Airborne resource limited heterogeneous unmanned aerial vehicle cluster dynamic alliance and cooperation task planning method

Abstract

The invention relates to an airborne resource limited heterogeneous unmanned aerial vehicle cluster dynamic alliance and cooperation task planning method. Firstly, a double-layer coding scheme is designed, an outer layer optimizes a grouping strategy of heterogeneous unmanned aerial vehicles, and an inner layer completes specific distribution and task planning of the unmanned aerial vehicles. Secondly, an unmanned aerial vehicle cooperation strategy based on the dynamic alliance idea is provided, the limitation of a traditional fixed cooperation relation is broken through, and a more flexible and efficient cooperation mode is achieved. In order to solve the problem that the task position in the actual disaster environment is difficult to accurately obtain, an investigation difficulty classification formula based on the shape of a detection area is constructed and is used for quantifying the task investigation difficulty in the disaster environment. In addition, a meta-knowledge migration strategy is improved, the evolutionary state is judged by calculating the distance between the population centroid and the optimal solution, the migration radius is determined in a self-adaptive mode, and efficient distribution scheme migration is achieved. In conclusion, according to the method, multiple strategies are fused, resource allocation is optimized and calculated, the heterogeneous unmanned aerial vehicle resource utilization rate is improved through a dynamic alliance mechanism, high-value tasks are scheduled preferentially in combination with an investigation difficulty classification formula, and finally the sub-population task planning efficiency is improved by using an improved meta-knowledge migration strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) mission planning, and in particular to a method for dynamic alliance and collaborative mission planning of a cluster of heterogeneous UAVs with limited onboard resources. Background Art

[0002] In recent years, national policies have explicitly called for accelerating the application of drones in areas such as emergency rescue, power line inspection, and ecological monitoring, driving rapid industry development. However, existing single drones often face challenges when performing missions, including insufficient endurance, limited coverage, and inadequate ability to cope with complex environments. This leads to low mission efficiency and significant resource waste. Furthermore, the mission planning phase of drone swarms lacks an effective collaborative optimization mechanism, hindering mission planning efficiency. Therefore, developing a more precise and efficient method for dynamic alliance and collaborative mission planning for heterogeneous drone swarms is crucial.

[0003] Currently, collaborative UAV mission planning frameworks can be categorized into two main approaches: centralized and hierarchical decoupled. Centralized approaches prioritize the tight coupling of task allocation and path planning, but suffer from low computational efficiency. Hierarchical decoupled approaches prioritize simplified problem solving, but cannot guarantee global optimality. Therefore, given limited onboard resources, the challenge of designing a more robust model that leverages the advantages of both centralized and distributed approaches, while simultaneously improving computational efficiency and ensuring global optimality, remains a pressing issue.

[0004] Furthermore, while the traditional fixed-role collaboration model for drone swarms can achieve basic task division and collaboration, it suffers from limitations in practical applications, such as inflexible task allocation. Furthermore, current research often treats task locations as known fixed coordinates, which is inconsistent with the real-world disaster environment. Furthermore, current knowledge transfer algorithms cannot be effectively applied to complex, dynamic drone mission planning problems. In summary, it is necessary to propose a heterogeneous drone alliance strategy that overcomes the limitations of traditional fixed-role collaboration, as well as a reconnaissance difficulty calculation method that is responsive to real-world disaster environments, to achieve optimal resource allocation and maximize benefits. Furthermore, developing an efficient knowledge transfer algorithm that is adaptable to drone mission planning scenarios is crucial. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for dynamic alliance and collaborative task planning of heterogeneous UAV clusters with limited onboard resources, so as to improve the efficiency of task planning of heterogeneous UAV clusters in scenarios with scarce onboard resources.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for dynamic alliance and collaborative mission planning of heterogeneous UAV swarms with limited onboard resources includes:

[0008] A two-layer coding scheme is proposed. The outer layer coding optimizes and determines the grouping scheme of heterogeneous drone clusters, while the inner layer coding mechanism optimizes each task subgroup to optimize the task execution order.

