A hierarchical unmanned cluster task planning method based on GP

Abstract

The application discloses a GP-based hierarchical unmanned cluster task planning method and device, equipment and medium, constructs a hierarchical task individual tree with structured semantics; the strong type genetic programming algorithm is used as a population evolution baseline, the hierarchical task individual tree is input into the strong type genetic programming algorithm to generate an initial population, the performance of each initial individual is evaluated, and an adaptability evaluation result is obtained; it is judged whether the termination condition is not met according to the adaptability evaluation result, each individual is selected, crossed and mutated by using the strong type genetic programming algorithm, redundant nodes in each individual are cleaned by using a pruning algorithm based on domain knowledge in the crossing and mutation operation process, and the optimal hierarchical task individual tree is converted into a task planning scheme of an unmanned cluster cooperative system. The application solves the technical problems of unclear task structure expression ability, imperfect feasibility guarantee mechanism, inefficient search and non-adaptability in the existing unmanned cluster cooperative task planning method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent unmanned systems, and more particularly to a hierarchical unmanned cluster task planning method and device based on GP, equipment and medium. BACKGROUND

[0002] With the wide application of intelligent unmanned systems in disaster response, environmental monitoring and urban security, multi-agent collaborative task planning has become an important direction of unmanned cluster system research. For example, in the emergency fire fighting scene, a collaborative system composed of a ground unmanned vehicle (UGV) carrying multiple unmanned aerial vehicles (UAVs) needs to quickly cover and dispose multiple fire points to achieve efficient, low-cost and collaborative fire fighting task scheduling.

[0003] In such tasks, the task has obvious hierarchical structure characteristics: the UGV as the upper layer command agent is responsible for sub-task allocation, path coordination and task parameter issuance, and the UAV as the lower layer executor undertakes specific target fire fighting tasks. In order to ensure scheduling efficiency and rational use of resources, the system needs to realize automatic planning and intelligent scheduling of tasks according to the task area, UAV energy limitation, fire point distribution and other conditions. However, the research on this kind of hierarchical multi-agent collaborative task planning problem still has the following technical difficulties: (1) The task modeling structure is not clear: traditional methods (such as centralized planning, graph search or linear programming) often cannot explicitly express the hierarchical structure of the task and the interaction between the upper and lower layers, lack of structural semantic support, resulting in that the scheduling scheme is difficult to explain and extend. (2) Weak task constraint processing capability: traditional optimization algorithms lack effective structure-level checking and repair mechanisms when facing "unmanned vehicle task allocation overload", "target point repeated coverage" or "uncovered target" and other feasibility problems, and are easy to generate illegal or inefficient scheduling schemes. (3) Large search space dimension and high redundancy: traditional optimization algorithms cannot actively avoid redundant areas in coding design, causing a lot of unnecessary search and slow convergence, especially in the multi-target multi-agent environment. SUMMARY

[0004] The present application aims to at least solve the technical problems of unclear task structure expression ability, imperfect feasibility guarantee mechanism, inefficient search and non-adaptive in the existing unmanned cluster cooperative task planning method. To this end, the first aspect of the present application proposes a hierarchical unmanned cluster task planning method based on GP, including the following execution steps: based on the domain knowledge of the hierarchical multi-agent system, five types of nodes of the unmanned cluster cooperative system constructed by the fire unmanned vehicle and the multi-unmanned aerial vehicle cluster are determined, and a hierarchical task individual tree with structured semantics is constructed based on the five types of nodes; the strong type genetic programming algorithm is used as the population evolution baseline, and the hierarchical task individual tree is input into the strong type genetic programming algorithm to generate an initial population, wherein the initial individual of each initial population is a hierarchical task individual tree, and each hierarchical task individual tree corresponds to the deployment scheme of each unmanned aerial vehicle cluster and the fire unmanned vehicle when executing the task, and a pre-defined fitness function is used to evaluate the performance of each initial individual to obtain a fitness evaluation result; whether the termination condition is met is judged according to the fitness evaluation result, if the termination condition is met, the optimal hierarchical task individual tree corresponding to the optimal fitness evaluation result is output; otherwise, the strong type genetic programming algorithm is used to select, cross and mutate each individual, and the pruning algorithm based on the domain knowledge is used to clean up the redundant nodes in each individual during the cross and mutation operation to generate the offspring hierarchical task individual tree, the performance of each population individual is evaluated by using the fitness function to obtain a new fitness evaluation result, the previous execution step is iteratively executed, and the optimal hierarchical task individual tree is converted into a task planning scheme of the unmanned cluster cooperative system.

[0005] Optionally, the five types of nodes include a total task node, a lower sub-task node, an interaction sub-task node, and an upper sub-task node, wherein the total task node, the upper sub-task node, the interaction sub-task node, and the lower sub-task node have a parent-child inheritance relationship, and the upper sub-task node includes a real sub-task node and a virtual sub-task node; the pruning algorithm based on domain knowledge is used to clean up redundant nodes in each individual in the process of crossover and mutation operation, including: obtaining an initial individual after crossover; selecting a sub-tree that is not a root node from each initial individual after crossover as a randomly generated new sub-tree to obtain a mutated individual based on a basic sub-tree mutation operator of a strong type genetic programming algorithm; calculating the number of lower intelligent agents currently participating in fire extinguishing according to the lower sub-task node of the mutated individual, wherein the lower sub-task node is used to determine the fire starting point corresponding to each lower intelligent agent, and the lower intelligent agent is a UAV; determining whether the number of lower intelligent agents is greater than the number of fire starting points, if greater, randomly generating an intermediate number within the range of the number of fire starting points, and determining whether the number of lower intelligent agents is greater than the intermediate number, if greater, randomly pruning the lower sub-task node until the number of lower intelligent agents is less than or equal to the number of fire starting points; wherein in the process of pruning the lower sub-task node, it is determined that all lower sub-task nodes under the interaction sub-task node are pruned, then the entire sub-tree with the real sub-task node as the root is pruned, and the virtual sub-task node is used to replace the real sub-task node to connect the total task node.

