Dynamic target allocation method and system based on improved DPSO algorithm

Abstract

The invention provides a dynamic target allocation method and system based on an improved DPSO algorithm, and belongs to the field of dynamic target allocation. The objective of the invention is to solve the problem of poor convergence and solving precision of a swarm intelligence optimization algorithm in the existing multi-missile cooperative combat target allocation process. According to the method, a traditional PSO (Particle Swarm Optimization) algorithm is discretized on the basis of a decimal coding and speed integerization mode, and resource constraints and integer constraints are ensured to be always met in a population updating process through a modular initialization and boundary value processing mode; a penalty function is introduced into a target function, and solutions violating strike efficiency constraints are continuously eliminated. In order to further improve the performance of the algorithm, domain expert knowledge is introduced in the solving process, and rapid and high-quality solving of the multi-missile cooperative combat target allocation problem is achieved.

Description

Technical Field

[0001] The present invention relates to the field of dynamic target allocation, and in particular to a dynamic target allocation method and system based on an improved DPSO algorithm. Background Art

[0002] With the increasing complexity of the modern battlefield environment, the traditional missile combat mode can no longer meet the needs of fast and efficient combat. Especially when facing multiple targets and complex battle situations, the strike effect of a single missile is limited, and it is difficult to achieve rapid response and efficient strikes. Therefore, how to improve the combat effectiveness of missiles has become the key to modern military research. The multi-missile cooperative combat strategy came into being. Through information interaction and collaborative cooperation between missiles, it broke through the limitations of a single missile system and significantly improved the strike accuracy and combat flexibility. This strategy enables missiles to select and destroy targets more quickly and accurately, achieving the maximum combat effect at a lower cost. In the modern battlefield, the challenge of target allocation is particularly prominent. The multi-target and multi-task strike requirements require the system to be able to quickly and accurately select and allocate targets, but in the existing multi-missile cooperative combat target allocation process, the convergence and solution accuracy of the swarm intelligence optimization algorithm are poor. To this end, the researchers proposed a multi-missile cooperative combat strategy based on a dynamic target allocation model, combined with an improved discrete particle swarm optimization algorithm, considering factors such as resource constraints and strike effectiveness, and greatly improving the efficiency and accuracy of target allocation. Summary of the invention

[0003] The technical problems to be solved by the present invention are:

[0004] In order to solve the problems of poor convergence and solution accuracy of swarm intelligence optimization algorithm in the existing multi-missile coordinated combat target allocation process.

[0005] The present invention adopts the following technical solutions to solve the above technical problems:

[0006] The present invention provides a dynamic target allocation method based on an improved DPSO algorithm, comprising the following steps:

[0007] S100. For combat missions, a dynamic target allocation model is established with the goal of completing strike missions while saving ammunition;

[0008] S200, through decimal coding and speed integerization method, the traditional particle swarm optimization algorithm is discretized, and through modular initialization and boundary value processing, it is used to ensure that the resource constraints and integer constraints of the population are always satisfied when updating;

[0009] S300, using an improved discrete particle swarm optimization algorithm to solve the model obtained in step S200, including introducing a penalty function into the objective function to gradually eliminate solutions that violate the strike effectiveness constraint; and introducing domain expert knowledge to reduce the randomness and uncertainty of the initial solution.

[0010] Further, in step S100, it specifically includes: according to the OODA observation-shooting-observation-shooting combat cycle, in a single engagement model, it is assumed that the Red side has m types of missiles, recorded as M1, M2, ..., M m The number of missiles of each type that can be used in this round is X1, X2, …, X m , the weight of each type of missile in this round of combat plan is w1, w2, … w m ;

[0011] The Red Army's combat planning matrix is ​​defined as:

[0012]

[0013] The matrix m rows and n columns represent the number of missiles and targets to be attacked, and the variable x ij Indicates the jth target to be hit T j The assigned missile type M i the number of

[0014] Define the lethality matrix of the red team against the blue team as:

[0015]

[0016] Among them, the variable r ij Indicates that each missile M i Can damage ij Enemy targets to be attacked T j ;

[0017] The loss of the blue side is equivalent to the sum of the attack effectiveness of the red side on it. If all the red side's missiles hit the target, the loss model of the j-th target to be hit by the blue side is obtained as follows:

[0018]

[0019] Considering resource constraints, strike effectiveness constraints and integer constraints, the target allocation optimization model for multi-missile coordinated combat is established as follows:

[0020]

[0021] Among them, α k It is a continuous damage factor;

[0022] The constraint set is used to limit the number of missiles of each type sent by the Red side to no more than the number of missiles available after deployment; the constraint set introduces a hyperparameter β that limits the maximum strike effectiveness, limiting the Red side to a maximum strike effectiveness of β times the number of targets; the constraint set is used to limit the Red side's decision variables x ij The value of is a non-negative integer;

[0023] Using the external penalty function method, the penalty for violating the jth constraint in the constraint set is defined as:

[0024]

[0025] The constraint set is introduced into the objective function in the form of penalty, and the solutions that violate the constraints are eliminated during iteration, and the new objective function is obtained as follows:

[0026]

[0027] Among them, σ represents the penalty factor, and σ<0; when a solution does not meet the jth constraint in the constraint set, it can be known that Since σ<0, The corresponding objective function value decreases, and is eliminated in the process of continuous iteration;

[0028] The target allocation model of multi-missile coordinated combat based on penalty function is obtained as follows:

[0029]

[0030] Further, in step S200, specifically including: in the multi-missile coordinated combat target allocation model established in step S100, the allocation plan of the Red side missiles is represented as an m-row and n-column matrix x=[x ij ] m×n , where the i-th row represents the number of missiles assigned to each target by the i-th missile;

[0031] A modular coding method is adopted and flattened to obtain a coding sequence;

[0032] The coding sequence has a total of m×n variables, where each m variables form a module, representing the i-th type of missile M i Treat the target T j (j=1,2,…,n) allocation scheme, decision variable x ij The position in the sequence is (i-1)×n+j.

[0033] Furthermore, in step S300, according to the constraint set established in step S100, the updated speed is approximately rounded to an integer to ensure that the updated decision variable is an integer.

[0034] Furthermore, in order to prevent the boundary constraints of the search space, the lower bound of a single decision variable is subjected to boundary absorption, and the lower bound of a single decision variable is subjected to modular boundary reflection based on the entire module, and the reflection rule is random reassignment.

[0035] Furthermore, in the process of introducing expert knowledge in the WTA field, missiles are allocated to targets with the greatest destructive effectiveness, including missile M i The damage efficiency of different targets is sorted, that is, the lethality matrix R n×m The elements of the i-th row of are sorted in descending order, and then replaced by a random number x ij The missiles are allocated to the target corresponding to each element in turn in the form of , and the remaining number of missiles X needs to be updated after each allocation is completed i , random numbers are generated as follows:

[0036] x ij =randi(0,X i )

[0037] The fitness function is consistent with the objective function in the multi-missile coordinated combat target allocation model:

[0038]

[0039] The present invention provides a dynamic target allocation system based on an improved DPSO algorithm. The system has a program module corresponding to the above steps and executes the steps in the above dynamic target allocation method based on the improved DPSO algorithm when running.

[0040] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of a dynamic target allocation method based on an improved DPSO algorithm when called by a processor.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1) The multi-missile coordinated combat strategy significantly improves combat effectiveness and flexibility through information exchange and coordinated actions between missiles when the performance of missiles is difficult to break through alone. This strategy enables missiles to select and destroy targets more quickly and accurately, thereby achieving the greatest combat effect at a lower combat cost.

[0043] 2) In view of the ever-changing battlefield combat environment and multiple targets to be hit, the multi-missile coordinated combat strategy can comprehensively consider the battlefield situation and quickly and accurately allocate targets according to certain rules. This adapts to the complexity and dynamics of the modern battlefield environment and improves the accuracy and efficiency of missile strikes.

[0044] 3) By establishing a dynamic target allocation model that maximizes enemy losses and minimizes own consumption, and proposing an improved discrete particle swarm optimization algorithm, a fast and high-quality solution to the multi-missile coordinated combat target allocation problem is achieved. This algorithm not only takes into account constraints such as weapon resources and strike effectiveness, but also introduces penalty functions and domain expert knowledge to further improve the performance and practicality of the algorithm.

[0045] 4) At the action level, the multi-missile coordinated combat strategy focuses on the coordinated guidance of missiles to ensure saturation strikes on targets. Through the coordinated research of each stage of missile flight, it provides better initial conditions for the terminal guidance stage of the missile, thereby improving the coordinated strike capability and hit accuracy of the missile.

