Weapon target distribution method and device for unmanned aerial vehicle fleet, equipment and medium

Abstract

The invention relates to the technical field of multi-unmanned aerial vehicle cooperative task planning, and provides a weapon target distribution method and device for an unmanned aerial vehicle group, equipment and a medium. The weapon target allocation method for the unmanned aerial vehicle fleet comprises the following steps: constructing a weapon target allocation model for cooperative combat of the unmanned aerial vehicle fleet, establishing a population set of an unmanned aerial vehicle set, a weapon resource consumption set and a static target set, and encoding individuals in the population set; in combination with constraint conditions and optimization targets in the weapon target allocation model, encoding individuals in the population set are initialized, and an initial population is generated; decomposing the initial population into a plurality of sub-populations; according to a neighborhood level relationship among the plurality of sub-populations, constructing neighborhood sub-populations among the plurality of sub-populations; screening the individuals in the neighborhood sub-population; and updating the neighborhood sub-populations based on a multi-population hierarchical selection mode to obtain a final population. And efficient and accurate weapon target distribution is realized, and the maximum attack income is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-UAV collaborative mission planning, and in particular to a weapon target allocation method, device, equipment and medium for an UAV fleet. Background Art

[0002] At present, in the relevant technologies, drone weapon target allocation is the key to obtaining the maximum attack benefit. In order to solve the static weapon target allocation problem, a multi-objective optimization algorithm is usually used to solve it. However, the traditional multi-objective optimization algorithm has the problems of reduced performance and low solution efficiency when solving the static weapon target allocation problem. Summary of the invention

[0003] A first aspect of an embodiment of the present invention provides a weapon target allocation method for a drone swarm, and the weapon target allocation method for a drone swarm includes: constructing a weapon target allocation model for cooperative combat of a drone swarm, the weapon target allocation model includes a drone set, a weapon resource consumption set and a static target set; optimizing and updating the allocation method between the drone set, the weapon resource consumption set and the static target set, specifically including: establishing a population set of drone sets, weapon resource consumption sets and static target sets, encoding individuals in the population set according to the correspondence between the drone set, the weapon resource consumption set and the static target set; initializing the encoded individuals in the population set in combination with the constraints and optimization objectives in the weapon target allocation model to generate an initial population; decomposing the initial population into multiple sub-populations; constructing neighborhood sub-populations between the multiple sub-populations according to the neighborhood hierarchical relationship between the multiple sub-populations; screening individuals in the neighborhood sub-populations; updating the neighborhood sub-populations based on a hierarchical selection method of multiple populations to obtain a final population.

[0004] In addition, the weapon target allocation method of the drone fleet in the above embodiment provided by the present invention may also have the following additional technical features:

[0005] In some embodiments, optionally, the allocation method between the drone set, the weapon resource consumption set and the static target set is optimized and updated, specifically including: based on the improved MOEA / D algorithm, the allocation method between the drone set, the weapon resource consumption set and the static target set is solved.

[0006] In some embodiments, optionally, the constraints in the weapon target allocation model include: the flight capability of the drone, the range and power of the weapon, the priority and destruction requirements of the target; the optimization objectives are one or more, and the optimization objectives include: maximizing the number of targets destroyed, minimizing drone losses, and balancing drone workload.

[0007] In some embodiments, optionally, the coded individuals in the population set are initialized, specifically including: using an incompletely random population initialization method to initialize the coded individuals in the population set to improve the fitness and evolutionary ability of the initialized individuals.

[0008] In some embodiments, optionally, the initial population is decomposed into multiple sub-populations, specifically including: using a multi-population decomposition method to decompose the initial population into multiple sub-populations, the weapon resource consumption corresponding to individuals in the same sub-population is the same, and has the same optimization goal.

[0009] In some embodiments, optionally, a neighborhood subpopulation is constructed between multiple subpopulations, specifically including: a neighborhood subpopulation design method based on neighborhood level, dynamically adjusting the neighborhood space size of each subpopulation according to the convergence and diversity level of the multiple subpopulations during the evolution process, and constructing neighborhood subpopulations to balance the convergence speed and diversity of the neighborhood subpopulations.

[0010] In some embodiments, individuals in the neighborhood subpopulation are optionally screened, specifically including: using a hierarchical non-dominated solution selection method to divide the individuals in the neighborhood subpopulation into different non-dominated levels, and giving priority to individuals with high non-dominated levels to meet the requirements for individual fitness in the neighborhood subpopulation and population diversity of the neighborhood subpopulation.

[0011] According to a second aspect of an embodiment of the present invention, there is provided a weapon target allocation device for a drone swarm, the weapon target allocation device for a drone swarm comprising: a model building unit, for building a weapon target allocation model for cooperative combat of a drone swarm, the weapon target allocation model comprising a drone set, a weapon resource consumption set and a static target set; a calculation and solution unit, for optimizing and updating the allocation method between the drone set, the weapon resource consumption set and the static target set, specifically comprising: establishing a population set of the drone set, the weapon resource consumption set and the static target set, encoding the individuals in the population set according to the correspondence between the drone set, the weapon resource consumption set and the static target set; initializing the encoded individuals in the population set in combination with the constraints and optimization objectives in the weapon target allocation model to generate an initial population; decomposing the initial population into multiple sub-populations; constructing a neighborhood sub-population between the multiple sub-populations according to the neighborhood hierarchical relationship between the multiple sub-populations; screening the individuals in the neighborhood sub-populations; updating the neighborhood sub-populations based on a hierarchical selection method of multiple populations to obtain a final population.

[0012] A third aspect of an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the weapon target allocation method for a drone fleet in the above-mentioned embodiment are implemented.

[0013] A fourth aspect of an embodiment of the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the weapon target allocation method for a drone fleet in the above-mentioned embodiment.