[0006] Optionally, the five types of nodes further include a parameter node; the total task node connects each upper sub-task node, and is used to determine the overall fire extinguishing task of the unmanned swarm cooperative system during fire extinguishing according to the upper sub-task node, and the execution order of the upper sub-task node is determined according to the connection order of the total task node; the real sub-task node of the upper sub-task node includes an interaction sub-task node and a parameter node, and the parameter node is used to store interaction parameters inside and outside the unmanned swarm cooperative system, wherein the interaction parameters include path coordinate parameters and energy consumption parameters, and the interaction sub-task node represents the fire starting point corresponding to each UAV and the cooperative path between the UAV group and the unmanned fire truck; the virtual sub-task node of the upper sub-task node is used as a placeholder when the number of real sub-tasks is less than the number of sub-nodes required according to the total task node.

[0007] Optionally, the hierarchical task variable tree is a four-layer structure, and the four-layer structure includes a root layer, an upper sub-task layer, an interaction sub-task layer, and a lower sub-task layer from top to bottom; wherein the total task node belongs to the root layer; the upper sub-task node belongs to the upper sub-task layer; the interaction sub-task node and the parameter node belong to the interaction sub-task layer; and the lower sub-task node belongs to the lower sub-task layer.

[0008] Optionally, before the obtaining the initial individuals after the crossover, the method further comprises: randomly selecting a plurality of initial individuals from the initial population based on a tournament selection operator; and randomly selecting a crossover point with the same type from each initial individual based on a conventional subtree crossover operator of the strong typed genetic programming algorithm, and exchanging subtrees with the crossover point as a root node between the initial individuals to obtain each individual after the crossover.

[0009] Optionally, before the generating the initial population, the method further comprises: predefining a node set and a node connection rule, wherein the node set and the node connection rule are used to generate a constraint condition to make the hierarchical task variable tree structure legal and consistent with the task constraint.

[0010] Optionally, the expression of the fitness function is:

[0011]

[0012] wherein, , , , represents the fire starting point the fire extinguishing task completion percentage, represents the number of actual deployed UAVs, represents the number of UAVs to be deployed, represents the total task time consumption of each UAV, represents the maximum total working time limit of all UAVs, represents the task completion degree, represents the resource occupation efficiency, represents the time consumption efficiency.

[0013] To achieve the above purpose, the second aspect of the present application provides a hierarchical unmanned cluster task planning method based on GP, characterized in that it comprises:

[0014] an individual number construction module configured to determine five types of nodes of an unmanned cluster cooperative system constructed by a fire unmanned vehicle and a plurality of unmanned aerial vehicle groups based on the domain knowledge of a hierarchical multi-agent system, and construct a hierarchical task individual tree with structured semantics based on the five types of nodes;

[0015] a fitness calculation module configured to use a strong typed genetic programming algorithm as a population evolution baseline, input the hierarchical task individual tree into the strong typed genetic programming algorithm to generate an initial population, wherein each initial individual of each initial population is a hierarchical task individual tree, and each hierarchical task individual tree corresponds to a deployment scheme of each unmanned aerial vehicle group and the fire unmanned vehicle when performing a task, and use a predefined fitness function to evaluate the performance of each initial individual to obtain a fitness evaluation result;

[0016] The task planning scheme generation module is configured to determine whether a termination condition is met according to the fitness evaluation result, output an optimal hierarchical task individual tree corresponding to the optimal fitness evaluation result if the termination condition is met, or perform selection, crossover and mutation operations on each individual by using the strong-type genetic programming algorithm, clean redundant nodes in each individual by using the pruning algorithm based on domain knowledge during the crossover and mutation operations, generate offspring hierarchical task individual trees, evaluate the performance of each population individual by using a fitness function, obtain a new fitness evaluation result, and iteratively perform the previous step, and convert the optimal hierarchical task individual tree into a task planning scheme of the unmanned cluster coordination system.

[0017] To achieve the above object, the third aspect of the present application provides an electronic device, characterized in that the electronic device comprises a processor and a memory, the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to implement the GP-based hierarchical unmanned cluster task planning method of any one of the above methods.

[0018] To achieve the above object, the fourth aspect of the present application provides a computer readable storage medium, characterized in that the storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by a processor to implement the GP-based hierarchical unmanned cluster task planning method of any one of the above methods.

[0019] The embodiment of the application provides a hierarchical unmanned cluster task planning method, device, equipment and medium based on GP, which has the following beneficial effects compared with the prior art: five types of nodes of an unmanned cluster cooperative system jointly constructed by a fire unmanned vehicle and a plurality of unmanned aerial vehicle groups are determined based on the domain knowledge of a hierarchical multi-agent system, and a hierarchical task individual tree with structured semantics is constructed based on the five types of nodes; a strong type genetic programming algorithm is used as a population evolution baseline, and the hierarchical task individual tree is input into the strong type genetic programming algorithm to generate an initial population, wherein the initial individual of each initial population is a hierarchical task individual tree, and each hierarchical task individual tree corresponds to a deployment scheme of each unmanned aerial vehicle group and the fire unmanned vehicle when performing a task; a predefined fitness function is used to evaluate the performance of each initial individual to obtain a fitness evaluation result; in the population initialization stage of the application, five types of task nodes are abstracted in combination with the domain knowledge according to the hierarchical characteristics of the HMAS task, and a node set structure with hierarchical semantics is constructed. By introducing a strong type genetic programming (STGP) mechanism, the node connection relationship and the syntax type are strictly limited in the individual construction process, so that each individual tree has clear upper and lower task division and reasonable scheduling logic in structure. This design avoids the generation of illegal structures such as task allocation conflicts, target redundant assignment and execution path breakage from the source, greatly reducing the occurrence of invalid combinations. Through the fusion of structure-level constraints and domain knowledge, the feasibility of the individuals in the initial population is greatly improved, the dimension of the search space is effectively compressed, the structure quality of the population and the initial fitness of the scheduling scheme are improved, and a more efficient and stable search starting point is provided for the subsequent evolution process; whether the termination condition is met is judged according to the fitness evaluation result, if the termination condition is met, the optimal hierarchical task individual tree corresponding to the optimal fitness evaluation result is output; otherwise, the strong type genetic programming algorithm is used to select, cross and mutate each individual, and a pruning algorithm based on domain knowledge is used to clean up redundant nodes in each individual during the cross and mutation operations to generate offspring hierarchical task individual trees, the performance of each population individual is evaluated by using the fitness function to obtain new fitness evaluation results, and the last execution step is iteratively executed, and the optimal hierarchical task individual tree is converted into a task planning scheme of the unmanned cluster cooperative system. The pruning mechanism of the application eliminates redundant leaf nodes and redundant branches, and dynamically shrinks the invalid structure space. This mechanism significantly improves the feasibility of individuals, ensures that the schemes generated in the evolution process meet the scheduling resource constraints and task logic constraints; the pruning operation of the mechanism effectively compresses the dimension of the search space and improves the convergence efficiency of the optimization algorithm; the pruning mechanism improves the structural adaptability and search performance of the algorithm on the basis of ensuring scheduling legality, and the application can ensure that the task planning scheme always maintains structural feasibility and reasonable scheduling during the evolution process, further solving the technical problems of unclear task structure expression ability, imperfect feasibility guarantee mechanism, low search efficiency and lack of self-adaptability in the existing unmanned cluster cooperative task planning method. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the present application, the drawings required to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0021] Figure 1 A flow chart of a GP-based hierarchical unmanned cluster task planning method provided for an embodiment of the present application;

