In this embodiment of the invention, step two involves iterating the initial task planning scheme multiple times to obtain the UAV swarm task planning scheme, including:

[0128] Step 21: Update the initial task planning scheme in one iteration to obtain the updated task planning scheme. Select the task planning scheme with the highest target value from all task planning schemes as the input for the second single iteration.

[0129] Step 22: Repeat the single iteration and calculate the probability that the task planning scheme is accepted. Based on the probability, obtain the UAV swarm task planning scheme. Specifically: update the task planning scheme of the first iteration in the second iteration to obtain the second iteration updated task planning scheme; after performing multiple iterations of the second iteration, the task planning scheme with the highest target value is output as the UAV swarm task planning scheme.

[0130] The embodiments of the present invention provide Each drone mission planning scheme is obtained after replanning the mission, timing, resources, and flight speed. Generate a new candidate solution. Each optimized solution serves as a new starting point for optimization (an optimized solution is an optimized task planning scheme; a drone swarm is a team of drones, and a drone task planning scheme is a solution). Based on the new starting point for the drone swarm optimization, the objective value of each drone task planning scheme is calculated. For any set of optimized solutions, from... Select the solution with the highest objective value from the candidate solutions and update the generated solution. Each solution serves as a new starting point for optimization.

[0131] The multiple iterations in step two of this embodiment are composed of multiple iterations of a single iteration. Please refer to [link / reference]. Figure 2 The single iteration process in this embodiment of the invention includes:

[0132] Step a: Update the task planning scheme based on prior knowledge to obtain an updated task planning scheme, where the task planning scheme is either the initial task planning scheme or a single-iteration updated task planning scheme.

[0133] Specifically, step a includes: calculating optimization terms based on prior knowledge of the UAV swarm, taking the initial task planning scheme or the single-iteration updated task planning scheme as the starting point for optimization, and replanning the task planning scheme through the probability terms of the corresponding optimization terms to obtain an updated task planning scheme; wherein, the prior knowledge includes the assigned tasks, the assigned resources, and the flight speed.

[0134] The optimization terms include the mission benefit range ratio, mission benefit segment length ratio, average mission benefit weight ratio, and time window penalty. The probabilistic terms include mission reassignment probability, mission timing scheduling probability, onboard resource reassignment probability, and UAV flight speed switching probability.

[0135] Specifically: Based on the optimized starting point of the UAV swarm, the ratio of the mission reward range length for each task is calculated to obtain the temporal scheduling probability of each task (a small mission reward range ratio for a UAV indicates that its flight range is too long or the mission reward is too low, and the UAV is more likely to be subject to mission reassignment; the probability of mission reassignment for UAVs is higher). (Higher). Using time-series scheduling probability, a task is selected. Under the condition of not violating the maximum flight time constraint of the UAV, a task is randomly selected, and the priority order of the two selected tasks is swapped.

[0136] Specifically: Based on the optimization starting point of the drone swarm, calculate the average task revenue weight ratio of each resource to obtain the resource redistribution probability. Using the resource redistribution probability, select one resource, and without violating the drone payload constraints, randomly select another resource and swap the drone numbers of the two selected resources.

[0137] Specifically, based on the optimization starting point of the drone swarm, the time window penalty for each drone is calculated to obtain the flight speed switching probability for each drone. Using the flight speed switching probability, a drone is selected. Under the condition of not violating the maximum flight time constraint of the drone, the task to be performed by the selected drone is randomly selected. If the time window penalty of the task is earlier than the expected time window, a smaller flight speed is switched; otherwise, a larger flight speed is switched.

[0138] Specifically, based on the optimization starting point of the drone swarm, the drones with the highest target values ​​are selected. One UAV mission planning scheme is used as an empirical solution (the empirical solution is the updated version). The solution with the highest objective value is among the top solutions. (Solution), update the accumulated experience of task allocation and resource configuration, calculate the probability of a drone being selected for task and resource reallocation, and replan the drone's tasks and resources. Based on the optimization starting point of the drone cluster, calculate the accumulated experience of task allocation to obtain the probability of a drone being selected for task reallocation. Randomly select a task, and use the probability of a drone being selected for task reallocation to select a drone, and assign the selected task to the selected drone. Based on the optimization starting point of the drone cluster, calculate the accumulated experience of resource configuration to obtain the probability of a drone being selected for resource reallocation. Randomly select a resource, and use the probability of a drone being selected for resource reallocation to select a drone, randomly select a resource loaded by the selected drone, and swap the drone numbers of the two selected resources. For Each UAV mission planning scheme generates a new candidate solution after the task and resources are reallocated. The optimized solution serves as the starting point for a new optimization.