[0014] The beneficial effects brought by the present invention are as follows:

[0015] It can be seen from the above scheme that the embodiment of the present invention provides a weapon target allocation method for a drone fleet, which can also be understood as a weapon target allocation method based on the improved MOEA / D algorithm. It can build a reasonable weapon target allocation model according to different combat environments and tasks, and use intelligent algorithms to help the drone system quickly, accurately and autonomously determine the target to be attacked. Specifically, the algorithm is improved by multiple population decomposition methods, neighborhood subpopulation design methods and hierarchical non-dominated solution selection methods to improve the solution performance and efficiency, so as to solve the problems of reduced performance and low solution efficiency when solving static weapon target allocation problems, and can quickly and efficiently complete the weapon target allocation process, achieve more efficient and accurate weapon target allocation, and obtain maximum attack benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of a method for allocating weapons targets for a drone fleet provided in an embodiment of the present application;

[0017] Figure 2 A schematic diagram of an individual coding structure in a weapon target allocation method for a drone fleet provided in an embodiment of the present application;

[0018] Figure 3 A schematic diagram of multiple group decomposition results in the weapon target allocation method for a drone group provided in an embodiment of the present application;

[0019] Figure 4 A schematic diagram of the neighborhood subpopulation distribution structure in the weapon target allocation method for a drone fleet provided in an embodiment of the present application;

[0020] Figure 5 A schematic diagram of an example of selecting a non-dominated solution that does not satisfy the POF constraint in the weapon target allocation method for a drone fleet provided in an embodiment of the present application;

[0021] Figure 6 A schematic diagram of the operation flow of a hierarchical selection method based on multiple groups in a weapon target allocation method for a drone group provided in an embodiment of the present application;

[0022] Figure 7 A bar graph showing the average time consumption of the weapon target allocation method for the drone fleet provided in the embodiment of the present application and three comparison algorithms;

[0023] Figure 8 A bar graph showing the weapon target allocation method for a drone fleet provided in an embodiment of the present application and the average number of iterations of three comparison algorithms;

[0024] Fig. 9 A structural block diagram of a weapon target allocation device for a drone fleet provided in an embodiment of the present application;

[0025] Fig.10 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application.

[0026] in, Fig. 9 and Fig.10 The corresponding relationship between the reference numerals and component names in the figure is:

[0027] 200 weapon target allocation device of UAV fleet, 202 model building unit, 204 calculation and solution unit, 300 electronic device, 302 processor, 304 memory. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Refer to the following Figures 1 to 10 , through specific embodiments and their application scenarios, the weapon target allocation method of the drone swarm, the weapon target allocation device of the drone swarm, the electronic device and the storage medium provided in the embodiments of the present application are described in detail.

[0030] A first aspect of an embodiment of the present invention provides a weapon target allocation method for a drone swarm, and the weapon target allocation method for a drone swarm includes: constructing a weapon target allocation model for cooperative combat of a drone swarm, the weapon target allocation model includes a drone set, a weapon resource consumption set and a static target set; optimizing and updating the allocation method between the drone set, the weapon resource consumption set and the static target set, specifically including: establishing a population set of drone sets, weapon resource consumption sets and static target sets, encoding individuals in the population set according to the correspondence between the drone set, the weapon resource consumption set and the static target set; initializing the encoded individuals in the population set in combination with the constraints and optimization objectives in the weapon target allocation model to generate an initial population; decomposing the initial population into multiple sub-populations; constructing neighborhood sub-populations between the multiple sub-populations according to the neighborhood hierarchical relationship between the multiple sub-populations; screening individuals in the neighborhood sub-populations; updating the neighborhood sub-populations based on a hierarchical selection method of multiple populations to obtain a final population.

[0031] The weapon target allocation method for a drone fleet provided by the present invention specifically relates to a weapon target allocation method based on an improved MOEA / D algorithm. The method can construct a reasonable weapon target allocation model according to different combat environments and tasks, and use an intelligent algorithm to help the drone system quickly, accurately and autonomously determine the target to be attacked. Specifically, the algorithm is improved by a multi-population decomposition method, a neighborhood subpopulation design method and a hierarchical non-dominated solution selection method to improve the solution performance and efficiency, so as to solve the problems of reduced performance and low solution efficiency when solving static weapon target allocation problems. The method can quickly and efficiently complete the weapon target allocation process, realize more efficient and accurate weapon target allocation, and obtain maximum attack benefits. The method is used to solve the problems of reduced performance and low solution efficiency when solving static weapon target allocation problems in traditional multi-objective optimization algorithms.

[0032] Specifically, see Figure 1 The embodiment of the present application provides a method for allocating weapon targets of a drone fleet, which may include the following steps:

[0033] S102, constructing a weapon target allocation model for coordinated combat of a drone fleet, wherein the weapon target allocation model includes a drone set, a weapon resource consumption set, and a static target set;

[0034] S104, establishing a population set of a drone set, a weapon resource consumption set, and a static target set, and encoding individuals in the population set according to the corresponding relationship between the drone set, the weapon resource consumption set, and the static target set;

[0035] S106, combining the constraints and optimization objectives in the weapon target allocation model, initializing the coded individuals in the population set to generate an initial population;

[0036] S108, decomposing the initial population into multiple sub-populations;

[0037] S110, constructing neighborhood subpopulations between the multiple subpopulations according to the neighborhood hierarchical relationship between the multiple subpopulations;

[0038] S112, screening individuals in the neighborhood subpopulation;

[0039] S114. Update the neighborhood subpopulation based on the hierarchical selection method of multiple populations to obtain the final population.

[0040] The weapon target allocation method of the drone swarm of the present application constructs a weapon target allocation model for the coordinated combat of the drone swarm, which comprehensively considers the complex relationship between the drone set, the weapon resource consumption set and the static target set. On this basis, the intelligent optimization algorithm is used, especially by improving MOEA / D (Multi Objective Evolutionary Algorithm based on Decomposition), to achieve an efficient solution to the weapon target allocation problem.

[0041] Specifically, the method first constructs a weapon target allocation model that includes a set of drones, a set of weapon resource consumption, and a set of static targets. This step provides a basic framework for subsequent allocation. Then, through encoding technology, the corresponding relationship between drones, weapon resource consumption, and static targets is converted into individuals in the population set, so that the optimization problem can be carried out in a discrete search space.