[0022] Figure 2 A fire-fighting task scene schematic diagram in a layered multi-agent system of a GP-based hierarchical unmanned cluster task planning method provided for an embodiment of the present application;

[0023] Figure 3 A structure schematic diagram of an individual tree of a GP-based hierarchical unmanned cluster task planning method provided for an embodiment of the present application;

[0024] Figure 4 A schematic diagram of generation of infeasible solutions in an individual tree crossover process of a GP-based hierarchical unmanned cluster task planning method provided for an embodiment of the present application;

[0025] Figure 5 An example task flow schematic diagram of a GP-based hierarchical unmanned cluster task planning method provided for an embodiment of the present application in task setting and

[0026] Figure 6 A structure block diagram of a GP-based hierarchical unmanned cluster task planning device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0028] The present specification provides method operation steps as described in the embodiments or flow charts, but can include more or less operation steps based on conventional or non-creative labor. In actual system or server product execution, the methods shown in the embodiments or drawings can be executed in sequence or in parallel (for example, parallel processor or multi-thread processing environment).​

[0029] To address the technical problems existing in the background technology of the prior art, this application proposes a hierarchical unmanned swarm task planning method based on genetic programming (GP) to solve the problems of difficulty in expressing task structure and low search efficiency in the task of minimizing resources and quickly covering multiple fire points in the "firefighting unmanned vehicle (UGV) - multiple unmanned aerial vehicle (UAV)" collaborative system.

[0030] This application, based on a hierarchical multi-agent system (HMAS), constructs a four-layer task variable tree with structured semantics, including: overall task node, upper-level sub-task node, interaction sub-task node, lower-level sub-task node, and parameter node. This structure uses domain knowledge extracted from the task execution process for node set design and type connection constraint modeling, ensuring strong expressiveness, clear structure, and valid connections in task coding, significantly improving the accuracy and systematic nature of task scheduling modeling.

[0031] In terms of algorithm design optimization, this application uses strongly typed genetic programming (STGP) as the evolutionary baseline. Through selection, subtree crossover, and mutation operations, individual trees are evolved and optimized to gradually generate higher-quality scheduling schemes. To address issues such as node redundancy, target duplication, and task allocation exceeding limits that may arise during crossover and mutation, a domain-knowledge-based pruning and repair mechanism is designed. This mechanism automatically prunes redundant subtrees and introduces virtual nodes to fill gaps, ensuring that the generated scheme maintains structural feasibility and scheduling rationality throughout the evolutionary process.

[0032] Assuming the unmanned swarm task planning scenario is a collaborative system for fire-fighting unmanned vehicles / drones (UGV / UAV), then the unmanned swarm task planning scenario is as follows: Figure 2 As shown, Figure 2 This is a firefighting scenario in HMAS. Figure 2 Includes 1 UGV (upper-level intelligent agent) and a maximum of The UAV (lower-level intelligent agent) is deployed. The UGV and its loaded drone initially reside at the red dot in the lower left corner. Orange asterisks indicate ignition points; the number of ignition points is [number missing]. In this scenario, the agents move at a constant speed, with UGVs typically moving slower than UAVs. The objective of task scheduling is to allow the drones to cover all fire points as quickly as possible with the minimum number of drones. Since the number of drones, n, is a decision variable in the task scheduling problem, different choices will lead to changes in the dimensionality of task parameters and cross-layer interaction parameters between upper and lower layer agents. This makes HMAS task scheduling a uniquely challenging problem with a variable-scale decision space.

[0033] The technical solutions of the present application will be further described below in combination with the drawings and examples:

[0034] Reference Figure 1 , Figure 1 The flowchart of the hierarchical unmanned cluster task planning method based on GP proposed in the present application can include the following execution process:

[0035] S10, based on the domain knowledge of the hierarchical multi-agent system, five types of nodes of the unmanned cluster cooperative system constructed by the fire unmanned vehicle and the multi-unmanned aerial vehicle cluster are determined, and a hierarchical task individual tree with structured semantics is constructed based on the five types of nodes;

[0036] Among them, the five types of nodes include overall task nodes, lower sub-task nodes, interaction sub-task nodes and upper sub-task nodes, wherein the overall task nodes, upper sub-task nodes, interaction sub-task nodes and lower sub-task nodes have parent-child inheritance relationship, and the upper sub-task nodes include real sub-task nodes and virtual sub-task nodes.

[0037] It should be noted that in the population initialization stage, in view of the hierarchical characteristics of the HMAS task, five types of task nodes including overall task, upper sub-task, interaction sub-task, lower sub-task and parameter node are abstracted in combination with domain knowledge, and a node set structure with hierarchical semantics is constructed. By introducing the strong type genetic programming (STGP) mechanism, the node connection relationship and syntax type are strictly limited in the individual construction process, ensuring that each individual tree has clear upper and lower task division and reasonable scheduling logic in structure. This design avoids the generation of illegal structures such as task allocation conflict, target redundant assignment and execution path breakage from the source, greatly reducing the occurrence of invalid combinations. Through the fusion of structure-level constraints and domain knowledge, the feasibility of individuals in the initial population is greatly improved, effectively compressing the dimension of the search space, improving the structure quality of the population and the initial fitness of the scheduling scheme, and providing a more efficient and stable search starting point for the subsequent evolution process.