[0009] A dynamic alliance strategy for heterogeneous UAVs is proposed to give full play to the performance advantages of heterogeneous UAVs and improve resource utilization;

[0010] Two mission reconnaissance difficulty calculation formulas based on the shape of the mission area are introduced to enable drones to prioritize high-value and low-difficulty missions, thus achieving reasonable resource allocation.

[0011] An adaptive meta-knowledge transfer method suitable for UAV mission planning scenarios is proposed. At the same time, an adjustment strategy for illegal solutions is given to effectively solve the mapping problem of transferred knowledge between different sub-problems.

[0012] Optionally, a two-layer encoding scheme includes:

[0013] The outer layer coding is used to optimize the grouping of heterogeneous drones to ensure the matching of tasks and drone functions; the inner layer coding, based on the outer layer decision, finds the optimal drone collaborative planning strategy and execution sequence, and feeds back the best benefits obtained from evolution to the outer layer to adjust the drone grouping.

[0014] Optionally, the dynamic alliance strategy of heterogeneous drones includes:

[0015] The strategy allows drones to flexibly adjust their collaborative relationships according to real-time mission requirements, and achieve optimal resource allocation and efficient mission execution by flexibly forming alliances with different drones of different types.

[0016] Optionally, two mission reconnaissance difficulty calculation formulas based on the shape of the mission area include:

[0017] To account for the uncertainty of obtaining the exact location of tasks in disaster scenarios, we use intervals to represent task locations, constraining them to lie within a quadrilateral defined by vertex coordinates. Quadrilaterals can be categorized as convex or concave based on their shape. Based on data such as the signal strength and side lengths of the quadrilateral's vertices, we can further calculate the detection difficulty of different tasks.

[0018] Optionally, an adaptive meta-knowledge transfer and constraint-based solution adjustment method applicable to UAV mission planning scenarios includes:

[0019] A new method for describing the evolutionary state of a population is proposed: the evolutionary trend of the population is determined by calculating the change in the Euclidean distance between the centroid of individuals in two consecutive generations and the optimal solution. Based on the changing pattern of this distance, a new migration operator is designed to achieve knowledge transfer. After the crossover operation of the inner-layer individuals is completed, the algorithm verifies the validity of the drone serial number. If the verification finds a violation of the constraints, the drone serial number is restored to the original corresponding drone serial number.

[0020] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0021] The present invention realizes hierarchical collaborative optimization of heterogeneous UAV grouping and mission planning by establishing a double-layer coding structure, adopts a dynamic alliance strategy of heterogeneous UAVs combined with an inner-layer meta-knowledge migration algorithm to improve resource utilization under limited airborne resources, and selects high-yield tasks by defining a mission reconnaissance difficulty assessment model in complex environments, thereby comprehensively improving mission execution efficiency and comprehensive benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings used in the embodiments. Obviously, the drawings described below only represent some embodiments of the present invention. Those skilled in the art can deduce other drawings based on these drawings without inventive work.

[0023] Figure 1 A flow chart of a method for dynamic alliance and collaborative mission planning of a cluster of heterogeneous UAVs with limited onboard resources provided by an embodiment of the present invention;

[0024] Figure 2 A structural diagram of a double-layer encoding method provided by an embodiment of the present invention;

[0025] Figure 3 The inner layer individual cross-variation flow chart provided by the embodiment of the present invention;

[0026] Figure 4 This is a flowchart of inner-layer task planning optimization provided by an embodiment of the present invention;

[0027] Figure 5 Flowchart for optimizing the outer UAV grouping scheme provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will provide a clear and complete description of the technical solution in conjunction with the accompanying drawings of the embodiments of the present invention. It should be pointed out that the described embodiments are only a part of the present invention and do not cover all possible embodiments. On the basis of the present invention, any other embodiments that can be obtained by any ordinary technician in this field without creative work are within the scope of protection of the present invention. The purpose of the present invention is to provide a method for dynamic alliance and collaborative task planning of heterogeneous drone clusters with limited onboard resources to improve resource utilization and task execution benefits.