[0042] In the initialization phase, the encoded individuals in the population set are initialized to generate the initial population, combining the constraints and optimization goals in the weapon target allocation model (such as maximizing attack benefits, minimizing weapon resource consumption, etc.). This step ensures that the individuals in the population have a certain degree of diversity and exploratory nature while meeting actual combat requirements.

[0043] Subsequently, the initial population is decomposed to obtain multiple subpopulations. These subpopulations are more evenly distributed in the search space, which helps the algorithm to find the optimal solution globally. According to the neighborhood hierarchical relationship between multiple subpopulations, neighborhood subpopulations are constructed. This step enhances the algorithm's search ability in the local range and helps to find more refined solutions.

[0044] In the screening stage, the individuals in the neighborhood subpopulation are screened and those with better fitness values ​​are retained. This step helps to reduce the computational burden of the algorithm and improve the efficiency of the solution. Finally, based on the hierarchical selection method of multiple populations, the neighborhood subpopulation is updated to obtain the final population. This step ensures that the algorithm can maintain the diversity and convergence of the population during the solution process by introducing a hierarchical selection mechanism.

[0045] The weapon target allocation method for the drone swarm provided by the present invention can achieve the following beneficial effects: by utilizing the multi-drone collaborative static weapon target allocation model modeled above and the solution of weapon target allocation based on the improved MOEA / D algorithm, when the drone swarms of the red and blue sides are engaged in combat, the red side's intelligent agent can quickly and efficiently complete the weapon target allocation process and achieve the total maximum attack benefit. It can also assist the red side's commanders in quickly judging the battlefield situation and improve the overall combat efficiency, and has a good application prospect.

[0046] Specifically, at present, drones are unmanned aerial vehicles that can fly through their own programs or remote control. Compared with manned aircraft, drones have the advantages of low cost, small size, high maneuverability, low risk, and flexible operation. They have been widely used in various fields, especially in high-threat scenario tasks. In the military field, drones can perform multiple tasks, including reconnaissance and surveillance, target search, data relay, fire strikes, and damage assessment. It can be seen that drones play a vital role in war.

[0047] Due to the complexity of the tasks performed by drones and the diversity of the surrounding environment, the combat capability, response capability and payload of a single drone can no longer meet the needs of modern military. If a drone fails in performing a mission, it will mean the failure of the entire operation, which may cause huge losses. Therefore, the research focus is placed on the collaborative execution of tasks by multiple drones. Through information interaction and capability complementarity between drones of different types and payloads, the advantages of each drone can be fully utilized, further improving the possibility of mission completion. In addition, the collaborative completion of tasks by multiple drones can shorten the time that drones are exposed to dangerous environments and reduce the risk of a single drone performing a mission. Therefore, the collaborative operation of multiple drones will become an inevitable requirement for future operations.

[0048] As one of the research contents of multi-unmanned coordinated combat, weapon target allocation is a hot topic in the current military combat field. Its purpose is to reasonably allocate our weapons to appropriate enemy targets in order to pursue the maximum attack benefits. With the continuous evolution of modern warfare technology, especially the emergence of various complex combat systems such as network-centric warfare, information and knowledge-centric warfare, the types and numbers of combat units are gradually increasing, and the situation changes on the battlefield are becoming more unpredictable. It is impossible to make a reasonable weapon target allocation plan based on the experience and response speed of the commanders. According to different combat environments and tasks, a reasonable weapon target allocation model is constructed, and intelligent algorithms are used to help the UAV system quickly, accurately and autonomously determine the target to be attacked, thereby maximizing combat efficiency and accuracy, and also improving the execution efficiency and combat effectiveness of the OODA (Observation Orientation Decision Action) loop.

[0049] For solving the static weapon target allocation problem, most of the traditional multi-objective optimization algorithms are designed for the problem of continuous Pareto (Pareto Front, optimal frontier or non-dominated solution set) frontier, while the Pareto frontier of the multi-objective static weapon target allocation problem is discrete, which reduces the performance and solution efficiency of the algorithm to a certain extent; at the same time, the Pareto frontier of the multi-objective static weapon target allocation problem is also constrained by multiple conditions, and the traditional multi-objective optimization algorithm cannot obtain a complete solution space. Therefore, the present invention improves the MOEA / D algorithm based on the different constraints of the Pareto frontier of the multi-objective static weapon target allocation problem, thereby improving the performance and efficiency of solving the multi-objective static weapon target allocation problem.

[0050] The technical solution of the present invention can be mainly divided into two parts: modeling of a multi-UAV cooperative static weapon target allocation model and solving the weapon target allocation based on the improved MOEA / D algorithm. First, the model of the multi-UAV cooperative static weapon target allocation problem is constructed to clarify that it is a type of multi-objective optimization problem, and it is improved on the basis of the MOEA / D algorithm. A multi-population decomposition method, a neighborhood subpopulation design method based on neighborhood level, and a hierarchical non-dominated solution selection method are proposed to improve the adaptability of the algorithm to the multi-UAV cooperative static weapon target allocation problem. At the same time, an incompletely random population initialization method is proposed to improve the algorithm convergence efficiency. The present invention utilizes the multi-UAV cooperative static weapon target allocation model modeled above and the solution of weapon target allocation based on the improved MOEA / D algorithm. When the red and blue drone groups are fighting, the intelligent body can quickly and efficiently complete the weapon target allocation process and achieve the total maximum attack benefit. It can also assist the red commander to quickly judge the battlefield situation and improve the overall combat efficiency. It has a good application prospect.

[0051] In some embodiments, optionally, the allocation method between the drone set, the weapon resource consumption set and the static target set is optimized and updated, specifically including: based on the improved MOEA / D algorithm, the allocation method between the drone set, the weapon resource consumption set and the static target set is solved.

[0052] In this embodiment, solving the allocation method between the drone set, the weapon resource consumption set and the static target set based on the improved MOEA / D algorithm is the core step in the optimization and updating process. This step not only inherits the advantages of the MOEA / D algorithm in multi-objective optimization problems, such as decomposition strategy, diversity preservation and convergence improvement, but also makes it more suitable for solving the specific problem of drone fleet weapon target allocation through a series of targeted improvement measures.