[0046] 5) The research and development of multi-missile coordinated combat strategies and related technologies have promoted the innovation and progress of military technology. This strategy not only improves the effectiveness and flexibility of missile strikes, but also provides new ideas and means for future military operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of a dynamic target allocation method based on an improved DPSO algorithm in an embodiment of the present invention;

[0048] Figure 2 This is a logical diagram of the Red Team's "shoot-observe-shoot" combat in an embodiment of the present invention;

[0049] Figure 3 Schematic diagram of the updating process of the PSO algorithm in a two-dimensional search space in an embodiment of the present invention;

[0050] Figure 4 Schematic diagram of a decimal integer coding sequence in an embodiment of the present invention;

[0051] Figure 5 The following is a scatter plot of the optimal fitness values ​​obtained by running the four algorithms 100 times respectively in the embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0053] In order to realize multi-target positioning in a low-computing-power and low-power platform such as a micro-robot, the present invention proposes a dynamic target allocation method based on an improved DPSO algorithm. First, for combat targets that complete strike missions while saving ammunition, the weapon resources and strike effectiveness constraints are considered to establish a dynamic target allocation model that maximizes enemy losses and minimizes one's own consumption. And an improved discrete particle swarm optimization algorithm is proposed for model solution. Based on decimal coding and speed integerization, the traditional PSO algorithm is discretized, and modular initialization and boundary value processing are used to ensure that resource constraints and integer constraints are always met during the population update process; by introducing a penalty function in the objective function, solutions that violate the strike effectiveness constraints are continuously eliminated. In order to further improve the performance of the algorithm, domain expert knowledge is introduced in the solution process to achieve a fast and high-quality solution to the multi-missile coordinated combat target allocation problem.

[0054] Specific implementation plan 1: Combine Figures 1 to 4 As shown, the present invention provides a dynamic target allocation method based on an improved DPSO algorithm, comprising the following steps:

[0055] S100. For combat missions, a dynamic target allocation model is established with the goal of completing strike missions while saving ammunition;

[0056] The attack and defense between the red and blue sides is usually a multi-round attack and defense process. The dynamic weapon target assignment (DWTA) problem for this process is a multi-stage decision-making problem. According to the classic OODA (observation, orientation, decision, action) combat cycle, the combat logic of the red side is "observe-shoot-observe-shoot", such as Figure 2 As shown;

[0057] In the above combat logic, the single "observation-shooting" process in each round of engagement belongs to the static weapon target assignment (SWTA) problem, so the DWTA problem can be decomposed into multiple SWTA problems for modeling and solving;

[0058] In the single engagement model, assume that the Red side has m types of missiles, denoted as M1, M2, …, M m The number of missiles of each type that can be used in this round is X1, X2, …, X m , the weight of each type of missile in this round of combat plan is w1, w2, … w m ,The weight represents the value of the missile, which depends on the cost of the missile and the cost of launching;

[0059] Assume that the blue team has n types of targets to be attacked, denoted as T1, T2, …T n The number of each type of target to be hit in this round is Y1, Y2, ..., Y n, whose weights in this round of combat plan are w′1,w′2,…w′ n ,The weight indicates the importance of the target to be attacked, which depends on the economic value, strategic position and military capability factors of the target;

[0060] In each round of engagement, the Red side needs to optimize the target allocation model and dispatch corresponding missiles to attack the target to be attacked. Therefore, the Red side's combat planning matrix is ​​defined as:

[0061]

[0062] The matrix m rows and n columns represent the number of missiles and targets to be attacked, and the variable x ij Indicates the jth target to be hit T j The assigned missile type M i the number of

[0063] Since each type of missile has different strike effects on the enemy's target, it is necessary to define the lethality matrix of the red side against the blue side as:

[0064]

[0065] Among them, the variable r ij Indicates that each missile M i Can damage ij Enemy targets to be attacked T j .