[0029] The purpose of the present invention is to provide a method for dynamic alliance and collaborative task planning of heterogeneous UAV clusters with limited onboard resources, so as to further improve the efficiency and execution benefits of task planning.

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] like Figure 1 As shown in Figure 1, the dynamic alliance and collaborative task planning method for heterogeneous UAV clusters with limited onboard resources includes:

[0032] Step 101: Initialize and generate multiple drone grouping schemes and inner-layer mission planning strategies based on the two-layer coding scheme. Specifically, it includes:

[0033] (1) In the outer optimization, the model uses integer coding to assign the same number of drones of each type to each task subgroup. Assume that a cluster of 15 drones includes 9 reconnaissance drones (numbered 1-9), 3 transport drones (numbered 10-12) and 3 communication drones (numbered 13-15). The number of drones of each type assigned to each task subgroup is the same, and the chromosomes are represented as <1, 3, 5, 10, 15, 2, 4, 9, 12, 14, 6, 7, 8, 11, 13>. According to the chromosome arrangement rule, when the number of subgroups is set to 3, then each 5-bit code represents a drone subgroup, and the specific grouping is as follows: the first subgroup consists of reconnaissance drones U1, U3, U5, transport drone U10 and communication drone U15; the second subgroup contains reconnaissance drones U2, U4, U9, transport drone U12 and communication drone U14; the third subgroup is equipped with reconnaissance drones U6, U7, U8, transport drone U11 and communication drone U13.

[0034] (2) In the inner optimization process, in order to optimize the drone allocation and task execution order at the same time, the model adopts a four-dimensional individual encoding method, and simultaneously determines the drone serial number in the individual and the execution priority of the task in the subgroup. Specifically: the first two dimensions record the serial numbers of the two drones performing the task, the third dimension stores the priority index s, and the fourth dimension marks the task serial number τ. Let the inner individual encoding matrix be Mx∈R n×4, where n is the total number of tasks, the encoding of the i-th task can be expressed as:

[0035] Mx i = i,1 , U i,2 , s i , τ i >

[0036] Among them, U i，1 , U i，2 To perform task T i The drone, s i is the task priority, τ i The task number.

[0037] Assuming that a task subgroup contains a tasks, the algorithm selects the appropriate drone based on the assigned drone group and the nature of the task, and fills in the corresponding dimensions. By sorting the task priority index s, the algorithm determines the execution order of the tasks, that is:

[0038] T g =(τ1, τ2, ..., τ a )

[0039] Where s1<s2<…<s a , T g Execution order for task subgroups.

[0040] In this two-layer coding scheme, the outer layer optimizes the grouping of heterogeneous drones to ensure that tasks and drone functions are aligned. Based on the outer layer's decisions, the inner layer searches for the optimal collaborative planning strategy and execution sequence for the drones, feeding back the resulting optimal gains to the outer layer to adjust the drone grouping. This two-way interaction allows for continuous iterative optimization of drone grouping and mission execution strategies, approaching the global optimal solution within a limited number of iterations. Figure 2 This is a structural diagram of the double-layer encoding method provided by an embodiment of the present invention.

[0041] Step 102: Calculate the knowledge transfer radius using the change in the Euclidean distance between the optimal solution location in the current main task population and the population center of mass. Specifically including:

[0042] Calculate the distance between the center of mass of the current population of the main task and the location of the optimal value. The formula is as follows:

[0043] csd=||x best -c||

[0044] Among them, x best ​The coordinates of the center of mass c are the mean of the two drone numbers assigned to the task in all scenarios. The quality distance (csd) of the current generation is denoted as δ, and the migration radius is defined as follows:

[0045]

[0046] Among them, δ g-1 The optimal distance represents the previous generation's optimal distance of the target population. By analyzing the optimal distances of two consecutive generations, the algorithm dynamically adjusts the migration radius to adapt to different stages of population evolution. When the optimal distance decreases, the algorithm multiplies it by a random number in (0, 1) to generate the migration radius of the population in the convergence state. Conversely, when the optimal distance of the current generation decreases, the algorithm multiplies the optimal distance by a random number in (1, 2) to generate the migration radius of the population in the exploration state.