[0038] In an embodiment of the present application, the hierarchical task variable tree is a four-layer structure, and the four-layer structure is root layer, upper sub-task layer, interaction sub-task layer and lower sub-task layer from top to bottom;

[0039] Among them, the overall task node belongs to the root layer; the upper sub-task node belongs to the upper sub-task layer; the interaction sub-task node and the parameter node belong to the interaction sub-task layer; and the lower sub-task node belongs to the lower sub-task layer.

[0040] Specifically, Figure 3 is a structural diagram of an individual tree. As Figure 3As shown, nodes are set according to the 4-layer structure of the individual tree, and then the individual tree is constructed by using the nodes. The nodes are set into 5 categories, and the 5 categories of nodes are: a general task node, an upper-layer sub-task node, an interaction sub-task node, a lower-layer sub-task node, and a parameter node.

[0041] In an embodiment of the present application, the five categories of nodes further include a parameter node; the general task node is connected with each upper-layer sub-task node, and is used for determining the overall fire extinguishing task of the unmanned swarm cooperative system during fire extinguishing according to the upper-layer sub-task nodes, and the execution sequence of the upper-layer sub-task nodes is determined according to the connection sequence of the general task node;

[0042] Exemplarily, the general task node As the root node of the variable tree, it is connected Each upper-layer sub-task node and organizes these sub-tasks into a general task, that is, represents an overall fire extinguishing task. The execution sequence of the upper-layer sub-task nodes depends on their connection sequence with the general task node.

[0043] The real sub-task nodes of the upper-layer sub-task nodes include an interaction sub-task node and a parameter node, and the parameter node is used for storing interaction parameters inside and outside the unmanned swarm cooperative system, wherein the interaction parameters include path coordinate parameters and energy consumption parameters, and the interaction sub-task node represents a corresponding fire point of each unmanned aerial vehicle and a cooperative path between the unmanned aerial vehicle group and the unmanned fire vehicle;

[0044] Exemplarily, the upper-layer sub-task node represents an upper-layer sub-task, and forms a tree structure with the general task node as a parent node. This type of node is divided into two categories: real sub-task nodes and virtual sub-task nodes. The real sub-task node includes two sub-nodes, the first one is an interaction sub-task node, and the other one is a parameter node for storing corresponding interaction parameters; the virtual sub-task node is used as a placeholder only when the number of required real sub-tasks is less than the number of required sub-nodes according to the root node. The upper-layer sub-task node corresponds to the area allocation or unmanned aerial vehicle task scheduling of the UGV. All upper-layer sub-task nodes form a set .

[0045] The virtual sub-task node of the upper-layer sub-task node is used as a placeholder when the number of real sub-tasks is less than the number of required sub-nodes according to the general task node.

[0046] Exemplarily, the interaction sub-task node is both a sub-node of the upper-layer sub-task node and a parent node of the lower-layer sub-task node, and is used for describing the cross-level interaction between the upper-layer and lower-layer intelligent agents. The interaction sub-task node embodies the cooperation and path interaction between the UGV and the UAV, and determines how many UAVs are deployed in the sub-task. All interaction sub-task nodes form an interaction sub-task node set .

[0047] Table 1 Experimental Node Set

[0048]

[0049] For example, the lower-level subtask nodes describe in detail the subtasks that the lower-level agent ultimately executes. Since the subtasks of the lower-level agent are typically simple and single-function, these nodes are terminal nodes (i.e., leaf nodes) of the variable tree. Each lower-level subtask node represents the ignition point ID corresponding to each UAV. All lower-level subtask nodes constitute the lower-level subtask node set. .

[0050] For example, as a terminal node, a parameter node is connected to the upper-level subtask node, providing coordinate parameters for the interactive subtask. Parameter nodes contain specific task parameters such as path coordinates and energy consumption. All parameter nodes constitute a parameter node set. .

[0051] Therefore, Given It can be automatically generated at that time. A variable tree covering upper-layer agent path planning (upper-layer sub-task nodes and parameter nodes), cross-level interaction (interaction sub-task nodes), and lower-layer agent task allocation (lower-layer sub-task nodes), wherein, Indicates population size. Node set. The definitions are shown in Table 1.

[0052] Based on the task scenario settings, the hierarchical task variable tree can be configured to follow a four-level structure: from top to bottom, the root level, the root level, and the root level. Upper-level sub-task layer Interactive Subtask Layer and lower sub-task layer Based on domain knowledge extracted from the execution process of HMAS general tasks, a node set containing 5 types of nodes is defined. To construct the variable tree, ensure full solution space coverage. In the process of constructing the hierarchical task variable tree using nodes, based on domain knowledge, it is clearly defined which type of node can be a parent node, which type of node must contain parameters or child nodes, and the semantic meaning of the node connection order (such as the task execution order).

[0053] Before generating the initial population, the method may further include the following execution process:

[0054] Predefined node sets and node connection rules are used to generate constraints to ensure that the hierarchical task variable tree structure is legal and conforms to task constraints.

[0055] That is, in the initialization phase, the processor needs to analyze the task scenario and participating agents, extract domain knowledge from the task characteristics and constraints, and use the domain knowledge to guide the decision variable representation to generate the initial population. The present application adopts strongly-typed GP (STGP) as the evolution baseline, integrates the domain knowledge of HMAS, and proposes a tree individual representation method to generate the initial population. In the initial population, the node set and connection rule are defined according to the task characteristics, ensuring that the generated individual tree structure is legal and meets the constraints. Each individual tree is an individual in the initial population, and each individual corresponds to a task scheduling scheme. In the process of constructing the individual tree using nodes, the processor can clearly specify which type of node can be used as a parent node, which type of node must contain parameters or child nodes, and the semantic meaning of node connection order (such as task execution order) according to the domain knowledge.

[0056] Next, after the initialization phase ends, the processor evaluates each individual in each generation population to obtain the corresponding fitness function value. If the termination condition is met, the optimal individual is saved as the task scheduling scheme; otherwise, offspring are generated through mutation, crossover and selection, and the optimal individual is further selected, wherein the crossover, mutation and selection operators basically inherit the traditional STGP method. Although the above crossover and mutation operations can ensure the feasibility of the connection of node types, they cannot ensure that the offspring meet the global task constraints of the underlying agents. Figure 4 Fig. 1 is a schematic diagram of an infeasible solution generated in the process of individual tree crossover.