[0066] The loss of the blue side is equivalent to the sum of the attack effectiveness of the red side on it. If all the red side's missiles hit the target, the loss model of the j-th target to be hit by the blue side can be obtained as follows:

[0067]

[0068] Taking into account resource constraints, strike effectiveness constraints and integer constraints, the target allocation optimization model for multi-missile coordinated combat is established as follows:

[0069]

[0070] Among them, α k It is a continuous damage factor;

[0071] In the above model, the constraint set is used to limit the number of missiles of each type sent by the Red side to not exceed the number of available missiles after deployment; the constraint set introduces a hyperparameter β that limits the maximum strike effectiveness, limiting the Red side to a maximum of β times the number of targets to avoid wasting resources by sending excessive forces; the constraint set is used to limit the Red side's decision variables x ijThe value of is a non-negative integer; however, considering the difficulty in solving multi-constraint optimization problems, the external penalty function method is used to define the penalty for violating the jth constraint in the constraint set as:

[0072]

[0073] Thus, the constraint set is introduced into the objective function in the form of penalty, and the solutions that violate the constraints are eliminated during the iteration; therefore, the new objective function is:

[0074]

[0075] Among them, σ represents the penalty factor, and σ<0; when a solution does not meet the jth constraint in the constraint set, it can be known that G j (x j )>0, since σ<0, therefore The corresponding objective function value decreases, and is eliminated in the process of continuous iteration;

[0076] Therefore, the multi-missile coordinated combat target allocation model based on penalty function is obtained as follows:

[0077]

[0078]

[0079] This type of WTA problem is a classic combinatorial optimization problem in the field of mathematics. This type of problem has the following characteristics: 1. The WTA problem belongs to an NP-complete problem, and its optimal solution must be solved by a complete enumeration method. When faced with a large-scale problem; 2. The dimension of the decision variable depends on the product of the equipment type and the target type, and the scale is huge; 3. It has a nonlinear objective function; 4. Due to changes in the battlefield, the constraints and the objective function have a certain randomness; 5. Its decision variables are discrete and non-continuous; The solution of this type of nonlinear integer programming problem is very complicated. When the problem scale is small, it is still feasible to obtain the optimal solution to the problem; but once the problem scale is enlarged, there is almost no deterministic optimization algorithm in the traditional branch and bound method, dynamic programming method and other precise algorithms that can calculate the optimal allocation plan of the problem within polynomial time; therefore, it is more appropriate to use an intelligent optimization algorithm to solve a less time-consuming suboptimal solution for the WTA problem; in view of this, the present invention proposes an improved discrete particle swarm optimization algorithm (IDPSO) for model solving;

[0080] S200, through decimal coding and speed integerization method, the traditional particle swarm optimization algorithm (PSO) is discretized, and modular initialization and boundary value processing are used to ensure that the resource constraints and integer constraints of the population are always met when updating;

[0081] The particle swarm optimization algorithm was originally inspired by the regularity of bird flock activities. It then simulated the process of bird flocks and fish swarms in nature, established a simplified model using swarm intelligence, and achieved model optimization and problem solving through information sharing and mutual cooperation between each individual in the swarm. Each bird in the bird swarm has a memory function, which can remember its position pBest (individual extreme value) closest to food. At the same time, through information sharing between clusters, it also knows the position gBest (global extreme value) of the bird closest to food in the entire cluster; therefore, in the process of searching for food, the bird flies towards its position pBest (individual extreme value) closest to food and the position gBest (global extreme value) of the bird closest to food in the entire cluster. In the PSO algorithm, each feasible solution is regarded as a bird, the optimal solution to be solved is regarded as food, and the distance of each bird from the food is regarded as fitness. Under the guidance of pBest and gBest, each solution moves closer to the optimal solution. The specific steps are as follows:

[0082] Assume that there are N particles forming a community in a D-dimensional search space, where the i-th particle can be expressed as:

[0083] X i =(x i1 ,x i2 ,…,x iD ),i=1,2,…,N

[0084] The flying speed of the i-th particle is also a D-dimensional vector, expressed as:

[0085] V i =(v i1 ,v i2 ,…,v iD ),i=1,2,…N

[0086] Each particle has a storage function. The optimal position searched by the i-th particle so far is called the individual extreme value, thus obtaining a D-dimensional vector:

[0087] p Best =(p i1 ,p i2 ,…,p iD ),i=1,2,…,N

[0088] Through information sharing within the entire particle swarm, a global extreme value is obtained:

[0089] g Best =(g1,g2,…,g D )

[0090] After finding these two optimal values, each particle updates its speed and adjusts its optimization direction through the following formula:

[0091] v ij (t+1)=v ij (t)+c1r1[p ij (t)-x ij (t)]+c2r2[g j (t)-x ij (t)]