[0047] Step 103: Obtain the meta-knowledge migration population based on the knowledge migration radius. This includes:

[0048] Based on the changing law of quality distance, this paper proposes a new migration operator calculation formula:

[0049]

[0050] Among them, cw and c t are the centroids of the migration population and the target population, The migrating individuals of the migrating population are composed of the best individuals from other populations.

[0051] Step 104: The inner layer individuals undergo crossover mutation, and the highest mission benefit of each drone grouping scheme is used as fitness feedback to the outer layer optimization. Specifically, it includes:

[0052] (1) First, in the inner layer optimization, the allocation of UAVs is based on a dynamic alliance strategy. The traditional fixed collaboration method may lead to an imbalance in resource allocation, making it difficult to meet the needs of all tasks and may increase the flight burden of the UAV formation. To this end, this algorithm dynamically adjusts the UAV collaboration objects according to the task requirements, realizes a richer UAV collaboration mode, and improves the flexibility and adaptability of task planning. Under the dynamic alliance strategy, the UAV U j The collaboration object changes with the task, namely:

[0053]

[0054] in, Indicates drone U j Execute Task T h The collaborating partner at that time, Indicates drone U jExecute Task T k But the above inequality is not always true. When task T h and T k For the same type of mission, UAV U j There is a certain probability that the same collaboration object will be selected, namely:

[0055] If T h With T k Same type

[0056] Dynamic alliances do not mean changing collaborators for each mission, but rather flexibly adjusting them based on mission requirements to improve mission planning adaptability and resource utilization efficiency. At the same time, the alliances formed by UAVs under the dynamic alliance strategy must still meet UAV allocation constraints.

[0057] (2) Calculation method for task reconnaissance difficulty. In practical applications, the location of the task point is often difficult to obtain accurately, involving factors such as environmental complexity and sensor errors. To characterize this uncertainty, this paper uses intervals to represent the task location and constrains it to be located within a quadrilateral defined by vertex coordinates. The vertex coordinates of each quadrilateral are expressed in interval form to more accurately describe the uncertainty of the task location.

[0058] Based on the differences in area shape, this study divides them into convex and concave quadrilaterals, and believes that the difficulty of detecting the specific coordinates of the target point in these two types of areas varies. Convex quadrilaterals are relatively simple and common. Inspired by the relative position model of the workshop, this paper uses the degree of center of mass offset to define the difficulty of detecting the exact coordinates of the task point in a convex quadrilateral area. Specifically, the detection difficulty is defined as follows:

[0059]

[0060] Among them, G is the coordinate position of the polygon centroid, O p is the coordinate of the pth vertex, d(G, O p ) represents the center of mass to vertex O p The Euclidean distance, D fi For task T i Reconnaissance difficulty. The farther the polygon's vertices are from the center of mass and the stronger the signal strength, the more difficult it is to detect the exact target coordinates. Concave quadrilaterals are more complex, and as the edges of the quadrilateral expand and the aspect ratio becomes extreme, the difficulty of detecting the target point increases. This article defines the difficulty of reconnaissance in concave quadrilaterals as follows:

[0061]

[0062] Among them, C is the task Ti The perimeter of the concave quadrilateral, L is the side length of the concave quadrilateral, d max and d min The maximum and minimum distances from the center of mass to any vertex of the graph are respectively, and w1, w2, and w3 are the weight coefficients of the perimeter of the concave quadrilateral, the ratio of the longest ratio to the shortest ratio, and the ratio of the farthest to the closest distances from the center of mass to the graph vertex, respectively. As the perimeter of the concave quadrilateral increases, the ratio of the longest side to the shortest side increases, and the ratio of the distance from the center of mass to the vertex becomes extreme, the difficulty of determining the exact location of the mission point also increases. Based on the above calculation, the difficulty of reconnaissance for each mission point is normalized according to the following formula. In the future, it can be used as the cost function of the drone to perform the mission by multiplying the difficulty coefficient:

[0063]

[0064] Where, T k is the quadrilateral shape of the task, T k =1 indicates a convex quadrilateral, T k =2 represents a concave quadrilateral; e represents the difficulty of investigation. The higher the difficulty of investigation, the greater the probability of failure of the task execution and the lower the probability of gaining benefits.