[0057] The above technical problems will be further explained as follows: Figure 4 As shown in Fig. 1, it is assumed that there is a constraint that requires the number of underlying sub-tasks represented by the green terminal node to be less than 4. Both parent individuals (A) and (B) meet this constraint, but after exchanging the subtree in the black dashed box, although the root types of the subtrees are the same and all the node positions are correct, the offspring (A') contains 4 green terminal nodes and violates the constraint. Mutation operation can also lead to infeasible solutions. In order to solve the problem of infeasible solutions caused by the crossover and mutation process, the present application designs a pruning scheme based on domain knowledge to modify these infeasible individuals, to clean up redundant nodes, to reconstruct the path order, and to improve the structure rationality and the feasibility of the scheduling plan.

[0058] Figure 1is a flowchart for solving the task planning model based on GP. As shown in FIG. 1, the process of solving the task planning model based on GP is as follows: first, the processor can extract the domain knowledge and construct the task level model. Then, in the population initialization stage, the node set design is guided by the domain knowledge to ensure that the generated individual tree is fully feasible and has no redundant decision space, and the fitness function is used to evaluate the fitness of the individual tree to obtain the fitness evaluation result, which is the content of step S20 of the present application.

[0059] Further, the search stage can specifically include the following execution process:

[0060] S20, the strong typed genetic programming algorithm is used as the population evolution baseline, and the hierarchical task individual tree is input into the strong typed genetic programming algorithm to generate an initial population, wherein the initial individual of each initial population is a hierarchical task individual tree, and each hierarchical task individual tree corresponds to a deployment scheme of each unmanned aerial vehicle group and fire unmanned vehicle when performing a task, and a pre-defined fitness function is used to evaluate the performance of each initial individual to obtain a fitness evaluation result;

[0061] Specifically, the first step of the search stage can be fitness evaluation:

[0062] The processor can evaluate each individual in each generation of population to obtain the corresponding fitness function value.

[0063] Wherein, the fitness function is shown in formula (1):

[0064] (1)

[0065] In the formula, ; ; ; represents the completion percentage of the fire extinguishing task of the fire point ; represents the number of actual deployed unmanned aerial vehicles; represents the number of unmanned aerial vehicles to be deployed; represents the total task time consumption of each UAV; represents the maximum total working time limit of all unmanned aerial vehicles.

[0066] Wherein, the function combines three core indicators task completion degree , resource occupation efficiency , and time consumption efficiency into one indicator, which can comprehensively evaluate the comprehensive performance of the scheduling scheme and avoid the deviation caused by single target optimization. The higher the fitness value is, the better the scheme is.

[0067] It is worth noting that the task completion degree The function is composed of two parts: a main part and an exponential decay part. The main part is a multiplication factor that directly drives the scheduling scheme to optimize towards higher target coverage. The exponential decay part A joint penalty mechanism of resource usage and time consumption is introduced to explicitly suppress redundant scheduling and delayed execution while ensuring task effectiveness, thus prompting the algorithm to generate more compact and efficient task allocation structures. In addition, the function structure has a normalization feature that can adapt to task parameter configurations of different scales and efficiently calculate without relying on complex simulations, facilitating integration into the evolutionary search process. Overall, this design can effectively guide the algorithm to balance between task completion, resource conservation, and execution efficiency, improving the optimization capability and practicality of the overall scheduling system.

[0068] In an embodiment of the present application, before the initial individuals after the cross are obtained, the method can further include the following execution process:

[0069] Randomly selecting multiple initial individuals from the initial population based on a tournament selection operator;

[0070] Illustratively, the processor uses the widely used tournament selection operator. Randomly select individuals from the population, and the best in each batch is selected into the next generation population.

[0071] Based on the conventional subtree crossover operator of the strongly typed genetic programming algorithm, randomly select a crossover point of the same type from each initial individual, and exchange the subtree with the crossover point as the root node between the initial individuals to obtain each crossed individual.

[0072] Illustratively, the processor follows the majority GP method and uses the conventional subtree crossover operator to randomly select a crossover point of the same type from the parent individuals and exchange the subtree with the point as the root.

[0073] S30, according to the fitness evaluation result, judge whether the termination condition is met, if the termination condition is met, output the corresponding optimal hierarchical task individual tree according to the optimal fitness evaluation result; otherwise, use the strongly typed genetic programming algorithm to select, cross, and mutate each individual, and use the pruning algorithm based on domain knowledge to clean up redundant nodes in each individual during the cross and mutation operation, generate the offspring hierarchical task individual tree, evaluate the performance of each population individual using the fitness function, obtain the new fitness evaluation result, and iteratively execute the previous execution step, and convert the optimal hierarchical task individual tree into a task planning scheme of the unmanned cluster coordination system. After executing the above pseudo code, the processor judges whether the termination condition is met,

[0074] In an embodiment of the present application, the pruning algorithm based on domain knowledge can include the following execution process:

[0075] obtaining the initial individuals after the crossover;

[0076] selecting, from each of the initial individuals after the crossover, a sub-tree other than the root node as a randomly generated new sub-tree based on a basic sub-tree mutation operator of the strongly typed genetic programming algorithm, to obtain the individuals after the mutation;

[0077] calculating the number of lower-level intelligent agents participating in fire extinguishing according to lower-level sub-task nodes of the individuals after the mutation, wherein the lower-level sub-task nodes are used to determine the fire points corresponding to the lower-level intelligent agents, and the lower-level intelligent agents are unmanned aerial vehicles;

[0078] judging whether the number of lower-level intelligent agents is greater than the number of fire points, if the number of lower-level intelligent agents is greater than the number of fire points, randomly generating an intermediate number within the range of the number of fire points, and judging whether the number of lower-level intelligent agents is greater than the intermediate number, if the number of lower-level intelligent agents is greater than the intermediate number, randomly pruning the lower-level sub-task nodes until the number of lower-level intelligent agents is less than or equal to the number of fire points;

[0079] wherein, in the process of pruning the lower-level sub-task nodes, if all the lower-level sub-task nodes under the interaction sub-task node are pruned, the entire sub-tree with the real sub-task node as the root is pruned, and the virtual sub-task node is used to replace the real sub-task node to connect the overall task node.