[0053] Specifically, the improvement of MOEA / D algorithm may include but is not limited to the following optimizations: Optimization of decomposition strategy: In view of the complexity of the weapon target allocation problem of drone fleet, a more sophisticated decomposition strategy is adopted to decompose the multi-objective optimization problem into a series of sub-problems with single objectives or fewer objectives. These sub-problems are both independent and interconnected, and together constitute a complete optimization framework. By solving these sub-problems, the global optimal solution can be gradually approached. Improvement of encoding and decoding mechanism: In view of the special relationship between drones, weapon resource consumption and static targets, a more reasonable encoding and decoding mechanism is designed. The encoding scheme needs to be able to accurately reflect the corresponding relationship between drones and targets, as well as the consumption of weapon resources. The decoding mechanism needs to be able to quickly generate a feasible weapon target allocation plan based on the encoding information. Adjustment of fitness function: According to the actual needs of the weapon target allocation problem, the fitness function is adjusted so that it can fully reflect the quality of the allocation plan. The fitness function may include multiple dimensions such as attack benefits, weapon resource consumption, and task completion time, and the advantages and disadvantages of the allocation plan are comprehensively evaluated by weighted summation or other methods. Optimization of neighborhood structure and selection strategy: In the MOEA / D algorithm, the neighborhood structure and selection strategy are crucial to the convergence speed and diversity of the algorithm. For the problem of weapon target allocation of drone swarms, a more reasonable neighborhood structure can be designed, such as a neighborhood partition method based on similarity or distance. At the same time, a more flexible selection strategy, such as hierarchical selection and elite retention, is adopted to ensure that the algorithm can maintain the diversity and convergence of the population during the solution process. Dynamic adjustment of algorithm parameters: During the operation of the algorithm, the algorithm parameters, such as crossover probability, mutation probability, neighborhood size, etc., are dynamically adjusted according to the current population distribution and optimization progress. Through parameter adjustment, the algorithm's solution efficiency and performance can be further improved. Based on the improved MOEA / D algorithm, the allocation method between drone sets, weapon resource consumption sets and static target sets is solved, which not only improves the algorithm's ability to solve complex problems, but also ensures the rationality and effectiveness of the allocation scheme.

[0054] In some embodiments, optionally, the constraints in the weapon target allocation model include: the flight capability of the drone, the range and power of the weapon, the priority and destruction requirements of the target; the optimization objectives are one or more, and the optimization objectives include: maximizing the number of targets destroyed, minimizing drone losses, and balancing drone workload.

[0055] In this embodiment, the constraints in the weapon target allocation model fully consider the flight capability of the drone, the range and power of the weapon, the priority and destruction requirements of the target. These constraints ensure the feasibility and actual combat effectiveness of the allocation scheme. At the same time, the optimization goal is set to maximize the number of target destruction, minimize the loss of drones, and balance the workload of drones. These goals together constitute the core pursuit of the weapon target allocation problem.

[0056] Specifically, the flight capability of the drone constrains the range of targets that the drone can reach and attack, while the range and power of the weapon determine the type and extent of targets that the drone can effectively destroy. The priority and destruction requirements of the targets are determined according to the battlefield situation and combat missions to ensure that key targets are given priority.

[0057] In terms of optimization goals, maximizing the number of targets destroyed is the primary task, which is directly related to the success or failure of the operation; minimizing the loss of drones is an important assessment indicator to ensure the continuous combat capability of the drone fleet; balancing the workload of drones helps to improve the overall combat efficiency and prevent individual drones from running out of resources or suffering losses in advance due to excessive tasks. By comprehensively considering these constraints and optimization goals, the weapon target allocation method in this embodiment can generate an allocation plan that is more in line with actual combat needs and realize efficient coordinated operations of drone fleets.

[0058] In some embodiments, optionally, the coded individuals in the population set are initialized, specifically including: using an incompletely random population initialization method to initialize the coded individuals in the population set to improve the fitness and evolutionary ability of the initialized individuals.

[0059] In this embodiment, an incompletely random population initialization method is used to initialize the coding individuals in the population set, aiming to improve the fitness and evolutionary ability of the initialized individuals. This initialization method combines problem-specific knowledge and heuristic information, making the generated initial population closer to the optimal solution, thereby accelerating the convergence process of the algorithm.

[0060] In some embodiments, optionally, the initial population is decomposed into multiple sub-populations, specifically including: using a multi-population decomposition method to decompose the initial population, and decomposing multiple sub-populations, in which the weapon resource consumption corresponding to individuals in the same sub-population is the same and has the same optimization goal.

[0061] In this embodiment, the use of a multi-population decomposition method to decompose the initial population is an effective strategy to improve the efficiency and quality of solving the weapon target allocation problem. Multiple sub-populations are obtained by decomposition, and the individuals in each sub-population have the same weapon resource consumption and optimization objectives, which helps the algorithm to conduct a deeper search in a local range while maintaining the diversity of the population.

[0062] In some embodiments, optionally, a neighborhood subpopulation is constructed between multiple subpopulations, specifically including: a neighborhood subpopulation design method based on neighborhood levels, and dynamically adjusting the neighborhood space size of each subpopulation according to the convergence and diversity levels during the evolution of multiple subpopulations to construct neighborhood subpopulations to balance the convergence speed and diversity of the neighborhood subpopulations.

[0063] In this embodiment, a neighborhood subpopulation design method based on neighborhood level is used to construct neighborhood subpopulations between multiple subpopulations. This strategy aims to balance the convergence speed and diversity of neighborhood subpopulations, thereby improving the efficiency and quality of solving the weapon target allocation problem.