[0065] (3) The crossover operation in the inner optimization is based on the drone serial number of the first two dimensions, and the mutation operation is based on the task priority index of the third dimension. During the crossover process, the algorithm randomly selects chromosomes with the same task properties and exchanges their first two-dimensional genes to generate a new drone allocation plan. Finally, the unreasonable drone serial numbers obtained by the crossover process are checked and modified. At the same time, the algorithm uses Figure 3 The inner-layer individual crossover and mutation flow chart provided in this embodiment of the present invention performs crossover and mutation optimization on the mission planning scheme. After each iteration, the algorithm calculates the mission planning scheme's benefits and selects the pop individuals with the highest benefits. After an inter_gen round of iterations, the algorithm finds the maximum mission execution benefit for each outer-layer drone grouping scheme. Figure 4 This is a flowchart of the inner-layer task planning optimization provided by an embodiment of the present invention.

[0066] (4) Constraint-based solution adjustment strategy. During the meta-knowledge transfer process, the transferred knowledge is obtained by multiplying the drone serial number of the task in other populations by a random number. Although the crossover operation is only performed between individuals with the same task nature, when the crossover individuals come from the population obtained by knowledge transfer, it is still possible that the drone serial number and the task type do not match. In addition, even if they match, the drone serial number may not belong to the subpopulation of the corresponding task type, thereby violating the task allocation constraint. Therefore, after the crossover operation of the inner individuals is completed, the algorithm needs to verify the legitimacy of the drone serial number. If the verification finds that the constraint is violated, it will be restored to the original corresponding drone serial number. Specifically, the correction steps are as follows:

[0067] ① Rounding up: First, the meta-knowledge obtained by transfer is rounded up to convert it into integer individuals.

[0068] ② Check crossover individuals: Cross the corrected individuals with the target population. If there are still illegal solutions after the crossover (for example, the drone type does not match the mission type or the drone does not belong to the group currently assigned to the mission subgroup), further adjustments are required.

[0069] ③ Adjustment strategy: For the detected illegal drone serial numbers, they should be adjusted to the corresponding allocation in the source individuals to ensure the feasibility of the solution.

[0070] Step 105: The outer optimization updates the drone grouping scheme based on fitness feedback and iterates until the termination condition is met. Specifically, it includes:

[0071] The outer optimization algorithm generates offspring individuals through random crossover (drone numbers within different subgroups) and passes them to the inner optimization module to calculate their fitness. After multiple rounds of iteration, the inner optimizer feeds the highest mission benefit obtained as the fitness value of the drone grouping scheme back to the outer optimizer. The outer optimizer merges the parent and offspring individuals, selects the POP drone grouping schemes with the highest fitness values, and returns to step 101 for the next round of optimization until the termination condition is met. Figure 5 This is a flowchart of outer-layer task planning optimization provided by an embodiment of the present invention.