[0080] In the specific execution process, the processor can first determine whether a termination condition is met according to the fitness evaluation result, output the optimal individual tree if the termination condition is met, and then determine the scheduling scheme; if the termination condition is not met, the selection, crossover and mutation operations are performed on the offspring individuals to generate a new individual tree.

[0081] wherein, in the process of generating the new individual tree, the processor can further use domain knowledge to prune and correct the infeasible individual tree, and perform structure reorganization on the individual tree after pruning and repairing, to generate a new individual tree that meets the constraints. Then, the processor can perform fitness evaluation on the newly generated individual tree to obtain a fitness evaluation result. Finally, the processor can determine again whether the termination condition is met according to the fitness evaluation result, and the cycle continues until the termination condition is met, to obtain the optimal individual tree and then obtain the task scheduling scheme.

[0082] For example, the processor can use the basic sub-tree mutation operator of the strongly typed genetic programming algorithm to replace the sub-tree with the randomly selected node as the root with a randomly generated new sub-tree. Wherein, the root node type of the new sub-tree is consistent with the original sub-tree to ensure the connection feasibility. Next, the processor first calculates the number of lower-level intelligent agents participating according to the lower-level sub-task nodes (lines 1-2 in the pseudo code). For the infeasible individual , randomly select and remove nodes until the condition is met If all the children of an interaction subtask node are removed, the whole sub-tree rooted at the upper subtask node will be pruned, and the connection position of the root node will be replaced by a virtual upper subtask node (lines 7-9 in the pseudo code). The introduction of the virtual node can cut off the redundant sub-tree without all the agents, thereby compressing the decision space and improving the search efficiency.

[0083] Exemplarily, in the process of executing the pruning algorithm of the field knowledge by the processor, the processor can set the number of layer nodes as , and the corresponding constraint of the number of nodes can be expressed as shown in expression (2):

[0084] (2)

[0085] wherein and constitute the feasible domain range.

[0086] It should be noted that, since only the interaction subtask node has an uncertain output dimension, this means that the number of lower subtask nodes is not constrained during the evolution process. Further considering the HMAS application scenario, the processor converts the global constraint of expression (2) into a limit on the number of participating lower agents, as shown in expression (3):

[0087] (3)

[0088] wherein is a knowledge-based converter for mapping to the number of participating lower agents .

[0089] Algorithm 1: Pruning operation pseudo code

[0090]

[0091] It should also be noted that, since the node type and link constraint have ensured that there is at least one lower subtask node for each interaction subtask node, only the case where exceeds the upper limit needs to be handled. The processor can reasonably prune the redundant lower subtask nodes by using expression (3), so that the infeasible offspring after crossover and mutation is converted into a feasible individual. This pruning operator not only ensures the feasibility of the solution, but also introduces a certain degree of randomness due to the random selection of the nodes to be pruned, which is beneficial to maintaining the population diversity. In the actual operation process, the pseudo code of the pruning operation can be seen in Algorithm 1.

[0092] The beneficial effects of the above steps are as follows: the processor dynamically shrinks the invalid structure space by performing the above cross and mutation steps, and in the mutation step, the redundant leaf nodes and branches are removed by using the pruning mechanism. The processor judges whether the upper limit of the resource constraint is exceeded by counting the number of lower task nodes in the individual tree and using the knowledge converter to map the number of actual lower intelligent agents participating. If there is a task overload, the pruning operation will randomly cut off the redundant lower task nodes to control the scheduling size, while automatically inserting virtual task nodes at the structure breaking position to make up for the position, maintaining the structure closure and grammatical integrity of the individual tree. This mechanism has multiple technical advantages: first, it significantly improves the individual feasibility ratio, ensuring that the schemes generated in the evolution process meet the scheduling resource constraints and task logic constraints; second, by actively removing redundant nodes and invalid task paths, the pruning operation effectively compresses the search space dimension, improving the convergence efficiency of the optimization algorithm; third, the introduction of virtual nodes enhances the scalability and flexibility of the scheduling structure, supporting dynamic changes in the number of tasks without damaging the syntax structure; finally, the randomized pruning strategy ensures constraint satisfaction while preserving the structural diversity of the population, which helps to improve the overall search capability and avoid falling into local optima. Therefore, on the basis of ensuring the legality of scheduling, the pruning mechanism effectively improves the structural adaptability and search performance of the algorithm.

[0093] Finally, the processor finally judges whether the fitness of the algorithm has no significant improvement or reaches the maximum number of generations, and then the processor outputs the current optimal individual and converts the individual tree into a collaborative scheduling scheme for the UGV and the UAV; otherwise, the processor returns to the initial step of the pseudo code and continues to execute the next generation evolution until the conditions described in the previous paragraph are met.

[0094] By way of example, the application also provides simulation results:

[0095] In the present application, the task planning scenario is set and . In this scenario, it is assumed that represents the percentage of fire extinguishing tasks completed at the fire point . To simplify the simulation, the initial value is 0, and each step of each unmanned aerial vehicle that is assigned and successfully arrives at the target will increase by 1. Since the energy of the unmanned aerial vehicle performing the fire fighting task is limited, the energy consumption of the unmanned aerial vehicle is represented by limiting the working time of each unmanned aerial vehicle to . The total working time of the unmanned aerial vehicle is limited to , and the unmanned aerial vehicle that exceeds the time limit cannot move or affect . The size of the scene is , and the time limit is set to In HMAS, the UGV / UAV moves at a constant speed. Given that the UGV speed is typically slower than the UAV speed, the UAV speed is set to [value missing]. The cruising speed of the UGV is Population size The mutation rate was 0.2 and the crossover rate was 0.8.