[0064] Specifically, the determination of neighborhood level: First, a neighborhood level is assigned to each subpopulation based on the similarities (such as weapon resource consumption, optimization objectives, etc.) and differences between subpopulations. The neighborhood level reflects the relative position of subpopulations in the search space and the possible interactions between them. Dynamic adjustment of neighborhood space size: During the evolution of multiple subpopulations, the size of the neighborhood space of each subpopulation is dynamically adjusted according to their convergence and diversity levels. If a subpopulation has converged to a local optimal solution and has low diversity, its neighborhood space can be appropriately expanded to introduce more different individuals to increase the breadth of the search. On the contrary, if a subpopulation is still in the exploration stage and has high diversity, its neighborhood space can be reduced to search the current area more deeply. Constructing neighborhood subpopulations: According to the determined neighborhood level and the dynamically adjusted neighborhood space size, a neighborhood subpopulation is constructed. The neighborhood subpopulation consists of individuals from other subpopulations that have similarities and differences with the current subpopulation, and they can exchange information and co-evolve. Balancing convergence speed and diversity: By constructing neighborhood subpopulations and dynamically adjusting the size of the neighborhood space during the evolution process, the algorithm can accelerate convergence to the global optimal solution while maintaining diversity. This is because individuals in the neighborhood subpopulation can learn and collaborate with each other to jointly explore better areas in the search space. The neighborhood subpopulation design method based on neighborhood level not only improves the convergence speed of the algorithm when solving the weapon target allocation problem, but also maintains the diversity of the population, avoiding the problem of premature convergence and falling into the local optimal solution.

[0065] In some embodiments, individuals in the neighborhood subpopulation are optionally screened, specifically including: using a hierarchical non-dominated solution selection method to divide the individuals in the neighborhood subpopulation into different non-dominated levels, and giving priority to individuals with high non-dominated levels to meet the requirements for individual fitness in the neighborhood subpopulation and population diversity of the neighborhood subpopulation.

[0066] In this embodiment, a hierarchical non-dominated solution selection method is used to screen individuals in the neighborhood sub-population, so as to meet the requirements on the fitness of individuals in the neighborhood sub-population and the population diversity of the neighborhood sub-population.

[0067] Specifically, individual stratification: First, according to the fitness values ​​of individuals in the neighborhood subpopulation, they are divided into different non-dominated levels. Individuals with high non-dominated levels indicate that they have better fitness and performance in the current search space.

[0068] The stratification process is based on the dominance and non-domination relationship between individuals. If an individual is not dominated by any other individual (that is, its fitness value is better than or equal to all other individuals, and is strictly superior to other individuals in at least one aspect), it belongs to the first non-dominated level. Then, new non-dominated individuals are continuously searched from the remaining individuals, and they are successively assigned to higher non-dominated levels. Give priority to individuals with high non-dominated levels: When selecting individuals, first select individuals with the highest non-dominated level. These individuals have the best fitness and performance in the current search space, so they are more likely to contain information about the global optimal solution. If the number of individuals in the first non-dominated level is not enough to meet the selection requirements, continue to select individuals in the second non-dominated level, and so on. Maintain population diversity: In the selection process, in addition to considering the non-dominated level, it is also necessary to pay attention to maintaining the diversity of the population. This can be achieved by introducing a certain amount of randomness during selection, or by using genetic algorithm operations such as mutation and crossover after selection to increase the diversity of the population. In addition, in order to avoid premature convergence and falling into a local optimal solution, a threshold can be set in the selection process. When the number of individuals in a non-dominated level exceeds the threshold, the individuals in this level are stopped from being selected, and the individuals in the next non-dominated level are selected. Iterative update: After a round of selection, the selected individuals are formed into a new neighborhood sub-population, and subsequent evolutionary operations (such as mutation, crossover, etc.) are performed. Then, according to the fitness value and diversity level of the new neighborhood sub-population, hierarchical non-dominated solution selection is performed again to iteratively update the neighborhood sub-population. By adopting the hierarchical non-dominated solution selection method, the weapon target allocation algorithm in this embodiment can quickly converge to the global optimal solution while maintaining population diversity.

[0069] like Fig. 9As shown, the second aspect of the embodiment of the present invention provides a weapon target allocation device 200 for a drone swarm, and the weapon target allocation device 200 for the drone swarm includes: a model building unit 202, which is used to build a weapon target allocation model for cooperative combat of a drone swarm, and the weapon target allocation model includes a drone set, a weapon resource consumption set and a static target set; a calculation and solution unit 204, which is used to optimize and update the allocation method between the drone set, the weapon resource consumption set and the static target set, specifically including: establishing a population set of drone sets, weapon resource consumption sets and static target sets, and encoding individuals in the population set according to the correspondence between the drone set, the weapon resource consumption set and the static target set; initializing the encoded individuals in the population set in combination with the constraints and optimization goals in the weapon target allocation model to generate an initial population; decomposing the initial population into multiple sub-populations; constructing a neighborhood sub-population between the multiple sub-populations according to the neighborhood hierarchical relationship; screening individuals in the neighborhood sub-population; updating the neighborhood sub-population based on the hierarchical selection method of multiple populations to obtain a final population.

[0070] like Fig.10 As shown, a third aspect of an embodiment of the present invention provides an electronic device 300, including a memory 304, a processor 302, and a computer program stored in the memory 304 and executable on the processor 302. When the processor 302 executes the program, the steps of the weapon target allocation method for the drone fleet in the above-mentioned embodiment are implemented.

[0071] A fourth aspect of an embodiment of the present invention provides a storage medium having a computer program stored thereon, which, when executed by the processor 302, implements the steps of the weapon target allocation method for a drone fleet in the above-mentioned embodiment.

[0072] In practical specific applications, the specific implementation methods are as follows:

[0073] (I) A multi-UAV collaborative static weapon target allocation model, that is, a weapon target allocation model for UAV swarm collaborative combat.

[0074] If the consumption of weapon resources is increased, the possibility of hitting and destroying the enemy target will increase, that is, the total expected value of the enemy target's survival will decrease, which means that the benefits of our attack mission will also increase. On the contrary, if the consumption of weapon resources is reduced, the total expected value of the enemy target's survival will increase, and the benefits of our attack mission will decrease, while also increasing our overall combat risk. Obviously, the consumption of weapon resources and the expected total value of our attack mission benefits are two contradictory optimization goals. In order to optimize these two conflicting goals at the same time, the benefit problem of multi-UAV coordinated attack missions should be transformed into a multi-UAV coordinated static weapon target allocation problem. The model of this problem is as follows:

[0075]

[0076] In the formula, f is an optimization target, which represents the expected total value of the attack mission benefits; g is another optimization target, which represents the consumption of weapon resources. M represents the total number of weapon resources, N represents the total number of targets; V i represents the value of the i-th target; p ij represents the probability of weapon j damaging target i; x ij is a decision variable, which indicates whether weapon j attacks target i, x ij It can be specifically expressed as:

[0077]

[0078] In the formula represents the probability that target i is not destroyed by weapon j, then represents the probability of target i being destroyed is a constraint, indicating that a weapon can only attack one target.