[0072] This embodiment can also be implemented by the following code:

[0073] enter:

[0074] Maximum number of iterations for outer optimization gen

[0075] The maximum number of iterations of inner optimization inter_gen

[0076] Outer population size POP

[0077] Inner population size pop

[0078] Output:

[0079] The most profitable heterogeneous drone swarm grouping and mission planning scheme

[0080] Initialize POP drone grouping plan

[0081] for k=1 to gen do

[0082] for m=1 to POP do

[0083] Generate pop mission planning solutions for the current drone grouping solution

[0084] for i=1 to inter_gen do

[0085] Calculate the meta-knowledge migration radius and obtain the knowledge migration population

[0086] Perform individual crossover and mutation and calculate offspring fitness values

[0087] Merge the parent and child generations and select the pop task planning scheme with the highest fitness value

[0088] end for

[0089] Calculate the benefits of the current optimal task planning scheme and feed back the optimal fitness value of the current grouping scheme to the outer optimizer

[0090] end for

[0091] Merge the parent and child generations and select the POP drone grouping scheme with the highest fitness value

[0092] Randomly generate crossover group numbers and crossover column numbers to produce new offspring individuals

[0093] end for

[0094] Output the highest-yield heterogeneous drone swarm grouping and mission planning solution

[0095] This invention provides a dynamic alliance and collaborative task planning algorithm for heterogeneous drone swarms under conditions of limited onboard resources in disaster scenarios. Traditional solutions have shortcomings in resource utilization and task planning efficiency, making it difficult to fully utilize the performance advantages of heterogeneous drones. This is particularly complex in scenarios with limited resources and diverse task requirements. To this end, this invention proposes a two-layer optimization framework that hierarchically processes drone grouping and task planning. The outer layer first performs grouping calculations on heterogeneous drones to provide reasonable initial conditions for task allocation; simultaneously, a dynamic drone alliance strategy is introduced to optimize the utilization of onboard resources. The inner layer employs a meta-knowledge-based population individual migration strategy, transferring high-yield task planning solutions from other task subgroups as knowledge, thereby optimizing the task execution sequence of the target population. Furthermore, the model fully considers the complexity of rescue operations in real-world disaster scenarios and incorporates the difficulty of mission surveying into the benefit evaluation system, helping drones prioritize high-yield tasks. This significantly improves task execution benefits in resource-constrained and high-complexity scenarios.

[0096] The embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the same or similar parts. The description of the disclosed system is brief because it corresponds to the disclosed method. For related content, please refer to the description of the method.

[0097] This article uses specific examples to explain in detail the principles and implementation methods of the present invention. These examples are primarily intended to deepen understanding of the present invention's methods and core concepts. Those skilled in the art may make corresponding adjustments in actual implementation and application based on the basic concepts of the present invention. Therefore, the content of this article should not be construed as limiting the present invention.

Claims

1. A method for dynamic alliance and collaborative task planning of heterogeneous UAV swarms with limited onboard resources, characterized by: include: A two-layer coding scheme is proposed. The outer layer coding optimizes and determines the grouping scheme of heterogeneous drone clusters, while the inner layer coding mechanism optimizes each task subgroup to optimize the task execution order. A dynamic alliance strategy for heterogeneous UAVs is proposed to give full play to the performance advantages of heterogeneous UAVs and improve resource utilization; Two mission reconnaissance difficulty calculation formulas based on the shape of the mission area are introduced to enable drones to prioritize high-value and low-difficulty missions, thus achieving reasonable resource allocation. An adaptive meta-knowledge transfer method suitable for UAV mission planning scenarios is proposed. At the same time, an adjustment strategy for illegal solutions is given to effectively solve the mapping problem of transferred knowledge between different sub-problems.