[0139] Specifically, based on the optimization starting point of the drone swarm, the target values ​​of all drone mission planning schemes are calculated. For any set of optimized solutions, the probability of the optimal solution being accepted from two candidate solutions is calculated, and this probability is used to select the optimal solution. The first solution serves as the starting point for the next iteration. If the maximum number of iterations has not been reached, the next iteration begins, and the next cycle of updates is executed. Otherwise, the iteration stops. The optimal UAV swarm task planning scheme is obtained based on the probability that the optimal UAV swarm task planning scheme is accepted. The probability that the above optimal solution is accepted in this embodiment of the invention is expressed as:

[0140]

[0141] In the formula, This represents the objective value of the optimal solution. This represents the objective value of the suboptimal solution. If the maximum number of iterations is reached, The target value is the globally optimal UAV mission planning scheme. The target value is the suboptimal global UAV mission planning scheme. This is the temperature coefficient, and its initial value is set to... In each iteration, The rate decreases if it passes through If, in a round of iterations, the current optimal solution has not improved, then... Reset to initial value .

[0142] More specifically: the expression for the mission benefit-range ratio in this embodiment of the invention is:

[0143] Q h ( T h , S h , G ( h )) = R ( T h , S h ) L [ G ( h )]

[0144]

[0145] L [ G ( h )] = ∑ i = 1 | T h | − 1 ‖ G ( h , σ i + 1 ) − G ( h , σ i ) ‖

[0146]

[0147] In the formula, Indicates drone Mission benefit-to-range ratio Indicates drone The set of assigned tasks Indicates drone The set of loading resources, Indicates drone Flight range, Indicates drone Total task revenue, Indicates drone The number of tasks assigned Indicates drone The number of resources loaded, For drones Carry resources Execute the task The rewards obtained from the task; L [ G ( h )] Indicates drone Total flight distance, Indicates drone Spatial location, Indicates drone Spatial location, Indicates drone Flight route, This indicates the flight path of the drone performing the second-to-last task. This indicates the flight path of the drone as it performs the last task.

[0148] The expression for the mission benefit segment length ratio is:

[0149]

[0150] In the formula, Indicates task Mission benefit segment length ratio Indicates drone The execution of the first One task, express The prerequisite tasks. Indicates drone The set of loading resources, Indicates the number of drones, Indicates drone Carry resources Execute the task The rewards obtained from the task Indicates drone Carry resources Execute the task The rewards obtained from the task.

[0151] The expression for the average task reward weight ratio is:

[0152]

[0153] In the formula, Representing resources Average task reward weight ratio Indicates drone , Indicates drone The set of assigned tasks Indicates drone The loaded first One resource, Indicates the first The resource number, Indicates drone Carry resources Mission rewards obtained along the flight path Representing resources The weight;

[0154] The expression for the time window penalty is:

[0155]

[0156] In the formula, Indicates drone Execute the task Time window penalty Indicates drone To perform the task, Indicates the execution of a task The earliest expected start time, Indicates the execution of a task The actual time of execution Indicates the execution of a task The latest expected end time.

[0157] More specifically:

[0158] The expression for the probability of task redistribution is:

[0159]

[0160] In the formula, This represents the probability of task redistribution for drones. This indicates the number of drones in a drone swarm. Indicates drone Mission benefit-to-range ratio Indicates drone The set of assigned tasks Indicates drone The set of loading resources;

[0161] The expression for the task timing scheduling probability is:

[0162]

[0163] In the formula, This indicates the timing probability of the drone performing the task. Indicates task Mission benefit segment length ratio Indicates drone The execution of the first One task, Represents any task , Indicates the number of tasks. Indicates the first The task number of each task. Indicates the first The task number of the task; if the task's reward segment length is too short, it means that the UAV's flight distance for performing the task is too long, and the probability of the task being subject to time-based scheduling increases.

[0164] The expression for the probability of airborne resource reallocation is:

[0165]

[0166] In the formula, This indicates that the drone carries the resource. The probability of redistribution received, Indicates the quantity of resources. Representing resources Average task reward weight ratio Indicates drone The loaded first One resource, Indicates the first The resource number, Indicates drone The resources carried;

[0167] The expression for the drone's flight speed switching probability is:

[0168]

[0169] In the formula, This indicates the probability of switching flight speeds for the drone. Indicates drone Time window penalties for all tasks This represents all the tasks of a single drone; This indicates the number of drones in a drone swarm. Indicates drone Time window penalties for all tasks Indicates drone The larger the time window penalty for a drone, the greater the probability that the drone's flight speed violates the time window constraint for the mission, and the higher the probability of the drone switching flight speeds.