[0079] (II) Solving the multi-UAV coordinated static weapon target allocation based on the improved MOEA / D algorithm, that is, optimizing and updating the allocation method between the UAV set, the weapon resource consumption set and the static target set. The specific steps are as follows:

[0080] 1. The encoding method of the individual. Figure 2 As shown, the present invention designs a decimal encoding method, which is specifically described as follows: For the problem of multi-UAV coordinated static weapon target allocation, the target set attacked by the weapon carried by UAV i can be expressed as where c i represents the number of weapon resources carried by drone i, t ij represents the target attacked by weapon j carried by drone i, are all random integers in [0,M]. When they are 0, it means that the weapon carried by the drone has not attacked the target. Therefore, the encoding of individuals in the population can be expressed as p k ={x k1 ,x k2 ,…,x kL}, k = 1, 2, ..., ps, where L is the number of drones; ps is the population size. The schematic diagram of its structure is as follows Figure 2 shown.

[0081] 2. Incompletely random population initialization method.

[0082] In the problem of multi-UAV coordinated static weapon target allocation, if a completely random method is used to initialize the population, there is a high probability that individuals with low fitness and weak evolutionary ability will be generated, thus affecting the convergence speed of the entire algorithm. In the present invention, a method of initializing the population in an incompletely random manner is proposed to initialize the individuals in the population so that the initial individuals have better fitness and higher evolutionary ability, which can improve the efficiency of the overall algorithm.

[0083] For the problem of multi-UAV coordinated static weapon target allocation, if one individual corresponds to all weapon resources of all UAVs attacking the same target, and another individual corresponds to different weapon resources of different UAVs attacking different targets, it is obvious that the latter will obtain higher attack benefits, that is, the individual has higher fitness and stronger evolutionary ability. Based on the above theory, the core idea of ​​the incompletely random population initialization method proposed in the present invention is to disperse the weapon resources of each UAV to attack different targets as much as possible. The pseudo code of the population initialization method is shown in Table 1:

[0084] Table 1 Pseudocode for incompletely random population initialization

[0085]

[0086] 3. Multiple population decomposition methods.

[0087] like Figure 3As shown, in multi-objective optimization problems, decomposition methods are usually used to decompose the problem into multiple scalar subproblems for solution. Although traditional decomposition methods (such as the Chebyshev method) work well in solving general multi-objective optimization problems, they cannot adapt to the characteristics of some specific problems. Therefore, for specific problems with different characteristics, designing corresponding decomposition methods can greatly improve the performance of the algorithm. Based on the above analysis, the present invention proposes a multi-population decomposition method to decompose a multi-objective static weapon target allocation problem. In this method, the population is decomposed into M sub-populations, where M is the total number of weapon resources. The weapon resource consumption corresponding to individuals in the same sub-population is the same, and has the same optimization goal f, that is, maximizing the expected total value of the attack mission benefit. When the total number of weapon resources is 8 and the number of targets is 4, the population decomposition results of the multi-objective static weapon target allocation problem are as follows: Figure 3 As shown, in Figure 3 In the scenario, there are 300 individuals in total. Since the number of weapon resources is 8, the population is decomposed into 8 sub-populations, and all individuals in the same sub-population correspond to the same weapon resource consumption. Therefore, the multi-target weapon target allocation problem is decomposed into a finite number of scalar sub-problems for solution.

[0088] 4. Neighborhood subpopulation design method based on neighborhood level.

[0089] In the traditional MOEA / D algorithm, each individual has T neighborhood individuals, and individuals in the neighborhood of the individual can be selected for mating operations, thereby improving the diversity of the population and accelerating the convergence speed of the algorithm. However, in the improved MOEA / D algorithm framework proposed in the present invention, the population is decomposed into a finite number of sub-populations based on the consumption of weapon resources, and the neighborhood space will be selected based on the sub-population. Therefore, the traditional neighborhood individual selection method cannot adapt to the improved MOEA / D algorithm.

[0090] In order to solve this problem, the present invention proposes a neighborhood subpopulation design method based on neighborhood level. In this method, the neighborhood space size of each subpopulation is no longer the same, but is determined according to the neighborhood level T. Assume that the population is decomposed into M subpopulations, that is, P = {s1, s2, ..., s M}, the neighborhood level is T, then the neighborhood subpopulation set B(i) of the i-th subpopulation is:

[0091]

[0092] Where: B(i) represents the neighborhood subpopulation of the ith subpopulation, s i represents the i-th subpopulation, T is the neighborhood level, and M is the number of subpopulations.

[0093] An example is given to further illustrate the specific implementation of this method. Still choose the scenario where the total number of weapon resources is 8 and the number of targets is 4. Let the neighborhood level T = 2, then the neighborhood sub-populations obtained by using this method are as Figure 4 shown. In Figure 4 , s1, s2, s3, s4, s5, s6, s7, and s8 are the sub-populations B(1), B(2), B(3), B(4), B(5), B(6), B(7), and B(8) decomposed based on the weapon resource consumption, respectively. The neighborhood sub-population sets corresponding to the above sub-populations are represented respectively. Each horizontal line in the figure corresponds to the sub-population it points to, and each vertical line and virtual circle combination represents a neighborhood sub-population set, which has intersections with the horizontal lines corresponding to the sub-populations included in the set. Therefore, the neighborhood sub-population sets of s1, s2, s3, s4, s5, s6, s7, and s8 are B(1) = {s1, s2, s3}, B(2) = {s1, s2, s3, s4}, B(3) = {s1, s2, s3, s4, s5}, B(4) = {s2, s3, s4, s5, s6}, B(5) = {s3, s4, s5, s6, s7}, B(6) = {s4, s5, s6, s7, s8}, B(7) = {s5, s6, s7, s8}, and B(8) = {s6, s7, s8}.