[0092] The speed update process of the PSO algorithm in the two-dimensional search space is as follows: Figure 3 As shown. The update speed formula is divided into three parts, where v ij (t) is the inertia term, which indicates the influence of the magnitude and direction of the velocity at the previous moment on the velocity at the current moment. Since the initial velocity of each particle is randomly initialized, this part has a certain exploration randomness. ij (t)-x ij (t) is the self-cognition item, which indicates the influence of one’s historical experience on speed. It is the vector from the current position to one’s optimal position. g ij (t)-x ij (t) is the social cognition term, which indicates the influence of the social history experience of the particle group on the speed, and is the vector pointing from the current position to the optimal position of the group; c1, c2 are learning factors, which measure the weight of the influence of the self-cognition term and the social cognition term on the exploration direction of the particle. Generally, c1=c2, indicating that the self-experience and social experience are equally important; r1, r2 are random numbers in the range of [0,1], which increase the randomness of each particle in the process of experience learning;

[0093] After that, each particle updates its position according to the following formula:

[0094] x ij (t+1)=x ij (t)+v ij (t+1)

[0095] Among them, t is the number of iterations, x ij (t) and x ij (t+1) represents the position of the i-th particle in the j-th dimension at the t-th and t+1-th iterations, respectively, and v ij (t+1) represents the optimized speed of the i-th particle in the j-th dimension at the t+1-th iteration. In order to enhance the particle's optimization ability, the particle speed is generally updated using the following formula:

[0096] v ij (t+1)=w·v ij (t)+c1r1[p ij (t)-x ij (t)]+c2r2[g ij (t)-xij (t)]

[0097] Among them, w is the inertia factor, which is used to measure the influence of its own flight inertia on the exploration direction of the particle. Since the inertia term has random exploration, the larger the inertia factor w, the more random the particle exploration is, and the stronger the global optimization ability is. The smaller the inertia factor w, the stronger the local optimization ability of the particle is. Therefore, the linear decreasing weight search strategy is currently adopted:

[0098]

[0099] In the formula, t represents the current iteration number, T max Indicates the maximum number of iterations, w min Indicates the minimum value of the inertia factor, w max Indicates the maximum value of the inertia factor; in the early stage of iteration, the inertia weight is large, the particle swarm exploration is more random, and it converges quickly to the vicinity of the global optimal point. As the number of iterations increases, the inertia weight decreases, the particle swarm conducts a more detailed local search, and finally converges to the global optimal point.

[0100] Since the WTA problem to be solved in this step belongs to a combinatorial optimization problem, its solution space is discrete, and the PSO algorithm is only applicable to optimization problems with a continuous solution space; therefore, this step refers to the optimization mechanism of the traditional PSO algorithm and proposes an improved discrete PSO algorithm for the WAT problem;

[0101] For the WTA problem, there is no need to perform complex transformations on the decision variables in the combinatorial optimization model. Instead, the decision variables are directly mapped to the solution sequence required by the algorithm in a modular form. The coding is convenient and easy to understand, and there is no need for decoding. The fitness of each particle can be directly solved, which effectively speeds up the algorithm solution speed.

[0102] In the multi-missile coordinated combat target allocation model established in step S100, the allocation plan of the Red side's missiles is represented as a matrix x=[x ij ] m×n , where the i-th row represents the number of missiles assigned to each target by the i-th missile. Considering that the number of missiles launched in each round must not exceed the total number of missiles, the decision variables in each row of the matrix affect each other during the optimization process. Therefore, a modular coding method is adopted and flattened to obtain the following: Figure 4 The coding sequence shown;

[0103] The coding sequence has a total of m×n variables, where each m variables form a module, representing the i-th type of missile M i Treat the target T j (j=1,2,…,n) allocation scheme, then the decision variable x ijThe position in the sequence is (i-1)×n+j;

[0104] S300, using an improved discrete particle swarm optimization algorithm (DPSO) to solve the model;

[0105] The purpose of population initialization is to construct a feasible solution as the initial point of algorithm iteration. Therefore, when initializing, the existing constraint set in the model must be fully considered. Among them, the constraint set requires that the number of missiles of each type sent by the Red side cannot exceed the number of available missiles after deployment. Therefore, it is necessary to initialize according to the module during initialization, and the sum of the elements in each module must not exceed the total number of corresponding missiles. The constraint set requires that the number of missiles that the Red side must send must be an integer greater than zero, so the variables in each module must be natural numbers during initialization. The constraint set requires that the Red side can only send a maximum of β times the strike effectiveness of the number of targets, but modular coding is based on the type of missiles for module division rather than the type of target. Considering this constraint set will complicate the initialization method and reduce the solution speed, so the constraint set is not restricted. At the same time, in the objective function, the constraint has been processed by the penalty function. Therefore, although there are some infeasible solutions in the initial solution, these infeasible solutions will be gradually eliminated due to the penalty function during the continuous iteration process.