[0170] The expression for the maximum flight time constraint of the UAV is:

[0171]

[0172] in, Indicates the number of tasks. Indicates task and tasks Collision-free distance For 0-1 decision variables, Indicates drone flight speed , Indicates drone Maximum flight time, Indicates the quantity of resources. Indicates drone Maximum flight time, Representing two-dimensional decision variables, Indicates drone Load resources ,otherwise , Indicates drone At flight speed The reduction in flight time per 100 meters of flight. Indicates drone , Indicates a collection of drones. Indicates the target value , This represents the target value.

[0173] The expression for the drone's payload constraint is:

[0174]

[0175] in, Representing two-dimensional decision variables, Indicates the quantity of resources. Indicates drone Maximum load capacity Indicates drone Maximum load capacity Indicates drone , This refers to a collection of drones.

[0176] Step b: Calculate the first target value of all schemes in the first update task planning scheme, and select local schemes from all schemes based on the first target value to form the second update task planning scheme.

[0177] Specifically, step b includes: obtaining all schemes of the one-time update task planning scheme, grouping all schemes into groups of N schemes, calculating the target value of all schemes, selecting the scheme with the highest target value as a local scheme to form the two-time update task planning scheme, and discarding the other groups of schemes.

[0178] It should be noted that the target values ​​for all schemes in the embodiments of the present invention are calculated using the following formula:

[0179]

[0180] In the formula, Indicates the target value. This represents the weighting coefficient of the total task revenue. This represents the total revenue generated from all missions performed by the drones. This represents the total flight time of all drones. This represents the weighting coefficient for total flight time. This represents the time window penalty coefficient. This represents the total time window penalty for all drones performing tasks; when the drone Loading load Execute the task hour, ,otherwise .in:

[0181] Total revenue from all drone missions The expression is:

[0182]

[0183] In the formula, This indicates the number of drones in a drone swarm. Indicates the number of tasks. Indicates the quantity of resources. Indicates that the drone carries resources Execute the task The rewards obtained from the task This represents a 0-1 decision variable.

[0184] Total flight time of all drones The expression is:

[0185]

[0186] In the formula, This indicates the number of drones in a drone swarm. Indicates the number of tasks. This refers to a drone. Indicates another drone, Indicates the first One task, Indicates drone Execute the task and tasks The actual time of execution For 0-1 decision variables, if the drone Execute tasks in sequence and tasks ,but ,otherwise . For the task The actual time of execution; Indicates drone Execute the task The actual time of execution Indicates drone Execute the task The actual time when it is executed.

[0187] Total time window penalties for all drone missions The expression is:

[0188]

[0189] In the formula, Indicates the number of tasks. Indicates another drone, Indicates task The earliest expected start time, Indicates the execution of a task The actual time of execution Indicates task The expected latest end time, if the task Not in the time window If the task is completed within the specified time window, its rewards will be subject to a limited time window penalty.

[0190] Step c: Based on experience, guide the update of the secondary update task planning scheme to obtain the tertiary update task planning scheme.

[0191] Specifically, step c includes: updating the UAV experience accumulation equation based on the second-update task planning scheme; calculating the probability of each UAV being selected for task reassignment and the probability of being selected for resource reassignment based on the UAV experience accumulation equation; and obtaining the third-update task planning scheme based on the probability of each UAV being selected for task reassignment and the probability of being selected for resource reassignment.

[0192] The UAV experience accumulation equation described in this embodiment of the invention includes accumulated experience in task allocation and accumulated experience in resource allocation; wherein:

[0193] The cumulative experience expression for task assignment is:

[0194]

[0195]

[0196]

[0197] In the formula, This represents the initial value for task allocation. Indicates the number of drones, Indicates drone Accumulated experience in task allocation Indicates the current iteration number. Indicates the learning rate. This indicates the number of current optimal solutions. Let 0-1 be the decision variable, if in the th... In the current task assignment results, the task Assigned to drones ,but ,otherwise Based on accumulated experience in task allocation, for tasks... Calculate the reassignment of tasks to each drone The probability of.

[0198] The cumulative empirical expression for resource allocation is:

[0199]

[0200]

[0201]

[0202] In the formula, This represents the initial value for resource allocation. Indicates the quantity of resources. This represents accumulated experience in resource allocation. Indicates the current iteration number. Indicates the learning rate. This indicates the number of current optimal solutions. Let 0-1 be the decision variable, if in the th... In the current resource allocation results, the task Assigned to drones ,but ,otherwise . Based on accumulated experience in resource allocation, for resources... Calculate the probability that each drone will be reconfigured.

[0203] Step d: Calculate the target value of all schemes in the three-time update task planning schemes, and select the single-iteration update task planning scheme based on the second target value. Among them, the single-iteration update task planning scheme that ends in the last iteration is the UAV swarm task planning scheme.