[0094] 5. Hierarchical non-dominated solution selection method.

[0095] In the traditional MOEA / D algorithm framework, the Pareto front solutions are selected by comparing the dominance relationships of feasible solutions. However, the traditional non-dominated solution selection method cannot take into account the constraints and characteristics of the POF in the multi-objective static weapon target allocation problem. In the multi-UAV cooperative multi-objective static weapon target allocation problem, the POF has two constraints, namely the quantity constraint and the target g-value constraint. Among them, the quantity constraint requires that the total number of Pareto front solutions is fixed, which is the total number of weapon resources; the target g-value constraint requires that the target g-values corresponding to each Pareto front solution are different. The quantity constraint problem has been initially solved during the population decomposition process, and the subsequent processing is relatively easy. However, the target g-value constraint problem cannot be handled by the traditional non-dominated solution selection method. An example of non-dominated solution selection that does not satisfy the POF constraints of the multi-objective static weapon target allocation problem is as Figure 5 shown.

[0096] Figure 5 describes a POF of the multi-objective static weapon target allocation problem when the total number of weapon resources is 6. A and B are respectively one of the solutions, and their corresponding objective function values are (f1, g1) and (f2, g2). According to the situation in the figure, it can be seen that f1 < f2 and g1 > g2, and solution B dominates solution A, that is At this time, if the traditional non-dominated solution selection method is used, solution A will be eliminated, and the POF in the example shown will lack solution A, which means that the weapon target allocation scheme when the weapon resource consumption corresponding to solution A is g1 cannot be obtained. When this happens, the obtained Pareto frontier solution will not meet the constraints of the multi-target static weapon target allocation problem on its quantity and target g value. In addition, when the population size is large, the traditional non-dominated solution selection method will also be very time-consuming, which will greatly reduce the convergence efficiency of the algorithm. Based on the above examples and analysis, the present invention proposes a hierarchical non-dominated solution selection method, and its specific process is shown in the pseudo code of Table 2.

[0097] Table 2 Pseudocode of hierarchical non-dominated solution selection method

[0098]

[0099] 6. Hierarchical selection method based on multiple populations.

[0100] In the traditional MOEA / D algorithm framework, each individual in the population is a scalar optimization problem, and converges toward the POF in a specific direction. In this process, evolutionary operations are continuously performed to generate new individuals, and the external population EP is updated. In the improved MOEA / D algorithm framework, the population is first decomposed into M sub-populations, and the individuals in each sub-population correspond to the same target g value. The multi-target static weapon target allocation problem is also decomposed into M scalar optimization sub-problems. Obviously, the traditional selection operation is no longer applicable. Therefore, based on the above analysis and the proposed hierarchical non-dominated solution selection method, the present invention proposes a hierarchical selection method based on multiple populations, which can be used to select individuals with higher fitness and greater evolutionary potential, thereby maintaining the diversity of the population.

[0101] When performing the selection operation, two important factors should be considered: the fitness of the individual and the diversity of the population. It can be seen that the distribution of POF in the multi-target static weapon target allocation problem is fixed. Based on this feature, when selecting individuals in the sub-population, it is necessary to maintain uniformity, so as to ensure that the distribution of the selected non-dominated solution is consistent with the distribution of POF. At the same time, uniform selection of individuals can also better maintain the diversity of the sub-population. The specific operation process of the hierarchical selection method based on multiple populations is as follows Figure 6 shown.

[0102] exist Figure 6 In the example, the selected individuals form a new population, where P is the size of the new population, M is the total number of weapon resources, n is the remainder of P divided by M, {x1,x2,···,x p} is the selected individual. In the first stage of the selection method operation process, the non-dominated solutions of each subpopulation in the current population are selected and placed in the new population, and then these selected individuals are removed from the current population. At this time, M individuals have been placed in the new population. In the second stage, the operation of the first stage is repeated in the population composed of the remaining individuals, and M individuals are selected again and placed in the new population. This process will be repeated q times, where q is the quotient of p divided by M, Figure 6 The penultimate stage in the process is the qth repetition of the operation. At this time, q×M individuals have been put into the new population. If n=0, this stage is the last stage and the selection process ends. If n≠0, n more individuals need to be selected and put into the new population.

[0103] Since this method selects non-dominated solutions in the subpopulation at each stage, it can be guaranteed that the selected individuals are those with the largest target f value in the subpopulation. At the same time, the selection operation at each stage selects non-dominated solutions in all subpopulations, and these non-dominated solutions contain all possible weapon resource consumption. Based on the above analysis, it can be seen that this selection method can meet the algorithm's requirements for individual fitness and population diversity.

[0104] For all algorithms involved in the present invention, the population size ps = M × 10, the crossover probability cp = 0.8, the mutation probability mp = 0.2, and the neighborhood level T = 3. At the same time, the termination condition of the algorithm is that the non-dominated solution in the current population remains unchanged in 50 generations of evolution. For each simulation test scenario, each algorithm is independently run 30 times.

[0105] The simulation test scenarios generated by the present invention are shown in Table 3:

[0106] Table 3 Simulation test scenarios for multi-UAV coordinated static weapon target allocation problem

[0107]

[0108] In the present invention, three different algorithms are selected for comparison to verify the overall performance of the improved MOEA / D algorithm. The three algorithms are NSGA-Ⅱ algorithm, MOEA / D-WS algorithm and MOEA / D-TCH algorithm.

[0109] The NSGA-Ⅱ algorithm is a typical evolutionary algorithm and one of the most commonly used and effective multi-objective optimization algorithms. It does not use a decomposition method, so it can be used to analyze whether the decomposition-based multi-objective optimization algorithm is more suitable for solving the multi-objective static weapon target allocation problem. At the same time, in order to control a single variable, the initialization method proposed in the present invention is added to the NSGA-Ⅱ algorithm. The MOEA / D-WS algorithm and the MOEA / D-TCH algorithm are two traditional MOEA / D algorithm frameworks, which use the weighted summation method and the Chebyshev method, respectively. Similarly, except for the decomposition method, the other parameters and related methods of these two algorithms are the same as those of the improved MOEA / D algorithm. For nine simulation test scenarios, the improved MOEA / D algorithm and the above three algorithms were independently run 30 times, and the statistical results are shown in Tables 4 and Figure 7 , Figure 8 shown.