2. The method for dynamic alliance and collaborative task planning of heterogeneous UAV clusters with limited onboard resources according to claim 1 is characterized in that: The dual-layer coding scheme specifically includes: (1) In the outer optimization, the model uses integer coding to assign the same number of various drones to each task subgroup. Assume that a cluster of 15 drones includes 9 reconnaissance drones (numbered 1-9), 3 transport drones (numbered 10-12) and 3 communication drones (numbered 13-15). The number of drones of each type assigned to each task subgroup is the same, and the chromosome is represented as <1, 3, 5, 10, 15, 2, 4, 9, 12, 14, 6, 7, 8, 11, 13>. According to the chromosome arrangement rule, when the number of subgroups is set to 3, then each 5-bit code represents a drone subgroup, and the specific grouping is as follows: the first subgroup consists of reconnaissance drones U1, U3, U5, transport drone U10 and communication drone U15; the second subgroup contains reconnaissance drones U2, U4, U9, transport drone U12 and communication drone U14; the third subgroup is equipped with reconnaissance drones U6, U7, U8, transport drone U11 and communication drone U13. (2) In the inner optimization process, in order to optimize the drone allocation and task execution order at the same time, the model adopts a four-dimensional individual encoding method, and simultaneously determines the drone serial number in the individual and the execution priority of the task in the subgroup. Specifically: the first two dimensions record the serial numbers of the two drones performing the task, the third dimension stores the priority index s, and the fourth dimension marks the task serial number τ. Let the inner individual encoding matrix be Mx∈R n×4 , where n is the total number of tasks, the encoding of the i-th task can be expressed as: Mx i = i，1 ,IN i，2 ,with i ,τ i >​ Among them, U i，1 , U i，2 To perform task T i drones, s i is the task priority, τ i The task number. Assuming that a task subgroup contains a tasks, the algorithm selects the appropriate drone based on the assigned drone group and the nature of the task, and fills in the corresponding dimensions. By sorting the task priority index s, the algorithm determines the execution order of the tasks, that is: T g =(τ1,τ2,...,τ a ) Where s1<s2<…<s a , T g Execution order for task subgroups. In this two-layer coding scheme, the outer layer optimizes the grouping of heterogeneous drones to ensure that tasks and drone functions are aligned. Based on the outer layer's decisions, the inner layer searches for the optimal collaborative planning strategy and execution sequence for the drones, feeding back the resulting optimal gains to the outer layer to adjust the drone grouping. This two-way interaction allows for continuous iterative optimization of drone grouping and mission execution strategies, approaching the global optimal solution within a limited number of iterations.

3. The method for dynamic alliance and collaborative task planning of heterogeneous UAV clusters with limited onboard resources according to claim 1 is characterized in that: The heterogeneous UAV dynamic alliance strategy specifically includes: In the inner layer optimization, the allocation of UAVs is based on a dynamic alliance strategy. The traditional fixed collaboration method may lead to an imbalance in resource allocation, making it difficult to meet the needs of all tasks and possibly increasing the flight burden of the UAV formation. To this end, this algorithm dynamically adjusts the UAV collaboration objects according to the task requirements, achieving a richer range of UAV collaboration methods and improving the flexibility and adaptability of task planning. Under the dynamic alliance strategy, UAVs j The collaboration object changes with the task, namely: in, Indicates drone U j Execute Task T h The collaborating partner at that time, Indicates drone U j Execute Task T k But the above inequality is not always true. When task T h and T k For the same type of mission, UAV U j There is a certain probability that the same collaboration object will be selected, namely: If T h With T k Same type Dynamic alliances do not mean changing collaborators for each mission, but rather flexibly adjusting them based on mission requirements to improve mission planning adaptability and resource utilization efficiency. At the same time, the alliances formed by UAVs under the dynamic alliance strategy must still meet UAV allocation constraints.