[0204] Specifically, step d includes: obtaining all schemes for the three updated task planning schemes; grouping all schemes into pairs; calculating the target value for all schemes; identifying the scheme with the higher target value in a group as the optimal scheme and the scheme with the lower target value as the suboptimal scheme; calculating the probability of the optimal scheme being accepted; sampling the value of whether the optimal solution is accepted from a binary distribution based on this acceptance probability; if the value of the optimal solution is 1, retaining the suboptimal scheme as the task planning scheme for the next iteration; otherwise, retaining the optimal scheme as the UAV swarm task planning scheme. It should be noted that the process of calculating the target value for all schemes in this step is consistent with the target value calculation in step b.

[0205] The application example of step two in this embodiment of the invention is as follows: The update of the task planning scheme is divided into four steps: (1) updating the task planning scheme based on prior knowledge; (2) local update; (3) updating the task planning scheme based on experience; (4) global update. Among them:

[0206] The task planning scheme is updated based on prior knowledge. Specifically, it is updated according to the search starting point. Prior knowledge refers to the task benefit range ratio, task benefit segment length ratio, average task benefit weight ratio, and time window penalty of all UAVs, calculated based on the task execution list, task time sequence list, resource loading list, and task execution flight speed list of all UAVs. Based on the prior knowledge, the probability of each UAV being selected for task reassignment, the probability of each task being scheduled in a time sequence, the probability of each resource being reassigned, and the probability of each UAV being selected to switch flight speed are calculated sequentially. For example, the probability of each UAV being selected for task reassignment: [0.1, 0.4, 0.5] means that the probability of UAV 1 being selected for task reassignment is 0.1, the probability of UAV 2 being selected for task reassignment is 0.4, and the probability of UAV 3 being selected for task reassignment is 0.5. The probability of each resource being reallocated: [0.4, 0.2, 0.4] means that the probability of resource 1 being selected for reallocation is 0.4, the probability of resource 2 being selected for reallocation is 0.2, and the probability of resource 3 being selected for reallocation is 0.4.

[0207] Specifically, (1) the task reassignment steps are as follows: based on the probability that each UAV is selected for task reassignment, sample one UAV. Randomly select the drone One task from the task list to be executed, and then a drone to be randomly selected. The selected task is added to the list of randomly selected drone tasks. (2) The timing scheduling steps are as follows: Based on the probability of timing scheduling for each task, sample a task, then randomly select a task, and exchange the priorities of the two tasks. (3) The resource reallocation steps are as follows: Based on the probability of each resource being reallocated, sample a resource, then select a resource, and exchange the drone numbers loaded with the two resources, that is, exchange the resources of the two drones. (4) The flight speed switching steps are as follows: Based on the probability of each drone being selected to switch flight speed, sample a drone, randomly select a task it is responsible for, if the execution time of the task is earlier than the expected start time of the time window, then switch to a smaller speed, otherwise switch to a larger speed. If the task has no time window, then do not switch. For example: There are 60 task planning schemes. Repeat the above steps (1)-(4) 5 times for each planning scheme to generate 60x5 new planning schemes. Adding the original 60 task planning schemes, there are a total of 60x(5+1) task planning schemes, which are regarded as a solution set.

[0208] The updates in this invention include partial updates, experience-based updates, and global updates. Specifically:

[0209] Local update: Based on the 60x(5+1) solutions generated by updating the task planning scheme based on prior knowledge, calculate the target value of all solutions. From the 60x(5+1) solutions, group them into sets of 6, select the solution with the highest target value among the 6 solutions and keep it, discarding the others, and select a total of 60 solutions as the new search starting point.

[0210] Experience-based updates: Updates are performed based on the new search starting point after the local update. Based on the drone's mission planning scheme, the accumulated experience equation for the drones is updated. The probability of each drone being selected for task reassignment and resource reassignment is calculated using this accumulated equation. The probabilities of a drone being selected for task and resource reassignment are represented by a two-dimensional table. For example, with 3 drones, 10 resources, and 15 tasks, the probability of a drone being selected for task reassignment is represented by a 15x3 two-dimensional table. The line indicates that three drones are targeting the mission. The probability of a drone being selected for task reallocation is represented by a 10x3 two-dimensional table. The statement indicated that three drones were targeting resources. The probability of being selected for repeated resource allocation. Specifically, (1) task reallocation includes: randomly selecting a task. According to the (2) Resource reallocation includes: randomly selecting a resource. According to the The probability of a drone being selected for resource reallocation is sampled, and a resource of the sampled drone is randomly selected. The drone numbers of the two resources are swapped, meaning the drones carrying the two resources exchange the selected resources. Based on the search starting point, a new task planning scheme is generated. For example, repeating the above steps once for 60 solutions yields 60 new task planning schemes, for a total of 60 x 2 task planning schemes.