[0110] Table 4 Statistical results of improved MOEA / D and three comparison algorithms

[0111]

[0112] In Table 4, Null means that the algorithm does not reach the termination condition within the specified time and no final result is output. Figure 7 and Figure 8 As shown, for scenarios 1 to 4, the average time consumption and average number of iterations of the improved MOEA / D algorithm are better than the three comparison algorithms; while for scenarios 5 to 9, the average time consumption and average number of iterations of the MOEA / D-WS algorithm are smaller, but this does not mean that the efficiency of the MOEA / D-WS algorithm is better than the other three algorithms, because the MOEA / D-WS algorithm has the problem of premature convergence. In addition, it can be seen that the average time consumption of the NSGA-Ⅱ algorithm is much higher than that of the other three algorithms. This is because the process of selecting non-dominated solutions by using non-dominated solution sorting and congestion sorting is very time-consuming. Based on the above analysis, the improved MOEA / D algorithm proposed in the present invention can more efficiently complete the solution of multi-UAV collaborative static weapon target allocation.

[0113] In this application, the term "plurality" means two or more than two, unless otherwise clearly defined. The terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; "connected" can be a direct connection or an indirect connection through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to the specific circumstances.

[0114] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "specific embodiments", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0115] The above are preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A weapon target allocation method for a drone fleet, characterized in that: The weapon target allocation method of the UAV fleet includes: Constructing a weapon target allocation model for coordinated combat of a drone fleet, wherein the weapon target allocation model includes a drone set, a weapon resource consumption set, and a static target set; Optimizing and updating the allocation method among the drone set, the weapon resource consumption set and the static target set, specifically including: Establishing a population set of the drone set, the weapon resource consumption set, and the static target set, and encoding individuals in the population set according to the corresponding relationship between the drone set, the weapon resource consumption set, and the static target set; Initializing the coded individuals in the population set in combination with the constraint conditions and optimization objectives in the weapon target allocation model to generate an initial population; Decomposing the initial population into multiple sub-populations; constructing neighborhood subpopulations between the plurality of subpopulations according to the neighborhood hierarchical relationship between the plurality of subpopulations; Screening individuals in the neighborhood subpopulation; The neighborhood subpopulation is updated based on a hierarchical selection method of multiple populations to obtain a final population.

2. The weapon target allocation method of a drone fleet according to claim 1, characterized in that: The optimizing and updating of the allocation method among the drone set, the weapon resource consumption set and the static target set specifically includes: Based on the improved MOEA / D algorithm, the allocation method among the UAV set, the weapon resource consumption set and the static target set is solved.

3. The weapon target allocation method of a drone fleet according to claim 1, characterized in that: The constraints in the weapon target allocation model include: the flight capability of the UAV, the range and power of the weapon, the priority and destruction requirements of the target; The optimization objectives are one or more, and the optimization objectives include: maximizing the number of targets destroyed, minimizing drone losses, and balancing drone workload.

4. The weapon target allocation method of a drone fleet according to claim 1, characterized in that: The initializing the coding individuals in the population set specifically includes: An incompletely random population initialization method is used to initialize the coding individuals in the population set to improve the fitness and evolutionary ability of the initialized individuals.

5. The weapon target allocation method of a drone fleet according to claim 1, characterized in that: Decomposing the initial population into a plurality of sub-populations specifically includes: The initial population is decomposed by using a multi-population decomposition method to obtain a plurality of sub-populations. The weapon resource consumption corresponding to individuals in the same sub-population is the same and has the same optimization goal.

6. The weapon target allocation method of a drone fleet according to any one of claims 1 to 5, characterized in that: The step of constructing a neighborhood subpopulation between the plurality of subpopulations specifically includes: The neighborhood subpopulation design method based on neighborhood level dynamically adjusts the neighborhood space size of each subpopulation according to the convergence status and diversity level of the multiple subpopulations during evolution, and constructs the neighborhood subpopulations to balance the convergence speed and diversity of the neighborhood subpopulations.

7. The weapon target allocation method of a drone fleet according to any one of claims 1 to 5, characterized in that: The screening of individuals in the neighborhood subpopulation specifically includes: A hierarchical non-dominated solution selection method is adopted to divide the individuals in the neighborhood sub-population into different non-dominated levels, and individuals with high non-dominated levels are preferentially selected to meet the requirements of individual fitness in the neighborhood sub-population and population diversity of the neighborhood sub-population.

8. A weapon target allocation device (200) for a drone fleet, characterized in that: The weapon target allocation device (200) of the drone fleet comprises: A model building unit (202) is used to build a weapon target allocation model for coordinated combat of a drone group, wherein the weapon target allocation model includes a drone set, a weapon resource consumption set, and a static target set; The calculation and solution unit (204) is used to optimize and update the allocation method between the drone set, the weapon resource consumption set and the static target set, specifically including: Establishing a population set of the drone set, the weapon resource consumption set, and the static target set, and encoding individuals in the population set according to the corresponding relationship between the drone set, the weapon resource consumption set, and the static target set; Initializing the coded individuals in the population set in combination with the constraint conditions and optimization objectives in the weapon target allocation model to generate an initial population; Decomposing the initial population into multiple sub-populations; constructing neighborhood subpopulations between the plurality of subpopulations according to the neighborhood hierarchical relationship between the plurality of subpopulations; Screening individuals in the neighborhood subpopulation; The neighborhood subpopulation is updated based on a hierarchical selection method of multiple populations to obtain a final population.

9. An electronic device (300), comprising a memory (304), a processor (302), and a computer program stored in the memory (304) and executable on the processor (302), characterized in that: When the processor (302) executes the program, the steps of the weapon target allocation method for a drone fleet described in any one of claims 1 to 7 are implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor (302), the steps of the weapon target allocation method for a drone fleet described in any one of claims 1 to 7 are implemented.

Images