[0106] In addition to considering the feasibility of the initial solution, the present invention reduces the randomness and uncertainty of the initial solution by introducing expert knowledge in the WTA field, and effectively improves the convergence speed of the algorithm by improving the quality of the initial solution; when initializing, missiles are preferentially assigned to targets with the greatest damage effectiveness; i For example, first rank the damage effectiveness of the missile on different targets, that is, the lethality matrix R n×m The elements of the i-th row of are sorted in descending order, and then replaced by a random number x ij The missiles are allocated to the target corresponding to each element in turn in the form of , and the remaining number of missiles X needs to be updated after each allocation is completed i , the random number generation method is shown in the formula:

[0107] x ij =randi(0,X i )

[0108] The initial solution constructed in this way can not only satisfy the constraint set, but also accelerate the convergence speed. The initialization pseudo code for the i-th module is as follows. The population initialization needs to loop this code m times;

[0109] Population initialization pseudocode

[0110]

[0111]

[0112] The fitness value is an important indicator that reflects the quality of the solution relative to the original objective function, and is also an important basis for selecting individual extreme values ​​pbest and global extreme values ​​gbest in the PSO algorithm. Here, the fitness function is consistent with the objective function in the multi-missile coordinated combat target allocation model, so the fitness of any solution is:

[0113]

[0114] S400, introducing a penalty function into the objective function, gradually eliminating solutions that violate the strike effectiveness constraint, thereby ensuring the rationality of the solution;

[0115] After testing, it is found that for the WTA problem, the update method of the PSO algorithm population and the processing method of boundary conditions are the key to whether it can learn the experience of excellent individuals in the population; due to the existence of the constraint set, all decision variables need to be kept as natural numbers; in the exploration process, in order to keep the updated variables as integers, the present invention approximates the updated speed to ensure that the updated decision variables are still integers;

[0116] At the same time, when the particles are exploring, they may be constrained by the boundaries of the search space. Although theoretically it will not affect the generation of the optimal solution, in practical applications, useless exploration outside the boundaries will cause a great waste of computing resources and seriously affect the convergence speed of the algorithm. Therefore, it is necessary to reasonably handle the boundary conditions to improve the performance of the algorithm. There are three common ways to handle boundary conditions, namely: penalty function, boundary absorption and boundary reflection, see Table 1.

[0117] Table 1 Boundary processing method of optimization algorithm

[0118]

[0119] In order to ensure that the decision variable is still a natural number, the lower bound of a single decision variable is processed by boundary absorption, that is: if x ij <0, then x ij =0; since the upper bound of a single decision variable is affected by other decision variables within the entire module, the processing of the upper bound of the decision variable is based on the entire module, using a modular boundary reflection method, and the reflection rule is random redistribution. If the sum of the decision variables within the module exceeds the total number of missiles that can be used by the module, the decision variables within the module are redistributed in accordance with the population initialization method of step S300.

[0120] Specific implementation scheme 2: The present invention provides a dynamic target allocation system based on an improved DPSO algorithm. The system has a program module corresponding to the above steps, and executes the steps in the above dynamic target allocation method based on the improved DPSO algorithm during operation.

[0121] The other combinations and connection relationships of this embodiment are the same as those of the first embodiment.

[0122] Specific implementation scheme three: The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of a dynamic target allocation method based on an improved DPSO algorithm when called by a processor.

[0123] The other combinations and connection relationships of this embodiment are the same as those of the first embodiment.

[0124] Simulation experiment

[0125] In order to test the performance of the improved DPSO algorithm of the present invention, the algorithm is compared and analyzed with the traditional discrete particle swarm optimization algorithm (DPSO) with decimal coding and the hybrid mutation DPSO algorithm for the single target allocation problem. Among them, Algorithm 1 expands the standard particle swarm optimization algorithm to the discrete search space, and solves the WTA problem by approximating the particle position after each update; Algorithm 2 introduces the concepts of mutation and crossover in the genetic algorithm into the standard particle swarm algorithm to form a hybrid mutation DPSO algorithm to solve the WTA problem. At the same time, in order to test the role of expert experience in accelerating the convergence speed of the algorithm, a group of comparison algorithms are added to improve DPSO2, that is, the improved DPSO algorithm with random initialization. The simulation case data is as follows:

[0126] Table 1 WTA simulation case combat data

[0127]

[0128] In order to ensure the fairness of the comparative experiments, all simulations were performed on the same computer, configured with Windows 10 operating system, 16GB memory and AMD Ryzen 54600H processor (3.00GHz), and all algorithms were programmed in MatlabR2021b. In the setting of simulation parameters, the variable population size and maximum number of iterations of the four algorithms were set to be the same, where N = 100, K max = 1000, the inertia weights of the particle swarm algorithm for speed and position update are all set to w max =0.8,w min =0.2, the acceleration factor is set to c c =c2=1.5.

[0129] In the comparative experiment, in order to avoid random errors caused by random numbers in the algorithm, it is necessary to conduct multiple experiments to increase the number of comparative samples. Therefore, based on the above case data and simulation parameter settings, 100 cycles of simulation are performed to obtain the maximum fitness value of each algorithm in each run. Figure 5 As shown in the scatter plot, in 100 cycles of simulation, the four algorithms can enable the particle population to find the optimal allocation solution within 100 times, and the corresponding maximum fitness value is 16.9211.

[0130] Table 2 Optimal allocation scheme for coordinated strike of Red Army missiles

[0131] <![CDATA[T1]]> <![CDATA[T2]]> <![CDATA[T3]]> <![CDATA[T4]]> <![CDATA[T5]]> <![CDATA[M1]]> 0 6 2 0 0 <![CDATA[M2]]> 0 0 0 0 6 <![CDATA[M3]]> 0 0 0 4 2 <![CDATA[M4]]> 0 0 10 0 0

[0132] In order to further analyze the performance of the above algorithms, the following three evaluation indicators are established: the average maximum fitness value (AMFV) of the algorithm, the average optimization time (ACT) of the algorithm, and the solution success rate (SR). The calculation formula is as follows:

[0133]

[0134]

[0135]

[0136] In the above formula, k max Represents the total number of algorithm cycles, MFV k and CT k Represents the maximum fitness value and the time taken to find the best solution in the kth cycle of the algorithm, MFV max It represents the maximum fitness value corresponding to the optimal allocation scheme. According to the above formula, the evaluation index of each algorithm can be calculated as shown in the following table.

[0137] Table 3 Evaluation indicators of each algorithm

[0138]

[0139] In Table 3, in addition to the above four algorithms, the mixed integer linear programming solver (Intlinprog) in MATLAB R2021b is also introduced for comparative experiments. The solver uses traditional solutions such as branch and bound and simplex method. After analysis, it can be seen that in terms of solution quality, the improved DPSO algorithm of the present invention and the Intlinprog solver are tied for first place, followed by the hybrid DPSO algorithm and the improved DPSO2 algorithm, and the DPSO algorithm is the worst. Combining the AMFV index of the solution quality of the reaction algorithm with the SR index of the success rate of the reaction algorithm optimization, it is found that the improved DPSO algorithm and the Intlinprog solver can find the optimal allocation plan 100% in 100 cycles, with strong stability and accuracy. The hybrid DPSO algorithm and the improved DPSO2 algorithm are slightly inferior to the above two algorithms, but they can also find solutions with a small gap with the optimal allocation plan. The DPSO algorithm has a large gap with other algorithms in terms of solution quality and solution success rate, which shows that in addition to extending the exploration space to discrete regions, it is necessary to make specific improvements to the WTA problem when extending the particle swarm algorithm to solve the WTA problem. In terms of solution time, the latter three algorithms are mainly compared, and it is found that introducing the experience of experts in the field in the initialization and boundary value processing of the algorithm has greatly improved the algorithm's solution performance and convergence speed.

[0140] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A dynamic target allocation method based on an improved DPSO algorithm, characterized in that: The following steps are involved: S100. For combat missions, a dynamic target allocation model is established with the goal of completing strike missions while saving ammunition; S200, through decimal coding and speed integerization method, the traditional particle swarm optimization algorithm is discretized, and through modular initialization and boundary value processing, it is used to ensure that the resource constraints and integer constraints of the population are always satisfied when updating; S300, using an improved discrete particle swarm optimization algorithm to solve the model obtained in step S200, including introducing a penalty function into the objective function to gradually eliminate solutions that violate the strike effectiveness constraint; And introduce domain expert knowledge to reduce the randomness and uncertainty of the initial solution.