[0096] The process of obtaining the optimal individual using a hierarchical unmanned swarm task planning method based on GP is as follows: Figure 5 As shown. (a) shows the UGV's travel path and status at t=1000s, and (b) shows the UGV navigating to the first deployment point at t=2400s. Two UAVs with targets 5 and 2 are deployed respectively. (c) is when the UAV cruises to the second deployment point at t=3700. The remaining five UAVs were deployed with targets of 3, 1, 1, 4, and 0. (d) At t=4035, the UGV was still at the second deployment point. The remaining five UAVs with targets 3, 1, 1, 4, and 0 have reached the ignition point. In the above process, the optimal individual's corresponding individual tree is: Root(Deploy([1.42,1.03], Deploy2(5,2)),Deploy([0.77,1.66], Deploy5(3,1,1,4,0)), Virtual, Virtual, Virtual, Virtual,Virtual, Virtual), with a fitness of 296.13. This indicates that the UGV will first navigate to the first deployment point. Two UAVs, designated as targets 5 and 2, were deployed respectively; the patrol then continued to the second deployment point. The remaining five targets were then deployed with UAVs of 3, 1, 1, 4, and 0 respectively. Figure 5 As can be seen, under this scheduling scheme, all UAVs can reach their corresponding targets within the energy limit, ultimately enabling HMAS to successfully cover all ignition points.

[0097] refer to Figure 6 Based on the above embodiments, this application also provides a hierarchical unmanned swarm task planning device based on GP to solve the same technical problem as any of the aforementioned method embodiments. The hierarchical unmanned swarm task planning device 100 based on GP may include an individual number construction module 1001, a fitness calculation module 1002, and a task planning scheme generation module 1003. The individual number construction module 1001 is used to determine five types of nodes in the unmanned swarm collaborative system jointly constructed by fire-fighting unmanned vehicles and multi-unmanned swarms based on the domain knowledge of the hierarchical multi-agent system, and to construct a hierarchical task individual tree with structured semantics based on the five types of nodes.

[0098] The fitness calculation module 1002 is configured to input the hierarchical task individual tree into a strong typed genetic programming algorithm as a population evolution baseline to generate an initial population, wherein an initial individual of each initial population is a hierarchical task individual tree, and each hierarchical task individual tree corresponds to a deployment scheme of each unmanned vehicle group and the fire unmanned vehicle when performing a task, and a predefined fitness function is used to evaluate the performance of each initial individual to obtain a fitness evaluation result.

[0099] The task planning scheme generation module 1003 is configured to determine whether a termination condition is met according to the fitness evaluation result, and if the termination condition is met, output a corresponding optimal hierarchical task individual tree according to an optimal fitness evaluation result; otherwise, perform selection, crossover and mutation operations on each individual by using the strong typed genetic programming algorithm, and clean redundant nodes in each individual by using a pruning algorithm based on domain knowledge in the crossover and mutation operation process to generate offspring hierarchical task individual trees, evaluate the performance of each population individual by using the fitness function to obtain a new fitness evaluation result, and iteratively perform the previous execution step, and convert the optimal hierarchical task individual tree into a task planning scheme of the unmanned swarm system.

[0100] It is worth mentioning that each module involved in the embodiment is a logical module. In actual application, one logical unit can be one physical unit, or a part of one physical unit, or realized by a combination of multiple physical units. In addition, in order to highlight the innovative part of the present application, units not closely related to solving the technical problems proposed in the present application are not introduced in the embodiment, but this does not mean that there are no other units in the embodiment.

[0101] In another embodiment provided by the present application, a device is also provided. The device includes a processor and a memory. The memory stores at least one instruction, at least one program, a code set or an instruction set. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the GP-based hierarchical unmanned swarm task planning method described in the embodiments of the present application.

[0102] In another embodiment provided by the present application, a computer readable storage medium is also provided. The storage medium stores at least one instruction, at least one program, a code set or an instruction set. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the GP-based hierarchical unmanned swarm task planning method described in the embodiments of the present application.

[0103] In the embodiments described above, all or some of the steps can be implemented by software, hardware or firmware, or any combination thereof. When implemented by software, all or some of the steps can be implemented in the form of one or more computer programs. The computer program can be stored in any computer readable medium, and when loaded into a computer system, can cause the computer system to perform one or more of the steps of the computer program. The computer readable medium can be a computer program product, a storage device, or a memory device. The computer readable medium can be a magnetic or optical disk, a magnetic tape, a semiconductor memory, a computer disk, or a computer disk drive. The computer readable medium can be a memory device such as a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, or a compact disk read-only memory (CD-ROM).

[0104] It should be noted that, in the present document, the terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between such entities or operations. Moreover, the terms "comprising", "containing", or any other similar term are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not necessarily include only those elements in the list, but can include other elements not expressly listed, or also include elements inherent in such process, method, article, or apparatus. Without more limitations, an element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0105] Each of the embodiments in the present specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, for the system embodiments, since they are basically similar to the method embodiments, they are described more simply, and the related parts can be referred to the part of the description of the method embodiments.

[0106] The above merely provides the preferred embodiments of the application, and not intended to limit the protection scope of the application. Any modification, equivalent replacement, improvement, and the like made within the principle and technical scope of the application shall fall into the protection scope of the application.