[0211] Global Update: Updates are performed based on the search starting point. For example: Given 60x2 solutions, calculate the target value of each set of 60x2 solutions in pairs. Within each set of target values, the solution with the higher target value is considered the optimal solution, and the solution with the lower target value is considered the suboptimal solution. Calculate the probability of the optimal solution being accepted. Sample the value of whether the optimal solution is accepted from a binary distribution based on this probability. If the value is 1, retain the suboptimal solution; otherwise, retain the optimal solution. Repeat this process for each pair of 60x2 solutions, calculating the probability of acceptance and retaining one solution. Finally, select 60 solutions as the new search starting point. From the generated 60 solutions, select the two solutions with the highest and second-highest target values ​​as the global optimal and suboptimal solutions, respectively. Calculate the probability of the global optimal solution being accepted. Sample the value of whether the global optimal solution is accepted from a binary distribution based on this probability. If the value is 1, retain the suboptimal global solution; otherwise, retain the global optimal solution. This generates a globally optimal task planning scheme, and the retained task planning scheme is recorded as the global optimal solution in the iterative loop. Increment the iteration count by 1, and check if the maximum number of iterations has been reached. If the maximum number of iterations has been reached, output the globally optimal solution as the optimal task planning scheme; otherwise, proceed to the next single iteration.

[0212] Repeat steps (1)-(4) above to update the task planning scheme based on prior knowledge, perform local updates, update the task planning scheme based on experience, and perform global updates. Each execution of steps (1)-(4) in sequence is considered as one iteration. The search starting point is the set of all task planning schemes for all UAVs.

[0213] Based on the above-disclosed method for UAV swarm task planning, this invention also discloses a UAV swarm task planning system. The UAV swarm task planning system includes a creation unit, a first iteration unit, a second iteration unit, and an Nth iteration unit. The creation unit randomly creates an initial task planning scheme for the UAV swarm. The first iteration unit iteratively updates the initial task planning scheme and outputs a first-iteration updated task planning scheme. The second iteration unit iteratively updates the first-iteration updated task planning scheme and outputs a second-iteration updated task planning scheme. The Nth iteration unit iteratively updates the N-1th iteration updated task planning scheme and outputs a UAV swarm task planning scheme; N≥3. Specifically, the iteration networks of the first iteration unit, the second iteration unit, and the Nth iteration unit are the same.

[0214] The iterative network described in this embodiment of the invention includes a prior knowledge update module, a first filtering update module, an experience-guided update module, and a second filtering update module. The prior knowledge update module obtains a first update task planning scheme based on the prior knowledge update task planning scheme. The first filtering update module calculates the first target value of all schemes in the first update task planning scheme and selects local schemes from all schemes based on the first target value to form a second update task planning scheme. The experience-guided update module obtains a third update task planning scheme based on the experience-guided update second update task planning scheme. The second filtering update module calculates the target value of all schemes in the third update task planning scheme and selects an X-times iterative update task planning scheme based on the second target value, where X ≥ 1.

[0215] This invention addresses the challenging task planning problem of unmanned aerial vehicle (UAV) swarms, which involves the coupling of multiple subproblems such as task allocation, timing scheduling, resource allocation, and speed selection. It randomizes the initial task planning scheme for UAVs, enriching the diversity of solutions. Based on the relationships between UAV tasks, resources, and flight speed, it fully extracts relevant task planning features to reflect the efficiency and profitability of UAV task execution. It correctly guides the solution direction of the UAV swarm task planning problem using timing scheduling probabilities, flight speed switching probabilities, and task and resource reallocation probabilities. A greedy strategy is used to locally update the UAV task planning scheme. Furthermore, accumulated experience in task allocation and resource allocation is used to redistribute UAV tasks and resources again. Experienced solutions with high task profitability are utilized to improve the effectiveness of UAV swarm task planning. Finally, an annealing mechanism is used to select the globally optimal UAV swarm task planning scheme. In summary, this invention solves the problems of large planning scale and complex decision-making in UAV swarm task planning based on the interrelationships between multiple subproblems. It uses prior knowledge and experience to guide the updating of task planning schemes, reduces flight time and time window penalties, and improves the profitability of UAV task execution.

[0216] The embodiments of the present invention are further illustrated by simulation experiments.

[0217] Experimental environment: The experimental operating system is Windows 11, the CPU model is Intel i7-13700H, the hardware memory is 16G, and the compilation language is Python 3.8.