2. A dynamic target allocation method based on an improved DPSO algorithm according to claim 1, characterized in that: In step S100, the following specifically includes: according to the OODA observation-shooting-observation-shooting combat cycle, in a single engagement model, it is assumed that the Red side has m types of missiles, recorded as M1, M2, ..., M m The number of missiles of each type that can be used in this round is X1, X2, …, X m , the weight of each type of missile in this round of combat plan is w1, w2, … w m ; The Red Army's combat planning matrix is ​​defined as: The matrix m rows and n columns represent the number of missiles and targets to be attacked, and the variable x ij Indicates the jth target to be hit T j The assigned missile type M i the number of Define the lethality matrix of the red team against the blue team as: Among them, the variable r ij Indicates that each missile M i Can damage ij Enemy targets to be attacked T j ; The loss of the blue side is equivalent to the sum of the attack effectiveness of the red side on it. If all the red side's missiles hit the target, the loss model of the j-th target to be hit by the blue side is: Considering resource constraints, strike effectiveness constraints and integer constraints, the target allocation optimization model for multi-missile coordinated combat is established as follows: st x ij ∈N,i=1,2,…,m,j=1,2,…,n Among them, α k It is a continuous damage factor; The constraint set is used to limit the number of missiles of each type sent by the Red side to no more than the number of missiles available after deployment; the constraint set introduces a hyperparameter β that limits the maximum strike effectiveness, limiting the Red side to a maximum strike effectiveness of β times the number of targets; the constraint set is used to limit the Red side's decision variables x ij The value of is a non-negative integer; Using the external penalty function method, the penalty for violating the jth constraint in the constraint set is defined as: The constraint set is introduced into the objective function in the form of penalty, and the solutions that violate the constraints are eliminated during iteration, and the new objective function is obtained as follows: Among them, σ represents the penalty factor, and σ<0; when a solution does not meet the jth constraint in the constraint set, it can be known that G j (x j )>0, since σ<0, therefore The corresponding objective function value decreases, and is eliminated in the process of continuous iteration; The target allocation model of multi-missile coordinated combat based on penalty function is obtained as follows: st x ij ∈N,i=1,2,…,m,j=1,2,…,n。 3. A dynamic target allocation method based on an improved DPSO algorithm according to claim 2, characterized in that: In step S200, the following specifically includes: in the multi-missile coordinated combat target allocation model established in step S100, the allocation plan of the Red side missiles is represented as a matrix x=[x ij ] m×n , where the i-th row represents the number of missiles assigned to each target by the i-th missile; A modular coding method is adopted and flattened to obtain a coding sequence; The coding sequence has a total of m×n variables, where each m variables form a module, representing the i-th type of missile M i For the allocation scheme of the target to be attacked Tj (j = 1, 2, ..., n), the decision variable x ij The position in the sequence is (i-1)×n+j.

4. The dynamic target allocation method based on the improved DPSO algorithm according to claim 3 is characterized in that: In step S300, according to the constraint set established in step S100, the updated speed is approximately rounded to an integer to ensure that the updated decision variable is an integer.

5. The dynamic target allocation method based on the improved DPSO algorithm according to claim 4 is characterized in that: In order to prevent the boundary constraints of the search space, the lower bound of a single decision variable is absorbed, and the lower bound of a single decision variable is based on the entire module to perform modular boundary reflection, and the reflection rule is re-random allocation.

6. The dynamic target allocation method based on the improved DPSO algorithm according to claim 5 is characterized in that: In the process of introducing expert knowledge in the WTA field, missiles are allocated to targets with the greatest destructive effectiveness, including missile M i The damage efficiency of different targets is sorted, that is, the lethality matrix R n×m The elements of the i-th row of are sorted in descending order, and then replaced by a random number x ij The missiles are allocated to the target corresponding to each element in turn in the form of , and the remaining number of missiles X needs to be updated after each allocation is completed i , random numbers are generated as follows: x ij =randi(0,X i ) The fitness function is consistent with the objective function in the multi-missile coordinated combat target allocation model:

7. A dynamic target allocation system based on an improved DPSO algorithm, characterized in that: The system has a program module corresponding to the steps of any one of claims 1 to 6, and executes the steps of the dynamic target allocation method based on the improved DPSO algorithm when running.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of the dynamic target allocation method based on the improved DPSO algorithm described in any one of claims 1 to 6 when called by a processor.

Images