Claims

1. A GP-based hierarchical unmanned swarm mission planning method, characterized in that, The method comprises the following execution steps: Based on the domain knowledge of the hierarchical multi-agent system, five types of nodes of the unmanned cluster coordination system jointly constructed by the fire unmanned vehicle and the multi-unmanned aerial vehicle group are determined, and a hierarchical task individual tree with structured semantics is constructed based on the five types of nodes; The strong type genetic programming algorithm is used as the population evolution baseline, and the hierarchical task individual tree is input into the strong type genetic programming algorithm to generate an initial population, wherein the initial individual of each initial population is a hierarchical task individual tree, and each hierarchical task individual tree corresponds to a deployment scheme of each unmanned aerial vehicle group and the fire unmanned vehicle when performing a task, and a pre-defined fitness function is used to evaluate the performance of each initial individual to obtain a fitness evaluation result; According to the fitness evaluation result, it is judged whether the termination condition is met, if the termination condition is met, the optimal hierarchical task individual tree corresponding to the optimal fitness evaluation result is output; otherwise, the selection, crossover and mutation operations are performed on each individual by using the strong type genetic programming algorithm, and the redundant nodes in each individual are cleaned up by using the pruning algorithm based on the domain knowledge in the crossover and mutation operation process to generate offspring hierarchical task individual trees, the performance of each population individual is evaluated by using the fitness function to obtain a new fitness evaluation result, and the last execution step is iteratively executed, and the optimal hierarchical task individual tree is converted into a task planning scheme of the unmanned cluster coordination system; The five types of nodes include a total task node, a lower sub-task node, an interaction sub-task node, a parameter node and an upper sub-task node, wherein the total task node, the upper sub-task node, the interaction sub-task node and the lower sub-task node have a parent-child inheritance relationship, and the upper sub-task node includes a real sub-task node and a virtual sub-task node; The pruning algorithm based on the domain knowledge in the crossover and mutation operation process for cleaning up the redundant nodes in each individual comprises: An initial individual after crossover is obtained; Based on the basic subtree mutation operator of the strong type genetic programming algorithm, a subtree other than the root node is selected from each initial individual after crossover as a randomly generated new subtree to obtain a mutated individual; The number of lower layer agents participating in fire extinguishing is calculated according to the lower sub-task node of the mutated individual, wherein the lower sub-task node is used to determine the fire points corresponding to each lower layer agent, and the lower layer agent is an unmanned aerial vehicle; It is judged whether the number of lower layer agents is greater than the number of fire points, if the number of lower layer agents is greater than the number of fire points, a middle number within the range of the number of fire points is randomly generated, and it is judged whether the number of lower layer agents is greater than the middle number, if the number of lower layer agents is greater than the middle number, the lower sub-task node is randomly pruned until the number of lower layer agents is less than or equal to the number of fire points; During the pruning of the lower sub-task node, if all the lower sub-task nodes under the interaction sub-task node are pruned, the entire subtree with the real sub-task node as the root is pruned, and the virtual sub-task node is used to replace the real sub-task node to connect the total task node; The hierarchical task variable tree has a four-layer structure, and the four-layer structure comprises a root layer, an upper sub-task layer, an interaction sub-task layer and a lower sub-task layer from top to bottom; The total task node belongs to the root layer; The upper sub-task node belongs to the upper sub-task layer; The interaction subtask nodes and the parameter nodes belong to an interaction subtask layer; The lower-layer subtask nodes belong to a lower-layer subtask layer.

2. The GP-based hierarchical unmanned swarm mission planning method of claim 1, wherein, The overall task node is connected with the upper-layer subtask nodes and is used to determine an overall fire extinguishing task of the unmanned cluster cooperative system during fire extinguishing according to the upper-layer subtask nodes, and an execution sequence of the upper-layer subtask nodes is determined according to a connection sequence of the overall task node; The real subtask nodes of the upper-layer subtask nodes include the interaction subtask nodes and the parameter nodes, and the parameter nodes are used to store interaction parameters inside and outside the unmanned cluster cooperative system, wherein the interaction parameters include path coordinate parameters and energy consumption parameters, and the interaction subtask nodes represent fire starting points corresponding to each unmanned aerial vehicle and a cooperative path between the unmanned aerial vehicle group and the unmanned fire vehicle; The virtual subtask nodes of the upper-layer subtask nodes are used as placeholders when the number of real subtasks is less than the number of subtask nodes required according to the overall task node.

3. The GP-based hierarchical unmanned swarm mission planning method of claim 1, wherein, Before the initial individuals after the crossover are obtained, the method further includes: randomly selecting multiple initial individuals from the initial population based on a tournament selection operator; randomly selecting crossover points with the same type from each initial individual based on a conventional subtree crossover operator of the strongly typed genetic programming algorithm, and exchanging sub-trees with the crossover points as root nodes between the initial individuals to obtain each individual after the crossover.

4. The GP-based hierarchical unmanned swarm mission planning method of claim 1, wherein, Before the initial population is generated, the method further includes: predefining a node set and a node connection rule, wherein the node set and the node connection rule are used to generate a constraint condition to make the hierarchical task variable tree structure legal and meet a task constraint.

5. The GP-based hierarchical unmanned swarm mission planning method of claim 1, wherein, The expression of the fitness function is: wherein, , , , denotes the fire point the percentage of completion of the fire extinguishing task, denotes the number of drones actually deployed, denotes the number of drones to be deployed, denotes the total task duration per UAV, denotes the maximum total working time limit for all drones, denotes the degree of task completion, denotes the resource occupation efficiency, denotes the time consumption efficiency.

6. A GP-based hierarchical unmanned swarm mission planning apparatus that applies the GP-based hierarchical unmanned swarm mission planning method according to any one of claims 1 to 5, characterized by includes: an individual number construction module configured to determine five types of nodes of an unmanned cluster cooperative system constructed by a fire unmanned vehicle and a multi-unmanned aerial vehicle group based on domain knowledge of a hierarchical multi-agent system, and construct a hierarchical task individual tree with structured semantics based on the five types of nodes; a fitness calculation module configured to use a strongly typed genetic programming algorithm as a population evolution baseline, input the hierarchical task individual tree into the strongly typed genetic programming algorithm to generate an initial population, wherein each initial individual of each initial population is each hierarchical task individual tree, and each hierarchical task individual tree corresponds to a deployment scheme of each unmanned aerial vehicle group and the fire unmanned vehicle during task execution, and evaluate the performance of each initial individual by using a predefined fitness function to obtain a fitness evaluation result; a task planning scheme generation module configured to determine whether a termination condition is met according to the fitness evaluation result, output a corresponding optimal hierarchical task individual tree according to an optimal fitness evaluation result if the termination condition is met, or perform selection, crossover and mutation operations on each individual by using the strongly typed genetic programming algorithm, clean redundant nodes in each individual by using a pruning algorithm based on domain knowledge during the crossover and mutation operations, generate offspring hierarchical task individual trees, evaluate the performance of each population individual by using the fitness function to obtain new fitness evaluation results, iteratively perform the previous execution step, and convert the optimal hierarchical task individual tree into a task planning scheme of the unmanned cluster cooperative system.

7. An electronic device, comprising: The electronic device includes a processor and a memory having stored therein at least one instruction, at least one program, a code set or instruction set, which are loaded and executed by the processor to implement the GP-based hierarchical unmanned cluster task planning method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The storage medium has stored therein at least one instruction, at least one program, a code set or instruction set, which are loaded and executed by the processor to implement the GP-based hierarchical unmanned cluster task planning method according to any one of claims 1-5.

Images