[0218] Experimental Test Cases: For the multi-subproblem coupled task planning problem proposed in this invention, there is currently no publicly available test set for comparative testing. To support the simulation effectiveness and performance testing of the embodiments of this invention, 12 test cases representing large, medium, and small-scale problems were generated through simulation, such as... Figure 5As shown, test cases 1-4 are test cases for small-scale task planning problems, while test cases 5-8 and 9-12 are test cases for medium- and large-scale task planning problems, respectively.

[0219] Evaluation metrics: To evaluate the experimental performance of the UAV swarm mission planning method based on prior knowledge and experience provided in this embodiment of the invention, the following metrics were selected: Figure 5 Test cases 1#, 5#, and 9# were used as energy absorption comparison test cases. The UAV swarm task planning method based on prior knowledge and experience will be abbreviated as KEG-HPA. To verify the performance of KEG-HPA, three algorithms—RSM, KG-HPA, and EG-HPA—were developed for comparative testing. Compared to KEG-HPA, KG-HPA randomly allocates UAV tasks and resources; compared to KEG-HPA, EG-HPA randomizes task timing, flight speed, and task and resource allocation; and compared to KEG-HPA, RSM uses a random strategy for UAV task planning. Figure 5 In test cases 1#, 5#, and 9#, RSM, KG-HPA, EG-HPA, and KEG-HPA were run 100 times each, with a maximum of 500 iterations, to compare the optimization effects. To further test the optimization effect of the method of this invention, in... Figure 5 Select one medium-sized test case (7#) and two large-scale test cases (9# and 12#), and run each of the four methods 30 times. Set the maximum number of iterations to 250.

[0220] The performance of each method is evaluated based on the target value during the optimization process: This represents the average optimization effect of the algorithm after 100 runs. Represents the average function, Indicates the first The optimal objective value for the next iteration. This represents the initial optimal objective value. k ∈ [ 0 , 500 ] , , Record the ratio of the optimal target value to the initial value in each iteration of the algorithm during a single run.

[0221] Experimental procedure: Taking test case 1# as an example, using... Figure 6 The heterogeneous UAV parameters shown, in the initial UAV swarm mission planning scheme constructed by KEG-HPA, indicate the flight trajectories of the UAVs executing their missions as follows: Figure 8 As shown. Through multiple iterations, the flight trajectory of the optimal UAV swarm mission planning scheme obtained by KEG-HPA is as follows. Figure 9As shown, drone 1 performs mission 14, drone 2 performs missions 8, 12, and 15, and drones 1-7, 9-11, and 13. Drone 1 carries resources 2 and 5, drone 2 carries resources 6, 7, and 8, and drone 3 carries resources 3, 9, and 10. The selected flight speeds for each mission are as follows: missions 4 and 13 use speed 1; missions 1, 5, 9, and 12 use speed 2; missions 3, 8, 10, 11, and 14 use speed 3; and missions 2, 6, 7, and 15 use speed 4. The actual flight times of the three drones are 42.93s, 114.91s, and 432.30s, respectively, with maximum flight time limits of 969.59s, 1051.93s, and 787.85s. Figure 8 , Figure 9 Based on the optimal drone swarm task planning scheme, the maximum flight time of each drone is satisfied, and the number of tasks performed by drone 3, which has the maximum flight time limit, is significantly higher than that of drone 1 and drone 2. The task allocation results of this experiment are reasonable.

[0222] In the experiments for test cases 1#, 5#, and 9#, four algorithms—KEG-HPA, KG-HPA, EG-HPA, and RSM—were compared, and the experimental results are as follows: Figure 7 As shown: Figure 7 This chart shows a comparison of the average optimization performance of small-scale algorithms. Figure 8 A comparison chart showing the average optimization performance of medium-scale algorithms; Figure 9 A graph showing the average optimization performance of large-scale algorithms; Figure 10 A comparison chart showing the distribution of target values ​​for small-scale algorithms; Figure 11 This chart shows a comparison of the distribution of target values ​​for medium-scale algorithms. Figure 12 A comparative chart showing the distribution of target values ​​in large-scale algorithms. Based on... Figures 7-9 It can be seen that KEG-HPA performs slightly worse than other algorithms in small-scale test cases, but its average optimization performance is better than the other three algorithms in medium and large-scale test cases. Because the number of drones, tasks, and resources involved in small-scale test cases is too small, KEG-HPA may not be able to fully demonstrate its advantages in small-scale drone swarm task planning scenarios. However, in large-scale drone swarm task planning scenarios, KEG-HPA consistently outperforms other algorithms on average during the optimization process, indicating that KEG-HPA has good performance in large-scale drone swarm task planning scenarios. Figures 13-15 It can be seen that, J [ s ] With the target value as the median, KEG-HPA's median in small-scale test cases is similar to that of other algorithms, but its median in medium- and large-scale test cases is greater than that of other algorithms, indicating that the optimization efficiency of the KEG-HPA algorithm is higher than that of the other three algorithms.