4. The method for dynamic alliance and collaborative task planning of heterogeneous UAV clusters with limited onboard resources according to claim 1 is characterized in that: The calculation formula for the task reconnaissance difficulty based on the shape of the task area specifically includes: Based on differences in regional shape, this study divides the task's quadrilateral area into convex and concave quadrilaterals, arguing that detecting the exact coordinates of the target point in these two types of areas presents different difficulties. Convex quadrilaterals are simpler and more common. Inspired by the relative position model of a workshop, this paper uses the degree of center of mass offset to define the difficulty of detecting the exact coordinates of the task point in a convex quadrilateral area. Specifically, the detection difficulty is defined as follows: Among them, G is the coordinate position of the polygon centroid, O p is the coordinate of the pth vertex, d(G, O p ) represents the center of mass to vertex O p The Euclidean distance, Df i For task T i Reconnaissance difficulty. The farther the polygon's vertices are from the center of mass and the stronger the signal strength, the more difficult it is to detect the exact target coordinates. Concave quadrilaterals are more complex, and as the edges of the quadrilateral expand and the aspect ratio becomes extreme, the difficulty of detecting the target point increases. This article defines the difficulty of reconnaissance in concave quadrilaterals as follows: Among them, C is the task T i The perimeter of the concave quadrilateral, L is the side length of the concave quadrilateral, d max and d min The maximum and minimum distances from the center of mass to any vertex of the graph are respectively, and w1, w2, and w3 are the weight coefficients of the perimeter of the concave quadrilateral, the ratio of the longest ratio to the shortest ratio, and the ratio of the farthest to the closest distances from the center of mass to the graph vertex, respectively. As the perimeter of the concave quadrilateral increases, the ratio of the longest side to the shortest side increases, and the ratio of the distance from the center of mass to the vertex becomes extreme, the difficulty of determining the exact location of the mission point also increases. Based on the above calculation, the difficulty of reconnaissance for each mission point is normalized according to the following formula. In the future, it can be used as the cost function of the drone to perform the mission by multiplying the difficulty coefficient: Where, T k is the quadrilateral shape of the task, T k =1 indicates a convex quadrilateral, T k =2 represents a concave quadrilateral; e represents the difficulty of investigation. The higher the difficulty of investigation, the greater the probability of failure of the task execution and the lower the probability of gaining benefits.

5. The method for dynamic alliance and collaborative task planning of heterogeneous UAV clusters with limited airborne resources according to claim 1 is characterized in that: The adaptive meta-knowledge transfer method applicable to UAV mission planning scenarios specifically includes: (1) Calculate the distance between the center of mass of the contemporary population of the main task and the location of the optimal value. The formula is as follows: csd=||x best -c|| Among them, x best The coordinates of the center of mass c are the mean of the two drone numbers assigned to the task in all scenarios. The quality distance (csd) of the current generation is denoted as δ, and the migration radius is defined as follows: Among them, δ g-1 The optimal distance represents the previous generation's optimal distance of the target population. By analyzing the optimal distances of two consecutive generations, the algorithm dynamically adjusts the migration radius to adapt to different stages of population evolution. When the optimal distance decreases, the algorithm multiplies it by a random number in (0, 1) to generate the migration radius of the population in the convergence state. Conversely, when the optimal distance of the current generation decreases, the algorithm multiplies the optimal distance by a random number in (1, 2) to generate the migration radius of the population in the exploration state. (2) Based on the changing law of quality-quality distance, this paper proposes a new calculation formula for the migration operator: P.S i tr Zβ*(P i s -c s )+c t Among them, c s and c t are the centroids of the migration population and the target population, P i s Migrant individuals of a migrating population. The individuals of the migrating population are composed of outstanding individuals from other populations. (3) Constraint-based solution adjustment strategy. During the meta-knowledge transfer process, the transferred knowledge is obtained by multiplying the drone serial number of the task in other populations by a random number. Although the crossover operation is only performed between individuals with the same task nature, when the crossover individuals come from the population obtained by knowledge transfer, it is still possible that the drone serial number and the task type do not match. In addition, even if they match, the drone serial number may not belong to the subpopulation of the corresponding task type, thereby violating the task allocation constraint. Therefore, after the crossover operation of the inner individuals is completed, the algorithm needs to verify the legitimacy of the drone serial number. If the verification finds that the constraint is violated, it will be restored to the original corresponding drone serial number. Specifically, the correction steps are as follows: ① Rounding: First, the meta-knowledge obtained through transfer is rounded to convert it into integer individuals. ② Check crossover individuals: Cross the corrected individuals with the target population. If there are still illegal solutions after the crossover (for example, the drone type does not match the mission type or the drone does not belong to the group currently assigned to the mission subgroup), further adjustments are required. ③ Adjustment strategy: For the detected illegal drone serial numbers, they should be adjusted to the corresponding allocation in the source individuals to ensure the feasibility of the solution.

Images