[0223] In the experiments for test cases 7#, 9#, and 12#, the experimental results are as follows: Figure 15 As shown. Figure 15 The horizontal axis represents the average optimal target value, with three sets of data, each containing four data points. From left to right, these represent the average optimal target values ​​for KEG-HPA, EG-HPA, KG-HPA, and RSM, respectively. Figure 15 It can be seen that the optimal target value distribution of KEG-HPA is significantly better than the other three algorithms, proving the effectiveness of the UAV swarm task planning method based on prior knowledge and experience in optimizing performance in medium and large-scale scenarios.

[0224] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

[0225] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0226] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for planning unmanned aerial vehicle (UAV) swarm missions, characterized in that, include: Step 1: Randomly create an initial mission planning scheme for the drone swarm; Step 2: The initial task planning scheme is iterated multiple times using a single iteration method to obtain a UAV swarm task planning scheme. The single iteration includes: first, updating the first planning scheme based on prior knowledge; second, selecting local schemes from all the schemes updated in the first iteration for a second update to obtain a second planning scheme; then, updating the second planning scheme a third time based on experience to obtain a third planning scheme; finally, selecting a fourth planning scheme from all the schemes in the third planning scheme; wherein, the first planning scheme is the initial task planning scheme or the single iteration-ending task planning scheme obtained in the previous iteration; the fourth planning scheme is the single iteration-ending task planning scheme or the UAV swarm task planning scheme.

2. The UAV swarm mission planning method according to claim 1, characterized in that, In step one, an initial task planning scheme is randomly created for the drone swarm, including: Obtain parameter information of the drone swarm; the parameter information includes drone information, mission information, and onboard resource information; The task planning information for each drone in the drone cluster is randomly assigned based on the parameter information; the task planning information includes the execution task information, onboard resource information, task timing priority, and drone flight speed for each drone. The initial task planning scheme is obtained by summarizing the task planning information.

3. The method for planning unmanned aerial vehicle (UAV) swarm missions according to claim 1, characterized in that, In step two, the initial task planning scheme is iterated multiple times using a single iteration method to obtain the UAV swarm task planning scheme, including: The initial task planning scheme is iterated once to obtain all task planning schemes updated in one iteration, and the task planning scheme with the highest target value is selected as the input for the next single iteration; Repeat the single iteration and calculate the probability of each task planning scheme being accepted, and obtain the UAV swarm task planning scheme based on the probability.

4. The UAV swarm mission planning method according to claim 1, characterized in that, The single iteration specifically refers to: First, the probability of the first planning scheme is calculated based on the prior knowledge of the drone swarm. Based on the probability, the task planning scheme is replanned to obtain an updated task planning scheme. The prior knowledge includes the assigned tasks, the assigned resources, and the flight speed. Secondly, all schemes of the one-time update task planning scheme are obtained and grouped. The target value of all schemes is calculated according to the group. The group with the highest target value is selected as a local scheme to form the second planning scheme. Then, based on experience, the probability of the second planning scheme is calculated, and the task planning scheme is re-planned based on the probability to obtain the third planning scheme; Finally, all schemes of the third planning scheme are grouped into groups of two, and their target values ​​are calculated for each group. Based on the target values, a group of schemes is selected from the groups, and the optimal and second-best schemes of the group are determined. The optimal or second-best scheme is then selected as the fourth planning scheme.

5. The UAV swarm mission planning method according to claim 4, characterized in that, The calculation of the probability of the first planning scheme based on prior knowledge of the UAV swarm includes: calculating optimization terms based on prior knowledge of the UAV swarm, and calculating probability terms corresponding to the optimization terms; wherein: The optimization terms include the mission benefit range ratio, mission benefit segment length ratio, average mission benefit weight ratio, and time window penalty; the probability terms include mission redistribution probability, mission timing scheduling probability, airborne resource redistribution probability, and UAV flight speed switching probability.

6. The UAV swarm mission planning method according to claim 5, characterized in that, The expression for the mission benefit-range ratio is: In the formula, Indicates drone Mission benefit-to-range ratio Indicates drone The set of assigned tasks Indicates drone The set of loading resources, Indicates drone Flight range, Indicates drone Total task revenue, Indicates drone The total flight distance; The expression for the mission benefit segment length ratio is: In the formula, Indicates task Mission benefit segment length ratio Indicates drone The execution of the first One task, express The prerequisite tasks. Indicates drone The set of loading resources, Indicates the number of drones, Indicates drone Carry resources Execute the task The rewards obtained from the task Indicates drone Carry resources Execute the task The rewards obtained from the task; The expression for the average task revenue weight ratio is: In the formula, Representing resources Average task reward weight ratio Indicates drone , Indicates drone The set of assigned tasks Indicates drone The loaded first One resource, Indicates the first The resource number, Indicates drone Carry resources Mission rewards obtained along the flight path Representing resources The weight; The expression for the time window penalty is: In the formula, Indicates drone Execute the task Time window penalty Indicates drone To perform the task, Indicates the execution of a task The earliest expected start time, Indicates the execution of a task The actual time of execution, Indicates task The latest expected end time; The expression for the task redistribution probability is: In the formula, This represents the probability of task redistribution for drones. This indicates the number of drones in a drone swarm. Indicates drone Mission benefit-to-range ratio Indicates drone The set of assigned tasks Indicates drone The set of loading resources; The expression for the task timing scheduling probability is: In the formula, This represents the timing probability of the drone performing the task. Indicates task Mission benefit segment length ratio Indicates drone The execution of the first One task, Represents any task , Indicates the number of tasks. Indicates the first The task number of each task. Indicates the first The task number of each task; The expression for the airborne resource reallocation probability is: In the formula, This indicates that the drone carries the resource. The probability of redistribution received, Indicates the quantity of resources. Representing resources Average task reward weight ratio Indicates drone The loaded first One resource, Indicates the first The resource number, Indicates drone The resources carried; The expression for the drone's flight speed switching probability is: In the formula, This indicates the probability of switching flight speeds for the drone. Indicates drone Time window penalties for all tasks This represents all the tasks of a single drone; This indicates the number of drones in a drone swarm. Indicates drone Time window penalties for all tasks Indicates drone .

7. The UAV swarm mission planning method according to claim 4, characterized in that, The formula for calculating the target value is: In the formula, Indicates the target value. This represents the weighting coefficient of the total task revenue. This represents the total revenue generated from all missions performed by the drones. This represents the total flight time of all drones. This represents the weighting coefficient for total flight time. This represents the time window penalty coefficient. This represents the total time window penalty for all drone missions; where: Total revenue from all drone missions The expression is: In the formula, This indicates the number of drones in a drone swarm. Indicates the number of tasks. Indicates the quantity of resources. Indicates that the drone carries resources Execute the task The rewards obtained from the task Represents a 0-1 decision variable; Total flight time of all drones The expression is: In the formula, This indicates the number of drones in a drone swarm. Indicates the number of tasks. This refers to a drone. Indicates another drone, Indicates the first One task, Indicates drone Execute the task and tasks The actual time of execution, 0-1 decision variables; Total time window penalties for all drone missions The expression is: In the formula, Indicates the number of tasks. Indicates another drone, Indicates task The earliest expected start time, Indicates the execution of a task The actual time of execution, Indicates task The latest expected end time.

8. The UAV swarm mission planning method according to claim 4, characterized in that, The probability of the second planning scheme is obtained through a cumulative UAV experience equation and experience-guided calculations. The cumulative UAV experience equation includes cumulative experience in task allocation and cumulative experience in resource allocation; wherein: The cumulative experience expression for task allocation is: In the formula, This represents the initial value for task allocation. Indicates the number of drones, Indicates drone Accumulated experience in task allocation Indicates the current iteration number. Indicates the learning rate. This indicates the number of current optimal solutions. Represents a 0-1 decision variable; The cumulative empirical expression for resource allocation is: In the formula, This represents the initial value for resource allocation. Indicates the quantity of resources. This represents accumulated experience in resource allocation. Indicates the current iteration number. Indicates the learning rate. This indicates the number of current optimal solutions. This represents a 0-1 decision variable.

9. A drone swarm mission planning system, based on the drone swarm mission planning method according to any one of claims 1-8, characterized in that, include: A creation unit randomly creates an initial task planning scheme for the drone cluster; A first iteration unit, wherein the first iteration unit iteratively updates the first planning scheme to obtain a first iteration updated task planning scheme; A secondary iteration unit, wherein the secondary iteration unit iteratively updates the task planning scheme of the first iteration update to obtain a secondary iteration update task planning scheme; The N-th iteration unit iterates and updates the task planning scheme through N-1 iterations to obtain the UAV swarm task planning scheme; N≥3； The iteration networks of the first iteration unit, the second iteration unit, and the Nth iteration unit are the same.

10. The UAV swarm mission planning system according to claim 9, characterized in that, The iterative network includes: The prior knowledge update module updates the first planning scheme to obtain an updated task planning scheme. The first filtering and updating module calculates and filters the first update task planning scheme to obtain the second update task planning scheme. An experience-guided update module updates the secondary update task planning scheme to obtain a tertiary update task planning scheme. The second filtering and updating module calculates and filters the three update task planning schemes to obtain an X-times iterative update task planning scheme, where X ≥ 1.

Images